Optical Biosensors: Present and Future

PREFACE

Since the birth of the field of optical biosensors, the pace of evolution of this field has been swift. While myriad reports have appeared describing applications and advancements in optical biosensor technology, few existing volumes are dedicated to a synopsis of this field. Since the development of optical biosensors mirrors the advances in the rapidly evolving telecommunications industry, we deemed the time to be ripe for such an opus. In order to catch the wave of this rapidly developing technology, we endeavored to focus both on the current state of the art and on technologies that will influence tommorrow's state of the art. We hope that this particular compendium of concepts will trigger new synapses to foma in the brains of our readers and yield even more innovation in the years to come. The history sections are included in order to recognize the contributions of the giants upon whose shoulders we stand--and we thank them for their creativity and pioneering spirit. These sections are comparatively short, not so as to minimize such contributions, but so that this book actually gets published in a single volume. According to the thematic focus on Present and Furore technology, the book is divided into two parts. In the first part, we compiled a list of the most outstanding optical biosensor technologies, while in the second part, the editors used their crystal ball to select the science we deem exciting and promising in terms of potential impact on biosensors. The optical biosensor technologies include two very different fiber optic biosensors, planar waveguides, and the displacement flow sensors, as well as sensors based on time-resolved fluorescence, electrochemiluminescence, surface plasmon resonance, resonant mirrors, and interferometry. The science for future technology development includes four different methods for producing new recognition elements (genetic engineering of proteins, chemical synthesis, combinatorial selection of nucleotide-based receptors, and molecular imprinting), two methods for immobilizing receptors on biosensors (sol gels and semi-synthetic membranes), two methods for producing very bright signals (PEBBLES and quantum dots), and soft lithography for surface patterning and microfluidics. We have asked leaders in each field to provide our readers with as thorough and objective a chapter as possible; they and their colleagues have been very patient with our nagging and nit picking and, as will be obvious to you, have put inordinant amounts of time into providing a conscientious review of their field. We tasked the authors to describe the underlying principles behind each technology, enumerate the types of applications for which it has been tested, provide their opinions about the advantages and disadvantages of their favorite vii

Preface biosensor (and the objectivity each has provided is admirable!), and philosophize on the future developments using that particular biosensor. The last section is intended to be fun for the readers as well as the authors; however, it is available for any clever venture capitalist to peruse as well. Finally, the editors intend this book to be a gift of gratitude to our colleagues in this rapidly expanding field. We appreciate the open sharing of ideas, the encouragement, and the competition that motivates us to greater effort. To work in the field of optical biosensors, one must be curious about biochemistry, chemistry, physics, and engineering and the possibilities ever present in the cracks between the disciplines. While information overload is a serious threat, boredom never is. Since it is absolutely impossible to be expert in all these fields, it behooves us to join forces with those who are. But even more than the ideas and accomplishments of our fellows, we delight in their personalities and camaraderie. Sincerely, Fran and Chris

viii

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 1

OPTRODE-BASED FIBER OPTIC BIOSENSORS (BIO-OPTRODE) ISRAEL BIRAN, PH.D. AND DAVID R. WALT, PH.D.

The Max Tishler Laboratory for Organic Chemistry Department Of Chemistry Tufts University, Medford, MA 02155 USA

Optrode-based fiber optic biosensors (bio-optrodes) are analytical devices incorporating optical fibers and biological recognition molecules. Optical fibers are small and flexible "wires" made out of glass or plastic that can transmit light signals, with minimal loss, over long distances. The light signals are generated by a sensing layer, which is usually composed of biorecognition molecules and dyes, coupled to the fiber end. Light is transmitted through the optical fibers to the sensing layer where different optical phenomena such as absorption or luminescence are used to measure the interactions between the analyte and the sensing layer. Bio-optrodes can be used for remote analytical applications including clinical, environmental, and industrial process monitoring. In the last decade, due to the rapidly growing use of fiber optics for telecommunication applications, new fiber optic technologies have been developed resulting in high-quality and inexpensive optical fibers that can be used for bio-optrode applications. Recent advancements in bio-optrode technologies include the development of nanoscale bio-optrodes, enabling measurements inside single living cells, and the development of multi-analyte and reagentless bio-optrodes. Although currently no biooptrodes are commercially available, it is expected that the development of advanced bio-optrode technologies will lead to commercially available devices for various analytical applications.

Biran and Walt

Figure 1. Schematic diagram of optrode system.

I. Principle of Operation The word "optrode" is a combination of the words "optical" and "electrode" and refers to a fiber optic based analytical device that can measure the concentration of a specific chemical or a group of chemicals in a sample of interest. The basic design of an optrode system is shown in Figure 1. The main components of an optrode are: (a) a light source; (b) an optical fiber to both transmit the light and act as the substrate for (c) the sensing material, which is usually immobilized to the surface of the end face of the fiber; and (d) a detector to measure the output light signal. Computers or microprocessors are used to control the optrode instrumentation and are employed to analyze the output signals. The "heart" of the optrode is the sensing element. When the sensing element interacts with the analyte, it undergoes physico-chemical transformations that change its optical properties. This transduction mechanism generates optical signals that can be correlated to the analyte concentration. The optical signals are measured by launching light from the light source through the optical fiber to the fiber end, where the sensing element is immobilized. The same fiber (Figure 1), or a different fiber (Figure 6), is used to guide the output light to the detector

Optrode-based Fiber Optic Biosensors

Core(nl) Cladding(n2)

Jaclket

Figure 2. Schematic diagram of an optical fiber shows core and clad structure.

(e.g., spectrophotometer, fluorometer) where the reflected, emitted or absorbed light is measured. Optrode biosensors or bio-optrodes are optrodes in which the sensing elements are of biological origin. Biological sensing elements, such as enzymes, nucleic acids, antibodies and cells, are immobilized on optical fibers and used for specific recognition of many different analytes (Cunningham, 1998; Kuswandi et al., 2001; Mehrvar et al., 2000; Wolfbeis, 2000). Since most biological sensing elements and most analytes do not possess intrinsic spectral properties, the biorecognition events are transduced to optical signals (e.g. changes in fluorescence or absorbance) by coupling optically responsive reagents to the sensing elements. For example, fluorescent dyes are used to label nucleic acids and convert the biorecognition interaction between two complementary DNA strands into a fluorescence signal. In another example, an indicator dye, which is optically sensitive to changes in H + concentrations, is used to transduce enzymatic activity that consumes or releases H § into an optical signal. The signals are generated on the fiber optic face and transmitted by the optical fiber to a remote measurement device. The small dimensions of bio-optrodes allow measurement in very small sample volumes, which make them suitable for various clinical applications (Meadows, 1996; Vo-Dinh and Cullum, 2000). Biooptrodes are also useful for different sensing applications in the industrial and environmental fields (Rogers and Mascini, 1998; Rogers and Poziomek, 1996; Marose et al., 1999; Mulchandani and Bassi, 1995; Scheper et al., 1996). In this section, optical fibers, their basic characteristics, and the optical methods used to transduce a biorecognition event to an optical signal are described. The instrumentation employed in optrode biosensors, the biological sensing elements, and the methods to immobilize them on the fiber optic surfaces are summarized.

Biran and Walt

l

n~ L:

n2

'

.

7

[Cladding ( a )

Cladding

(b)

Core

(c) [

"-

-

n~

"

Acceptance cone

Core

Figure 3. Propagation of light through the optical fiber occurs when the total internal reflection condition exists at the interface between the core, (nl), and clad, (n2) such that nl > n2. (a) Light entering the fiber is totally internally reflected (TIR), if the light angle is greater than the critical angle (pc. (b) Light will be partially reflected and partially refracted, if the light angle is less then the critical angle ~oc.(c) Light will propagate in TIR only when the entering light angle is within the acceptance cone angle (~0m)range.

1.1. O p t i c a l f i b e r c h a r a c t e r i s t i c s a n d use in b i o - o p t r o d e s

Optical fibers are small and flexible "wires" made out of glass or plastic that can transmit light signals, with minimal loss, for long distances. Optical fibers are remarkably strong, flexible and durable and therefore can be used in harsh and hazardous environments. Optical fibers are non-electrical, which make them highly suitable for applications where the presence of electric current is detrimental (e.g., in-vivo monitoring inside a patient body). In the last decade, due to the rapidly growing use of fiber optics for telecommunication applications, new fiber optic technologies have been developed resulting in high-quality and inexpensive optical fibers that can also be used for sensing applications. Optical fibers can transmit multiple optical signals simultaneously, thereby offering multiplexing capabilities for sensing.

Optrode-based Fiber Optic Biosensors Optical fibers consist of a core with a refractive index, n~, surrounded by a cladding with a lower refractive index, n2 (Figure 2). The difference in the refractive indices between the core and the cladding enables the core-clad interface to effectively act as a mirror such that a series of internal reflections transmits the light from one end of the fiber to the other as shown in Figure 3 (a). Light undergoes total internal reflection (TIR) at the core-clad interface if two basic conditions are fulfilled: (a) The light strikes the cladding at an angle greater than the critical angle, (Pc,(Figure 3 (a) and 3 (b)). The critical angle is defined by the ratio between the clad and the core refractive indices, as shown in Equation (1): sin (p~ - n 2 / n 1 (1) (b) The angles of the light entering the fiber should be within the acceptance cone as shown in Figure 3 (c). The acceptance cone angle, (am, depends on the refractive indices of the core and the clad and also on the refractive index of the medium from which the light enters the fiber, no.

sin(Pm =

(2) l't o

Another important parameter that defines the fiber's light collection efficiency is the numerical aperture (NA). This parameter is related to the acceptance cone's angle and is given by: NA = n0sinfo m

(3)

A high NA indicates a wide acceptance cone and better light gathering capabilities of the fiber. A typical NA value for a high quality glass fiber is 0.55, but fiber NAs as high as 0.66 or as low as 0.22 have been used for sensing. Optical fibers are usually made out of plastic and glass and have many different configurations, formats, shapes, and sizes. Glass fibers are the most commonly used fibers in optrode biosensors. Glass optical fibers can transmit light in the visible and near-infrared regions of the optical spectrum (400 n m < ~, < 700 nm) and are therefore suitable for measuring fluorescence signals generated by most fluorescent dyes. For applications in which light in the UV region is required, quartz (pure silica) is used as the fiber's core material and doped silica (with a lower refractive index) is used as the cladding material. For most fiber opticbased biosensors, optical fibers with diameters ranging from 50 to 500 ~tm are employed.

Biran and Walt

Figure 4. Optical fiber bundle fabrication and its use for imaging. (a) Fiber bundles are constructed from thousands of individual single fibers that are fused together. (b) Coherent bundles can be used for imaging (Pantano and Walt, 1995). Reprinted with permission from the American Chemical Society.

Recently, fiber optic bundles (Figure 4(a)) comprising thousands of identical single fibers each with a diameter of a few micrometers, were employed for biooptrodes. The fibers can be bundled in a coherent or random fashion. In coherent fiber bundles, the position of each fiber on one end is identical to its position on the other end. These fibers were originally designed for imaging applications as shown in Figure 4(b) and are also often called "optical imaging fibers". Imaging fibers are suitable for multi-analyte optrode biosensor design (Healey and Walt, 1995; Healey et al., 1997a; Michael et al., 1998; Steemers and Walt, 1999; Walt, 2000) since each small individual fiber in the bundle can carry its own light signal from one end of the bundle to the other. Moreover, optical imaging fiberbased biosensors can be used for sensing and imaging simultaneously, providing remote spatial sensing capabilities (Walt, 1998).

1.2. Optical phenomena employed for biosensing in bio-optrodes In bio-optrodes, dyes are coupled to the biological sensing element and transduce the biorecognition events to an optically detectable signal. Different optical 10

Optrode-based Fiber Optic Biosensors phenomena, including fluorescence, luminescence and absorption, are employed for monitoring these optical changes. In this section, the basic principles of these phenomena and their use in bio-optrodes are described.

Fluorescence is commonly used in bio-optrodes. Fluorescence occurs when molecules are excited at a specific wavelength and re-emit radiation at a lower energy, i.e., a longer wavelength. The absorption of the excitation light promotes the molecule's energy from its ground state to a higher energy state. The molecule emits fluorescent light when it returns to the ground state. Each fluorescent molecule has a unique fluorescence spectrum since the excitation and emission occur only at distinct energy levels corresponding to particular wavelengths. The characteristic fluorescence spectrum of particular molecules allows multiple fluorescent dyes to be used simultaneously in a single analytical assay. In fluorescence-based bio-optrodes, the fluorescence signals are measured by transmitting the excitation light through an optical fiber and measuring the light emission using a detector. Usually the increase or decrease in fluorescence intensity is measured and then correlated to the analyte concentration. For example, when a fluorescent-labeled antibody is used as the sensing element, the fluorescence intensity is proportional to the amount of antigen (analyte) bound to the optical fiber. One method for measuring fluorescence lifetime is frequencydomain. In this method, sinusoidally modulated light is used to excite the fluorescent molecule. The resulting emission light also oscillates at the same frequency. The emission light is phase shifted (delayed) and demodulated with respect to the excitation light because of the finite lifetime of fluorescence. The phase shift is expressed as a phase angle from which the lifetime can be determined using simple relationships between the modulation frequency and the degree of demodulation. The concentration of analyte that induces changes in the molecule's fluorescence lifetime can be determined by measuring phase angle values (Thompson et al., 1996). A decrease in fluorescence intensity due to quenching can also used for sensing. In this case, the biorecognition event causes a decrease in fluorescence (quenching) of the fluorescent reporter molecule. The fluorescence decrease is related to the analyte concentration. For example, a dye that undergoes fluorescence quenching when the pH decreases can be coupled to an enzymatic reaction that converts a substrate into an acidic product and results in a pH drop. Thus, the decrease in fluorescence can be correlated to the analyte concentration (see also Section 1.4.1). Fluorescence quenching is also one manifestation of another fluorescence phenomenon used for sensing in bio-optrodes -fluorescence resonance energy transfer (FRET). This phenomenon occurs when two distinct fluorophores are present. If the emission spectrum of one fluorophore overlaps with the excitation spectrum of a second fluorophore, and the two fluorophores are in proximity (<100/~ ), then the excited fluorophore (donor) can transfer energy non-radiatively to the second fluorophore (acceptor). There are two types of acceptors. Quenchers are acceptors that are not fluorescent and therefore 11

Biran and Walt cause the donor simply to decrease its fluorescence emission intensity. Acceptors can also be fluorescent dyes that accept the energy non-radiatively from the donor, and then re-emit the energy at specific emission wavelength. This energy transfer results in an increase in light emission by the acceptor and a decrease in light emission from the donor. When an energy transfer pair of fluorophores is used to label two interacting molecules (e.g., antibody-antigen, enzyme-substrate), they can be used for sensing. Recently, both the donor and the acceptor molecules were incorporated into single biological molecules such as proteins (Hellinga and Marvin, 1998) and nucleic acids (e.g., molecular beacons) (Tyagi and Kramer, 1996; Tyagi et al., 2000). When these sensing molecules are in their native conformation, the donor and the acceptor are in proximity and therefore low fluorescence signals from the donor are obtained. When the molecule interacts with the analyte, conformational changes occur that separate the donor and the acceptor molecules and cause an increase in the fluorescence from the donor (see Section 3.3). The most commonly used fluorescent molecules in bio-optrodes are organic dyes. Recently self-fluorescent proteins have also been used. The sources of these proteins are marine organisms such as the jellyfish Aequorea victoria that produce the green fluorescent protein (GFP) (Chalfie et al., 1994). When GFP is excited, it emits light at a lower energy and therefore at a higher wavelength. GFP is highly fluorescent, with a quantum efficiency of approximately 80% and is very stable to heat and pH (5.5-12). The GFP has been expressed in different cell types (bacteria, yeast, mammalian, plant) and used as reporter gene for different cellular events (Naylor, 1999). In order to allow monitoring of several cellular events simultaneously, several GFP mutants have been developed each with unique excitation and emission wavelengths. Cells expressing fluorescent proteins, and also the purified proteins have been used for constructing different bio-optrodes (see Sections 1.4.1 and 3.3).

Time-resolved fluorescence spectroscopy is another phenomenon used in biooptrodes. This method is based on the fluorescent molecule's excited state lifetime. The light intensity emitted from a molecule excited by a short pulse of light decays exponentially with time. The decay time pattern is unique for each molecule and can be used for analytical purposes. Barker et al. (1999) used this method to improve the performance of a bio-optrode for nitric oxide detection. A different light emission phenomenon used in bio-optrodes is chemiluminescence. In contrast to fluorescence, chemiluminescence is produced when a chemical reaction yields an excited species that emits light as it returns to its ground state. The use of chemiluminescence in biosensors, including fiber optic-based biosensors, was recently reviewed (Aboul-Enein et al., 2000; Gubitz et al., 2001). In many bio-optrodes, the chemiluminescence of luminol is used to generate the light signal. The reaction between luminol and H2Oz produces a

12

Optrode-based Fiber Optic Biosensors

Figure 5. Design of flow-cells incorporating bio-optrodes (Kuswandi et al., 2001). Reproduced with permission of the Royal Society of Chemistry.

luminescence signal and is also catalyzed by certain ions or molecules (e.g., MnO42-, Iz, Cu2+). This reaction can be used, for example, in enzyme-based biooptrodes in which the enzymatic reaction generates H202 (see Section 3.3). Enzymes such as horseradish peroxidase can also catalyze or induce a chemiluminescence reaction by producing H2Oz. In addition, alkaline phosphatase and 13-galactosidase can be used to label biological sensing elements such as antibodies or nucleic acids. In the presence of a 1,2-dioxetane substrate (Bronstein et al., 1996), these enzymes catalyze light formation proportional to the analyte concentration. Chemiluminescence-based bio-optrodes are usually used in conjunction with flow cells. An optical fiber with an immobilized sensing element is placed inside the flow cell and transmits the light signals to the detector (Figure 5). Bioluminescence is a biological chemiluminescent reaction. Many organisms produce bioluminescence for signaling, self-protection, mating, attracting prey and finding food (Campbell and Sala-Newby, 1993). The bioluminescence reaction is catalyzed by the enzyme luciferase and requires the presence of oxygen. The bioluminescent substrate used in this reaction is called luciferin. Different luciferin molecules are used by different organisms. For example, 13

Biran and Walt aldehydes and flavins are used by bacteria and imidazolopyrazines are employed by some fish and squid. B ioluminescence can be applied for analytical measurements in two ways: (1) One can detect cellular events inside living cells by fusing the luciferase gene (e.g., the luc gene coding for firefly luciferase or the lux gene coding for the bacteria Vibrio fischeri luciferase) to the gene of interest. The in-vivo activity of the selected gene can be detected by monitoring the luminescence signal (LaRossa, 1998). (2) Alternatively, one can use purified recombinant luciferase and synthetic luciferin substrates for ex-vivo detection assays for analytes such as ATP, NADH and FMN (Blum et al., 1993). In biooptrodes, the cells or the purified enzymes are immobilized on the fiber tip and the luminescence signals are transmitted through the fiber to the detector.

Absorption is a simpler process than fluorescence and has also been used in biooptrodes. Absorption is a process in which light energy is absorbed by an atom or a molecule, promoting the molecule from the ground energy state to a higher energy excited state. The resulting energy is dissipated non-radiatively (i.e., thermally) to the medium when the excited state relaxes to the ground state. The absorbance changes are related to the concentration [C] via the Beer-Lambert relationship:

A = log(Io/I)= e . [ C ] . l

(4)

where A is the optical absorbance, and Io and I are the intensities of transmitted light in the absence and presence of the absorbing species respectively, 1 is the effective path length, and e is the molar absorption coefficient. In practice, optical fibers are used to measure absorbance by transmitting light through the fiber to the-sensing layer and measuring changes in the scattered light. Alternatively, light is transmitted through one arm of bifurcated optical fiber to the sensing region and reflected light signals are measured through the other arm of the fiber (Figure 6 (b)). In a different configuration, two fibers are placed with one fiber facing the other creating an optical cell in which the distance between excitation and collection fiber is the pathlength.

1.3. Optrode biosensor (bio-optrode) design and instrumentation Different bio-optrode system designs have been used and recently reviewed (Kuswandi et al., 2001; Mehrvar et al., 2000) The design of bio-optrodes is similar to chemical optrode design and two basic configurations are used: (a) a single fiber is used to transmit the light from the light source to the sample region and back to the detector, as shown in Figure 1, or (b) multiple fibers are used in which one fiber is employed to transmit the light to the sample region and the other fiber or fibers are used to transmit light from the sample region to the detector, as shown in Figure 6 (a). For the second configuration, the most common format is a bifurcated fiber. Bifurcated fibers are fabricated by fusing 14

Optrode-based Fiber Optic Biosensors

)

(a)

()

(b)

,,

Lighi Sourcei

)'------~.etec;or ]

,

ight Source [

J ~--'--~Detector. [

(c)

Sensing layer

Figure 6. Design principle of a bio-optrode. (a) Two fibers: one carries light to the sensing layer and one carries the signal to the detector. (b) Bifurcated fiber: the biosensing layer is placed on the fused end of the fiber (c) The biosensing layer is placed on the central fiber and the surrounding fibers are used to collect the light signals.

two fibers on one end leaving the other ends free. The sensing elements are immobilized on the fused side and the other ends of the bifurcated fiber are connected to the light source and to the detector as shown in Figure 6 (b). In a different configuration, multiple fibers comprising one central fiber surrounded by several fibers are employed. The central fiber carries the immobilized sensing elements and is connected to the light source; the surrounding fibers collect the output light signals and transmit them to the detector (Figure 6 (c)). The light sources used for bio-optrodes should provide sufficient light intensity within the sensor wavelength operating range. In addition, the light output should be stable over long time periods since light fluctuations may add noise to the measurement and reduce the sensor sensitivity. The different light sources used in bio-optrodes and their characteristics are summarized in Table 1. In most fiber optic biosensor systems, the light transmitted from the sensing element (output light) is measured by using photon detection devices, which 15

Biran and Walt Table 1. Li ;ht sources Wavelength Characteristics (nm) IR/NIR, visible High power output, bulky, expensive, 200-300 used together with wavelength selection device. 200-1000 470-1300 LOW power Output, high stability, long life, robust, compact size, inexpensive. 377, ~188-568, Monochromatic, very high power 633 output, directional, bulky, expensiv e . 800-904 High power output, long life, narrow spectral band, inexpensive, compact .size. i

Type Ill

Tungsten lamp Deuterium lamp

ii

[]

i

i

I

Ill

....

LEDs

Laser (N2, Ar § He~ Ne) Laser Diodes

I

Detector t~'l~e Photomultipliers (PMT)

Photodiodes (PD) Charge-coupled devices Avalanche photodiodes

Table .2. Light detectors Advantages Sensitive, fast, low noise, internal amplification, compact. Fast, robust, compact, inexpensive Very sensitive, can be used for. imaging Lower noise than PD, fast, sensitive, can tolerate intense illumination

Ill l

I

Limitations Need t~orhigh power voltage supply, destruction by over exposure High noise, no internal amplifier . Slow, expensive, need for a cooling system. More expensive than PD

absorb photons and convert them into electrical signals. Several photon detectors are available as shown in Table 2. 1.4. Biological sensing elements Bi0-optrodes are constructed by immobilizing biological recognition components, such as enzymes, antibodies, nucleic acids, or cells to optical fibers. In nature, interactions between biological molecules, such as receptor-ligand, antibody-antigen or two complementaryDNA strands, are highly specific. Some of these recognition molecules can be purified and used in fiber optic biosensors. Moreover, by using genetic engineering, the original recognition element's structure can be modified and designed for a specific analytical application (Hellinga and Marvin, 1998). Biological sensing compounds can be divided into 16

Optrode-based Fiber Optic Biosensors two main categories based on their bioactivity: biocatalysts (enzymes and cells), and bioaffinity molecules (antibodies, receptors, and nucleic acids).

1.4.1. Biocatalyst-based optrodes. Enzymes are proteins that selectively bind and catalyze the conversion of a substrate to product. Enzymes are used as sensing elements in bio-optrodes based on their ability to bind specific substrates (e.g., the analyte) and catalyze their conversion into an optically detectable product (Kuswandi et al., 2001). The optical signal obtained, e.g., absorbance or fluorescence, is proportional to the product concentration and consequently, to the analyte concentration. Products that possess intrinsically optical properties can be measured directly, but the most common enzymatic reactions products, such as H § ammonia, oxygen, carbon dioxide and hydrogen peroxide, do not possess optical properties and are therefore measured indirectly by using indicators (Wolfbeis, 1997). The indicators change their optical properties when interacting with these products. For example, fluorescein is a pH indicator and its emission intensity can be correlated to changes in H + concentration. Other indicators employed in enzyme bio-optrodes were recently reviewed (Kuswandi et al., 2001). An interesting example that demonstrates the simple fabrication and function of enzyme-optrodes is the one used for glucose detection based on the enzyme glucose oxidase. Glucose oxidase catalyzes the oxidation of glucose with oxygen to produce gluconolactone and H202.

Glucose + 0 2

Glucose oxidase .

v

Gluconic acid

+H202

Two approaches have been employed to determine the glucose concentration with the enzyme: (1) measuring the amount of oxygen consumed in the enzymatic reaction using a ruthenium complex as an indicator (Rosenzweig and Kopelman, 1996a, 1996b) or (2) measuring the amount of H202 produced using a chemiluminescence indicator (Marquette et al., 2000). In many cases, a sequence of enzymatic reactions is required to detect a specific analyte. In order to fabricate bio-optrodes for detection of such analytes, two or three enzymes are immobilized together on the optical fiber in such a way that sequential reactions can occur. The first enzyme catalyzes the conversion of the analyte to a product that serves as a substrate for subsequent enzymatic reactions that eventually convert the initial analyte to an optically detectable product (Michel et al., 1998a, 1998b). Using this methodology, analytes that could not be detected in a single reaction step can be detected. In addition, coimmobilizing two enzymes can achieve signal amplification through enzyme recycling systems as shown in Figure 7 (Zhang et al., 1997). 17

Biran and Walt

-.. -

"

' . ~..iFiberfi~ .

NADH~

. "..i..:"...."...".!ii.:I.......i.i......."ii[..i....................

.

S

Pyruvate~

S

H202

Biocatalytic Layer NAD §

Lactate S

L ~ ~ Pyruvate O~

NADH

02 Bulk Solution

Figure 7. Schematic diagram of signal amplification using a dual-enzyme bio-optrode. Pyruvate is detected using lactate dehydrogenase (LDH) and lactate oxidase (LDO), which are co-immobilized on a fiber optic tip. Pyruvate concentration is determined by measuring NADH fluorescence. Pyruvate and NADH diffuse from the bulk solution into the enzyme layer, LDH catalyzes the formation of lactate and NAD§ during the reduction of pyruvate. LDO then catalyzes the regeneration of pyruvate causing additional consumption of NADH by the LDH-catalyzed reaction. Thus, the signal obtained using a dual-enzyme system is higher then when a single enzyme is used (Zhang et al., 1997). Reprinted with permission from Elsevier Science.

Inhibition of enzymatic reactions can also be used as a sensing mechanism in bio-optrodes (Freeman and Bachas, 1992). In this approach, the inhibitor is the analyte and the measured signal is the decrease in enzymatic activity. One example is detection of organophosphate and carbamate pesticides using an enzyme inhibition-based optrode. The bio-optrode is based on the inhibition of acetylcholinesterase (ACHE) by organophosphate pesticides. The enzyme is coimmobilized together with a pH sensitive dye at the fiber's distal end. The substrate acetylcholine is hydrolyzed by AChE causing a change in the local pH and thereby the fluorescence signal. The inhibition of this reaction can be correlated to the pesticide concentration in the sample (Doong and Tsai, 2001; Hobel and Polster, 1992). In living cells, cellular functions are carried out by enzymes that simultaneously catalyze numerous biochemical reactions. Some enzymatic activities that occur in cells have been applied for bio-optrode fabrication. Although enzymes can be isolated and purified, their activity outside the cells is usually reduced compared to their activity within the cells where they function in an optimum environment containing all the necessary cofactors. Whole cell biocatalysts are particularly advantageous when the detection is based on a sequence of multiple enzymatic 18

Optrode-based Fiber Optic Biosensors reactions. These enzymatic cascade reactions are very difficult and complicated to accomplish ex-vivo by coimmobilizing the enzymes but are relatively straightforward when employing whole cells. In practice, whole cells that produce Unique or enhanced enzymatic activity and can transform the analyte (substrate) into detectable products or cells that produce cellular responses such as changes in oxygen consumption are immobilized on optical fibers (Preininger et al., 1994). The methods for detecting the products in cell-based fiber optic biosensors are similar to those employed in enzyme optrodes. In a more recent approach, cells are genetically engineered to over-express specific enzymes involved in the analytical measurement. An example of this approach is the use of E. coli cells that were engineered to over-express the enzyme organophosphorus hydrolase (OPH) on their outer cell membrane (Mulchandani et al., 1998). This enzyme catalyzes the hydrolysis of organophosphorus pesticides to form a chromophoric product that can absorb light at a specific wavelength. The cell optrode is fabricated by immobilizing the cells on a bifurcated fiber optic tip and using a photomultiplier detection system to measure the light signals. Although the specificity of whole cell optrodes is reduced compared to enzyme optrodes, cells are very simple to use and obtain (e.g., growing the cells for a few hours), and there is no need for purification steps, which makes cell bio-optrodes inexpensive to assemble. A different approach for sensing with whole cells, which does not directly involve biocatalysis, is based on utilizing genetic responses and signal transduction mechanisms in living cells (Daunert et al., 2000; Kohler et al., 2000; Naylor, 1999). Cells may express a specific gene or set of genes when a specific molecule (e.g. analyte) is present in the cell's environment. By fusing reporter genes, encoding for optically detectable enzymes or proteins (e.g., luciferase, [3galactosidase, GFP) to the responsive gene, the genetic response is measured and correlated to the analyte concentration. For a more detailed description of this approach, see Chapter 10 in this book.

1.4.2. Bioaffinity-based optrodes. The natural high selectivity of antibodies, receptors and nucleic acids make them very powerful sensing elements for recognizing their binding partners. Such bioaffinity optrodes are used as probes because the recognition reaction is essentially irreversible. The bio-optrode sensing elements must be regenerated or recharged before the probe can be used to make another measurement. In many cases, a probe-based bio-optrode configuration involves the use of a permanent fiber optic and a disposable sensing layer that can be placed on the fiber optic's distal end (Figure 8). lmmuno/receptor optrodes are a major group of bioaffinity fiber optic biosensors based on transducing antibody-antigen (analyte) interactions into an optical signal that is proportional to the antigen concentration. Monoclonal antibodies that can recognize a specific antigenic epitope region (i.e., a specific spatial

19

B iran and Walt

Figure 8. Configuration of probe-based bio-optrodes with disposable biosensing elements. (a) Biorecognition sensing molecules immobilized on membrane, which is held by a screw cap on the optical fiber tip. (b) Disposable glass slide with gel entrapped enzyme. (c) Nylon membrane, with immobilized sensing molecules, attached to the fiber using an O-ring (Kuswandi et al., 2001). Reproduced with permission of the Royal Society of Chemistry.

structure on the antigen molecule) or polyclonal antibodies that recognize different antigenic epitopes are used in immuno-optrodes. Several detection schemes are employed; the simplest scheme involves the detection of intrinsically fluorescent analytes such as polynuclear aromatic hydrocarbons (PAHs) (Vo-Dinh, et al., 2000). Antibodies are immobilized on the fiber surface and a fluorescence signal is obtained when the analyte (antigen) binds to the optrode's surface as shown in Figure 9(a). A competition assay is a more generalized detection scheme that can be applied to any antibody antigen pair. The detection is based on competition for the antibody binding site between the antigen present in the sample (analyte) and an externally added fluorescent-labeled antigen as shown in Figure 9 (b). A known concentration of fluorescent-labeled antigen is added and captured by an antibody, which is immobilized on the optical fiber surface. The fluorescence signal obtained is measured and set as the initial signal. To perform an analysis, the same fluorescent-labeled antigen concentration is mixed with a sample containing an unknown antigen concentration. When this mixture is analyzed using the bio-optrode, the resulting fluorescence signal obtained is lower than the initial signal because of competition with the labeled antigen in the sample. The relative decrease in the initial signal is proportional to the analyte concentration in the sample. Using this detection scheme, bio-optrodes for the detection of different analytes have been developed (Wittmann et al., 1996; Zhao et al., 1995). 20

Optrode-based Fiber Optic Biosensors

(a)

(b) Optical fiber

~-~ Optical

(c) Optical fiber

Optica

' ~ Self-fluorescingantigen 0 Antigen Antibody

~

Fluorescentdye ,.

Figure 9. Schematic principle of immuno bio-optrodes. (a) Detection of intrinsicallyfluorescent molecules using immobilized antibodies. (b) Competition assay using a fluorescent-labeled antigen. (c) Sandwich immunoassay using an immobilized antibody and a fluorescent-labeled antibody.

The preferred detection scheme is the sandwich immunoassay, which involves the use of two antibodies. The first antibody is immobilized to the fiber and used to capture the antigen while a second antibody, which is labeled by a fluorescent dye or enzyme, is used to generate the signal (Figure 9(c)). The competition and sandwich assays require using labeled antigens or antibodies. Fluorescent molecules and enzymes are employed for labeling using different chemistries (Wortberg, 1997). Very low concentrations of enzymes can be detected based on their enzymatic activity. The enzymes used for labeling, such as alkaline phosphatase, catalyze the conversion of a non-fluorescent substrate to a fluorescent product and can be detected by monitoring the fluorescent signal generated (Michael et al., 1998). Other enzymes, such as horseradish peroxidase, can catalyze chemiluminescence reactions and are detected by monitoring the emitted light signals (Aboul-Enein et al., 2000; Diaz et al., 1998; Gubitz et al., 2001; Spohn et al., 1995). Enzyme labeling is more sensitive than fluorescent dye labeling since the signal is amplified by the enzymatic reaction. Another new technology to increase the labeling efficiency 21

Biran and Walt

Figure 10. Principle of DNA fiber optic biosensors. (a) Single strand DNA probe molecules, with a sequence complementary to the target DNA sequence, are immobilized onto the fiber. (b) The fluorescent-labeled sample DNA molecules are first dehybridized and the fiber is dipped into the sample solution. (c) After hybridization, the complementary strands of the target DNA are attached to the probe DNA on the fiber and a fluorescence signal is obtained. of biological molecules was recently proposed and will be discussed in Chapter 17.

Nucleic acid-based optrodes are the second major group of bioaffinity-based optrodes. Nucleic acid base pairing is used as the sensing mechanism in biooptrodes for nucleic acid detection. The presence of a specific DNA sequence, the "target", among millions of other different sequences is detected by hybridization to its complementary DNA sequence, the "probe", which is immobilized on the optical fiber, as shown in Figure 10. In a typical assay, the target DNA is first amplified and fluorescently labeled using fluorescent primers and the polymerase chain reaction (PCR). The resulting fluorescent double stranded DNA molecules are dehybridized (usually by heating) (Figure 10 (b)) and then allowed to rehybridize (by cooling) to the single strand DNA probe molecules immobilized on the fiber surface (Figure 10 (c)). The excess DNA molecules are washed away, and if the complementary target DNA sequence is present in the sample, a fluorescence signal is detected (Ferguson, et al., 1996). For example, the target sequence can be a unique sequence found only in specific 22

Optrode-based Fiber Optic Biosensors pathogenic bacteria (Iqbal et al., 2000; Pilevar et al., 1998). The target DNA can be easily extracted from water, wastewater or clinical samples, and the presence of pathogenic microorganisms can be determined by the bio-optrode. Recently, new groups of nucleic acid molecules, such as aptamers (Lee and Walt, 2000) and molecular beacons (Liu et al., 2000; Steemers et al., 2000), were incorporated as sensing molecules into bio-optrodes. These different DNA sensing schemes can be multiplexed by fabricating an array of hundreds to thousands of probes as will described later in Section 3.2.

1.5. Sensing element immobilization Immobilization of sensing biomolecules to the optical fiber is a key step in biooptrode development. A good immobilization method should be simple, fast and durable but, more importantly, it should be gentle so the biological molecule being immobilized can retain its biochemical activity. In addition, biological recognition elements are often coimmobilized together with indicator dyes, so that ideally the immobilization method should be suitable for both molecules. In some cases, the recognition compounds are immobilized directly to the optical fiber surface. Alternatively, the molecules are first immobilized on membranes, such as cellulose acetate or polycarbonate that are later physically attached to the optical fiber (Figure 8). There are three main methods for immobilizing a biological sensing compound: adsorption/electrostatic, entrapment, and covalent binding. A schematic representation of these methods is shown in Figure 11. Adsorption immobilization methods involve adsorbing the sensing material onto a solid surface or polymer matrix. Sensing materials can be adsorbed directly on the fiber optic end. This immobilization method is very simple; however the adsorbed molecules tend to gradually leach from the solid support, decreasing sensing performance and/or lifetime. In order to overcome leaching problems, the solid support surface may first be modified with complementary functional groups. For example, a hydrophobic surface can be prepared to immobilize a hydrophobic species. Electrostatic interaction can also be employed for immobilization. This immobilization scheme is based on interaction between oppositely charged molecules. For example, an optical-fiber surface can be coated with a positively charged layer (i.e., using poly-L-lysine) that interacts electrostatically with negatively charged recognition molecules (Figure 11 (a)). The electrostatic immobilization method is very easy and highly reproducible but may be affected by changes in the medium pH or by changes in other ion concentrations. Entrapment immobilization involves physical entrapment of sensing biomolecules within a porous matrix (Figure 11 (b)). The biomolecules are suspended in a monomer solution, which is then polymerized to a gel causing the molecules to be entrapped. Such polymers can be either thermally or

23

Biran and Walt

Figure 11. Schematic diagram of three different immobilization techniques employed in biooptrodes. (a) Absorption/electrostatic. (b) Entrapment. (c) Covalent immobilization.

photochemically initiated and attached to the fiber surface by dip-coating procedures (Healey et al., 1995). The immobilized molecules usually do not leach out of the matrix and can retain their biochemical activity. Polyacrylamide gels are most commonly used for entrapment immobilization, although agarose and calcium alginate gels have also been used (Polyak et al., 2001). One important limitation of this approach is the slow diffusion rates of the analytes and products through the immobilization matrix, which increases the bio-optrode response time. Optically transparent sol-gel glasses are also used for biological sensing molecule entrapment as described in Chapter 14 (Dunn et al., 1998, 2001; Jordan et al., 1996). Sol-gel glasses are produced by hydrolysis and polycondensation of organometallic compounds, such as tetraethyl orthosilicate (Si(OCH3)4). The sensing biomolecules are added to the reaction mixture during the formation of the sol or gel. Sol-gel glasses prepared by this method contain interconnected pores formed by a three-dimensional SiO2 network. As a result, the biomolecules and dyes are trapped but small analytes can readily diffuse in and out of the pores. The main advantages of the sol-gel glass immobilization method are the chemical, photochemical and mechanical stability of the immobilized layer. 24

Optrode-based Fiber Optic Biosensors Disadvantages of sol-gel glass immobilization are the slow response times in aqueous media and the fragility of thin sol-gel glass films compared with polymer films. Functional groups in the sensing biomolecules can be covalently bound to reactive groups on the surface of optical fibers allowing robust immobilization (Figure 11 (c)). The fiber surface can be chemically modified using silanization reactions (Weetall, 1993). For example, the tiber surface can be aminosilanized to form amine functional groups on the fiber surface followed by reaction w i t h COOH groups on the enzyme or antibody. Amine-modified surfaces can also covalently bind to the biomolecule's amine groups using bifunctional crosslinkers such as glutaraldehyde. Covalent immobilization methods are usually more complicated and time-consuming compared with the other immobilization techniques, but are very reliable since the biomolecules and dyes are not likely to leach out. It should be noted that covalent binding might change the biomolecule activity. In some cases, if the binding occurs at crucial sites (e.g., an enzyme active site or an antibody binding site), activity can be lost completely. To avoid such inactivation, substrate, inhibitors and other effectors are often included in the immobilization medium to protect the active or binding site of the biomolecules. In recent years, new techniques have been developed which enable the immobilized molecule's orientation on the sensing surface to be controlled resulting in an increase in the immobilization efficiency (Sackmann, 1996). A more generalized and widely used binding method involves the use of avidinbiotin chemistry (Wilchek and Bayer, 1990). The fiber surface can be modified with biotin groups and bind avidin-modified biomolecular conjugates or vice versa. This method is very attractive since many biotin -~ or avidin-labeled enzymes, antibodies and nucleic acids are commercially available.

2. History Optical fiber-based biosensors evolved from chemical optrodes. The first optical fiber-based chemical sensor was developed by Lubbers and Opitz (1975). Their device was designed to measure CO2 and 02 and was used in biological fluids. A few years later, biological molecules were coupled to the optical fiber-based chemical sensors and bio-optrodes were formed. One of the first bio-optrodes involved coupling the enzyme glucose oxidase to an Oz optrode to fabricate a glucose biosensor (Arnold, 1985). In the following years, many bio-optrodes with different recognition molecules were developed and reported in several books (Blum et al., 1994; Wolfbeis, 1991), and reviews (Aboul-Enein et al., 2000; Fraser, 1995; Mehrvar et al., 2000; Rabbany et al., 1994; Wolfbeis, 2000). Although the bio-optrode basic configuration has not changed much from the one proposed by Lubbers and Opitz (1975), new types of optical fibers, optical instruments, biorecognition molecules and indicators have been integrated into 25

Brian and Walt bio-optrodes. These materials, combined with new immobilization techniques and advanced optical approaches, led to the development of more sophisticated, selective and sensitive bio-optrodes. Advances in two fields influenced biooptrode development in the last decade. First, development of new fiber optic technologies that were developed for telecommunication applications. Second, advances in molecular biology techniques allow specific biorecognition molecules to be designed. Integration of technologies from these two fields has led to the development of advanced bio-optrode technologies such as multianalyte bio-optrodes, reagentless bio-optrodes and nano bio-optrodes.

3. Advanced Bio-Optrode Technologies and Applications In this section, a few examples of new bio-optrode technologies and applications will be described. Although many novel and interesting papers related to biooptrode developments have been published in recent years, we focus here on a few examples that emphasize the diversity of existing bio-optrode technologies. In addition, a few examples of bio-optrode applications in the industrial, environmental and clinical fields will be described.

3.1. Nano bio-optrodes One of the most exciting advances in bio-optrode development is the miniaturization of sensors to submicron dimensions. Nanotechnology facilitates research in this field and leads to development of new nano bio-optrodes (Cullum and Vo-Dinh, 2000). The main importance of such biosensors is their ability to monitor biomolecule concentrations inside a single living cell and thereby expand our knowledge about complex intracellular process. In order to prepare nano bio-optrodes, optical fibers a few nanometers in diameter are fabricated. The fabrication process involves pulling optical fibers with an initial diameter of a few microns using a modified micropipette puller optimized for optical fiber pulling. After pulling, tapered fibers are formed with typical distal end (tip) diameters of 20-80 nm. This technique was used by Kopelman and coworkers to make a nano fiber optic chemical sensor for monitoring intracellular pH inside living cells (Tan et al., 1992). Changes in pH were measured by immobilizing a pH sensitive dye to the fiber tip. The same design was used to prepare an enzyme-based nano bio-optrode for nitric oxide detection (Barker et al., 1998). Fluorescently labeled cytochrome c', which undergoes conformational changes in the presence of NO, was immobilized to the fiber tip. Changes in NO concentrations were correlated to changes in the energy transfer between cytochrome c' and the fluorescent dye.

26

Optrode-based Fiber Optic Biosensors

Figure 12. Nano bio-optrodes. (a) Fabricating a nano fiber optic tip. An optical fiber is heated and pulled and a tapered end with submicron diameter is formed. The tapered fiber side walls are then coated with a thin metal layer, using thermal evaporation, in order to prevent excitation light leakage. Biorecognition molecules can be immobilized on the fiber tip (Vo-Dinh, et al., 2000). Reprinted with permission from Nature Biotechnol. (b) Scanning force micrograph (SFM) of nano fiber (Vo-Dinh, et al., 2001). Reprinted with permission from Elsevier Science.

An antibody-based nano bio-optrode for the fluorescent analyte benzo[a]pyrene tetrol (BPT) was also fabricated for detection inside a single living cell (Vo-Dinh et al., 2000). The nano bio-optrode was prepared by coating the tapered fiber's outside walls with a thin silver, gold or aluminum layer using a vacuum evaporator as shown in Figure 12 (a). In this system, the fiber is held at an angle relative to the metal vapor, resulting in a coating on the side of the fiber and leaving the tip uncoated. This coating prevents light leakage from the fiber's walls and helps to get maximum light intensity to the fiber tip. The fiber's uncoated tip surface was then silanized in order to covalently attach anti-BPT antibodies. The final nano bio-optrode tip diameter was 200 to 300 nm. Biooptrodes of this size have several advantages over larger bio-optrodes including fast response time and higher sensitivity. Using BPT nano bio-optrodes, BPT concentrations as low as --300 zeptomoles were detected (Vo-Dinh et al., 2001). 27

Biran and Walt

Figure 13. Measurements inside a single live cell using a nano bio-optrode. (a) The optical measurement system. (b) A nano bio-optrode inside a single cell (Vo-Dinh et al., 2001). Reprinted with permission from Elsevier Science.

The optical measurement system used with the nano bio-optrode is shown in Figure 13(a). Laser light is transmitted through the fiber and used to excite the captured BPT molecules. Changes in fluorescence signals due to the presence of bound BPT molecules are transmitted through the microscope objective and measured using a PMT. Using this experimental set-up, BPT molecules inside single living cells were measured. The fiber's tip was inserted into the cell (Figure 13(b)) and incubated for 5 minutes inside the cells to allow the antibodies to bind the antigen (BPT). The fiber was then removed from the cell and the fluorescence signal obtained from the bound BPT was immediately measured. Concentrations as low as 9.6 x 1011M were measured inside the cells. The ability to measure concentrations of specific analytes inside single living cells with nano bio-optrodes can lead to a better understanding of many cellular processes such as transport mechanisms through cellular membranes, signal transduction pathways, complex enzymatic reactions and even gene expression. 28

Optrode-based Fiber Optic Biosensors

3.2. Multi-ana|yte sensing One of the main challenges of any sensor device is to detect several analytes simultaneously. Multi-analyte sensing is important for clinical, environmental, and industrial analysis. For example, measuring the presence of proteins, antibodies, DNA sequences, antibiotics, viruses and bacteria in single blood samples can provide physicians with rapid and comprehensive information about a patient's medical condition. Several approaches have been described for multianalyte bio-optrode fabrication (Anderson et al., 2000; Healey et al., 1997a; Li and Walt, 1995; Michael et al., 1998). The conventional approach to preparing multi-analyte sensors is to simply bundle multiple individual optical sensors. In this approach to multi-analyte sensing, several optical fibers are assembled, each containing a different immobilized biorecognition molecule on a single fiber bundle. This approach was used to fabricate multi-analyte biosensors for detecting different DNA target sequences simultaneously (Ferguson et al., 1996). Eight optical fibers, each with a different immobilized DNA probe, were bundled together as shown in Figure 14 (a). The bundled fiber's distal end was inserted into the sample solution containing a fluorescein isothiocyanate -labeled oligonucleotide with a sequence complementary to one of the probe sequences. The fluorescence signals were measured from the fiber's proximal end. Figure 14 (b) shows that when only one target sequence is present, a signal is obtained only from the fiber (bright signal) that contains the complementary probe sequence, while the rest of the fibers in the bundle do not respond. When several target sequences were present, signals from several fibers carrying the complementary probes were observed (Figure 14 (b)). In different work, the specificity of this approach was demonstrated (Healey et al., 1997b). Two probes were prepared, one that was complementary to the HRas oncogene sequence and a second probe containing a similar sequence but with a single base-pair mismatch. When the hybridization reaction was performed at low temperature, both sequences hybridized to the probe, but at high temperature, only the wild type sequence hybridized. This experiment shows that these sensors can be used for point mutation detection. The same sensor configuration can be applied for different sensing elements such as antibodies, enzymes or whole cells. This approach, theoretically, is not limited in the number of individual fibers (each with a different sensing chemistry) that can be used simultaneously; however the array size grows as more sensing elements are added. An alternative approach involves fabrication of discrete sensing regions, each containing different bio-sensing elements, at precise spatial locations on an imaging fiber's distal end (Figure 15 (a)). The sensing regions are formed using photopolymerization techniques (Pantano and Walt, 1995). The imaging fiber's proximal end is first prefunctionalized with a polymerizable silane. The fiber is 29

Biran and Walt

Figure 14. Multianalyte bio-optrode for oligonucleotide detection. (a) Schematic diagram of bio-optrode setup. Individual optical fibers, each with a specific immobilized oligonucleotide probe sequence are bundled together. The fiber's distal end is incubated with the sample and the signals obtained at the proximal end are measured using a CCD detector. (b) Fluorescence images acquired after incubating the multianalyte bio-optrode in solutions containing different target sequences. Image F show the bio-optrode response to the presence of three different targets in the sample (Ferguson, et al., 1996). Reprinted with permission from Nature Biotech.

then dipped into a solution containing monomer, cross-linker, indicators, photoinitiator and the sensing biomolecules. Using a pinhole, light is focused onto a small area (-30 ~tm in diameter) on the imaging fiber's proximal end. Light travels through the imagin fiber, from the illuminated are at the proximal end to the distal end. At the distal end, the light activates a photinitiator and the polymer layer is formed only at the illuminated area (Figure 15(b)). For the formation of the next sensing polymer, light is focused on a different area at the proximal end and the fiber's distal end is dipped into a polymerization solution containing different sensing biomolecules. Initially this approach was used to fabricate a multi-analyte sensor for pH, CO2 and 02 by forming sensing regions with different fluorescent dyes on a single optical imaging fiber face (Ferguson et al., 1997). Based on this initial work, multi-analyte biosensors for detecting penicillin and pH were developed (Healey and Walt, 1995; Healey et al., 1997a). This sensor incorporated two sensing regions; in one region the enzyme penicillinase was immobilized together with a pH indicator, and in the second region only the pH indicator was immobilized. In the presence of penicillin, the penicillinase activity results in the formation of H § and therefore a decrease in the local pH in the polymer's microenvironment. By simultaneously monitoring pH changes in both sensing regions (with and without 30

Optrode-based Fiber Optic Biosensors

Figure 15. Multianalyte bio-optrode with different biosensing elements immobilized in polymers attached to an imaging fiber. (a) Setup of photopolymerization procedure used to fabricate the bio-optrode. Reprinted from Pantano and Walt (1995) with permission from the American Chemical Society. (b) Scanning force micrograph of immobilized sensing polymer on an imaging fiber (Ferguson et al., 1997). Reprinted with permission from Elsevier Science.

the enzyme), the changes related to the enzymatic activity can be discriminated from pH changes in the bulk solution. Thus, this dual sensor is able to detect penicillin and can account for changes in the solution pH (Figure 16). A similar approach was used to fabricate glucose and 02 biosensors. The enzyme glucose oxidase was used and the depletion of 02 in the presence of glucose was monitored (Li and Walt, 1995). A separate sensor for 02 was also prepared on the same imaging fiber. When the glucose sensor signals were compared with the signals obtained from the sensing region that contained only the 02 indicator, the concentration of glucose could be determined. Both biosensors can be used to determine the analyte concentrations in different environments. In addition, they can provide information about both the biochemical analytes and pH or 02 concentrations using a single imaging optical fiber. A possible future application for such biosensors may be for in-vivo multi-analyte sensing, where early changes in drug levels, glucose, 02, and pH are important. 31

B iran and Walt

Figure 16. Imaging fiber-based penicillin and pH bio-optrode. (a) Response of biooptrode, similar to the one described in Figure 15, with penicillin-sensitive polymer regions (containing the enzyme penicillinase) and pH-sensitive polymer regions. When the penicillin concentration is increased, only the fluorescence intensity from the penicillin-sensitive polymer regions increases. (b) Bio-optrode response to penicillin (solid squares) and pH (empty squares) are shown in the left plot. The difference between the buffer pH and the microenvironmental pH at the penicillin-sensitive polymer is shown in the right plot (Healey and Walt, 1995). Reprinted with permission from the American Chemical Society.

In both of these approaches (sensor bundling or photopolymerization), when more then twenty optical fibers or polymer regions are required, the bundle of fibers becomes too big or the photopolymerization protocol becomes complicated. A new approach that overcomes this limitation was recently proposed (Michael et al., 1998; Walt, 2000). This approach is based on using the unique characteristics of optical imaging fibers (see Section 1.1). Imaging fibers consist of thousands of optical fibers coherently bundled together, with each individual fiber maintaining its ability to carry its own light signal from one end of the fiber to the other. Thus, by attaching a sensing material to the individual fiber's distal end, an array of thousands of sensing elements can be constructed on the tip of a single imaging fiber array. In practice, microwells are fabricated on the end of each individual fiber by selectively etching the fiber cores. This 32

Optrode-based Fiber Optic Biosensors

Figure 17. High-density multianalyte bio-optrode composed of microsphere array on an imaging fiber. (a) Scanning force micrograph (SFM) of microwell array fabricated by selectively etching the cores of the individual fibers composing the imaging fiber. (b) The sensing microspheres are distributed in the microwell. (c) Fluorescence image of a DNA sensor array with ~ 13,000 DNA probe microspheres. (d) Small region of the array showing the different fluorescence responses obtained from the different sensing microspheres (Walt, 2000). Reprinted with permission from the American Association for the Advancement of Science. process results in the formation of a high density microwell array on the imaging fiber tip as shown in Figure 17(a). The sensing elements are prepared by immobilizing fluorescent indicators and/or biorecognition molecules to the microsphere surfaces. The microspheres and microwells are matched in size such that the microspheres can be distributed into the microwells to form an array of sensing elements (Figure 17(b)). When different biorecognition molecules are immobilized on different microspheres, the array can be used to detect multiple analytes. A CCD detector is used to monitor and spatially resolve the fluorescence signals obtained from each microsphere (Figure 17(c) and (d)). Imaging and data analysis software are used to calculate the analyte concentrations. These sensor arrays are prepared by randomly distributing the microspheres into the wells. In order to allow multi-analyte sensing, the location of each sensing microsphere must be determined. The microsphere registration process involves 33

Biran and Walt

Figure 18. Randomly ordered array bio-optrode. (a) Schematic representation of the biorecognition elements immobilized on different sets of encoded microspheres (APalkaline phosphatase). The microspheres were encoded using three different ratios of two fluorescence dyes: indodicarbocyanine (DilC) and Texas red cadaverine (TRC), both dyes are excited at 577 nm and emit at 670 nm and 610 nm respectively. (b) The three rnicrospheres types are mixed and randomly distributed into the microwell array. Fluorescence responses in the presence of the AP fluorogenic substrate, avidin-FITC and biotin-FITC are shown on the top images. The identity of each microsphere was determined by calculating the emission ratio 670 nm/610 nm obtained using 577 nm excitation light (bottom images) (Michael et al., 1998). Reprinted with permission from the American Chemical Society.

using one of several encoding/decoding schemes. When the microspheres are prepared, each type of microsphere is modified such that it carries a unique optical marker in addition to the biorecognition element. This marker can be a fluorescent dye or a combination of several different fluorescent dyes. Different markers are used for the different microsphere types, allowing each of the microspheres carrying a certain type of biomolecule to be encoded with a unique optical signature. For example, as shown in Figure 18, three types of microspheres were prepared by immobilizing the enzyme alkaline phosphatase to one group of microspheres, avidin to the second group, and biotin to the third group (Figure 18(a)). Each type was encoded with different concentrations of 34

Optrode-based Fiber Optic Biosensors

Figure 19. Molecular beacons (MB) structure. (a) The hairpin structure is formed due to the complementary sequences near the 3' and 5' ends. The single strand "loop" contains the probe sequence. In this configuration, the fluorophore and quencher are in proximity and therefore no fluorescence signal is produced. (b) When a target sequence binds, the MB structure changes causing separation of the fluorophore and quencher resulting in a fluorescence signal change. two fluorescent dyes. When the three microsphere types were mixed and randomly distributed into the microwell array, their location could be determined by applying the appropriate excitation and emission wavelengths to establish the different fluorescent markers on each bead. This biosensor was used for multianalyte detection of fluorescein diphosphate, biotin-FITC and avidin-FITC, as shown in Figure 18(b). For each analyte, several different microspheres produced fluorescence emission signals, indicated by the bright spots on the array images. These images demonstrate two main advantages of this technology. First, the presence of replicates of each microsphere type provides statistically significant results and reduces the possibility of both false negatives and false positives. Second, averaging signals from many identical individual sensing elements results in higher signal/noise ratios. This multi-analyte biosensor design was also used to develop a DNA biosensor with the ability to detect 25 different fluorescently labeled DNA sequences simultaneously (Ferguson et al., 2000). Another biosensor, comprising microspheres with different immobilized molecular beacons, was used to detect three different unlabeled DNA sequences (Steemers et al., 2000). Recently, microspheres with immobilized antibodies 35

Biran and Walt were used for simultaneous detection of the clinically important drugs digoxin and theophylline (Szurdoki et al., 2001). Multi-analyte bio-optrodes are in the first stages of research and development. Due to their importance for many analytical applications, it is expected that research efforts will continue to advance the capabilities of such sensors.

3.3. Reagentless bio-optrodes for homogeneous assay One limitation common to many bio-optrode technologies is the need to add external reagents to the analytical assay. For example, when antibodies are used as recognition molecules in a sandwich assay, there is a need to add secondary labeled antibody in order to measure the analyte concentration (Figure 9 (c)). The same requirement applies to a competition assay where a labeled antigen is used (Figure 9 (b)). Most nucleic acid bio-optrodes are based on pre-labeling the target sequence with fluorescent dye. The necessity to add reagents complicates the assay procedure and limits the acceptance of bio-optrodes as standard and simple analytical tools. Therefore, many research efforts have concentrated on developing bio-optrodes for "mix and measure" assays where no reagents are added. In this section, several approaches for reagentless (also called homogeneous) bio-optrode fabrication will be described. One approach for reagentless bio-optrode fabrication is based on monitoring conformational changes in the biorecognition molecule following analyte binding. The conformational changes are usually detected using FRET as the transduction mechanism. In one example, molecular beacons (MB) wereused to detect unlabeled DNA sequences (Steemers et al., 2000). Molecular beacon structures consist of single stranded DNA in a hairpin configuration with a fluorophore and quencher attached to opposite termini (Tyagi and Kramer, 1996). The molecule's 3' and 5' ends are complementary to one another and form the hairpin structure. The probe sequence, which is complementary to the target sequence, is located in the center (Figure 19(a)). In the absence of target, the fluorophore and quencher are within the requisite energy transfer distance, resulting in fluorescence quenching (Figure 19(a)). Upon target binding, a conformational change occurs, the hairpin separates (denatures) and the fluorescence signal increases (Figure 19(b)). Using an imaging fiber-based MB bio-optrode, three different sequences from mutant genes related to cystic fibrosis were simultaneous detected (Steemers et al., 2000). The multi-analyte imaging fiber-based bio-optrode was prepared as previously described in Section 3.2. Each type of MB probe was immobilized to beads that were encoded with unique optical signatures. The resulting three types of beads were randomly distributed into a microwell array and used for the analysis of three different target sequences simultaneously.

36

Optrode-based Fiber Optic Biosensors

W

0

.

i .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

._

.~..

-

-_

_:,

_.

_

-

....

.

8 -5 \

N

o CM

2 0 0 nM Co

~Q %

-IO

%

I'-% %

-15

I~uM H64C- CY3-a~poCA FIBER: 20OHM CORE SILICA 5% GEL, 0.7ram THICK EXC514nm EM .57Ohm+

J

-20

-25t

0

.... i t

_

% %

'o 2,,uM Co

J

__n

..... n

....... I

Z

3

4

5

,

-

I ......

6

TtME a MINUTES

Figure 20. Reagentless Co 2§ bio-optrode. The enzyme carbonic anhydrase was labeled with a fluorescent donor and immobilized to the fiber optic distal end. When Co 2+ binds to the enzyme, the donor's fluorescence is quenched and the signal decreases (Thompson et al., 1996). Reprinted with permission from Elsevier Science.

In a similar manner, donor and acceptor molecules can be incorporated into proteins and used as reporters for substrate binding events. In one approach, the enzyme carbonic anhydrase, which binds metal ions with high affinity and selectivity, was used to fabricate Zn 2§ Co 2§ and Cu 2§ bio-optrodes (Thompson et al., 1996; Thompson and Jones, 1993). Donor molecules, such as Cy-5 or Cy-3 dyes, were bound to primary amines in the protein, using N-hydroxysuccinimide esters of the dyes as modification reagents. The acceptor molecules in this case were the Co 2+ and Cu 2§ analytes themselves, which exhibit weak d-d absorbance bands at long wavelengths. Thus, upon analyte binding, a decrease in the donor fluorescence was observed. The decrease was measured by monitoring the timedependent phase angle change at a fixed frequency upon binding of the metal ion. Results for two different concentrations of Co 2§ are shown in Figure 20. The fiber configuration included an entrapped enzyme in a polyacrylamide layer immobilized to the tip of an optical fiber. 37

Biran and Walt

(a)

...(a.~!...

ct)P

AMP

. . . . . . . . . . . . . .

AND/OR 2 A D P ~ - - -

ADP

=

adenylLate kinase

~ - -~enylate. kinase .....~ -

+

ATP

-'~f~;-phosphoc'zcadnc

creatine

cz'~atinc

kinase

AI'P + l~iferin + O 7 ----i-u~cifera~.-7---~

oxylur

+ AMP § PPi + CO z

lisht (~.,,~ ,~ 560 nm)

(b) . . . .

r ~me

.

-

,..

~,,

I

t20 0

: ........ ,,%- . . . . . .*.,.--,-,., ..%.:;.*-,.**.,,,-'.+.+ . . . .

R

...1 -2SD

o o

.St

ff

1'l'~al= perk-",.l : 2.~ lurers :,= 3"2 m e a n valttt= f ~ ) : 1(15, O ]R.LU l.mmlllnJ d~vlmion (SD) ~f e l i n e

O

. O

. 4

.

m

~

. ~

. i~

: 5 . 6 3 ;RI.,IJ

dnv~.|cm

. t6

Number

: 5.3r

qS

.

. 20

24

28

32

36

o4' a s s a y s

Figure 2 I. Reagentless bio-optrode for AMP, ADP and ATP detection. (a) Schematic of the enzymatic reactions employed for the measurements. (b) Repetitive measurements of ATP using controlled released of luciferin from acrylic microspheres. The light intensity was measured after each injection of 25 pmol ATP. The self-contained bio-optrode reproducibility over three hours (32 repetitive injections) is shown (Michel et al., 1998a). Reprinted with permission from Elsevier Science.

A related approach for fabricating reagentless enzyme-based biosensors is based on transducing conformational changes occuring upon substrate binding into FRET signals. Proteins such as calmodulin, maltose binding protein and phosphate binding protein undergo conformational changes upon substrate binding and were used to prepare such biosensors (Hellinga and Marvin, 1998). Using genetic engineering, two FRET fluorescent groups (acceptor and donor) are bound to two different cysteine residues that are spatially located such that conformational changes, due to analyte binding, result in a FRET signal change. 38

Optrode-based Fiber Optic Biosensors A different example for a reagentless enzyme-based bio-optrode was recently described (Michel et al., 1998a). The sensor was designed to detect the threeadenylate nucleotides (ATP, ADP, AMP) using a three-enzyme reaction sequence. Three enzymes were used: adenylate kinase, creatine kinase and luciferase. The enzymes were compartmentalized in such a way that the product of the first reaction would be accessible to serve as the substrate for the subsequent reactions shown in Figure 21 (a). The final indicator reaction for all three analytes involves the luciferase reaction. In previous bio-optrode designs the cosubstrate for this reaction, luciferin, was externally added to the flow cell. In the new bio-optrode design, luciferin was incorporated into acrylic microspheres. When the microspheres were immobilized together with the enzymes on the fiber surface they slowly released the luciferin allowing continuous detection for 3 hours (Figure 21 (b)). This approach is generic for the controlled release of cosubstrates or cofactors, which can be used in different enzyme-based bio-optrodes.

3.4. Environmental applications Many bio-optrodes have been proposed for use in environmental applications (Marty et al., 1998; Rogers and Gerlach, 1999; Rogers and Poziomek, 1996; Schobel et al., 2000). For remote monitoring, only the fiber tip containing the biorecognition element has to be located at the measurement site (e.g., lakes, rivers, sewage streams) while the optical signal detection instrumentation can be located in a protected location away from the site. Optical fibers are small in diameter and flexible, and therefore can be located in places inaccessible to other sensing devices. In addition, the optical fiber's durable structure protects it from harsh environmental conditions. At present, environmental bio-optrodes are still in the research and development stage with most of the research focused on detection scheme development and optimization. Several antibody-based bio-optrodes have been described for detecting pesticides such as terbutryn (Bier et al., 1992), parathion (Eldefrawi et al., 1995) and imazethapyr (Wong et al., 1993). One example is a bio-optrode for the detection of 2,4-dichlorophenoxyacetic acid (2,4-D) in water (Wittmann et al., 1996). In this system, an optical fiber with an immobilized analyte (2,4-D) was placed into a flow cell. The assay procedure involved several steps: (1) The fiber was incubated with fluorescently labeled monoclonal antibody for 2,4-D and the initial (maximum) fluorescence signal was measured. The fiber was then washed with buffer; (2) The sample was incubated with fluorescently labeled monoclonal antibody for 2,4-D; (3) The fiber was incubated with the sample-labeled antibody solution mixture; (4) The fiber was washed and the signal was measured. When a high concentration of analyte (>1000 ~tg/L) was present in the sample, a low signal was obtained because most of the antibodies were occupied with the sample analyte and could not bind to the 2,4-D immobilized on the fiber. This bio-optrode was used to measure concentrations ranging between 0.2-100 lxg~, a 39

Biran and Walt concentration range suitable for environmental applications where the permitted level of 2,4-D in drinking water cannot exceed 0.1 ~tg/L. The sensing layer could be regenerated by washing with proteinase K. This procedure enabled the biooptrode to be used for more than eight weeks and more than 500 successive measurements. Such bio-optrodes have the potential to be useful for on-line analysis of drinking water and to serve as warning devices for hazardous pesticide contamination. Enzyme bio-optrodes for environmental applications have also been developed. The most common approach employs enzyme inhibition as the sensing mechanism. The inhibition of the enzyme acetylcholinesterase (ACHE) with its substrate, acetylcholine, by an organophosphate pesticide was used in several sensors (Doong and Tsai, 2001; Eldefrawi et al., 1995; Xavier et al., 2000), and was also described in Section 1.4.1. A different enzyme-based bio-optrode that uses a chemiluminescence reaction for detection of phenolic compounds was recently described (Ramos et al., 2001). This bio-optrode is based on the enhancement of the luminol-H202-horseradish peroxidase chemiluminescence reaction by phenolic compounds. Using this biooptrode, p-iodophenol, p-coumaric acid, and 2-naphthol were detected in concentrations as low as 0.83 ~VI, 15nM, and 48 nM respectively. The biooptrode was fabricated by entrapping the enzyme in a sol-gel layer; the gel was prepared directly on the fiber tip. The assay was performed by inserting the fiber with the immobilized enzyme into a test tube containing the analyte, luminol and H202. The chemiluminescence intensity maximum at 5 min was the output signal. Whole cells were also used for environmental bio-optrode construction. Recombinant E. coli cells over-expressing the enzyme organophosphorus hydrolase were immobilized to an optical fiber and used to detect organophosphate nerve agents, as was described above (Section 1.4.1) (Mulchandani et al., 1998). The bio-optrode detection limits were 3 ~M for paraoxon and parathion and 5 ~M for coumaphos. The sensor was stable for over a 1-month period and used for over 75 repeated measurements. Using a different approach, in which the cell's genetic response was used as the sensing mechanism, a whole cell bio-optrode was used for detection of naphthalene and salicylate (Heitzer et al., 1994; Ripp et al., 2001). The sensing was performed by Pseudomonas fluorescens HK44 cells carrying a plasmid containing a genetic fusion between the nahG gene, which is responsive (i.e., induced) to the presence of naphthalene and salicylate, and the IuxCDABE reporter gene, coding for the enzyme luciferase. B ioluminescence was produced when the cells were exposed to either naphthalene or salicylate. The cells were immobilized onto the surface of a liquid light guide or an optical fiber by using strontium alginate. The bio-optrode tip was placed in a measurement flow-cell 40

Optrode-based Fiber Optic Biosensors that simultaneously received a waste stream solution and a maintenance medium. A rapid increase in bioluminescence was obtained when one of the analytes was present in the waste stream. Real environmental samples of pollutant mixtures containing naphthalene were tested using this system. High bioluminescence was obtained when aqueous solutions saturated with JP-4 jet fuel or aqueous leachates from contaminated soil were tested. Using a similar approach, Ikariyama and coworkers (1997) used recombinant E. coli cells carrying genes responsive to the presence of aromatic compounds fused to the tuc (coding for firefly luciferase) reporter gene. The cells were immobilized to an optical fiber and the remote sensing device was used to measure aromatics in the part per billion concentration range. Recently, a bio-optrode based on the same idea, was reported in which recombinant E. coli cells produced bioluminescence in response to the presence of genotoxic agents (Polyak et al., 2001). This bio-optrode was able to detect as low as 25 ~tg/L of mitomycin C in less then two hours. The main importance of genetic response based bio-optrodes for environmental analysis is the information they provide about the bioavailability of the analytes. This parameter is very important and helps to decide how to treat the polluted site and which remediation strategies to employ.

3.5. Clinical applications The development of bio-optrodes for clinical applications is another promising field and is focused on two types of applications; (a) in-vivo detection inside a patient, (b) ex-vivo detection when clinical samples are analyzed at the patient's bedside. The in-vivo bio-optrodes would enable continuous monitoring of important analyte concentrations and would dramatically improve clinical procedures such as heart bypass surgery and critical care procedures in patients with compromised respiratory conditions. The optical fiber's small diameter, flexibility, nontoxic nature, durability and lack of direct electrical connections make them highly suitable for in-vivo applications. Moreover, optical fibers have already proven to be valuable for in-vivo clinical applications such as endoscopic procedures and laser power transmission for surgical procedures. For example, endoscopes are used in endoscopic surgery for gall bladder removal and for chest and knee surgery. In principle, bio-optrodes can be coupled to such devices and used to provide analytical information during endoscopic surgeries. At present, such bio-optrodes (or other in-vivo chemical optical sensors) have not been implemented because of blood compatibility problems in which a thrombus (clot) forms around the sensor tip and affects the measurement accuracy. The second clinical application for bio-optrodes is ex-vivo diagnostics, mainly in critical care situations. Most diagnostic tests are presently performed in a 41

Biran and Walt centralized laboratory. Samples must be collected with the attendant transport, storage and chain-of-custody issues. The remote location of the laboratory delays the medical diagnosis. In order to provide rapid diagnostic tests, analytical devices, such as bio-optrodes, can be used to bring the laboratory closer to the patient. These point-of care devices should be sensitive, selective, self contained (no need to add reagents), and simple to operate. They should also be small in size in order to be conveniently located near the patient. In addition, it is preferable that the sampling unit in contact with the sample (e.g., blood, urine) be disposable. B io-optrode devices of this type are still not commercially available, but there are similar chemical-based fiber optic sensor devices used routinely in clinics. In these devices, fluorescent dyes are use as indicators for monitoring blood gases (PO2, PCOz) and pH. In one device, the immobilized dyes are incorporated into a disposable apparatus that is inserted into an extracorporeal blood circuit on one side and connected to a fiber bundle on the other (Owen, 1996). These sensors are mainly used to monitor blood gases during open-heart surgery. Another device is used for a paracorporeal measurement at the patient's bedside (Martin et al., 1994). The sensors are placed into an external tube connected to an arterial blood line. Blood samples are periodically and automatically pumped into the tube, analyzed by the sensors, and then returned to the blood line. In this way, the blood can be monitored semi-continuously without requiring blood samples to be taken from the patient. It should be possible to incorporate bio-optrodes into such devices and use them to monitor other clinically important analytes. For both in-vivo and ex-vivo applications, the first step in bio-optrode development is to establish sensitive sensing mechanisms that can be used to recognize specific analytes in a complex sample such as blood, urine or other human fluids. Many examples of bio-optrodes directed for clinical applications have been proposed (Meadows, 1996; Vidal et al., 1996; Vo-Dinh and Cullum, 2000). Several glucose bio-optrodes, based on the enzyme-catalyzed reaction of glucose with glucose oxidase have been prepared or proposed (Li and Walt, 1995; Marquette et al., 2000; Moreno-Bondi et al., 1990). A submicrometer glucose bio-optrode has been prepared (Rosenzweig and Kopelman, 1996a, 1996b). In this bio-optrode, the consumption of molecular oxygen is measured by the fluorescence quenching of ruthenium complexes and serves as a reporter for glucose concentration. The enzyme and the indicator are immobilized in an acrylic polymer support on the fiber tip. The bio-optrode response is very fast (2 seconds) and concentrations as low as 1 x 1015 mol were detected. Recently, a bio-optrode for myoglobin was described (Hanbury et al., 1997). This self-contained antibody-based bio-optrode is clinically important since it can serve as a method for monitoring the extent of myocardial infarction. Myoglobin was detected by immobilizing a fluorescently labeled monoclonal antibody in polyacrylamide gel on the tip of an optical fiber (Figure 22 (a)). Cascade Blue was used both as the fluorescent labeling agent and as a FRET donor molecule. 42

Optrode-based Fiber Optic Biosensors When myoglobin was captured by the antibody, fluorescence energy transfer occurred between Cascade Blue and the myoglobin heme group (acceptor). Fluorescence quenching of Cascade Blue was measured and correlated to the myoglobin concentration (Figure 22 (b)). The polyacrylamide gel was optimized to serve as a size selective filter allowing only low molecular mass molecules to penetrate and interact with the antibodies. Using this approach, myoglobin (16,500 Da) could be discriminated from hemoglobin (bigger than 70 kDa). The size selection was necessary since antibodies for myoglobin can also bind hemoglobin due to similar antigenic determinants. As shown in Figure 22 (b), a significant response was obtained when the bio-optrode was incubated with myoglobin but no response was obtained with hemoglobin (93 nmol/L). The detection limit of this bio-optrode was 5 nmol/L (83 ~tg/L), which is near the clinical decision limit for myocardial infarction diagnostics. The limitation of using the gel layer is the increased response time due to the low diffusion rate through the gel layer. In addition, when the gel layer was used, the bio-optrode response was irreversible even when the bio-optrode was incubated in a solution containing a high concentration of myoglobin antibodies (Figure 22 (b)). Irreversible sensor responses could limit the use of this bio-optrode for continuous monitoring applications. Another example of a bio-optrode for heart related disease diagnostics is the Ddimer antigen bio-0ptrode (Grant and Glass, 1999). D-dimer antigen is formed when vascular occlusions are treated with a thrombolytic agent in order to lyse the clot. This treatment involves inserting a microcatheter at the occlusion site and injecting thrombolytic agents. Although thrombolytic therapy can help in preventing strokes, it suffers from several limitations that bio-optrodes can help to overcome. One limitation of this approach is that it is difficult to know if the occlusion occurred from an atherosclerotic plaque or a thrombus. Using the biooptrode, initiation of D-dimer antigen formation following the injection of a small amount of thrombolytic agent would indicate the presence of a thrombus clot. If D-dimer antigens are not detected after the thrombolytic agent injection, it would be an indication that the occlusion is caused by an atherosclerotic plaque and alternative treatment would be required. In addition, in the case of a thrombus clot, the bio-optrode can be used for on-line monitoring of the thrombolytic agent dosage needed to dissolve the clot by monitoring the D-dimer antigen formation during the procedure. The bio-optrode principle of operation is similar to the myoglobin bio-optrode described above. In this case, the antibody is labeled with fluorescein and fluorescence quenching is observed when Ddimer antigen binds. The fluorescently labeled antibody is immobilized in a solgel on an optical fiber tip. The bio-optrode was used to detect D-dimer antigen in human plasma and in whole blood; the detected concentration range was 0.56 - 6 ktg/ml, which is within the clinically relevant range. The limitations of this biooptrode are its poor reversibility and the short lifetime of the immobilized antibodies (four weeks).

43

Biran and Walt

Figure 22. Reagentless bio-optrode for myoglobin detection. (a) Bio-optrode setup. (b) Myoglobin bio-optrode responses. The bio-optrode responses in PBS buffer (0-80 min.) and after incubations with hemoglobin (Hb), myoglobin (Mb) and unlabeled myoglobin antibody (Ab) are shown (Hanbury et al., 1997). Reprinted with permission from Clin. Chem.

In addition to the bio-optrodes described above that are directed to in-vivo applications, several bio-optrodes for point-of-care applications have also been described. These bio-optrodes offer a miniature design and a fast response time for analytes such as bilirubin (Li and Rosenzweig, 1997), cholesterol (Marazuela et al., 1997), and D-amino acids (Zhang et al., 1995).

3.6. Industrial applications including bioprocess monitoring Cell culture-based bioprocesses are very complex to control since they are sensitive to minor changes in the chemical composition of the fermentation medium. Therefore, tools for on-line monitoring of different analyte concentrations during the bioprocess are highly desirable. Bio-optrodes offer 44

Optrode-based Fiber Optic Biosensors

-I..............x'"I- x

~o.o

,

,

,,,

I'

~

-

. _ .

:113.0'

o 20.0

tj gl 0

10.0"

tb3

0.0 0

50

100

I(~

800

~

"~" ~-'" ~0 400

800

process U m e / b

"~ il-............I &

.~ t.5. o

1,0, &&

o o

0.5,

0.0 I 0

,r I

IIo

tI~

160

pl'oceII

2o0

. . . .

II~

9

. . . .

~I~

9

~

2

. . . .

w

UlXle/h

Figure 23. Five-channel FIA-based enzyme bio-optrode for continuous monitoring of lactate, glucose, glutamine, ammonia and glutamate (from top to bottom) during animal cell cultivation. Chemiluminescence measurements were employed to determine the analyte concentrations using five optical fibers, each with a different immobilized enzyme (Spohn et al., 1995). Reprinted with permission from Elsevier Science.

several advantages in such applications. The ability to use optical fibers directly inside fermentors (in situ) eliminates the need to periodically remove samples for analysis in a remote analytical laboratory. Once inside the fermentor, biooptrodes can be used for sensitive and selective on-line monitoring of different analytes. Fermentation substrates and products such as proteins, antibodies and antibiotics can be monitored. Other parameters related to the biological status of the cells, such as cell viability and activity, can also be measured. The ability to perform this measurement from a remote location (e.g., central control room) without using wires offers another important advantage. The main obstacle, which prevents wide use of in-situ bio-optrodes (or any type of biosensor) is the 45

Biran and Walt need to sterilize the bio-optrode, which may damage the sensing biomolecules. For this reason, most bio-optrodes for bioprocess monitoring use a flow system in which a sample of the medium is taken from the fermentor and is delivered to the sensor (Dremel et al., 1992; Marose et al., 1999; Mulchandani and Bassi, 1995; Scheper et al., 1996). In one example, a FIA-based enzyme bio-optrode system was used for simultaneous detection of five different analytes (Spohn et al., 1995). Glucose, lactate, glutamate, glutamine, and ammonia were detected in samples that were removed during animal cell cultivation. The system is based on chemiluminescence detection and consists of five optical fibers, each with a different immobilized enzyme. Each fiber is inserted into a different flow cell and, when the sample is injected, each fiber's response is measured. The results are combined and the concentrations of the different analytes are determined. Figure 23 shows results from a 350-hour monitoring of an animal cell culture medium. In a similar way, antibody-based or nucleic acid-based bio-optrodes can be used to monitor different bioprocesses.

4. Advantages and Limitations of Bio-optrode Technology B io-optrodes offer several advantages over other biosensing technologies based on the unique characteristics of optical fibers. The optical fiber's small dimensions, flexibility, and their ability to transmit optical signals for long distances allows them to be used for remote sensing in places where other biosensors cannot be used. In addition, their ability to function without any direct electrical connection to the sample makes them safer than electrochemical biosensors. B io-optrodes are intrinsically simpler than electrode-based biosensors since no reference electrodes are needed. Moreover, the development of new biorecognition molecules, such as those containing FRET-based dyes, enables the fabrication of self-contained bio-optrodes where no addition of reagent is needed. B io-optrodes based on imaging fibers offer additional advantages since they allow multiplexing with rnulti-analyte sensing capabilities. It is expected that as new optical technologies are developed for telecommunication applications, they will be adopted for bio-optrodes. These technologies include miniaturization of light sources, detectors and optical fiber components (Kostov and Rao, 2000). Furthermore, new developments in the area of nanotechnology should eventually enable development of new bio-optrodes at the nanometer scales. Nevertheless, bio-optrode technologies suffer from several drawbacks. Some of these drawbacks are common to all biosensor devices, with the most difficult one being the poor stability of the biological recognition molecules. Such molecules tend to be sensitive to pH or temperature changes and therefore have short 46

Optrode-based Fiber Optic Biosensors lifetimes. Another important limitation is the high cost of some of the purified biological sensing materials. Regeneration of sensing biomolecules is usually problematic and, in most cases, fresh biorecognition molecules are required for each assay. In addition, there are several problems related to the immobilization process including loss in activity, leaching of reagents, and the decreased response time due to slow diffusion of analytes through the irrkrnobilized layer. Several other bio-optrode limitations are related to the nature of optical fibers. Since light signals are the measured parameter, bio-optrodes are sensitive to ambient light interference and precautions must be taken either to exclude light or to employ optical designs with lock-in detection capabilities. In most biooptrodes, there is a need to use indicator dyes in order to transduce the biorecognition events. The dyes have to be immobilized together with the biomolecules and therefore complicate the bio-optrode fabrication. In addition, the dyes may leach from the immobilization matrix or may lose their characteristics because of photobleaching.

5. Potential for Improving Performance or Expanding Current Capabilities As with any sensing or monitoring device, the ideal bio-optrode should be specific, sensitive, simple to fabricate and use, well adapted to the measurement environment (e.g., detect specific analytes in a complex sample), reliable and self-contained. When used as a sensor, it should be operated in a continuous and reversible manner. When used as a probe, it should include a simple and disposable unit that contains the sensing molecules. In addition, for many applications, bio-optrodes should be small, able to detect multiple analytes simultaneously, and enable measurements in remote sites. At present, no biooptrode device has achieved all these ideal performance capabilities. Nevertheless, based on new bio-optrode technologies currently under development, it is expected that the next generation of bio-optrodes will come closer to achieving these goals. The development of new bio-optrode technologies and devices is highly dependent on advances in several different fields. Advances in biology, chemistry, materials science, optics, electrical engineering, mechanical engineering, and computer engineering, are expected to inspire new bio-optrode technology development. In this section, a few new key technologies and their future impacts on the bio-optrode field are discussed.

5.1. New optical fibers and instrumentation Optical fibers have attracted attention mainly due to their use in telecommunications. New technologies have been developed for fabricating optical fibers with very efficient light transmission capabilities. Fibers can 47

Biran and Walt transmit extremely high amounts of data when used in both single or bundle format. These characteristics will advance the development of real-time multianalyte bio-optrodes for various analytical applications. Improved, smaller, and less expensive light sources and detectors are driving consumer electronics. Integration of these components into bio-optrodes can lead to miniaturization and commercialization of bio-optrode devices (Kostov and Rao, 2000). Among the different possible light sources, light emitting diodes (LEDs) are very attractive to use in bio-optrodes. LEDs are very small, cover the entire visible spectrum, produce optical power in the range of 0.1-5 mW, and have a very long life (100,000 hours) and low cost (-$2). Once LEDs at a particular wavelength have been demonstrated and commercialized, laser diodes are usually available within a few years. Laser diodes have higher power output and are nearly monochromatic whereas LEDs have a relatively broad spectral emission output. Another interesting new light source is the scintillation light source that can be used as a high stability light source for the UV and blue region (Potyrailo et al., 1998). These sources are based on long-lived radioisotopes in scintillation crystals, which convert the radioactive emission (typically beta particles) into light emission. These sources are extremely stable, can be used without external power sources, and have an expected life of 20 years. In recent years, new generations of miniaturized and improved light detectors, such as photodiodes (PD) and photomultiplier tubes (PMT), have been developed (Kostov and Rao, 2000). The most sensitive detectors are avalanche photodiodes. CCD chips are also rapidly developing; bigger chips with higher signal/noise ratios, wider dynamic ranges, and lower dark currents have been developed. In addition, CCD detectors have been miniaturized and integrated into small devices. Image intensifiers have been integrated into CCD cameras to increase light detection sensitivity. Although CCD chip prices have been dramatically reduced, scientific grade CCD cameras are still very expensive (-$20,000). A competing technology to CCD is the complementary metal oxide semiconductor (CMOS) technology. Recent developments in this technology have demonstrated light detection capabilities similar to CCD detectors. The advantages of this technology are lower cost, simpler fabrication process and the ability to use it for very fast image acquisition (32,000 pictures per second) because frame transfers are not required as all the processing is done on chip. Both CCD and CMOS technologies will most likely be integrated into future bio-optrode devices.

5.2. New biological recognition elements The "heart" of any bio-optrode is the biological recognition element that initiates the detection process by its interaction with the analyte. Development of new biological recognition elements will increase the number and types of analytes that can be detected by bio-optrodes. New advances in molecular biology techniques allow the design or selection of new recognition molecules. For 48

Optrode-based Fiber Optic Biosensors example, using phage display technology it is possible to screen and identify a single chain antibody (scFv) with specificity for almost any analyte (Hoogenboom et al., 1998). The recombinant scFv molecule is a smaller version of an antibody molecule containing the antigen binding site. Genetically modified bacteriophages, each presenting a unique scFv molecule on its surface, are used in the screening process. Phage presenting scFvs with higher affinities to the analyte are selected. The system is designed in such a way that the sequence coding for the scFv presented in the selected phage can be readily identified. This sequence is placed in an expression vector and E. coli cells are used to produce large quantities of the selected scFv molecule. This process is very powerful since it allows antibodies to be identified and isolated in a very short time (Goldman et al., 2000). Moreover, once the antibody is found, it takes only a few days to produce it in large quantities. In addition, a scFv molecule can be specifically designed to be used in biosensor devices by adding immobilization capabilities to the molecule at the genetic level. For example, scFvs were fused to the CBD (cellulose binding domain) resulting in scFv molecules that can be easily immobilized to a cellulose membrane (Berdichevsky et al., 1999). Using molecular biology techniques, several other biorecognition molecules have been designed. Genetic fusion between antibody molecules and the green fluorescent protein resulted in self-fluorescent antibodies, also called fluorobodies, and eliminates the need to label the antibody with a fluorescent dye. Other genetically modified molecules designed for biosensing are described in Chapter 10. Biological recognition elements isolated using combinatorial approaches, such as aptamers, are described in detail in Chapter 12. Biomimetic polymer materials for bio-optrode applications are described in Chapters 11 and 13.

5.3. Imaging and biosensing The coupling of chemical and biosensing capabilities to optical fiber-based imaging devices (e.g., endoscopes) is expected to attract much attention in the future. Optical imaging fibers are used in endoscopes since they can carry images from one end of the fiber to the other due to the coherent nature of the fibers. These imaging capabilities can be utilized to simultaneously image and measure local analyte concentrations with micron-scale resolution (Michael and Walt, 1999). The imaging fiber's distal face is coated with an analyte-sensitive layer (typically a biorecognition molecule and fluorescent indicator), which produces a microsensor array capable of spatially resolving analyte concentrations. The concept is shown in Figure 24. For example, an acetylcholine imaging fiber biooptrode was fabricated by coating the imaging fibers with a polymer layer containing the enzyme acetylcholinesterase and fluorescein (Bronk et al., 1995). The fiber was used both to visualize the tobacco hornworm and to measure acetylcholine release from the neural ganglion (Walt, 1998). The acetylcholine

49

B iran and Walt

Figure 24. Combined imaging and biosensing concept. The technique provides the ability to both view tissue slices or individual cells and measure the release or consumption of different analytes using fluorescence techniques. release was measured followed electrical stimulation of the worm's sensory nerve. First, the fiber was placed over the neural ganglion. A white light image was taken to visualize the neural ganglion morphology. The imaging fiber biooptrode was then switched to the fluorescence mode and the neural ganglion was stimulated. Higher intensity signals were obtained from the neural ganglion regions that released acetylcholine. The ability to observe the location of neurotransmitter release, with microscale spatial resolution, can provide a powerful tool for neuroscience researchers. In a similar approach, one day the technology may be used for in-vivo applications. For example, a suspected cancer tumor could be examined based both on its morphology and its response to specific antibodies immobilized on the imaging fiber tip.

5.4. Data analysis New bio-optrode technologies are expected to provide a large amount of analytical information from each measurement. For example, multi-analyte biooptrodes will measure the concentration of many different analytes simultaneously. When such measurements are performed in continuous fashion (e.g., multiple measurements every second), it will generate a high data volume. 50

Optrode-based Fiber Optic Biosensors

(a) ~,7 F=5

0

i'~"'V'l"

0

(b)

"

20

'

I "

40

'

"

I

'

61:)

'"

|-~

' ~ S m M

10

'

80

" ]1

'

"

I

'

'

"

I

" '

'

I

'

r"

|

'~'

100 120 140 160 160 200

tlme [sl ..........

................. ,oo

r.,. ~

0

20

40

_

~0

~SmM

eo

100 120 time Is]

140 180

180 2oo

Figure 25. Penicillin detection using a FIA-based bio-optrode. (a) Fluorescence signals obtained in response to different penicillin concentrations. (b) Fluorescence signals obtained with different buffer ion concentrations and a fixed penicillin concentration (5 g/L). Neural network analysis was applied to analyze the relationships between these two responses (Muller et al., 1997). Reprinted with permission from Elsevier Science.

In order to acquire, analyze, and save such high data volumes, sophisticated software should be developed or adapted from other high volume applications. The most significant computerized task is the data analysis because it may affect the specificity, sensitivity, and reproducibility of the bio-optrode. Analysis of biosensor measurements may be complicated due to the high variability in the activity of biorecognition molecules and because the measurements are usually performed in a complex sample matrix. It was previously shown that even in a simple FIA biosensor system for measurement of a single analyte in bioprocess samples, there is a need to use advanced computational analysis in order to improve the sensitivity, selectivity, and reproducibility of the measurement (Hitzmann et al., 1998). For example, in typical bioprocess samples, the pH or ion concentrations change during cultivation, which may affect the biosensor's enzyme activity or antibody binding properties. In addition, inhibitors and proteases can be produced during the bioprocess, which will affect protein-based bio-optrodes, or nucleases can be produced, which will affect nucleic acid-based 51

Biran and Walt biosensors. In order to overcome these problems, multivariate evaluation techniques such as neural networks have been used. Scheper and coworkers (Muller et al., 1997) applied the neural network approach to improve the analysis of signals obtained from a FIA bio-optrode for penicillin. The need to use the neural network approach was due to the sensitivity of the measurement to changes in buffer ion concentration (Figure 25). The neural network was used to simultaneously evaluate the ion and penicillin concentrations from a single measurement based on characteristic signal shape variations. The shape characteristics were thought to be useful because in a preliminary experiment, multiple measurements of the same sample showed reproducible signal shapes. The results from this neural network showed errors of less then 11% for six different penicillin concentrations measured at five different ion concentrations. This simple example demonstrates the power of such computing techniques in the analysis of bio-optrode signals. It is clear that such techniques will be essential for analyzing signals from multi-analyte bio-optrodes. Advanced computational methods have recently been used for the analysis of chemical sensor arrays (Jurs et al., 2000), and it is expected that they will be adapted to the analysis of bio-optrode measurements. The generation of high amounts of information from future multi-analyte bio-optrodes is expected to shift the emphasis from signal measurement to data analysis.

6. References Aboul-Enein, H. Y., R. I. Stefan, J. F. van Staden, X. R. Zhang, A. M. GarciaCampana and W. R. G. Baeyens, 2000, Critical Rev. Anal. Chem. 30, 271. Anderson, G. P., K. D. King, K. L. Gaffney and L. H. Johnson, 2000, Biosens. Bioelectron. 14, 771. Arnold, M. A., 1985, Anal. Chem. 57, 565. Barker, S. L. R., R. Kopelman, T. E. Meyer and M. A. Cusanovich, 1998, Anal. Chem. 70, 971. Barker, S. L. R., H. A. Clark, S. F. Swallen, R. Kopelman, A. W. Tsang and J. A. Swanson, 1999, Anal. Chem. 71, 1767. Berdichevsky, Y., E. Ben-Zeev, R. Lamed and I. Benhar, 1999, J. Immunol. Meth. 228, 151. Bier, F. F., W. Stocklein, M. Bocher, U. Bilitewski and R. D. Schmid, 1992, Sens. Actuators B-Chem. 7, 509. Blum, L. J., S. M. Gautier and P. R. Coulet, 1993, J. Biotechnol. 31,357. B lum, L. J., S. M. Gautier and P. R. Coulet, 1994, In Food B iosensor Analysis, Eds., G. Wagner and G. G. Guilbault, Marcel Dekker, New-York, pp. 101. Bronk, K. S., K. L. Michael, P. Pantano and D. R. Walt, 1995, Anal. Chem. 67, 2750. 52

Optrode-based Fiber Optic Biosensors Bronstein, I., C. S. Martin, J. J. Fortin, C. E. M. Olesen and J. C. Voyta, 1996, Clin. Chem. 42, 1542. Campbell, A. K. and G. Sala-Newby, 1993, In Fluorescence And Luminescence Probes For Biological Activity, W. T. Mason, Eds., Academic Press, London, pp. 58-79. Chalfie, M., Y. Tu, G. Euskirchen, W. W. Ward and D. C. Prasher, 1994, Science 263,802. Cullum, B. M., and T. Vo-Dinh, 2000, Trends Biotechnol. 18, 388. Cunningham, A. J., 1998, Introduction To Bioanalytical Sensors, John Wiley & Sons, Inc, New York, pp. 260-277. Daunert, S., G. Barrett, J. S. Feliciano, R. S. Shetty, S. Shrestha and W. SmithSpencer, 2000, Chem. Rev. 100, 2705. Diaz, A. N., M. C. R. Peinado and M. C. T. Minguez, 1998, Anal. Chim. Acta 363, 221. Doong, R. A. and H. C. Tsai, 2001, Anal. Chim. Acta 434, 239. Dremel, B. A. A., S. Y. Li and R. D. Schmid, 1992, Biosens. Bioelectron. 7, 133. Dunn, B., J. Cox, E. Lan and J. I. Zink, 2001, Abstr. Pap. Am. Chem. Soc. 221, 473-COLL. Dunn, B., J. M. Miller, B. C. Dave, J. S. Valentine and J. I. Zink, 1998, Acta Materialia 46, 737. Eldefrawi, M. E., A. T. Eldefrawi, N. A. Anis, K. R. Rogers, R. B. Wong and J. J. Valdes, 1995, In Immunoanalysis of Agrochemicals, American Chemical Society, Washington, pp. 197-209. Ferguson, J. A., T. C. Boles, C. P. Adams and D. R. Walt, 1996, Nature Biotechnol. 14, 1681. Ferguson, J. A., B. G. Healey, K. S. Bronk, S. M. Barnard and D. R. Walt, 1997, Anal. Chim. Acta 340, 123. Ferguson, J. A., F.J. Steemers and D.R. Walt, 2000, Anal. Chem. 72, 5618. Fraser, D., 1995, Med. Device Technol. 6, 28, 34. Freeman, M. K. and L. G. Bachas, 1992, Biosens. Bioelectron. 7, 49. Goldman, E. R., M. P. Pazirandeh, J. M. Mauro, K. D. King, J. C. Frey and G. P. Anderson, 2000, J. Mol. Recognition 13, 382. Grant, S. A. and R. S. Glass, 1999, IEEE Trans. Biomed. Eng. 46, 1207. Gubitz, G., M. G. Schmid, H. Silviaeh and H. Y. Aboul-Enein, 2001, Crit. Rev. Anal. Chem. 31,167. Hanbury, C. M., W. G. Miller and R. B. Harris, 1997, Clin. Chem. 43, 2128. Healey, B. G., S. E. Foran and D. R. Walt, 1995, Science 269, 1078. Healey, B. G. and D. R. Walt, 1995, Anal. Chem. 67, 4471. Healey, B. G., L. Li and D. R. Walt, 1997a, Biosens. Bioelectron. 12, 521. Healey, B. G., R. S. Matson and D. R. Walt, 1997b, Anal. Biochem. 251,270. Heitzer, A., K. Malachowsky, J. E. Thonnard, P. R. Bienkowski, D. C. White and G. S. Sayler, 1994, Appl. Environ. Micro. 60, 1487. Hellinga, H. W. and J. S. Marvin, 1998, Trends Biotechnol. 16, 183. Hitzmann, B., A. Ritzka, R. Ulber, K. Schongarth and O. Broxtermann, 1998, J. Biotechnol. 65, 15. 53

Biran and Walt Hobel, W. and J. Polster, 1992, Fresenius J. Anal. Chem. 343, 101. Hoogenboom, H. R., A. P. de Bruine, S. E. Hufton, R. M. Hoet, J. W. Arends and R. C. Roovers, 1998, Immunotechnol. 4, 1. Ikariyama, Y., S. Nishiguchi, T. Koyama, E. Kobatake, M. Aizawa, M. Tsuda and T. Nakazawa, 1997, Anal. Chem. 69, 2600. Iqbal, S. S., M. W. Mayo, J. G. Bruno, B. V. Bronk, C. A. Batt and J. P. Chambers, 2000, B iosens. B ioelectron. 15,549. Jordan, J. D., R. A. Dunbar and F. V. Bright, 1996, Anal. Chim. Acta 332, 83. Jurs, P. C., G. A Bakken and H. E. McClelland, 2000, Chem. Rev. 100, 2649. Kohler, S., S. Belkin and R. D. Schmid, 2000, Fresenius J. Anal. Chem. 366, 769. Kostov, Y. and G. Rao, 2000, Rev. Sci. Instruments 71, 4361. Kuswandi, B., R. Andres and R. Narayanaswamy, 2001, Analyst 126, 1469. LaRossa, R. A., 1998, In B ioluminescence Methods And Protocols, Vol. 102, Humana Press, Totowa, New Jersey, pp. 85-299. Lee, M. and D. R. Walt, 2000, Anal. B iochem. 282, 142. Li, L. and D. R. Walt, 1995, Anal. Chem. 67, 3746. Li, X. P. and Z. Rosenzweig, 1997, Anal. Chim. Acta 353,263. Liu, X. J., W. Farmerie, S. Schuster and W. H. Tan, 2000, Anal. Biochem. 283, 56. Lubbers, D. W. and N. Opitz, 1975, Pflugers Archiv-Eur. J. Physiol. 359, R145. Marazuela, M. D., B. Cuesta, M. C. MorenoBondi and A. Quejido, 1997, Biosens. Bioelectron. 12, 233. Marose, S., C. Lindemann, R. Ulber and T. Scheper, 1999, Trends B iotechnol, 17, 30. Marquette, C. A., A. Degiuli and L. J. Blum, 2000, Appl. Biochem. Biotechnol. 89, 107. Martin, R. C., S. F. Malin, D. J. Bartnik, A. Schilling and S. C. Furlong, 1994, Proc. SPIE, 2131. 426. Marty, J. L., B. Leca and T. Noguer, 1998, Analysis 26, M144. Meadows, D., 1996, Adv. Drug Deliv. Rev. 21, 179. Mehrvar, M., C. Bis, J. M. Scharer, M. Moo-Young and J. H. Luong, 2000, Anal. Sci. 16, 677. Michael, K. L., L. C. Taylor, S. L. Schultz, and D. R. Walt, 1998, Anal. Chem. 70, 1242. Michael, K. L. and D. R. Walt, 1999, Anal. Biochem. 273, 168. Michel, P. E., S. M. Gautier-Sauvigne and L. J. Blum, 1998a, Talanta 47, 169. Michel, P. E., S. M. Gautier-Sauvigne and L. J. Blum, 1998b, Anal. Chim. Acta 360, 89. Moreno-Bondi, M. C., O. S. Wolfbeis, M. J. Leiner and B. P. Schaffar, 1990, Anal. Chem. 62, 2377. Mulchandani, A. and A. S. Bassi, 1995, Crit. Rev. Biotechnol. 15, 105. Mulchandani, A., I. Kaneva and W. Chen, 1998, Anal. Chem. 70, 5042. Muller, C., B. Hitzmann, F. Schubert and T. Scheper, 1997, Sens. Actuators BChem. 40, 71. 54

Optrode-based Fiber Optic Biosensors Naylor, L. H., 1999, Biochem. Pharmacol. 58, 749. Owen, V. M., 1996, Biosens. Bioelectron. 11, R5. Pantano, P. and D. R. Walt, 1995, Anal. Chem. 67, A481. Pilevar, S., C. C. Davis and F. Portugal, 1998, Anal. Chem. 70, 2031. Polyak, B., E. Bassis, A. Novodvorets, S. BelkSn and R. S. Marks, 2001, Sens. Actuators B-Chem. 74, 18. Potyrailo, R. A., S. E. Hobbs and G. M. Hieftje, 1998, Anal. Chim. Acta 367, 153. Preininger, C., I. Klimant and O. S. Wolfbeis, 1994, Anal. Chem. 66, 1841. Rabbany, S. Y., B. L. Donner and F. S. Ligler, 1994, Critical Rev. Biomed. Eng. 22, 307. Ramos, M. C., M. C. Torijas and A. N. Diaz, 2001, Sens. Actuators B-Chem. 73, 71. Ripp S., D. E. Nivens, Y. Ahn, C. Werner, J. Jarrell, J.P. Easter, C. D. Cox, R. S. Burlage and G. S. Sayler, 2001, Environ. Sci. Technol. 34, 846. Rogers, K. R. and C. L. Gerlach, 1999, Environ. Sci. Technol. 33, 500A. Rogers, K. R. and M. Mascini, 1998, Field Anal. Chem. Technol. 2, 317. Rogers, K. R. and E. J. Poziomek, 1996, Chemosphere 33, 1151. Rosenzweig, Z. and R. Kopelman, 1996a, Anal. Chem. 68, 1408. Rosenzweig, Z. and R. Kopelman, 1996b, Sens. Actuators B-Chem. 36, 475. Sackmann, E., 1996, Science 271, 43. Scheper, T. H., J. M. Hilmer, F. Lammers, Muller, C.M. Reinecke, 1996, J. Chromatog. A 725, 3. Schobel, U., C. Barzen and G. Gauglitz, 2000, Fresenius J. Anal. Chem. 366, 646. Spohn, U., F. Preuschoff, G. Blankenstein, M. R. Kula, D. Janasek and A. Hacker, 1995, Anal. Chim. Acta 303, 109. Steemers, F. J., J. A. Ferguson and D. R. Walt, 2000, Nature B iotechnol. 18, 91. Steemers, F. J. and D. R. Walt, 1999, Mikrochim. Acta 131, 99. Szurdoki, F., K. L. Michael and D. R. Walt, 2001, Anal. Biochem. 291,219. Tan, W. H., Z. Y. Shi, S. Smith, D. Bimbaum and R. Kopelman, 1992, Science 258, 778. Thompson, R. B., Z. F. Ge, M. Patchan, C. C. Huang and C. A. Fierke, 1996, Biosens. Bioelectron. 11,557. Thompson, R. B. and E. R. Jones, 1993, Anal. Chem. 65,730. Tyagi, S. and F. R. Kramer, 1996, Nature Biotechnol. 14, 303. Tyagi, S., S. A. E. Marras and F. R. Kramer, 2000, Nature B iotechnol. 18, 1191. Vidal, M. M., I. Delgadillo, M. H. Gil and J. Alonso-Chamarro, 1996, Biosens. Bioelectron. 11,347. Vo-Dinh, T., J. P. Alarie, B. M. Cullum and G. D. Griffin, 2000, Nature Biotechnol. 18, 764. Vo-Dinh, T. and B. Cullum, 2000, Fresenius J. Anal. Chem. 366, 540. Vo-Dinh, T., B. M. Cullum and D. L. Stokes, 2001, Sens. Actuators B-Chem. 74, 2. Walt, D. R., 1998, Accounts Chem. Research 31,267. 55

Biran and Walt Walt, D. R., 2000, Science 287, 451. Weetall, H. H., 1993, Appl. Biochem. Biotechnol. 41,157. Wilchek, M. and E. A. Bayer, 1990, Avidin-Biotin Technology, Academic Press, San Diego. Wittmann, C., F. F. Bier, S. A. Eremin and R. D. Schmid, 1996, J. Ag. Food Chem. 44, 343. Wolfbeis, O. S., 1991, In Fiber Optic Chemical Sensors And Biosensors, Vol. 2, CRC Press, Boca Raton, FL. pp 193-257. Wolfbeis, O. S., 1997, In Optical Fiber Sensors - Applications, Analysis, And Future Trends, Vol. 4, B. Culshaw and J. Dakin, Eds., Artech House, N orw ood, pp. 53-107. Wolfbeis, O. S., 2000, Anal. Chem. 72, 81R. Wong, R. B., N. Anis and M. E. Eldefrawi, 1993, Anal. Chim. Acta 279, 141. Wortberg, M., M. Orban, R. Renneberg and K. Cammann, 1997, In Handbook of Biosensors and Electronic Noses: Medicine, Food, and the Environment, Ed., E. Kress-Rogers, CRC Press, Boca Raton, pp. 369. Xavier, M. P., B. Vallejo, M. D. Marazuela, M. C. Moreno-Bondi, F. Baldini and A. Falai, 2000, Biosens. Bioelectron. 14, 895. Zhang, W., H. Chang and G. A. Rechnitz, 1997, Anal. Chim. Acta 350, 59. Zhang, Z. J., Z. L. Gong and W. B. Ma, 1995, Microchem. J. 52, 131. Zhao, C. Q., N. A. Anis, K. R. Rogers, R. H. Kline, J. Wright, A. T. Eldefrawi and M. E. Eldefrawi, 1995, J. Ag. Food Chem. 43, 2308.

56

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER2

EVANESCENT WAVE FIBER OPTIC BIOSENSORS

CHRIS ROWE TArrT, PH.D. AND FRANCES S. LIGLER, D.PH~., D.Sc.

Center for Bio/Molecular Science & Engineering, Naval Research Laboratory, Washington, DC 20375 USA

When light is launched down a waveguide placed in contact with a lower refractive index material, if conditions are present to allow total internal reflection o f this light, an electromagnetic component of the light extends out from the surface of the waveguide into the lower index medium. This electromagnetic field, the evanescent wave, has a limited penetration depth and can be used to specifically excite fluorophoresbound to, or in close proximity to, the waveguide surface. Evanescent wave fiber optic biosensors have been developed utilizing the limited penetration depth of the evanescent wave to detect a variety of analytes. These sensors are able measure optical events at the fiber's surface with relatively little interference from the bulk solution. The ability of these instruments to detect analytes rapidly and specifically, even in the presence of complex sample matrices, has been demonstrated both under laboratory conditions and in the field.

1. Technical Concept Fiber optic biosensors utilize two distinct assay configurations for signal generation and measurement: the optrode configuration, discussed in exquisite detail in the preceding chapter, and the evanescent wave configuration, the focus of this chapter. Both configurations rely on the same principle of total internal reflection (TIR) for light propagation and guiding. However, while optrodes use the light shining out the end of the fiber to generate a signal either at the distal face of the fiber or in the medium near the fiber' s end, evanescent wave sensors rely on the electromagnetic component of the reflected light at the surface of the fiber core to excite only the signal events localized at that surface. The penetration depth of the light into the surrounding medium is much more 57

Taitt and Ligler restricted than for optrodes while the surface area interrogated is much larger in comparison to optrodes of equal diameter. The result is that evanescent wave biosensors require immobilization of the biological recognition molecules onto the longitudinal surface of the optical fiber core, primarily measure binding events, and are relatively immune from interferents in the bulk solution. 1.1. T h e e v a n e s c e n t w a v e

As described in more detail in Chapter 1, TIR is observed at the interface between two dielectric media with different indices of refraction. TIR is described by Snell's law: nl/n2 sin01 = sin02,

(1)

where n~ and n2 are the refractive indices of the fiber optic core and the surrounding medium, respectively; 0~ is the incident light angle through the fiber optic core; and 02 is the angle of either the light refracting into the surrounding medium or the internal reflection back into the core. Total internal reflection requires that n~ > n2 and occurs when the angle of incidence is greater than the critical angle, 0c, defined as 0c = sin -1 (n2/nl)

(2)

This parameter must be considered when designing any biosensor based on optical fibers. However, while Snell's law describes the macroscopic optical properties of waveguides, it does not account for the electromagnetic component of the reflected light, known as the evanescent wave. The evanescent wave is an electric field that extends from the fiber surface into the lower index medium and decays exponentially with distance from the surface, generally over a distance of 100 to several hundred nanometers (Figure 1). For multimode waveguides, the penetration depth, dp, the distance at which the strength of the evanescent wave is 1/e of its value at the surface, is approximated by: dp = ~ / (4g[n12sin20 - n22]~),

(3)

where n~ and n2 are refractive indices of the optical fiber and surrounding medium, respectively, and 0 is the angle of incidence (Harrick, 1967). The importance of the evanescent wave is its ability to couple light out of the fiber into the surrounding medium, thereby providing excitation for fluorophores bound to or in close proximity to the fiber core surface. This confined range of excitation is one of the major factors responsible for the relative immunity of evanescent wave-based systems to the effects of matrix components or interferents beyond the reaction surface. 58

Evanescent Wave Fiber Optic Biosensors 0

Cladding

0
,

0

Core ~anescent 0 0

Figure 1. Evanescent wave biosensor using partially clad fiber.

A crucial factor in both evanescent excitation and coupling of fluorescence emission back into the fiber core is the waveguide parameter of the optical fiber, or V-number. The V-number is a dimensionless factor that determines how many modes a fiber can support, and is defined as: V = (2rcr/X)(nlLn22)w,

(4)

where r is the radius of the optical fiber, nl is the refractive index of the fiber, and n2 is the refractive index of the surrounding medium or cladding. For uniform mode distribution, the power present in the evanescent wave (i.e., coupling out of the fiber) decreases with increasing V-number. On the other hand, the coupling of fluorescence emission back into the fiber from the surface increases with Vnumber (Thompson, 1991). Love and Button (1988) originally suggested that dipoles close to the surface could emit approximately 2% of their radiated power (fluorescence) into modes coupled back up the fiber; however, Polerck3~ et al. (2000) have more recently analyzed films of dipoles and concluded that a much higher proportion of the radiation can couple into guided modes. Fiber optic probes have been constructed from both unclad and partially clad fiber cores. Additional light may be lost from long, partially clad fibers, due to a V-number mismatch. Light propagating through a waveguide is limited to specific modes that are functions of the wavelength and the physical characteristics of the waveguide. If the refractive index of the cladding is different from that of the medium surrounding the declad sensing region, according to Equation 4, a V-number mismatch occurs, and the fluorescent emission is not guided into the core but enters the clad region and is not 59

Taitt and Ligler transmitted to the detection optics. In order for the light guided in the unclad region of the core to be propagated in the clad region of the core, the following constraint must be satisfied:

raq _ r~l [(n~o 2 - ncl2)/(nco2-naq2)] ~6, where nco is the refractive index of the fiber core, n~l is the refractive index of the cladded fiber, naq is the refractive index of the aqueous medium, raq is the radius of the declad sensing region, and rc~ is the radius of the cladded region. The mechanism for holding an unclad fiber in place can have the same effect as cladding. Einink et al. (1990) described the phenomenon discussed above in terms of skew rays (higher order modes). By using a holder that minimized contact with the fiber core, the sensitivity of the fiber optic biosensor increased. Lackie (1992) patented a mounting system for an unclad fiber where the holder at one end of the fiber had the same refractive index as the fiber itself in order to avoid a V-number mismatch. Glass (1989) patented a similar mounting system for unclad fiber probes with the same goal in mind. Erb and Downward (1998) coated the ends of an unclad fiber with a Teflon having the same refractive index as the aqueous samples (n = 1.3) to prevent the V-number mismatch and subsequent loss of light in higher order modes where the fiber probe touched the holder. Using partially clad fibers, it was demonstrated that decreasing the fiber radius to eliminate the V-number mismatch increased the light coupling efficiency from the sensing region into the clad region and subsequent transmission of the fluorescence through the rest of the fiber (Thompson and Kondracki, 1990; Thompson and Villaruel, 1991; Golden et al., 1992, 1994a,b; Anderson et al., 1994a,b; Anderson and Golden, 1995; Gao et al., 1995). Whereas lower order modes are concentrated near the center of the fiber core, higher order modes are distributed more towards the outer edge of the core and penetrate further into the surrounding medium. Thus, higher order modes will effect greater evanescent excitation of surface-bound fluorophores. However, emitted fluorescence coupled into back into the fiber is also carded in higher order modes. In the case of the untapered fibers, the fluorescence in these higher order modes is preferentially lost into the cladding. With tapering, the higher order modes are converted into the lower order modes; these lower order modes concentrate near the center of the fiber core and can easily become propagated modes which reach the detector. By altering the geometry of the sensing region (Figure 2), immunoassay sensitivities could also be increased up to 80-fold (Anderson et al., 1993).

60

Evanescent Wave Fiber Optic Biosensors

-i

_ I]1 . . . . . . .

III

IU .

.

.

.

.

.

.

.

.

.

. . . . . . .

III II

. . . . .

B .

.

..

..

.. .

.

.

.

.

.

.

-

J

.

.

.

.

.

.

.

.

II

-

_

.

U

D

--

II111

.

II.

.

.

.

Figure 2. Schematic of untapered and tapered fibers. A) Fiber probe with cladding stripped away from core. B) Step-tapered core to create V-number match between sensing and cladded regions. C) Continuously tapered core to maximize excitation light at surface. D) Combination tapered core to reach a V-number match quickly and subsequently maintain maximum excitation light at the fiber surface.

1.2. Optical fibers Optical fibers possess a number of related physical characteristics that can be used to distinguish them, such as radius, refractive index, V-number (described above), and the numerical aperture (NA); the last is a measure of the difference in refractive indices between the fiber core and the cladding and is related to the V-number: NA = (nlZ-n2Z)'~

(6)

Monomode fibers, such as those used in the telecommunications industry, typically have a core size of 5-10 jxm and propagate only a single mode at any given wavelength. As they have lower numerical apertures, they have a higher percentage of total power present outside the core; greater than 50% of the power can be in the media surrounding the monomode fibers with low V-numbers (Gloge, 1971) - hence, more power to excite fluorophores via the evanescent field. Single mode fibers have been used for evanescent sensing by a number of groups (Lew et al., 1986; Villaruel et al., 1987; Carlyon et al., 1992; Hale et al., 1996), but are not generally given much attention for sensing purposes; their small core radii render them extremely fragile and difficult to handle. Furthermore, the intensity of power present in their evanescent fields is such that many fluorophores rapidly photobleach. For these reasons, most evanescent wave sensors utilize multimode silica or plastic fibers for light transmission. Multimode fibers possess the advantages of good light transmission over short 61

Taitt and Ligler Multimode fibers possess the advantages of good light transmission over short and medium distances with a wide variety of optical components. While evanescent excitation of surface-bound fluorophores is not as efficient in the monomode fibers, ease of use, lower rates of photobleaching, and increased coupling efficiency are strong advantages of multi-mode fibers. And most importantly, the larger, multimode fibers produce more surface area for the immobilization of the biological recognition molecules. Fibers composed of fused silica offer excellent optical transmission from the near UV range to the near-IR range of wavelengths and have low intrinsic fluorescence. In our studies, however, we found that the amount of intrinsic fluorescence in fused silica fibers was a significant problem for the discrimination of fluorescence below 600 nm (Ligler et al., 1995). Photobleaching the fibers immediately prior to conducting a fluorescence assay eliminated the background fluorescence; operating at wavelengths above 600 nm avoided the problem altogether (Shriver-Lake et al., 1995b). Diameters of silica multimode fiber cores generally range from 50 gm up to 1.5 mm. Furthermore, fibers made from this material are resistant to most biological buffer systems. However, methods required to resolve V-number mismatches (i.e., tapering) require time-consuming and often hazardous procedures that are difficult to reproduce from batch to batch, e.g., etching with hydrofluoric acid. Plastic fibers, on the other hand, can be injection molded to fit the user's and instrument's requirements (Slovacek et al., 1992; King et al., 1999; Saaski and Jung, 2000). Furthermore, dopants can be added to change the refractive index over a wide range. The most commonly used plastic fibers are composed of polymethylmethacrylate (PMMA) or polystyrene. While plastic fibers have a very limited range of temperatures at which they can be used in comparison to silica fibers, this range of temperatures (-30~ to 80~ is currently sufficient for study of most biological systems. The chief problem with plastic fibers is the limited spectral range for which they can be used. This limitation is due to the high attenuation in the red and near-IR spectrum f r o m - C H absorption bands (Figure 3). This problem can be partially circumvented by doping the plastics with deuterium; deuterium replaces t h e - C H absorption band with a - C D band, thereby increasing the range of useful wavelengths. These deuterium-doped plastic fibers, however, tend to lose their optical transmission over time (Boisd6 and Harmer, 1996). In practice, an additional problem with the use of plastic fibers is the variation in the starting materials used for molding; plasticizers and other additives which fluoresce at visible wavelengths may be included in company-proprietary formulations. As a result, each batch of starting material must be screened for the resulting optical properties at the specific wavelengths intended for use.

62

Evanescent Wave Fiber Optic Biosensors

10,000

polycarbonate

"7 %

% %,

E ',,z' 1313 '1:3

!

%

I i

%j%

1,000

I #

polystyrene

v

t~ m

o

..Q I.i... ,, , . . . , , .

Deuterated PMMA

/

PMMA

100

4

50500

600 700 Wavelength (nm)

800

Figure 3. Spectra of various plastics used in optical fibers. Adapted from Boisd6 and Harmer (1996).

1.3. Instrumentation In addition to the optical fiber probe coated with a biological recognition molecule, the fiber optic biosensor consists essentially of a fluorimeter that is capable of discriminating excitation light from emitted fluorescence and a fluid transfer system to move samples and reagents over the fiber probe. The components and configurations for these two systems depend on the geometry of the fiber and the degree of automation of the sensor, 1.3.1. The fluorimeter. The fluorimeter portion of the biosensor can be divided into the excitation and emission collection components. Generally, lenses are used to focus light from the source onto or into the fiber and line filters are included in the excitation path if necessary. The first fiber optic biosensors used halogen lamps (Block and Hirschfeld, 1987; Glass, 1989; Block et al., 1990) or 63

Taitt and Ligler xenon lamps (Kooyman et al., 1987; Lipson et al., 1992) as light sources; these were soon replaced with lasers with more uniform excitation wavelengths (Thompson and Ligler, 1988; Golden et al., 1992; Hale et al., 1996). The development of hydrophilic, near infrared dyes (Shriver-Lake et al., 1995b, Wadkins et al., 1995) and diode lasers (Golden et al., 1994, 1997; Choa et al., 1996) meant that the advantages of laser excitation could be implemented in small, low cost devices. An additional advantage of the diode lasers is that the mechanical choppers originally included in order to discriminate fluorescence from excitation and ambient light could be eliminated and the diode lasers pulsed to achieve the same effect. The ability to discriminate a weak fluorescent signal above the background excitation light is the most critical feature of the fluorimeter. Choppers or pulsed lasers provide one mechanism to accomplish that discrimination; excitation line filters and high quality emission filters are also crucial. However, perhaps the most interesting factor in accomplishing the discrimination is the variety of geometries employed by different groups to separate the excitation and emission paths physically. Figure 4 depicts several of these strategies. Strategies A, B, and C rely on coupling the light into the evanescent wave from the center of the core and recovering the fluorescent light from the higher order modes. Appropriate lenses and filters are included with each of these configurations. While the first fluorescence immunoassays performed on the surface of a waveguide involved collection of the fluorescent signals perpendicular to the waveguide (Kronick and Little, 1975), nearly all fiber optic biosensors collect the emitted light out the end of the fiber. This means that the signal is integrated over the active surface of the fiber core and focused on a single detector, usually either a photomultiplier tube or photodiode. In a notable exception to end-face collection, Fang and Tan (1999) determined that they could detect individual fluorescent molecules using evanescent excitation and signal collection normal to the waveguide using a microscope equipped with an intensified charge coupled device. In this case, there was no intent to integrate the signal from multiple biological recognition events. 1.3.2. The fluidics. In addition to the fluorometer, the fluidics component for delivering the sample and reagents to the fiber probe is an integral part of the biosensor. Primary considerations include the mechanism for holding the fiber probe in place within the fluid stream, the capillary-type chamber surrounding the probe with its inlet and outlet ports, and the associated pumps and valves for automated assays in biosensors intended for applications other than research.

64

Evanescent Wave Fiber Optic Biosensors

Figure 4. Strategies for separation of excitation and emission light paths. A) A traditional dichroic mirror is used to separate excitation and emission based on the difference in wavelengths. The excitation light may pass through the mirror while the emission light is reflected onto a detector (left) or, in a scheme that has generally proven more effective, the stronger excitation light is reflected onto the end of the fiber while the emitted light passes straight through. B) The excitation light passes through a hole in an off-axis parabolic mirror while the emission light is reflected by the mirror onto a detector (Thompson and Levine, 1992). C) A fiber bundle is used between a high numerical aperture sensing probe and the optics. The 635 nm excitation light is channeled down the central silica fiber while the emitted higher wavelength fluorescence is coupled back up the surrounding plastic fibers (Golden et al., 1997). D) The fiber is illuminated using a light source normal to the fiber, and the fluorescence is detected at the distal end (Kooyman et al, 1987; Ligler et al, 2002).

65

Taitt and Ligler

Flow out

Flow...._in

To detector Figure 5. Optical fibers within capillary fluidics chambers. Adapted from Anderson et al. (1994a).

Several capillary chambers with individual fluid inlets and outlets have been described as part of a fiber optic biosensor system and deliver fluid to the surface of the fiber (Block and Hirschfeld, 1987; Glass, 1989; Slovacek and Love, 1992; Anderson et al., 1993; Meserol, 1996; Neel and Lyst, 1997). Where the capillaries are intended for use with unclad fiber probes, significant attention has been given to the mechanism for holding the probe in the center of the capillary in order to prevent the V-number mismatch problems described above. Using a single fiber in a capillary tube, Oroszlan et al. (1993) reported the first fully automated assay system based on an optical fiber and computerized the operation of over 200 sequential 15-30 minute assays over a single fiber probe. The simple capillary system described by Anderson and colleagues (1993) (Figure 5) was the first example connecting multiple probes in order to perform simultaneous analyses on a single sample for multiple analytes (Shriver-Lake et al., 1998; Bakaltcheva et al., 1998). The addition of samples and reagents over four fibers in series was automated in a system based on the Analyte 2000 fiber optic biosensor (Research International, Woodinville, WA) and included commercially available pumps and valves. This automated prototype was tested for its ability to collect and identify aerosolized bacteria while airborne in a small, unmanned plane (Ligler et al., 1998; Anderson et al., 1999) (Figure 6). This was the first demonstration that samples could be both collected and tested without manual operations. The demonstration of a fully automated biosensor was the impetus for the commercial development of an automated portable system. Hitherto fore, the only commercial systems were the ORD device made by Block and his associates and the Analyte 2000. Both devices were basically small fluorimeters designed 66

Evanescent Wave Fiber Optic Biosensors

Figure 6. Analyte 2000 with automated fluidics for remote sensing (Ligler et al., 1998).

to measure fluorescence emitted from one or four probes, respectively; neither system included an automated fluidics component and both were designed primarily for use by the research community. Research International subsequently developed an automated portable system designed to flow fluids through a disposable coupon containing four optical probes (King et al., 1999; Saaski and Jung, 2000; Saaski, 2000). The fiuidics system went through several iterations, primarily due to necessary improvements in the mechanism for fluid transport. The first three prototype instruments, the Mantis, SOF-FOWG and RAPTOR, used a pneumatic fluidic system based on small air pumps, pinch valves, and pressurized fluid reservoirs. In theory, if no samples ever went through the pumps, the pumps could not clog. However, the membranes in the valves proved unreliable and variations in the rate of fluid flow affected the assay performance. In the latest version of the system, the RAPTOR-Plus, the pneumatic fluidics have been replaced with a peristaltic pump-driven system based on small, custom-made pumps. This system is proving to be very reliable in terms of long term operation (> 1 year to date). The disposable coupon that contains the fiber probes has been altered only to remove the on-board valves related to the pneumatic fluidics and to accommodate minor improvements in the fiber probes themselves. The coupon holds four polystyrene probes, which are inserted after being coated with the recognition biomolecule and dried. The probes are designed to integrate a combination tapered sensing region with a lens for signal collection and a tab for gluing the probe into the coupon (Figure 7) (Saaski and Jung, 2000). The coupon automatically aligns the probe so that it is in the middle of the fluid channel (Saaski, 2000) and the light emitted from the end of the fiber is focused onto a collection lens in the permanent portion of the biosensor. 67

Taitt and Ligler

Figure 7. Schematic (A) and photograph (B) of injection-molded optical fiber probes and fluidics coupon for use with the RAPTOR-Plus.

1.4. Detection parameters A number of different optical parameters, such as absorbance, fluorescence emission spectral shifts, fluorescence resonant energy transfer (FRET), and bioluminescence, have been used for detection/quantification with fiber optic optrodes and are reviewed in Chapter 1. However, the number of parameters measured using evanescent wave fiber optic sensors is much more limited. While absorbance has been directly measured in evanescent wave chemical sensors (Buerck et al., 2001) and in an evanescent immunosensor based on long period fiber Bragg gratings (DeLisa et al., 2000), fluorescence intensity has been the property most often used in evanescent wave biosensors. The most notable exception, surface plasmon resonance (Jorgensen and Yee, 1993), is reviewed in Chapter 7. In affinity-based biosensors, measurement of fluorescence requires that the analyte of interest or the immobilized recognition element have either intrinsic fluorescence that can be distinguished from the background or that either species be labeled with a fluorescent tag. When neither species is intrinsically fluorescent or can be labeled, use of a fluorescent "tracer" reagent is required; this tracer is most often a recognition molecule directed against an epitope not involved in the binding of analyte to the recognition element immobilized on the surface. Fluorescence quenching may be used in evanescence-based sensors, but only if quenching does not occur in competition with photobleaching. 68

Evanescent Wave Fiber Optic Biosensors The availability of many new fluorophores allows greater distinction between specific signals and background signals from the sample matrix by increasing the signal-to-noise ratios. Near-infrared dyes, in particular, have the advantage of emitting at longer emission wavelengths, where (background) fluorescence from naturally occurring compounds is minimal. The cyanine dyes also have high quantum yields, are easy to conjugate under mild conditions, and are significantly more resistant to photobleaching than rhodamine and fluorescein dyes. These dyes also have the advantage of exciting in the range of wavelengths emitted by photodiode lasers, which are lightweight, long-lived, and inexpensive. Molecular Probes (Eugene, OR) has recently made available a series of AlexaFluor dyes that are marketed for their limited self-quenching. These dyes can be used to create more highly labeled proteins (up to eight dye molecules per protein molecule) than is feasible using the cyanine dyes without decreased fluorescence due to self-quenching (G.P. Anderson, unpublished data). Another method for substantially reducing the background of evanescent wave assays is use of time-resolved fluorescence. While this technology is not yet in widespread use in the biosensor field due to requirements for high intensity, single-photon, pulsed lasers and sophisticated data analysis, two groups have demonstrated time-resolved fluorescence detection using evanescent wave fiber optic technology. Bock et al. (1995) used evanescent excitation to detect binding of analytes to an optical fiber and was able to distinguish between antibodies labeled with two fluorophores that differed only in their fluorescence lifetimes. Browne and coworkers (1996), using sol-gel-coated fibers, were able to distinguish between encapsulated fluorophores localized in different regions on the optical fiber. While this latter work did not utilize a biologically-based system, the authors demonstrated the use of time-resolved detection as a means of resolving the fluorescence kinetics of various fluorophores, as well as the spatial arrangement of the fluorophores on the fiber. 1.5. Immobilization of biomolecules on optical fiber probes A necessary step in the development of evanescent fiber optic biosensors was the development of methods for immobilizing the biomolecules on the surface of the fiber in such a way as to maximize the density of the receptors and to maintain their function. Two major approaches have been explored extensively: direct attachment to the fiber core and attachment of the receptor to an intermediate biomolecular thin film (i.e., through avidin or a lipid monolayer). Direct attachment was the first approach developed. The early work of Hirshfeld (Hirschfeld, 1984; Hirschfeld and Block, 1985) suggests modifying the fiber surface with silanes and using any of a variety of chemistries or crosslinkers to attach proteins. Bhatia and colleagues (Bhatia et al., 1989; Eigler et al., 1991) describe an attachment chemistry based on thiol silanes and the heterobifunctional crosslinker 7-maleimidobutyryloxy-succinimide, which has 69

Taitt and Ligler subsequently become very widely used in many types of biosensors. This chemistry has the advantage that unmodified thiols oxidize to form sulfonates (Bhatia et al., 1989) which are very hydrophilic and may prevent subsequent nonspecific adsorption of the biomolecule to the glass (and also denaturation). Feldman et al. (1995) subsequently published an independent analysis of the activity of antibodies immobilized using the Bhatia method. Willamson et al. (1989) published a comparison of direct attachment methods based on toluenesulfonyl chloride, chloroformate, and an amino silane plus glutaraldehyde. While the last method is very simple and was used for several years, it has fallen out of favor for direct binding of proteins; the number of attachment sites for the proteins is probably too high and subsequent mobility of the proteins may be too restricted for optimal functionality. Herron and colleagues (1996) modified the amino silane-glutaraldehyde method in a way that solves this problem. They attached the hydrogel polymethacryloyl hydrazide or polyethylene glycol (PEG) molecule to the glutaraldehyde, providing a hydrophilic spacer between the hydrophobic amino silane and the protein. Fab' fragments could be coupled to the hydrogel using a linker between the free thiol of the antibody fragment and the hydrazido group of the hydrogel. Any protein could be linked to the PEG after oxidation. While nonspecific adsorption is effective in attachment of proteins to many surfaces, investigators working with glass or silica fibers found that the density of attached antibodies was highly variable, binding activity was often lost, and loss of protein by leaching was problematic. However, Rogers et al. (1989, 1991) found that nonspecific adsorption of the larger acetylcholine receptor was a very satisfactory method to immobilize the protein and maintain its binding function. With the use of plastic fibers instead of glass or fused silica, King et al. (2000) has found that nonspecific adsorption of antibodies to polystyrene is reliable in terms of maintaining antibody density and function, even after long periods of storage or extended use (<40 assays). In addition to direct immobilization, recognition molecules have been immobilized through an intermediate layer that prevents the non-specific adsorption of the recognition molecule to the fiber core and permits greater control of the number of attachment sites on the biomolecule. Avidin has been linked to the surface by the methods of Bhatia et al. (1989) and Herron et al. (1996), as well as though covalently immobilized biotin (Abel et al., 1996; Liu and Tan, 1999). This subsequently serves as a covalently bound substrate for the attachment of biotinylated recognition molecules.

70

Evanescent Wave Fiber Optic Biosensors

Figure 8. Detection of B. globigii using antibodies immobilized onto optical fibers via avidin bridge (gray bars) vs. direct covalent attachment (black bars). Adapted from Ligler et al. (1998).

The improvement in sensitivity using fused silica fibers coated with avidin has been demonstrated by Anderson and colleagues (Narang et al, 1997a; Ligler et al., 1998) (Figure 8). A similar approach was also demonstrated using Protein A or Protein G as an intermediate layer (Anderson et al., 1997). Protein A and Protein G are microbiaIly produced IgG-binding proteins with different affinities for the Fc region of IgGs of different species. The antibodies immobilized via Protein A or Protein G produced brighter signals in direct assays than the same antibodies covalently attached to the fiber probes; however, there was no improvement in the limits of detection. The subsequent work of this group with fused silica fibers used the avidin chemistry as it is more generally applicable to a variety of assays than the Protein A or Protein G approach. In an interesting twist on the use of an intermediate protein layer, Mauro et al. (1996) genetically engineered a DNA-binding protein to include a Protein G terminus and bound this receptor to IgG-coated fiber probes. In addition to protein layers, organic films fabricated using the LangmuirBlodgett method have been deposited on optical fibers. Krull et al. (1988) pioneered this approach using lipids both for direct sensing and for membrane receptor supports. They demonstrated the response of an embedded fluorophore to membrane perturbation by phloretin and valinomycin. Zhao and Reichert (1992) doped the lipid films with a biotinylated lipid to serve as an anchor for subsequent attachment of avidin; this method allowed very fine control over the density of the biotin and the association of the avidin. Avidin layers attached to 71

Taitt and Ligler immobilized biotin can also serve as the substrate for the attachment of another layer containing biotinylated receptors (Piunno et al., 1995; Duvenek et al., 1996; Abel et al., 1996). While not a biosensor per se, Matsuo et al. (2000) continued to use Langmuir-Blodgett films (arachidic acid and tetrabromophenol blue) on optical fibers as a evanescent sensing layer for albumin, using the dye to pH change. The complexity of using the Langrnuir-Blodgett method and problems with reproducibility, however, make it highly unlikely that this approach will yield a widely used or commercial product.

1.6. Molecular recognition and detection Optical biosensors require a biological recognition element with the following qualities: affinity and specificity for the analyte of interest; stability when incorporated into the biosensor; and ability to cause a detectable change in optical properties upon recognition of analyte. The recognition of analyte may take the form of a catalytic reaction, such as in catalytic sensors, or may be one or a series of binding events, such as used in immunosensors and other affinitybased sensors. Since the use of catalytic biomolecules with evanescent wave fiber optic sensors is so limited (acetylcholinesterase, Rogers et al., 1991), we will discuss only affinity sensors in detail here. The ability to regenerate the receptor to its active state after a binding event is a commendable, but not a necessary, asset. Regardless of the type of biological recognition molecule, specific recognition is formatted so that a fluorescent complex is either formed or deleted at the surface of the optical biosensor in the presence of analyte. The change in fluorescence of membrane-embedded dyes is a semi-selective recognition mechanism that modifies the environment of the dye upon partition of the analyte into the membrane (Krull et al., 1988; Matsuo et al. 2000), and thus we will categorize it ouside the mainstream of specific recognition processes. The specific assays fall into three major formats" direct binding, competitive, and sandwich assays (Figure 9). Direct binding assays employ a biological recognition molecule immobilized on the surface of the optical fiber (Figure 9A). The immobilized biomolecule then binds an analyte that is intrinsically fluorescent, carries a label, or changes the fluorescence of the immobilized biomolecule upon binding. Examples in which the analyte is itself fluorescent have been used primarily for demonstrating the sensitivity of the system (Graham et al., 1992; Abel et al., 1996; Mauro et al, 1996) or for determining the kinetics of binding (Orvedahl et al, 1991; Zhao and Reichert, 1992; Vijayendran et al, 1999). Alternatively, the analyte and all

72

Evanescent Wave Fiber Optic Biosensors

Figure 9. Formats for affinity assays. A. Direct binding assay. A fluorescent analyte or nonspecifically stained analyte binds to a recognition molecule immobilized on the fiber probe. B. Competitive imrnunoassay. Labeled and unlabeled analyte compete for binding to the immobilized recognition biomolecule. C. Sandwich immunoassay. A fluorescent complex forms when the immobilized recognition molecule binds to the analyte and the fluorophore-labeled tracer binds to the analyte at the surface of the optical fiber.

similar moieties in the Solution can be made fluorescent and the specificity of the receptor given the responsibility of binding only the target moiety. For example, all bacteria can be stained and only the target species bound by the antibody on the probe (Ligler et al., 1993, 1996). Similarly, the biorecognition molecule can be a DNA that binds its target and is subsequently stained using ethidium bromide (Piunno et al., 1995) or changes its fluorescence (Liu and Tan, 1999). In a variation of the direct binding assay, fluorescence is associated with the receptor unless the analyte is bound. Pandey and Weetall (1995a) use double stranded DNA stained with an intercalating dye as a receptor. When a compound 73

Taitt and Ligler is present that is capable of intercalating into DNA, the dye is displaced and the fluorescence reduced. Competitive assays have been used in immunoassays for small molecules containing a single epitope (Oroszlan et al., 1993; Devine et al., 1995; ShriverLake et al., 1995a; Thompson and Maragos, 1996). However, competitive formats have proven quite useful with other types of receptors and binding molecules. For example, the antibiotic polymyxin B was been used to detect endotoxin using a competitive assay (James et al., 1996; Ligler and James, 1999). In assays using the acetylcholine receptor, a neurotoxin was detected if labeled abungarotoxin was prevented from binding to the acetylcholine receptor (Rogers et al., 1989). Erb et al. (2001b) have developed a competitive assay for compounds with estrogenic activity using the estrogen receptor. Kleinjung et al. (1998) demonstrated the detection of adenosine using RNA as the binding moiety. Competition assays can be configured in two ways: In the first format, unlabeled and labeled analytes compete for binding to recognition molecules immobilized on the surface of the waveguide; the decrease in fluorescent signal will be proportional to the amount of unlabeled species in the mix (Figure 9B). The second format for a competition assay requires that the recognition molecule be present in solution and an analog of the analyte be immobilized on the optical fiber; this assay format has recently been used to test urine samples for cocaine and its metabolites (Nath et al., 1999). This setup results in competition between the immobilized analyte and analyte free in solution for binding to the fluorescent recognition molecule. In this case also, the amount of analyte in the sample is directly proportional to the decrease in fluorescent signal compared to controls. By far the largest number of assays reported, however, have been sandwich assays. Sandwich assays (Figure 9C) are commonly used for large molecules and require a second (fluorescent) recognition species. This fluorescent "tracer" is used to detect an analyte that has been "captured" by a recognition element on the fiber surface. This type of assay most often utilizes antibodies as both "capture" and "tracer" elements. The ability to detect proteins (e.g. Sutherland et al., 1984; Ogert et al, 1992; Duvenek et al., 1995; Moreno-Bondi et al., 2000; Spiker et al., 1998) is well documented. However, sandwich immunoassays have also proven effective for the detection of bacteria (e.g., Ligler et al., 1993; Pease et al., 1995; and De Marco et al., 1999), viruses (Lee and Thompson, 1996; King et al., 2000), and protozoa (Anderson and Rowe-Taitt, 2001). For many applications, it is useful to have a sensing probe that can be regenerated and reused. For most of the competitive immunoassays, where the binding of the analyte to the probe is univalent, the bound analyte or labeled analog can be removed without destroying the functionality of the receptor. The simplest method for removing the bound molecule is with mild organic solvents (Oroszlan et al., 1993 (200 assays/fiber probe); Shriver-Lake et al., 1995b). Larger molecules which bind to immobilized receptors by multiple sites are more 74

Evanescent Wave Fiber Optic Biosensors difficult to remove without irreversibly denaturing the immobilized biomolecule. Oroszlan et al. (1992), Wijesuria et al. (1994), and Anderson et al. (1997) use rather harsh chaotropic agents and pH extremes to accomplish such regeneration, but the number of times that the biomolecule will tolerate such abuse, if at all, is limited. Moreno-Bondi et al. (2000) describe the use of sonication for separating the CA15-3 antigen from immobilized antibody; however, only 65% of the antigen was removed, and the number of binding molecules for which sonication is a useful approach may be limited. King et al. (2000) uses the fiber optic probes to perform as many as 40 assays without regeneration; however, this group is working in an application area where positives should be rare and once the binding sites on the fiber probe are filled, the probe is replaced. Reagents other than antibodies have different requirements for regeneration. An extremely harsh regeneration protocol involved treatment of the AChR biosensor (Rogers et al., 1989, 1991, 1992) with 1% SDS for use in multiple sequential competition assays; however, the receptor survived the regeneration process for extended periods of use. Evanescent wave sensors based on DNA hybridization or that have oligonucleotides as recognition elements can generally also withstand relatively harsh regeneration conditions. Regeneration procedures can involve simply melting the duplex or triplex DNA (Uddin et al., 1997; Piunno et al., 1995; Abel et al., 1996) or may involve chemical treatments with low salt buffers (Rogers et al., 2001), 50% urea (Duveneck et al., 1996; Abel et al., 1996), or 90% formamide (Liu and Tan, 1999; Watterson et al., 2001). Both thermal and urea-based regeneration procedures have been demonstrated to extend fiber optic biosensor uses up to 400 sequential assays (Abel et al., 1996).

2. History The concept of the evanescent wave was first introduced by Hirschfeld (1965) in the mid-1960' s. Ten years elapsed before this concept was put into practice for immunoassays by Kronick and Little (1975). These researchers attached haptens to a planar waveguide and detected binding of evanescently excited fluorescent antibodies both in direct binding and competitive assay formats. At about the same time, Hesse (1974) patented the use of a fiber optic-based sensor for 02 and iodide. However, the combination of evanescent wave and fiber optic technologies did not occur until 10 years after Hesse's patent. The first evanescent wave fiber optic immunosensor, developed by Hirschfeld (1984) was further optimized by Andrade et al. (1985) and Sutherland et al. (1984). These latter groups demonstrated the greater efficiency of in-line fluorescent detection (versus perpendicular detection), whereby both excitation light and emitted fluorescence could be propagated in the waveguide. During the late 1980's, research on evanescent wave systems focused on the development of the immobilization chemistry and the integration of lasers for 75

Taitt and Ligler better discrimination of excitation and emission. In the early 1990's, the advent of near infrared dyes and diode lasers instigated a movement toward small, portable systems that maintained the sensitivity of the laboratory systems with their heavy lasers. For the first time, the promise of a sensitive, robust system for field use started driving fiber optic biosensors toward applications in environmental monitoring, food safety, and point-of-care clinical applications. Since the early 1990's, the majority of publications in the field of evanescent fiber optic biosensors have described new biochemical assays, targeted toward the detection of analyte in complex, mostly unprocessed samples, with some discussion of further refinements to the optical and fluid delivery systems. In short, the majority of evanescent wave fiber optic biosensors are still used in laboratory settings, using instruments and protocols that require both a controlled environment and trained operators. A recent addition to the evanescent wave fiber optic field is a completely automated, portable system capable of use by an untrained user out of doors. This instrument, Research International's RAPTORPlus, will be discussed in greater detail in the next section.

3. State of the Art The majority of recent work in evanescent wave fiber optic biosensors has been in one of the following broad areas: 1) immunoassay development and analyte detection in real-world samples; 2) semi-selective recognition schemes for detection of classes of analytes; 3) detection of nucleic acid sequences, and 4) extending the capabilities of the instrumentation (miniaturization, automation, multi-analyte detection, continuous monitoring).

3.1. Immunoassays for specific analyte recognition The vast majority of assays performed using evanescent wave fiber optic biosensors are immunoassays. There has been a great deal of research recently into expanding the breadth of analytes that can be detected using antibody-based techniques, as well as methods for improving sensitivity of these assays and testing of analytes in non-homogeneous matrices. Since only 1995, the plethora of analytes for which these biosensor immunoassays have been developed include the following: viruses (Lee and Thompson, 1996, King et al., 2000); gram-negative bacteria (Pease et al., 1995, King et al., 2000; Anderson et al., 2000; DeMarco et al., 1999; Tims et al., 2001); gram-positive bacteria, both vegetative and spore forms (Anderson et al., 2001); protozoa (Anderson and Rowe-Taitt, 2001); toxins (Tempelman, et al., 1996; James, et al., 1996; Thompson and Maragos, 1996; Narang et al., 1997b; King et al., 1999); physiological markers of health or disease (Cat et al., 1995; Anderson et al., 1996; Nath et al., 1997; Rowe et al., 1998; Anderson et al., 1998; 2001; Spiker

76

Evanescent Wave Fiber Optic Biosensors Table 1. Evanescent fiber optic biosensors for detection in complex matrices. ii

t

i

Analyte Cocaine, benzyoylecgonine, and cocaethylene Cocaine TNT TNT, RDX

Sample matrix Urine

Reference Nath et al., 1999

Coca leaf extract River water, bilge water Groundwater

Bacillus globigii

Air samples

Ricin

River water, urine river water blood, plasma, serum

Topporada et al., 1997 Shriver-Lake et al., 1995a Shriver-Lake et al., 1997 van Bergen et al., 2000 Ligler et al., 1998 Anderson et al., 1999 Ogert et al., 1993 Narang et al., 1997b Cao et al., 1995 Anderson et al., 1996 Rowe et al., 1998 Rowe-Taitt, unpublished King et al., 1999

Yersinia pestis F1 antigen

D-Dimer Staphylococcal enterotoxin B Staphylococcal enterotoxin B Specific antibody Giardia Burkholderia cepacia Lipopolysaccharide (LPS) Hormones, cytokines E. coli O157:H7 E. coli O157:H7 ....

plasma whole blood Clay, topsoil, pollen, smoke extracts Serum, urine, ham extract Serum Fecal extracts; pond, river and sea water Ground water 20% plasma plasma ground beef extract

...................unpasteur!zed apple juic e

Templeman et al., 1996 Nath et al., 1997 Anderson et al., 1998 Anderson and Rowe-Taitt, 2001 Pease et al., 1995 James et al., 1996 Erb et al., 200 lb DeMarco et al., 1999; DeMarco and Lim, 2002a, .......DeMarco and Lim, 2002b

and Kang, 1999; Erb et al, 2001b); serum antibodies (Anderson et al., 1998), hormones (Erb et al., 2001a); explosives (Shriver-Lake et al., 1997, 1998; Bakaltcheva et al., 1999); and drugs of abuse (Devine et al., 1995; Topporada et al., 1997; Nath et al., 1999). In all of these cases, the intent is to develop an assay that functions in the presence of complex sample matrices. The lack of sensitivity of the evanescent wave detection to fluorophores more than a few hundred nanometers from the surface, coupled with the use of dyes that excite above 600 nm where few natural compounds fluoresce, makes the evanescent sensor particularly useful for measuring the formation of a specifically labeled complex without prior separation of the analyte from other sample components. Sample matrices successfully tested for analyte detection include: clinical fluids (blood, serum,

77

Taitt and Ligler

Figure 10. Schematic of neurotoxin detection by acetylcholinesterase sensor. plasma, sputum, urine, fecal extracts); soil extracts; food homogenates; aerosolized particles (pollen, clay, smokes); and water from various sources, including sea water (Table 1). In most cases, while a diminution or an increase in signal was observed in the presence of the various interferents, the presence of the interferent did not produce false-positive or false-negative results in terms of the ability to detect the analyte. Most effects could be attributed to aggregation of the analyte or increased viscosity of the sample solution/suspension. The assays are indeed proving to be relatively immune from sample interference.

3.2. Semi-selective recognition schemes While antibodies recognize a single analyte, and perhaps very closely related molecules or organisms sharing a common epitope, enzymes and other types of protein receptors may be useful for detecting families of compounds. The number of assays based on enzymes and membrane receptors using evanescent fiber optic biosensors is limited but sufficient for proof of principle. Eldefrawi and his colleagues (Rogers et al., 1992) developed an enzyme-based fiber optic biosensor to detect various neurotoxic compounds (Figure 10). This group immobilized fluorescein-labeled acetylcholinesterase (ACHE), the target enzyme of the neurotoxins, onto optical fibers. This enzyme catalyses the following reaction: Acetylcholine ...... > choline + acetic acid

(7)

In the presence of choline, the reaction above causes the pH to drop; as fluorescein is a pH-sensitive dye, the consequent dip in pH resulted in quenched fluorescence and a low fluorescence signal. In the presence of an AChE-specific neurotoxin, however, the enzymatic reaction was inhibited, causing the pH to remain neutral and fluorescein emission to remain detectable. This group was able to detect paraoxon, echotiophate, bendiocarb, and methomyl. 78

Evanescent Wave Fiber Optic Biosensors Receptors and binding proteins have also been used as recognition elements. Rogers et al. (1991, 1992) used acetylcholine receptors (AchR) to detect various cholinergic agonists and antagonists. In a competitive assay, fluorescein-labeled bungarotoxin in solution competed with unlabeled agonist (or antagonist) to determine inhibition constants for acetylcholine, carbamylcholine, nicotine, Naja naja toxin, bungarotoxin, and tubocurarine. Constants for the antagonists (toxins) were similar to Ka values reported with radioligand assays using solubilized preparations of AChR. Erb et al. (2001b) developed an assay for estrogenic compounds using a competition between an immobilized ligand and the test compound for binding of a labeled recombinant human estrogen receptor. They were able to determine relative binding affinities for a variety of both estrogenic and anti-estrogenic compounds including beta-estradiol, estrone, estriol, diethylstilbestrol, zearalenone, and tamoxiphen. Several other groups have utilized the natural affinity of binding molecules for analytes of interest. James and coworkers (James et al., 1996; Ligler and James, 1999) developed a biosensor based on the high affinity of the antibiotic polymyxin B for the lipopolysaccharide (LPS) component of gram-negative bacterial cell walls. After immobilizing polymyxin B onto optical fibers, the authors were able to detect 25 ng/ml LPS spiked into samples containing up to 20% plasma; assays were complete within 30 seconds. 3.3. Detection of nucleic acid sequences

Since Squirrel and his colleagues (Graham et al., 1992) first described a hybridization assay on the surface of an optical fiber probe, the use of evanescent fiber optic biosensors to detect specific nucleic acid sequences has been steadily increasing. These assays include an intriguing variety of formats and target analytes; the oligomer may be the target analyte, the receptor, or both. Mauro and coworkers (Mauro et al., 1996; Campbell, 1995) developed a rapid evanescent wave sensor for detection of fluorescent polymerase chain reaction (PCR) products by engineering a chimeric recognition molecule. The chimeric molecule was composed of the DNA-binding domain of a yeast transcriptional regulator protein (N-terminus) with the IgG-binding domain of Protein G (Cterminus). This protein was then immobilized onto IgG-coated optical fibers (via the Protein G domain-IgG interaction) and was used to detect nanomolar concentrations of PCR product encoding the yeast DNA operator sequence in less than 10 minutes; multiple sequential analyses could be performed with only a 1-minute regeneration time. While this biosensor was not designed to reduce the time for PCR amplification, it provided a significant reduction in the time required to detect the PCR products.

79

Taitt and Ligler A great deal of work has been published on detection of specific nucleic acid sequences using hybridization reactions performed on evanescently excited optical fibers. Many of these DNA sensors utilize fluorescently labeled sequences (e.g., labeled PCR products) to demonstrate binding of DNA or RNA samples to immobilized oligonucleotides (Graham et al., 1992; Abel et al., 1996; Pilevar et al., 1998). These "direct" detection schemes require time-consuming labeling steps prior to the biosensor assay, but the labeling can be concurrent with the amplification. In a different labeling approach, multiple groups have utilized intercalating dyes such as ethidium bromide to monitor hybridization of duplex (Piunno et al., 1994, 1995; Uddin et al., 1997; Duveneck et al., 1996) and triplex DNA (Uddin et al., 1997). In one of the few instances of a displacement assay performed on an evanescent wave biosensor, Pandey and Weetall (1995) generated response curves for the displacement of ethidium bromide from DNA duplexes immobilized on optical fibers by other intercalating compounds; this biosensor could therefore be used as a generic sensor for aromatic carcinogens. Temperature-sensitive DNA-hybridization experiments provide the ability of evanescent wave sensors to characterize DNA/DNA or DNA/RNA interactions. By measuring the effect of temperature on fluorescence intensity of hybrids treated with ethidium bromide, Uddin and coworkers (1997) were able to demonstrate formation of both parallel and antiparallel triplex DNA complexes. Melting profiles of duplex DNAs were also used recently by Rogers et al. (2001) to assess the extent of radiation-induced damage to an oligonucleotide probe. Real-time measurements have also been utilized by Pilevar and coworkers (1998) in a "sandwich"-type assay to detect Helicobacter pylori rRNA. Total RNA prepared from H. pylori was hybridized to immobilized oligonucleotides complementary to H. pylori rRNA. A subsequent addition of fluorescent detector probe, complementary to a closely related site, allowed detection of the H. pylori RNA. Recently, Watterson and coworkers (2001) used an evanescent wave system in a series of elegant experiments to dissect the different processes involved in nucleic acid hybridization versus adsorption. Using a stopped-flow system for injection and mixing, this group followed adsorption of complementary and noncomplementary DNA sequences to DNA probes immobilized on optical fibers. They determined that adsorption occurred at a significantly higher rate than hybridization, but that this non-selective adsorption did not inhibit selective interaction between complementary sequences. Surprisingly, their results indicated that the presence of high concentrations of non-complementary DNA reduced the response time of the sensors for hybridization of complementary DNA present in the same sample.

80

Evanescent Wave Fiber Optic Biosensors

a

B

--_ -

,

complementary DNA binds, i Opens loop

Figure 11. Schematic of "reagentless" detection of DNA sequences using molecular beacons. A) The molecular beacon is synthesized to possess a 5'-fluorophore, a 3'quencher, a 3'-biotin, and a stem-and-loop structure; the loop domain of the stem-andloop structure is complementary to the sequence of interest. In the absence of target sequence, the stem-and-loop structure is closed and fluorescence emission is quenched. B) When target sequence is present, the stem-and-loop structure opens up, allowing the complementary loop DNA to hybridize with the target sequence. The open structure increases the distance between the 5'-fluorophore and the 3'-quencher, alleviating the quenching; fluorescence intensity increases.

A novel recognition scheme for "reagentless" detection of DNA sequences by an evanescent wave sensor was recently described by Liu and Tan (1999). The authors synthesized oligonucleotide "molecular beacons" that contained a stemand-loop structure; the loop region was complementary to the target sequence and the two ends were labeled with either a fluorophore (5') or a quencher (3') (Figure 11). The molecular beacon was immobilized on an avidin-coated optical fiber by an avidin-biotin bridge; the 3' end of the molecular beacon (close to the quencher) was modified to contain a biotin. In the absence of the target sequence, the stem-and-loop structure of the molecular beacon was conserved, leading to background levels of fluorescence. If target DNA (or RNA) with a complementary sequence was added to the system, hybridization of the target DNA to the complementary sequence in the loop region caused the stem-andloop structure to open up. This resulted in movement of the fluorophore and quencher arms away from each other. The increased distance between the ends no longer allowed efficient quenching, and fluorescent emission was observed. 81

Taitt and Ligler

3.4. Extending the capabilities of the instrumentation The first evanescent fiber optic biosensor (Hirschfeld, 1984; Hirschfeld and Block, 1985) was made with a halogen lamp in order to keep the device small and was sold by ORD (Waltham, MA) incorporated into a small metal suitcase. The subsequent laser-based breadboard devices were all significantly larger, and the advent of commercial diode lasers helped to maintain the advantages of the collimated, narrow bandwidth excitation and small size. The Analyte 2000 was the first device that was not only portable, but that interrogated more than one fiber probe. This instrument contains four integrated "daughter" cards holding four individual pulsed 635 nm diode lasers and four photodiodes. Fiber bundles (Figure 4C) connect the optics to the fiber probes and the electronic signals from the Analyte 2000 are analyzed on a laptop computer. The configuration of this system allowed four separate assays to be performed simultaneously, either in parallel or in series, depending on the configuration of the fluidics components assembled by the user. This instrument was used in many of the early demonstrations of multi-analyte detection using evanescent wave technology (Table 2). Table 2 also shows a number of "semi-selective" fiber optic sensors that have been used for detection of multiple analytes using a single recognition element. These sensors utilized recognition elements that bound to multiple ligands that were often structurally related or had similar modes of action. These latter sensors can be used for detection of multiple analytes, but identification of the specific analyte detected may require additional, more-specific analyses. In addition to miniaturization, automation has been a major thrust in the move toward commercialization of biosensors and has facilitated use of the biosensors in the field. Oroszlan et al. (1993) automated the biosensor so that multiple assays could be used to screen large numbers of samples for clinically relevant analytes. In order to perform continuous monitoring remotely, Ligler et al. (1998) also developed a computer controlled fluidics module to deliver samples and reagents to the four probes attached to an Analyte 2000 (Figure 6). ShriverLake et al. (1998) employed a similar version of this fluidics module in order to run multiple assays simultaneously in the field; automation not only enabled the continous monitoring and multi-analyte capabilities, but it also increased the reproducibility of the assays.

82

Evanescent Wave Fiber Optic Biosensors Table 2. Evanescent wave fiber optic biosensors used for multi-analyte detection. Analytes

,

~ ....

TNT, RDX Ovalbumin, B..globigii Ovalbumin, Erwinia herbicola, B. globig, ii, MS2 SEB, ricin, Francisella tularensis, B. g.lobigii 0valbumin, B. anthracis, Giardia, F. tularensis Ricin, SEB, cholera toxin, ovalbumin

,_i. Biosens~

_ .Specifi c Identification Antibody: Analyte 2000

I Reference

Antibody..! .Analyte 2000 Antibody: RAPTOR

Bakaltcheva et al., 1999 van Bergen et. al., 2000_ Anderson et al., 1999 King et al., 2000

Antibody: RAPTOR

Anderson et al., 2000

Antibody: RAPTOR

Anderson and Rowe-Taitt, 2001; Anderson et al., 2001 Anderson et al., 2001

Antibody: RAPTOR

Semi-selective Analyses Remazol brilliant blue, EtBr-duplex DNA anthraquinone-2,6-disulfonic (displacement reaction)" acid, decacycliene, DAPI . ORD device Acetylcholine, Acetylcholinesterase: carbamylcholine, nicotine, ORD device Naja naja toxin, bungarotoxin, tubocurarine Cocaine, benzyoylecgonine, Antibody: ORD device cocaethylene, cocaine Antibody: Analyte 2000 metabolites Acetylcholine, AchR: ORD device benzoylcholine, benzylacetate

Pandey and Weetall, 1995 Rogers et al., 1991, 1992

Devine et al., 1995 Nath et al., 1999 Rogers et al., 1989

There are two evanescent fiber optic biosensors currently on the market intended for use by non-technically trained individuals. The first is a single-use system designed primarily for clinical applications (Erb et al., 2001a, 2001b) known as the Endotect TM (ThreeFold Sensors, Ann Arbor, MI; http://ic.net/--tfs) (Figure 12). It uses single fibers snapped into a disposable cartridge and measures binding rates in competitive assays. The second system is a portable, fully automated sensor capable of either continuous monitoring or measuring discrete samples repeatedly using a single coupon containing four plastic probes. The RAPTOR-Plus (Research International, Woodinville, WA; http://ww.w.resrchintl.com) weighs about 12 pounds, includes a display to convey the qualitative results, and has an RS232 port to download the quantitative data to a laptop or for remote transmission (Figure 13). Both binding rates and total amount of bound labeled antibody are evaluated in order to minimize the possibility of a false positive response (Anderson et al., 2000). The optics are essentially the same as in the Analyte

83

Taitt and Ligler

Figure 12. The EndotectTM system. 2000, but the fluidics, plastic probes and coupon, and battery operation are a step forward. The RAPTOR-Plus has been hardened for military operation, but is adaptable for civilian applications as well. While the studies published using the RAPTOR-Plus and previous versions of the device (designated as RAPTOR in Table 2) describe the analyses of samples individually introduced into the sensor, the RAPTOR-Plus has the capability to control a portable air sampler (the SASS, Research International) and analyze the effluent samples automatically at userdefined intervals. Since the results can be stored or transmitted remotely, no operator need be present during the monitoring period. Once a large positive 84

Evanescent Wave Fiber Optic Biosensors

Figure 13. The RAPTOR-Plus and coupon containing four fiber probes. signal is obtained, however, and the binding sites on the fiber probes are occupied, the coupon containing the probes does need replacement.

4. Advantages and Limitations Like all optical biosensors, fiber optic sensors have several advantages over sensors that utilize other signal transduction methods. No direct electrical connection to the transduction system is required. Therefore, optical sensors are immune to many interfering electrochemical and electromagnetic effects that plague sensors based on electrochemical transduction, such as electromagnetic radiation, corona discharge, radiofrequency interferences, corrosion, shock/vibrations, and harmonic induction. Furthermore, the lack of requirement for radiolabels or hazardous organic materials for sample preparation or signal generation (such as benzene and other organic solvents) render optical sensors more "user-friendly" than other methods that do. In addition, over the years, with the increased availability and quality of small-scale light sources, the size and price of optical biosensors have continued to decrease. 85

Taitt and Ligler Fiber optic biosensors, in particular, have reaped the benefits of the telecommunications boom that has led to the commercial availability of inexpensive, high quality optical fibers. The use of optical fibers in biosensors takes advantage of the versatility of these components. System configurations can be changed to suit the user's specifications. Shortened fibers can be used in sensors designed for portability, while fibers of extended length allow assays to be performed in remote or hazardous locations. To date, evanescent fiber probes have not been explored for in vivo applications due to the requirement for additional reagents in sandwich and competitive assays. Preininger et al. (2000) demonstrated that binding molecules could be encapsulated in sol gels on the surface of the optical fiber probe and generate a signal. While the assay used was a fluorescent immunoassay, the use of assays using enzymes, molecular beacons, or artificial receptors (Chapters 10-13) in a sol gel or polymer matrix at the surface may be appropriate for in vivo applications. Sensors based on evanescent wave excitation have the additional advantage of surface-specific detection. The limited penetration depth of the evanescent wave allows spatial separation of surface-bound fluorescent complexes from those present in the bulk solution, obviating the need for washing steps; in practice, however, washing steps are often included to minimize any effects from the tracer reagents present within the evanescent field. Use of the evanescent field thus allows real-time measurement of surface interactions to be performed both in homogeneous and non-homogeneous samples. Experimental realization of this capability has occurred only recently (Table 1). Furthermore, diffusional barriers exacerbated by viscous sample matrices have also been minimized by incorporation of flow into the sensors' design. In optical sensors, the speed with which results are available is limited by the rate of molecular interaction, not by the signal transduction. This advantage is further improved by the surface-selective nature of the evanescent wave, which allows the capability of real-time measurements. While real-time measurement of surface binding events is possible with evanescent wave sensors, this capability has not received widespread use; even so, however, assay times for evanescent wave fiber optic sensors are at least an order of magnitude shorter than those from ELISAs, considered the gold standard for bioassays. Real-time measurement has been demonstrated in evanescent wave fiber optic biosensors based on both antibody binding and nucleic acid hybridization (and melting). Recent experiments utilizing time-resolved fluorescence have further demonstrated the potential of such real-time measurements for determination of binding constants and enhanced discrimination between different fluorescent species. Creation of true multi-analyte evanescent wave fiber optic biosensors is a relatively recent phenomenon (Table 2). With smaller optical components readily available, the only limitation to the number of biochemical assays that 86

Evanescent Wave Fiber Optic Biosensors can be performed simultaneously (in series or in parallel) will depend only on the compatibility of the biochemical reagents and possible limitations inherent in sensors of increased complexity, e.g., increased cost, size, maintenance. A key disadvantage of evanescent wave systems is the lack of power within the evanescent wave for generating fluorescence. While single-mode fibers have up to 50% of their energy present outside the fiber core, photobleaching of surfacebound fluorophores and efficient coupling of emitted fluorescence back into the fiber remain problematic. Moreover, the surface area of monomode fibers available for coupling chemical recognition elements is low. Multi-mode fibers, on the other hand, have a greater surface area available for immobilization of recognition elements. The ratio of the surface area on an evanescent fiber probe compared to the surface area at the end of a fiber of the same radius, r, is 2L/r, where L is the length of the sensing region; since fibers with a radius of 0.1-0.5 mm and lengths of 40-100 mm are frequently described, this ratio typically varies from 400-800. In addition to the large surface area for immobilization of the biological recognition molecules, with appropriate tapering and/or V-number matching, multimode fibers have improved in their excitation and emission collection efficiencies. As with all sensors based on biological recognition elements, stability and regeneration of evanescent wave fiber optic biosensors must be characterized for each recognition system. Long-term storage also depends of stability of the recognition elements, but storage times of over 12-18 months have been demonstrated for antibodies immobilized on fiber optic probes and lyophilized fluorescent antibody reagents (Ligler et al, 1992; King, et al., 2000). While regeneration of hybridization-based sensors is simply a matter of "melting" the complexes, regeneration of sensors utilizing protein recognition elements only seems useful for monovalent analytes, due to inactivation or denaturation of the recognition elements when the analyte is bound to immobilized molecules at several epitopes. At this time, long-term monitoring is feasible only if a positive is a rare event (and no regeneration is required) or if the association between the analyte and recognition molecule is readily reversible.

5. Future of Evanescent Fiber Optic Biosensors There are opportunities for improving each part of the evanescent fiber optic biosensor while still utilizing the advantages of this technology. We will discuss the areas that we predict will offer such opportunities in terms of the biorecognition molecules, the fluidics, tile waveguide materials and optics, and systems integration. The optical biosensor is only as good as the molecular recognition molecule it uses. As discussed in several other chapters (10-13) in this book, new types of 87

Taitt and Ligler recognition molecules are being developed using different forms of combinatorial biochemistry. Single chain Fv antibodies have been created with high affinities, molecularly imprinted polymers are showing increased specificity, and Kleinjung et al. (1998) have already used RNA aptamers with fiber optic-based biosensors. Goldman et al. (2000) also demonstrated the use of phage-displayed peptides as detector reagents. After several rounds of biopanning to isolate M13 constructs expressing surface-bound peptides with affinity to staphylococcal enterotoxin B (SEB), the phage were labeled with a fluorescent dye and used as detection reagents in an evanescent wave fiber optic biosensor. While these peptides were able to distinguish between optical fibers coated with SEB from those coated with streptavidin, they were unable to detect SEB bound to the fibers in a sandwich assay. Clearly these technologies for production of new receptors are not yet straightforward. However, they do show promise not only for obtaining molecules with designer specificity, but also with enhanced stability for extended use or exposure to samples incompatible to most antibodies (e.g., high salt, organic solvent). The work of Mauro et al. (1996) using a chimeric Protein G-DNA binding protein also suggests an additional approach to improving the biological recognition molecule. The Protein G binding site provided a means for easily immobilizing the molecule and also stabilizing the binding protein. Using molecular engineering, one could design and construct receptors with ligand binding moieties, a stabilization motif, and a region designed to facilitate immobilization. Furthermore, building on approaches like the molecular beacon concept, one could theoretically also introduce a site that would produce a fluorescence signal upon analyte binding. And just to make the molecule perfect, why not add a site that would cause the receptor to release bound analyte and return to its nonfluorescent state? The second area for improving evanescent fiber optic biosensors is in the area of controlling the transport of sample and reagents to the recognition molecules on the fiber probes. Several approaches have been used to overcome diffusion limitations on analyte transport. The fluidics systems developed for the Analyte 2000 and RAPTOR use flow to expose the fiber probe to a maximum amount of analyte; the RAPTOR can even slosh the sample back and forth over the probe if desired. However, the low Reynolds number flow in the capillaries surrounding the fiber probes means that mixing of fluids at the fiber surface is very limited. Recently, Zhou et al. (1998) described a method for increasing the local analyte concentration in a fiber optic biosensor by sonication. The authors demonstrated that the ultrasonic standing-wave chamber concentrated Salmonella typhimurium into parallel layers or in a column along the axis of the testing cell. Sonication had no effect on the antibodies, but forced cells (and cells bound to polystyrene microspheres) to move to the axis of the test cell. Fluorescent signals were an order of magnitude higher in tests utilizing the acoustic standing-wave than those when the ultrasonic treatment was absent. Another alternative approach for 88

Evanescent Wave Fiber Optic Biosensors increasing transport of analyte to the surface might include the use of passive microfluidic structures such as those discussed in Chapter 18. The third area for improvement involves the optical components. The probes could be fabricated from materials that have absolutely no fluorescence and the molds could be made so that the probes are absolutely smooth and the geometry optically perfect. Diode lasers at lower wavelengths would make new labels useful, including the quantum dots discussed in Chapter 17. One could envision a single UV diode laser exciting a probe fabricated with a new UV-transparent plastic and measuring the formation of complexes containing different quantum dots, discriminated using a single holographic emission filter. Altematively, a bundle of fiber probes could be excited using a single array of VIXELS and the emission collected using a CMOS detector coated with a microlens array. The exciting advances in the solid state optics will open new doors for smaller sensors. And finally, the importance of system integration cannot be understated. In this chapter, we have discussed the sensing biochemistry, fiber probes, optics and fluidics as separate elements of the sensor. However, small size and effective automation rely on the ability to make these components work together. Furthermore, the more functions that a single structure can perform, i.e., waveguiding and fluid containment (Ligler et al., 2002), the more efficient the system might be. Research in the area of microfluidics is already producing sensors on a single substrate that includes a light source and photodectors. Systems integration has the potential, more than ever before, to produce monolithic structures with multiple functionalities. The biosensor of the future might resemble a pack of cards, with single card constituting a multianalyte probe.

6. Acknowledgements The authors thank the Office of Naval Research for supporting the preparation of this chapter.

7. References Abel, A.P., M.G. Weller, G.L. Duveneck, M. Ehrat and H.M. Widmer, 1996, Anal. Chem. 68, 2905. Anderson, G.P., K.A. Breslin and F.S. Ligler, 1996, ASAIO J. 42, 942. Anderson, G.P., J.P. Golden, L.K. Cao, D. Wijesuria, L.C. Shriver-Lake and F.S. Ligler, 1994a, IEEE Eng. Med. Biol., 13, 358. Anderson, G.P., J.P. Golden and F.S. Ligler, 1994b, IEEE Trans. B iomed. Eng., 41, 578. 89

Taitt and Ligler Anderson, G.P. and J. Golden, 1995, US Patent No. 5,430,813. Anderson, G.P., J.P. Golden and F.S. Ligler, 1993, Biosens. Bioelectron. 8, 249. Anderson G.P., M.A. Jacoby, F.S. Ligler and K.D. King, 1997, Biosens. Bioelectron. 12, 329. Anderson, G.P., K.D. King, L.K. Cao, M. Jacoby, F.S. Ligler and J. Ezzell, 1998, Clin. Diagn. Lab. Immunol. 5,609. Anderson, G.P., K.D. King, D.S. Cuttino, J.P. Whelan, F.S. Ligler, J.F. MacKrell, C.S. Bovais, D.K. Indyke and R.J. Foch, 1999, Field Anal. Chem. Technol. 3, 307 Anderson, G.P., K.D. King, K. L. Gaffney and L. H. Johnson, 2000, Biosens. Bioelectron. 14, 771. Anderson, G.P. and C.A. Rowe-Taitt, 2001, SPIE 4206, 58. Anderson, G.P., C.A. Rowe-Taitt and F.S. Ligler, 2001, Proc. 1st Conf. Point Detect. Chem. Biol. Defense, Oct. 2000, Williamsburg, VA. Anderson, G.P., L.C. Shriver-Lake, J. P. Golden and F.S. Ligler, 1992, Proc. SPIE 1648, 39. Andrade, J.D., R.A. Van Wagenen, D.E. Gregonis, K. Newby and J.N. Lin, 1985, IEEE Trans. Electron. Devices 32, 1175. Anis, N.S., J. Wright, K.R. Rodgers, R.G. Thompson, J.J. Valdes and M.E. Eldefrawi, 1992, Anal. Lett. 25,627. Bakaltcheva, I.B., F.S. Ligler, C.H. Patterson and L.C. Shriver-Lake, 1999, Anal. Chim. Acta 399, 13. Bakaltcheva, I.B., L.C. Shriver-Lake and F.S. Ligler, 1998, Sens. Actuators B. Chem. 51, 46. Bhatia, S.K., L.C. Shriver-Lake, K.J. Prior, J.H. Georger, J.M. Calvert, R. Bredehorst and F.S. Ligler, 1989, Anal. Biochem. 178,408. Block, M.J. and T.B. Hirschfeld, 1986, US Patent No. 4,582,809. Block, M.J. and T.B. Hirschfeld, 1987, US Patent No. 4,716,121. Block, M.J., S.J. Lackie and T.R. Glass, 1990, US Patent No. 4,909,990. Bock, D., K. Galla, M. Martin, J. Wolfrum and S. Seeger, 1995, Sens. Actuators B Chem. 29, 293. Boisd6, G. and A. Harmer, 1996. Chemical and Biochemical Sensing with Optical Fibers and Waveguides, Artech House, Boston, 389 pp. Browne, C.A., D.H. Tarrant, M.S. Olteanu, J.W. Mullens and E.L. Chronister, 1996, Anal. Chem. 68, 2289. Buerck, J., S. Roth, K. K.raimer, M. Schloz and N. Klaas, 2001, J. Hazard. Mater. 83, 11. Campbell, J.R., 1995, US Patent Reg. No. H1398. Cao, L.K., G.P. Anderson, F.S. Ligler and J. Ezzell, 1995, J. Clin. Microbiol. 33, 336. Carlyon, E.G., C.R. Lowe, D. Reid and I. Bennion, 1992, Biosens. Bioelectron. 7, 141. Choa, F.S., M.H. Shih, A.R. Toppozada, M. Block and M.E. Eldefrawi, 1996, Anal. Lett. 29, 29.

90

Evanescent Wave Fiber Optic Biosensors DeLisa, M.P., Z. Zhang, M. Shiloach, S. Pilevar, C.C. Davis, J.S. Sirkis and W.E. Bentley, 2000, Anal. Chem. 72, 2895. DeMarco, D.R., E.W. Saaski, D.A. McCrae and D.V. Lim, 1999, J. Food Protect. 62, 711. DeMarco, D.R. and D.V. Lim, 2002a, J. Food Protect., in press. DeMarco, D.R. and D.V. Lira, 2002b, J. Rapid Meth. Automation Microbiol., in press. Devine, P.J., N.A. Anis, J. Wright, S. Kim, A.T. Eldefrawi and M.E. Eldefrawi, 1995, Anal. B iochem. 227, 216. Duveneck, G.L., P. Oroszlan, A.P. Abel, B. Klee, V. Steiner, M. Ehrat, D. Gygax and H.M. Widmer, 1996, Proc. SPIE 2631, 14. Einink, R.G., H.E. de Bruijn, R.P.H. Kooyman and J. Greve, 1990, Anal. Chim. Acta 238, 317. Eigler, F.S., J. Georger, S.K. Bhatia, J. Calvert, L.C. Shriver-Lake and R. Bredehorst, 1991, US Patent No. 5,077,210. Erb, J. and J. Downward, IV, 1998, US Patent No. 5,854,863. Erb, J.L., J.G. Downward, J.R. Erb-Downward and U. Otho, 2001a, US Patent No. 6,251,688. Erb, J.L., J.G. Downward, J.R. Erb-Downward and J.L. Wittliff, 2001b, US Patent No. 6,300,082. Fang, X.H. and W.H. Tan, 1999, Anal. Chem. 71, 3101. Feldman, S.F., E.E. Uzgiris, C.M. Penney, J.Y. Gui, E.Y. Shu and E.B. Stokes, 1995, Biosens. Bioelectron. 10, 423. Gao, H.H., Z. Chen, J. Kumar, S.K. Tripathy and D.L. Kaplan, 1995, Proc. SPIE 34, 3465. Glass, T.R., 1989, US Patent 4,844,869. Gloge, D., 1971, Appl. Opt. 10, 2252. Golden, J.P., G.P. Anderson, S.Y. Rabbany and FS Ligler, 1994, IEEE Trans. Biomed. Eng. 41,585. Golden, J.P., E.W. Saaski, L.C. Shriver-Lake, G.P. Anderson and F.S. Ligler, 1997, Opt. Eng. 36, 1008. Golden, J.P., L.C. Shriver-Lake, G.P. Anderson, R.B. Thompson and F.S. Ligler, 1992, Opt. Eng. 31, 1458. Golden, J.P., L.C. Shriver-Lake, N. Narayanan, G. Patonay and F.S. Ligler, 1994, Proc. SPIE 2138, 241. Goldman, E.R., M.P. Pazirandeh, J.M. Mauro, K.D. King, J.C. Frey and G.P. Anderson, 2000, J. Molec. Recog. 13,382. Graham, C.R., D. Leslie and D.J. Squirrel, 1992, Biosens. Bioelectron. 7,487. Hale, Z.M., F.P. Payne, R.S. Marks, C.R. Lowe and M.M. Levine, 1996, Biosens. Bioelectron. 11,137. Harrick, N.J., 1967, Internal Reflection Spectrosopy, Wiley, New York, 327 pp. Herron, J.N., D.A. Christensen, K.D. Caldwell, V. Janatova, S.-C. Huang and H.K. Wang, 1996, US Patent No. 5,512,492. Hesse, H.C., 1974, East German Patent 106,086. Hirschfeld, T., 1965, Can. J. Spectrosc. 10, 128. 91

Taitt and Ligler Hirschfeld, T., 1984, US Patent No. 4,447,546 Hirschfeld, T.E. and M.J. Block, 1984, US Patent No. 4,447,546 Hirschfeld, T.E. and M.J. Block, 1985, US Patent No. 4,558,014. James, E.A., K. Schmeltzer and F.S. Ligler, 1996, Appl. Biochem. Biotechnol. 60, 189. Jorgenson, R.C. and S. S. Yee, 1993, Sens. Actuators B 12, 213. King, K.D., G.P. Anderson, K.E. Bullock, M.J. Regina, E.W. Saaski and F.S. Ligler, 1999, Biosens. Bioelectron. 14, 163. King, K.D., J.M. Vanniere, J.L. LeBland, K.E. Bullock and G.P. Anderson, 2000, Environ. Sci. Technol. 34, 2845. Kleinjung, F., S. Klussmann, V.A. Erdmann, F.W. Scheller, J.P. Furste and F.F. Bier, 1998, Anal. Chem. 70, 328. Kooyman, R.P.H., H.E. de Bruijn and J. Greve, 1987, Proc. SPIE 798, 290. Kronick, M.N. and W.A. Little, 1975, J. Immunol. Meth. 8, 235. Krull, U.J., R.S. Brown, R.F. DeBono and B.D. Hougham, 1988, Talanta 35, 129. Lackie, S.J., 1992, US Patent 5,152,962. Lee W.E. and H.G. Thompson, 1996, Can. J. Chem. 74, 707. Lew, A., C. Depeursinge, F. Cochet, H. Berthou and O. Parriaux, 1986, Proc. SPIE, 514, 71. Ligler, F.S., G. P. Anderson, P.T. Davidson, R.J. Foch, J.T. Ives, K.D. King, G. Page, D.A. Stenger and J.P. Whelan, 1998, Environ. Sci. Technol. 32, 2461. Ligler, F.S., M. Breimer; J,P. Golden, D.A. Nivens, J.P. Dodson, T.M. Greene, D.P. Haders and O.A. Sadik, 2002, Anal. Chem. 73, in press. Ligler, F.S., K.A, Breslin, L.K. Cao and G.P. Anderson, 1995, Proc. SPIE 2388, 16. Ligler, F.S., J.P. Golden, L.C. Shriver-Lake, R.A. Ogert, D. Wijesuria and G.P. Anderson, 1993, Immunometh. 3, 122. Ligler, F.S. and E.A. James, 1999, US Patent Reg. No. H1775. Ligler, F.S., L.C. Shriver-Lake and R.A. Ogert, 1992, in Proceedings of Biosensors '92, ed. A.P.F. Turner, Elsevier, Oxford, pp. 308-315. Ligler, F.S., L.C. Shriver-Lake and D.C. Wijesuriya, 1996, US Patent No. 5,496,700. Lipson, D.G., N.G. Loebel, K.D. McLeaster and B. Liu, 1992, IEEE Trans. Biomed. Eng. 39, 886. Liu, X. and W. Tan, 1999. Anal. Chem. 71, 5054. Love, W.F. and L.J. Button, 1988, Proc. SPIE, 990, 175. Matsuo, H., S. Kuniyoshi, K. Kudo and K. Tanaka, 2000, Synthetic Metals, 115, 37. Mauro, J.M., L.K. Cao, L.M. Kondracki, S.E. Walz and J.R. Campbell, 1996, Anal. Biochem. 235, 61. Meserol, P., 1996, US Patent No. 5,492,674. Moreno-Bondi, M.-C., J. Mobley, J.-P. Alarie and T. Vo-Dinh, 2000, J. Biomed. Opt. 5,350. 92

Evanescent Wave Fiber Optic Biosensors Narang, U., G.P. Anderson, K.D. King, H.S. Liss and F.S. Ligler, 1997a, Proc. SPIE 2980, 187. Narang, U., G.P. Anderson, F.S. Ligler and J. Burans, 1997b, Biosens. Bioelectron. 12, 937. Nath, N., M. Eldefrawi, J. Wright, D. Darwin and M. Huestis, 1999, J. Anal. Toxicol. 23,460. Nath, N., S.R. Jain and S. Anand, 1997, Biosens. Bioelectron. 12, 491. Neel, T.G. and J.E. Lyst, Jr., 1997, US Patent No. 5,639,668. Ogert R.A., L.C. Shriver-Lake and F.S. Ligler, 1993, Proc. SPIE 185, 11. Oroszlan, P., C. Thommen, M. Wehrli, G. Duveneck and M. Ehrat, 1993, Anal. Meth. Instrumentation 1, 43. Oroszlan, P., S. Wicar, G. Teshima, S.-L. Wu, W.S. Hancock and B.L. Karger, 1992, Anal. Chem. 64, 1623 Orvedahl, D.S., W.F. Love and R.E. Slovacek, 1991, Proc. SPIE 1487, 187. Pandey, P.C. and H.H. Weetall, 1995, Anal. Biochem. 67, 787 Pease, M.D., L. Shriver-Lake and F.S. Ligler, 1995, in B iosensor and Chemical Sensor Technology. ACS Symposium Series 613, eds. K.R. Rogers, A. Mulchandani and W. Zhou, American Chemical Society, Washington DC, pp. 33-43. Pilevar, S., C.C. Davis and F. Portugal, 1998, Anal. Chem. 70, 2031. Piunno, P.A.E., U.J. Krull, R.H.E. Hudson, M.J. Damha and H. Cohen, 1994, Anal. Chim. Acta 288, 205. Piunno, P.A.E., U.J. Krull, R.H.E. Hudson, M.J. Damha and H. Cohen, 1995, Anal. Chem. 67, 2635 Preininger, C., A. Mencaglia and F. Baldini, 2000, Anal. Chim. Acta 403, 67. Polereck3% L., J. Hamrie, and B.D. MacCraith, 2000, Appl. Opt. 39, 3968. Rogers, K.R., N.A. Anis, J.J. Valdes and M.E. Eldefrawi, 1992, In Biosensor Design and Application, Eds. P.R. Mathewson and J.W. Finley, American Chemical Society, Washington, pp.165-173. Rogers, K.R., M.E. Eldefrawi, D.E. Menking, R.G. Thompson and J.J. Valdes, 1991, Biosens. Bioelectron. 6, 507. Rogers, K.R., J.J. Valdes and M.E. Eldefrawi, 1989, Anal. Biochem. 182, 353. Rogers, K.R., A. Apostol, S.J. Madsen and C.W. Spencer, 2001, Anal. Chim. Acta 444, 51. Rowe, D.A., J.S. Bolitho, A. Jane, P.G. Bundesen, D.B. Rylatt, P.R. Eisenberg and F.S. Ligler, 1998, Thromb. Haemost. 79, 94. Saaski, E.W., 2000, US Patent No. 6,082,185. Saaski, E.W. and C.C. Jung, 2000, US Patent No. 6,136,611. Shriver-Lake, L.C., K.A. Breslin, P.T. Charles, D.W. Conrad, J.P. Golden and F.S. Ligler, 1995, Anal. Chem. 34, 2431. Shriver-Lake, L.C., B.L. Donner and F.S. Ligler, 1997, Environ. Sci. Technol. 31,837. Shriver-Lake, L.C., J.P. Golden, G. Patonay, N. Narayanan and F.S. Ligler, 1995, Sens. Actuators B Chem. 29, 25.

93

Taitt and Ligler Shriver-Lake, L.C., N.A. Naz and F.S. Ligler, 1998, In Current Protocols in Field Analytical Chemistry, V. Lopez-Avila, Ed., John Wiley & Sons, New York, p. 2E 1-2E 12. Slovacek, R.E. and W.F. Love, 1992, US Patent No. 5,156,976. Slovacek, R.F., W.F. Love, T.A. Cook, R.L. Schulkind and I.M. Walczak, 1992, US Patent No. 5,346,715. Spiker, J.O. and K.A. Kang, 1999, Biotechnol. Bioeng. 66, 158. Spiker, J.O., K.A. Kang, W.N. Drohan and D.F. Bruley, 1998, Adv. Exp. Med. Biol. 454, 681 Sutherland, R.M., C. Dahne, J.F. Place and A.S. Ringrose, 1984, Clin. Chem. 30, 1533. Templeman, L., K.D. King, G.P. Anderson and F.S. Ligler, 1996, Anal. Biochem. 233, 50. Thompson, R.B., 1991. In Topics in Fluorescence Spectroscopy, Vol. II: Principles, J.R. Lakowicz, Ed., Plenum, New York, pp. 345-365. Thompson, R.B. and M. Levine, 1992, US Patent No. 5,141,312. Thompson, R.B. and F.S. Ligler, 1988, Proc. SPIE 904, 27. Thompson, R.B. and L. Kondracki, 1990, Proc. SPIE 1204, 35. Thompson, V.S. and C.M. Maragos, 1996, J. Agric. Food Chem. 44, 1041. Thompson, R.B. and C.A. Villaruel, 1991, US Patent No. 5,061,857. Tims, T.B., S.S. Dickey, D.R. DeMarco and D.V. Lim, 2001, Am. Clin. Lab. 20, 28. Topporada, A.R., J. Wright, A.T. Eldefrawi, M.E. Eldefrawi, E.L. Johnson, S.D. Emche and C.S. Helling, 1997, Biosens. Bioelectron. 12, 113. Uddin, A.J., P.A.E. Piunno, R.H.E. Hudson, M.J. Damha and U.J. Krull, 1997, Nucl. Ac. Res. 25, 2635. van Bergen, S.K., I.B. Bakaltcheva, J.S. Lundgren and L.C. Shriver-Lake, 2000, Environ. Sci, Technol. 34, 704. Vijayendran, R. A., F.S. Ligler and D.E. Leckband, 1999, Anal. Chem. 71, 5405. Villaruel, C.A., D.D. Dominguez and A. Dandridge, 1987, Proc. SPIE 798, 225 Wadkins, R.M., J.P. Golden and F.S. Ligler, 1995, Anal. Biochem. 232,73. Watterson, J.H., P.A.E. Piunno, C.C. Wust, S. Raha and U.J. Krull, 2001, Fresenius J. Anal. Chem. 369, 601. Wijesuriya, D., K. Breslin, G. Anderson, L. Shriver-Lake and F.S. Ligler, 1994, Biosens. Bioelectron. 9, 585. Williamson, M.L., D.H. Atha, D.J. Reeder and P.V. Sundaram, 1989, Anal. Lett. 22, 803. Zhao, S. and W.M. Reichert, 1992, Langmuir 8, 2785. Zhao, C., P. Pivarnik, A.G. Rand and S.V. Letcher, 1998, Biosens. Bioelectron. 13,495.

94

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 3

PLANAR WAVEGUIDES FOR FLUORESCENCE BIOSENSORS

1KIM SAPSFORD,PH.D., 2CHINSROWETAITT, PH.D. AND2FRANCESS. LIGLER,D.PHIL., D.Sc. 1Center for Bioresource Development, George Mason University, Fairfax, VA 22030-4444, USA 2Center for Bio/Molecular Science and Engineering, Naval Research Laboratory, Washington, DC 20375-5348, USA

Total internal reflection fluorescence (TIRF) is the process whereby fluorophores that are either attached to or in close proximity to the surface of a waveguide are selectively excited via an evanescent wave. The use of a planar waveguide allows the immobilization of multiple capture biomolecules and the possibility therefore of multianalyte detection on a single substrate. Planar waveguide TIRF has been used in the measurement of a variety of analytes including hormones, toxins, bacteria and viruses, leading to applications in areas such as environmental monitoring, clinical diagnostics and military defense. Analytes have been measured both in buffer and in complex matrices, such as whole blood, nasal secretions and soil suspensions. Detection limits both in buffer and complex matrices have been comparable. The continued development and miniaturization of the sensor instrumentation has led to systems that are fully automated and portable and would be highly competitive with current techniques upon transition to the commercial market.

1. Technical Concept Although total internal reflection fluorescence (TIRF) coupled with planar waveguides is a relatively new development for biosensor applications, its history is deeply rooted in the field of fiber optics, where much of the surface chemistry and principles have been studied. TIRF is a means of selectively exciting the fluorescence emission of fluorophores present near the surface of a waveguide and is relatively immune to bulk matrix effects. This technique has 95

Sapsford, Rowe Taitt, and Ligler found numerous applications in the field of biosensors, immunosensors and sensors for DNA analysis.

in particular

Like other biosensors, fluorescence biosensors based on planar waveguides consist of two important features: the molecular recognition element and the signal transduction mechanism (Hall, 1990; Vo-Dinh and Cullum, 2000). The molecular recognition element can take the form of a biomolecule (antibody, enzyme or nucleic acid), biological system (membranes, tissues or whole cells) or biomimetic (synthetic bioreceptors) and imparts specificity to the system. The signal transduction mechanism is the process by which the biochemical recognition event is converted into a measurable signal, the intensity of which is proportional to the analyte concentration. In this case, the molecular recognition event results in the immobilization of the fluorophore at the surface of the planar waveguide, TIRF is used to excite the fluorophore, and the resulting fluorescence is measured.

1.1. Signal transduction and amplication There are three techniques which can be grouped under the principle of reflectance: attenuated total reflectance (ATR) which monitors alterations in the IR, visible and UV-regions; surface plasmon resonance (SPR) which measures variations in refractive index; and total internal reflection fluorescence (TIRF) which monitors changes in fluorescence (Bradley et al., 1987; Lu et al., 1992; Wadkins et al., 1998; Chittur, 1998; Plowman et al., 1998). Fluorescence-based biosensors, which use planar substrates as waveguides, fall under the last category. The basic arrangement of the TIRF system is shown in Figure 1. At the interface of two media with different refractive indices, described by Equation 1, (nx: where nx is either nl or nz) incident light from the higher refractive index medium will be partly refracted and partly reflected, nx = [(ei~ / ~o~to)]a~2

(1)

where ei and eo are the permittivity of light in the dielectric and a vacuum, respectively, and ~i and lXoare the permeability of light in the dielectric and a vacuum, respectively. However, when the angle of incidence is greater than the critical angle (0~), given by Equation 2, the phenomenon of total internal reflection is observed, whereby all the light is reflected and none refracted. 0c = sin "1 (n2/nl) where nl > n2

96

(2)

Planar Waveguides

Figure 1. The basic experimental arrangement of a system based on the principle of reflectance.

Under the condition of total internal reflection, a standing wave of the electromagnetic field is set up at the point of reflection. This standing wave is known as the evanescent wave and penetrates into the lower refractive index medium with an exponential decay. The depth at which the intensity of either the electric or magnetic component of the electromagnetic wave drops to 1/e of its original value is known as the depth of penetration and is described by Equation 3; where n21 = nz / nl. dp = (~'vac / 2nnl) [1 / (sinZO - n212)]1/2

(3)

As shown in Equation 3, dp is dependent on the incident angle (0), the refractive indices of the dielectrics (n~ and n2), and the wavelength @) of the light. Typical penetration depths range between 100 and 300 nm. It is this evanescent wave, generated at the interface between the two dielectrics, that interacts with the surface species immobilized at the interface. The intensity of the reflected light will decrease after each reflection due to the presence of absorbing material at the surface of the waveguide. This decrease in the intensity of the reflected light can be measured using the detection system described previously in Figure 1 and is typically used in interferometric techniques. When the evanescent light excites a fluorophore on the planar waveguide, the resulting fluorescence emission can be measured either as shown 97

Sapsford, Rowe Taitt, and Ligler

Figure 2. Detection of emitted fluorescence at right angles to the waveguide interface.

in Figure 1 or, as is more common, at right angles to the waveguide interface, as shown in Figure 2. Coherent light in the form of lasers is typically used as the excitation source in TIRF studies. The exact choice of the laser is dependent upon the fluorescent label used; the most commonly used lasers are the argon-ion (488 nm) laser for fluorescein and a helium-neon (633 nm) or diode laser (635 nm) for the cyanine dye Cy5. A number of devices have been used in the detection of the resulting fluorescence emission, in particular charge coupled device (CCD) cameras (Silzel et al., 1998; Plowman et al., 1999; Feldstein et al., 1999), photomultiplier tubes (PMT) (Schult et al., 1999), photodiodes (Brecht et al., 1998) and a single photomultiplier tube (Lundgren et al., 2001; Schuderer et al., 2000). There has been much research into improving the optics and sensitivity of the TIRF instrumentation represented in Figure 2. Most of the final systems described consist of a number of similar components such as the light source and detector and also a variety of focusing lenses to improve detector response (Duveneck et al., 1995; Herron et al., 1996, 1997; Golden, 1998; Feldstein et al., 1999). Golden (1998) used a 2-dimensional graded index lens (GRIN) to focus the fluorescence from the planar waveguide onto a CCD. The GRIN lens provided a shorter working distance than a standard lens with a concomitant decrease in overall instrument size. The introduction of bandpass and longpass filters was found to improve the rejection of scattered laser light and hence reduce the background of the system (Feldstein et al., 1999). 98

Planar Waveguides Unfortunately a side effect of using bulk waveguides and collimated light is the production of sensing "hot spots" along the planar surface which occur where the light beam is reflected, illuminating only discrete regions. These hot spots have been successfully utilized as sensing regions by Brecht and coworkers in the development of an immunofluorescence sensor for water analysis (Brecht et al., 1998; Klotz et al., 1998). Feldstein et al. (1999, 2000) overcame this problem by incorporating a line generator and a cylindrical lens to focus the beam into a multi-mode waveguide which included a propagation and distribution region prior to the sensing surface. This resulted in uniform lateral and longitudinal excitation at the sensing region. Another method of achieving uniform longitudinal excitation of the sensing region is to decrease the waveguide thickness (Herron et al., 1996). When the thickness of the waveguide is much greater than the wavelength of the reflected light, the waveguide is referred to as an internal reflection element (IRE). However, if the thickness of the waveguide is decreased such that it approaches the wavelength of the incident light, the pathlength between the points of total internal reflection become increasingly shorter. At the thickness where the standing waves, created at each point of reflection, overlap and interfere with one another, a continuous streak of light appears across the waveguide, and the IRE becomes known as an integrated optical waveguide (IOW) (Plowman et al., 1998). Integrated optical waveguides, used frequently in TIRF studies, are mono-mode and prepared by depositing a thin film of high refractive index material onto the surface of a glass substrate. These thin films are typically 80 - 160 nm in thickness and consist of inorganic metal oxide compounds such as tin oxide (Duveneck et al., 1995), indium tin oxide (Asanov et al., 1998), silicon oxynitride (Plowman et al., 1999) and tantalium pentoxide (Duveneck et al., 1997; Pawlak et al., !998). The light is coupled into these IOWs via a prism or grating arrangement. Studies by Brecht and coworkers compared IRE- and IOW-based waveguides and concluded that the integrated optics significantly improved the sensitivity of the system by a factor of 100 (Brecht et al., 1998).

99

Sapsford, Rowe Taitt, and Ligler

Figure 3. Schematic representation of the three main categories of solid-phase immunoassays. A) Direct non-competitive assay between the immobilized antibody and the antigen in solution. B) Competitive assay between the fluorescently labeled (known concentration) and unlabeled (unknown concentration) antigen for binding sites on the immobilized antibody. C) Sandwich assay, in which the amount of immobilized antigen is determined by passing a second fluorescently labeled antibody over the surface.

1.2. The molecular recognition element The different types of binding events that are typically monitored via fluorescence measurements include antibody-antigen interactions, nucleic acid hybridization (DNA/RNA) and receptor-ligand binding. These are often utilized in affinity-based biosensors (Rogers, 2000). Although a number of these biomolecules contain some intrinsic fluorescence in the form of amino acid residues or cofactors, often this fluorescence may have poor intensity and low quantum yields. Therefore, extrinsic fluorescent probes are incorporated into one of the binding partners. The introduction of an extrinsic fluorescence probe, such as rhodamine, coumarin, cyanine, or fluorescein dyes, allows both site and spectral selection. Assays were originally performed using radiolabeled species; however, fluorescent markers are now the more common label of choice due to safety, longer shelf lives, lower costs and ease of preparation and disposal. The use of fluorescent markers is also favored over enzyme labels due to increased stability and because no additional substrates are required.

1.2.1. Immunoassays. To date, antibody-antigen binding interactions are the most well characterized systems employed in sensors based on TIRF. The assays carried out using antibody-antigen systems can be divided into three main categories: direct, competitive and sandwich immunoassays. As can be seen from Figure 3, the direct assay (Figure 3A) is the simplest method to perform; however, it requires that the antigen contain some form of intrinsic fluorescence that can be detected. In the absence of a fluorescent antigen, preferred formats are competitive and sandwich assays (Figure 3B, 3C). In the competitive assay 100

Planar Waveguides (Figure 3B), a fluorescent-labeled antigen competes with a non-labeled antigen for binding sites on the immobilized antibody (Brecht et al., 1998; Schult et al., 1999). Hence, the resulting fluorescence signal is inversely proportional to the unlabeled antigen concentration. Competitive formats are especially useful in the detection of molecules not large enough to possess two distinct epitopes (e.g., haptens). Sandwich assays (Figure 3C), on the other hand, require relatively large antigens; the antigen is bound to the immobilized capture antibody at one epitope and is detected by a fluorescent-labeled tracer antibody bound to a different epitope (Silzel et al., 1998; Rowe et al., 1999; Schult et al., 1999; Plowman et al., 1999). Sandwich and direct assays produce a fluorescence signal that is directly proportional to the amount of bound antigen. Another type of assay not represented in Figure 3 is that of displacement: here immobilized antibodies are first saturated with a fluorescently labeled antigen. Upon introduction of the unlabeled antigen, displacement of the labeled antigen occurs and is then measured (Rabbany et al., 1994). However, displacement assays have not as yet been demonstrated in planar waveguide TIRE formats. 1.2.2. DNA and mRNA analysis. The hybridization between complementary DNA or mRNA strands has become increasingly studied using planar waveguide TIRF (Duveneck et al., 1997; Budach et al., 1999; Schuderer et al., 2000). Duveneck et al. (1997) coupled 16-mer oligonucleotides to (3glycidyloxypropyl) trimethoxysilane (GOPTS)-functionalized tantalum pentoxide waveguide chips using a hexaethylenglycol linker. The specific binding of Cy5-1abeled complementary oligonucleotide was then detected using TIRF and the surface regenerated using 50% urea. The system demonstrated good stability and excellent signal reproducibility during repeated cycles of binding and regeneration. Budach and coworkers (1999) immobilized two different 5'-amino oligonucleotides in a checkerboard pattern on tantalium pentoxide waveguides functionalized with GOPTS using ink-jet printing. One of the Cy5-1abeled complementary oligonucleotides was then passed over the waveguide surface and the fluorescence increase of the spots monitored. There was no measurable cross-reactivity between the non-complementary oligonucleotides with detection limits of 50 fM achieved for each analyte. The surface was regenerated numerous times, using urea, without any apparent loss in activity. Schuderer et al. (2000) opted to immobilize biotinylated 18-mer oligonucleotides onto glass surfaces functionalized with adsorbed avidin using a flow cell assembly; this would allow the immobilization of different capture biomolecules to the same waveguide should it be required. The fluoresceinlabeled complementary oligonucleotide was then passed over the surface and the fluorescence increase in each channel monitored using a single photomultiplier, which was scanned over the surface. Detection limits ranged between 3 and 10 fmol with surface regeneration achieved using sodium hydroxide. The main disadvantage to this system was the need to scan the detector over the surface, making it unreliable for the determination of fast binding kinetics. DNA

101

Sapsford, Rowe Taitt, and Ligler hybridization has numerous biosensing applications in areas such as the detection of polymorphisms or mutations in gene sequences, as markers for disease, and in drug development (Debouck and Goodfellow, 1999).

1.2.3. Membrane receptor-ligand interactions. Ligand binding to receptors other than antibodies and DNA has been much less studied in the literature. Currently, only a limited number of studies describing receptor-ligand binding using planar waveguide TIRF have been published. Schmid et al. (1997, 1998) immobilized the green fluorescent protein and serotonin receptor, Rowe-Taitt et al. (2000a) studied gangliosides as receptors and Pawlak et al. (1998) investigated the membrane enzyme Na,K-ATPase in membrane fragments. One major problem with studying receptor-ligand binding has been the immobilization of the receptor protein which is often membrane-bound in its natural environment; hence immobilization onto a sensor surface should mimic this environment to ensure that the receptor remains active (Pawlak et al, 1998). Vogel's group successfully immobilized fully functional green fluorescent protein (GFP) and serotonin receptors onto planar waveguides in defined orientations (Schmid et al., 1997, 1998). GFP and serotonin were genetically modified to contain poly-histidine tags at the C- and N-terminus respectively. These tags bound specifically to waveguides coated with thiosilane and nitrilotriacetic acid (NTA)-bound metal ions; binding of the his-tagged protein was reversible. Activity of the immobilized serotonin receptor was demonstrated through binding of a fluorescent ligand and competitive binding studies (Schmid et al., 1998). Pawlak et al. (1998) created multi-layers on top of the planar waveguide. First, a long-chain alkyl phosphate monolayer was used to make the metal oxide surface hydrophobic. Lipid vesicles were then adsorbed to the monolayer to form a lipid film. Membrane fragments containing the ATPase were finally adsorbed onto the lipid biolayer for use in a fluorescence quenching assay. Rowe-Taitt et al. (2000) immobilized gangliosides as receptors for toxins by immobilizing a long-chain alkyl silane to create a hydrophobic surface for adsorption of the ganglioside; presumably the fatty acid tail of the ganglioside intercalated into the silane film; the resulting receptor immobilization was quite stable to rinsing, dehydration and rehydration. Both direct binding and sandwich assays using different gangliosides patterned on the waveguide were demonstrated. When successful, receptor-ligand binding studies offer applications in the pharmaceutical industry for drug development and for investigating membrane processes. 1.3. Immobilization of the biomolecule to the transducer

The various methods in which the biological component of a biosensing system can be immobilized onto the surface of the transducer include physical adsorption, covalent immobilization, and entrapment in polymer matrices (Hall, 1990). One important requisite of all immobilization techniques is that the integrity of the biomolecule be preserved and that the active site remain 102

Planar Waveguides accessible to the binding partner. Each immobilization procedure has various advantages and disadvantages; however, physical adsorption and covalent binding to functionalized surfaces are the most commonly used in TIRF measurements. Physical adsorption of a biomolecule to a surface occurs via dipole-dipole interactions, van der Waals forces or hydrogen bonding, depending on the nature of the substrate surface and the adsorbate. However, physical adsorption in general is not only strongly influenced by changes in the ambient conditions, such as pH and the solvent used, but may also be a reversible process. Furthermore, adsorption may not provide as high density of immobilized biomolecules as covalent immobilization (Ulbrich et al., 1991). Physical adsorption is generally unspecific, random and multi-orientated in nature, often resulting in the inaccessibility of the active binding site. Covalent immobilization provides surfaces with reproducibly attached biomolecules at relatively high densities (2 ng/mm2; Bhatia et al., 1989). Most methods of covalent immobilization involve the activation of the surface (e.g., silane or thiol self-assembled monolayers), followed by covalent linkage of the biomolecule either directly or using a crosslinker. Self-assembled monolayers (SAMs) are extensively used in the construction of artificial biomolecular surfaces due to simplicity of preparation and the adaptability of the resulting surface chemistry. The surface of the SAM can be engineered by incorporating different functionalities into the tail group, such that hydrophobic or hydrophilic surfaces which can undergo further chemical reactions are produced. The use of SAMs in the development of biosensors and biosurfaces has been reviewed by a number of groups (Wink et al., 1997; Mrksich et al., 2000; Ferretti et al., 2000). Although the majority of research has been concerned with the immobilization of biomolecules onto noble metal surfaces, an increasing number of groups are interested in the use of glass substrates for immobilization (Bhatia et al., 1989; Brizzolara et al., 1994; Lee and Saavedra, 1996; Sojka et al., 1999; Pirrung et al.,

2ooo). There are a number of different planar surfaces used in the immobilization of biomolecules for study with TIRF. These include simple bulk waveguides such as glass, silica and polystyrene and the slightly more complicated IOW waveguides such as tantalium pentoxide (Ta205). There are likewise a variety of different surface chemistries used to modify these waveguides such that immobilization of the biomolecule can be facilitated. A number of examples of these systems, which were characterized using TIRF, are given in Table 1. It can be seen from Table 1 that silanization of the waveguide, whether it be the bulk glass or an integrated optical waveguide, is a popular method of functionalising the surface for further chemistry. Also the avidin-biotin interaction is extensively used in the immobilization of biotinylated molecular recognition elements. This

103

Sapsford, Rowe Taitt, and Ligler Table 1. Various planar waveguides and immobilization chemistries for biomolecules as studied by TIRF i

Waveguide Bulk silica (antibody/antigen)

100 nm tantalium pentoxide (Ta2Os) on glass (DNA) 350 nm indium tin oxide (ITO) on quartz (antibody/antigen) 150 nm Ta205 on glass (receptor/ligand)

Surface Chemistry (1) Adsorbed avidin/bi0tinylated antibodies (2) 10 nm polymethacrylate/couple to the antibodies via reactive thiol groups Glycidyloxypropyltrimethoxy silane/ hexaethyleneglycol linker/3' end of capture oligonucleotides

3-(N-(2-aminoethyl)-aminopropyl) trimethoxysilane/sulfo-NHS-LCbiotin Mono-C16 alkyl phosphate monolayers/spread liquid vesicle (1palmitoyl-2-oleoyl-sn-glycero-3phosphocholine)/Na, K-ATPase membrane disks labeled with fluorescein (3-mercaptopropyl) trimethoxysilane/ Glass slides NTA-maleimide/loading Ni +2 (receptor/ligand) solution/His-tagged receptor NeutrAvidin adsorbed directly/ Polystyrene film biotinylated antibodies (antibody/antigen) Polymethylmethacrylate (1) Sandwich assay; adsorbed NeutrAvidin/biotinylated antibodies, sensor chips (2) Competitive assay; Adsorbed (antibody/antigen) antibodies Dimethyldiclorosilane/antibody Silicon oxynitride on adsorption glass (antibody/antigen) (3-Mercaptopropyl) trimethoxysilane/ Glass slides (antibody/antigen) N-(~,-maleimidobutyryloxy) succinimide ester / NeutrAvidin/biotin labeled antibodies Adsorbed avidin/biotin labeled Glass slides capture oligonucleotides (DNA) Octadecyltrichlorosilane/ganglioside Glass slides adsorption (receptor/ligand)

104

Reference Herron et al., 1993 and 1996 Duveneck et al., 1997

Asanov et al., 1998 Pawlak et al., 1998

Schmidt et al., 1998 Silzel et al., 1998 Schult et al., 1999

Plowman et al., 1999 Rowe et al., 1999a

Schuderer et al., 2000 Rowe-Taitt et al., 2000a

Planar Waveguides non-covalent protein-ligand interaction is commonly used in the production of multilayers, often involving the use of both covalent and non-covalent interactions (Rowe et al., 1999a; Birkert et al., 2000).

2. History The evanescent wave excitation of a surface-bound fluorophore has been studied for a number of years in the form of fiber optic technology. A number of the researchers currently involved in developing planar waveguide TIRF previously carried out much of their initial research in the field of fiber optics, as summarized in Table 2. There are a number of advantages to the use of a planar waveguide including the relative ease of preparation and integration into fluidic systems. Early researchers in the field, and also some more recent investigators, immobilized capture biomolecules uniformly over the planar surface and monitored the fluorescence intensity either as a function of time or the concentration of the labeled binding partner (Herron et al., 1993; Duveneck et al., 1997; Brecht et al., 1998; Pawlak et al., 1998; Schult et al., 1999). However, the most important advantage of using a planar waveguide is the possibility of creating patterns of immobilized biomolecules leading to multiple, parallel assays on a single waveguide. A number of techniques have been used in the creation of patterned biomolecular assemblies on planar surfaces as reviewed by Blawas and Reichert (1998a). In terms of fluorescence studies, the production of these patterned surfaces has been investigatedusing either the fluorescence microscope or TIRF instrumentation. The patterns can be created using photolithography, or by depositing the recognition molecules in physically separate locations on the waveguide. A method for the photolithographical patterning of proteins on surfaces was described by Bhatia et al. (1992, 1993). Ultraviolet light was used to pattern (3mercaptopropyl) trimethoxysilane on a glass surface. Exposed regions of the surface became protein-resistant through the conversion of the thiol group to a sulfonate species, while the masked areas went on to bind the biomolecule. This proved to be a convenient method of creating high resolution patterns (less than 10 nm in width) of immobilized capture biomolecules. Unfortunately this method had the disadvantage that only a single biomolecule could be patterned. Photopatterning of the surface has also been utilized by Schwarz and coworkers. Here photoablation of polymer substrates was used to produce avidin patterns on the exposed substrate (Schwarz et al., 1998). Likewise Wadkins et al. (1997) used glass slides coated with a photoactivated optical adhesive and a mask to create wells in the gel layer upon light exposure. A different biomolecule could then be covalently attached to the exposed glass in each well prior to removal of the polymer. A similar method was adopted by Guschin et al. (1997) and 105

Sapsford, Rowe Taitt, and Ligler Table 2. Groups involved in the development of TIRF-based biosensors initially with fiber optic and later with planar waveguides Leaders

"Fiber Optic R'efe'rences

Herron, Christensen, Reichert (USA)

Lin et al., 1988 Yoshida et al., 1988

Duveneck, Ehrat, Neuschafer (Switzerland)

Abel et al., 1996 Duveneck et al., 1996

Duveneck et al., 1997 Pawlak et al., 1998 Budach et al., 1999

Ligler (USA)

Thompson and Ligler, 1988 Golden et al., 1992 King et al., 1999

Ligler et al., 1998 Wadkins et al., 1998 Rowe-Taitt et al., 2000b

Bilitewski (Germany)

Beir et al., 1992

Schuderer et al., 2000

-

Planar Waveguide ...... References Herron et al., 1993 Herron et al., 1996 Plowman et al., 1999

Arenkov et al. (2000) for the immobilization of oligonucleotides; however, the biomolecules were immobilized to the gel pads rather than the glass. B lawas and coworkers (1998b) used the caged-biotin-bovine serum albumin (BSA) compound, methyl cz-nitropiperonyloxy-carbonyl-biotin-BSA, to pattern glass surfaces. First, caged-biotin-BSA was adsorbed onto the glass slide: second, the slide was exposed to 353 nm light through a mask, which effectively removed the cage surrounding the biotin in unmasked areas. Third, streptavidin selectively bound to the irradiated regions of the surface. Finally, biotinylated capture antibody was bound to the streptavidin. The disadvantage of this method is that there was some cross-contamination of the first immobilized streptavidin with the biotinylated antibody intended for immobilization to a streptavidin bound in a second or third repetition of the process. Two additional methods of photolithographically patterning proteins on planar waveguides use polyethylene glycol (PEG) moieties to prevent non-specific protein adsorption. In the method developed by Conrad et al. (1997, 1998), the photochemically active silane, o-nitrobenzyl polyethylene glycol trichlorosilane, was attached to a glass waveguide. Photo-oxidation of the PEG-terminated silane through a mask cleaved the carbamate to yield an amino terminal PEG and a surface bound o-nitrosobenzaldehyde. The antibody was reacted with the onitrosobenzaldehyde in a Schiff's base reaction forming a stable amide linkage. The process was repeated for the addition of different antibodies to additional 106

Planar Waveguides spots on the waveguide. Similarly, Liu et al. (2000) tethered a benzophenone photophore through a PEG spacer to a maleimide group. After exposure to light, antibodies were immobilized and an assay conducted on the polystyrene waveguides. Spacer lengths of five ethylene glycol groups proved optimal for maximizing the signal-to-background ratio. The use of ink jet printing is another popular choice for the production of patterned biomolecular surfaces. Silzel et al. (1998) ink jet printed either the capture antibodies or the protein avidin in 200 ktm diameter zones on the surface of polystyrene films. Biotinylated antibodies were later immobilized on the avidin spots. A checkerboard pattern of two different oligonucleotides was produced by Budach and coworkers (1999) using the ink jet printing of capture biomolecules onto a Ta205 waveguide using (3-glycidoxypropyl) trimethoxysilane. Physically isolated patterning using flow cells constructed from a variety of materials, including polydimethylsiloxane (PDMS) (Feldstein et al., 1999), a rubber gasket (Plowman et al., 1999), a Teflon block fitted with O-rings (Schuderer et al., 2000) and a microfluidics network made of silicon (Bernard et al., 2001), have been used. Typically the flow cell, containing a number of channels, was temporarily attached to the surface of the planar waveguide and each channel filled with a solution of the capture biomolecule, as shown in Figure 4A. In this example (Feldstein et al., 1999), the resulting waveguide was patterned with stripes of immobilized biomolecules (Figure 4B). The sample and fluorescently labeled antibody were then passed over the surface using a second flow cell orientated perpendicular to the immobilized capture biomolecule channels, as shown in Figure 4B. The beauty of the physically isolated patterning technique is its ability to immobilize a number of different capture biomolecules, i.e. one type in each flow cell channel, onto a single surface, creating an array of recognition sites with no possibility of cross contamination. The disadvantage of the approach is that it may be difficult to scale it up for manufacture of large numbers of waveguides. Yet it is easy to use for investigations with no special photolithographic or arraying equipment required.

3. State of the Art

As described in the previous sections, much of the initial investigation into the use of planar waveguides in TIRF biosensors has centered on both instrumentation development and reproducible immobilization of the capture biomolecules. However, once these systems have been optimized, the question of application becomes the driving force behind further development.

107

Sapsford, Rowe Taitt, and Ligler

Figure 4. The patterning of capture biomolecules using flow cells (adapted from Feldstein et al., 1999). A) A multichannel flow cell is pressed onto the planar waveguide and each channel filled with a solution of the capture biomolecule. B) Sample and fluorescent tracer antibody are passed over the waveguide surface perpendicular to the immobilized capture biomolecule channels using a second flow cell.

Due to the vast number of biological systems which respond to a variety of analytes, the number of potential applications for biosensors in general is huge including medical diagnostics and healthcare, environmental monitoring, process monitoring in the chemical, food and beverage industries, and military defense. With respect to papers published in the field of planar waveguide TIRF biosensors, environmental monitoring, clinical diagnostics and military defense seem to be the most popular. Conventional methods for the measurement of analytes are often slow, multi-step, complex, and require skilled technicians and specialized laboratory equipment. The driving force behind the development of biosensors is the possibility of quick, cost effective, user friendly, field analytical technologies that have sensitivity and specificity comparable to the laboratory measurements. A number of original studies, along with some more recent ones, have investigated the use of a single capture biomolecule-analyte assay. Such systems include DNA hybridization assays (Duveneck et al., 1997; Schuderer et al., 2000) and antibody-antigen studies for the human pregnancy hormone human chorionic gonadotropin (Herron et al., 1993; Schult et al., 1999), the pesticide pollutant atrazine (Brecht et al., 1998), and the asthma drug theophylline (Schult et al., 1999). The majority of these systems were found to have comparable or slightly better detection limits than the conventional methods used to measure the analyte, such as ELISA. However, as previously stated, one of the major advantages of using a planar substrate is the ability to create arrays of different capture biomolecules for multi-analyte sensing. It is the development of this microarray, multi-analyte based technology that will give the resulting biosensor the edge over a number of 108

Planar Waveguides Table 4. Summaryof research groups currently involved in the patterning of multiple capture biomolecules onto planar waveguides for study with TIRF i

i.

"Singleor MultiAnalyte Detection Multi

Limit of Detection and Assay Time 15 ng/mL 2h

References

Ricin, Y. pestis F1, staphylococcal enterotoxin B, ovalbumin, mouse and human IgG, D-dimer, B. globigii, MS2 bacteriophage, cholera toxin, botulinum toxoids A and B, B. anthracis, F. tularensis, B. abortus

Single and Multi

Typically ng/mL range: see various papers.

Wadkins et al., 1998; Ligler et al., 1998; Golden et al., 1999; Feldstein et al., 1999; Rowe et al., 1999a, b; Rowe-Taitt et al., 2000a, b, c

Budach et al.

16-Mer and 22-mer oligonucleotides

Single

-- pM 12 min

B udach et al., 1999

Plowman et al.

Various IgG Creatin kinase MB Cardiac troponin I Myoglobin

Single and Multi

-- ng / mL 5-20 min

Plowman et al., 1999

Zeller et al.

Mouse and rabbit IgG

Single

Not reported 60 min / analyte

Zeller et al., 2000

....Research ' Analytes Measured Group

i

Silzel et al. Ligler et al.

' Four different'human IgG subclasses

i

iiii

i iiiii

<14 min

Silzel et al., 1998

current laboratory-based measurements. There are to date at least five research groups involved in the immobilization of patterns of multiple capture biomolecules onto planar waveguides, although only three of these groups have demonstrated multi-analyte measurements, summarized in Table 4. Zeller and coworkers (2000) have developed quite a unique TIRF system in which the planar waveguide consists of multiple, single pad, sensing units. Each of these single pads has its own laser light input, background suppression and coupling of the fluorescence emission to the detector. The authors demonstrated 109

Sapsford, Rowe Taitt, and Ligler a two-pad sensing device in which one pad was modified with mouse IgG and the other with rabbit IgG. However, only one Cy5-1abeled antibody directed against each immobilized antigen was assayed at a time. It was suggested that the laser light could be split into spatially different parts in multianalyte measurements, using multiple single sensing pads. Such a process would probably involve a number of optical components, and therefore the robustness of the system for use outside the laboratory is still in question. The performance of single and multiple analyte assays was compared by Plowman et al. (1999) using IgGs of different species and antibodies specific for cardiac proteins (creatin kinase MB, cardiac troponin I and myoglobin). Studies also investigated the effects of using polyclonal versus the more specific monoclonal antibodies during assays. Results suggest that polyclonal antibodies are more prone to cross-reactivity in the multi-analyte assay format; therefore, monoclonal antibodies were the capture biomolecule of choice when available. Also, the single analyte assay format was found to have lower detection limits than its multi-analyte assay counterpart. However, both sets of detection limits reported were in the clinically significant ranges set for the three cardiac proteins. Interestingly, however, a study by Rowe et al. (1999b) demonstrated the use of mixtures of polyclonal antibodies of different species with no cross-species interactions observed; furthermore, the same sensitivity was achieved with mixtures of the fluorescent tracer antibodies as with single antibody assays. The group including Ligler, Golden and coworkers have produced the greatest number of publications in the field of planar waveguide TIRF, in particular with respect to multi-analyte detection. The group f'trst published papers on TIR.F studies in 1998 for the detection of three analytes ricin, Yersinia pestis F1 and staphylococcal enterotoxin B (SEB) (Wadkins et al., 1998), using the instrumentation shown in Figure 5. The long-term aim was the development of a fully automated instrument geared towards portability and low cost. In this particular study, the antigens and the Cy5 tracer antibodies were added sequentially and the slide imaged. Later that year, the use of a PDMS flow cell for the patterning of capture biomolecules was developed and the immunoassay, run prior to imaging, carried out using a fixed polymethylmethacrylate (PMMA) flow cell aligned perpendicular to the patterned antibody channels (Ligler et al., 1998). Not only was simultaneous detection of analytes demonstrated but also the detection of Y. pestis F1 in clinical fluids such as whole blood, plasma, urine, saliva and nasal secretions. This study was further extended in 1999 to the measurement of SEB and D-dimer in clinical fluids, all with detection limits suitable for clinical analysis requirements (Rowe et al., 1999a). The impact of potential environmental interferents has also been addressed (Rowe-Taitt et al., 2000c). Analyte samples of Bacillus anthracis, Francisella tularensis LVS, Brucella abortus, SEB, cholera toxin and ricin were assayed in the presence and absence of interferents such as sand, clay, pollen, and smoke 110

Planar Waveguides

Figure 5. Basic instrumentation which makes up the array biosensors developed by Ligler, Golden, and co-workers at the NRL (adapted from Wadkins et al., 1998).

extracts, and results were compared. No false positive or false negative responses were caused by the potential interferents; however, in some cases the signal amplitude was affected. In order to further develop the array technology, an automated image analysis program was developed, and some of the optical and fluidic components were miniaturized (Feldstein et al., 1999). To test the ability of the array biosensor to measure three diverse classes of analytes, assays were carried out using bacterial, viral and protein analytes (Rowe et al., 1999b). Single or multi-analyte samples were run though the assay channels followed by either individual Cy5-1abeled tracer antibodies or a mixture of polyclonal tracer antibodies. The array biosensor was used in the study of 126 blind samples and the automated image analysis proved reliable in the discrimination of fluorescent signals, with detection limits in the mid ng/mL range, similar to ELISA results. This approach using mixtures of tracer antibodies was later extended to monitor the six biohazardous analytes B. anthracis, F. tularensis, B. abortus, botulinum toxoids, cholera toxin and ricin, demonstrating simultaneous analysis of six samples for six analytes in 12 min (Rowe-Taitt et al., 2000b, c). Recent studies have concentrated on realizing the goal of a fully automated array biosensor system and have involved the combination of an automated fluidics system with an automated image analysis program (Feldstein et al., 1999, 2000; Rowe-Taitt et al., 2000b). The further investigation of various methods with which to produce reliable methods of attaching the flow cell manifold to the glass sensing surface in the automated array resulted in two separate designs, either of which could be used depending on the application required (Leatzow et al., 2001). Also to decrease the size and improve the robustness of the fluidics system for use in the field, a fluidics cube was designed for automated handling. 111

Sapsford, Rowe Taitt, and Ligler The cube was made from a thermoplastic and contained the reservoirs and channels required for both the sample and the reagents (Dodson et al., 2001). Currently the array biosensor can fit inside a box of dimensions 31 x 23 x 30 cm; this includes the excitation source, sample, optics, CCD camera, and the fluidics system and pumps. The standard PC has also been replaced with a single board Pentium computer and keyboard, with a LCD screen placed in the lid of the box (Ligler et al., 2001). The use of planar waveguide TIRF in the detection of multiple analytes has clearly been demonstrated as well as the ability to miniaturize the instrumentation. The technologies developed in the laboratories of Herron, Ehrat and Ligler are all under commercial development by different small businesses.

4. Advantages and Limitations As with any other measurement process, there are always associated advantages and limitations (for a review, see Schobel et al., 2000). The introduction of a fluorescent probe to a biomolecule has the advantage of allowing both site and spectral selection. Fluorescent labels have longer shelf lives, lower costs and greater safety than radiolabels. The use of a fluorescent probe is also favored over enzyme labels typically used in ELISA, for stability reasons and because additional substrates are not required. As mentioned in previous sections, limits of detection for TIRF are often equivalent to the benchmark standard ELISAs, and yet the TIRF immunoassay can be carried out in a much quicker time, e.g., 10 minutes (TIRF) versus 2 h (ELISA). The only major disadvantage of TIRF systems is the requirement for a fluorescent assay component (analyte or "tracer" molecule), which generally involves labeling and should ideally be selective such that the label does not interfere with the binding interactions of the biomolecule. Surface plasmon resonance (SPR) is a technique commonly used in the study of binding events and does not require any labeling of the sample. However, it is susceptible to problems such as low sensitivity and increased backgrounds due to non-specific adsorption. One of the major limitations of SPR, set by the signal-to-noise ratio of the detection system, is the measurement of low molecular weight species, typically less than 5000 Da (Garland, 1995), although this limit of detection is improving. Often, the larger of the two binding partners is required to become the analyte with the smaller component immobilized as the "capture" element. This can limit experimental design. Hence, TIRF is far superior to SPR in this respect. Due to the nature of its measurement (refractive index change), SPR is also highly sensitive to ambient temperature variations and is more susceptible to bulk matrix effects than TIRF.

112

Planar Waveguides One inherent advantage of using evanescent wave fluorescence is its surface -~ specific nature. Only fluorophores within the field of the evanescent wave are excited to give fluorescence emission. This renders the technique relatively immune to matrix effects such as fluorophores in the bulk, particulates, and turbidity so binding of the fluorophore can be measured in real time (Sapsford et al., 2001). The use of the planar waveguide has allowed the patterning of multiple capture biomolecules to a single surface and hence has lead to the possibility of simultaneous assays for multiple analytes. The number of analytes that can be identified simultaneously is limited only by the reproducibility of the method for making the patterns, the area of the waveguide which can be evenly illuminated and imaged, and the affinity and specificity of available recognition molecules.

5. Potential for Expanding Current Capabilities Planar waveguide TIRF is a sensitive technique with limits of detection typically in the mid-to-low picomolar region (Duveneck et al., 1997; Plowman et al., 1999; Rowe et al., 1999a). All reported data were generally found to be comparable with standard measurement techniques, such as ELISA, or within clinically or environmentally set limits. A number of changes have been made to various parts of the system in order to improve the sensitivity. Ligler and co-workers, for example, switched from a room temperature CCD camera to a thermoelectrically cooled version and observed a decrease in the limit of detection from 5 ng/mL (Wadkins et al., 1998) to 1 ng/mL (Rowe et al., 1999a). There are a variety of techniques for immobilization of biomolecules to the surfaces of planar waveguides. Most of those mentioned earlier do not control the orientation of the molecule on the surface. Biotinylation and attachment through an immobilized avidin does permit careful control of the number of sites though which the biomolecule is attached. While this helps maintain the mobility of a receptor molecule and minimizes steric hinderance, it does not control the location of the attachment site on the molecule. Non-covalent attachment of antibodies via binding by Protein A or Protein G or covalent attachment through the carbohydrate side chain limit the attachment site to the Fc region of an antibody; thus there is no attachment in the binding site. Fab' antibody fragments have been linked via the thiol group at the opposite end of the molecule from the antigen binding site (Lu et al., 1995; Huang et al., 1996). The methods for preparing antibodies to be immobilized through either the carbohydrate or the thiol groups result in a significant loss of antibody during the processing procedure but generally produce highly functional surfaces. An additional method for controlling the attachment site in a biomolecule is by genetically engineering unique attachment sites on the biomolecule's surface, a technique known as site directed mutagensis (McLean et al., 1993; Vigmond et 113

Sapsford, Rowe Taitt, and Ligler al., 1994). However, this is generally a complicated multi-step procedure not altogether favorable from a commercial viewpoint. Another process which may reduce the sensitivity of TIRF measurements is nonspecific binding of the fluorescent species to non-sensing regions of the surface, increasing the background signal. Currently researchers commonly use either BSA and/or polyoxyethylene-sorbitan monolaurate (Tween) to block the surface and minimize nonspecific adsorption after immobilization of the capture biomolecule. Another possibility is the introduction of monolayers resistant to nonspecific adsorption of proteins in the non-sensing regions of the planar waveguide. Monolayers consisting of oligoethylene oxide terminal groups have found to be highly protein resistant in SPR (Silin et al., 1997; Lahiri et al., 1999) and TIRF studies (Conrad et al., 1997). This approach, however, may not prevent nonspecific adsorption to the sensing regions of the planar waveguide (Conrad et al., 1997). The production of increasingly smaller sensing spots on the planar surface not only results in increased multi-analyte capabilities but also, according to Ekins, should increase the sensitivity of the system (Ekins, 1995, 1998). Ekins suggested that under static or equilibrium conditions diffusion rates play an important role in the determination of the association rate. It is thought that the smaller the spot area, the lower the diffusion constraints become on the rates of association, and hence higher signal-to-noise ratios are attained in shorter time periods. However, in a number of TIRF studies using planar waveguides, flow cells and flow conditions are used which minimize diffusion limitations. Sensitivity therefore becomes a function of the method of detection, rather than an issue of the spot size (Sapsford et al., 2001). The power of the larger scale arrays has been demonstrated using both DNA chips (Brown and Botstein, 1999; Lockhart and Winzeler, 2000; Gullans, 2000; Knight, 2001) and antibody chips (MacBeath and Schreiber, 2000; Haab et al., 2001). These chips are currently being scanned using confocal microscopy. However, a recent study by Duveneck and coworkers (2001) comparing confocal microscopy and TIRF methods demonstrated that TIRF offered a more sensitive measurement than current scanning confocal microscopy. TIRF biosensors such as that of RoweTaitt et al. (2000a) are also sufficiently sensitive to measure such assays (unpublished data), but the methods used to deposit the spots in high density arrays need to be improved to attain reproducible surface concentrations of the capture biomolecules. Integrated optical waveguides (IOW) are also found to increase the sensitivity of the system. These waveguides are very promising for clinical and drug screening applications (Duveneck et al., 2001). However, as pointed out by Feldstein et al. (1999), the increased constraints and requirements of the optical Components would reduce the robustness of the system from a commercial point of view, should the device be intended for field applications (Brecht et al., 1998). 114

'

Planar Waveguides Asanov and coworkers (1998) produced an IOW consisting of an indium-tin oxide layer (a transparent conductor) and used it to investigate biotin-streptavidin binding. They found that electrochemical polarization of the waveguide dissociated the biotin-streptavidin complex effectively regenerating the surface. Regeneration of the surface could be useful for commercial applications as an altemative to disposable substrates, depending on the relative cost of the additional system components and the number of times regeneration can be carried out without loss of surface integrity. As mentioned previously, successful regeneration of DNA chips has been carried using urea, without apparent loss of activity (Duveneck et al., 1997; Budach et al., 1999). Herron, Christensen and their colleagues (1997) developed a molded plastic device that included both the planar waveguide and the lens for distributing the light across the waveguide as a single piece. The array biosensor fluidics system and flow manifold described by Dodson et al. (2001) and Leatzow et al. (2001), respectively, could also be molded in a single plastic piece. Such plastic parts could be integrated along with a molded lens array for focusing the fluorescence onto the image camera and freeze dried fluorescence reagent into a disposable sensing element. The published work using planar waveguides for fluorescence biosensors have used only limited types of recognition elements, primarily DNA, antibodies and avidin. Studies describing the immobilization and use of different kinds of recognition molecules, such as receptors, in planar waveguide TIRF systems are limited to a handful of papers (Pawlak et al., 1998; Rowe-Taitt et al., 2000a; Schmid et al., 1998). A number of receptors have shown promise for use in fiber optic, SPR and resonant mirror sensors (Rogers et al., 1989; Fisher and Tjamhage, 2000; Lang et al., 2000; Altin et al., 2001; Puu, 2001). The methods used in these studies may also prove successful for the immobilization of functional receptors to planar waveguides for use in TIRF biosensors. There are many other membrane receptors which could be studied such as the nicotinic acetylcholine receptor for detection of organophosphate nerve agents. The use of nucleic acids for the detection of non-nucleic acid ligands such as proteins or small organic molecules is another relatively unexplored field although these systems have been studied with other biosensors. The use of DNA aptamers for analyte detection has recently been extended to use in planar waveguide TIRF biosensors (Potyrailo et al., 1998). Potyrailo and coworkers (1998) studied the binding of thrombin to a fluorescently labeled, immobilized DNA aptamer by monitoring changes in the evanescent-wave-induced fluorescence anisotropy. DNA and RNA aptamers can be synthesized for a number of target analytes with specificities comparable with the corresponding antibodies (Li and Li, 2000; Wang et al., 2001). DNA and RNA aptamers, as well as catalytic DNAs (Li and Li, 2000), show great promise for use in planar

115

Sapsford, Rowe Taitt, and Ligler waveguide biosensors as they can be synthesized; a later chapter by Ellington and coworkers describes these molecules. The use of "molecular beacons" could be an interesting development in planar waveguide TIRF. These systems make use of fluorescent probe-fluorescent quencher pairs and could be utilized with either the probe and quencher species immobilized on different binding pairs or, as studied by Liu and Tan (1999), on the same molecule. Liu and Tan synthesized an oligonucleotide such that it possessed a stem and loop structure with the 5' labeled quencher end and the fluorophore-labeled 3' end in close proximity. Addition of the target DNA separated the fluorophore and quencher ends, resulting in increased fluorescence. There are also a number of enzymes, such as oxidases and reductases, which react with a variety of clinically and environmentally important analytes. Enzymes are used for signal amplification in ELISAs, but have also been used as biological recognition elements in a wide variety of biosensors (reviewed by Wilson and Hu, 2000); although they have not as yet been utilized as capture biomolecules in planar waveguide TIRF studies. Though the majority of enzyme biosensors use an electrode as the signal transducer, some biosensors utilize enzymes that produce or consume an optically detectable species. Fiber optic systems using a number of luminescence-generating enzymes have been described (as reviewed by Rowe-Taitt and Ligler, 2001). Furthermore, enzymes have been used in a number of optical fiber measurements in which the target analyte is indirectly determined. Here the enzyme is either labeled or coimmobilized with the fluorescent probe. Typically the enzyme activity, upon introduction of the analyte, produces changes in the concentration of another species, such as oxygen, water, ammonia, or pH. These changes can result in either the increase or decrease in the fluorescence intensity of the probe. For example, oxygen quenching of ruthenium fluorescence was the basis of a fiber optic glucose biosensor (Moreno-Bondi et al., 1990). In addition, a fiber optic sensor for acetylcholinesterase (AchE) inhibitors utilizes fluorescein labeled AchE immobilized onto the surface. Enzyme activity with its substrate acetylcholine results in the production of protons and a quenching of fluorescein fluorescence intensity (Rogers et al., 1991). However to date, enzyme based recognition and detection systems have not been used widely in planar waveguide TIRF studies. The limitations of using enzyme-based systems in planar waveguide TIRF studies may lie in the requirement for a surface associated fluorophore. Research by Pawlak and coworkers may lead the way to immobilizing fully functional enzymatic recognition elements. This chapter was concerned with the technique of TIRF that utilizes planar waveguides and has discussed its background, history and current developments. The technique has shown great potential as a probe for a number of clinical, environmental and military applications. Prototypes exist for an automated, stand alone device which has shown potential in a number of fields (Ligler et al., 2001). However, with the continuing development of smaller electronics and 116

Planar Waveguides components, such as the CMOS camera currently used in the confocal microscopy studies of microarrays (Vo-Dinh et al., 1999; Askari et al., 2001), the potential for even smaller, handheld planar waveguide TIRF devices could become reality. The speed of signal transduction and relative immunity to matrix effects and other interfering influences, which are often problematic for other types of transduction methods, are two key advantages of planar waveguide TIRF biosensors. These advantages, and the additional benefit of spatially distinct sensing regions, are enabling these systems to gain advantage over other singleanalyte sensing systems. Although the majority of studies have centered on antibody-antigen type systems, expansion of the work concerned with DNA/mRNA and receptor-ligand binding interactions, as well as expansion into the use of enzymes and other types of recognition biomolecules could lead to a much wider field of applications. In conclusion, the future looks bright for planar waveguide TIRF both in optical biosensing and other areas of science, such as investigating biological processes.

6. Acknowledgements This work was supported by the Office of Naval Research and ~the National Aeronautics and Space Administration. The views are those of the authors and do not reflect opinion or policy of the U. S. Navy or Department of Defense.

7. References Abel, A. P., M. G. Weller, G. L. Duveneck, M. Ehrat and H. M. Widmer, 1996, Anal. Chem. 68, 2905. Altin, J. G., F. A. J. White and C. J. Easton, 2001, B iochim. B iophys. Acta 1513, 131. Arenkov, P., A. Kukhtin, A. Gemmell, S. Voloshchuk, V. Chupeeva and A. Mirzabekov, 2000, Anal. Biochem. 278, 123. Asanov, A. N., W. W. Wilson and P. B. Oldham, 1998, Anal. Chem. 70, 1156. Askari, M., J. P. Alarie, M. Moreno-Bondi and T. Vo-Dinh, 2001, Biotechnol. Prog. 17, 543. Bernard, A., B. Michel and E. Delamarche, 2001, Anal. Chem. 73, 8. Bhatia, S. K., L. C. Shriver-Lake, K. J. Prior, J. H. Georger, J. M. Calvert, R. Bredehorst and F. S. Ligler, 1989, Anal. Biochem. 178, 408. Bhatia, S. K., L. C. Shriver-Lake, L. Kondrackil, B. R. Ratna, R. B. Thompson and F. S. Ligler, 1990, Abstr. Pap. Am. Chem. Soc. 199, 96. Bhatia, S. K., J. J. Hickman and F. S. Ligler, 1992, J. Am. Chem. Soc. 114, 4432. Bhatia, S. K., J. L. Teixeira, M. Anderson, L. C. Shriver-Lake, J. M. Calvert, J. H. Georger, J. J. Hickman, C. S. Ducley, P. E. Schoen and F. S. Ligler, 1993, Anal. Biochem. 208, 197.

117

Sapsford, Rowe Taitt, and Ligler Bier, F. F., W. Stocklein, M. Bocher, U. Bilitewski and R. D. Schmid, 1992, Sens. Actuators B Chem. 7, 509. Birkert, O., H.-M. Haake, A. Schutz, J. Mack, A. Brecht, G. Jung and G. Gauglitz, 2000, Anal. Biochem. 282, 200. Blawas, A. S. and W. M. Reichert, 1998a, Biomater. 19, 595. Blawas, A. S., T. F. Oliver, M. C. Pirrung and W. M. Reichert, 1998b, Langmuir 14, 4243. Bradley, R. A., R. A. L. Drake, I. A. Shanks, A. M. Smith and P. R. Stephenson, 1987, Phil. Trans. R. Soc. Lond. B 315, 143. Brecht, A., A. Klotz, C. Barzen, G. Gauglitz, R. D. Harris, G. R. Quigley, J. S. Wilkinson, P. Sztajnbok, R. Abuknesha, J. Gascon, A. Oubina and D. Barcelo, 1998, Anal. Chim. Acta 362, 69. Brizzolara, R. A. and B. C. Beard, 1994, J. Vac. Sci. Technol. A 12, 2981. Brown, P. O. and D. Botstein, 1999, Nat. Genet. Suppl. 21, 33. Budach, W., A. P. Abel, A. P. Bruno and, D. Neuschafer, 1999, Anal. Chem. 71, 3347. Chittur, K. K., 1998, Biomater. 19, 357. Conrad, D. W., A. V. Davis, S. K. Golightley, J. C. Bart and F. S. Ligler, 1997, Proc. SPIE 2978, 12. Conrad, D. W., S. K. Golightley and J. C. Bart, 1998, U. S. Patent 5,773,308. Debouck, C. and P. N. Goodfellow, 1999, Nat. Genet. Suppl. 21, 48. Dodson, J. M., M. J. Feldstein, D. M. Leatzow, L. K. Flack, J. P. Golden and F. S. Ligler, 2001, Anal. Chem. 73, 3776. Duveneck, G. L., D. Neuschafer and M. Ehrat, 1995, International Patent GolN 21/77, 21/64. Duveneck, G. L., P. Oroszlan, A. P. Abel, B. Klee, V. Steiner, M. Ehrat, D. Gygax and H. M. Widmer, 1996, Proc. SPIE 2631, 14. Duveneck, G. L., M. Pawlak, D. Neuschafer, E. Bar, W. Budach, U. Pieles and M. Ehrat, 1997, Sens. Actuators B 38-39, 88. Duveneck, G. L., A. P. Abel, M. A. Bopp, G. M. Kresbach and M. Ehrat, 2001, Anal. Chim. Acta, in press. Ekins, R. P., 1995, Principles in Nuclear Medicine, N. H. Wagner, Z. Szabo, J. W. Buchanan Murphy eds., WB Saunders, Philiadelphia, 247. ,...... Ekins, R. P., 1998, Clin. Chem. 44, 2015. Feldstein, M. J., J. P. Golden, C. A. Rowe, B. D. MacCraith and F. S. Ligler, 1999, J. Biomed. Microdevices 1:2, 139. Feldstein, M. J., B. D. MacCraith and F. S. Ligler, 2000, US Patent 6,137,117. Ferretti, S., S. Paynter, D. A. Russell, K. E. Sapsford and D. J. Richardson, 2000, Trends Anal. Chem. 19, 530. Fisher, M. I. and T. Tjamhage, 2000, Biosens. Bioelectron. 15,463. Garland, P. B., 1995, Q. R. Biophys. 29, 91. Golden, J. P., L. C. Shriver-Lake, G. P. Anderson, R. B. Thompson and F. S. Ligler, 1992, Opt. Eng. 31, 1458. Golden, J. P., 1998, US Patent 5,827,748.

118

Planar Waveguides Golden, J. P., C. A. Rowe, M. J. Feldstein and F. S. Ligler, 1999, Proc. SPIE 3602, 132. Gullans, S. R., 2000, Nat. Genet. Suppl. 26, 4. Guschin, D., G. Yershov, A. Zaslavsky, A. Gemmell, V. Shick, D. Proudnikov, P. Arenkov and A. Mirzabekov, 1997, Anal. Biochem. 250, 203. Haab, B. B., M. J. Dunham and P. O. Brown, 2001, Genome Biol. 2, 4.1. Hall, E. A. H., 1990, Biosensors, Open University Press, Milton Keynes. Herron, J. N., K. D. Caldwell, D. A. Christensen, S. Dyer, V. Hlady, P. Huang, V. Janatova and H.-K. Wang, A.-P. Wei, 1993, Proc. SPIE 1885, 28. Herron, J. N., D. A. Christensen, K. D. Caldwell, V. Janatova, S.-C. Huang and H.-K. Wang, 1996, US Patent 5,512,492. Herron, J. N., D. A. Christensen, H.-K. Wang and K. D. Caldwell, 1997, US Patent 5,677,196. Huang, S.-C., K. D. Caldwell, J.-N. Lin, H.-K. Wang and J. N. Herron, 1996, Langmuir 12, 4292. King, K. D., G. P. Anderson, K. E. Bullock, M. J. Regina, E. W. Saaski and F. S. Ligler, 1999, Biosens. Bioelectron. 14, 163. Koltz, A., A. Brecht, C. Barzen, G. Gauglitz, R. D. Harris, G. R. Quigley, J. S. Wilkinson and R. A. Abuknesha, 1998, Sens. Actuators B 51, 181. Knight, J., 2001, Nature 410, 860. Lahiri, J., L. Isaacs, J. Tien and G. M. Whitesides, 1999, Anal. Chem. 71,777. Lang, S., J. Xu, F. Stuart, R. M. Thomas, J. W. Vrijbloed and J. A. Robinson, 2000, Biochem. 39, 15674. Leatzow, D. M., J. M. Dodson, J. P. Golden and F. S. Ligler, 2001, B iosens. B ioelectron., in press. Lee, J. E. and S. S. Saavedra, 1996, Langmuir 12, 4025. Li, J., Y. Li, 2000, J. Am. Chem. Soc. 122, 10466. Ligler, F. S., D. W. Conrad, J. P. Golden, M. J. Feldstein, B. D. MacCraith, S. D. Balderson, J. Czarnaski and C. A. Rowe, 1998, Proc. SPIE 3258, 50. Ligler, F. S., J. P. Golden, C. A. Rowe-Taitt, J. P. Dodson, 2001, Proc. SPIE 4252, 32. Lin, N. J., J. D Andrade, D. E. Yoshida, J. N. Herron and H. R. Yen, 1988, Abstr. Pap. Am. Chem. Soc. 196, 119. Liu, X. and W. Tan, 1999, Anal. Chem. 71, 5054. Liu, X. H., H. K. Wang, J. N. Herron and G. D. Prestwich, 2000, Bioconj. Chem. 11,755. Lockhart, D. J. and E. A. Winzeler, 2000, Nature 405, 827. Lu, B., C. Lu and Y. Wei, 1992, Anal. Lett. 25, 1. Lu, E. J., J. Xie, C. Lu, C. Wu and Y. Wei, 1995, Anal. Chem. 67, 83. Lundgren, J. S., N. Watkins, D. Racz and F. S. Ligler, 2000, B iosens. Bioelectron. 15,417. MacBeath, G. and S. L. Schreiber, 2000, Science 289, 1760. McLean, M. A., Stayton, P. S. and S.G. Sligar, 1993, Anal. Chem. 65, 2676. Moreno-Bondi, M. C., O. S. Wolfbeis, M. J. P. Leiner and B. H. P. Schaffer, 1999, Anal. Chem. 62, 2377. 119

Sapsford, Rowe Taitt, and Ligler Mrksich, M., 2000, Chem. Soc. Rev. 29, 267. Pawlak, M., E. Grell, E. Schick, D. Anselmetti and M. Ehrat, 1998, Faraday Discuss. 111,273. Pirrung, M. C., J. D. Davis and A. L. Odenbaugh, 2000, Langmuir 16, 2185. Plowman, T. E., S. S. Saavedra and W. H. Reichert, 1998, Biomaterials 19, 341. Plowman, T. E., J. D. Durstchi, H. K. Wang, D. A. Christensen, J. N. Herron and W. M. Reichert, 1999, Anal. Chem. 71, 4344. Potyrailo, R. A., R. C. Conrad, A. D. Ellington and G. M. Hieftje, 1998, Anal. Chem. 70, 3419. Puu, G., 2001, Anal. Chem. 73, 72. Rabbany, S. Y., B. L. Donner and F. S. Ligler, 1994, Crit. Rev. Biomed. Eng. 22, 307. Rogers, K. R., J. J. Valdes and M. E. Eldefrawi, 1989, Anal. Biochem. 182, 353. Rogers, K. R., C. J. Cao, J. J. Valdes, A. T. Eldefrawi and M. E. Eldefrawi, 1991, Fundam. Appl. Toxicol. 16, 810. Rogers, K. R., 2000, Mol. Biotechnol. 14, 109. Rowe, C. A., S. B. Scruggs, M. J. Feldstein, J. P. Golden and F. S. Ligler, 1999a, Anal. Chem. 71,433. Rowe, C. A., L. M. Tender, M. J. Feldstein, J. P. Golden, S. B. Scruggs, B. D. MacCraith, J. J. Cras and F. S. Ligler, 1999b, Anal. Chem. 71, 3846. Rowe-Taitt, C. A., J. J. Cras, C. H. Patterson, J. P. Golden and F. S. Ligler, 2000a, Anal. Biochem. 281, 123. Rowe-Taitt, C. A., J. P. Golden, M. J. Feldstein, J. J. Cas, K. E. Hoffman and F. S. Ligler, 2000b, Biosens. Bioelectron. 14, 785. Rowe-Taitt, C. A., J. W. Hazzard, K. E. Hoffman, J. J. Cras, J. P. Golden and F. S. Ligler, 2000c, Biosens. Bioelectron. 15,579. Rowe-Taitt, C. A. and F. S. Ligler, 2001, Handbook of Fiber Optic Sensing Technology, J. M. Lopez-Higuera, ed., John Wiley & Sons, in press. Sapsford, K. E., Z. Liron, Y. S. Shubin and F. S. Ligler, 2001, Anal. Chem., in press. Schmid, E. L., T. A. Keller, Z. Dienes and H. Vogel, 1997, Anal. Chem. 69, 1979. Schmid, E. L., A.-P. Tairi, R. Hovius and H. Vogel, 1998, Anal. Chem. 70, 1331. Schobel, U., C. Barzen and G. Gauglitz, 2000, Fresenius J. Anal. Chem. 366, 646. Schuderer, J., A. Akkoyun, A. Brandenburg, U. Bilitewski and E. Wagner, 2000, Anal. Chem. 72, 3942. Schult, K., A. Katerkamp, D. Trau, F. Grawe, K. Cammann and M. Meusel, 1999, Anal. Chem. 71, 5430. Schwarz, A., J. S. Rossier, E. Roulet, N. Mermod, M. A. Roberts and H. H. Girault, 1998, Langmuir 14, 5526. Silin, V., H. Weetall and D. J. Vanderah, 1997, J. Col. Interface Sci. 185, 94. Silzel, J. W., B. Cercek, C. Dodson, T. Tsay and R. J. Obremski, 1998, Clin. Chem. 44, 2036. 120

Planar Waveguides Sojka, B., P. A. E. Piunno, C. C. Wust and U. J. Krull, 1999, Anal. Chim. Acta 395,273. Szurdoki, F., K. L. Michael and D. R. Walt, 2001, Anal. Biochem. 291,219. Thompson, R.B. and F.S. Ligler, 1998, Proc. SPIE 1204, 35. Ulbrich, R., R. Golbik and A. Schellenberger, 1991, Biotechnol. Bioeng. 37, 280. Vigmond, S. J., M. Iwakura, F. Mizutani and T. Katsura, 1994, Langmuir 10, 2860. Vo-Dinh, T., J. P. Alarie, N. lsola, D. Landis, A. L. Wintenberg and M. N. Ericson, 1999, Anal. Chem. 71,358. Vo-Dinh, T. and B. Cullum, 2000, Fresenius J. Anal. Chem. 366, 540. Yoshida, D. E., D. A. Christensen, J. D. Andrade and W. M. Reichert, 1988, J. Electrochem. Soc. 135, C390. Wadkins, R. M., J; P. Golden and F. S. Ligler, 1997, J. Biomed. Optics 2, 74. Wadkins, R. M., J. P. Golden, L. M. Pritsiolas and F. S. Ligler, 1998, Biosens. Bioelectron. 13,407. Wang, L., C. Carrasco, A. Kumar, C. E. Stephens, C. Bailly, D. W. Boykin and W. D. Wilson, 2001, B iochem. 40, 2511. Wilson, G. S. and Y. Hu, 2000, Chem. Rev. 100, 2693. Wink, T., S. J. van Zuilen, A. Bult and W. P. Van Bennekom, 1997, Analyst 122, 43R. Zeller, P. N., G. Voirin and R. E. Kunz, 2000, Biosens. Bioelectron. 15, 591.

121

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 4

FLOW IMMUNOSENSOR

ANNEW. KUSTERBECK

Center for Bio/Molecular Science and Engineering Naval Research Laboratory Washington, DC 20375-5348 USA

The flow immunosensor combines the selectivity and sensitivity of traditional immunoassays with a non-equilibrium displacement reaction that allows rapid analysis of small-molecular weight compounds within minutes. Working assays have been developed for a wide range of molecules including explosives, drugs of abuse and environmental contaminants. Accurate determinations of analyte concentrations can be made on site, thus providing immediate feedback to field managers and law enforcement personnel. Side-by-side comparison of these measurements with laboratory instruments (HPLC, GC/MS) has demonstrated the accuracy and precision of the method. Commercial versions of the flow immunosensor have been engineered that integrate fluidics, electronics and computer control into a portable instrument. More recently, advanced laboratory prototypes of the biosensor have been fabricated to improve lowend detection, extend the applications to underwater sensing, enhance field ruggedness and assist in the manufacturing process. The evolution of this technology from laboratory prototype to field applications is presented in this chapter.

1. Principles of Operation Immunosensors have become a predominant form of biosensor due primarily to the ready availability of antibodies and their well-understood characteristics. Many of the problems of selective ligand binding and molecular recognition have been solved by nature, since an antibody can selectively bind an antigen in the presence of multiple compounds. The immunoassay format also has a long tradition of use in the medical and pharmaceutical research communities and is 123

Kusterbeck accepted by them as a standard detection platform. One biosensor format that has emerged for applications in a variety of new detection areas is the displacement flow immunosensor. This immunoassay system has been variously described in the literature as a non-conventional immunoassay system (Ghindilis et al., 1998) or as a category of chromatographic immunoassays (Hage and Nelson, 2001). In any case, the aim in developing the method has been to provide improved, faster and more efficient detection. The resulting automated or semi-automated system provides a more rapid response time without the need for incubation steps, reagent addition or user intervention. In the flow immunosensor, displacement immunoassays are incorporated into a device that is set up under continuous buffer flow. The essential components of the system are antibodies immobilized on a solid support, antigen analogs that are labeled with a reporter molecule, and the associated hardware needed to establish a controlled flowing system. Variations in any one of these basic elements allow the system to be adapted for different testing scenarios. Preparation for an assay first involves covalently immobilizing a monoclonal antibody (Ab) to a prepared surface. In parallel, reporter molecules are chemically coupled to a target antigen/analyte (Ag*). After purification, this labeled molecule is allowed to react with the bound Ab until equilibrium is reached, generally at least 2 hours or overnight. To perform an assay, the ANAg* substrate is placed in a support column and put under a buffer flow. When a sample containing the analyte of interest is injected into the flow stream, the labeled antigen molecules are displaced into the ~buffer and measured downstream by a detector. Figure 1 provides a simplified schematic of flow immunosensor operation. The integrated peak area of the reporter molecules can be calculated and is proportional to the number of analyte molecules injected, within apredetermined linear range for each antibody. The entire process is complete within 30-90 seconds, depending on the flow rate and the hardware used. Multiple sample injections can be made on a single prepared substrate without the need for additional reagents or regeneration, though high analyte concentrations (>10-20% of substrate capacity) will degrade the sensor response more rapidly. Figure 2 illustrates a typical dose-response curve observed for increasing concentrations of the compound RDX. As seen in this graph, the response is linear between 5 and 500 ng/ml. At higher concentrations, the response is not linear, but still indicates that a sample contains significant level of analyte.

124

Flow Immunosensor

Figure 1. Schematic of flow immunosensor operation. Step 1: Monoclonal antibodies are covalently attached to a solid substrate and incubated with labeled antigen analog (Ab/Ag* Substrate). Step 2: Sample is injected into the flow stream and labeled antigen is displaced. Step 3: Displaced labeled antigen and excess sample move down stream, where a detector measures the label.

Immobilization of the antibody is generally accomplished using known covalent chemistries. Tresyl chloride linkages (Kusterbeck et al., 1990; Wemhoff et al., 1992; Whelan et al., 1993) have been used as well as proprietary commercial chemistries supplied on membranes (Rabbany et al., 1999). Recently, silane chemistries and the immobilization procedure first described by Bhatia et al. (1989) have all been found to work exceptionally well for attaching antibodies to glass surfaces (Narang et al., 1997; Charles et al., 2000). The reporter molecule or label attached to the antigen analog has predominantly been a fluorophore, though radiolabels have also been used (Kusterbeck et al., 1990). Fluorophores tested in the flow irnmunosensor include fluorescein (Judd et al., 1995) and the cyanine-based near-infrared dye Cy5 (Bart et al., 1997a). In all cases, the important factor in synthesizing the fluorescent conjugate has been the inclusion of a spacer molecule of sufficient length to prevent fluorescence quenching. One strategy, which was successfully used, was the attachment of multiple fluorophores to an insulin-A chain carrier peptide (Bredehorst et al., 125

Kusterbeck 1991). More recently, carbon spacers of varying lengths have been employed to separate the cyanine dye Cy5 and modify the affinity of the Ab/Ag interaction. The solid substrates used for immobilization have included Sepharose gels (Kusterbeck et al., 1990; Whelan et al., 1993; Ogert et al., 1992), glass beads (controlled pore glass or solid) (Yu et al., 1996), bis-acrylamide/azolactone copolymer beads (Judd et al., 1995; Bart et al., 1997b), activated nylon membranes (Rabbany et al., 2000; Charles et al., 2000), fused silica capillaries (Narang et al., 1997; Charles et al., 1999), and polymethyl methacrylate (Holt et al, 2002). Table 1 summarizes the number of ways in which the system has been configured to achieve the appropriate assay parameters and detection thresholds for different analytes.

2. History The flow immunosensor has evolved over the past 15 years from an idea to working commercial prototype instruments. The many steps taken in this evolution include proof-of-principle experiments, development of a laboratory device, basic studies of antibody behavior, field testing of a laboratory breadboard, engineering of commercial prototypes, and demonstration/validation of the commercial instruments to gain regulatory approval and establish equivalency with accepted analytical methods.

2.1. Assay development Responding to the need of law enforcement agencies in the mid-1980's for a simple rapid method to detect illicit substances, including explosives and drugs of abuse, the Naval Research Laboratory (NRL) undertook the development of antibody-based biosensors. In that same time frame, Warden and coworkers (1987) demonstrated a "hit and run" assay for T-2 toxin using bound antigen on a substrate and displaced Fab' antibody fragments after short incubation periods.

126

Flow lmmunosensor

14 12

i " d~

4

i

O

0

"

i

-~lliO0

Figure 2. Representative flow immunosensor dose response curve. Increasing concentrations of the cocaine metabolite, benzoylecgonine (BE), were introduced to the system. The relative fluorescence intensity was measured for each injection. Values represent the standard error of the mean for 3 injections.

Wanting to conserve the expensive antibodies, NRL immobilized the antibody on the substrate and attempted to displace labeled antigen in a real-time reaction. This work led to a new displacement immunoassay system, the flow immunosensor. Initial proof-of-principle experiments were performed using the well-studied 2,4-dinitrophenol (DNP) molecule as a model antigen for the common explosive 2,4,6-trinitrotoluene (TNT). In these studies, an anti-DNP monoclonal antibody was covalently attached to an agarose gel (Kusterbeck et al., 1990). An 1125 label was conjugated to DNP via an insulin A-chain spacer molecule developed specifically for coupling a single small molecule to multiple labels (Bredehorst et al., 1991). The antibody/labeled antigen gel was placed in a small column, flow was started and samples were applied. Detection was accomplished by measuring the level of radioactivity in the eluted fractions. The next transition for the flow immunosensor came with the demonstration of a working assay using the fluorophore fluoroscein isothiocyanate (FITC) rather than a radiolabel. The signals achieved were comparable to radioactivity but did not involve the hazards associated with handling radiolabels. 127

Kusterbeek

Table 1. Analytes and formats reported for the flow immunosensor Analyte -"

Formats 13sed '"

DNP

Beads

TNT

Sample Sources Lab standards

References ....

Agarose gel

Lab standards Groundwater Air collection

Whelan et al 1993 Bart et al., 1997b Rabbany et al., 1999

Membranes

Groundwater Soil

Charles et al., 2000 Gauger et al., 2001

Capillary

Groundwater

Narang et al., 1997

Sol-gel on polymethyl methacrylate

Lab standards

Holt et al., 2002

RDX

Agarose gel Membranes Capillary

Groundwater Groundwater Lab standards

Bart et al., 1997b Charles et al., 2000 Charles et al., 1999

PETN

Copolymer beads Agarose gel

Lab standards

Judd et al., 1995

Lab standards Urine

Ogert et al., 1992 Yu et al., 1996

Cortisol

Agarose gel

Blood, plasma

Kapstein et al., 1997

Alachlor

Agarose gel

Lab standards

Lehotay et al., 1994

PCBs

Agarose gel

Lab standards Oil samples

Charles et al., 1995 Charles et al., 1995

Benzoyl ecognine

128

Kusterbeck et al., 1990

III I

Flow Immunosensor After work with the DNP model system, efforts focused on detection of the explosive TNT. Because monoclonal antibodies specific for TNT were not commercially available, custom hybridomas were produced and screened for activity in a displacement immunoassay performed in a 96-well plate. The fluorophore FITC was coupled to a TNT analog, trinitrobenzene (TNB). With adjustment in buffer composition and flow rate, detection sensitivities for TNT were achieved at the low ng/ml level (Whelan et al., 1993). Over the next ten years, the displacement assay concept was extended to antigens that were of interest in other real-world scenarios. Assays were demonstrated for the explosives TNT (Whelan et al., 1993), hexahydro-l,3,5-trinitro-l,3,5 triazine (RDX) (Bart et al., 1997a; Charles et al., 2000) and pentaerythritol tetranitrate (PETN) (Judd, et al., 1995). For the drugs of abuse, assays were validated for benzoylecognine (Ogert et al., 1992), marijuana, and opiates (unpublished data). Appropriate conditions were also determined for the environmental contaminant polychorinated biphenyls (PCBs) (Charles et al., 1995) and the pesticide alachlor (Lehotay et al., 1994). Finally, efficacy of the flow immunosensor concept was shown with the therapeutic drug cortisol (Kaptein et al., 1997). Sample matrices tested have included air samples, laboratory standards in buffer and solvents, environmental groundwater and soil samples, and body fluids, including urine, saliva and blood serum (Table 1). The samples were generally tested without pretreatment, though filtration was reported in several cases as a means of removing large particulates that could impede flow. The criteria established to validate individual assays were a good linear response over several orders of magnitude, demonstration of a detection level significantly below a regulatory threshold, and accuracy and precision with a statistically significant number of samples. In addition, acceptance by the regulatory communities required good agreement when compared with previously validated standard laboratory analysis by High Performance Liquid Chromatography (HPLC), Gas Chromatography/Mass Spectrometry (GC/MS), or Gas Chromatography/Electron Capture Detector (GC/ECD). Sample collection, one of the most critical aspects of field analysis, has been accomplished in many ways. Initially, the Midwest Research Institute (MRI) (Kansas City, MO) Spincon collector was used to test TNT displacement immunoassay performance with air samples. The TNT vapors/particulates produced by a vapor generator or solid block of TNT were pulled from the air

129

Kusterbeck

Figure 3. Commercial prototypes developed based on flow immunosensor technology. Shown on the left is the IMPACT Test System manufactured by Lifepoint, Inc. for on site measurements of drugs of abuse. On the right is the FAST 2000 engineered by Research International (RI) for environmental monitoring of explosives (photos courtesy of Lifepoint, Inc. and RI).

and into a water sample by the Spincon. Though the MRI instrument was relatively inefficient (similar to other air samplers), the prototype flow immunosensor was able to detect TNT (or DNT) when present in the collected samples at levels above 5 ng/ml. These values were independently confirmed using GC/ECD analysis. Later air tests have employed the Team Technologies (Boston, MA) SCAEP air sampler and the Research International (Woodinville,

WA) SASS. Other sample collection strategies used include surface swipes with cotton swabs and elution of bound material, solid phase extraction after extended exposure to contaminated air or water, collection of body fluids (blood, urine, saliva), and simple grab samples of water or process effluents. The last method is by far the easiest and most cost effective for on-site work. Perhaps the most difficult collection approach has been the extraction of analytes from soil samples. Gauger et al. (2001) demonstrated that a 3-minute solvent extraction using acetone, with subsequent drying of the sample and resuspension in buffer, could be used successfully to measure explosives levels in soil.

2.2. Hardware development In parallel with assay development, the design, fabrication and engineering of flow immunosensor instrumentation has occurred. In the simplest form, the laboratory prototype device consisted of off-the-shelf parts, including small disposable plastic columns, flexible polypropylene tubing, a standard HPLC 130

Flow Immunosensor injector, a peristaltic pump and a fluorescence detector with a flow cell (Wemhoff et al., 1992; Whelan et al., 1993). Recording of the signal was accomplished using a simple analog chart recorder. A version of this system was used in a series of field tests conducted in 1995 at two U.S. military bases (Bart et al., 1997). These tests provided the first true demonstration of the ability of the flow immunosensor to measure environmental contaminants on site. Though workable, this early system was cumbersome and required lengthy setup and skilled users. Simple instruments with minimal user input were needed. Accordingly, in cooperation with private industry, a flow immunosensor prototype device was configured that incorporated small commercial in-line valves and a dedicated laptop computer with custom-designed software. After establishing and demonstrating a workable design, engineering of commercial prototypes was undertaken. For both drug detection and explosives detection, years of development were required before production prototypes were fabricated. Early systems were unreliable and gave a number of false positives and/or false negatives. Frequent problems were encountered with the fluidics or computer software. The culmination of the hardware development has been the engineering of second and third generation instruments designed to improve flow immunosensor performance. Shown in Figure 3 are commercial prototypes, the IMPACT Test System (Lifepoint, Inc., Rancho Cucamonga, CA) and the Research International (Woodinville, WA) FAST 2000 instrument. These standalone instruments are capable of simultaneous analysis of multiple drugs of abuse (IMPACT) or explosives (FAST) within minutes. 2.3. Basic kinetic studies

Another critical aspect of flow immunosensor development has been an examination of the basic kinetics of antigen/antibody interactions. Because the displacement event relies on the dissociation rate of the antigen from the antibody, numerous studies have examined the effect of flow rate (Wemhoff et al., 1992), immobilized antibody density (Rabbany et al., 1994), antibody hetergeneity (Selinger and Rabbany, 1997; Rabbany et al., 1997), and more recently, mass transport considerations and fluid dynamics (Holt et al., 2000). In these studies, a number of theoretical equations unique to the flow immunosensor were derived and experimentally tested. The most important findings are highlighted below. A starting point for any kinetic study is the well-documented antibody/antigen behavior in solution (Bersofsky et al., 1984). At equilibrium in a solution, the binding constants are defined as: K = k~/~ = [AbAg]/([Ab][Ag])

131

(1)

Kusterbeck where Ab is free antibody; Ag is free antigen, AbAg is the antibody/antigen complex, ka is the association rate constant and La is the dissociation rate constant. In contrast to these solution interactions, the properties of immobilized antibodies were not well understood. However, the initial equation for the displacement event following antigen injection can be defined as: AbAg* + Ag ~

AbAg + Ag*

(2)

where Ab/Ag* is the immobilized antibody/labeled antigen complex, Ag is the injected antigen, Ab/Ag is the immobilized antibody/unlabeled antigen complex and Ag* is the displaced labeled antigen. With subsequent injections of unlabeled antigen, more antibody binding sites are depleted of labeled antigen and the efficiency of the reaction changes significantly. This undissociated fraction, termed 0, represents the difference between the total amount of labeled antigen initially bound tothe solid support and the amount displaced after each addition of unlabeled antigen (Wemhoff et al., 1992). A second equation was then derived to measure the displacement efficiency (De) for subsequent injections as follows: De = (displaced A g * ) , 1

(3)

(loadedAg) 0 Finally, looking at the displacement reaction as a first order rate process led to an equation that could be used to calculate the "apparent" dissociation rate, or lq as a function of time, t. ln0 kd = t (4) m

- - - . . , . - . , . -

This early analysis, with supporting experimental work, demonstrated that flow immunosensor assays were dependent on flow rate. Low flow rates allowed a higher displacement efficiency, with more labeled antigen released and lower detection levels achieved. As expected, higher flow rates increased the displacement rate and improved response time, but decreased low-end detection (Wemhoff et al., 1992). These results demonstrated that values for a given Ab/Ag pair could be predicted and outlined successful approaches for adjusting the detection threshold of the flow immunosensor. Rabbany and coworkers (1995) extended this work to show how the density of immobilized antibody affected the performance of the flow immunosensor assays. Yu et al. (1996) provided a mathematical solution to account for loss of signal molecules that allowed the system to be recalibrated with subsequent sample injections. Later repetitive displacement experiments also illustrated how 132

Flow Immunosensor multiple samples could be tested on a single Ab/Ag* substrate with extended periods of flow without significant loss of signal intensity (Rabbany et al., 2000). Finally, Holt et al. (2000) investigated the fluid dynamics surrounding the displacement immunoassay. All formats of the assay prior to 2000 (bead, membrane, capillary) were compared based on physical parameters of the system, antibody densities, flow rate and length of washing. The computational models developed in this study showed that accurate measurements could be made in an immunoassay under flow conditions and clearly defined the direct relationship between assay performance and flow immunosensor operating parameters.

3. Applications/Implementation of Technology 3.1. Aviation security As mentioned previously, law enforcement agencies represented the initial impetus for NRL flow immunosensor efforts. Aviation security programs at the Federal Aviation Administration (FAA) supported early development of the TNT sensor for pre-flight screening of cargo holds, baggage compartments and passenger cabins of airplanes. In the early field tests, the NRL prototype could detect TNT (or DNT) particles present on the airplane when a MRI Spincon was able to collect the explosives into water at levels above 5 ng/ml. These values were independently confirmed using GC/ECD analysis. Further assay development was pursued for the other explosive molecules of importance in terrorist activities, including RDX and PETN, the primary components of plastic explosives. Reliable assays were demonstrated for all the compounds of interest (Whelan et al., 1993; Judd et al., 1995; Bart et al., 1996). True. implementation of the technology, however, has been problematic due to regulatory constraints, instrument costs, passenger impatience/non-acceptance of screening delays, and the extremely low levels of explosives available to collect and analyze.

3.2. Drug detection A second target of opportunity for the flow immunosensor has been with the law enforcement community. The U.S. Customs Service and the Drug Enforcement Agency (DEA) often require rapid confirmation of illicit substances at border checkpoints and other field sites. As with explosive compounds, most drugs of abuse are small molecules, making them good candidates for flow immunosensor applications. Accordingly, flow immunosensor assays have been developed and tested for cocaine and its major metabolite, benzoylecognine (Ogert et al., 1992). The data in this study showed excellent agreement with the "gold standard" analytical method, GS/MS and led directly to collaboration with private industry. Under a cooperative research and development agreement (CRADA) between 133

Kusterbeck NRL and US Drug Testing (now Lifepoint, Inc.), a commercial prototype was built and tested (Yu et al., 1996). From the late 1990's to the present time, the flow immunosensor developed under this program has demonstrated accurate, quantitative assays for the five National Institutes of Drug Abuse (NIDA-5) drugs: cocaine, opiates (heroin/morphine/codeine), phencyclidine (PCP), amphetamines/methamphetamines, and tetrahydro-cannabinol (THC, marijuana). In addition, the company has completed validation of their instrument and received FDA 510K approval of the NIDA assays. Commercially available disposable cassettes can accommodate a number of immunoassays and allow for the analysis of up to 10 analytes simultaneously. Potential markets and customers of this technology include on-site drug screening, law enforcement personnel, and medical testing laboratories. The company expects to market the IMPACT Test System instruments in late 2001 (Lifepoint, Inc., First Quarter Report, 2001). A related drug detection application of the flow immunosensor is the monitoring of patient levels of therapeutic drugs in a continuous mode. Kaptein et al. (1997) used flow immunosensor technology to develop a rapid sensitive assay for continuous monitoring of cortisol levels in patient sera. 3.3. Environmental monitoring and field demonstrations

Perhaps the most investigated field of use for flow immunosensor technology is on-site environmental analysis. In the late 1980's, environmental regulators and field managers at Superfund remediation Sites were looking for ways to improve site characterization of contaminated sites and reduce costs involved with longterm testing and monitoring (Rogers et al.~ 1996). The emergence of immunochemical methods for environmental testing and the acceptance in 1993 of nine immunoassays for inclusion in the U.S. EPA Solid Waste Testing Methods was the first step in this direction (Van Emon, 1996). Flow immunosensor assays previously developed for explosives detection in air samples were easily adapted to groundwater and soil analysis. With the support of the EPA, the flow immunosensor was used in an extensive series of tests at U.S. military bases designed to validate on site instrument performance. Preliminary flow immunosensor data from laboratory tests of explosives standards in uncontaminated groundwater showed a high degree of accuracy in the RDX and TNT assays. Accuracy values for samples in buffer ranged from 93% to 99% when known concentrations were analyzed. Calculations for precision ranged from as low as 6% to as high as 15%. An assessment of the flow immunosensor in initial experiments with field samples also suggested that the method was comparable to most immunoassay or HPLC methods when samples were free of other interfering compounds.

134

Flow Immunosensor

T N T Contour Plots

Figure 4. Contour plots showing TNT concentrations at UMDA as measured by the flow immunosensor and HPLC (Bart et al., 1996). Reprinted with permission of the American Chemical Society.

In fact, the method could very effectively map both the extent of contamination and concentration of explosives in contaminated aquifers (Bart et al., 1996). Figure 4 shows a contour plot generated from analysis of TNT levels in groundwater monitoring wells as Umatilla Army Depot Activity (UMDA), Umatilla, Oregon. As seen in this plot, the flow immunosensor field test results were almost identical to those obtained from laboratory HPLC analysis. Similar agreement with laboratory methods was observed in later field tests using the commercial FAST 2000 flow immunosensor. Figure 5 illustrates the agreement between HPLC laboratory data and the FAST 2000 on-site determinations for RDX in groundwater monitoring wells at UMDA. In general, the data showed that FAST 2000 immunosensor measurements for individual explosives were not as affected by matrix interferants at higher concentrations, though matrix effects were apparent at the low end. Overall, for the five field sites used in field validation studies, the FAST 2000 was able to detect TNT and RDX with reasonable accuracy. The method was also predictive for two of the three field sites selected for detailed statistical analyses using the T-test. For most samples, the standard deviations observed were < 10% and the comparison with HPLC results was good. These conclusions echo those of Hennion and Barcelo (1998) who suggest that the most appropriate use of immunoassay 135

Kusterbeck 1200

ryz

0 . 7 3 x - 41.59

= 0.81

1000

,.J

-

~.,

800

o

600 -

O')

-

I-

1/)
400

-

200

-

I1.

| .

I

I

I

I

I

i

0

200

400

600

800

1000

HPLC

1200

(.u,g/L)

Figure 5. Composite of results for groundwater field samples from Umatilla Army Depot. Calculations of RDX concentrations were performed on identical sample splits using the FAST 2000 on site (replicate analyses, n=7) and an EPA-certified laboratory for HPLC analysis. Note: Values below the detection limit have been omitted.

methods is primarily as a preliminary field screening technique. If possible, quantitative measurements of field sites can be made for later laboratory confirmation. Van Emon and Gerlach (1998) also determined that immunochemical methods, like the flow immunosensor, are well suited to field analysis because they tolerate high sample loading, are sensitive, inexpensive and require no hazardous materials or solvents. As seen earlier, flow immunosensor assays have been developed for additional environmental contaminants including the pesticide alachlor (Lehotay, 1994) and PCBs (Charles et al., 1995). These assays have yet to be validated and field tested. 3.4. A d v a n c e d

development

A breakthrough in flow immunosensor technology occurred in the late 1990's when Narang and coworkers (1997) immobilized antibodies inside fused silica capillaries, and with this new substrate, performed a displacement immunoassay. By using this format, they were able to lower the detection limit 2-3 orders of magnitude below that previously observed for flow immunoassays (1 part per trillion versus 1 part per billion). These additional orders of magnitude in 136

Flow Immunosensor sensitivity opened the way to detection of trace chemicals in environmental samples (Narang et al., 1998; Charles et al., 2000) as well as improving the ability of the system to meet regulatory guidelines and decrease matrix effects. The ongoing development of this format into a plastic coupon with machined internal channels (Holt et al., 2001) should improve the ruggedness and extend the utility of the method to other field applications.

4. Advantages and Limitations Immunoassays and immunochemical field methods have inherent advantages that have been reviewed extensively (Van Emon and Gerlach, 1998; Rogers, 1995; Hennion and Barcelo, 1998). Hage and Nelson (2001) recently concluded that assays performed in flow provide direct, simple and rapid detection of analyte, highlighting a unique advantage of the flow irranunosensor format. Because flow is continuous, non-reactive compounds are removed from the surface and subsequent samples can be injected without resetting the sensor. Another advantage of the flow immunosensor is that it is field portable and can be carried on-site for real-time analysis of the various analytes. Results are obtained within minutes rather than days or weeks. If there is a question about a particular sample, it can be analyzed again immediately. In addition, a single substrate can be used for multiple tests (typically >20), depending on analyte concentration. The ability to alter formats to fit a particular test scenario or application is also a major strength. For example, based on flow rate studies, Rabbany et al. (1995) determined that slowing the instrument flow rate improved low-end detection. Increased flow rates made the assay less sensitive, but extended the system lifetime. Similarly, changes can be made in the solid substrate, the reporter molecule, and the selected Ab/Ag pair to change assay detection limits. Finally, unlike competitive flow immunoassays systems recently described for cocaine and opiates (O'Connell et al., 1999; Eldefrawi et al., 2001), the flow immunosensor signal is seen as rise from baseline, not a decrease in signal. It is a direct detection of molecular recognition. Throughout development of the technology, a major obstacle has been incorporation of effective sampling methods. For many relevant applications, air is the preferred medium. Because the flow immunosensor works in a water medium, samples collected by air samplers must be partitioned into an aqueous medium. This can be done in two ways. The antigen may be pulled directly into water using a cyclone sampler or charged particle collector, like the SCAEP and Spincon samplers described earlier. A second method involves solid phase

137

Kusterbeck

Figure 6. Schematic of next-generation FAST flow immunosensor. Built by Research International, the instrument is an integrated system with features that include on-board battery power, remote control capabilities and autosampling in the underwater environment.

extraction, in which the sample is trapped on a coated surface such as a selective filter for later extraction. These preparative methods have been extremely effective for air and water samples (Barshick and Griest, 1998; Durrach et al., 1998). Charles et al. (2001) have recently shown that this method can also enhance flow immunosensor perfomance and improve the accuracy twofold in comparison to previous work. Nevertheless, the process is cumbersome, costly, and time consuming. A related problem with operation in the field is the negative effect of environmental matrices on flow immunosensor performance. Routinely, little-tono sample preparation is used. Thus, reliable analysis of a sample depends upon a given water sample having little to no effect on the antibody/antigen interaction or the fluorescence intensity. This is decidedly not the case for a given percentage of field samples, where detrimental effects on system accuracy and precision are seen with samples from selected field sites. Similar negative effects on immunoassay performance have been observed with samples high in salt, picric acid, or humic acids (Myers et al., 1994). A final limitation of the technology is the degradation or loss of signal with time and repeated use. Because the labeled antigen is constantly being displaced from the immobilized antibody, there is a measurable loss of assay response (Rabbany 138

Flow Immunosensor et al., 1995). Also, standard deviations begin to increase and low-end detection is lost. This makes it critical that calibration steps be included any immunosensor protocol.

5. Future Potential

Future advances in the flow immunosensor technology will involve improvements in assay components and hardware/software innovations. The ideal flow immunosensor for field use would contain integrated reference/calibration capabilities and could be run for extended periods (days to weeks) with unattended use. To achieve these goals, the system will need to be rugged, reliable, and capable of remote operation. A step in this direction is the newly engineered version of the flow immunosensor, the Unmanned Underwater Vehicle (UUV) FAST unit. Shown in Figure 6, the advanced system is selfcontained, has a small footprint (6-7 inches in diameter) and does auto-sampling using appropriate pumps and valves. The instrument can be operated with a simple push button menu or remotely using embedded microprocessors. A second promising research effort is the development of a "flow immunosensor on a chip" reported by Holt et al. (2002). In this format, the substrate can be regenerated so that the system lifetime is extended and low detection limits are maintained. In a sensor that uses a flow-based displacement event for detection, the limiting step is not only the kinetics of the binding event, but the actual engineering of the device as well. Thus, if the detector is sufficiently distant from the solid substrate, there may be peak broadening as well as a very real delay in response. While not necessary for applications involving environmental cleanup, any lag time may interfere with "real time" monitoring in an underwater sensor using autonomous vehicles or in process control scenarios. Screening of large areas would be precluded because the time between sample collection and analysis would be too great. For process monitoring, the instrumental analysis time would need to be matched to the reaction rate. This could be achieved by implementing kinetic determinations that use only the rate of response to calculate analyte concentration. Key factors in this approach would involve significantly improved fluidics and increased signal sampling rates. To improve detection limits and reduce or eliminate matrix effects on the immunoassay, the immunosensor could incorporate on-line solid-phase extraction and preconcentration. Yu et al. (2001) have recently demonstrated such a system and are testing this modification on a microfluidic device that uses monolithic porous polymers. With respect to receptor binding and signal transduction, genetic engineering and molecular imprinting are being explored to gain stable molecular recognition 139

Kusterbeck elements with appropriate affinities that withstand solvents and work despite the extremes in salt, pH and temperature found in environmental samples. Phage displayed-peptides, single chain antibodies (ScFv) and synthetic aptamers (Lee and Walt, 2000) that have specificity for the target molecules might also play a role in future systems. Alternative reporter molecules that amplify a signal or eliminate the need for fluorescence detector are also possible. Surface engineering of biomaterials and the use of specific coupling chemistries to improve antibody attachment and retain a higher level of activity on the substrate have the potential to increase the rate and/or amplitude of signal transduction. A possible future area of application is in monitoring food and water quality. There has been an increased interest in maintaining a safe water supply and detecting contaminants in food, including meat and milk. Standard immunoassays with ELISAs or kits have been used previously (Aga and Thurman, 1997) and antibodies have been reported for more than 35 common pesticides (Hennion and Barcelo, 1998); therefore the starting materials for assay development are readily available. The ultimate challenge for acceptance of the flow immunosensor technology (and other immunoassay-based methods) is certification by regulatory agencies for onsite use. Before this can occur, the field method must be accurate, must meet all data quality objectives, and must provide consistent analytical performance in a variety of samples (Van Emon and Gerlach, 1998). The need to analyze large numbers of samples may also require high sample throughput. Despite more than a decade of intense research, including demonstration/validation field studies, prototype fabrication, manufacturing by private industry and improvements to assay performance, true commercialization of the flow immunosensor has yet to occur. This slow technology transfer mirrors that seen with the rest of the biosensor community (Weetall, 1999). The reality is that transition remains a signficant barrier, despite good technology. At the present time, the flow immunosensor can be considered a useful tool for semiquantitative field analysis that can provide valuable information to on site decision makers.

6. Acknowledgements The author thanks Lisa Shriver-Lake and Paul Charles for technical assistance and advice. This work was supported by~ the Environmental Security and Technology Certification Program (ESTCP) and the Office of Naval Research. The views expressed here are those of the author and do not reflect those of the U.S. Navy or the Department of Defense.

140

Flow lmmunosensor 7. References

Aga, D.S. and E.M. Thurman, 1997, in Immunochemical Technology for Environmental Applications, D. Aga and E.M. Thurman, eds., American Chemical Society, Washington, D.C., ACS Symposium Series 657, 1. Barshick, S.A. and W.H. Griest, 1998, Anal. Chem. 70, 3015. Bart, J.C., L.L. Judd, K.E. Hoffmanl A.M. Wilkins, P.T. Charles and A.W. Kusterbeck, 1996, Am. Chem. Soc. Symp. Ser. 657, 210. Bart, J.C., L.L. Judd and A.W. Kusterbeck, 1997a, Sens. Actuators A Physical: 38, 411. Bart, J.C., L.L. Judd, K.E. Hoffman, A.M. Wilkins, P.T. Charles and A.W. Kusterbeck, 1997b, Environ. Sci. Technol. 31, 1505. Berzofsky, I.A and I.J. Berkower, 1984, in Fundamental Immunology, W.E. Paul, ed., Raven Press, New York, 595. Bhatia, S.K., L.C. Shriver-Lake, K.J. Prior, J.H. Georger, J.M. Calvert, R. Bredehorst and F.S. Ligler, 1989, Anal. Biochem. 178,408. Bredehorst, R., G.A. Wemhoff, A.W. Kusterbeck, P.T. Charles, R.B.Thompson, F.S. Ligler and C.W. Vogel, 1991, Anal. Biochem. 193, 272. Charles, P.T., D.W. Conrad, M.S. Jacobs, J.C. Bart and A.W. Kusterbeck, 1995, Bioconjugate Chem. 6, 691. Charles, P.T. and A.W. Kusterbeck, 1999, Biosens. Bioelectron. 14, 387. Charles, P.T., P.R. Gauger, C.H. Patterson, Jr. and A.W. Kusterbeck, 2000, Environ. Sci. Technol. 34, 4641. Charles, P.T., B.M. Dingle, S.V. Bergen, P.R. Gauger, C.H. Patterson Jr. and A.W. Kusterbeck, 2001, Field Anal. Chem. Technol. 5, 1. Ghindilis, A.L., P. Atanasov, M. Wilkins, and E. Wilkins, 1998, Biosens. Bioelectron. 131, 113. Durrach, M.R., A. Chutjuan and G.A. Plett, 1998, Environ. Sci. Technol. 32, 1354. Eldefrawi, M.E., A.N. Nath, N.A. Anis, M.S. Bangalore, K.P. O'Connell, R.P. Schwartz and J.Wright, 2000, Appl. Biochem. Biotech. 87, 25. Gauger, P.R., D.B. Holt, C.H. Patterson, Jr., P.T. Charles, L. C. Shriver-Lake and A.W. Kusterbeck, 2000, J. Haz. MatI. 83, 234. Hafner, F.T., R. A. Kautz, B.L. Iverson, R.C. Tim and B.L. Karger, 2000, Anal. Chem. 72, 5779. Hage, D.S. and M.A. Nelson, 2001, Anal. Chem.73, 199A. Hennion, M. C. and D. Barcelo, 1998, Anal. Chim. Acta 362, 3. Holt, D.B., S.Y. Rabbany, A.W. Kusterbeck and F.S. Ligler, 1999, Rev. Anal. Chem. 18, 1999. Holt, D.B., A.W. Kusterbeck and F.S. Ligler, 2000, Anal. Biochem. 287, 234. Holt, D.B., P.R Gauger, A.W. Kusterbeck and F.S. Ligler, 2002, Biosens. Bioelectron. 17, 95. Judd, L.L., A.W. Kusterbeck, D.W. Conrad, H. Yu, H.L. Myles Jr. and F.S.Ligler, 1995, Proc. SPIE 2388, 198.

141

Kusterbeck Kaptein, W.A., J.J. Zwaagstra, K. Venema, M.H.J. Ruiters and J. Korf, 1997, Sens. Actuators B 45, 63. Kusterbeck, A.W., G.A. Wemhoff, P.T. Charles, D.Yeager, R. Bredehorst, C.W. Vogel and F.S. Ligler, 1990, J. Immunol. Meth. 135, 191. Lee, M. and D.R. Walt, 2000, Anal. Biochem. 282, 142. Lehotay, S.J., S.F. Wright, A.B. Demilo and L.H. Stanker, 1994, Proc. 21 st Annual Fed. Anal. Chem. Spec. Soc. Conference, St. Louis, MO. Myers, K.F., F. McCormick, B.G. Strong, P. Thome and T. Jenkins, 1994, Cold Regions Research Engineering Laboratory Technical Report IRRP-94-4. Narang, U., P.R. Gauger and F.S Ligler, 1997, Anal. Chem. 69, 2779. Narang, U., P.R. Gauger, A.W. Kusterbeck and F.S. Ligler, 1998, Anal. Biochem. 255, 13. O'Connell K.P., J. J. Valdez, N.L. Azer, R.P. Schwartz, J. Wright, and M.E. Eldefrawi, 1999, J. Immunol. Meth. 225, 157. Ogert, R.A., A.W. Kusterbeck, G.A.Wemhoff, R.Burke and F.S.Ligler, 1992, Anal. Lett. 25, 1999. Rabbany, S.Y., A.W. Kusterbeck, R. Bredehorst and F.S. Ligler, 1994, J. Immunol. Meth. 168, 227. Rabbany, S.Y., A.W. Kusterbeck, R. Bredehorst and F.S.Ligler, 1995, Sens. Actuators B 29, 72. Rabbany, S.Y.R. Piervincenzi, L.J. Judd, A.W. Kusterbeck, R. Bredehorst, K. Hakansson and F.S. Ligler, 1997, Anal. Chem. 69 175. Rabbany, S.Y., W.A. Marganski, A.W. Kusterbeck and F.S. Ligler, 1998, Biosens. Bioelectron. 13, 939. Rabbany, S.Y., W.J. Lane, W. A. Marganski, A.W. Kusterbeck and F.S. Ligler, 2000, J. Immunol. Meth 246, 69. Rogers, K. R., 1995, Biosens. Bioelectron. 10, 533. Rogers, K.R. and C.L. Gerlach, 1996, Environ. Sci. Technol. 30, 486A. Selinger, J.V. and S.Y. Rabbany, 1997, Anal. Chem. 69, 170. Van Emon, J.M. and C. L. Gerlach, 1996, in Environmental Immunochemical Methods, J.M.Van Emon, L.L. Gerlach and J.C. Johnson, eds., American Chemical Society, Washington, DC, 2. Van Emon J.M. and C. L. Gerlach, 1998. Anal. Chim. Acta 376, 55. Warden B.A., K. Allam, A. Sentissi, L.J. Cecchini and R.W. Giese, 1987, Anal. Biochem. 162, 363. Weetall, H. H., 1999, Biosens. Bioelectron. 14, 237. Wemhoff, G. A., S.Y. Rabbany, A.W. Kusterbeck, R.A. Ogert, R. Bredehorst and F.S. Ligler, 1992, J. Immunol. Meth. 156, 223. Whelan, J. P., A.W. Kusterbeck, G.A. Wemhoff, R. Bredehorst, and Ligler, 1993, Anal. Chem: 3561.F.S. Yu, K., A.W. Kusterbeck, J.P. Whelan, M. Hale, F.S. Ligler, 1996, Biosens. Bioelectron. 11,725. Yu, Cong, M.H. Davey, F. Svec and J.M.J. Frechet, 2001, Anal. Chem. 73, 5088.

142

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 5

FLUORESCENCE LIFETIME BIOSENSORS

RICHARD B. THOMPSON' PH.D. Department of Biochemistry and Molecular Biology University of Maryland School of Medicine Baltimore, Maryland 21201 USA

Fluorescence lifetime biosensors transduce recognition of the analyte by a biologicallyderived molecule as a change in fluorescence lifetime. While typically more complex instrumentally than optical biosensors which readout a steady state intensity, they have useful advantages for certain applications. Among these advantages are straightforward transduction design, facile calibration, relative freedom from artifact, broad dynamic range, and ready adaptation to fiber optics.

1. Principles of O p e r a t i o n

Many fluorescence-based biosensors have been described wherein recognition of an analyte by a biomolecule is transduced as a change in emitted fluorescence; many are described in chapters elsewhere in this book. A small subset of these transduce the presence or level of the analyte as a change in fluorescence lifetime; it is these with which we shall concern ourselves. The fluorescence lifetime is a fundamentally different parameter than the more familiar fluorescence intensity or polarization. Formally, the lifetime is the average amount of time a fluorophore spends in the excited state between absorption of a photon and its emission as fluorescence; the lifetime is the reciprocal of the emissive rate. An ensemble of fluorophore molecules excited with an instantaneous flash (8-pulse) of light will exhibit fluorescence decaying in an exponential fashion until undetectable (Figure 1). From a theoretical standpoint, the decay is described by Equation 1: I(t) = Io e -t/~

(1)

143

Thompson

~ .

Io/e

o

0.10

T =

z o.01 ILl

O3 ILl

=o

0 E3 ._J It..

_

l _ .

0

_

l

.

_

I

10 20 TIME, NANOSECONDS

_

30

Figure 1. Concept of fluorescence lifetime. Following a brief (picosecond) excitation flash, the fluorescence emission decays in an exponential fashion; the time required to decay to 1/e times the initial intensity is the lifetime. where I is intensity, t is time and z is the lifetime, which is also the time required for the intensity to decay to 1/e times its initial value, I0. It is important to note that the lifetime is a property of a particular fluorophore molecule, like its molecular weight or formula. It is also important to note that emission from a particular fluorophore is a random event, which may occur picoseconds after excitation, or at a time corresponding to many lifetimes later. Under ideal conditions, the lifetime of a fluorophore will be described adequately by a single exponential decay like that in Figure 1. More commonly for fluorophores in solution, the time-dependent decay of emission will not be adequately described by a single exponential (emissive rate); rather, a variety of processes can result in the observation of a so-called multiexponential decay, where the decay is better described by a sum of exponentials: I(t) = I0 ~: ~i e U'

(2)

Detailed study of such multiexponential decays provides insight into their physical causes, especially in the case of emission from tryptophanyl residues in proteins. Similarly, fluorescence decay kinetics have provided insight into a host of biophysical processes, including exciton migration in photosynthesis, 144

Fluorescence Lifetime Biosensors observation of conformational changes in macromolecules, and the dynamics of membranes. For the purposes of fluorescence lifetime-based sensing, these details are of only passing interest. Typical lifetimes for allowed transitions (e.g., singlet-singlet and triplet-triplet) for organic molecules are in the nanosecond regime. More precisely, observed lifetimes are seldom longer than that of pyrene, which is a few hundred nanoseconds, or shorter than that of POPOP, a scintillation fluor, at approximately one nanosecond. Several molecules exhibit shorter lifetimes due to intramolecular quenching processes (Thompson and Gratton, 1988; Kawski et al., 1991), but they also have commensurately reduced quantum yields (see below). Important exceptions to this range of lifetimes include transition metal complexes exhibiting metal-to-ligand charge transfer (MLCT) transitions such as the Ru(II) polypyridyls, and certain complexes of the lanthanide ions, Eu, Tb and Sm, which all have long lifetimes extending up into the millisecond time range (Demas and Crosby, 1971). Molecules producing emission from nonallowed transitions (e.g., triplet-to-singlet) termed phosphorescence also have long lifetimes, but this is observed in solution only under special conditions. Importantly, the phenomena described below used to transduce analyte recognition as a change in lifetime work almost irrespective of the lifetime of the fluorophore, and irrespective of whether the decay is single- or multiexponential.

1.1. Transduction approaches Several different means have been used to transduce analyte recognition as a change in fluorescence lifetime. Of course, many classical fluorescent indicators for metals and other analytes exhibit a change in fluorescence intensity upon binding the analyte (White and Argauer, 1970; Femandez-Gutierrez and Munoz de la Pena, 1985), and frequently (although by no means always) this also results in a change in lifetime. The most common approach is quenching, wherein the analyte recognition results in a diminution of the fluorescence intensity. The degree of quenching is expressed as a function of analyte (quencher) concentration using the Stem-Volmer equation: I0 / I = 1 + k [Q]

(3)

where I's are intensity, k is the Stern-Volmer quenching constant, and [Q] is the concentration of the quenching species. The inverse of the intensity plotted as a function of quencher concentration is ideally a straight line with slope equal to the quenching constant (Figure 2). As a rule of thumb, an efficient quencher able to diffuse rapidly in solution (or if the fluorophore has a lifetime in the tens of nanoseconds or more) can quench the fluorophore's emission fifty percent when the quencher is present near millimolar concentrations. To a first approximation, 145

Thompson ,

. . . . . .

_

F0 = 1.0 O.8

5

LLI r

z

0.6

I,i

0.4

0

D J

0.2 ..

-,-.,.,,_.._

t.L

0

_

I

!._

2

1_

.

l

!

4

I

-

6

_

-

J_,

!_.

2

r" QUENCHER 11

_J_

J

-

4

J

!

-6'

{i QUENCHER ]

Figure 2. Stern-Volmer fluorescence quenching. The left panel depicts the fluorescence intensity as a function of quencher concentration for a classical Stern-Volmer quencher; the right panel depicts a Stern-Volmer plot of the same data.

quenching processes can be divided into "dynamic" or "collisional" quenching, where the quenching species diffuses to collide with the fluorophore while in its excited state, and "static" quenching, wherein the quencher makes a complex with the fluorophore while it is in the ground state, quenching it immediately and completely so no fluorescence is observed from such complexes. For our purposes the distinction is important, since dynamic quenching results in a reduction in lifetime, and static quenching does not since the complexes are "dark" and do not contribute to the emission. The reduction in apparent lifetime in collisional quenching arises because the diffusing quencher effectively selects against those fluorophores which individually spend a longer time in the excited state; essentially, the longer the fluorophore spends before emitting, the greater its chances of colliding with a quencher and losing its energy before it can emit. Fortunately perhaps, the simplified picture discussed above does not describe many quenching processes in detail, nor their effect on fluorescence decay (Lakowicz, 1999). In particular, most quenching processes do not require physical contact of the quencher with the fluorophore, but rather act through space, sometimes over significant distances (several nanometers). Equally important is the corollary that not all complexes need be dark, but rather can exhibit emission (and a reduced lifetime) even when saturated. Consequently, for most fluorescence lifetime biosensors based on forming a noncovalent, often reversible complex between the biologically-derived recognition molecule and the analyte, binding of the analyte results in a reduction in fluorescence lifetime (and intensity) due to some quenching process, as depicted in Figure 3. Measurements of intensity (and lifetime) as a function of analyte concentration thus display saturation, which is not observed with classical collisional quenching. We presume (as is usually the case) that the binding of analyte is described simply by the Law of Mass Action such that the fractional saturation of 146

Fluorescence Lifetime Biosensors

RECEPTOR

" 0

ANALYTE

COMPLEX

FLUORESCENCE

FLUOROPHORE INTENSITY

~

INTENSITY

LIFETIME

~

LIFETIME

PHASE ANGLE ~

PHASE ANGLE

MODULATION

MODULATION

~,

~k

Figure 3. Generic fluorescence biosensor scheme based on quenching. The recognition molecule or receptor labeled with a fluorophore in the absence of bound analyte exhibits high intensity and long lifetime, with concomitant large phase angle and low modulation. Upon binding the analyte, the intensity decreases along with the lifetime and phase angle, whereas the modulation increases.

the binding site is simply a function of the analyte concentration. An example of this behavior is depicted in Figure 4, which shows that increasing concentrations of any of several different metal ions result in reduced intensity. Note that the quenching in this instance cannot be collisional, since given the short lifetime of the fluorescent label and the modest concentration of the metal ions (picomolar to nanomolar), the ion would have to diffuse faster than possible to effect such a reduction in intensity. Several quenching mechanisms are known and may be exploited for biosensor development depending on the chemistry of the analyte and the recognition molecule, or receptor. These include spin-orbit coupling enhanced by the presence of high-Z atoms, electron transfer, quenching by unpaired electrons (radicals), quenching by proton transfer, electron exchange, quenching by paramagnetic species (Lakowicz, 1999), and finally quenching by F6rster energy transfer, which will be dealt with separately below. Most of these mechanisms are short range, requiring contact or close approach (< 0.5 nm) of the quencher to 147

Thompson

apo-

v) I--510

-

~

EM. 5 5 0

nm

o-~O--~ o

_ . .

~D >I1:

CA ] / - N 6 7 C - ABD ( T )

EXC.430nm, 9

.

8

rn n~

~2 z w bz w

,

4

o

x~,

~

w 2 0 _J tl.

0,.O_o \\

o

z w U m

I

-13

I

-12

!

-11

,O..r

!

-10

I

, I

I

I

-9

-8

-7

-6

"

,I

-5

LOG [ M = * ] , M O L A R

Figure 4. Fluorescence intensities of ABD-t-labeled N67C-apo carbonic anhydrase as a function of free Cu(II) (open diamonds), Zn(II) (filled diamonds), Cd(II) (open circles), Ni(II) (triangles), and Co(II) (filled circles) concentration. the fluorophore to deexcite the latter. Occasionally, the analyte will have chemical features which can be exploited for quenching-based sensors: for instance, the iodine atoms in the hormone triiodothyronine are efficient quenchers by the so-called heavy atom effect. In general, the fluorophore will be chosen to maximize the efficiency of the particular quenching process. Thus, the efficiency of quenching by electron transfer depends critically on the standard oxidation potential of the fluorophore with respect to the reducing/oxidizing species (Rehrn and Weller, 1970). If the analyte itself is unlikely to be a good quencher, it may be possible to have it compete for binding to the receptor with a modified form of the analyte which is. Among quenching mechanisms, the most useful for lifetime biosensing is Ft~rster transfer, which is a dipolar coupling of excited donor and acceptor molecules if they have a good match (overlap) between the donor emission and the acceptor absorption (Forster, 1948). Like other dipole couplings, the efficiency of the transfer increases as the inverse sixth power of the distance; the propensity of a given donor to transfer to a given acceptor is thus a function of distance, spectral

148

Fluorescence Lifetime Biosensors

3.2i

~ 2_...8

39 5 0 M H z

5

<

u.. 2 4 33 .......

2.0

[

I

25

.I

~

,

-I

75 50 EOSIN-A 9 (Alg/ml)

,,,

. . . . . . . . . . . .

n

, LO

!

~

100

-

30

Figure 5. Phase angle (right scale) and apparent fluorescence lifetime (left scale) of FITC-IgG as a function of eosin-labeled antigen concentration.

overlap, quantum yield of the donor, and an orientation factor. For a donoracceptor pair these factors are conveniently expressed as the F0rster distance, R0, at which distance transfer is 50% efficient; this number can be readily calculated. While Ftirster transfer is commonly observed as sensitized fluorescence of the acceptor, from the standpoint of the donor it is a quenching process that competes with emission of photons from the donor, and always results in a decline in the lifetime. An early example of an antibody-based lifetime biosensor was described by Ozinskas et al. (1993) (Figure 5). In this case the energy transfer immunoassay of Ullman (Ullman and Schwarzberg, !981) is adapted for readout by a change in fluorescence lifetime instead of intensity. The advantage of energy transfer as a quenching mechanism in biosensor design is the ease with which the energy transfer can be controlled to optimize the response of the transducer. For instance, with a biomolecule receptor binding an analyte which is itself an energy transfer acceptor, the position of the fluorescent donor attachment to the receptor and its degree of overlap may be chosen to maximize the change in lifetime (Thompson et al., 1996a). A comparable level of control by design is difficult to achieve using other mechanisms.

149

Thompson

z f ~b

E XC ITATION

:>(.z3 z

LLJ

T ==4.5

ns

f'-

MHz

60

_

,

I

O

,,.

I

TIME ~.~ "~ 2 ~" • TO= m

=

m

,

.

.

.

I

-

30

NANOSECONDS

( MODULATION

(-l'~n .d ~

.

20

10

FREQUENCY

) --- 2 ~ ' f

)I~

Err'fISSION

/ rrt E X C I T A T I O N

~

'

A I C 'B / C -

9 I

TO--.rm "r e < ' r "r

FOR m

SHOULD

MONOEXPONENTIAL

FOR

MULTIEXPONENTIA APPROXIMATELY

L EQUAL

1/~,~

Figure 6. Concept of frequency domain lifetime measurement. The excitation (solid line) is sinusoidally modulated at some frequency in the megahertz range; the emission is phase shifted and demodulated to a degree determined by the lifetime or lifetimes of the sample.

In other cases, binding of the analyte to the receptor does not result in quenching of a fluorophore at all, but rather an increase in intensity and (sometimes) lifetime. Thus some of the saccharide biosensors based on bacterial periplasmic binding proteins exhibit substantial increases in intensity (Li and Cass, 1991), and likely, lifetime. One of the fluorescent-labeled carbonic anhydrase variants (apo-N67C-ABD) also displays a nine-fold increase in intensity upon zinc binding (Thompson et al., 1999), but a fifty percent increase in fluorescence lifetime, which is quite usable for sensing purposes. The disparity between intensity increase and lifetime increase upon zinc binding indicates the presence of a static quenching component when the protein is in the apo-form that is somehow negated when zinc binds. Certain fluorophores such as ANS display enhanced emission intensity and lifetime in hydrophobic solvents and in putatively hydrophobic environments when bound to proteins (Slavik, 1982), but reliably coupling the effect to binding of an analyte is difficult. 1.2. Measurement of lifetimes A precondition for a lifetime-based sensor is obviously the ability to measure the lifetime, or some parameter related to the time-dependent fluorescence decay. While a thorough review of lifetime measurement is beyond the scope of this chapter, suffice it to say that there are two main approaches: time and frequency 150

Fluorescence Lifetime Biosensors

I- - - " - - /

-

9~

#J ~" 1 i

"

~: -~.--~::

-.

PERCENT "~ ..-..x~,, -,,, t.sns 4.ons '~-',,'~,", 0

100

x\ x x

/ 213o r ~ I 50

70 50

I

1

s;~ ULATEO

1 ~'0

/

| x

0.8

~'~-

,,:,,. ',/.,~~

7

',

'd,','~,////

" \"

,

'

5

\

="

_

~:

',, , "

g - 0.2

0

0.0

1

to

Ioo

FREQUENCY, MHz

Figure 7. Frequency-dependent phase shifts (solid lines) and modulations (dashed lines) for 4.0 nsec mono-exponential decay (curve 1), 1.5 nsec (curve 5), and mixtures thereof (curves 2, 3, and 4). domain; only the latter is widely used in lifetime biosensing. The time domain measurement embodies determining the time-dependence of the fluorescence emission following a brief flash of excitation, as illustrated in Figure 1. While many time domain methods have been described, the most popular and by far the most accurate and sensitive is time-correlated single photon counting (TCSPC) (Birch and Imhof, 1991). TCSPC has found little favor in sensing because, unless pulsed diode lasers or extremely expensive mode-locked lasers are used for excitation, acquisition of a single decay curve can take hours. By comparison, frequency domain instruments (phase fluorometers) provide rapid readout of two lifetime-related parameters, the phase and modulation. In frequency-domain fluorometry, the sample is excited with light whose amplitude is sinusoidally modulated at a frequency close to the reciprocal of the lifetime: typically, about 1 - 300 MHz. The fluorescence emission is perforce also modulated at the same frequency, but it is phase-delayed and demodulated to a degree that (for a single exponential) is a simple function of the lifetime (Figure 6). If the phase and modulation are measured over a hundred-fold range of modulation frequencies and plotted as a function of the logarithm of the frequency, one obtains a data set like those depicted in Figure 7. The curves in Figure 7 are depicted for a 1.5 nsec monoexponential decay, a 4 nsec monoexponential decay, and three different mixtures of the 1.5 and 4 nsec decays, corresponding to partial saturation of the transducer binding site. The phase angles increase monotonically from zero with frequency, asymptotically 151

Thompson

FRACTION 1.0

0.8

1.5nsec 0.6

0.4

COMPONENT 0.2

0.0 0.9

7O

.' ..... ;....... tO Ld t.d LM C3

J_J

t.9

50

_

70MHz

"-,,

!

-J

0.8

~. I I I

0.7

--Z 0 I--

40

0.6

< ..J E3 C~ 0

,:30

0.5

:~

Z ,< LIJ

tn ,<

I 13.

20

-0

-

j .... 0.2

FRACTION

I 0.4

SIMULATED I J0.6 O.B

4.0 nsec

0.4 1.0

COMPONENT

Figure 8. Phase angles (solid line) and modulations (dashed line) as a function of the fraction of 4.0 nsec component (bottom scale) in a mixture of 4.0 and 1.5 nsec emitters (Thompson and Patchan, 1995a). approaching 90 degrees. The modulations are nearly a mirror image of the phases, monotonically declining from 1.0 and asymptotically approaching zero. Note that the shapes of the modulation and phase curves for the two monoexponentials, 1.5 and 4 nsec, are identical - - they are merely the same curve shifted in frequency space. By comparison, mixtures of the two exhibit curves which cannot be superimposed on a monoexponential decay. These data are an example of what would be obtained for a lifetime biosensor transducer which exhibited a four nanosecond lifetime in the free state, and a 1.5 nsec lifetime in the bound state. Such data sets can be fitted to assumed decay laws, and the proportions and lifetimes of the components determined by an iterative process (Straume et al., 1991). For a sensor transducer which exhibits differing lifetimes in the analytebound and analyte-free states, it is unnecessary to do this. Rather, plotting the phase angles and modulations for the different mixtures at some suitable modulation frequency (70 MHz in Figure 8) leads to curves like those shown in Figure 8. Experimentally, one measures the phase angles and modulations as a function of analyte concentration instead of fraction bound, but the results appear similar. 152

Fluorescence Lifetime Biosensors

-

l

i

iq-lf~Ti

1.25/~M 50

_

r

MOPS

J..... i w 1 1 1 1 1

pH 7 . O j

'l

i

I m'w=1~ I

i-

~

BICINE-Co=+,25'C

i

I =J~i,

- 0.74

....o EXC. 4 4 2 nm t

EM. KV 5 2 O h m

/ o " o

,-, 4 6 ul 9 w Lit IZ

0 42 tuJ r~ ..I 0 z

1 0 rnM

~-

F 131C "CA - ABD- T

0,70 B.

0 9

I:I

.

- 0.66

-

38

0.62

-

- 0.58

I.u 3 4 v) . ,< 3: _ a. 3 0

L I

z _o .j

_

26

J o

o 0 :E

- 0.54

O_ t " - " ~l ' -l - o t - Fll n! i, 10

_

t_ i

i Li j t , l _ _ I

t

100

i ] i t ,[[

_ .t

,

i , a ~J_ 0 . 5 0

a000

[ g o '§ 1, N A N O M O L A R

Figure 9. Phase angles (filled circles) and modulations(open circles) as a function of Co(II) concentration for apo- F131C-ABD-T carbonic anhydrase.

An example of such data is depicted in Figure 9, wherein the phase and modulation of apocarbonic anhydrase labeled with ABD-T at a cysteinyl residue replacing phenylalanine-131 is shown as a function of free Co(H) ion concentration. At concentrations well below the metal ion KD (about 100 nanomolar), the phase angle roughly corresponds to the unquenched fluorescence lifetime, whereas at high metal ion concentrations, the phase and modulation correspond to the partially quenched, reduced lifetime of the bound state. Note that complete quenching of the label by bound metal ion would result in no emission from the bound state, thus no perturbation of the apparent lifetime and derived phase angles and modulations. Phase fluorometers are now available from several manufacturers. A schematic description of such an instrument is shown in Figure 10. The basic configuration and heterodyning approach were introduced by Spencer and Weber (1969) and are used in subsequent instruments. Key components include: the light source, which may be a laser, lamp, LED or electroluminescent source; and the modulator, which modulates the light intensity at the frequency 03 set by the frequency synthesizer and is typically a (broadband) Pockels cell or a (narrowband) acousto-optic device. Importantly, light sources such as LED's, laser diodes, and electroluminescent devices (Lippitsch et al., 1988; Thompson et 153

Thompson

MODULATOR

F'REOUENC~

~

oE-TEc~a~d, ~ ....

l EM's~'~ ....... I

DETECTOR

'

/

DATA A'At.~S,S

........... <,."~t.. ~ " ~ / / \

Figure 10. Schematicof multifrequencyphase fluorometer.

al., 1992) can be directly modulated, avoiding the expense and complexity of the modulator altogether. The phase and modulation of a small portion of the excitation are monitored by a reference detector, while the bulk of the light excites the sample mounted on a turntable together with a reference fluorophore of known lifetime. Errors in phase and modulation arising from path length differences thus cancel out. Another key feature of the fluorometer is the heterodyne detection, wherein the PMT detector gain is modulated at a frequency (co +LSc0) slightly offset from the modulation frequency (Birks and Dyson, 1961). This mixing in the detector causes signals to be produced at the sum (203 + A03) and difference (A03) of the two as well as the fundamental; these signals maintain the same phase and modulation relationships between the two detectors. Thus the sum and fundamental frequency signals, being at high frequencies, are easily discarded using a low pass filter, leaving the difference frequency signal at perhaps 25 Hz carrying all the information, and whose phase and modulation are trivial to measure with high precision. As long as the synthesizers in Figure 10 maintain their A03 frequency offset, the signal can be measured at modulation frequencies ranging from kHz to GHz. Recently, phase fluorometers have appeared which are significantly smaller, cheaper, and simpler than the research instruments available heretofore. These devices are typically single frequency, with directly modulated semiconductor or laser sources, and semiconductor detectors (Levy et al., 1997; Harms and Rao, 1999). One commercial version is approximately five-fold cheaper and resides in a package at least three-fold smaller than the current research instruments. This suggests that battery-powered, portable field instruments are not far off. 154

Fluorescence Lifetime Biosensors

PROTEN I

IIMMOBILIZATtON

9

9

9

9

~

- - ~ 1 ~ ? ~ 1 ] .- -

-

C - -

0'-,", .

.

GMBS

~ - [c..z :~-~--o..,-.

N ~ FLUORE iSCENT-LABELED CA M2

Figure 11. Schematic of immobilization method of Bhatia et al (1989).

In our definition (Thompson and Walt, 1994), a sensor is a device capable of continuously (or quasi-continuously) monitoring some (usually liquid) matrix for the presence of the analyte. Thus for many scenarios, such as monitoring flowing blood or the ocean, mixing the fluorescent-labeled protein (or other transducer) in the matrix is impractical. The obvious solution is to immobilize the transducer molecule on some substrate to be inserted in the matrix, and interrogated optically. Fortunately, many approaches have been developed for this purpose (Cass and Ligler, 1998). One that we have used with some success was developed by Bhatia et al. (1989) and is diagrammed in Figure 11. A silica substrate is derivatized with a thiol silane, and then the protein transducer molecule is coupled through the use of a heterobifunctional crosslinker. The protein is stable under refrigeration for many months using this approach (Thompson et al., 2000). 155

Thompson 1.3. Measurement through optical fiber Fluorescence lifetime-based biosensing is particularly well suited to remote sensing through optical fiber (Thompson, 1991). Fluorescence-based fiber optic sensors determine the analyte in situ by using an optical fiber with the transducer molecule(s) immobilized or entrapped at its distal end to conduct excitation light to the transducer and fluorescence emission from the transducer back to the detector; in many instances, a single fiber is used for both purposes. Essentially, the analysis is brought to the sample instead of vice versa; this has particular advantages for monitoring analytes in inaccessible, hazardous, or remote environments, or intracorporeally (Peterson et al., 1980). Use of separate fibers to carry excitation to and emission from the transducer makes the instrumentation simpler, since coupling the excitation source and the detector to their respective fibers is trivial. However, there can be difficulty in assuring that there is good registration between the volume illuminated by the excitation fiber(s) and light collected by the emission fiber(s); high numerical aperture (N.A.) fibers are helpful in this regard. Walt and colleagues (Barnard and Walt, 1991) have used endoscopic fiber bundles together with an imaging detector successfully for some years, but not for lifetime-based biosensing. Using a single fiber to conduct both makes for a somewhat more complex instrument, but permits the use of long lengths of (relatively inexpensive) telecommunications fiber, permits the use of pigtails so that the transducer can be readily changed if it fails, and simplifies the construction of the sensing tip. The following discussion will focus on the single fiber system. Measuring lifetimes .through fibers is well known tothe art, and straightforward in the frequency domain (Bright, 1988; Thompson and Lakowicz, 1993). For the single fiber systems, o n e adapts the phase fluorometer optics to launch the excitation into the proximal e n d o f the fiber, and collects the fluorescence coming back to the detector. The optical configuration for doing this is depicted in Figure 12,wherein the modulated~excitation passes through a hole in a n offaxis paraboloid and is focused on the proximal end of the fiber by an objective; fluorescence returning is not quite recollimated by the objective, reflects off the off-axis paraboloid into the detector. Properly speaking, what the instrument measures is the change in phase angle and modulation of the transducer, since the long path of the fiber introduces an arbitrary phase shift and demodulation to the emission. Fortunately, this may be corrected for by calibrating in solutions that give a known value of lifetime of the transducer; with modem synthesizers, the phase angle drift is small over several days.

156

Fluorescence Lifetime Biosensors

FtI~ER O#T!C PHABE FL!UORO~'ts

{.X#LOOED VIEW

.r

Figure 12. Schematic of fiber optic phase fluorometer (Thompson et al., 1996b).

2. History The history of lifetime-based biosensing is modest, as fluorescence lifetime measurements were applied to sensing of any kind in only the last twenty years. Temperature was measured by changes in the phosphorescence lifetime of rare earth phosphors, and ultimately this technology was adapted to fiber optic use (Grattan and Zhang, 1994). The first lifetime-based chemical sensors were developed in the late 80's (Lippitsch et al., 1988; Carraway et al., 1991) to measure oxygen by quenching of ruthenium polypyridyl complexes; this continues to be an active area of work. The Lakowicz group subsequently described several lifetime-based chemical sensors using classical indicator systems (see review by Szmacinski and Lakowicz, 1994). They probably described the first fluorescence lifetime biosensor when they demonstrated a lifetime-based immunoassay using energy transfer. However, their approach could not be used for continuous monitoring, given that it was a competitive assay with (probably) the typical effective irreversibility of antibody binding (Ozinskas et al., 1993). The Lakowicz group has also developed non-antibody lifetime biosensors, in particular for glucose (Lakowicz and Maliwal, 1993). It is widely appreciated that a glucose sensor capable of continuous, in vivo monitoring of glucose levels is much preferable to intermittent (1 - 2 times per day) sampling, as is done with the current electrochemical biosensor; moreover, such a sensor might be used to control an insulin pump, creating an "artificial pancreas". Tolosa et al. (1999) 157

Thompson

~oo 0 ,uM Glucose I~uM 4,uM 8~uM

z

_<2 .J

75 0

~E "~

Global

50

Fit

"_.t

~D

z .<

n

25

L

0 O-~

1 0 "'~

10 ~

~0 ~

;02

1.0~

FREQUENCY (MHz)

Figure 13. Frequency-dependent phase shifts and modulations for varying concentrations of glucose. From Tolosa et al. (1999).

adapted the glucose/galactose binding protein (GGBP) from E. coli to glucose sensing by introducing a cysteinyl residue to replace the glutamine at position 26, which is close to the saccharide binding site; the cysteinyl residue was derivatized with an ANS analog to form ANS26-GGBP. They found that although the fluorescence intensity dropped 50% upon saturation with glucose (at 8 ~tM) there wasno appreciable change in the lifetime. They therefore adopted an approach developed in the Lakowicz lab, wherein the cuvette containing the sample and fluorescent indicator is coated with a long lifetime MLCT probe in an impervious film, whose intensity and lifetime do not change upon addition of glucose. Under these conditions, the emission is a mixture of fluorescence from the MLCT probe (whose intensity is constant) and the ANS26-GGBP, whose intensity varies with glucose concentration. Because the proportion of emission from the ANS26-GGBP varies with glucose concentration, the apparent time dependence (or frequency-dependent phases and modulations) also vary with glucose concentration, and these can be related to the glucose concentration (Figure 13). The modulation shows the largest changes at around 1 MHz, making it possible to quantitate glucose at concentrations in the range of 1 to 10 gM (Figure 14). This is a substantially lower level than that required for glucose sensing in blood, but it illustrates the principle. 158

9

Fluorescence Lifetime Biosensors

0.9

Glucose

Sensing

Based

on Modulation at 2.1 MHz Q

N

3: ~E

(M I'--

z 0.8 0 p,. < ..,J E) D 0 :E

0.007 ~ fr! k

.;'.,

II il |1 II II )1

|1 af

i

-~, r-'-/I c = 0 . 2 p M

0.7

--j

I

0

I

_t

2

....

4

I

6

..........

i

8

__

10

[GLUCOSE ] , MICROMOLAR

Figure 14. Modulation as a function of glucose concentration at 2.1 MHz (Tolosa et al., 1999).

The group has also demonstrated a fluorescence lifetime biosensor for glutamine that utilizes similar principles, but uses an E. coli periplasmic binding protein labeled with the relaxation probe acrylodan at a cysteinyl residue replacing serine-179 (Dattelbaum and Lakowicz, 2001). The fluorescence intensity declines about three-fold upon saturation with glutamine, and the average lifetime declines from 2.28 nsec to 0.97 nsec, resulting in a very usable change of 18 degrees in phase angle at 110 MHz and an apparent Ko of 0.7 paVl (Figure 15). One hopes this sensor molecule can be modified to detect glutamate as well, in view of the high scientific interest in glutamate as a neurotransmitter and practical interest in monosodium glutamate as a food additive. Presumably future versions of these sensors will appear with the proteins immobilized onto a substrate for continuous monitoring, and with labels excitable with visible light sources. Lifetime-based biosensors for analysis of metal ions have employed carbonic anhydrase as a transducer molecule. It is widely appreciated that for many practical applications, a chemical sensor must be selective even m o r e than sensitive because of the complexity of the matrices in which the analyte is found. 159

Thompson

20 18 ~~3t)

-o

16

GInBP S179C-acrylodan A{ = 18 degrees 110 MHz

14 12

--~ 10 c

<

8

~ c-a..

6

4

0

0.01

0.1

1

Glutamine ( ~ M ) Figure 15. Phase angles at 110 MHz for glutamine-binding protein labeled with acrylodan at position S 179C as a function of glutamine concentration. Examples of this are blood serum and sea water, where a sensor for metal ions at trace levels (ppb and below) must not only respond to those levels (roughly 50 nM and below), but be immune to interference from species present at millionfold higher concentrations, like Ca(N) and Mg(II). Samples submitted to classical analytical techniques such as ICP-MS or GFAAS typically undergo some preprocessing or separation, which is not feasible for a sensor intended for a real-time response. Inasmuch as most classical metallofluorescent indicators are poorly selective (Fernandez-Gutierrez and Munoz de la Pena, 1985), the well-known selectivity (Lindskog and Nyman, 1964) of carbonic anhydrase U (CA) for binding Cu(II) and Zn(II) in its active site was adapted to the development of a fluorescence-based sensor. In particular, wild type human apocarbonic anhydrase II has high affinities for Cu(II) (KB = 0.1 pM), Zn(II) (KB = 4 pM), Cd(II) (KD = 100 pM), Ni(II) (KD = 100 nM), and Co(H) (KD = 2 ~xM), with no measurable binding of other metals, particularly Ca(H) and Mg(II), at millimolar levels (McCall, 2000). Figure 16 highlights a key advantage of metal ion biosensors based on carbonic anhydrase. By making modest changes in the protein structure, the afffmity and selectivity of the transducer can be usefully modified. For instance, the T93/S95/V97 variant exhibits thousand-fold better sensitivity to Cu(U), and fifty-fold worse sensitivity to Zn(II); clearly, this variant could be used under conditions where much better discrimination against Zn(II) 160

Fluorescence Lifetime Biosensors

i6 712:

14

,s .J O

:g

------ C u

-

12

" - - C u

.....

EU ----

Zn

---

Cu

-

L..J

10

.---

Co

-

...,

-.--- Cd

....==.-

Zn

---

Co

Cu

Cd

o

T

. . . .

Fe

-----Co

~ Zn

--

;

--- ---

V

-

-----

Ba

LO (= ",e" --

O1 O m I

~

-~~

Cd ----

Rb

_

Zn

Mn

~

, ~ -

Ca

----

Co

Sr

----

Ni

___

l~n

----

Zn

~ -

----

C d Ni

----

Co

Ba

- - - Z n Cd

- -

--"Ni

-

Sr

SEA

WATER (TOTALS; BRULAND )

t

FURA-2 FLUORESCENT INDICATOR

CA]I

l

I

CA]].

CAI[

CA]I

WILD

VARIANT

VARIANT

VARIANT

TYPE

Tg 3 $ g B V g 7

HZ1,gN

-

Hg4 N

Figure 16. Selectivity of sensor transducers. Concentrations of metal ions in sea water (left column) and apparent KD'S (other columns) are shown from (right to left) CA variants H94N, H119N, T93/S95/V97, wild type CA, and the fluorescent indicator Fura2.

is necessary than is available from the wild type protein (Hunt and Fierke, 1997). Other variants have been developed which exhibit a more rapid response to changes in zinc concentrations than the wild type enzyme (Huang et al., 1996). For zinc detection, the pioneering work of Chen and Kernohan (1967) was exploited wherein they showed that a sulfonamide inhibitor, dansylamide, bound to apo-carbonic anhydrase with a seven-fold increase and 100 nm blue shift in the fluorescence emission, as well as a change in the fluorescence lifetime. Thompson and Jones (1993) showed that the apoprotein bound dansylamide negligibly in the absence of zinc, and that the zinc concentration could be related to the ratio of emissions in the blue and green portion of the spectrum (Figure 17). In fact, the dansylamide also undergoes an increase in fluorescence lifetime upon binding to the protein from less than 3 nsec to 22 nsec; this large change is very suitable for lifetime-based sensing. In particular, at 22 MHz the phase shift is as great as 40 degrees, an unusually large value (Figure 18). 161

Thompson hot

o-

CA

DAB S'tLAMIO[..

.

KO=

,t,

.

.

-.

1+1:1 . . . . . . . .

+ Z:.3 4.

.

.

@

.

~ '8

. . . . . .

pl~ .--,': ~ :aa ~ .

I~

* Z!~ 2+ \%/

/

CA /~

. . . . . . . . . . . .

,~00

~,~:.~ ?:

i'~-~_

--

500

600

WAV~ L . E N ~ i r H + N A N 0 , M E " I ' E R $

Figure 17. Schematic of zinc determination with dansylamide and apo-carbonic anhydrase.

70~

2 2 MHz

6O _ 03 l..d l.JJ o: 0 uJ c)

/~'~--'+-t

..--f'/

__/!

.../,/

50

~It

520

550

,,, 4 0 .J (,9 m <.

3O U3 03 < I o_

480 , - - - ~ , 20

/./D/o//o,~..... o//i1 x1

_

I

-lo

9

_

I. . . . . .

-9

o

I

__

-8

LOG [ Z n ] ,

-

,l,_

_

-7

I

l

-6

-5

MOLAR

Figure 18. Zinc concentration-dependent phase angles of apo-CA and dansylamide measured at 22 MHz at emission wavelengths of 440 nm (filled circles), 460 nm (open triangles), 480 nm (filled squares), 500 nm (open squares), 520 nm (filled triangles), and 550 nm (open circles) (Thompson and Patchan, 1995a). 162

Fluorescence Lifetime Biosensors AZOSULFAMIDE ( AZO ) ( ACCEPTOR )

n....~HNO,S-~N=N" "r" Y

OH NHAc

,.f

/

N

(CH2)~

/

NH C=O (2)

,

J

3~MAX= 508 nm r-

-

I

I-

C6 FLUORESCEIN CA ( DONOR ) EMISS.AT 510 nrn

~2zos,.,,.,~,.,,o~

I

/

~ \~/FLUORESCE'N

i

\Y~ ~,,,,,~,o,~

\

-

-

I

~

6OO 5OO WAVELENGTH HO" ~

" 0 " ~.'- ~"0

Figure 19. Schematic of energy transfer-based zinc sensing (Thompson and Patchan, 1995b).

Szmacinski and Lakowicz (1993) discovered in measuring pH using indicators whose emission or excitation shifted and lifetime changed that the dynamic range of the measurement could be greatly expanded by choosing the excitation or emission wavelength to favor the desired form. Thus, they found that ordinarily one can determine pH accurately perhaps one pH above and below the pKa of the indicator. By shifting the excitation or emission to favor the protonated form of the indicator, they could see a smaller proportion accurately and extend the measurement to higher pH; in some cases, the same indicator could accurately measure pH over more than five units. The same approach can be used for measuring zinc over a broad range, by shifting the emission wavelength to favor the bound or free forms of dansylamide (Figure 18). While dansylamide has the desirable features enumerated above, it nevertheless has drawbacks. Among these is its modest absorbance at short wavelengths (e = 3300 M -1 cm "1 at 330 nm), which makes it difficult to detect at low levels and inconvenient for fluorescence microscopy. By comparison, fluorescein has twenty-fold better absorbance, and is conveniently excited by the argon laser in the visible at 488 nm. The issue was finding a fluorescent aryl sulfonamide that exhibited a large change in fluorescence upon binding to the zinc in the active site; at that time, none was known, although others were developed later. However, colored (nonfluorescent) arylsulfonamides such as azosulfamide were known, and this was used in an energy transfer-based lifetime sensing approach, as depicted schematically in Figure 19. In this case, a fluorescent label is 163

Thompson , ~

"

~ HN

Ko~0.1 pM

Cuz+

Cu2L~.~ ENERG Y

Cu-CA

-~.~.RANSFER

A"SOIRPTION// ~4"--.~

/

y

[

: ', / ',

/

"

I

\

%

.,J

l

I

I

,I

!

600 700 WAVELENGTH, NANOMETERS

CY-5-CA ~,Exc=655 nm, .~,EM=670nm ( DIODE LASER)

Figure 20. Schematic of energy transfer-based copper sensing.

attached to the apo-CA polypeptide backbone and exhibits some lifetime. If the colored arylsulfonamide is present at micromolar concentration (above its KD), it will not bind appreciably to the apo-form, and consequently it will little affect the label's emission. In the presence of zinc bound in the active site, the azosulfamide binds to the protein and is thus brought into close enough proximity (< 20A) to serve as an energy transfer acceptor for the fluorescent label, thereby quenching it partially and reducing its lifetime commensurately. The virtue of this approach is that azosulfamide absorbs in the visible, permitting a variety of fluorophores to be used which otherwise would be insensitive to the presence of the zinc (Thompson and Patchan, 1995b). Site-directed mutagenesis of the carbonic anhydrase molecule to insert cysteinyl residues as attachment points for labels enables the response of the transducer to zinc binding to be controlled and predicted exactly using Ft~rster theory (Thompson et al., 1996a). An issue with all these transduction approaches is the requirement for a separate aryl sulfonamide. This is an issue because the small molecule may not have the same solubility or diffusion properties as the protein, particularly inside cells. Moreover, only zinc (and to a much lesser extent, Cd and Co) promotes binding of aryl sulfonamides to the holoprotein, and consequently ions such as Cu(II) and Ni(II) are not detected by such approaches. Thus a slightly different approach was taken, wherein the weak d-d absorbance bands observed in certain transition metal complexes of the protein are used as energy transfer acceptors. Thus, when Cu(II), Co(H) or Ni(II) bind to apo-CA, they exhibit absorbance mainly in the visible and near-IR, with extinction coefficients of the order of 100 M 1 crn1 , which is so weak as to be difficult to observe and useless analytically. However, for a fluorophore positioned closely enough and chosen for a good spectral 164

Fluorescence Lifetime Biosensors

60 -00

b9 W t~

a:po0

91

NG.7 C -OG ;MHZ.

3 0 ' ~'C

ill

O0 0 0

0

50

-

-0

0

0

0

tu .J

O0

4o

0

3O tl.

In

O

0

Z

0

Ii

0 II

-,

ifli i

0

Hi

1.0 "~"

I 0 -~;'

1 0 -=~

10 -~

I0 '6

1: M ~*] i,. M O L A R Figure 21. Phase angles of apo-N67C-Oregon Green apo-CA at 91 MHz as a function of Cu(II) (open diamonds), Ni(II)(filled squares), and Co(II) (open circles) concentration.

overlap, such metal centers serve admirably as energy transfer acceptors capable of reducing the label lifetime (Figure 20). Again, use o f the Ft~rster theory enables one to optimize the response, although if the fluorophore is close enough other quenching mechanism may come into play. This is surely the case for Cd, since it has no d-d absorbance bands. An example of the response of such a transducer is depicted in Figure 21. In this case, the label is Oregon Green conjugated at position 67 of carbonic anhydrase through an inserted cysteinyl residue; the labeled apoprotein exhibits a lifetime of approximately 3 nsec, which is reduced to 600 psec when Cu(U) is bound (data not shown), resulting in a large change in phase angle. More recently, we have employed labels excitable with visible laser diodes, for reasons that will be discussed hereunder. Also, some labels are quenched by other mechanisms such that they are quenched even by Zn and Cd, which typically are spectroscopically silent (Thompson et al., 1998). Thus ABD-T, when conjugated to the N67C position of apo-CA, exhibited quenching with all of Cu(H), Zn(II), Co(H), Cd(II), and Ni(II) (Figure 4). Finally, some labels, if placed very close to the binding sites, are quenched by other agents which are present there, but which are displaced by the metal. Thus ABD when conjugated to position 67 exhibits a nine-fold increase in intensity upon binding Zn(II), but less than two-fold upon binding Cu(II). This may be attributed to the Zn(II) ion displacing some quenching agent (such as water) upon binding, with a concomitant increase in intensity; in the case of Cu(II), the increase due to the displacement is offset by its own quenching of the label by other mechanisms. 165

Thompson

- 0,050

I

o,.o

-

3oo

-"

o,o~o T

z::)

9,2 =I.- 9.0 .,.,.,

<

8,8

i-

Z w

- O,or,5

8,6

-= 8,~

LI98:-,.F,68o

~,/-

/

290

,IJ

25m FIBEROPTIC EXC. 660 rim: ~I

sEAE"'WATER

9

o.ooi

, , ..... ~

....... ,

0.0~

- 280 o

X

......... i

0..1

Z

0.030 o

).,.

~ 0.02~

= / .6. .%, . ,o

I~ODEL / / 200 M,,=_..,o"

0~0,35 I It:

-i 2;.0 1

[ 26o

.~.._. ............. '-"

~,0

10

"~;'"'* ....-'"

tO0

0:020

<

0 o

z.

~

0.01S

z.50 .z 1/=

0,010

[ C u ~*].,PICOMOLA R

Figure 22. Phase angles (open circles), intensities (filled circles), and modulations (open squares), for immobilized apo-L198C-Alexa Fluor 660 apo-CA at 200 MHz measured through 25 meters of optical fiber.

3. S t a t e of the A r t

An example of the state of the art is the recent use of a fluorescence lifetimebased metal ion biosensor to measure changes in free Cu(ID concentration in the ocean at the picomolar level, in real time. The basic sensor configuration was as described above, with two exceptions. First, the labeled variant was L198CAlexa Fluor 660. This transducer exhibits a very substantial change in phase and modulation as a function of free copper concentration at 200 MHz, with an apparent KD of 1 picomolar. Immobilizing the transducer on the end of the fiber, as described above, provided a working transducer with similar phase response, but reduced modulation response, which we attribute to imperfect optical filtration (Figure 22). .... Working at 160 nm longer excitation and emission wavelengths reduced the attenuation of the light passing through the optical fiber approximately 80-fold. While the length of fiber used was only 25 meters, the imperceptible attenuation and strong signal made it apparent that a longer length of fiber could have been used had it been necessary. The second difference was that the excitation was provided by a laser diode whose amplitude was directly modulated, instead of externally modulating a CW laser as depicted in Figure 12. Use of the laser has several advantages, in that direct modulation is cheaper, simpler, and has less loss, in addition to the modest cost, high efficiency, and high reliability of the laser itself (Melles Griot 56 DOL series). 166

Fluorescence Lifetime Biosensors

i

310- - -

8.2

o

0 O. O

~ 8.0

300 ~

O

9.

O O

\

O

O

0

0

"

.-'-78

,

0

29,0

3 L M

O:::~

Z

~

280

7.6

a

0,030

D

0.02'5

Z O

r

w

c~ 7.1, Z W

0,020

~ 7, 2 ~ 7.0

w .O'1

\

-EO~5

O

- 0,010

0

60

120

180

2~0

300

TIMEt MINUTES

Figure 23. Phase angles (open circles), intensities (filled circles) and modulations (open squares) of sea water in Eel Pond Inlet at 200 MHz as a function of time after the beginning of ebb tide.

The site for this experiment was a copper-contaminated inlet on the Massachusetts coast. This site has been well-characterized (Moffett et al., 1997); in particular, tidal outflow from the inlet exhibits a substantial increase in free copper concentration over a period of a few hours. Thus, the inlet provided a good test of the sensor with a substantial change in copper concentration (albeit at picomolar levels) occurring over a short period in the very complex sea water matrix. The Alexa Fluor 660-L198C apo-CA-labeled tip of the fiber (following calibration with a sea water model) was immersed in the inlet shortly after the beginning of the ebb tide, and the phase, modulation, and fluorescence intensity measured at 15 minute intervals. The results of the experiment are shown in Figure 23. As expected, a decrease in intensity was observed, but the variable apparent background level made intensities useless for quantitation. The phase angles were too noisy (due to the low modulation) to be of use; however, the measured modulations confirmed the intensity trend and were quantitatively valid, indicating an increase in free Cu(II) from approximately 0.1 pM to 1 pM over about four hours. A measurement by an entirely different technique (CLEASV) of a sample taken a few hours later at the same site indicated 4 pM free Cu(II), in excellent agreement. The low modulation levels are attributed to inadequate optical filtration, the need for which is well-recognized in this optical configuration (Thompson et al., 1990). The real-time nature of the measurement and its extreme sensitivity suggest the potential capabilities of the technology. 167

Thompson

4. Advantages and Limitations Probably the only real limitations to the approach of lifetime-based biosensing are the complexity and expense of the instrumentation; perhaps with the recent introduction of less expensive commercial instruments, these factors will decline in importance. Part of the issue is the general lack of familiarity with lifetime measurements on the part of many investigators, and the perception that they are difficult to carry out and interpret. While more complex than typical molecular biology techniques, lifetime measurements are certainly no more complex than modem multinuclear NMR techniques, and in any case may be learned in a short course. In our view, the principal advantages of lifetime measurements are a combination of the virtues of lifetime-based sensing ~with those of biosensing transduction. In particular, lifetime-based sensing offers several advantages over simple intensitybased sensors, including facile calibration, insensitivity to indicator washout or bleaching, and insensitivity to excitation variation or pathlength differences. Add to this the potential for a broad dynamic range, usability with fiber optics for remote sensing, and the capability for imaging. Combine with this the virtues of biosensors, including high selectivity and sensitivity, as well as the ability to detect many different analytes, particularly using energy transfer approaches, and one has potentially a very powerful family of techniques indeed.

5. Potential for Improved Performance or Expanded Capabilities Probably the most exciting prospect for fluorescence lifetime biosensing is in imaging applications, particularly of cells and living tissues by microscopy. As discussed above, the matchless selectivity and high sensitivity of biosensors generically makes them well suited for chemical sensing in biological systems, either for research purposes or clinical diagnosis. Certainly the ability to image and quantify calcium ion fluxes in cells has revolutionized our understanding of calcium's roles in biology, which underscores the value of fluorescence microscopy using ratiometric fluorescent indicators such as Fura-2 and Indo-1. However, such indicators are available for only a handful of analytes; potentially one might wish to image fluxes and levels of many analytes, including scores of small molecules, and hundreds of macromolecules. Furthermore, ratiometric indicators remain difficult to synthesize with the desired spectral, selectivity, and sensitivity properties. By comparison, lifetime-based biosensor transducers are often easier to configure, particularly using energy transfer. Furthermore, lifetime-based fluorescence microscopes are now commercially available (vanderOrd et al., 2001), suggesting that this sensing approach will become more widespread. It should be noted that the picosecond/femtosecond pulses of the titanium sapphire lasers used for multiphoton microscopy are ideal sources for lifetime-based microscopy, and it is probably easier to do lifetime-based high 168

Fluorescence Lifetime Biosensors resolution imaging with these systems than to try to adapt a conventional confocal microscope to lifetime-based measurements. Finally, the difficulty of ratiometric measurements in a confocal or two-photon format suggests that lifetime-based sensing will become a preferred method.

6. A c k n o w l e d g m e n t s

The author would like to thank his co-investigators whose names are given in the references for their help and many fruitful discussions; the Office of Naval Research, National Science Foundation, and National Institutes of Health for support; and Mrs. Krystyna Gryczynska for the figures.

7. R e f e r e n c e s

Barnard, S. M. and D.M. Walt, 1991, Nature 353, 338. Bhatia, S. K., L.C. Shriver-Lake, K.J. Prior, J. Georger, J.M. Calvert, R. Bredehorst and F.S. Ligler, 1989, Anal. Biochem. 178,408. Birch, D. J. S. and R. E. Imhof, 1991, In Topics in Fluorescence Spectroscopy Volume 1" Techniques, Ed. J.R. Lakowicz, Plenum Press, New York, pp. 1 - 96. Birks, J. B. and D.J. Dyson, 1961, J. Scient. Instrum. 38,282. Bright, F. V., 1988, Proc. SPIE 909, 23. Carraway, E. R., J.N. Demas, B.A. DeGraff and J.R. Bacon, 1991, Anal. Chem. 63,337. Cass, A. and F.S. Ligler, Eds., 1998, Immobilized Biomolecules in Analysis: A Practical Approach, Oxford University Press, Oxford. Chen, R. F. and J. Kernohan, 1967, J. Biol. Chem. 242, 5813. Dattelbaum, J. D. and J.R. Lakowicz, 2001, Anal. Biochem. 291, 89. Demas, J. N. and G.A. Crosby, 1971, J. Am. Chem. Soc. 93, 2841. Fernandez-Gutierrez, A. and A. Munoz de la Pena, 1985, In Molecular Luminescence Spectroscopy, Part I: Methods and Applications, Vol. 77, Ed., S.G. Schulman, Wiley-Interscience,_ New York, pp. 371-546. Forster, T., 1948, Ann. Phys. 2, 55. Grattan, K. T. V. and Z.Y. Zhang, 1994, In Topics in Fluorescence Spectroscopy Volume 4: Probe Design and Chemical Sensing, Ed., J.R. Lakowicz, Plenum, New York, pp. 335-376. Harms, P. and G. Rao, 1999, Proc. SPIE 2602, 52. Huang, C.-C., C.A. Lesburg, L.L. Kiefer, C.A. Fierke and D.W. Christianson, 1996, Biochem. 35, 3439. Hunt, J. A. and C.A. Fierke, 1997, J. Biol. Chem. 272, 20364. Kawski, A., I. Gryczynski, K. Nowaczyk, P. Bojarski and J. Lichacz, 1991, Z. Naturforsch. 46a, 1043.

169

Thompson Lakowicz, J. R., 1999 Principles of Fluorescence Spectroscopy, Kluwer Academic / Plenum Publishers, New York. Lakowicz, J. R. and B.P. Maliwal, 1993, Anal. Chim. Acta 272, 155. Levy, R., E.F. Guignon, S. Cobane, E. St. Louis and S. Fernandez, 1997, Proc. SPIE 2980, 81. Li, Q. Z. and A.E.G. Cass, 1991, Biosens. Bioelectron. 6, 445. Lindskog, S. and P.O. Nyman, 1964, Biochim. Biophys. Acta 85,462. Lippitsch, M. E., J. Pusterhofer, M.J.P. Leiner and O.S. Wolfbeis, 1988, Anal. Chim. Acta 205, 1. McCall, K. A., 2000, In Department of Biochemistry, Duke University, Durham, NC, 190 pp. Moffett, J. W., L.E. Brand, P.L. Croot and K.A. Barbeau, 1997, Limnol. Oceanogr. 42, 789. Ozinskas, A., H. Malak, J. Joshi, H. Szmacinski, J. Britz, R.B. Thompson, P. Koen and J.R. Lakowicz, 1993, Anal. Biochem. 213,264. Peterson, J. I., S.R. Goldstein, R.V. Fitzgerald and D.K. Buckhold, 1980, Anal. Chem. 52, 864. Rehm, D. and A. Weller, 1970, Isr. J. Chem. 8,259. Slavik, J., 1982, Biochim. Biophys. Acta 694, 1. Spencer, R. D. and G. Weber, 1969, Ann. NY Acad. Sci. 158, 361. Straume, M., S.G. Frasier-Cadoret and M.L. Johnson, 1991, In Topics in Fluorescence Spectroscopy, Volume 2: Principles, Ed., J.R. Lakowicz, Plenum, New York, pp. 177 - 240. Szmacinski, H. and J.R. Lakowicz, 1993, Anal. Chem. 65, 1668. Szmacinski, H. and J.R. Lakowicz, 1994, In Topics in Fluorescence Spectroscopy Vol. 4: Probe Design and Chemical Sensing, Ed., J.R. Lakowicz, Plenum, New York, pp. 295 - 334. Thompson, R. B., 1991, In Topics in Fluorescence Spectroscopy Vol. 2: Principles, Ed., J.R. Lakowicz, Plenum Press, New York, pp. 345-365. Thompson, R. B., J.K. Frisoli and J.R. Lakowicz, 1992, Anal. Chem. 64, 2075. Thompson, R. B., Z. Ge, M.W. Patchan and C.A. Fierke, 1996a, J. Biomed. Opt. 1,131. Thompson, R. B., Z. Ge, M.W. Patchan, C.-C. Huang and C.A. Fierke, 1996b, Biosens. Bioelectron. 11,557. Thompson, R. B. and E. Gratton, 1988, Anal. Chem. 60, 670. Thompson, R. B. and E.R. Jones, 1993, Anal. Chem. 65.730. Thompson, R. B. and J.R. Lakowicz, 1993, Anal. Chem. 65, 853. Thompson, R. B., M. Levine and L. Kondracki, 1990, Appl. Spectrosc. 44, 117. Thompson, R. B., B.P. Maliwal, V.L. Feliccia, C.A. Fierke and K. McCall, 1998, Anal. Chem. 70, 4717. Thompson, R. B., B.P. Maliwal and C.A. Fierke, 1999, Anal. Biochem. 267, 185. Thompson, R. B. and M.W. Patchan, 1995a, J. Fluoresc. 5, 123. Thompson, R. B. and M.W. Patchan, 1995b, Anal. Biochem. 227, 123. Thompson, R. B. and D.R. Walt, 1994, Nav. Res. Rev. 46, 19.

170

Fluorescence Lifetime Biosensors Thompson, R. B., H.H. Zeng, M. Loetz and C. Fierke, 2000, Proc. SPIE 3913, 120. Tolosa, L., I. Gryczynski, L.R. Eichhorn, J.D. Dattelbaum, F.N. Castellano, G. Rao and J.R. Lakowicz, 1999, Anal. B iochem. 267, 114. Ullman, E. F. and M. Schwarzberg, 1981, Syva Company, U.S. vanderOrd, C. J. R., C.J. deGrauw and H.C. Gerritsen, 2001, Proc. SPIE 4252, 119. White, C. E. and Argauer, R. J., 1970, Fluorescence Analysis: A Practical Approach, Marcel Dekker, Inc., New York.

171

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 6

ELE CTRO CHEMILUMINE S CEN CE

MARK M. RICHTER, PH.D.

Department of Chemistry, Southwest Missouri State University Springfield, MO 65804-0089 USA Electrochemiluminescence (ECL) is the process where species generated at electrodes undergo electron transfer reactions to form excited states that emit light. Application of a voltage to an electrode in the presence of an ECL luminophore such a s Ru(bpy)3 2+ (where bpy = 2,2 '-bipyridine) results in light emission and allows detection of the emitter at very low concentrations (_<1011 M). By employing ECL-active species as labels on biological molecules, ECL has found application in commercial instruments for immunoassays and DNA analyses. Commercial systems have been developed that use ECL to detect many clinically relevant analytes with high sensitivity and selectivity. The principles, history, applications, advantages, limitations and possibilities for improving the performance of this technology are outlined in this chapter.

1. Principles of Operation A wide variety of methods exist for the detection of chemical and biological analytes of interest. One of the most versatile, and one that has been commercially developed for the clinical diagnostic market, is Electrochemiluminescence (also called Electrogenerated Chemiluminescence and abbreviated ECL). ECL is a means of converting electrical energy into light (radiative energy). It involves the production of reactive intermediates from stable precursors at the surface of an electrode. These intermediates then react under a variety of conditions to form excited states that emit light. It is important to distinguish ECL from chemiluminescence (CL). Both ECL and CL involve the production of light by species that can undergo highly energetic electron transfer reactions. However, luminescence in CL is initiated by the mixing of reagents and controlled by careful manipulation of fluid flows. In ECL, luminescence is initiated and controlled by switching an electrode voltage. 173

Richter In little more than three decades, ECL has moved from being a "laboratory curiosity" to a useful analytical technique. The first ECL reactions were investigated in the early 1960s on systems composed of polyaromatic organic compounds in highly purified and deaerated nonaqueous solvents such as acetonitrile (CH3CN) and dimethylformamide (DMF). This early work was followed by studies on luminescent inorganic species, mainly Ru(bpy)32+ (Tokel and Bard, 1972). Much of this early work has been reviewed (Faulkner and Bard, 1977; Faulkner and Glass, 1982). Adding certain species, called coreactants, made it possible to observe ECL in aqueous solution (Rubinstein and Bard, 1981). These developments have resulted in a wide range of analytical applications for ECL (Knight and Greenway, 1994), including commercial applications in clinical diagnostic assays (Bard and Whitesides, 1993; Blackburn et al., 1991). 1.1. Annihilation ECL Traditionally, ECL was generated via annihilation, where the electron transfer reaction is between an oxidized and reduced species, both of which are generated at an electrode by alternate pulsing of the electrode potential. A

+

e

---->

A~

(reduction at electrode)

(1)

A

-

e

--->

A "+

(oxidation at electrode)

(2)

A'

+

A .+

A*

~

A

A*+ A +

hv

(excited state formation) (light emission)

(3) (4)

where hv is a photon of light. For example, the potential of the working electrode is quickly changed between two different values in order to generate the reduced, A ~ and oxidized, A .+, species (Equations 1 and 2, respectively) that will react near the electrode surface to form the emissive state, A* (Equation 3). A typical reaction involves 9,10-diphenylanthracene (DPA, Figure 1). ECL is generated when a double potential step is applied to an electrode (typically platinum), producing the radical cation (DPA "+) upon anodic oxidation and the radical anion (DPA ~ upon cathodic reduction. The resulting electro-generated products can then react and undergo annihilation (i.e., Reaction 3) to produce an excited state (DPA*) that is then able to emit light. DPA - e ---->DPA "+

(5)

DPA + e- ---~D P A '

(6)

174

Electrochemiluminescenee DPA ~ + DPA ~ ---) DPA + DPA*

(7)

DPA* --~ DPA + hv

(8)

For DPA the emission maximum ()~max) OCCURS at about 512 nm and the ECL spectrum (i.e., a plot of ECL emission versus wavelength) is identical to DPA photoluminescence. This indicates that the ultimate product of charge transfer, and hence luminescence, is the lowest singlet DPA species, 1DPA*. There is considerable evidence that the reactions outlined in Equations (7) and (8) are oversimplifications, and that several mechanistic steps intervene between the electron transfer and photon emission steps. The identification of these mechanisms has been a focus of past work in the field and are thoroughly reviewed elsewhere (Faulkner and Bard, 1977; Faulkner and Glass, 1982). DPA, and several other polyaromatic hydrocarbons, were among the first complexes studied using ECL (Hercules, 1964; Chandross and Visco, 1964; Santhanam and Bard, 1965; Bader and Kuwana, 1965) since they were known to undergo chemically and electrochemically reversible one electron oxidation and reduction at easily attainable potentials (e.g., < 2V vs SHE where SHE = standard hydrogen electrode) and display photoluminescence efficiencies (~m = number of photons emitted per photons absorbed by the compound) of near unity. These qualities, as well as the stability of the oxidized and reduced forms of the complexes (i.e., the radical anions and cations) and their ability to undergo electron transfer reactions, are still used as criteria to determine whether a compound shows promise as an ECL luminophore. DPA is an example of an "energy sufficient" system. That is, the enthalpy, AH~ of the electron transfer reaction in Equation (3) is larger than the energy required to produce the excited singlet state from the ground state (Equation 4). In DPA, one of the products of the reaction is therefore produced with excess energy that can be emitted as light (e.g., 1DPA*). When the luminescence is err&ted by a species in an excited singlet state this process is known as the singlet or S-route. Since the free energy (AG~) available for exciting the product is the energy available from the redox reaction producing the ground-state products (i.e., Equations 1 and 2), AH~ can be calculated from the reversible standard potentials of the redox couples: AH~ = AG~ + TAS ~ = (E~

-- E~

-

0. leV

where E~ and E~ represent the cyclic voltammetric peak potentials for the oxidative and reductive redox couples, respectively. For example, in DPA:

175

Richter

DPA

rubrene

3

(H3C)2N'-~.... ~N(CHa)2 ,3

thianthrene

TMPD

O

I

NH2

bpy

0

luminol Figure 1. Structures of compounds. 176

Electrochemiluminescence DPA + e---~ D P A '

E~

= -1.89 V vs SCE

(9)

D P A - e - - ) DPA "+

E~

= +1.35 V VS SCE

(10)

A+/Dr,A and E~ are the cyclic voltammetric peak potentials for the oxidation and reduction of diphenylanthracene, respectively and SCE is the saturated calomel reference electrode (Faulkner and Bard, 1977). This results in an enthalpy for the electron transfer reaction of 3.14 eV. When we compare this to the energy of the emitted light in DPA obtained spectroscopically ('~,m~x = 512 nm or 3.00 eV) we see that the emitting state is accessible and XDPA* may be populated directly in the reaction.

E~

Another example of an energy sufficient system is the inorganic species Ru(bpy)32+ (where bpy = 2,2'-bipyridine, Figure 1), whose ECL was first reported in 1972 (Tokel and Bard, 1972). Ru(bpy)3 z+ - e ---> Ru(bpy)33+

E~

+ 1.2 V vs SCE

(11)

Ru(bpy)3 z+ + e ----> Ru(bpy)3 +

E Ru2+/l+=-1.4 V vs SCE

(12)

o

Ru(bpy)33+ + Ru(bpy)3+--+ Ru(bpy)32+* + Ru(bpy)32+

(13)

R u ( b p y ) 3 2+* "---) R u ( b p y ) 3 2+ + hv (-2.1 eV, 610 nm)

(14)

The excited state formed in the ECL reaction is the same as that formed during photoexcitation. In photoexcitation (i.e., photoluminescence), an electron is excited from metal-based drc orbitals to ligand-based zr* orbitals (a metal-toligand charge transfer (MLCT) transition). The excited electron then undergoes intersystem crossing to the lowest triplet state of Ru(bpy)32+* from where emission occurs (Demas and Crosby, 1968; Van Houton and Watts, 1975). Since the photoluminescent and ECL spectra are nearly identical, the emission process in ECL involves the MLCT state of Ru(bpy)3 ~+. This state may be formed if an electron is transferred to the n* orbital of one of the bipyridine ligands. Ru(bpy)32+* can then decay to the ground state, producing the same luminescence as obtained from photoluminescence spectroscopy (Tokel and Bard, 1972). The reaction of Equation (14) has about 2.6 V of free energy to place into an excited product. Since the charge-transfer triplet lies at 2.1 eV, it is readily accessible to the homogenous redox process (Luttmer and Bard, 1981; Glass and Faulkner, 1981). No other excited states of appreciable lifetime can be populated so the opportunities for unusual kinetics and alternate mechanistic pathways in the Ru(bpy)32+ system are reduced. Ru(bpy)32§ is perhaps the most thoroughly studied ECL active molecule (TokelTakvoryan et al., 1973; Wallace and Bard, 1979; Itoh and Honda, 1979; Luttmer 177

Richter and Bard, 1981; Glass and Faulkner, 1981). This is for a number of reasons, including its strong luminescence and solubility in both aqueous and nonaqueous media at room temperature, and its ability to undergo reversible one-electron transfer reactions at easily attainable potentials, leading to stable reduced and oxidized species. However, the ECL of several other inorganic compounds have been reported (Hemingway et al., 1975; Faulkner and Glass, 1982; Knight and Greenway, 1994; Richter and Bard, 1996; Richter et al., 1998). This is not surprising, since many inorganic compounds display the electrochemical and spectroscopic qualities required of ECL luminophores. Furthermore, the overall ECL efficiency (photons produced per redox event) is a product of the photoluminescence quantum yield and the efficiency of production of the excited state. Ru(bpy)3 z+ has a photoluminescent quantum efficiency (~m) of 0.0682 (Caspar and Meyer, 1983; Van Houten and Watts, 1976) and an ECL efficiency (r of 0.0500 (Rubinstein and Bard, 1981; White and Bard, 1982). Under certain conditions, the reaction of Equation (13) produces the emitting chargetransfer triplet with an efficiency approaching 100% (Wallace and Bard, 1979; Itoh and Honda, 1979) and is comparable with photoluminescence data (Demas and Crosby, 1971; Meyer, 1978) showing that about 5% of the excited states produce luminescence. This demonstrates that photoluminescence and electrochemical data can be used to predict compounds that may show promise as ECL light emitters. It is also possible to use two different precursors to generate an emissive state. A

+

e

---->

A~

(15)

D

-

e-

--~

D .§

(16)

A'

+

D .§

----)

m~+ D (or A + D*)

(17)

---)

A (or D) + hv

A* (or D*)

(18)

A classic example is the electron transfer reaction and subsequent chemiluminescence between the anion radical of DPA and the cation radical of N,N,N',N'-tetramethyl-p-phenylenediamine (Wurster's Blue (TMPD), Figure 1) in dimethylformamide. DPA + e --->DPA ~

E~

TMPD - e ~ TMPD "+

E~

= -1.89 V vs SCE = +0.24 V vs SCE

(19) (20)

D P A + TMPD + --+ 1DPA* + TMPD

(21)

~DPA* ~ DPA + hv

(22) 178

Electrochemiluminescence where tDPA* is the excited singlet state whose excess energy is emitted as light. In theory, light may be emitted from either 1DPA* or TMPD (e.g., either A* or D* in Equation 17). However, the emitted light generated via ECL is identical to DPA photoluminescence, indicating that 1DPA* is the ultimate product of charge transfer. This is not surprising, since the energy states of TMPD are inaccessible and hence TMPD cannot luminesce (Faulkner and Bard, 1977). The fact that the DPA/TMPD system undergoes ECL at all is surprising since the enthalpy for the electron transfer reaction is 2.03 eV, much less than that required to reach the emitting singlet excited state for DPA of 3.00 eV. Therefore, the ultimate emitter is not directly populated. Such systems are called "energy deficient", and a more complicated scheme than that depicted in Equations 21 - 22 is required to generate the emitter. The most commonly accepted explanation involves triplet intermediates, the so-called triplet or T-route. A'- + D'+ ___.>3A* + D

(23)

3A* + 3A* ---) 1A* + A (triplet-triplet annihilation)

(24)

Reaction (24) is generally called "triplet-triplet annihilation' (Faulkner and Bard, 1977) where the energy from two-electron transfer reactions is pooled to provide sufficient energy to form the singlet excited state (Parker, 1968; Birks, 1970). For the DPA/TMPD system: TMPD ~ + DPA ~ --~ TMPD + 3DPA*

(25)

3DPA* + 3DPA*---~ DPA + ~DPA*

(26)

Many ECL reactions with different precursors follow this route and several examples are given in the literature (Faulkner and Bard, 1977; Knight and Greenway, 1994). The T-route may also operate in energy-sufficient systems. Although analytical uses for ECL are possible with annihilation systems (e.g., display devices), most annihilation systems require the use of rigorously purified and deoxygenated nonaqueous solvents since the available potential range in water is too narrow to generate the required energetic precursors. In essence, the stability range for the electrochemical oxidation and reduction of water is too small to conveniently generate both species (i.e., the radical anion and cation) needed for annihilation ECL. For example, Ru(bpy)32§ is oxidized at a Pt electrode to form Ru(bpy)33§ at about 1.2 V vs SCE. Ru(bpy) 32+ is reduced at a Pt electrode in aqueous solution to form Ru(bpy)3§ at about -1.4 V, a potential not easily attainable at Pt electrodes in aqueous solution without the evolution of large amounts of hydrogen gas. The products of water oxidation and/or reduction interfere with the annihilation reaction (Equation 3) such that little to no light is observed. Background ECL observed in such systems has been 179

Richter attributed to a light-emitting reaction between Ru(bpy)33§ and O H ions, rather than the annihilation of the + 1 and +3 ruthenium complex (Rubinstein and Bard, 1981). The ECL systems shown above generate oxidized and reduced species at a single electrode. It is also possible to obtain emission at two different electrodes that are sufficiently close to allow the electro-generated reactants to interdiffuse and undergo annihilation (e.g., Equation 3) (Maloy et al., 1971; Brilmyer and Bard, 1980). For example, a rotating ring disk electrode (RRDE) can be employed. One reactant, such as A ~ can be generated at the central disk and the other, A ~247 generated at the ring. These are then swept together by diffusion and convection (Bard and Faulkner, 1980), resulting in a ring of light on the inner edge of the ring electrode (Faulkner and Bard, 1977; Maloy and Bard, 1971). Other experiments have employed dual-working electrode systems (Brilmeyer and Bard, 1980; Bartelt et al., 1992; Fiaccabrino et al., 1998a,b) with interdigitated electrodes, thin-layer geometry or flowing streams to move the reactants together. 1.2. Coreactant ECL ECL can also be generated in a single potential step utilizing a coreactant. A coreactant is a species that, upon oxidation or reduction, produces an intermediate that can react with an ECL luminophore to produce excited states. Usually, this occurs upon bond cleavage of the coreactant to form strong oxidants or reducants. For example, the oxalate ion (C2042) is believed to produce the strong reductant CO2~ upon oxidation in aqueous solution (Rubinstein and Bard, 1981): - e --~ [C204 ~

C2042

~

C 0 2 ~ q- C 0 2

(27)

The oxidizing potential that leads to CO2~ may also oxidize an ECL luminophore (e.g., D where D is, for example, Ru(bpy)32§ D

-

e"

__~

D .§

(28)

D ~247 and CO2~ may then react to produce an excited state capable of emitting light.

180

Electrochemiluminescence

H+

TPrA'+-~ TPrA"

q~A

9

r

~D

Ru(bpy) e~,, Ru(bpy)

L._

2+*

3

hv (620 nm) Figure 2. Proposed mechanism for Ru(bpy)32+/TPrA ECL system.

CO2" + D "+ --+ D* + CO2

(29)

D*

(30)

-->

D + hv

Oxalate is often referred to as an "oxidative-reductive" coreactant due to its ability to form a strong reducing agent upon electrochemical oxidation. Unlike annihilation schemes where a double potential step (e.g., oxidation followed by reduction) is required to generate the highly energetic precursors, in coreactant ECL the electrode typically only oxidizes or reduces the reagents in a single 181

Richter potential step. For example, in the oxalate system the electrode oxidizes both the oxalate and the ECL reactant D; the reducant, CO2"- is then generated upon bond cleavage of oxalate via Equation (27). This strategy is used in most analytical and biotechnology applications, with the reactant D being Ru(bpy)32+. This methodology has allowed the generation of ECL in aqueous solution, a great advantage in terms of analytical applications. Without this ability, it is doubtful whether ECL would have moved beyond the laboratory phase. Another example of an "oxidative-reductive" system is the commercially important Ru(bpy)32+/TPrA system (TPrA = tri-n-propylamine). As with the oxalate system, this involves the production of a strong reductant (presumably TPrA ~ by an initial oxidation sequence (Leland and Powell, 1991; McCord and Bard, 1991). Ru(bpy)32+- e

~

TPrA - e ---) [TPrA'] +

Ru(bpy)33+

(31)

--~ T P r A ' + H §

Ru(bpy)33+ + TPrA" ---) Ru(bpy)32+* +

products

Ru(bpy)32+" ---) Ru(bpy)32+ + hv

(32) (33) (34)

ECL is produced upon concomitant oxidation of Ru(bpy)32§ and TPrA (Figure 2). Electrochemical studies of various aliphatic amines have indicated a possible reaction pathway for the oxidation of TPrA (Smith and Mann, 1969). Upon oxidation, the short lived TPrA radical cation (TPrA ~ is believed to lose a proton from an c~-carbon to form the strongly reducing intermediate TPrA ~ This radical can then reduce Ru(bpy)33§ to Ru(bpy)32+*. Other reaction mechanisms for production of the excited state have also been proposed. For example, reduction of Ru(bpy)32§ to Ru(bpy)31§ by TPrA', followed by annihilation: Ru(bpy)31+ + Ru(bpy)33+

~ Ru(bpy)32+ + Ru(bpy)32+*

(35)

Although the details of the coreactant ECL mechanism (Equations 31 - 34) to generate light emission are still under study (Zu and Bard, 2000; Kanoufi et al., 2001) the origin of the light emission from Ru(bpy)32+ has been well documented (Glass and Faulkner, 1981; Faulkner and Glass, 1982; Leland and Powell, 1991; McCord and Bard, 1991). Since the photoluminescent and ECL spectra are nearly identical, the emission process in ECL involves the MLCT state of Ru(bpy)32+. This state may be formed if the reducing agent (i.e., TPrA ~ transfers an electron to the rt* orbital of one of the bipyridine ligands. Ru(bpy)32+* can then decay to the ground state, producing the same luminescence as obtained from photoluminescence spectroscopy. Solution phase co-reactant ECL using TPrA

182

Electrochemiluminescence and Ru(bpy)32§ is quite sensitive, with sub-picomolar detection limits achieved (Leland and Powell, 1991; Blackburn et al., 1991). Other systems use coreactants that are reduced to generate reactive species (i.e., "reductive-oxidative" coreactants). For example, in the case of peroxydisulfate (SZO82-), reduction produces the strong oxidant SO4~ that then undergoes an electron-transfer reaction with an ECL luminophore like Ru(bpy)3 z+ to generate light (White and Bard, 1982; Bolletta et al., 1981) as shown below. Ru(bpy)32+ + e ----~Ru(bpy)3 +

(36)

SzO82- + e ----~ SO4" + SO4z

(37)

Ru(bpy)3 + + SO4" ---~ Ru(bpy)32+* + SO4z-

(38)

As well of being of practical interest, ECL reactions of this type also demonstrate the intermediacy of species such as TPrA ~ and CO2". However, the mechanisms of these reactions are still not well understood.

2. History The first detailed studies on ECL were begun in the mid-1960s but interest in light emitted during electrolysis was generated much earlier. In 1927 Dufford and co-workers (Dufford et al., 1927) observed emission at an anode by applying between 500 and 1500 V to a cathode in a solution of Grignard compounds in anhydrous ether. The reaction conditions in these experiments were not very well defined and it is doubtful whether this process was actually ECL. At such high potentials, electroluminescence (the direct injection and removal of charge with the formation of an electron-hole pair), or electrode processes were probably responsible for the observed ligh t emission. However, this initial report was followed two years later when Harvey published experiments on luminol (2,3aminophthalhydrazide) (Figure 1) in aqueous/alkaline solution (Harvey, 1929). The potentials used to generate anodic light emission were much lower than those used by Dufford and coworkers (2.8 V vs 500-1500 V, respectively). Several groups followed up on this (Bemanose et al., 1947; Vojir, 1954; Kuwana et al., 1963), and in fact the luminol system continues to generate interest (Vitt et al., 1991; Haapakka, 1982; Haapakka and Kankare, 1980; van Dyke and Cheng, 1989). Practical applications of the luminol system are hampered by a number of factors, most notably the high non-specific ECL (i.e., background) - possibly due to the formation of oxygen at the anode in aqueous solution, followed by chemiluminescent reactions involving o x y g e n - and the extremely basic conditions (pH > 11) that are needed to generate sufficient light emission.

183

Richter In the mid 1960s several research groups (Hercules, 1964; Chandross and Visco, 1964; Santhanam and Bard, 1965; Bader and Kuwana, 1965) decided to study in detail the luminescence generated during electrolysis of polyaromatic hydrocarbons (e.g., anthracene, diphenylanthracene, thianthrene, rubrene) in aprotic media, Figure 1. In essence, they wished to see if excited states could be generated electrochemically as well as photochemically. It was observed, both visually and spectroscopically, that the radiation emitted by sweeping to both negative and positive potentials (annihilation pathway, Equations 1 - 4 ) was often identical to that generated during photoluminescence, indicating formation of the excited singlet state. Throughout the sixties and seventies work continued on the polyaromatic hydrocarbons, and was eventually extended to other systems, most notable among them the ruthenium chelates. Since the discovery that Ru(bpy)32§ is photoluminescent (Paris and Brandt, 1959), a large body of literature has appeared aimed at understanding both the ground and excited state properties of Ru(bpy)32§ Os(bpy)32§ and their polyazine derivatives (Demas and Crosby, 1971; Sutin and Creutz, 1978; Meyer, 1978; Barigelletti et al, 1991; Roundhill, 1994). Therefore, it is not surprising that these compounds have also played an important role in the development of ECL. The first report of ECL in a metal chelate was in 1972 (Tokel and Bard, 1972), in which the excited state of Ru(bpy)32§ was generated in aprotic media by annihilation of the reduced, Ru(bpy)3 l§ and oxidized, Ru(bpy)33§ species (Equations 11 - 14). The original coreactant, and thus the first report of ECL in aqueous solution, was oxalate ion (C2042; equations 27 - 29) (Rubinstein and Bard, 1981). Subsequently, other species were shown to act as coreactants, among them peroxydisulfate ($208~; Equations 3 6 - 38) and tri-n-propylamine (TPrA; Equations 31 - 34). The discovery of TPrA (Leland and Powell, 1991) allowed efficient ECL not only in aqueous media, but also at physiological pH. Following the first report on TPrA, other species containing amine groups were proposed, among them many biologically important analytes (e.g., alkylamines, NADH, antibiotics, L,D-tryptophan, glucose, erthromycin, valine, HIV- gag gene). The list is quite extensive, and compilations up to 1998 have been published (Knight and Greenway, 1994; Knight, 1999). To date, ECL has found use in studying the properties of both organic and inorganic systems (Faulkner and Bard, 1977; Knight and Greenway, 1994). These include polyaromatic hydrocarbons (Faulkner and Bard, 1977; Faulkner and Glass, 1982; Richards and Bard, 1995), exciplexes (Hemingway, Park and

184

Electrochemiluminescence ......

2+

O

~~

Ru

X Figure 3. Ru(bpy)32§ NHS ester for ECL labeling of proteins and nucleic acids.

Bard, 1975; Prieto et al., 2001), polymer assemblies (Rubinstein et al, 1983; Downey and Niemann, 1992; Richter et al., 1994), transition metal complexes incorporating such metals as Ru, Os and Pt (Faulkner and Glass, 1982; Knight and Greenway, 1994; Tokel and Bard, 1972; Vogler and Kunkeley, 1984; Kim et al., 1985; Richter et al., 1998) as well as rare earth chelates (Hemingway et al., 1975; Richter and Bard, 1996), to name a few. Ru(bpy)32§ is perhaps the most thoroughly studied ECL active molecule (Tokel-Takvoryan et al., 1973; Wallace and Bard, 1979; Itoh and Honda, 1979; Luttmer and Bard, 1981; Glass and Faulkner, 1981) and, as with other ECL systems, there was particular emphasis on characterizing the nature of the excited state, discerning the mechanisms by which these states were formed and determining the efficiency of excited state formation. Various techniques were used and are still being used, including detailed electrochemical studies, spectroscopic and spin-resonancemeasurements as well as magnetic field effects (Faulkner and Bard, 1977, 2001; Faulkner and Glass, 1982). In the early 1980s Bard and Whitesides (Bard and Whitesides, 1993; Bard and Whitesides, 1994) developed a method for the binding of Ru(bpy)32§ to biological molecules of interest (e.g., antibodies, proteins, nucleic acids). The interest in using Ru(bpy)32§ stems from its rather unique properties. Namely, it emits and is soluble at room temperature in aqueous, fluid, solution, and undergoes reversible one-electron transfer reactions at easily attainable potentials. Also, the ligands provide synthetic versatility. For example, by attaching N-hydroxysuccinimide (NHS) ester to one of the bipyridine ligands (Figure 3), the ECL label can bind to 185

Richter substances containing free amino groups. The amino acid will attack the carboxylate ester, leading to displacement of N-hydroxysuccinimide. IGEN International, Inc. began developing ECL for use in biosensor analyses in the early 1980s. In the early 1990s, the prototype ORIGEN | Analyzer was introduced by IGEN International, Inc. The ORIGEN is an ECL-based immunoassay system, incorporating Ru(bpy)32+-tagged antibodies and is engineered to be a biomedical research tool for immunoassays and DNA probes. The ORIGEN instrument is semi-automated, and incorporates a flow injection system to allow rapid and reproducible determinations of single samples. The detector is a photomultiplier tube positioned above the working electrode, and light from the electrode is recorded and integrated for each measurement. Roche Diagnostics licensed the ECL-Ru(bpy)32§ technology from IGEN International, Inc. in 1992, and subsequently produced the first fully automated instruments (Elecsys | 1010 and 2010), built upon the same flow cell design as its predecessor. The Etecsys systems were launched in Europe in 1996 for use in clinical and reference laboratories (those handling large volumes of samples), and in the U.S. starting in January of 1997 following FDA approval. Assays that have been developed for these systems include alpha-fetoprotein, digoxin, thyrotopin, protein and steroidal hormones, cytokines, and various antibodies, to name a few.

3. State of the Art 3.1. Analytical applications of ECL Coreactant ECL has been used in a wide range of analytical applications (Knight and Greenway, 1994; Bard et al., 2000). Since ECL emission intensity is usually proportional to the concentration of the emitter (Cruser and Bard, 1967) or coreactant (Leland and Powell, 1991), ECL can be used in the analysis of various species. For example, the system of interest is introduced into an electrochemical cell, a voltage is applied to an electrode and the light intensity and/or ECL spectrum is measured. ECL in such systems is very sensitive since photoncounting methods can be used to measure very low light levels. For example, ECL from Ru(bpy)32§ has been used to measure the concentrations of coreactants such as oxalate and peroxydisulfate to levels as low as 10-13 M (M = mol L 1) (Ege et al., 1984). In fact, the ability of Ru(bpy)32+ to undergo "oxidativereductive" ECL in the presence of coreactants has led to the selective determination of oxalate in synthetic urine samples (Rubinstein et al., 1983), and Ru(bpz)32+ (where bpz = bipyrazine) has been used for the determination of peroxydisulfate with nanomolar (nM) detection limits (Yamashita et al., 1991). Since the intensity of ECL is a function of both the coreactant and the emitter, ECL can be used to analyze for both. In these examples, ECL was measured in the presence of high, pre-determined concentrations of ECL emitters. These 186

Electrochemiluminescence types of experiments can then be used as a means to assay for compounds that act as coreactants including a variety of amines (Noffsinger and Danielson, 1987; Knight and Greenway, 1994; Bard et al., 2000). ECL assays for amines find many applications since amine groups are prevalent in numerous biologically and pharmacologically important compounds including alkyl-amines, antibiotics, antihistamines, opiates, nicotinamide, and the reduced form of NADH (i.e., adenine dinucleotide) (Danielson et al., 1989; Knight and Greenway, 1994; Knight and Greenway, 1996). In general, these compounds contain no chromophore and therefore cannot undergo luminescence unless an ECL-active compound such as Ru(bpy)32§ is present. As a general rule, the ECL signal from alkylamine coreactants follows the order: 3~176 1~ (Leland and Powell, 1991). Primary amines have been detected using Ru(bpy)3 2§ coreactant ECL after prior derivatization with divinylsulfone (CH2-CH-SO2-CH=CH2). The primary amines undergo a cycloaddition reaction resulting in the formation of acyclic tertiary amines (Uchikura et al., 1993) that then act as efficient coreactants. Other examples of cyclic amines that undergo ECL include nictone, atropine and sparteine (Uchikura and Kirisawa, 1991). It is also possible to quantitatively measure amino acids, peptides and proteins such as proline and valine. In fact, detection limits of 20 pM for proline (He et al., 1990) and 30 pM for valine (Brune and Bobbitt, 1991) using flow injection techniques have been achieved. Although the ability of numerous amines to act as coreactants makes ECL a versatile technique for their detection it also makes selectivity for the presence of a specific amine problematic. More recently, Xu and Dong have obtained high selectivity for the measurement of chlorpromazine, a commonly prescribed dopamine inhibitor. Using Ru(bpy)32+ as the ECL luminophore and chlorpromazine as an oxidative-reductive coreactant, selectivity was achieved by preconcentration of the chlorpromazine at a lauric acid-modified carbon paste electrode with a detection limit of 3.1 x 10.9 M (Xu and Dong, 2000). ECL has also been used to monitor enzymatic reactions. In such systems, the reaction is often coupled to the generation or consumption of an ECL coreactant. A good example is the coenzyme nicotinamide adenine dinucleotide (NADH). NADH contains an amine moiety that acts as a coreactant for Ru(bpy)32§ However, the oxidized form (NAD+) is not a coreactant (Downey and Nieman, 1992). Since numerous NADH-dependent enzymes are known, this allows for the detection of a variety of analytes including glucose (Jameison et al., 1996). Another application of ECL is the detection of ~-lactamase activity (Liang et al., 1996). Pencillin and its derivatives do not act as coreactants with Ru(bpy)3 z§ to produce ECL. However, 13-1actamase catalyzed hydrolysis of pencillin forms a molecule with a secondary amine that can act as a coreactant. The efficiency of the ECL process has been increased by covalent attachment of a ~-lactamase substrate to a Ru(bpy)32§ derivative (Liang et al., 1996). The ECL of aminopeptidase and esterase cleavage products have also been reported by covalently attaching such species as ligands to bis(bipyridine)ruthenium (II). 187

Richter (bpy)zRu 2+ has little to no intrinsic ECL, but attachment of a third ligand leads to enhanced ECL (Dong and Martin, 1996). Similar methods have been used in detector cells for high-performance-liquidchromatography (HPLC). These also involve the ECL of Ru(bpy)3 z+ for the detection of species that act as coreactants, such as amino acids, amines, and NADH (Jackson and Bobbitt, 1994). One technique that has acheived picomole detection limits uses post-column ECL detection. A solution of Ru(bpy)32+ is steadily injected into the solution stream containing separated species coming from the HPLC column. The mixed stream flows into an electrochemical flow cell where the ECL reaction occurs and emission can be measured (Holeman and Danielson, 1994). Ru(bpy)32+ can also be immobilized in a thin film of polymer (e.g., Nation) deposited on the working electrode (Rubinstein and Bard, 1980). This eliminates the need for a constant stream of Ru(bpy)32+. In this technique, ECL results when a species that can act as a coreactant is in the solution coming from the HPLC column and reacts with the immobilized Ru(bpy)32+ in the detector cell (Downey and Nieman, 1992). ECL in flowing streams has also provided information about the hydrodynamics in the detector cell (Schultz et al., 1996). In the methods described above ECL was measured in the presence of high, predetermined concentrations of ECL emitters. ECL can also be used to analyze an emitting species (eg., Ru(bpy)32+) that often serves as a label on a molecule of interest. The coreactant, typically TPrA, is present in high concentrations so the amount of luminescence depends on the concentration of the ECL emitter present in the assay. Since the emitters are bound to the analyte of interest, the amount of luminescence can be correlated with the concentration of the analyte. The most frequently used ECL-active label is Ru(bpy)32+ for reasons discussed in preceding sections. Also, the emission is intense, fairly stable and the emission intensity is proportional tc concentration over several orders of magnitude (e.g., 10 -7 to 1013M) (Ege et al., 1984; Leland and Powell, 1991). By attachment of a suitable group to the bipyridine moieties (Figure 3), Ru(bpy)32+ can be linked to biologically interesting molecules, such as antibodies or DNA, were it serves as a label for analysis in an analogous manner to radioactive or fluorescent labels (Bard and Whitesides, 1993; Blackburn et al., 1991). The most common and, arguably, the most important commercial application to date for ECL is its use in diagnostic assays. These applications typically use

188

Electrochemiluminescence

Figure 4. Representation of ECL "Sandwich" (antibody-antigen/analyte-antibody) assay.

ECL emitters as labels in affinity binding assays that attach the ECL emitter to the analyte of interest (Blackburn et al., 1991). The label is physically linked to one of the binding partners in the assay and provides the means for detecting the coupling of the binding partner to the analyte. Several classes of binding partners 189

Richter are used including antibody/antigen, enzyme/inhibitor, carbohydrate/lectin, and nucleic acid/complementary nucleic acid (Wild, 1994). Commercial instruments are available for ECL assays of antibodies, antigens, and DNA (Blackburn et al., 1991; Yang et al., 1994; Hoyle, 1994) and are currently based on the use of magnetic bead technology. The use of magnetic beads for immunomagnetic separations are well known and has been thoroughly reviewed elsewhere (Uhlen et al., 1994; Olsvik et al., 1994; Safarikova and Forsythe, 1995; Bruno, 1998a). In the context of ECL, the use of magnetic beads allows for the separation of the analyte and ECL label onto a solid support (i.e., the bead) followed by collection of the labeled beads on an electrode surface. Most magnetic beads used in ECL systems are paramagnetic (i.e., magnetic only in the presence of an external magnetic field) and consist of a core of magnetite (Fe304) surrounded by a polystyrene shell. These micron-sized particles may be purchased (e.g., Dynal Corp., Lake Success,NY) with preconjugated streptavidin or a variety of surface immobilization chemistries including amines, hydrazides and long chain alkyl linkers, to name a few. Since the "sandwich assay" format is often used for ECL affinity binding assays our discussion will center on it. The principles of a typical sandwich assay for an antigen are outlined in Figure 4. Magnetic beads modified by attaching an antibody for a particular antigen of interest (e.g., prostrate specific antigen, PSA), the sample of interest, and Ru(bpy)32+ - labeled antibodies are mixed. If the antigen of interest is present it acts as a bridge to form the "sandwich" structure, and the antibody labeled with ECL luminophore becomes attached to the magnetic bead. If no antigen is present, the labeled antibody does not attach to the bead. These labeled and unlabeled beads are then flushed into an ECL flow cell, where they are captured on the working electrode by positioning a magnet beneath the working electrode. The beads are washed to remove any unattached Ru(bpy)32§ - labeled antibodies as well as other reaction components, and a solution of the appropriate composition containing a coreactant (usually TPrA) is pumped into the cell. The concentration of the coreactant is kept constant and high (e.g., > 50 mM) to maximize the sensitivity of the detection and to prevent fluctuations in concentration of the coreactant from changing the ECL. The electrode is then swept to positive potentials to initiate ECL and the intensity of the emitted light is measured with a photomultiplier tube. In these measurements, the number of ECL labels on the solid phase is directly proportional to the concentration of analyte. The magnetic beads are then washed from the cell, which is cleaned and made ready for the next sample. Currently, more than forty assays, including those for tumor and cardiac markers, analytes relevant to infectious diseases, fertility therapies, and thyroid diseases, are commercially available (Bard et al., 2000).

190

Electrochemiluminescence 2+

Ho~O~i/O~

,,~

T

N

Figure 5. Ru(bpy)32§ - phosphoramadite linker for ECL labeling of nucleic acids.

ECL coupled with magnetic bead separation has also been used to develop assays for a variety of biotoxoids that are important for both food industry and military applications. For example, assays for bovine leutenizing hormone show greater sensitivity with ECL than standard radioimmunoassays (Deaver, 1995). Also, several authors have reported extremely sensitive ECL assays for bacterial species in a variety of matrices. Species such as such as anthrax (Bacillus anthracis) (Gatto-Menking et al., 1995; Bruno and Yu, 1996; Bruno and Kiel, 1999), Escherichia coli O157 and Salmonella typhimurium (Yu and Bruno, 1996; Yu, 1996) have been reported with limits of detection and assay sensitivity equal to or greater than conventional assays using flow cytometry, enzyme linked immunosorbant assays (ELISA) and radioallergosorbent test (RAST). Most ECL assays reported in the literature have used an N-hydroxysuccinimide (NHS) ester linked Ru(bpy)32§ conjugate ("TAG NHS Ester" available from IGEN International Inc., Figure 3) or a Ru(bpy)32+ phosphoramidite - conjugate (Figure 5)) to label analytes. Using these labels, the ECL of HW 1 gag gene has been reported with detection limits of < 10 to 30 gene copies (Blackburn et al., 1991; Kenten et al., 1992). Coupling ECL with polymerase chain reaction (PCR) amplification has lowered the detection limit of HIV 1 gag DNA to less than five copies (Schutzbank and Smith, 1995). Other assays and applications incorporating both PCR and ECL for nucleic acid based analyses have been 191

Richter reported (Stem, H.J et al., 1995; Heroux and Szczepanik, 1995; Wilkinson et al., 1995; Motmans et al., 1996; Gudibande et al., 1992; Van Gemen et al., 1994) including the quantitation of Varicella zoster DNA in whole blood, plasma and serum (De Jong et al., 2000) and the detection of viable oocysts of Cryptosporidium parvum (Baeumner et al., 2001). This will undoubtedly continue to be an area of intense research activity. Although Ru(bpy )3 2+ is the most widely used ECL luminophore, assays have also been developed for a range of antioxidants that quench anthracene sensitized ECL upon electrolysis of sodium citrate, methanol and dissolved oxygen (Chmura and Slawinski, 1994). Indole and tryptophan have also been shown to generate ECL upon electrolysis in the presence of hydrogen peroxide, with detection limits of 0.1/,tM for both indole and tryptophan (Chen et al., 1997) and sensors have been reported for a range of alcohols and saccharides since hydroxyl compounds have been shown to generate ECL directly (Egashira et al., 1996). Diaminotoluene isomers form weakly electrochemiluminescent compounds in the presence of Au § and Cu 2§ ions (Bruno and Cornette, 1997). Since aminoaromatic compounds like diaminotoluene are often associated with the degradation of explosives such as TNT, this approach may find use in military applications. Surprisingly, ECL can also be obtained from blood extracted from tunicates ("sea-squirts"), a class of marine invertebrates (Bruno et al., 1997). Although the exact source of the low-level ECL is not known, it is believed to emanate from metal-ion tunichrome complexes since synthetic analogues of the tunichrome chromophore showed a 10 fold ECL enhancement when complexed to Hg 2+. 3.2. Instrumentation

Experiments focused on annihilation ECL of radical ions are carried out in fairly conventional electrochemical apparatus. However, cells, electrodes and experimental procedures must be modified to allow electrogeneration of two reactants, rather than one, while taking into account constraints imposed by optical measurement equipment and the exclusion of stray light (i.e., "light-tight" experiments). In addition, one must pay scrupulous attention to the purity of the s01vent/supporting electrolyte system, especially with organic systems (e.g., polyaromatic hydrocarbons). Water and oxygen are particularly harmful to these experiments since they can quench ECL. Thus, cells and electrodes are constructed to allow transfer of solvent and degassing on high-vacuum lines or in inert-atmosphere ("glove") boxes. Electrochemical apparatus for coreactant ECL are, in many instances, identical to those used in annihilation ECL (Knight and Greenway, 1994). However, the 192

Electrochemiluminescence constraints of working with nonaqueous systems (e.g., vacuum lines) are alleviated. The earliest experiments were carried out in electrochemical batch cells designed to fit into optical spectrophotometer chambers (Rubinstein and Bard, 1981). As the development of coreactant ECL for use in diagnostics and for flow injection and liquid chromatographic applications increased, many ECL flow cell configurations were developed (Hill et al., 1986; Downey and Nieman, 1992; Sakura and Imai, 1988) . Electrode configurations, cells incorporating them and experimental details for both annihilation and coreactant systems have been thoroughly reviewed (Knight and Greenway, 1994; Faulkner and Bard, 1977; Bard and Faulkner, 1980). The ORIGEN TM analyzer (IGEN International, Inc) was the first commercial instrument that used ECL (Carter and Bard, 1990; Blackburn et al., 1991; Kenten et al., 1991; Kenten et al., 1992). This analyzer provides highly sensitive and precise assays in an automated format. It employs a flow injection system that allows rapid and reproducible deter~nations of sequential samples. The detector is a photomultiplier tube positioned directly above the working electrode so that light from the electrode can be recorded and integrated during each measurement. Typically, the assays use magnetic beads as a solid support and Ru(bpy)32+frPrA as the label and coreactant, respectively. However, the instrument may also be operated without the use of beads or other solid supports. A personal computer controls the instrument and aids in the processing and storage of data. The sample and assay reagents are combined in plastic tubes and agitated in a carousel to mix the sample and reagents and allow assay binding reactions to go to completion. The sample solution containing magnetic beads is then flushed automatically into an ECL cell, where they are captured on the working electrode by applying a magnetic field. The beads are washed, a solution containing coreactant is pumped into the cell and ECL is induced. Following light measurement, the beads are washed from the cell and the cell cleaned in preparation for the next measurement. A typical read and clean cycle requires approximately 1 minute. Roche Diagnostics, a licensee of IGEN Intemational, Inc.'s technology, has developed the ELECSYS TM instrument for conducting immunoassays in centralized hospital and reference laboratories. The fundamental technology and operation of this highly automated instrument is similar to the ORIGEN TM analyzer. The instrument can operate in random access mode and has the capability to produce Short Turn .Around Time (STAT) samples. Several other instrument configurations have been developed (Rozhitskii, 1992) including one that incorporates an RIU)E for both electrochemical and ECL analyses. ECL instrumentation has also followed the trend that has developed in the past decade towards smaller cells and electrodes (i.e., microfabrication). For example, an ECL system approximately 1/20 th the size of the original ORIGEN TM instrument has been developed (the TRICORDER | detection system, IGEN International, Inc.). This system is self-contained and has an accuracy and 193

Richter sensitivity equal to the ORIGEN TM analyzer. A miniaturized ECL cell has also been developed based on the ECL of Ru(bpy)3 ~§ for the determination of peptides having proline at the amino terminal (Egashira et al., 2000). Several other cells and devices have also been reported. For example, a thin layer flow cell using a planar optical waveguide coated with an indium-tin oxide (ITO) layer was developed. This ITO layer was then modified by covalently attaching glucose oxidase, and ECL measured from luminol at the end of the waveguide at the cell's edge (Kremeskotter et al., 1995). The generation of Ru(bpy)32§ ECL on interdigitated gold microelectrodes mounted above a photodiode has been observed, with a limit of detection for Ru(bpy)32§ of 0.5 txM (Fiaccabrino et al., 1998a). An ECL cell with gold and optically transparent indium tin oxide coated glass electrodes incorporating a photodiode has also been fabricated (Hsueh et al., 1996) for use in the quantification of DNA labeled with Ru(bpy)32+. Reported limits of detection were 1 nM, with a cell volume of 85 ktL. Cell volumes as low as 100 nanoliters have also been reported (Arora et al., 1997) with Ru(bpy)32+ ECL (detection limit 5 x 10"13 M) using a cell composed of a sandwich of poly(methylmethacrylate) layers containing two platinum thin film electrodes, connected to a conventional flowing systems using photomultiplier tube detection. The ultimate goal of much of this work is the development of miniaturized analytical systems, or the so called "lab on a chip" technology (Haswell, S.J., 1997). Fundamental studies aimed at the development of sensors and probes have also been reported. An especially active area has been the development of fibre optic probes for ECL analyses (van Dyke and Cheng, 1989; Kuhn et al., 1990; Egashira et al., 1992). For example, a gold-coated fibre optic probe for the measurement of Ru(bpy)32§ in the presence of peroxydisulfate has been developed (Kuhn et al., 1990). Also, a miniaturized fibre optic sensor has been developed and applied to the determination of oxalate using Ru(bpy)32+ (Egashira et al., 1990) in real urine samples with a limit of detection of 3 x 10.5 M. The ultimate goal of many of these studies is the development of portable devices for use in point-of-care clinical analyses (e.g., in hospital examining rooms) (Bard et al., 2000) and environmental applications (Bruno, 1998a).

4. Advantages and Limitations As with other measurements based on the emission of light (e.g., photoluminescence, chemiluminescence), ECL labels have distinct advantages over detection methods such as radioactivity. For example, they are sensitive, nonhazardous, inexpensive, diagnostic of the presence of a particular label, linear over a wide range and incorporate simple and relatively inexpensive equipment. When compared to such light emission techniques as photoluminescence (PL) and chemiluminescence (CL), ECL also displays certain desirable qualities. In PL, excited state formation occurs upon absorption of electromagnetic radiation: 194

Electrochemiluminescence R + hv --> *R

(39)

The versatility of this technique lies in the number of species able to luminesce, the quantum efficiency of emission, and the ability to incorporate these molecules into a wide variety of formats. Unfortunately, this versatility also leads to limitations. For example, in clinical situations, typical biological fluids containing analyte may also contain a large number of potential luminophores. In ECL, for a complex to emit it must meet several stringent criteria, including stable redox chemistry and the ability to undergo energetic electron or energy transfer. Of course, this advantage of ECL is also a potential limitation, in that the number of efficient ECL labels is diminished. CL involves the generation of excited states due to an energetic chemical reaction. In a typical CL reaction, reagents are pumped separately to the reaction site. In ECL, on the other hand, production of reagents occurs electrochemically in-situ from passive precursors, allowing spatial and temporal localization of the emission near the electrode. This results in enhanced sensitivity since the optics used for light detection can be focused on a relatively small area. Furthermore, amplification is possible in ECL due to the turnover of reactants at or near the electrode surface resulting in sub-picomol detection limits and a linear dynamic range of greater than six orders of magnitude. Ru(bpy)32+ - labels are also extremely stable and can be stored for over 1 year at room temperature in the dark. Often, the small size of the ECL luminophore allows multiple labels to be attached to the same molecule without affecting the stability, immunoreactivity or hybridization of the probes. However, as with any electrochemical process, stringent cleaning of the electrodes is required prior to and after each run to ensure reproducibility.

5. Potential for Expanding Current Capabilities Solution phase co-reactant ECL is quite sensitive, with sub-picomolar detection limits achieved (Leland and Powell, 1991). When the ECL luminophore is bound to a magnetic particle (the particle can then be captured on the surface of the electrode prior to electrochemical stimulation), detection limits as low as 1018M are attainable (Blackburn et al., 1991; Kenten et al., 1991). However, there are many systems where even greater sensitivity is needed, such as in environmental (where pre-concentration of samples is often necessary) and molecular diagnostics applications, where the detection of as few as 10 molecules would eliminate the need for sample amplification (e.g., via the polymerase chain reaction). One approach has been to vary the properties of the ECL luminophore. For example, Ru(bpy)32+ has an ECL efficiency of 0.050 (Rubinstein and Bard, 1981b; Glass and Faulkner, 1981), or ~ 5 % of the Ru(bpy)32+ molecules that undergo electron transfer generate emission. With the goal of increasing the magnitude of ECL emission, and therefore increasing ECL sensitivity and lowering detection limits, the ECL of the bimetallic ruthenium system

195

Richter [(bpy)zRu]z(bphb) 4+ (bphb = 1,4-bis(4'-methyl-2,2'-bipyridin-4-yl)benzene) was studied (Richter et al., 1998). The ligand bphb is capable of binding two independent metal centers through a "bridging ligand" framework. This bimetallic species produced more intense emission ( 2 - 3 fold) than Ru(bpy)32+ in aqueous and nonaqueous solution using annihilation and coreactant methods. A key point to this study was that for enhanced ECL to be possible in multimetallic assemblies, there must be small electronic coupling between metal centers via the bridging ligand so that the metal centers are electronically isolated or "valence trapped" (Robin and Day class I Systems; Robin and Day, 1967). This work has recently been extended to dendrimeric systems containing eight Ru(bpy)3 z+ units at the periphery (Zhou and Roovers, 2001) of a carbosilane dendrimer platform. Preliminary experiments indicated that the ECL of the Ru(bpy)3 z+ dendrimer is five times that of the reference monometallic species. As with the bimetallic study, spectroscopic and electrochemical studies show that the Ru(bpy)32+ units don't interact in either the ground or excited state showing that ECL (and photoluminescence) emission can be amplified by using multimetallic species. Multimetallic compounds such as these show much promise for use in analytical applications. However, it has yet to be shown whether these types of labels will change nucleic acid hybridization or affinity binding of antigens and antibodies in diagnostic applications. Another approach to improving ECL sensitivity that has met with limited success has been to vary the nature of the coreactant. Numerous amine-based coreactants have been studied, including primary, secondary and tertiary systems (Leland and Powell, 1991), and attempts made to understand the electron donating and withdrawing properties that might lead to optimum coreactant efficiency (Knight and Greenway, 1996). To date TPrA still provides the optimum ECL in the Ru(bpy)32+ system (Leland and Powell, 1991; McCord and Bard, 1991). Yet another approach is to add a species to the solution that will facilitate excited state formation and/or lead to increased quantum yields for emission. For example, the ECL intensity of Ru(bpy)32+ increased slightly (< 5%) in the presence of benzene (Dixon et al., 1993). The reason for the increase is unclear, but the excited states of ruthenium and osmium polypyridyl systems are sensitive to the nature of the environment and are able to detect subtle changes in solution composition (Meyer, 1978). The presence of benzene may lead to decreased interactions between the hydrophobic luminophore and the solvent media, resulting in increased ECL efficiency. Although Ru(bpy)32+ has many properties that make it an ideal ECL luminophore for sensitive and selective analytical methods, it would be useful to have other ECL labels that span a wide range of wavelengths so that simultaneous determination of several analytes in a single sample is possible. For example, Ru(bpy)32+ has a broad emission spectrum stretching from about 500 to 700 nm (Lr~x " 620 nm), and this can be a disadvantage in applications where an ECL internal standard or multianalyte determinations are desired. The ECL of a series of europium chelates, cryptates and mixed-ligand chelate/cryptate complexes were studied (Richter and Bard, 1996) since many trivalent 196

Electrochemiluminescence lanthanides display high photoluminescence efficiencies, large Stokes' shifts (-300 nm) and narrow emission spectra (Crosby et al., 1961; Sinha, 1971). ECL appears to occur by a different mechanism than transition metal systems via a "ligand-sensitization" route, where ECL occurs in the organic ligands with subsequent transfer to the f-orbitals of the metal centers. Although it was clear from this work (Richter and Bard, 1996) that the ligands play an integral role in rare earth ECL, very low ECL efficiencies were observed in nonaqueous solvents with little to no ECL observed in aqueous media. ECL in aqueous solution has been observed for the polyaromatic hydrocarbons 9,10-diphenylanthracene-2sulfonate (DPAS) and 1- and 2-thianthrenecarboxylic acid using TPrA as a coreactant and for DPAS using peroxydisulfate as a reductive-oxidative coreactant (Equation 37) (Richards and Bard, 1995). These complexes emit in the blue and green regions of the spectrum (e.g., ~,max(DPAS) = 430 nm) making them attractive as complementary labels to Ru(bpy)32+. ECL is not only limited to the visible region of the spectrum. Recently, the first report of near-infrared electrochemiluminescence was described (Lee et al., 1997) for a heptamethine cyanine in acetonitrile using coreactants. Undoubtedly, work will continue in these areas to find new molecules and coreactants to improve and expand ECL past its current capabilities. Solubilization of Ru(bpy)32§ in aqueous nonionic surfactant solutions leads to significant, and potentially useful, changes in the electrochemiluminescence (ECL) properties (McCord and Bard, 1991; Workman and Richter, 2000). For example, increases in both ECL efficiency (>8-fold) and duration of the ECL signal were observed in surfactant media upon oxidation of Ru(bpy)32§ and TPrA (Workman and Richter, 2000). However, the mechanism of the surfactant effect is still unclear. The effect of surfactants on Ru(bpy)32+/TPrA (Workman and Richter, 2000; Zu and Bard, 2001) and other ECL systems (e.g., Ru(dp-bpy)32§ and Ru(dp-phen)32§ (dp-bpy = 4,4'-biphenyl-2,2'-bipyridyl and dp-phen = 4,7diphenyl-l,10-phenanthroline) (McCord and Bard, 1991), Os(bpy)32§ (Ouyang and Bard, 1988), and the heptamethine cyanine dye IR-144 (C56H73NsOsS2)) (Lee et al., 1998) were attributed to strong hydrophobic interactions between the ECL luminophore and micellized surfactant. However, recent work (Zu and Bard, 2001) on the effects of electrode hydrophobicity on ECL indicate that adsorption of Triton X-100 (polyoxyethylene(10) isooctylphenyl ether) on Pt and Au electrodes renders the surface more hydrophobic, facilitating coreactant oxidation and leading to increased ECL intensities in the Ru(bpy)32+/TPrA system. A recent study of the effects of nonionic chain lengths on Ru(bpy)32+ / TPrA ECL (Factor et al., 2001) confirm these results. Although the effects of micelles and discrete complexation of the surfactants with Ru(bpy)32+ and TPrA cannot be ruled out, these studies indicate that increases in ECL intensity are probably due to changes in electrode hydrophobicity upon formation of a surfactant adsorption layer and less likely due to micelle interactions (Zu and Bard, 2001; Factor et al., 2001). The precise mechanism of the surfactant effect is still under study, but the dramatic increases in ECL intensity, coupled with work 197

Richter on more efficient ECL labels and coreactants may have profound impacts on the sensitivity of ECL for a variety of applications. Ultrasonic enhancement of ECL has also been investigated. Ultrasonic irradiation on coreactant and annihilation ECL of Ru(bpy)32§ in aqueous oxalate solutions and in acetonitrile solutions, respectively, increases the electrochemiluminescence yield over 100%, results in highly stable and reproducible ECL signals and leads to less electrode fouling (Walton et al., 1992; Malins et al., 1997). The dramatic increases appear to be due to agitation of the system, leading to greater mass transport across the electrode double layer and less dependence on diffusion to get material to the electrode surface. Also, it is speculated that the degassing effects of sonication reduce the aggregation of gas bubbles at the electrode surface and prevent the formation of passivating films. The mechanisms of ECL under sonification are the same as conventional conditions, but the high reproducibility of the signal has allowed the measurement of ECL quenching via oxygen to be measured with greater precision than previously possible (Malins et al., 1997). Environmental applications for ECL are also being explored. For example, the increased ECL emission of Ru(bpy)32+ in the presence of benezene has led to the proposal of using it to detect aromatic hydrocarbon pollutants (Dixon et al., 1993). Other environmental applications include the detection of toxic metal ions (Bruno, 1998b; Taverna et al., 1998), metal ions bonded to aminoaromatics (Bruno and Cornette, 1997), and environmentally important ethoxylate surfactants containing amine groups (Alexander and Richter, 1999). Quenching of Ru(bpy)32+/TPrA electrochemiluminescence in the presence of phenols, hydroquinones, catechols and benzoquinones (McCall et al., 1999; McCall and Richter, 2000) has opened up the possibility of,using ECL to detect these environmentally, biologically and pharmacologically important classes of compounds. The generation of ECL at micro- and ultramicroelectrodes has been known for some time (Collinson and Wightman, 1993; Maness and Wightman, 1995). In fact, an ultramicroelectrode was used to observe individual reaction events of DPA annihilation ECL in nonaqueous solution (Collinson and Wightman, 1995) showing the extreme sensitivity possible with ECL. ECL with microelectrodes has recently been coupled to Scanning Probe techniques such as Scanning Electrochemical Microscopy (SECM) to image surfaces (Fan et al., 1998; Maus et al., 1999). High frequency ECL has also been used to image the surfaces of microelectrodes (Wightman et al., 1998). More recently, ECL has been used as a light source for near-field scanning optical microscopy using ultramicroelectrodes with effective diameters from 1 ~m to less than 100 nm (Zu et al., 2001). This technique was used to image an interdigitated array with resolution comparable to that observed via near-field scanning optical microscopy (NSOM). Using ECL for near-field imaging appears to have several 198

Electrochemiluminescence advantages over NSOM. In NSOM, a metal-coated fibre optic probe is used and this leads to fundamental resolution limits due to the finite skin depth of the metal coating. In ECL, tip preparation is easier since standard techniques can be used to generate nanometer sized electrodes (Zu et al., 2001). In addition, ECL does not require a laser so there is no heating of samples and tip from absorption of light on metal coatings. Therefore, this approach looks promising for nearfield optics in solution. As mentioned above, many recently developed methods of ECL analysis include chromatographic or other separation steps and the use of Ru(bpy)32+ ECL in HPLC has been reviewed (Lee, 1997). More recently, ECL has been coupled with capillary electrophoresis (CE) (Dickson et al., 1997; Forbes et al., 1997; Tsukagoshi et al., 1997; Bobbitt and Jackson, 1997; Gillman et al., 1994). Since biochemical applications of ECL have rapidly expanded in the past decade, coupling ECL with CE should provide rapid, efficient and versatile methods of separating and detecting biochemicals using extremely small sample volumes. Annihilation ECL of Ru(bpy)32+ has recently been reported in aqueous solution containing no electrolyte (Fiaccabrinno et al., 1998b). This is possible using a microfabricated interdigitated carbon dual-electrode system. Each electrode is biased to form the reduced, Ru(bpy)3 +, or oxidized, Ru(bpy)33+, species. The electrodes are in close enough proximity (2 lxm width and spacing) that the simultaneously produced reactants can diffuse together and undergo annihilation (Equation (13)). Also, carbon is used as the electrode material to prevent formation of water oxidation and reduction products (e.g., oxygen) that tend to quench ECL emission. Annihilation ECL has been used to study aluminum quinolate/triarylamine and related organic complexes used as light emitting diodes (Gross et al., 2000; Anderson et al., 1998), sol-gel composites containing Ru(bpy)3 z+ (Sykora and Meyer, 1999), and diode-like chemiluminescence in frozen concentrations gradients of the ruthenium polymer poly-[Ru(vbpy)3](PF6)2 (Maness et al., 1996). ECL in sol-gel derived glasses (Collinson et al., 2000), nation-silica composite films (Khramov and Collinson, 2000) and gel-entrapped Ru(bpy)32+ (Collinson et al., 1999) using coreactants has also been observed with the potential for using both coreacant and annihilation ECL in display device technology. Although commercial applications for these systems have yet to materialize, they have opened up fascinating areas for both fundamental and applied studies. This chapter has centered on the background and history of ECL, and its development into a biomedical research and clinical diagnostic tool. However, ECL is a versatile detection methodology and is being developed as a sensor and probe for other applications. Since the first detailed studies, over 1000 papers, patents and book chapters have appeared on ECL, ranging from the very applied to a focus on the underlying science. With the interest in using ECL reactions as the basis for highly sensitive and selective analysis, the prediction made by 199

Richter Faulkner and Glass that "Continued research in this area will probably stress the development of ECL as a probe rather than as an end in itself' (Faulkner and Glass, 1982) has come to fruition. One wonders what the next 30 years hold for ECL but whatever the improvements and new aspects of ECL that emerge, ECL will continue to show promise for optical biosensing and in other areas of science and technology.

6. References

Alexander, C. and M.M.Richter, 1999, Anal. Chim. Acta 402, 105. Anderson, J.D., E.M. McDonald, P.A. Lee, M.L. Anderson, E.L., Ritchie, H.K. Hall, T. Hopkins, E.A. Mash, J. Wang, A. Padias, S. Thayumanavan, S. Barlow, S.R. Marder, G.E. Jabbour, S. Shaheen, B. Kippelen, N. Peyghambarian, R.M. Wightman and N. R. Armstrong, 1998, J. Am. Chem. Soc. 120, 9646. Arora, A., A.J. DeMello and A. Manz, 1997, Anal. Commun. 34, 393. Bader, J.M. and T. Kuwana, 1965, J. Electroanal. Chem. 10, 104. Baeummer, A.J., M.C. Humiston, R.A. Montagna and R.A. Durst, 2001, Anal. Chem. 73, 1176. Bard, A.J., J.D. Debad, J.K. Leland, G.B. Sigal, J.L. Wilbur and J.N. Wohlstadter, 2000, Encyclopedia of Analytical Chemistry, Ed. R.A. Meyers, Wiley, Chichester, pp 9842-9849. Bard, A.J. and L.R. Faulkner, 1980, Electrochemical Methods, Wiley, New York, pp 624-626. Bard, A.J. and L.R. Faulkner, 2001, Electrochemical Methods Fundamentals and Applications Second Edition, Wiley, New York, pp 736 - 745. Bard, A.J. and C.M.Whitesides, 1993, U.S. Patent 5,21,605. Bard, A.J. and C.M. Whitesides, 1993, U.S. Patent 5,238,808. Bard, A.J. and G.M. Whitesides, 1994, U.S. Patent 5,310,687. Barigelletti, F., L. De Cola, V. Balzani, R. Hage, J.G. Haasnoot, J. Reedijk and J.G. Vos, 1991, Inorg. Chem. 30, 641. Bartelt, J.E., S.M. Drew and R.M. Wightman, 1992, J. Electrochem. Soc. 139, 70. Bernanose, A., T. Bremer and P. Goldfinger, 1947, Bull. Soc. Chim. Belg., 56, 269. B irks, J.B., 1970, Photophysics of Aromatic Molecules, Wiley, New York. Blackburn, G.F., H.P. Shah, J.H. Kenten, J. Leland, R.A. Kamin, J. Link, J. Pterman, M.J. Powell, A. Shah, D.B. Talley, S.K. Tyagi, E. Wilkins, T-G Wu and R.J. Massey, 1991, Clin. Chem. 37, 1626. Bobbit, D.R. and W.A. Jackson, 1997, US Patent 5,614,073. Bolletta, F., A. Rossi and V. Balzani, 1981, Inorg. Chim. Acta 53, L23-L24. Brilmyer, G.H. and A.J. Bard, 1980, J. Electrochem. Soc. 127, 104. Brune, S.N. and D.R. Bobbitt, 1991, Talanta 38, 803.

200

Electrochemiluminescence Bruno, J.G., 1998a, Recent Research Developments in Microbiology. Vol. 1, Ed. S.G. Pandalai, Trivandrum Publishing, India, pp. 25-46. Bruno, J.G., 1998b, J. Biolumin. Chemilumin. 13, 139. Bruno, J.G., S.B. Collard and A.R.J. Andrews, 1997, J. Biolumin. Chemilumin. 12, 155. Bruno, J.G., S.B. Collard, D.J. Kuch and J.C. Comette, 1996, J. Biolumin. Chemilum 11,193. Bruno, J.G. and J.C. Comette, 1997, Microchem. J. 56, 305. Bruno J.G. and J.L. Kiel, 1999, Biosens. Bioelectron. 14, 457. Bruno, J.G. and H. Yu, 1996, Appl. Environ. Microbiol. 62, 3474. Carter, M. and A.J. Bard, 1990, Bioconj. Chem. 1,257. Caspar, J.V., T.J. Meyer, 1983, J. Am. Chem. Soc. 105, 5583. Chandross, E.A. and R.E. Visco, 1964, J. Am. Chem. Soc. 86, 5350. Chen, G.N., R.E. Lin, Z.F. Zhao, J.P Duan and L. Zhang, 1997, Anal. Chim. Acta 341, 251. Chmura, J. and J. Slawinski, 1994, J. Biolum. Chemilumin. 9, 1. Collinson, M.M., B. Novak, S.A. Martin and J.S. Taussig, 2000, Anal. Chem. 72, 2914. Collinson, M.M., J.S. Taussig and S.A. Martin, 1999, Chem. Mater. 11, 2594. Collinson, M.M. and R.M. Wightman, 1993, Anal. Chem. 65, 2576. Collinson, M.M. and R.M. Wightman, 1995, Science 268, 1883. Crosby, G.A., R.E. Whan and R.M. Allire, 1961, J. Chem. Phys. 34, 743. Cruser, S.A. and A.J. Bard, 1967, Anal. Lett. 1, 1I. Danielson, N.D., L. He, J.B. Noffsinger and L. Trelli, 1989, J. Pharm. Biomed. Anal. 7, 1281. Deaver, D.R., 1995, Nature 377, 758. De Jong, M.D., J.F.L. Weel, T. Schuurman, P.M.E. Wertheim-van Dillen and R. Boom, 2000, J. Clin. Microbiol. 38, 2568. Demas, J.N. and G.A. Crosby, 1968, J. Mol. Spectrosc. 26, 72. Demas, J.N. and G.A. Crosby, 1971, J. Am. Chem. Soc. 93, 2841. Dickson, J.A., M.M. Ferris and R.E. Milofsky, 1997, J. High Resolut. Chromatogr. 20, 643. Dixon, S.B., J. Sanford and B.W. Swift, 1993, Principles and Practices for Petroleum Contaminated Soils, Eds. E.J. Calabrese and P.T. Kosteki, Lewis Publishers, pp 85-99. Dong, L. and M.T. Martin, 1996, Anal. BioChem. 236, 344. Downey, T.-M. and T.A. Nieman, 1992, Anal. Chem. 64, 261. Dufford, R.T., D. Nightingale and L.W. Gaddum, 1927, J. Am. Chem. Soc. 49, 1858. Egashira, N., N. Kondoh, Y. Kurauchi and K. Ohga, 1992, Denki Kagaku 60, 1148. Egashira, N., H. Kumasako and K. Ohga, 1990, Anal. Sci. 6, 903. Egashira, N., Y. Nabeyama, Y. Kurauchi and K. Ohga, 1996, Anal. Sci. 12, 793. Egashira, N., J. Piao, E. Hifumi and T. Uda, 2000, Bunseki Kag 49, 1029. Ege, D., W.G. Becker and A.J. Bard, 1984, Anal. Chem. 56, 2413. 201

Richter

Factor, B., G. Muegge, S. Workman, E. Bolton, J. Bos and M.M. Richter, 2001, Anal. Chem. submitted. Fan, F.-R.F., D. Cliffel and A.J. Bard, 1998, Anal. Chem. 70, 2941. Faulkner, L.R. and A.J. Bard, 1977, Electroanalytical Chemistry Vol. 10, Ed. A.J. Bard, Marcel Dekker, New York, pp 1-95. Faulkner, L.R. and R.S. Glass, 1982, Chemical and Biological Generation of Excited States; Eds. A. Waldemar, and C. Giuseppe, Academic Press, New York, ch. 6. Fiaccabrino, G.C., N.F. de Rooij and M. Koudelka-Hep, 1998, Anal. Chim. Acta 359, 263. Fioccabrino, G.C., M. Koudelka-Hep, Y-T Hsueh, S.D. Collins and R.L. Smith, 1998, Anal. Chem. 70, 4157. Forbes, G.A., T.A. Nieman and J.V. Sweedler, 1997, Anal. Chim. Acta, 347, 289. Gatto-Menking, D.L., H. Yu, J.G. Bruno, M.T. Goode, M. Miller, and A.W. Zulich, 1995, Biosens. Bioelectron. 10, 501. Gilman, S.D., C.E. Silverman and A.G. Ewing, 1994, J. Microcolumn Sep., 6, 97. Glass, R.S. and L.R. Faulkner, 1981, J. Phys. Chem. 85, 1160. Gross, E.M., J.D. Anderson, A.F. Slaterbeck, S. Thayumanavan, S. Barlow, Y. Zhang, S.R. Marder, H.K. Hall, M. Flore Nabor, J.-F. Wang, J.-F., E.A. Mash, N.R. Armstrong and R.M. Wightman, 2000, J. Am. Chem. Soc. 122, 4972. Gudibande, S., J.H. Kenten, J. Link, K. Friedman and R.J. Massey, 1992, J. Molec. Cell. Probes 6, 495. Haapakka, K.E., 1982, Anal. Chim. Acta 139, 229. Haapakka, K.E. and J.J. Kankare, 1980, Anal. Chim. Acta 118, 333. Harvey, N., 1929, J. Phys. Chem. 33, 1456. Haswell, S.J., 1997, Analyst 122, R 1. He, L., K.A. Cox and N.D. Danielson, 1990, Anal. Lett. 23, 195. Hemingway, R.E., S.-M. Park and A.J. Bard, 1975, J. Am. Chem. Soc. 95,200. Hercules, D.M., 1964, Science 143, 808. Heroux, J.A. and A.M. Szczepanik, 1995, PCR Methods Appl. 4, 327. Hill, E., E. Humphreys and D.J. Malcolme-Lawes, 1986, J. Chromatogr. 370, 427. Holeman, J.A. and N.D. Danielson, 1994, J. Chromatogr. 679, 277. Hoyle, N.R., 1994, J. B iolumin. Chemilumin. 9, 289. Hsueh, Y.T., R.L. Smith and M.A. Northrup, 1996, Sens. Actuators B Chem. 33, 110. Itoh, I. and K. Honda, 1979, Chem. Lett. 99. Jackson, W.A. and D.R. Bobbitt, 1994, Anal. Chim. Acta. 285,309. Jameison, F., R.I. Sanchez, L. Dory, J.K. Leland, D. Yost and M.T. Martin, 1996, Anal. Chem. 68, 1298. Kanoufi, F., Y. Zu and A.J. Bard, 2001, J. Phys. Chem. B. 105, 210. 202

Electrochemiluminescence Kenten, J.H., J. Casadei, J. Link, S. Lupold, J. Willey, M.J. Powell, A. Rees and R.J. Massey, 1991, Clin. Chem. 37, 1626. Kenten, J.H., S. Gudibande, J. Link, J.J. Wiley, B. Curfman, E.O Major and R.J. Massey, 1992, Clin. Chem. 38, 873. Khramov, A.N. and M.M. Collinson., 2000, Anal. Chem. 72, 2943. Kim, J., F.-R.F. Fan, A.J. Bard, C.-M. Che and H.B. Gray, 1985, Chem. Phys. Lett. 121,543. Knight, A.W., 1999, Trends in Anal. Chem. 18, 47. Knight, A.W. and G.M. Greenway, 1994, Analyst 119, 879. Knight, A.W. and G.M. Greenway, 1996, Analyst 121,101R. Kremeskotter, J., R. Wilson, D.J. Schiffrin, B.J. Luff and J.S. Wilkinson, 1995, Meas. Sci. Technol. 6, 1325. Kuhn, L.S., A. Weer and S.G. Weber, 1990, Anal. Chem. 62, 1631. Kuwana, T., B. Epstein and E.T. Seo, 1963, J. Phys. Chem. 67, 2243. Leland, J.K. and M.J. Powell, 1991, J. Electroanal. Chem. 318, 91. Lee, S.K., M.M. Richter, L. Strekowski and A.J. Bard, 1997, Anal. Chem. 69, 4126. Lee, W.Y., 1997, Mikrochim. Acta 127, 19. Liang, P., L. Dong and M.T. Martin, 1996, J. Am. Chem. Soc. 118, 9198. Luttmer, J.D. and A.J. Bard, 1981, J. Phys. Chem. 85, 1155. Malins, C., R. Vandeloise, D. Walton and E. VanderDonckt, 1997, J. Phys. Chem. 101, 5063. Maloy, J.T., K.B. Prater, K.B. and A.J. Bard, 1971, J. Am. Chem. Soc. 93, 5959. Maloy, J.T. and A.J. Bard, 1971, J. Am. Chem. Soc. 93, 5968. Maness, K.M., R.H. Terrill, T.J. Meyer, R.W. Murray and R.M. Wightman, 1996, J. Am. Chem. Soc. 118, 10609. Maness, K.M. and R.M. Wightman, 1995, J. Electroanal. Chem. 396, 85. Maus, R.G., E.M. McDonald and R.M. Wightman, 1999, Anal. Chem. 71, 4944. McCall, J., C. Alexander and M.M. Richter, 1999, Anal. Chem. 71, 2523. McCall, J. and M.M. Richter, 2000, Analyst 125,545. McCord, P.M. and A.J. Bard, 1991, J. Electroanal. Chem. 318, 91. Meyer, T.J., 1978, Acc. Chem. Res. 11, 94. Motmans, K., J. Raus and C. Vandevyer, 1996, J. Immunol. Meth. 190, 107. Noffsinger, J.B. and N.D. Danielson, 1987, Anal. Chem. 59, 865. Noffsinger, J.B. and N.D. Danielson, 1987, J. Chromatogr. 387, 520. Olsvik, O., T. Popovic, E. Skjerve, K.S. Cudjoe, E. Homes, J. Ugelstad and M. Uhlen, 1994, Clin. Microbiol. Rev. 7, 43. Ouyang, J. and A.J. Bard, 1988, Bull. Chem. Soc. Jpn. 1988, 61. Paris, J.P. and W.W. Brandt, 1959, J. Am. Chem. Soc. 81, 5001. Parker, C.A., 1968, Photoluminescence of Solutions, Elsevier, Amesterdam. Prieto, I., J. Teetsov, M.A. Fox, D.A. Vanden Bout and A.J. Bard, 2001, J. Phys. Chem. A., 105,520. Richards, T.C. and A.J. Bard, 1995, Anal. Chem. 67, 3140. Richter, M.M. and A.J. Bard, 1996, Anal. Chem. 68,2641.

203

Richter Richter, M.M., A.J. Bard, W. Kim and R.H. Schmehl, 1998, Anal. Chem. 70, 310. Richter, M.M., F.-R.F. Fan, F. Klavetter, A.J. Heeger and A.J. Bard, 1994, Chem. Phys. Lett. 226 115. Robin, M.B. and P. Day, 1967, Adv. Inorg. Chem. RadioChem. 10, 247. Roundhill, D.M., 1994, Photochemistry and Photophysics of Coordination Complexes, Plenum, New York, ch. 5. Rozhitskii, N.N., 1992, J. Anal. Chem. USSR, 47, 1288. Rubinstein, I. and A.J. Bard, 1980, J. Am. Chem. Soc. 102, 6641. Rubinstein, I. and A.J. Bard, 1981, J. Am. Chem. Soc. 103,512. Rubinstein, I., C.R. Martin and A.J. Bard, 1983, Anal. Chem. 55, 1580. Safarikova, S.M. andS.J. Forsythe, 1995, J. Appl. Bacteriol. 78, 575. Saji, T. and A.J. Bard, 1977, J. Am. Chem. Soc. 99, 2235. Sakura, S. and H. Imai, 1988, Anal. Sci. 4, 9. Santhanam, K.S.V. and A.J. Bard, 1965, J. Am. Chem. Soc. 87, 139. Schultz, L.L., J.S. Stoyanoff and T.A. Nieman, 1996, Anal. Chem. 68, 349. Schutzbank, T.E. and J. Smith, 1995, J. Clin. Microbiol. 33, 2036. Sinha, A.P.B., 1971, Spectrosc. Inorg. Chem. 2, 255. Smith, P.J. and C.K. Mann, 1969, J. Org. Chem. 34, 1821. Stem, H.J., R.D. Carlos and T.E. Schutzbank, 1995, Clin. Biochem. 28,470. Sutin, N. and C. Creutz, 1978, Adv. Chem. Ser. 168, 1. Sykora, M. and T.J. Meyer, 1999, Chem. Mater. 11, 1186. Taverna, P.J., H. Mayfield and A.R.J. Andrews, 1998, Anal. Chim. Acta 373, 111. Tsukagoshi, K., K. Miyamoto, E. Saito, T. Nakajima, K. Hara and K. Fujinaga, 1997, Anal. Sci. 13, 639. Tokel, N. and A.J. Bard, 1972, J. Am. Chem. Soc. 94, 2862. Tokel-Takvoryan, N.E., R.E. Hemingway and A.J. Bard, 1973, J. Am. Chem. Soc. 95, 6582. Uchikura, K. and M. Kirisawa, 1991, Anal. Sci. 7, 803. Uchikura, K., M. Kirisawa and A. Sugii, 1993, Anal. Sci. 9, 121. Uhlen, M., E. Homes and O. Olsvik, Eds. 1994, Advances in Biomagnetic Separation Eaton Publishing. van Dyke, D.A. and H.Y. Cheng, 1989, Anal. Chem. 61,633. Van Gemen, B., R. Van Beuningen, A. Nabbe, V. Van Strijp, S. Jurriaans, P. Lens, R. Schoones and T. Kievits, 1994, J. Virol. Meth. 49, 157. Van Houten, J. and R.J. Watts, 1976, J. Am. Chem. Soc. 98, 4853. Vitt, J.E., D.C. Johnson and R.C. Engstrom, 1991, J. Electrochem. Soc. 138, 1637. Vogler, A. and H. Kunkeley, 1984, Ang. Chem. Int. Ed. Engl. 23,316. Vojir, V., 1954, Collect. Czech. Chem. Comun.19, 872. Wallace, W.L. and A.J. Bard, 1979, J. Phys. Chem. 83, 1350. Walton, D.J., S.S. Phull, D.M. Bates, J.P. Lorimer and T.J. Mason, 1992, Ultrason. 30, 186. White, H.S. and A.J. Bard, 1982, J. Am. Chem. Soc. 104, 6891. 204

Electrochemiluminescence Wightman, R.M., C.L. Curtis, P.A. Flowers, R.G. Maus and E.M. McDonald, 1998, J. Phys. Chem. B, 102, 9991. Wild, D.Ed., 1994, The Immunoassay Handbook, Macmillan Press Ltd.. Wilkinson, E.T., S. Cheifetz and S.A. De Grandis, 1995, PCR Methods Appl. 4, 363. Workman, S. and M.M. Richter, 2000, Anal. Chem. 72, 5556. Xu. G. and S. Dong, 2000, Anal. Chem. 72, 5308. Yamashita, K., S. Yamazaki-Nishida, Y. Harmia and A. Segawa, 1991, Anal. Chem. 63, 872. Yang, H., Y.K. Leland, D. Yost and R.J. Massey, 1994, Bio/Technol. 12, 193. Yu, H., 1996, J. Immunol. Meth. 192, 63. Yu, H. and J.G. Bruno, 1996, Appl. Environ. Microbiol., 62, 587. Zhou, M. and J. Roovers, 2001, Macromolec. 34, 244. Zu, Y. and A.J. Bard, 2000, Anal. Chem. 72, 3223. Zu, Y. and A.J. Bard, 2001, Anal. Chem., in press. Zu, Y., Z. Ding, J. Zhou, Y. Lee and A.J. Bard, 2001, Anal. Chem. 73, 2153.

205

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 7

SURFACE PLASMON RESONANCE BIOSENSORS

JIOAHOMOLA, PH.D., SINCLAIRS. YEE, PH.D., ANDDAVID MYSZKA, PH.D.* Department of Electrical Engineering, University of Washington, Seattle, WA 98105, USA Center for Biomolecular Interaction Analysis, University of Utah, Salt Lake City, UT 84132, USA

Surface plasmon resonance (SPR) biosensors exploit special electromagnetic w a v e s surface plasmon-polaritons - to probe changes in refractive index at surfaces of metals. SPR biosensors can therefore be used to monitor the interaction between an analyte in solution and its biospecific partner immobilized on the metal surface without the use of labels. Major application areas include detection of low levels of biological analytes and study of biomolecular interactions. In the past ten years, SPR biosensor technology has been commercialized and SPR biosensors have become a central tool for characterizing and quantifying biomolecular interactions both in life science and pharmaceutical research.

1. Technical Concept Surface plasmon resonance (SPR) biosensors use surface plasma waves to probe biomolecular interactions occurring at the surface of a sensor. This chapter introduces surface plasma waves, presents methods for their excitation and interrogation, and discusses the concept of SPR optical biosensors.

1.1. Surface plasma waves I. I. 1. Surface plasmon-polaritons at a plane interface between a semi-infinite dielectric and a metal. Surface plasmon-polaritons or surface plasma waves

occur at the surfaces of metals, which behave like nearly free electron plasmas. 207

Homola, Yee, and Myszka Let us consider a plane interface composed of two semi-infinite homogeneous, isotropic media and choose the coordinate system so that the metal (dielectric function aM) occupies the region z<0, while the dielectric medium (dielectric constant co) occupies the region z>0 (Figure 1). Electromagnetic field modes supported by this geometry can be found by solving Maxwell's equations in each medium and applying boundary conditions at the interface. We are looking for modes guided by the interface, and thus the magnitudes of all their field vectors have to be decreasing with increasing distance from the interface. In general, stratified media with plane boundaries can support transverse electric (TE) modes (in which the direction of propagation and magnetic and electric intensity vectors form an orthogonal triad and the electric vector is parallel to the interface) and transverse magnetic (TM) modes (where the magnetic vector is perpendicular to the direction of propagation of the wave and parallel to the plane of interface) (Figure 1). TE-polarized surface modes can not exist in this geometry if the materials involved are non-magnetic (Boardman, 1982). Therefore, only TMmodes can be supported by the metal-dielectric interface. In complex notation, electric and magnetic intensity vectors of a TM-polarized mode are: Dielectric" 2

E = (E~,O,E z) = 1 , 0 ,

Aexp[-croz + i(flx-cot)],

(1)

r H = (O,H 0) = (0,1,0) -/Aa)e~ exp[-croz + i ( f l x - cot)] Y' c'%/~ Metal: E = (E~,0,E.). =ll,O,-ifl)Aexp[crMz+crMj i(flx-cot)], r

H = (O,H,,O) = (0,1,0 9

c

o~u/z

,4, - :

-

e.

(2)

exp[crM z + i(flx - cot)]

where co is the angular frequency, fl is the propagation constant, c is the speed of light in vacuum, /~ is the permeability of vacuum, and A is a normalization constant proportional to the energy carried by the mode (Boardman, 1982). Application of boundary conditions requiring continuity of tangential components of electric and magnetic intensity vectors (Hy and Ex) at the interface yields an equation for the propagation constant (Boardman, 1982):

co,[ trMe~

(3)

f l = c ~t eM + e o

208

Surface Plasmon Resonance Biosensors

~Z

E

Dielectric

Y

Metal Figure 1. Transverse magnetic (TM) wave at a metal-dielectric interface.

The propagation constant is generally a complex number because the dielectric function of metal em is a complex function of the angular frequency. Equations (1), (2), and (3) represent a true surface plasma wave (SPW) that propagates along the interface and decays exponentially in a direction perpendicular to the direction of propagation, if the real part of eM is negative and its absolute value is smaller than Eo. At optical wavelengths, this condition is fulfilled for several metals of which gold and silver are the most commonly used (Figure 2). The real part of the propagation constant is related to the effective refractive index N - the quantity commonly used in waveguide optics - in the following manner:

N =CRe{fl}=

l~emeMe~ eD}

(4)

where Re{ } denotes the real part of a complex number. The imaginary part of the propagation constant is related to the modal attenuation b (in dB/cm if [3 is given in l/m): b = Im{fl} In 0.2 10 = Im

e~e~

0.2co

(5)

[~e~, + e o j clnlO

where Im{ } denotes the imaginary part of a complex number. Spectral dependencies of the effective refractive index and mode attenuation for SPWs supported by silver and gold surfaces are shown in Figure 3. As follows 209

Homola, Yee, and Myszka

~ ---

" "" "" " -,..,,.,...,, .,, ,.. ,,..

-20

.,---- Re{~)

-40

Gold Silver

8

6

- "-~

,...,

(9

4E

-60

tr

-80

"

~

600

0 1 m { e 4 " - ' .-.7. ~ . ' - .: 2

700

800

9O0

1000

Wavelength [nm]

Figure 2. Dielectric function of gold and silver. Data taken from Palik (1985) and Ordal et al. (1983).

a)

b) f

S P W on gold S P W on silver

~) 1.7

~ ....

S P W on gold S P W on silver

105

1.6

{ ,o.

"~ 1.5

>*,

g

1.4 10 s

1.3

i

600

.

.

.

.

|

700

.

.

.

.

.

i

.

.

.

.

800 W a v e l e n g t h

i

900

.

.

.

.

i

!

1000

600

700

800

900

1000

W a v e l e n g t h [nm]

[nm]

Figure 3. Surface plasmon-polaritons: effective refractive index and attenuation. The effective refractive index (a) and attenuation (b) of a surface plasmon-polariton as a function of the wavelength for a surface plasmon-polariton propagating along the interface between a metal (gold or silver) and a non-dispersive dielectric (refractive index = 1.32).

from Figure 3, the effective index of an SPW increases with decreasing wavelength. The effective refractive index of the SPW supported by a gold surface is larger that that of the SPW supported by silver because the real part of the dielectric constant of gold is smaller than that of silver. The mode attenuation of the SPW supported by a gold boundary is larger than that of the SPW supported by a silver surface as the imaginary part of the dielectric constant of gold is larger than that of silver. The field profile of a surface plasmon-polariton is illustrated in Figure 4. 210

Surface Plasmon Resonance Biosensors

a)

b) 1.0

.

.~ 0.8

.

.

.

.

.

.

.

.

.

.

.

.

~

~

1.0

~ 0.8

~"~

0.6

~~

0.4

t--

0.4

0.2

"0

0.2

-

0.6

._z,

~ i.: !~_~~L _ _ _ _ - ' ~ _ _ _ - ' . . . .

0.0

c~

Metal -0.2

-0.2

Dielectric

i

0.0

0.2

0.4

~.~,=o.o

,

~

O)

0.6

z-coordinate ,[p.m]

0.8

1.0 ~

-0.2

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

z-coordinate [p.m]

Figure 4. Surface plasmon-polariton: field pattern. Spatial distribution of the magnetic intensity for a surface plasmon-polariton at the interface between gold and a nondispersive dielectric (refractive index = 1.32) in the direction perpendicular to the interface, calculated for the wavelength of 630 nm (a) and 850 nm (b).

The field decay in the direction perpendicular to the direction of propagation may be characterized by means of the penetration depth, Lp. The penetration depth is defined as the distance from the interface at which the amplitude of the field has fallen to 1/e of its value at the surface and is related to field parameters CtD and ctM as follows:

Lt,o = 1/Re{ere } ,

LoM : 1/Re(a M}

(6)

Similarly, the attenuation of an SPW in the direction of propagation can be characterized by means of the propagation length, which is defined as the distance in the direction of propagation at which the energy of the wave decreases by a factor of 1/e.

L = l/[2Im{ fl}]

(7)

Characteristics of SPWs supported by surfaces of gold and silver are given in Table 1. As follows from Equation (7), Table 1, and Figure 3, SPWs propagating along the surface of silver are less attenuated than those propagating along the surface of gold. SPWs propagate with high attenuation and the attenuation increases with decreasing wavelength. As follows from Figure 4, the electromagnetic field of an SPW is distributed in a highly asymmetric fashion and majority of the field (usually more than 90 per cent) is concentrated in the dielectric medium.

211

Homola, Yee, and Myszka Table 1. Major characteristics of surface plasmon-polaritons at the interface between metal (silver or gold) and a non-dispersive dielectric (refractive index = 1.32) for two different wavelengths.

t

Metal Wavelength [nm] Penetration depth into metal [nm] Penetration depth into dielectric [nm] Propagation length [~m] "'

......

Silver 630 24 219 19

f

Gold 85()

850 23 443

630

29

25 4oo

57

2

24

-

1.1.2. Surface plasmon-polaritons on a thin metal film surrounded by dielectric media. A thin metal film, surrounded by dielectric media may support surface plasmon-polaritons at both the interfaces. For thin metal films there is coupling between surface plasmon-polaritons associated with each boundary, giving rise to mixed m o d e s - symmetric and antisymmetric surface plasmon-polaritons. These modes are found as solutions of Maxwell's equations in each medium, which satisfy boundary conditions at both the interfaces (Burke et al., 1986). Figure 5 shows the effective refractive index and attenuation of these two modes as a function of the thickness of the metal film. Clearly, if the metal film is rather thick (-- 100 nm), the propagation constants of the two modes are almost identical. As the metal thickness decreases, the two modes become more different. The symmetric surface plasmon-polariton exhibits an effective refractive index and attenuation which both decrease with decreasing metal film thickness, while the effective refractive index and attenuation of the antisymmetric surface plasmon-polariton increase with decreasing thickness of the metal film. Thus, the symmetric surface plasmon-polariton exhibits a lower effective refractive index and attenuation than its antisymmetric counterpart. Therefore, the first mode is often referred as a long-range surface plasmonpolariton while the other is referred as a short-range surface plasmon-polariton, a reference to their relative attenuation and propagation lengths. The dispersion properties of symmetric and antisymmetric surface plasmonpolaritons (Figure 6) are rather different from those of the traditional SPWs supported by a single interface (Figure 3). It should be noted that the effective refractive index of the symmetric mode exhibits much lower 'dispersion' (dependence on the wavelength) than that of the antisymmetric mode. This is because the symmetric surface plasmon exhibits a much weaker electromagnetic field in the metal film, which is a strongly dispersive medium, Figure 7. As illustrated in Figure 7, the distribution of magnetic intensity (and the transverse electric intensity) is symmetric and antisymmetric with respect to the

212

Surface Plasmon Resonance Biosensors

10= '

Antiaymmelric SPW

~

10~ 103 .Q

o

g

10t 10~

/I

~ y m m a ~ r i e SPW

10.2 Z 1.60

|

~

1.40

Symmetric SPW 1..._~

D

-

.

. ~, - - - - r - - " T - ' ~ .

.

,

. .

i

2D

=

.

9

9

40

~

,

,

r

.

.

--_'.

.

. ,

. ,

80

SO

~o0

Film thickness [nmJ

Figure 5, Surface plasmon-polaritons on a thin metal film, Effective refractive index and mode attenuation as a function of the metal fiim thickness for symmetric and antisymmetric surface plasmon-polaritons propagating along a thin gold film embedded between two identical non-dispersive dielectrics (refractive index = 1.32); wavelength = 800 nm. Note the drop in the attenuation for the symmetric bound mode as the metal thickness approaches zero.

~ i~~----..__~ ...... 48,

7OO

nO0

g00

1000

WavelonO~la,-nl

Figure 6. Surface plasmon-polaritons on a thin metal film. Effective refractive index and mode attenuation as a function of the wavelength, calculated for symmetric and antisymmetric surface plasmon-polaritons propagating along a thin gold film embedded between two identical non-dispersive dielectrics (refractive index = 1.32), gold film thickness = 20 nm.

center of the metal film for symmetric and antisymmetric surface plasmonpolaritons, respectively. The field of the symmetric SPW penetrates deeper into the dielectric media than that of the antisymmetric SPW.

213

Homola, Yee, and Myszka

a) ,--,

1.0

~

0.8

b) 1.0

0.8 0.6

,_...

"~ 0.4

: E ~"

0.6

c..~_ "o ,~

0.4

.~

.~ tt~

0.2

Im{Hy}

~ o.o-

0.2

~

-0.2

~

-0.4

~

-0.6

ImlHy}

O.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

-0.8 -0.2

-1 .o

. . . .

i

-o5

. . . .

i

. . . .

0.o

z-coordinate

I

O.S

. . . .

, -1.0 -I .0

1.o

[,p.m]

,

,

,

l

-0.5

. . . .

l

. . . .

0.0 z-coordinate

,

0.5

. . . .

1.0

[~m]

Figure 7. Surface plasmon-polaritons on a thin metal film: field patterns. Spatial distribution of the magnetic intensity of a) symmetric and b) antisymmetric surface plasmon-polaritons propagating along a thin gold film embedded between two identical non-dispersive dielectrics (refractive index = 1.32), gold film thickness = 20 nm, wavelength = 800 nm.

1.2. Optical excitation of surface waves A light wave can couple to a surface plasma wave at a metal-dielectric interface if the component of light's wavevector that is parallel to the interface matches that of the surface plasma wave. As the propagation constant of a surface plasma wave at a metal-dielectric interface (Eq. 3) is larger than that which can be provided by the component of the wavevector of light in the dielectric, SPWs cannot be excited directly by light incident onto a smooth metal surface. In order to allow for excitation of a SPW by a light wave, the light's wavevector needs to be enhanced to match that of the surface plasma wave. The light's wavevector can be increased by passing the light through a medium with a refractive index higher than that of the dielectric medium at the boundary at which the SPW is to be excited, or by roughness of the metal surface (statistical or regular).

1.2.1. Excitation of surface plasmon-polaritons using prism couplers. The enhancement of a light wave's wavevector by passing light through an optically denser medium is illustrated in Figure 8. The effective refractive index of an SPW at the metal (gold) - dielectric (water) interface ranges from 1.35 (at a wavelength of 1000 nm) to 1.68 (at a wavelength of 600 nm) while the effective refractive index of light in the dielectric does not exceed 1.335. A light wave propagating in the dielectric with a higher refractive index, such as BK7 glass, can provide a component of the normalized wavevector parallel to the metal surface with a value between 0 and about 1.52, depending on the angle between the direction of propagation of the light wave and the metal surface. Therefore, 214

Surface Plasmon Resonance Biosensors

1.7 x| 9"o 1.6 .c_

Surlace plasma wave at Id-water interface

1.5

V / / / ~

./~CC//.////

;////.kieht wave in ~ (

fllass

i,. .> 1.4

1.:3 600

9

700

800

9

. ight wave in water;

900

1000

W a v e l e n g t h [nm]

Figure 8. Effective refractive index of a surface plasmon-polariton as a function of the wavelength for a surface plasmon-polariton propagating along the interface between a gold and water. Normalized wavenumbers provided by light waves in a B K7 glass and water shown for comparison.

the condition for coupling light into the SPW can be fulfilled for wavelengths long than 630 nm. The coupling between a light wave in the high refractive index dielectric medium and an SPW at the metal - low-refractive index dielectric interface can be established by the total internal reflection method. A light wave passes through a high refractive index prism and is totally reflected at the prism base generating an evanescent wave penetrating a metal film in the Kretschmann configuration (Figure 9a) or a dielectric layer in the Otto configuration (Figure 9b). This evanescent wave propagates along the interface with the propagation constant, which can be adjusted to match that of the SPW by controlling the angle of incidence. Thus, the matching condition can be fulfilled allowing the evanescent wave to be coupled into the SPW. Assuming that the prism has only a minor influence on the propagation constant of the SPW at the interface of a metal and a low refractive index dielectric, the coupling condition may be expressed as"

(8)

~-Tsin(O) = Rel, ]- e~t''e~ l L~/e,, + eo J

where 0 denotes the angle of incidence, ep, cMand co denote dielectric functions of the prism, the metal film and the dielectric medium (ep > eo). As discussed in Section 1.2.4, this condition is fulfilled for 'thick' metal films. (For gold at optical wavelengths, in this approximation the 'thick' gold films are 50 nm or more). 215

Homola, Yee, and Myszka

Figure 10. Excitation of surface plasma waves by guided modes of optical waveguides, a) excitation of SPW on the inner boundary of a thin metal film, and b) excitation of SPW on the outer metal boundary through a dielectric buffer layer.

1.2.2. Excitation of surface plasmon-polaritons using optical waveguides. Similarly, a surface plasma wave can be excited by a light wave guided by an optical waveguide. This approach is illustrated in Figure 10. Light propagates in a waveguide in the form of guided modes. The electromagnetic field of a guided mode is concentrated in the waveguiding layer. A fraction of the optical energy propagates in the form of an evanescent wave in the low-refractive index medium surrounding the waveguiding layer. In the section of the waveguide containing an SPW-active metal film, this evanescent wave can excite an SPW at the outer (in Figure 10a) or inner (Figure 10b) surface of the metal film. Assuming that the waveguide influences the propagation constant of the SPW only slightly, the coupling condition for a guided mode and a surface plasma wave can be expressed as follows: (9)

216

Surface Plasmon Resonance Biosensors

Figure 11. Excitation of surface plasma waves at the surface of a diffraction grating.

where Ney denotes the effective refractive index of the waveguide mode, and ~v~ and CDdenote dielectric functions of the metal and the dielectric medium.

1.2.3. Excitation of surface plasmon-polaritons using grating couplers. A surface plasma wave may also be excited by a light wave with its wavevector increased by the wave's interaction with surface roughness of the metal film. Most commonly used configurations for SPW excitation on rough metal surfaces are based on diffraction gratings (Figure 11). ff a light wave is made incident on a periodically distorted surface of a diffraction grating, a series of waves directed away from the surface at different angles is produced (Hutley, 1982). The components of the wavevector of these diffractiongenerated light waves parallel to the interface are:

k x + mG =km

(1 O)

where m is the diffraction order (integer), kx is the component of the wavevector of the incident light along the grating surface, G is the grating wavevector, and k~m is the wavevector of the diffracted light wave. In case of shallow gratings, the coupling condition may be expressed as: e ~ o sin(0)+ m-- = +Re A [~/eM +eo

(Ii)

where 0 is the angle of incidence of the light wave, A denotes the pitch of the grating (A= 2n/G), eM and eo denote dielectric functions of the metal and the dielectric medium, and ~ denotes the free-space wavelength. 217

Homola, Yee, and Myszka

1.2.4. Excitation of surface plasmon-polaritons: energy transfer. In the process of optical excitation of surface plasma waves, a portion of the energy of the light wave is transferred into the energy of a surface plasma wave and dissipated in the metal film. Assuming IRe(eM)l>>eoand IRe(eM)l>>Im(eM),the reflectivity for the Kretchmann geometry of the attenuated total reflection method may be expressed as (Raether, 1983): R(k x) = 1 -

4FiFr~a ~kx - fl+~2 + ( F i +

1-',..d

)2

'

(12)

where

k.,. = ~

co sin(O), r+ = Re {/3 + Aft}, F, = Im {fl}, Fr,,a = Im {Aft}. (13) s

0 is the angle of incidence, cv is the dielectric constant of the prism, co is the angular frequency, c is the velocity of light in vacuum, kx is the component of the wavevector of the light wave which is parallel with the interface, and fl is the propagation constant of the SPW as given by Equation (3)./Yi characterizes the attenuation of the surface plasma wave at the metal-dielectric interface due to the dissipation in the metal. The complex quantity Aft characterizes the influence of the prism on the propagation constant of the surface plasma wave. Its imaginary part, Frad, characterizes the attenuation of the surface plasma wave due to the coupling of the SPW to optical radiation in the prism. As follows from Equation (12), the reflectivity exhibits a Lorentzian dip located at the angle Om~ given by the following condition: --sin(0,~,,) =/3+.

(14)

r

In the approximation of thick metal films, Re{fl}>>Re{dfl} and the right-hand side of Equation (14) is equal to the propagation constant of the surface plasmon at the metal-dielectric interface. This suggests that the reflectivity minimum occurs when the component of the wavevector of the incident light wave parallel to the interface matches the propagation constant of the SPW. The depth of the reflectivity dip can theoretically range from 0 to 100 per cent. The reflectivity minimum reaches zero if ~ = ~ , indicating that all the energy of the incident light wave can be lost due to the SPW's excitation if the parameters of the structure and incident light wave are chosen properly. It can be shown by the examination of Equation (12) that the width of the dip is proportional to ~ + ~ad and thus to the total loss the surface plasma wave exhibits. Therefore the excitation of surface plasma waves on a metal with low loss gives rise to a narrow dip in the TM-wave reflectivity and vice versa. For a given metal, the reflectivity minimum and the width of the dip cannot be minimized at the same time, as ~ decreases with decreasing metal film thickness while ~,d exhibits an opposite trend. In order to produce deep and narrow reflectivity dips, one has to 218

Surface Plasmon Resonance Biosensors

a) 1.0 ;

b' ......

SPW..

....

0.8

270 240

1.0

210 180

0.8

150

150,-~ 120 .~

9~'~"0.6

180

120 90

. ~ 0.6

"13

6o ~' 30

13-

0

0.2

-30

0

0.2 -3o

-60 0.0 - '

, 50

.

.

.

.

.

.

.

.

.

55

-90 60

0.0

~

49.0

Angle of incidence [deg]

,-

'~-

49.2

.

.

.

.

.

49.4

.

.

.

.

.

49.6

.

.

.

.

.

49.8

.

.

60

50.0

A n g l e of i n c i d e n c e [deg]

Figure 12. Reflectivity for a transverse magnetic wave in: a) Kretschmann (three-layer) geometry consisting of an SF14 glass prism (refractive index = 1.65), a gold layer (thickness = 50 nm), and a low refractive index dielectric medium (refractive index = 1.32), and b) four-layer geometry consisting of an SF14 glass prism, a low refractive index buffer (refractive index = 1.32, thickness = 1200 nm), a thin gold layer (20 nm), and a low-refractive index dielectric medium (refractive index- 1.32), wavelength - 800 nm. Note the difference in the x-axis scale.

use very thin metal films (<20 nm) and reduce the coupling of the SPW to optical radiation in the prism by introducing a buffer layer between the prism and the metal film. Such a structure allows for the excitation of long-range surface plasma waves. Drops in the reflectivity due to the excitation of a surface plasrnon in the traditional three-layer Kretschmann configuration and excitation of longrange surface plasma waves in the four-layer configuration are illustrated in Figure 12. Besides the change in the amplitude, the light wave exciting a surface plasma wave undergoes also a change in phase (Figure 12). 1.3. Optical sensor based on surface plasma waves

1.3.1. Surface plasma waves as refractive index probes. When a surface plasma wave propagates along the metal-dielectric interface, its field penetrates into the dielectric (with penetration depth given by Equation (6)), probing the dielectric medium. Any change in the dielectric's optical properties results in a change in the characteristics of the SPW, especially its propagation constant. The relationship between variations in optical parameters of the dielectric and the propagation constant of the SPW can be established using the perturbation theory (Snyder and Love, 1983). According to the perturbation theory, a small change in the profile of the refractive index of the dielectric, n(z) ---->n'(z), produces a change in the propagation constant of the surface plasma wave, 13 --->13': 219

Homola, Yee, and Myszka

Figure 13. Surface plasma waves probing (a) a homogeneous medium and (b) a thin dielectric layer.

k2 IIn a(z) - n2(z)]H~(z)dz P'BB+~-, 2fl iHZy(z)d z

(15)

where k is the free-space wavenumber and Hy(z) is the spatial dependence of magnetic field of the surface plasma wave. Using Equation (15), one can show that a small change in the refractive index of a homogeneous dielectric medium, An, (An <
(16)

where k denotes the free-space wavenumber. Thus, the change in the effective refractive index of the SPW produced by a change in the bulk refractive index of the dielectric medium is approximately equal to the refractive index change. If the refractive index perturbation is caused by the presence of a thin dielectric layer with the thickness, d, much smaller than the penetration depth of the SPW, Lpo, (d < Lpdl0) (Figure 13b), the corresponding change in the real part of the propagation constant of the SPW can be expressed as follows: ns ks _ 2)d, Re{Aft}--- ~Re{e.] I (n, ~ n~

(17)

where ny and n,. denote the refractive index of the thin dielectric layer and the refractive index of the dielectric medium interfacing the layer, respectively (Parriaux et al., 1990). Thus, the change in the SPW propagation constant by the presence of a thin dielectric layer is proportional to the thickness of the layer and the difference between the squares of the refractive indices of the dielectric layer and that of the surrounding medium. Using Equation (17), one can show that a 220

Surface Plasmon Resonance Biosensors change in the refractive index of such a thin layer, An, produces a change in the real part of the propagation constant of the SPW: 2n,,nlkZd

(18)

As follows from Equation (18), the induced change in the propagation constant of the SPW is proportional to the refractive index change and the thickness of the dielectric layer. Consistent with Equation (15) and (16), the factor F (F < 1) in Equation 18 accounts for the fact that the thin dielectric layer is probed by only a fraction of the field of the SPW. 1.3.2. Optical sensors based on surface plasmon-polaritons. Optical sensors based on resonant excitation of surface plasma waves, often referred to as surface plasmon resonance (SPR) sensors, are optical devices that exploit the sensitivity of the propagation constant of an SPW to refractive index to measure changes in the refractive index or changes in non-optical quantities that can produce changes in the refractive index. A change in the refractive index to be measured produces a change in the propagation constant of the SPW, which results in a change in the characteristics of the light wave interacting with the SPW (see Section 1.2.4). Based on which characteristic of the light wave interacting with SPW is measured, SPR sensors can be classified as follows: SPR sensors with angular modulation. The component of the light wave's wavevector parallel to the metal surface matching that of the SPW is determined by measuring the coupling strength at multiple angles of incidence of the light wave and determining the angle of incidence yielding the strongest coupling (Figure 14a, upper plot). The wavelength of the light wave used to excite an SPW is fixed. SPR sensors with wavelength modulation. The component of the light wave's wavevector parallel to the metal surface matching that of the SPW is determined by measuring the coupling strength at multiple wavelengths and determining the wavelength yielding the strongest coupling (Figure 14b). The angle at which the light wave is incident onto the metal film is kept constant. SPR sensors with intensity modulation. The change in the intensity of the light wave interacting with the SPW is measured (Figure 14b). Both the angle at which the light wave is incident onto the metal film and its wavelength are kept constant. SPR sensors with phase modulation. The shift in phase of the light wave interacting with the SPW is measured (Figure 14b, lower plot). Both the angle at which the light wave is made incident onto the interface and its wavelength are kept constant.

221

Homola, Yee, and Myszka

a) 1.0

b .,~ ,.

..._ . _ . .

0.8

1.0 ",

"- 0.6 r

no

.... ',

0.4

o=I

, .'''"

"

'*

{

_1200

!' ~ ~ n o no= 1.32

= 1.32

no

= 1.35

.

0.8

..

.............. 48 50 52

,

JSt

0.6

0.4

,oo' ~_ g

71:_35_ . . . .

, ................ 54 56 58

60

0

0.2

n

, . 1-100 62

0.0 600

Angle of incidence [deg]

700

eoo

900

1000

W a v e l e n g t h [nm]

Figure 14. a) Reflectivity and phase for a transverse magnetic (TM) light wave exciting surface plasma waves in the Kretschmann geometry (SF14 glass prism- 50 nm thick gold layer - dielectric) as a function of the angle of incidence for two different refractive indices of the dielectric; wavelength = 682 nm. b) Reflectivity for a TM wave in the same geometry as a function of the wavelength for two different refractive indices of the dielectric; angle of incidence = 54 deg.

SPR sensors with polarization modulation. The amplitude and phase of the TM polarized wave interacting with the SPW change if the propagation constant of the SPW changes. TE-polarized light wave does not interact with SPWs and thus exhibits no resonant amplitude and phase variations. Therefore, the polarization state of the incident light wave consisting of both the polarizations would also be sensitive to variations in the propagation constant of the SPW. 1.3.3. Surface plasmon resonance (SPR) biosensors. In principle, any phenomenon which gives rise to a change in the refractive index at the surface of the SPW-active metal film can be observed and quantified by means of an SPR sensor. SPR biosensors are SPR sensing devices that incorporate biomolecules, which recognize and are able to interact with selected analytes. These biomolecular recognition elements are immobilized on the SPR sensor surface. When a solution containing analyte molecules is brought into contact with the SPR sensor, analyte molecules in solution bind to the recognition elements on the sensor surface, producing an increase in the refractive index at the sensor surface. This change produces a change in the propagation constant of the SPW and is eventually measured by measuring a change in one of the characteristics of the light wave interacting with the SPW (Figure 15). 1.3.4. Surface plasmon resonance sensors: main performance characteristics. The main performance characteristics of SPR biosensors include sensitivity, resolution and the lowest detection limit. Sensor sensitivity is the ratio of the change in sensor output to the change in the value of the measurand. In SPR 222

Surface Plasmon Resonance Biosensors

Resonant wavevector's / ' i component along ] ] the interface

Binding of target molecules to sensor surface Surface refractive index change SPW propagation constant change

I

/I

9 1 A Phase (Z~q)r) ,h I . Cha..n.g..e....'.m... ...................//[ [~ t--IN~['"'cia'arac'ienstics of light ~ Intensity (Air)

-

_

Polarization ,

Figure 15. Surface plasmon resonance (SPR) biosensor 9principle of operation.

biosensors, the sensor output is, depending on the modulation method used, the coupling angle of incidence, wavelength, intensity, phase, and polarization of the light wave interacting with the SPW, and the measurand is the concentration of analyte. Mathematically, the sensitivity is the slope of the calibration curve Output=Output(Measurand): S=

0Output -~,-0P OMeasurand 0c

(19)

where P is the output of the sensor and c is the concentration of target analyte. As the output of an SPR biosensor is usually a non-linear function of the concentration of analyte, the sensor's sensitivity is not a constant, but varies with the analyte concentration. The sensitivity of an SPR biosensor can be decomposed into two components - sensitivity to the refractive index change produced by the binding of analyte to a biospecific coating on the sensor's surface, SRI, and the efficiency, E, with which the presence of analyte at a given concentration, c, is converted into the change in the refractive index, n ' (20)

S - OP On = SINE. On Oc

The efficiency, E, depends on the ability of the biospecific coating to capture analyte molecules and on the properties of the target analyte, and thus is application dependent. The refractive index sensitivity, Sin, can be expressed as:

SR I "-

OP 0Re{fl} ORe{fl} On

__

(21)

SiS2

223

Homola, Yee, and Myszka

1.60 L ' l ~

'

"'

1.55 I\~'~.~,~ SPW (no= 1.32) ~ 1.50I._~~SPW(n~ \ 145 r-.~,

>~

__._

Lw (e= 56 deg)

"~ - i- - ~ ~ ( _ o

1.3600.I5~

700

800

=,_4_~eg_ ! 900

1000

Wavelength [nm] Figure 16. Spectral dependence of the effective refractive index of a surface plasma wave (SPW) at gold - dielectric interface for two different refractive indices of the dielectric medium. Also shown is the spectral dependence of the component of the wavevector of an light wave (LW), which is parallel to the SPW-active metal surface, calculated for two different angles of incidence.

The first term, S I, depends on the modulation method and the method of excitation of the SPW and can be determined if the coupling condition is known for the sensor geometry (examples are Equations (8), (9) and (11)). $2 is independent of the modulation method and the method of excitation of the SPW and describes the sensitivity of SPW's propagation constant to the refractive index change. It can be easily determined from Equations (16) (17), and (18), depending on whether the refractive index change occurs in the vicinity of the sensor' s surface or through the whole depth of the SPW' s field. The sensitivity of sensors based on the attenuated total reflection method and angular modulation decreases with the increasing wavelength and, except for very short wavelengths, depends on the wavelength rather weakly (Homola et al., 1999a). On the contrary, the sensitivity of SPR sensors with wavelength modulation increases rapidly with the increasing wavelength. This difference in the spectral properties of refractive index sensitivities is illustrated in Figure 16, which shows the coupling between a light wave (LW) in a prism and a surface plasma wave (SPW) at the metal-dielectric interface for two different refractive indices of analyte. For a dielectric refractive index equal to 1.32, the coupling condition between the SPW and LW is fulfilled for an angle of incidence of 54 degrees and a wavelength of 682 nm. A spectrally homogeneous increase in the refractive index of the dielectric by 0.03 produces an increase in the effective refractive index of the SPW of 0.037 (at a wavelength of 650 nm) to 0.032 (at a wavelength of 1000 nm). To reestablish the coupling condition, the angle of incidence has to be increased by about 2 degrees, or the wavelength has to be increased by about 73 nm. The sensitivity of an angle-modulated sensor is proportional to the change in 224

Surface Plasmon Resonance Biosensors the effective refractive index of the SPW, which does not change much with the operating wavelength. However, the sensitivity of the wavelength-modulated sensor depends on the relative dispersion of the surface plasma wave and the light wave, which varies substantially with the wavelength. As the dispersions of SPW and LW are more similar at longer wavelengths, a change in the effective refractive index of the SPW produces a larger shift in the coupling wavelength at longer wavelengths. For example, for an angle of incidence of 56 degrees, the coupling condition is fulfilled for a wavelength of 646 nm and the refractive index induced change in the coupling wavelength is about 32 nm, and thus is smaller by a factor of more than two than for the sensor operating at wavelength operating at the angle of incidence of 54 degrees. Detailed studies of the sensitivity of SPR sensors have been previously published (Kooyman et al., 1988; de Bruijn et al., 1992; Yeatman, 1996; Homola, 1997a; Homola et al., 1999a, Johansen et al., 2000a). Resolution is the smallest increment in the measurand that can be resolved by the sensor. Herein, we shall use this term for describing the ability of SPR instrumentation to resolve changes in the refractive index. We shall use the lowest detection limit to describe the lowest concentration of analyte that can be measured by the SPR biosensor. Both the resolution and lowest detection limit of an SPR biosensor are ultimately limited by the accuracy with which the SPR sensor's output can be determined, which is limited by the stability of sensor's baseline (the sensor response stability over a period of time if no analyte is present) and noise of the sensor output. The noise is determined by noise properties of the components of the sensor system (Kolomenskii et al., 1997) and the data processing method used (Chinowsky et al., 1999a).

2. History In 1902 Wood, observing the spectra of a continuous light source using a reflection diffraction grating, noticed narrow dark bands in the spectrum of diffracted light (Wood, 1902). Theoretical analysis carried out by Fano (1941) led to the conclusion that these anomalies are associated with surface plasma waves supportable by a grating. In 1968 Otto pointed out that surface plasma waves may be excited using attenuated total reflection (Otto, 1968). In the same year, Kretschmann and Raether (1968) reported excitation of surface plasma waves in another configuration of the attenuated total reflection method. Following the pioneering work of Otto, Kretschmann, and Raether, research in surface plasma waves has gathered momentum and broadened (see, for example, Boardman, 1982; Raether, 1983). The potential of surface plasma waves for characterizing thin films (Pockrand, 1978) and studying processes at metal interfaces (Gordon et al., 1980) was recognized in the late seventies. In 1983, Nylander and Liedberg exploited surface plasma waves excited in the Kretschmann geometry of attenuated total reflection (ATR) for gas detection and 225

Homola, Yee, and Myszka biosensing (Nylander et al., 1982; Liedberg et al., 1983). In the following years, SPR sensors based on diffraction gratings (Cullen et a1.,1987/88) and planar optical waveguides (Kreuwel et al., 1987) were demonstrated, while angular and spectral modulation methods were introduced to SPR sensing (Matsubara et al., 1988; Zhang and Uttamchandani, 1988). The potential of the SPR method for spatially resolved measurements was recognized, and the first surface plasmonpolariton microscopes were reported (Yeatman and Ash, 1987; Rothenh~iusler and Knoll, 1988). In the early nineties, the first fiber optic SPR sensors were reported (Jorgenson and Yee, 1993; Dessy and Bender,1994; Homola, 1995a). In an effort to further increase the sensitivity of SPR sensors, phase modulated SPR sensors based on heterodyne phase measurement (Nelson et al., 1996) and interferometry (Nelson et al., 1996; Nikitin et al., 1999) were introduced in late nineties. Advances in the development of SPR sensor instrumentation opened up possibilities for SPR sensor applications in the measurement of physical, chemical and especially biological quantities. The development of SPR sensor technology for biological applications received a great spur in 1990, when the first commercial SPR biosensor was launched by Biacore International AB. In the years following, Biacore has developed a range of laboratory SPR instruments based on the Kretschmann geometry of the attenuated total reflection method and angular modulation (Biacore, 2001). These sensors systems have enabled the use of SPR biosensor technology for biomolecular interaction kinetic analysis, affinity measurements, screening and concentration assays. SPR sensors have been developed by several companies. These include the IBIS system developed by British Windsor Scientific Ltd. (United Kingdom) (Windsor, 2001), SPR-670 and SPR-CELLIA systems by Nippon Laser and Electronics Laboratory (Japan) (Nippon Laser and Electronics Laboratory, 2001), an SPR system developed by Johnson & Johnson Clinical Diagnostics (United Kingdom), and the Spreeta SPR sensor developed by Texas Instruments Inc. (USA) (Texas Instruments, 2001). These SPR sensors are based on the Kretschmann geometry of the ATR method and angular modulation. An SPR probe based on a multimode optical fiber and wavelength modulation was developed by EBI Sensors Inc. (USA), which was later acquired by BIAcore. Current research into SPR biosensor hardware focuses on further improving detection limits, increasing throughput, and reducing sample volumes as well as the size of the SPR sensors to yield portable field-use analytical devices. The applications of SPR biosensors in life science and pharmaceutical research continue to evolve along with the instrumentation. Two rapidly expanding applications involve the use of biosensors in small molecule screening for drug discovery as well as the development of surface chemistries that allow stable lipid surfaces within the biosensor to support membrane biology.

226

Surface Plasmon Resonance Biosensors

Figure 17. SPR sensor based on angular modulation.

3. State of the Art

An SPR biosensor instrument consists of an optical system for excitation and interrogation of surface plasma waves, a biospecific coating incorporating a biomolecular recognition element which interacts with analyte molecules in a liquid sample, and a fluidic system comprising flow-cell or cuvette for sample confinement at the sensing surface and systems for sampling and sample delivery. This section reviews the state of the art in the development of the two key components of SPR biosensors - optical instrumentation and biospecific coatings, and discusses the main application areas for SPR biosensors. 3.1. Surface plasmon resonance sensor instrumentation 3. I. 1. SPR sensors based on prism couplers. Attenuated total reflection (ATR) in

prism couplers has been used widely for the excitation of surface plasma waves in SPR sensors. Prism-based SPR sensors based on the Kretschmann geometry of the attenuated total reflection method and all the main modulation approaches have been demonstrated. These include SPR sensors based on intensity modulation (Nylander et al., 1982; Vidal et al., 1993), angular modulation (Matsubara et al., 1988; Liedberg et al., 1993), wavelength modulation (Zhang and Uttamchandani, 1988; Homola et al., 1995b), phase modulation (Nelson et al., 1996; Nikitin et al:, 1999), and polarization modulation (Kruchinin and Vlasov, 1996). Angular and wavelength modulation have been used most frequently as they rely on multipoint measurements yielding more robust measurements than simple intensity and phase modulation approaches relying on single-point measurement. 227

Homola, Yee, and Myszka

Figure 18. SPR sensor based on spectral modulation.

Figure 17 shows an example of a prism-based SPR sensor with angular modulation. A convergent beam of monochromatic light is launched into a prism coupler and made incident onto a thin metal film. The angular component of light which fulfills the coupling condition (Equation (8)) excites a surface plasma wave at the outer boundary of the metal film. The coupling produces a narrow dip in the angular spectrum of the reflected light. The position of this dip is measured with a position sensitive photodetector (e.g., a charge coupled device or a photodiode array). This configuration has been developed into numerous research laboratory SPR sensor systems and several commercial SPR sensor instruments (see Section 2). The bulk refractive index sensitivity of these devices is typically 50-100 degrees/RIU (RIU - refractive index unit), depending on the optical properties of the prism and the SPW-active metal layer, and the wavelength (Homola et al., 1999a). The best laboratory SPR sensors based on prism coupling and angular modulation attain refractive index resolution better than 3• -7 RIU (Karlsson and St~hleberg, 1995). Several attempts to miniaturize this SPR sensor configuration have been reported. The sensor reported by Foster et al. (1994) replaces a conventional prism with a thin rectangular glass substrate with an SPR active metal film and integrates a source of monochromatic light and a CCD array into a portable system. Another system integrates a light source, a detector and an optical system for excitation and interrogation of the SPW into a single component (Melendez et al., 1996). This system uses a divergent light beam and attains a refractive index resolution of 5• 6 RIU (Elkind et al., 1999). Wavelength modulation-based SPR sensors (Figure 18) use a collimated beam of polychromatic light to excite an SPW on a thin metal film. A spectral dip due to the excitation of SPW is observed using a wavelength sensitive photodetector (e.g., spectrograph, monochromator). A refractive index sensitivity of these sensors increases with the coupling wavelength and attains about 7100 nm/RIU 228

Surface Plasmon Resonance Biosensors

Lightsource

Fiber, ~ ~ - - - - ( ~ opticcoupler / //. "--"_L "ff ( ~ ~ ~ C~176 A

"1 I spectrometer ~ Ii ~ ~ IIi~:~ ot 700 t" ,00I [O? Computer{

]/

'--.,___ ~ _

W/t (/I-\1'v-L I

900j I

:1.] ] Sample~

~ ~

f

input/outputlight I 1/-x Metalfilm

i/.A// /~"

~ Surfaceplasma waves

Figure 19. SPR sensor using a miniature retroreflective sensing element. (Reproduced from Cahill et al., 1997, with kind permission from Elsevier Science.)

around 800 nm; a typical refractive index resolution is around 3x10 "6 (Pfeifer et al., 1999). Miniaturized versions of the wavelength modulated SPR sensors based on retroreflective designs in which the reflected light with an encoded SPR dip in its spectrum is collected by the same optics originally used to launch light into the prism have been developed (Cahil et al., 1997; Stemmler et al., 1999) (Figure 19). A sensor combining angular and spectral modulations has been proposed by Karlsen et al. (1996). In this sensor, SPWs in different locations along a thin glass lightguide are excited by white light incident at different angles. Light beams associated with different angles are separated at the output of the lightguide and spectrally analyzed. The attenuated total reflection method is very attractive for the development of multichannel SPR sensing devices. The most straightforward approach to SPR multichannel sensing uses 'parallel' sensing channels. In this approach, surface plasmons are excited simultaneously in multiple areas which are arranged perpendicularly to the direction of propagation of SPWs. Light reflected from each area is separately analyzed to yield information about SPR in every location. Wavelength (Nenninger et al., 1998) and angular (Lt~f~ts et al., 1991) modulation-based SPR sensors with up to four parallel sensing channels have been demonstrated (Figure 20). An alternative approach to multichannel sensing uses sensing elements of special designs to excite SPW in different channels by means of different spectral 229

Homola, Yee, and Myszka

Figure 21. SPR dual-channel sensors based on spectral discrimination of sensing channels, a) Spectral discrimination of sensing channels by means of altered angles of incidence (Homola et al., 2001a). b) Spectral discrimination of sensing channels by means of a high refractive index overlayer (Homola et al., 1999b).

regions of the incident light. This spectral separation of optical signals from different sensing channels may be accomplished by serially exciting SPW in different areas of the sensing element by light waves of different angles of incidence, (Figure 21a) (Homola et al., 2001a) or by employing an overlayer which shifts the coupling wavelength for a part of the sensing surface to longer wavelengths, (Figure 2 lb) (Homola et al., 1999b). Another approach to multichannel SPR sensing is based on the same idea as the surface plasmon microscopy (Rothenh~iusler and Knoll, 1988). This approach, sometimes referred as SPR imaging, uses a collimated beam of monochromatic 230

Surface Plasmon Resonance Biosensors

Figure 22. SPR multichannel sensor based on SPR imaging. light to illuminate the entire surface to simultaneously couple light to surface plasma waves in multiple areas. The strength of the coupling depends on the refractive index at the sensor surface in each area. When the reflected light beam is imaged onto a two-dimensional detector array, light reflected from each area of the sensing surface is incident onto a different area of the detector (Figure 22). Analysis of intensity distribution yields information on refractive index in each area (Berger et al., 1998; Nelson et al., 1999). In general, the use of prism couplers provides SPR sensors with a number of attractive features such as optical system simplicity, robustness, versatility and potential for multichannel sensing. Although high-performance SPR sensing devices of this type tend to be rather bulky laboratory bench-top instruments, high-performance portable SPR sensing devices will likely be developed in the near future. 3.1.2. SPR sensors based on grating couplers. Light diffraction at the surface of a diffraction grating has been used in SPR sensors to a lesser extent than attenuated total reflection in prism couplers mainly because grating-based SPR sensors require advanced modeling and optimization tools (Hutley, 1982; Chandezon et al., 1982; Moharam and Gaylord, 1986) and complex fabrication procedures. Typical implementations of the grating-based SPR sensors include intensity-modulated (Cullen et al., 1987/88; Cullen and Lowe, 1990; Lawrence et al., 1996) and wavelength-modulated (Vukusic et al., 1992; Jory et al., 1994) SPR sensing devices. A wavelength-modulated SPR sensor based on a silvercoated grating demonstrated by Jory et al. (1994) was shown to exhibit refractive index sensitivity of about 1000 nm/RUI. An intensity-modulated SPR sensor based on a gold-coated grating reported by Cullen (Cullen et al., 1987/88) shows a refractive index sensitivity of about 900 percent/RIU. An interesting modification of the wavelength modulation approach using an acousto-optic modulator has been proposed by Jory et al. (1995) (Figure 23). In this approach,

231

Homola, Yee, and Myszka

signal detector reference detector cubic beam white light source r ' " ' " ' " ~ % l L " . . . . . . . . 1--'1 .splitter " " " ~ LIJlpolarizer AOTFjt ~ \

'~Ir~ !

"' '= =' -

"

' , , ' - ' . , ,,= = ,

l

I~II=

!

Ugrating

L.'.

9

, " , aperture

.........

|

analogue divider

~-

-,~

--

......

X

I

amplifier

!

/

lock-in i

|

ll l

i

' i

computer-

signal generator

Figure 23. SPR sensor based on a grating coupler and an acousto-optical modulator. (Reproduced from Jory et al., 1995, with kind permission from IOP Publishing Limited.)

the acousto-optic modulator is used to control the wavelength of a light beam incident on a diffraction grating. Modulation of the light's wavelength around the minimum reflectivity produces a modulated intensity output which is proportional to the differential of the response with respect to wavelength. By locking to the zero differential corresponding to the SPR reflectivity minimum and monitoring the acousto-optic modulator's frequency, the SPR minimum position is measured with the accuracy of about 5x10 4 nm. This leads to a grating-based SPR sensor refractive index resolution below 106 RIU (Jory et al., 1995). The grating coupler has been also used for the excitation of SPWs in an SPR sensing device based on a Schottky-barrier semiconductor structure, as demonstrated by Nikitin et al. (1994, 1997). A refractive index resolution of 1• 10.5 RIU has been accomplished with this SPR sensor. Grating-based sensors offer several advantageous features over other configurations of SPR sensors. Gratings for SPR biosensors can be produced in plastic substrates (Lawrence et al., 1996) by high volume replication techniques (Gale, 1997). Precise control of the thickness of the SPW-active metal film is not required for grating-based SPR sensors. However, it may be considered a drawback that the light beam needs to pass through a sample and therefore the sample and the flow-cell need to be optically transparent.

232

Surface Plasmon Resonance Biosensors

Figure 25. SPR sensor based on an integrated optical waveguide and wavelength modulation. (Reproduced from Dost~ilek et al., 2001, with kind permission from Elsevier Science.)

3.1.3 SPR sensors based on integrated optical waveguides. Research into integrated optical waveguide SPR sensors was pioneered by researchers at the University of Twente in the late eighties (Lambeck, 1992; additional references cited therein). Since then, SPR sensing devices using slab (Layers et al., 1994) and channel (Harris and Wilkinson, 1995) single-mode integrated optical waveguides have been reported. These sensing devices exhibit a rather limited operating range. The sensor's operating range can be adjusted by using waveguides fabricated in low refractive index glass (Harris and Wilkinson, 1995), waveguides with a buffer layer (Layers et al., 1994), a high refractive index overlayer (Ctyrok3~ et al., 1997), or a multilayer (Weiss et al., 1996). Integrated-optical SPR sensors exhibit a refractive index sensitivity of about 2000 nm/RIU (Homola et al., 1997b). An integrated optical SPR sensor with a reference arm compensating for variations of light levels in the waveguide input has been described by Mouvet et al. (1997) (Figure 24). An alternative approach based on a channel integrated optical waveguide and wavelength modulation has

233

Homola, Yee, and Myszka

Collimator Fiber

Optical fiber

connector

Lamp Beam du____mp

, Fiber optic

spectrograph

Sensor

Mode scrambler SPR-active metal (silver or gold) .

:

"ver

,.or

Figure 26. SPR probe based on a multimode optical fiber. (Reproduced Jorgenson and Yee, 1994, with kind permission from Elsevier Science.)

been demonstrated by Dosuilek et al. (2001) (Figure 25). This sensor was demonstrated to exhibit a refractive index sensitivity of 2100 nm/RIU and resolution of lx106 R1-U. Integrated optical waveguide SPR sensors hold potential for the development of multichannel sensing devices with on-chip referencing and multianalyte detection capability. Challenges for this SPR sensor configuration are in the development of efficient and inexpensive method for coupling light in and out of the integrated optical waveguide. 3.1.4. SPR sensors based on optical fibers. Direct excitation of surface plasma waves by modes of optical fibers presents an interesting approach to the development of miniature SPR sensing devices. The first SPR sensor based on a multimode optical fiber was reported by Jorgenson in 1993 (Jorgenson and Yee, 1993). This wavelength modulation-based SPR sensor uses a conventional polymer clad silica (PCS) fiber with partly removed cladding and a metal film deposited symmetrically around the exposed section of fiber core. This approach yields a miniature optical fiber SPR probe with an interaction area about 10 mm long. This sensor is able to detect refractive index variations with a resolution up to 5• .5 RIU within its operating range of 1.2 - 1.4 RIU (Jorgenson and Yee, 1993). The operating range of the sensor may be adjusted for sensor applications in the refractive index range 1 - 1.7 RIU by using a thin high-refractive index

234

Surface Plasmon Resonance Biosensors

Figure 27. SPR probe based on a single-mode optical fiber. (Reproduced from Slavik et al., 1998, with kind permission from Elsevier Science.)

dielectric overlayer and high refractive index core fibers (Jorgenson and Yee, 1994) (Figure 26). A similar geometry, in which the sensing area is formed not at the tip but in the middle of an optical fiber, has.been used for the development of an intensity modulation-based SPR sensor (Trouillet et al., 1996; Ronot-Trioli et al., 1996). In this configuration, a collimated monochromatic light beam is launched into a fiber in such a fashion that only modes with propagation constants within a narrow range are efficiently excited. Variations in the refractive index of sample are determined by measuring the transmitted optical power. The sensitivity of this SPR sensor is somewhat reduced because numerous fiber modes, which are incident on the metal surface at slightly different angles, take part in the excitation of the SPW. The reported sensor resolutions of 8x10 5 RIU for a goldbased sensor and of 5x104 RIU for a silver-based sensor are good owing to a relatively low system noise level, which allows for the resolution of intensity changes as small as 0.2 percent (Trouillet et al., 1996; Ronot-Trioli et al., 1996). SPR sensors based on single-mode optical fibers have also been reported (Dessy and Bender, 1994; Homola, 1995a). These sensors employ an optical fiber with a locally removed cladding and an SPW-active metal film. A guided mode propagates in the fiber and excites an SPW at the outer surface of the metal film if the two modes are closely phase-matched. This sensor can be designed to operate as a transmissive sensor (Homola, 1995a) or a fiber optic probe (SlavN et al., 1998) (Figure 27). An intensity-modulated SPR sensor of this design has been demonstrated to exhibit sensitivity of 3900 dB/RIU in the refractive index range 1.4193-1.4104 and 2300 dB/RIU in the operating range 1.3302-1.3422 which correspond to a resolution better than 2x104 RIU, if variations in light intensity of 1 percent can be resolved (Slavl"k et al., 1999). A wavelength modulated-version of this SPR sensor has been shown to attain a refractive index sensitivity of 3100 nm/RIU and a resolution of 5X10 7 RIU (Slav~ et al., 2001). Performance of this sensor is however limited by the sensor's sensitivity to the polarization state of the fiber mode in the interaction region of the sensor, which limits its actual resolution to 3x104 RIU. This sensitivity may be dramatically reduced by employing polarization-maintaining fibers (Homola et al., 2001b) 235

Homola, Yee, and Myszka yielding sensor resolutions of 2x10 5 RIU and 2x10 6 RIU for the SPR sensors with intensity and wavelength modulations, respectively. SPR fiber optic sensors present the highest degree of miniaturization of SPR sensor technology. The main challenge for SPR fiber optic sensors remains the development of multichannel devices with referencing capabilities. 3.2. Biosensor surface chemistries

Various biomolecular interactions have been used in SPR biosensors. These include antigen-antibody, hormone-receptor, protein-protein, DNA-DNA, and DNA-protein interactions. An essential requirement of SPR biosensors is that one of the interacting partners must be immobilized to the sensor surface. The first SPR biosensors were functionalized by simple surface adsorption of antibodies on a gold film (Liedberg et al., 1983). However, attaching biomolecules such as proteins directly to metal surfaces often denatures the molecule and leads to a loss of binding activity. Moreover, this approach provides no control over the orientation of the antibodies. Several immobilization chemistries that provide desired chemical properties for stable and defined binding of ligands have been developed. One approach is based on covalent attachment of the biological recognition element to the metal film via a linker layer. This can be performed by forming a self-assembled monolayer of thiol molecules (e.g., alkanethiols) with suitable reactive groups on one end of the molecule and a gold-complexing thiol on the other (Duschl et al., 1996). Then the recognition element can be attached to the thiols, forming a monolayer of ligand molecules. Another approach uses a hydrogel matrix composed of carboxylmethylated dextran chains to yield a two-dimensional matrix for ligand attachment (L6fhs et al., 1990). The matrix provides a hydrophilic environment conducive to maintaining the activity of biomolecules. The non-crosslinked dextran is also thought to maintain much of the entropic properties of the immobilized macromolecule and allow for diffusion within the local environment. In addition, the matrix increases the sensitivity of the SPR biosensor by providing an extended volume within the field of the surface plasma wave. Carboxyl groups on the dextran are easily modified using standard coupling chemistries allowing proteins to be attached via surface-exposed amine, carboxyl, sulfhydryl, and aldehyde groups. Modified versions of dextran have also been used for different applications, including shorter dextran to characterize larger molecular weight analytes and lower-charge dextran to reduce nonspecific binding caused by ionic interactions. Alternatively, metal surfaces may be functionalized by thin polymer films to which ligands may by coupled via amino groups (Nakamura et al., 1997). A rapidly evolving field of biosensor applications involves the use of membrane or lipid surfaces as the binding substrate. In fact, the solid phase nature of biosensor technology makes it ideal

236

Surface Plasmon Resonance Biosensors

Figure 28. Surface plasmon resonance biosensing- main types of detection formats.

for characterizing the interactions of molecules with membrane surfaces. Selfassembled lipid layers are routinely generated using gold-thiol coupling chemistry, but additional surface designs have been used to create stable hybrid lipid bilayers as well as suspended bilayer systems (Cooper and Williams, 1999). Several groups have used these surfaces to measure specific interactions of proteins with phopholipid head groups (Santagata et al., 2001; Evans and MacKenzie, 1999) as well as to characterize lateral interactions between proteins bound to a lipid interface (Lahiri et al., 1999). 3.3. Applications of surface plasmon resonance biosensors There are two major application areas for SPR biosensing devices -detection of biological analytes and biophysical analysis of biomolecular interactions.

3.3.1. SPR biosensor-based detection of bioanalytes. To date, SPR biosensors have been demonstrated for detection of various analytes. Various detection formats have been employed in SPR biosensing in order to optimize performance of SPR biosensors for specific applications. The main detection formats include direct, sandwich and competitive assays (Figure 28). In the direct detection format, a sample containing target molecules is brought into contact with the sensor surface coated with respective biomolecular recognition elements (e.g., antibodies). Binding of analyte molecules to antibodies produces an increase in the refractive index at the sensor surface. The SPR sensor instrument translates this change into a change in sensor response. A sandwich assay consists of two steps. In the first step, the analyte molecules bind to antibodies immobilized on the sensor surface as in the direct detection format. In the second step, the sensor is incubated with a solution containing secondary antibodies, which bind to the previously captured analyte, enhancing the specific sensor response. Another two-step detection format is based on competitive reaction of analyte in solution and analyte immobilized on the surface with free antibodies. In the first step of the competitive assay, a sample is mixed with antibodies. Analyte molecules in 237

Homola, Yee, and Myszka the sample bind to antibodies and block their binding sites. Then, the sample with antibodies is brought to sensor surface so that the unbound antibodies can bind to analyte molecules immobilized on the sensor surface. In this type of assay, the SPR biosensor measures the concentration of unbound antibodies, which can be used to calculate the concentration of target analyte. In general, the choice of detection format depends on the specifics of the application (size of target analyte molecules, binding characteristics of available biomolecular recognition element, range of analytes concentrations to be measured). Direct detection is usually preferred in applications when binding of analyte at appropriate concentrations produces a sufficient direct response. If necessary, the lowest detection limits of the direct SPR sensors can be improved by using a sandwich assay. The secondary antibodies may also be coupled to large particles such as latex particles (Severs, 1993) and gold beads (Leung, 1994) to further enhance the SPR sensor response Competitive assays are used in SPR biosensor-based detection of low concentrations of small analytes, where direct binding does not produce a measurable sensor response. Small molecules that have been detected by SPR biosensors include the herbicides atrazine (Minunni and Mascini, 1993) and simazine (Mouvet et al., 1997; Harris et al., 1999), and the drugs morphine (Sakai et al., 1998) and methamphetamine (Sakai et al., 1999). Minunni and Mascini used an SPR sensor and competitive (binding inhibition) assay to detect atrazine. Monoclonal antibodies against atrazine were mixed with the sample containing atrazine and the free antibody concentration was determined by exposing the sample to the atrazine derivative-coated SPR biosensor. A detection limit of 0.05 ng/ml was achieved (Minunni and Mascini, 1993). Simazine detection in water samples was demonstrated using an integrated optical SPR sensor and binding inhibition assay (Mouvet et al., 1997; Harris et al., 1999). The lowest detection limit for simazine was determined to be 0.1 ng/ml (Harris et al., 1999). Morphine detection based on a binding inhibition assay was reported by Miura et al. (1997), who detected morphine in concentrations down to 0.1 ng/ml. Sakai et al. (1999) developed an SPR biosensor for detection of methamphetamine. Figure 29 shows the calibration curve for their SPR biosensor and methamphetamine concentrations ranging from 0.1 ng/ml to 10 ~tg/ml. As their SPR biosensor uses a binding inhibition assay and the unbound antibodies rather than the molecules of methamphetamine are detected by the sensor, the sensor response is inversely proportional to methamphetamine concentration. The lowest detection limit of the SPR biosensor was 0.1 ng/ml (Sakai et al., 1999). Examples of medium size molecules that have been detected by SPR biosensors are staphylococcal enterotoxin B (Rasooly, 2001; Homola et al., 2001c) and human choriogonadotropin (Dost~ilek et al., 2001). Detection of staphylococcal enterotoxin B was performed using both angular (Rasooly, 2001) and wavelength

238

Surface P l a s m o n R e s o n a n c e Biosensors

0.20

0.15

0.I0

0.05

0 -!

10o

101

102

103

l04

M A concentration I ppb

Figure 29. Detection of methamphetamine (MA) using SPR biosensor and binding inhibition assay. Coupling angle change as a function of MA concentration; anti-MA antibody concentration- 20 ppm. (Reproduced Sakai et al., 1999, with kind permission from Elsevier Science).

12

1.2

9 - Direct response (PBS) ,---, 10 . --O--- - Amplified response (PBS) E .= r cO D.

cn

1.0

8

0.8

6

0.6

4

0.4

2

0.2

Q.

E <

'r . . . . .

-0.0 i,|l

|

1

i

i

~ t,l,,!l

.

.

.

.

lO

S E B concentration [ng/ml]

Figure 30. Detection of staphylococcal enterotoxin B (SEB) using SPR biosensor. Equilibrium SPR sensor response to different concentrations of SEB in buffer (PBS), direct and sandwich detection modes (Homola et al., 2001c).

(Homola et al., 2001c) modulation-based SPR sensing devices. The lowest achieved detection limits were 10 ng/ml (Rasooly, 2001) and 5 ng/ml and 0.5 ng/ml for direct detection mode and sandwich assay mode, respectively (Homola et al., 2001c) (Figure 30). SPR biosensor-based detection, followed by matrixassisted laser desorption/ionization time-of-flight mass spectrometry, made it possible to detect staphylococcal enterotoxin B in milk and mushroom samples at levels of 1 ng/ml (Nedelkov et al., 2000). Direct detection of human 239

Homola, Yee, and Myszka

10000

9000 T

'

ee00 9o0o, ,'-,

6000

~,"

.r

$r

Ig

~m

Start Inject -

zo00

Ignd l mJ _eel_] | |

(llneterln)

1 0 ~ 84

~ .

,

.

.

.

.

=**

.

.

i

,N

m

i

..o

Sl=rt inject w,~ (Antibody)

n,

,,,

m

,o~

_

tin

-~--

i~

,

j,

i.,

jso0

Time (sec)

Figure 31. Detection of Escherichia coli O157:H7 using SPR biosensor and sandwich assay. Curve 1 represents sensor response to E.coli, 2 - S. typhimurium, 3- Y. enterocolitica; bacteria concentration is about 5x 109. (Reproduced from Fratamico et al., 1998, With kind permission from Kluwer Academic Publishers.)

choriogonadotropin by means of an integrated optical SPR sensor was demonstrated by Dost~ilek et al. (2001). The lowest detection limit was 2 ng/ml. Bacterial pathogens Escherichia coli O157:H7, Salmonella enteritidis and Listeria monocytogenes are examples of large analytes that have been targeted by SRP biosensor technology. Detection of E. coli O157:H7 was performed by Fratamico et al. (1998), who used a Biacore SPR sensor and sandwich assay. They used monoclonal antibodies immobilized on the sensor surface for capturing E. coli and polyclonal secondary antibodies for enhancing the specific sensor response (Figure 31). The lowest detection limit for E. coli was established at 5x107 cfu/ml. Direct detection of S. enteritidis and L. monocytogenes at concentrations down to 106 cfu/ml was demonstrated by Koubov~i et al. (2001). In order to optimize detection limits, SPR biosensors usually use biomolecular recognition elements exhibiting rather high affinities to the target analyte. However, the strong interaction between the biomolecular recognition elements and target analyte does not allow the analyte to dissociate if the concentration of analyte in the sample drops. Thus, such biosensors cannot perform continuous monitoring of concentration of the target. The feasibility of SPR biosensors for continuous monitoring was investigated by Ohlson et al. (2000) who demonstrated continuous detection of biopharmaceuticals and antibody titres in patient sera using used low-affinity antibodies. 240

Surface Plasmon Resonance Biosensors

Figure 32. Binding of 4-carboxy-benzenesulfonamide to carbonic anhydrase II immobilized on a BIACORE 2000. Triplicate injections of 4-carboxybenzenesulfonamide were made from a concentration of 20 uM to 40 rdvl in two fold dilutions. The experimental data (black lines) were globally fit to a simple A+B=AB mechanism.

3.3.2. SPR biosensors for biophysical analysis of biomolecular interactions. The key advantages of SPR biosensors are the lack of labeling requirements and the ability to monitor binding reactions in real-time. These advantages allow the analysis of interaction of nearly any binding system, and it is therefore no surprise that SPR biosensors have become a mainstay of both life science and pharmaceutical research. Biosensors can be used to measure binding kinetics directly, including association rates (ka) and dissociation rates (lq) as well as the equilibrium constant (KD) for complex formation. Instruments that can collect binding data at a variety of temperatures make it possible to extract the temperature dependence of the rate constants, which provides detailed information about the enthalpic (AH) and entropic (AS) thermodynamic parameters for the transition state and complex (Roos et al., 1999; Myszka, 2000). The quality of data available from biosensors is high enough to support global data analysis as a method of extracting rate constants (Morton, 1998). In global analysis, the association and dissociation phase data acquired for a set of analyte concentrations over the same ligand surface are fit simultaneously using one set of rate constants. Global fitting results in a more robust evaluation of the shared parameter values, which should be independent of the concentration of the analyte and the surface density of the immobilized ligand. The statistical behavior of the parameter estimates is also improved (Morton et al., 1995). For example, Figure 32 shows the binding data collected for the small molecule 4carboxy-benzenesulfonamide interacting with carbonic anhydrase II immobilized on to the surface of a B IACORE 2000 biosensor. The experimental data were globally fit to a simple 1:1 interaction model (A+B=AB) to determine the reaction rate constants (ka = 4.8 +_.0.2 x 104 M'ls 1 and kd = 0.0365 +__0.0006 s-l). The kinetic rate constants provide more detailed information about a binding 241

Homola, Yee, and Myszka interaction than equilibrium constants (Ko). For example, two different systems may have the same affinity, but vastly different rate constants, giving them different biological properties. Association and dissociation rate constants that can be measured withcurrent SPR biosensor instruments (Biacore, 2001) range from lxl03 to lxl07 M~s ~ and 1 to lxl0 6 s~, respectively, allowing the determination of equilibrium dissociation constants (KD) in the range of i mM to 1 pM. From a practical standpoint, commercial biosensors have been most often used to characterize the binding activity of monoclonal antibodies. Biopharmaceutical companies routinely use SPR biosensor technology to determine the kinetics of antibody/antigen interactions during initial screening as well as to provide quality control during production. B iosensor analysis has also become an important tool for characterizing engineered proteins and for identifying the roles of side chains within protein interfaces. Examples include mapping ligand/receptor interfaces to identify hot spots of activity (Pearce et al., 1996) and dissecting the roles of residues in complex association and dissociation (Katsamba et al., 2001). While kinetic analysis has been a major emphasis of biosensor use, there are other important applications. B iosensors are often used in a qualitative format to identify overlapping binding sites for antibodies. Typically, these epitope mapping experiments are done by capturing an analyte on a surface using one antibody and testing if other antibodies are capable of recognizing the bound antigen (Daiss and Scalice, 1994). In addition, because reactants can be added sequentially in time, it is possible to characterize molecular assembly processes to identify the order of association of subunits in a large complex. There are also examples of using SPR to monitor the refolding of a protein after acid denaturation. A change in response is observed as an immobilized protein transitions from an unfolded to a folded state due to changes in its dielectric constant (Sota et al., 1998).

4. Advantages and Limitations of SPR Biosensors

SPR biosensors belong to a family of thin film refractometry-based sensors such as the grating coupler (Nellen and Lukosz, 1990), the resonant mirror (Buckle et al., 1993) (Chapter 8), and the integrated optical interferometer (Heideman et al., 1993) (Chapter 9), which all measure refractive index changes produced by biomolecular interactions occurring at the surface of the sensor. These sensors exhibit similar fundamental advantages and limitations. The main advantage of these approaches is their ability to detect molecular interactions directly without the use of radioactive or fluorescent labels, which makes it possible to observe biomolecular interactions in 'real time'. Another advantage is the versatility of these techniques - analyte molecules do not have to exhibit any special properties such as fluorescence or characteristic absorption and scattering bands. The main 242

Surface Plasmon Resonance Biosensors limitation of these biosensors is in the specificity of detection, which is solely based on the ability of the biomolecular recognition coating to recognize and capture target analyte molecules while preventing the non-target molecules from adsorbing on the sensor surface, which would produce a false refractive index change resulting in false sensor response. Currently, SPR biosensors present the most advanced thin film refractometric biosensor technology, both in terms of development of sensor instrumentation and applications. A large variety of SPR sensor platforms - from miniature SPR fiber optic probes to robust bench-top laboratory units - is available to meet special requirements of a multitude of biosensing applications. Existing SPR biosensors are somewhat limited in terms of number of sensing channels they can support, which restricts their ability to analyze large number of samples in parallel fashion or simultaneously detect multiple analytes in complex samples. SPR biosensor arrays based on SPR imaging can overcome this limitation; however, current SPR imaging instruments rely on intensity modulation which makes their resolution inferior to SPR sensors based on angular and wavelength modulation. The best SPR sensor systems exhibit refractive index resolution in the low 10.7 RIU range, making it possible to detect small and medium size analytes in concentrations well below 1 ng/ml, which is sufficient for a wide range of analyses. For example, minimum lethal doses for most potent proteinaceous toxins are about 100 ng (Gill, 1982). The ability of SPR biosensors to detect larger analytes such are bacteria and cells are, however, less satisfactory, with detection limits on the order of 10 6 - 10 7 cfu/ml. Given infectious doses for many bacterial pathogens ranging from 1 0 - 104 cells (Ryan, 1990), the current detection limits obtainable with SPR biosensors are not adequate. The main factors limiting the ability of SPR biosensors to detect bacteria include: the size of a bacterium (for a bacterium of typical size of ~ 1 ktm, the bulk of the bound cell is situated outside the evanescent field of an SPW), low concentration of the target antigen (relative to total cellular material), and slow diffusion of bacterial cells to the sensor surface (Perkins and Squirrell, 2000). In the field of biophysical analysis of biomolecular interactions, SPR biosensors allow for monitoring of both weak and strong interactions (KD ranging from 1 mM to 1 pM). Compared to traditional biophysical techniques which are used to measure binding interactions (titration calorimetry, ultracentrifugation, stopped flow, and column chromatography), SPR biosensors typically require less material and have a higher throughput.

243

Homola, Yee, and Myszka

5. Potential for Expanding Current Capabilities Improvements in the detection capabilities of SPR biosensors are desired to enable direct detection of biomolecular interactions involving small molecules or extremely low analyte concentrations. In order to accomplish this goal, approaches to improving sensitivity and resolution of SPR sensor instruments are being developed. One promising approach to the development of high-sensitivity SPR sensors is based on the exploitation of phase properties of light coupled to SPW. A phase modulation SPR sensor has been developed by Nelson et al., (1996), who measured the phase of the light wave using the optical heterodyne technique. This approach demonstrated the potential of phase-modulated SPR sensors; however, it resulted in a rather bulky and complex instrument requiring an acousto-optic modulator and sophisticated data processing. Another interesting approach to enhancing sensitivity of wavelength modulation SPR sensors is based on altering dispersion of the SPW to match that of the light wave exciting the SPW (see Figure 16). This can be accomplished by employing longrange surface plasma waves (LRSPW) (Nenninger et al., 2001), which exhibit lower dispersion than classical SPWs supported by a metal-dielectric interface (see Figures 3 and 6). As the long-range surface plasma waves penetrate deeper into the probed sample, the use of these LRSPW will benefit biosensor geometries in which the biomolecular interaction to be observed occurs within rather thick layers (e.g., in systems using extended coupling matrices (Liedberg et al., 1993)). Besides yielding the sensitivity enhancement, long-range surface plasmons produce very narrow angular or spectral dips more than 10-fold narrower than those produced by conventional SPWs (Figure 12). This makes it possible to determine the spectral position of the SPR dip with a high accuracy, and thus yields an improved resolution for sensors with wavelength and angular modulation. The resolution of SPR sensors can also be improved by optimization of sensor data processing. Noise in SPR sensing devices (Kolomenskii et al., 1997) and its propagation through various sensor data processing algorithms are being investigated (Johnston et al., 1997; Chinowsky et al., 1999a; Johansen et al., 2000b). This research effort is expected to lead to optimized sensor data processing methods yielding sensor output with highest signal-to-noise ratios and aid design optimization of SPR sensing devices. Referencing approaches are being investigated for improving the specificity of SPR biosensors. In conventional SPR systems for biophysical analysis of biomolecular interactions, one of the 'parallel' sensing channels (specific channel) is typically treated with ligand binding analyte, while the other (reference channel) is treated with the coupling chemistry in the absence of added ligand or coupled with a protein that is known not to interact with the analyte under analysis. Although such sensors deal with pure samples in controlled laboratory environments, changes in the background refractive index introduced by exchanging buffers or by injecting high concentrations of analyte in the study 244

Surface Plasmon Resonance Biosensors of low-affinity interactions cannot be entirely compensated. The uncompensated non-specific response (typically several per cent of the background refractive index change), in particular the transient part following introduction of the background refractive index change, can cause errors in the determination of kinetic constants of biomolecular interactions (Ober and Ward, 1999). Compensation for non-specific effects is even more crucial in SPR biosensors for detection of bioanalytes in crude samples and out-of-laboratory environments. Development of robust referencing methods for discriminating the specific response due to the binding of analyte from the non-specific response due to adsorption of non-target molecules on the sensor surface or sample refractive index variations remains one of the main challenges in development of practical biosensors for detection of bioanalytes in the field. SPR biosensor-based detection of bioanalytes with compensation for sample refractive index variations has been demonstrated in several configurations, including an SPR sensor with two parallel sensing channels (Karlsson and Sthhleberg, 1995; O'Brien et al., 1999; Ober and Ward, 1999), an SPR sensor with two serial sensing channels (Homola et al., 2001a), and an SPR sensor combined with a critical angle refractometer (Chinowsky et al., 1999b), providing background refractive compensation of 95-99 percent. The search for methods for simultaneous compensation for non-specific binding and background refractive index variations, allowing for application of SPR biosensors to analysis of crude samples continues (Homola et al., 200 ld). Approaches to multichannel SPR sensing are under investigation to enhance throughput of SPR sensors and provide them with multi-analyte detection capability. Advances in miniaturization of SPR sensing devices enable the number of sensing channels in conventional parallel-channel SPR sensors to be expanded. Recently, a prototype of an eight-channel SPR sensor has been developed by Biacore and tested by a consortium of investigators referred to as Foodsense (2001). Traditional angle and wavelength modulated SPR sensors can benefit from combining.paraUel channel, architecture, (Figure 20)(Nenninger et al., 1998; Lt~fhs et al., 1991) with the spectral discrimination of sensing channels (Figure 21) (Homola et al., 1999b, 2001a), which can double the number of sensing channels. Sensors based on SPR imaging (Thiel et al., 1997) can provide a large number of sensing channels. Scientists at Biacore have recently demonstrated the ability to immobilize 315 protein spots on a sensor chip (Figure 33). Current work is focused on developing instrumentation able to read SPR responses from each spot independently and simultaneously. From biological application standpoint, the field of membrane biology could benefit from advances in immobilization methods. There is a tremendous need to develop methods incorporating specific transmembrane receptors, such as seventransmembrane systems, onto the biosensor surface. Recently, Salamon et al. (2000) have reconstituted transmembrane receptors on synthetic lipid layers over

245

Homola, Yee, and Myszka

Figure 33. Prototype protein array chip from Biacore. 315 discreet spots of protein were immobilized to create the protein array. Figure provided by Biacore.

a silver surface. Using a plasmon-coupled waveguide sensor, they were able to detect specific conformational states of the receptor upon the binding of various agonist and antagonist ligands. Routine methods for studying membraneassociated receptors would open up entirely new avenues to study the interactions of these receptors in a more native environment. Advances in the development of new SPR biosensor hardware will be driven by the needs of the consumer. A large untapped market for biosensors lies in industries that require high sensitivity as well as high throughput. Food and environmental analysis could benefit greatly from the real-time aspects of biosensor analysis, once the throughput has been increased. The pharmaceutical industry, which was fast to adopt optical biosensors as tools to characterize biopharmaceutical products, is beginning to implement the technology in small molecule screening, as well as general bioavailability analysis. SPR biosensors could also play an important role in defense, where fast, portable and durable units are needed for early detection and identification of hazardous agents in the field. Development of these systems will require significant advances in miniaturizing detection systems, coupling them with microfluidics as well as specialized sampling systems. Given their extremely wide capabilities and everevolving applications, we envision the use of SPR biosensor technology will continue to expand as a modern bioanalytical tool.

6. References Berger, C. E. H., T. A. M. Baumer, R. P. H. Kooyman and J. Greve, 1998, Anal. Chem. 70, 703. Biacore, 2001, Biacore website: http://www.biacore.com. Boardman, A.D., 1982, Electromagnetic Surface Modes, John Wiley & Sons, Chichester, pp. 1-76. Buckle, P. E., R. J. Davies, T. Kinning, D. Yeung, P. R. Edwards, D. PollardKnight and C. R. Lowe, 1993, Biosens. Bioelectron. 8, 355. Burke, J. J., G. I. Stegeman and T. Tamir, 1986, Phys. Rev. B 33, 5186. Cahill, C. P., K. S. Johnston and S. S. Yee, 1997, Sens. Actuators B 45, 161. 246

Surface Plasmon Resonance Biosensors Chandezon, J., M. T. Dupuis, G. Comet and D. Maystre, 1982, J. Opt. Soc. Am. 72, 839. Chinowsky, T.M., L. S. Jung and S. S. Yee, 1999a, Sens. Actuators B 54, 89. Chinowsky, T. M., A. Strong, D. U. Bartholomew, S. Jorgensen, T. Notides, C. E. Furlong and S. S. Yee, 1999b, Proc. SPIE 3857, 104. Choi, K., W. Seo, S. Cha and J. Choi, 1998, J. Biochem. Mol. Biol. 31,101. Christensen L. L., 1997, Anal. Bio~hem. 249, 153. Cooper M. A. and D. H. Williams, 1999, Anal. Biochem. 276, 36. Ctyrok3~, J., J. Homola, and M. Skalsk3~, 1997, Electr. Lett. 33, 1246. Cullen, D.C., R. G. Brown and C. R. Lowe, 1987/88, Biosens. Bioelectron. 3, 211. Cullen, D.C. and C. R. Lowe, 1990, Sens. Actuators B 1,576. Daiss, J. L. and E. R. Scalice, 1994, Methods 6, 143. Danelian E, A. Karl6n, R. Karlsson, S. Winiwarter, A. Hansson, S. Lt~f~s, H. Lennem~is and M. D. H~im~il~iinen, 2000, J. Med. Chem. 43, 2083. de Bruijn, H. E., B. S. F. Altenburg, R. P. H. Kooyman and J. Greve, 1992, App. Opt. 31,440. Dessy, R. E. and W. J. Bender, 1994, Anal. Chem. 66, 963. Dost~ilek, J., J. Ctyrok2~, J. Homola, E. Brynda, M. Skalsk3~, P. Nekvindov~i, J. ~pirkov~i, J. ~kvor and J. Schri3fel, 2001, Sens. Actuators B 76, 8. Duschl, C., A. Sevin-Landais and H.Vogel, 1995, Biophys. J. 70, 1985. Elkind, J: L., D. I. Stimpson, A. A. Strong, D. U. Bartholomew and J. L. Melendez, 1999, Sens. Actuators B 54, 182. Evans, S. V. and C. R. MacKenzie, 1999, J. Mol. Recogn. 12, 155. Fano, U., 1941, J. Opt. Soc. Am. 31, 213. Foodsense, 2001, Foodsense website: http://www.slv.se/foodsense2. Foster, M. W., D. J. Ferrel and R. A. Lieberman, 1994, Proc. SPIE 2293, 122. Fratamico, P. M., T. P. Strobaugh, M. B. Medina and A. G. Gehring, 1998, Biotechnol. Tech. 12, 571. Frostell-Karlsson A., A. Remaeus, H. Roos, K. Andersson, P. Borg, M. H~im~ilainen and R. Karlsson, 2000, J. Med. Chem. 43, 1986. Gale, M. T., 1997, Microelectron. Eng. 34, 321. Gaus, K. and E. A. H. Hall, 1998, Biosens. Bioelectron. 13, 1307. Gill, D.M., 1982, Microbiol. Rev. 46, 86. Gordon II, J. G. and S. Ernst, 1980, Surf. Sci. 101,499. Harris, R. D. and J. S. Wilkinson, 1995, Sens. Actuators B29, 261. Harris, R. D., B. J. Luff, J. S. Wilkinson, J. Piehler, A. Brecht, G. Gauglitz and R. A. Abuknesha, 1999, Biosens. Bioelectron. 14, 377. Heideman R. G., R. P. H. Kooyman and J. Greve, 1993, Sens. Actuators B 10, 209. Homola, J., 1995a, Sens. Actuators B 29, 401. Homola, J., G. Schwotzer, H. Lehmann, R. Willsch, W. Ecke and H. Bartelt, 1995b, Proc. SPIE 2508, 324. Homola, J., 1997a, Sens. Actuators B 41,207.

247

Homola, Yee, and Myszka Homola, J., J. t~tyrok3~,M. Skalsk3~, J. Hradilov~i and P. Kolfi~ov~i, 1997b, Sens. Actuators B 38-39, 286. Homola, J., I. Koudela and S. Yee, 1999a, Sens. Actuators B 54, 16. Homola, J., H. B. Lu and S. S. Yee, 1999b, Electr. Lett. 35, 1105. Homola, J., J. Dostfilek and J. t~tyrok3~, 200 l a, Proc. SPIE 4416, 86. Homola, J., M. Piliarik, R. Slavfk and J. Ctyrok2~, 2001b, Proc. SPIE 4416, 82. Homola, J., J. Dostfilek, S. Chen, A. Rasooly, S. Jiang and S. S. Yee, 200 l c, Proc. SPIE 4416, 280. Homola, J., H. B. Lu, G. G. Nenninger J. Dostfilek and S. S. Yee, 200 ld, Sens. Actuators B 76, 403. Hutley, M. C., 1982, Diffration Gratings, Academic Press, London. Johansen, K., R. Stalberg, I. Lundstrom and B. Liedberg, 2000a, Meas. Sci. Technol. 11, 1630. Johansen, K., H. Arwin, I. Lundstrom and B. Liedberg, 2000b, Rev. Sci. Inst. 71, 3530. Johnston, K. S., S. S. Yee and K. S. Booksh, 1997, Anal. Chem. 69, 1844. JOnsson U, et al., 1991, Biotech. 11,620. Jorgenson, R. C. and S. S. Yee, 1993, Sens. Actuators B 12, 213. Jorgenson, R.C. and S. S. Yee, 1994, Sens. Actuators A 43, 44. Jory, M. J., P. S. Vukusic and J. R. Sambles, 1994, Sens. Actuators B 17, 1203. Jory, M. J., G. W. Bradberry, P.S. Cann and J. R. Sambles, 1995, Meas. Sci. Technol. 6, 1193. Karlsen, S., R., K. S. Johnston, S. S. Yee and C. C. Jung, 1996, Sens. Actuators B 32, 137. Karlsson, R. and R. Stfihleberg, 1995, Anal. B iochem. 228, 274. Katsamba P. S., D. G. Myszka and I. A. Laird-Offringa, 2001, J. Biol. Chem. 276, 21476. Kolomenskii, A. A., P. D. Gershon and H. A. Schuessler, 1997, Appl. Opt. 36, 6539. Kooyman, R. P. H., H. Kolkman, J. van Gent and J. Greve, 1988, Anal. Chim. Acta 213, 35. KoubovL V., E. Brynda, L. KarasovL J. ,~kvor, J. Homola, J. Dostfilek, P. Tobi~ka and J. Ro~ick~, 2001, Sens. Actuators B 74, 100. Kretschmann, E. and Raether, H., 1968, Z. Naturforsch. 23A, 2135. Kreuwel, H., P. Lambeck, J. V. Gent andT. Popma, 1987, Proc. SPIE 798, 218. Kruchinin, A. A. and Y. G. Vlasov, 1996, Sens. Actuators B 30, 77. Lahriri, J., L. Isaacs, J. Tien and G. M. Whitesides, 1999, Anal. Chem. 71,777 Lambeck, P.V., 1992, Sens. Actuators B 8, 103. Lavers, C. R. and J. S. Wilkinson, 1994, Sens. Actuators B 22, 75. Lawrence, C. R., N. J. Geddes, D. N. Furlong and J. R. Sambles, 1996, Biosen. Bioelectron. 11,389. Leung, P.T., D. Pollard-Knight, G. P. Malan and M. F. Finlan, 1994, Sens. Actuators B 22, 175. Liedberg, B., C. Nylander and I. Lundstr6m, 1983, Sens. Actuators 4, 299. Liedberg, B., I. Lundstrom and E. Stenberg, 1993, Sens. Actuators B 11, 63. 248

Surface Plasmon Resonance Biosensors LOfgts, S., M. Malmqvist, I. R0nnberg, E. Stenberg, B. Liedberg and I. Lundstrt~m, 1991, Sens. Actuators B 5, 79. Matsubara, K., S. Kawata and S. Minami, 1988, Appl. Opt. 27, 1160. Melendez, J., R. Carr, D. U. Bartholomew, K. Kukaskis, J. Elkind, S. S. Yee, C. Furlong and R. Woodbury, 1996, Sens. Actuators B 35-36, 212. Minunni, M. and M. Mascini, 1993, Anal. Lett. 26, 1441. Miura, N, K. Ogata, G. Sakai; T. Uda and N. Yamazoe, 1997, Chem. Lett. 8, 713. Moharam, M. G. and T. K. Gaylord, 1986, J. Opt. Soc. Am. A3, 1780. Morton T. A., D. G. Myszka and I. M. Chaiken, 1995, Anal. Biochem. 227, 176. Morton, T. A. and D. G. Myszka, 1998, Meth. Enzymol. 295,268. Mouvet, C., R. D. Harris, C. Maciag, B. J. Luff, J. S. Wilkinson, J. Piehler, A. Brecht, G. Gauglitz, R. Abuknesha and G. Ismail, 1997, Anal. Chim. Acta 338, 109. Myszka, D. G, 2000, Meth. Enzymol. 323,325. Nakamura, R., H. Muguruma, K. Ikebukuro, S. Sasaki, R. Nagata, I. Karube and H. Pedersen, 1997, Anal. Chem. 69, 4649. Nedelkov, D., A. Rasooly and R. W. Nelson, 2000, Int. J. Food Microb. 60, 1. Nellen, P. M. and W. Lukosz, 1990, Sens. Actuators t3 1,592. Nelson, B. P., A. G. Frutos, J. M. Brockman and R. M. Corn, 1999, Anal. Chem. 71, 3928. Nelson, S.G., K. S. Johnston and S. S.Yee, 1996, Sens. Actuators B35-36, 187. Nenninger, G. G., J. B. Clendenning, C. E. Furlong and S. S. Yee, 1998, Sens. Actuators B 51, 38. Nenninger, G. G., P. Tobi~ka, J. Hornola and S. S. Yee, 2001, Sens. Actuators B 74, 145. Nikitin, P. I. and A. A. Beloglazov, 1994, Sens. Actuators A 41-42, 547. Nikitin, P. I., A. A. Beloglazov, M. V. Valeiko, J. A. Creighton, A. M. Smith, N. A. Sornmerdijk and J. D. Wright, 1997, Sens. Actuators B 38, 53. Nikitin, P.I., A. A. Beloglazov, A. V. Kabashin, M. V. Valeiko and V. E. Kochergin, 1999, Sens. Actuators B 54, 43. Nippon Laser and Electronics Laboratory, 2001, Nippon Laser and Electronics Laboratory website: http://www.nle-lab.co.jp/English/ZO-HOME.htm Nylander, C., B. Liedberg and T. Lind, 1982, Sens. Actuators 3, 79. Ober, R. J. and E. S. Ward, 1999, Anal. Biochem. 271, 70. O'Brien, M. J. II, S. R. J. Brueck, V. H. Perez-Luna, L. M. Tender and G. P. Lopez, 1999, Biosen. Bioelectron. 14, 145. Ohlson, S., C. Jungar, M. Strandh and C. F. Mandenius, 2000, Trends Biotechnol. 18, 49. Ordal, A. M., L. L. Long, R. J. Bell, S. E. Bell, R. R. Bell, R. W. Alexander and C. A. Ward, 1983, Appl. Opt. 11, 1099. Otto, A.,1968, Z. Physik, 216, 398. Palik, E. D., 1985, Handbook of Optical Constants of Solids, Academic Press, London, 804 pp. Parriaux, O. and G. Voirin, 1990, Sens. Actuators A 21-23, 1137.

249

Homola, Yee, and Myszka Pearce K. H. Jr., M. H. Ultsch, R. F. Kelley, A. M. de Vos and J. A. Wells, 1996, Biochem. 35,10300. Perkins, E.A. and D. J. Squirrell, 2000, Biosen. Bioelectron. 14, 853. Pfeifer, P., U. Aldinger, G. Schwotzer, S. Diekmann and P. Steinrticke, 1999, Sens. Actuators B 54, 166. Pockrand, I., 1978, Surf. Sci. 74, 237. Raether, R., 1988, Surface Plasmons on Smooth and Rough Surfaces and on Gratings, Springer- Verlag, Berlin, 136 pp.. Rasooly, A., 2001, J. Food Protect. 64, 37. Ronot-Trioli, C., A. Trouillet, C. Veillas and H. Gagnaire, 1996, Sens. Actuators A 54,589. Roos, H., R. Karlsson, H. Nilshans and A. Persson, 1999, J. Mol. Recognit. 11, 204. Rothenhtiusler, B. and W. Knoll, 1988, Nature 332, 688. Ryan, K.J., 1990 In: Medical Microbiology: An Introduction to Infectuous Diseases, Sherris, J. C., Ed., Elsevier, New York, p. 357. Sakai, G., K. Ogata, T. Uda, N. Miura and N. Yamazoe, 1998, Sens. Actuators B 49,5. Sakai, G., S. Nakata, T. Uda, N. Miura and N. Yamazoe, 1999, Electr. Acta 44, 3849. Salamon Z., S. Cowell, E. Varga, H. Yamamura, V. Hruby and G. Tollin G, 2000, Biophys. J. 79, 2463. Santagata S., T. J. Boggon, C. L. Baird, W. S. Shan, C. Gomez, D. G. Myszka and L. Shapiro, 2001 Science 292, 2019. Severs, A. H. and R. B. M. Schasfoort, 1993, Biosens. Bioelectron. 8, 365. Shimomura, M., Y. Nomura, W. Zhang, M. Sakino, K. H. Lee, K. Ikebukuro and I. Karube., 2001, Anal. Chim. Acta 434, 223. Slav~, R., J. Homola and J. Ctyrok37, 1998, Sens. Actuators B 51, 311. Slav~, R., J. Homola and J. Ctyrok37, 1999, Sens. Actuators B 54, 74. Slav~, R., J. Homola and J. Ctyrok2~, 2001, Sens. Actuators B 74, 106. Snyder, A. W. and J. D. Love, 1983, Optical Waveguide Theory, Chapman and Hall, London, 734 pp. Sota, H., Y. Hasegawa and M. Iwakura, 1998, Anal. Chem. 70, 2019. Stemmler, I., A. Brecht and G. Gauglitz, 1999, Sens. Actuators B 54, 98. Texas Instruments, 2001, Texas Instruments website: http://www.ti.com. Thiel, A. J., A. G. Frutos, C. E. Jordan, R. M. Corn and L. M. Smith, 1997, Anal. Chem. 69, 4948. Trouillet, A., C. Ronot-Trioli, C. Veillas and H. Gagnaire, 1996, Pure App. Opt. 5,227. Vidal, M. M. B., R Lopez, S. Aleggret, J. Alonso-Chamarro, I. Garces and J. Mateo, 1993, Sens. Actuators B 11,455. Vukusic, P. S., G. P. Bryan-Brown and J. R. Sambles, 1992, Sens. Actuators B 8, 155. Weiss, M. N., R. Srivastava, H. Groger, P. Lo and S. F. Luo, 1996, Sens. Actuators A 51, 211. 250

Surface Plasmon Resonance Biosensors Windsor, 2001, Windsor Scientific website: http://www.windsor-ltd.co.uk Wood, R.W., 1902, Phil. Magm. 4, 396. Yeatman, E. M. and A. Ash, 1987, Electron. Lett. 23, 1091. Yeatman, E. M., 1996, Biosens. Bioelectron.ll, 635. Zhang, L. M. and D. Uttamchandani, 1988, Electr. Lett. 23, 1469.

251

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 8 THE RESONANT MIRROR OPTICAL BIOSENSOR

TIM KINNING AND PAUL EDWARDS, PH.D. Thermo Labsystems, Saxon Way, Bar Hill, Cambridgeshire, UK.

The resonant mirror optical biosensor is a waveguide-based sensor that uses the evanescent field to detect refractive index changes on or close to the sensing surface. As such, in theory, any molecular ~nteraction can be followed in real-time without the requirement for labeling. To accomplish this, one of the interactants must be attached to the sensor surface and the binding of its partner must produce a measurable change in refractive index. As there is no significant stipulation on the types of interactions investigated with this biosensor, the literature base is very broad. The uses of this biosensor are reviewed with respect to selected literature published within the last two to three years. Future advances in both the instrumentation and the consumables, together with the limitations of the technology, are also discussed within the chapter

1. Technical Concept The resonant mirror biosensor ( R M ) u s e s an evanescent field to probe interactions occurring within a few hundred nanometers of a surface. These sensors are designed to measure binding in real-time between 'entities', usually, but not exclusively, of biological origin. These 'entities' can be anything from molecules of a few hundred Daltons to whole cells. Measurement selectivity is created by immobilisation of one of the interactants of interest at the surface; the exponential decay of the evanescent field strength prevents interactions in the bulk milieu from having any influence. There are two main types of optical evanescent sensors, those based on surface plasmon resonance (SPR), and those based on waveguiding techniques. SPR sensors (Liedberg et al., 1983) consist of a thin metallic layer (typically Au or Ag) of about 50 nm deposited directly onto a glass substrate. This glass device could be in the form of a prism or grating, or be capable of making intimate contact with such a structure. The measurement principle is based on excitation 253

Kinning and Edwards of surface plasmons at the boundary between the metal film and the sensing layer. As a result, incident light is adsorbed, resulting in a dip in the reflected light intensity. The angle of incidence of the exciting light is extremely sensitive to any changes occurring in the sensing film at the metal-film boundary. Sensing can then be achieved by monitoring the position of the dip in intensity (or intensityat a fixed position) of the reflected light. RM instruments are based on a waveguide structure (RM) (Cush et al. 1993; Davies and Pollard-Knight, 1993), where the evanescent field results from the propagation of light along the waveguide. This waveguide is made from a metal oxide. Changes at the sensing layer, at the top surface of the waveguide, alter the conditions under which the light couples in and travels along the guide. Monitoring these changes with time forms the basis of the measurement, as described in Section 1.1. Unlike SPR devices, the system is looking for a peak of intensity, not a dip. For the RM to be of use, there has to be a means of adding to and removing samples from the sensing surface. In the case of the IAsys RM devices, the RM is mounted into a rnicro-cuvette device to form an integrated experimental unit, as described in Section 1.2. 1.1. The resonant mirror structure

The RM uses the optical phenomenon of an evanescent field to detect changes in refractive index or mass close to the sensor surface. Evanescent fields are generated whenever light undergoes total internal reflection (TIR) at the boundary between high and low refractive index materials. The electric field strength of the evanescent field does not disappear discontinuously, but penetrates approximately one wavelength into the low refractive index layer, decaying exponentially. The evanescent field is enhanced by the resonant structure shown in Figure 1. It is called a resonant mirror because the structure forms a resonant cavity, or waveguide, that is an almost perfect reflector of light. This structure is composed of a low refractive index coupling layer and a high refractive index resonant layer, which acts as both a waveguide and the sensing surface. (Cush et al. 1993; Davies and Pollard-Knight, 1993). At the resonant angle, the evanescent field strength is enhanced, such that light passes from the prism through the low refractive index coupling layer and propagates in the high refractive index waveguide layer at the same angle. The sample forms another low refractive index layer above the high refractive index resonant layer. Therefore, propagating light in the high refractive index waveguide layer will generate an evanescent field extending into the sample. With all optical parameters within the system being fixed, the resonant angle is sensitive to changes in refractive index at the ligand attachment surface. Due to the exponential decay of the evanescent field, only changes close the sensor surface will affect the resonant angle; changes in the bulk will not. To resolve 254

Resonant Mirror Biosensors

Figure 1. Structure of the resonant mirror. Light at the resonant angle (solid line) couples into the resonant structure and propagates along it, emerging out of the structure to fall onto the detector array. Light not at the resonant angle (dotted lines) does not couple into the structure. The light that has travelled within the resonant structure undergoes a phase change, enabling only this light to pass through the output polariser. Adapted from Cush et al. (1993) and Davies and Pollard-Knight (1993).

the resonant angle, the optics are set up such that light that is coupled into the waveguide undergoes a 90 ~ phase change. A polarizer is inserted prior to the detector, such that only light propagated within the waveguide layer, and 'affected' by interactions at the surface, reaches the detector. Light at all other angles is not detected. What is in fact detected is an angle, specifically the angle of the laser beam at the time the light reaches the detector array. Laser light is scanned over an angle of 14 ~ the middle 10 ~ of which is used for detection purposes. The other 4 ~ is used to ensure the laser is moving at a constant speed over the measurement region. An encoder attached to the laser arm enables the position of the arm to be known at any time. As the resonant angle changes with time, the position of the arm at that time is obtained, resulting in a display of angle versus time.

1.2. The sample cuvette Integral to the RM system marketed as the IAsys is the stirred cuvette. The RM block structure (Figure 1) is mounted onto a cuvette structure (Figure 2) such that the sensing surface is exposed to sample contained within the cuvette's lumen. Sample volume within each cuvette chamber is approximately 80 }.tl for the dualwell version and 200 ~tl for the single-well version. After insertion into 255

Kinning and Edwards

Figure 2. Line drawing and cross-section of IAsys twin well cuvette illustrating the position of the two sample cells in respect of the stirrer and the aspiration tubes.

an instrument, a vibro-stirring mechanism and a vacuum aspiration system are placed within each lumen of the cuvette. A biocompatible coating on the cuvette body prevents non-specific binding. Sample or other reagents may be added to each chamber by pipette (via a flap located on the instrument) or by a syringedriven auto-sampler. Samples and other solutions no longer required can be removed by pipette or by a vacuum-driven aspiration system to a waste bottle. Samples required for further analysis can be removed using a pipette or autosampler. The lack of a complicated sample delivery system ensures maximum flexibility to experimental design and reduces the surface area open to non-specific adsorption of samples, minimising carry-over. Furthermore, complex, viscous samples such as serum can be used without any complication arising from blockages of tubing. To ensure efficient mass transport of material to the sensor surface and rapid equilibration of the system (Glaser, 1993), a vibrating stirrer assembly is inserted into the cuvette after placement in the instrument. The vibration-driven stirrer operates using a single motor to ensure that both cells are stirred equally. As the stirrer vibrates, liquid is forced towards the surface of the sample chamber and back up through the centre of the stirrer cone (Figure 3). The stirrer frequency is fixed at 126 Hz. However, adjustments can be made to the amplitude via the instrument operating software. Where flow is an advantage, for example in the monitoring of samples from chromatography columns (Bracewell et al., 1998), the system can be modified such that samples can be continuously flowed into the cuvette via inlet pipes just above the block surface (see Figure 4). A further set of pipes, set at a fixed height above the cuvette surface, enables liquids to be removed via a pump set at a flow rate faster than the in-rate. The difference between the height of the inlet 256

Resonant Mirror Biosensors

Direction of liquid movement

I00

80 &

0,0 t 0

............

/ / ~ / / / /

/ / / / / Block surface

Figure 3. Action of the IAsys vibro-stirrer. The stirrer vibrates at a fixed frequency at an amplitude controlled using the application software. Here the block surface is the sensing surface.

and outlet pipes defines the volume of sample resident within the cuvette. To ensure any sample injected is rapidly mixed within the cell, the stirrers are left in position and allowed to operate in the normal fashion. As refractive index measurements are temperature sensitive, the cuvette is clamped tightly against Peltier-driven thermal pads when inserted in the instrument. Good thermal conductivity is ensured between the Peltier pads and the sample by the configuration of the cuvette body. Once an experiment is completed, the cuvette can be removed for storage in a refrigerator. However, cuvette bodies are also available that enable the RM block to be removed from the cuvette body and remounted later. This arrangement is ideal for developing unique sensor surfaces or for coupling the biosensor to another technique such as atomic force microscopy (Fisher and Tjamhage, 2000).

1.3. Sensor surfaces To ensure the specificity of the measurement, a recognition element or ligand is immobilised at the sensor surface. To facilitate this immobilisation, the RM 257

Kinning and Edwards

Figure 4. Flow cell modification to IAsys instrumentation.

matrix surfaces consist of a layer of immobilised carboxymethylated dextran (CMD), forming a three-dimensional matrix on the surface. The carboxymethyl residues enable simple coupling of amino-containing molecules through succinimidyl ester chemistries. The main advantage of this type of surface is that the 3-dimensional nature of the matrix keeps immobilised ligands away from direct contact with the surface, potentially resulting in greater activity. Furthermore, the hydrophilic nature of the surface generally reduces non-specific binding. There are several disadvantages with this type of surface. First, it is possible to overload the matrix with ligand. Second, particles, cells, or large molecules may be size-excluded. Third, for kinetic experiments, molecules bound to sites farthest from the surface may sterically hinder binding of other molecules to sites within the matrix (Edwards et al., 1995; Schuck, 1996). This disadvantage is further exacerbated by the exponential decay of the evanescent field as one moves away from the surface; molecules binding at the outer limit of the dextran layer may well not be fully 'sensed'. Last, due to the random nature of the immobilisation chemistry, the orientation of immobilised recognition molecules will be not be uniform. Planar surfaces can be derivatised with carboxylate, amino, hydrophobic and biotin functionalities. Carboxylate surfaces, as with CMD surfaces, enable simple coupling of amine-containing molecules through the succinimidyl ester chemistry. Amino surfaces can be activated with cross-linkers such as glutaraldehyde or bis[sulfosuccinimidyl]suberate. Both of these immobilisation chemistries enable large molecules or cells to more closely approach the sensor surface. 258

Resonant Mirror Biosensors Hydrophobic surfaces can be used for the immobilisation of lipids from solution, obviating the requirement for liposomes (Athanassopoulou et al., 1999). The surface is simply exposed to the required mixture of lipids in solvent, the solvent is then diluted out by the addition of buffer, and the lipids are forced to assemble on the surface. Biotin-derivatised surfaces possess the advantages described but have the further advantage of enabling the simple capture of an avidin. Surface-bound avidin is then used to capture biotinylated molecules of interest. Most biotinylation of peptides and nucleic acids results in a single point of attachment of the biotin; thus, when used as the capture entity, all the molecules will have the same orientation, simplifying kinetic interpretation. It is also possible to re-use the surface by completely removing the avidin-captured molecule. This method has been widely used for immobilisation of oligonucleotides and DNA (Grunden et al., 1999; De Zutter et al., 2001), antibodies and other proteins (Charalambous and Feavers, 2000; Salek-Ardakani et al., 2000; Kobayashi et al., 2001; Sheng et al., 2000), and peptides (Holaska et al., 2001). A further variable with the cuvette is the surface area exposed. Where a relatively large amount of material is required, for example in the case of ligand fishing, large surface area cuvettes are available. These have approximately four times the surface area available as the standard versions.

2. History The RM derived out of a UK Department of Trade and Industry Collaborative grant in 1987 to Fisons Instruments, University of Cambridge Institute of Biotechnology and GEC Marconi. After proof of concept, in 1990 Fisons Instruments took on the project in order to create a commercial product from the research. The company set up to exploit the technology was initially known as Fisons Applied Sensor Technology (FAST). FAST became Affinity Sensors in 1996. During 1998, Thermo Electron acquired the majority of Fisons Instruments companies. Internal re-organisations merged the company with Labsystems, finally becoming Thermo Labsystems in 2001. Initial designs used a separate waveguide and coupling prism. RM devices were fabricated by coating a low refractive spacer layer and metal oxide layer onto thin glass sheets. These were placed onto a prism mounted in the optical unit; coupling of the prism and RM was achieved through an index matching fluid. A sample chamber, with an attached gasket, was clamped in place over the RM device to form a cuvette. In the final production version, the RM structure, prism and cuvette were integrated into the design shown in Figure 2.

259

Kinning and Edwards The first commercial instrument, launched in 1993, was a single channel instrument with manual sample injection and a peristaltic pump driven aspiration system. A dual channel system with auto-sampler and vacuum aspiration system, the IAsys AUTO+, was launched in 1996. In 1997, a two-channel manual system, again with a vacuum aspiration system, was launched. Software advances enabling real-time decision-making was introduced in 1998 for the AUTO+, to become the AUTO+ Advantage. An improved sensitivity version of all the instruments was introduced in 2000, together with the addition of vacuum aspiration for the single channel IAsys.

3. State of the Art The selectivity afforded to the instrument is via the attachment of one of the interactants to the sensor surface. This method of selectivity allows the researcher to monitor a vast array of differing interactions. The only proviso is that one of the interactants must be attached to the sensor surface in a manner that allows binding activity to be retained.

3.1. Protein-protein interactions The initial use of optical biosensors was for the investigation of protein-protein interactions and this is still by far the largest application area. The RM is ideally suited to monitoring these interactions as the interactants are often of large molecular weight (greater than 5 kDa) thus producing large instrument responses. The varied uses of the RM within the field of protein-protein interactions are highlighted in the applications below.

3.1.1. Signal Transduction.

Shenolikar et al. (2001) investigated the dimerisation of Na+/H+ exchanger regulatory factor, NHERF. NHERF is a 55 kDa protein that mediates signal transduction at the plasma membrane after binding to its receptors and ion transporters. Purified recombinant His-NHERF was immobilised to a CMD surface and the interaction of soluble His-NHERF investigated. Data analysis was complicated by the ability of the NHERF to form dimers and as such bind in a non-l" 1 stoichiometry. Despite these issues, an estimated dissociation equilibrium constant, Ko, of between 1 and 10/~M was determined. The authors proposed that NHERF dimers play a role in hormonal signalling. Jostock et al. (2001) used the RM to investigate the interaction between interleukin-6 (IL-6) and the signal transducer gpl30. The cytokine IL-6 must bind to its receptor (IL-6R) and cause the dimerisation of gpl30 in order to perform its signal transduction functions. Naturally occurring soluble gp130 was immobilised on CMD and the interaction of the IL-6/IL-6R complex in the form of the fusion protein Hyper-IL-6 monitored. From the results, it was concluded 260

Resonant Mirror Biosensors that sgpl30 is the natural inhibitor of the agonistic IL-6/slL-6R complex. It was also found that sgp 130 blocked IL-6 responses from soluble slL-6R but not from membrane-bound IL-6R. These findings suggest that sgpl30 could be used in blocking such IL-6 responses found in disease states such as Crohn's disease.

3.1.2. Diagnostics. The specific interaction of anti-neutrophil cytoplasmic antibodies (ANCA) with proteinase 3 (PR3) can be used as a diagnostic aide in Wegeners granulomatosis (WG). Griffith et al. (2001) immobilised PR3 to a CMD surface and bound ANCA from patients. ANCA from different patients recognised the same or closely related epitopes on PR3, indicating that this system could be used as a rapid method for identification of ANCA in patient samples for WG given its general applicability. Conrath et al. (2001) illustrated the specificity of bispecific antibody constructs using camel single domain antibody constructs as building blocks. Bispecific antibody constructs were assembled using single antibody domains with specificity to lysozyme and either Nme-A or c~-amylase. The RM allowed the stability and affinity of these constructs to be determined. The study demonstrated that the constructs remained active over a period of 28 hours, even after storage at 37~ Furthermore, there was no discernible difference between the affinity of the constructs compared to the single domain antibody fragments. Parsons et al. (2001) investigated the binding of laminin to recombinant lutheran glycoproteins containing the Fc domain of human IgG (Lu-Fc). Lutheran glycoproteins are thought to contribute to vasoocclusion in sickle cell disease. The Fc region of Lu-Fc was bound to the RM surface via CMD-immobilized anti-Fc antibody. Subsequent laminin binding showed that the Lu-Fc/laminin binding is specific with a high affinity. The first three N-terminal immunoglobulin superfamily domains were shown to mediate the laminin binding.

3.1.3. Protein Folding. Tsukahara et al. (2000) identified a novel variant of the human heat shock cognate protein 70, HSC70, named heat-shock cognate protein 54 (HSC54) and characterised the binding of HSC54 to a cochaperone p60. The interaction of immobilised p60 with differing concentrations of HSC54 and HSC70 was monitored at 37~ It was found that both the association and dissociation rate constants were similar. These results suggested that HSC54 may function as an endogenous inhibitory regulator of HSC70 by competing with the cochaperones.

3.1.4.

Immune Response. Tamm-Horsfall protein (THP) is one of the most abundant proteins present in normal human urine. It interacts with various leucocytes, including neutrophils, monocytes and lymphocytes, as well as with complement lq (Clq). Rhodes (2000) investigated the interaction of THP with C lq on both planar carboxylate- and amino-derivatized surfaces. Regeneration 261

Kinning and Edwards of the system was problematic regardless of which of the partners was immobilized or which surface was used. To overcome these issues, the author used the equilibrium titration methodology where additional ligate is added to the cuvette without regeneration (Hall and Winzor, 1997). Another route involved the electrostatic nature of the carboxylate surface. Here positively charged Clq was electrostatically attached to the carboxylate surface, washed with low ionic strength buffer and the binding of THP monitored. Complete regeneration was afforded by increasing the ionic strength of the buffer, thereby dissociating C lq from the surface. Differing concentrations of THP were therefore investigated by re-attaching C lq for each binding experiment. All these approaches yielded a KD of between 10 and 15 nM. The electrostatic experiments suggested that the association rate constant was 1.25 x 105 M-~s~. The equilibrium titration on the carboxylate surface showed a marked difference in the KD for the interaction in low ionic buffer (12.4 nM) compared to physiological (360 nM). Dissociation data from the equilibrium titration on the amino surface suggest that this difference may be due to a significant change in the dissociation rate constant. A dissociation rate constant of 5 x 10.3 s~ was observed in low ionic strength buffer, while at physiological concentration, a rate of 1 x 10-1 s1 was observed. These studies suggest that at least a portion of this interaction involves electrostatic forces between the positively charged Clq and the negatively charged THP molecule with nanomolar affinity. The interaction of a model peptide (PS 1CT3) with monomeric IgM antibodies obtained by reduction of IgM was investigated. PS 1CT3 was chosen as a model peptide as it represents a T-dependent immunogen. Using the peptide as immobilized ligand, Nakra et al. (2000) investigated the affinities of different monomeric IgM preparations. Furthermore, thermodynamic properties of the interaction were investigated by monitoring the interactions at various concentrations and constructing an Arrhenius plot. From these data, it was suggested that there was a relationship between the primary recognition of the peptide and the chemical nature of the epitope recognised. This relationship way derive from enthalpic contributions. Salek-Ardakani et al. (2000) used RM to show that heparin binds interleukin-10 (IL-10), an 18 kDa glycoprotein cytokine. B iotinylated heparin-albumin was attached to a planar biotin surface via streptavidin and the binding of IL-10 monitored. An affinity of 54 nM was calculated. Given the fact that the physiological relevance of IL-10 binding to heparin can be expected to depend on the affinity of the interaction, it was postulated that such binding might take place in a physiological environment and that this interaction may play an important factor in the biology of IL-10.

3.1.5. Vaccine development. Charalambous and Feavers (2000) investigated the potential of peptide mimics of the outer membrane lipooligosaccharide (LOS) of Neisseria meningitidis, to interact with a LOS-specific antibody. The peptide 262

Resonant Mirror Biosensors mimics were identified and enriched from a coliphage display library (M13) with a LOS-specific monoclonal antibody using the RM instrument. Several rounds of panning were performed. Bound phage was eluted from the immobilized ligand by glycine, neutralised, amplified by transfection and finally sequenced. The RM was then used to determine the affinity of these peptides for the LOSspecific antibody. Given the high affinity, the authors suggested that a vaccine based upon peptide mimics might have better immunological memory and fewer safety concerns than LOS. In a previous paper, Charalambous et al. (1999) investigated the specificity and affinity of two antibodies raised against the same LOS idiotype which are routinely used in immunological typing. Biotinylated LOS was attached to a planar biotin sensor surface via preloaded streptavidin and the kinetics of the two antibodies were determined. There was a marked difference in affinities between the two antibody preparations (7.5 nM and 233 nM). This difference was the result of differing association rate constants and not differing dissociation rate constants. These data, in combination with ELISA and gene sequencing, supported the conclusion that this difference is likely due to the two antibodies recognizing different epitopes as a result of the different immunization protocols used in their production. Patients with anti-phospholipid antibodies (aPL) are classified as having antiphospholipid syndrome (APS). Such aPLs have been associated with a variety of disorders including myocardial infarctions. Initially, it was thought that aPL were directed to phospholipids, but it has now been found that other phospholipid binding proteins are relevant antigens. Kobayashi et al. (2001) showed specific interactions between [32-glycoprotein I (132-GPI), a major antigen recognized by aPL, oxidised low density lipoprotein (oxLDL), and anti-132-GPI auto-Ab. The authors immobilised 132- GPI to a RM surface and allowed LDL or oxLDL to bind. LDL failed to interact with the attached [32-GPI but ox-LDL bound with high specificity. Furthermore, attachment of biotinylated anti-132-GPI auto-Ab revealed that binding of oxLDL was observed only in the presence of 132-GPI. This observed specificity allowed the authors to purify a ligand specific for 132GPI from extracted lipids of oxLDL. Camelids can produce antibodies devoid of light chains and CH1 domains. As such camelid heavy-chain variable domain (VHH) fragments are small in size (a single immunoglobulin domain) and their levels of expression and solubility are far higher than those of Classical Fab or Fv. However, due to their small size and monomeric nature, their ability to bind haptens has been questioned. Spinelli et al. (2000) produced VHH fragments to a dye hapten. Attachment of dye to a RM cuvette allowed the specific nature and affinity of the anti-RR6 fragment to be demonstrated. Values of 1.5 x 105 M -~ s-~ and 3.3 x 10.3 s~ were determined for the association and dissociation rate constants respectively. Combining these rate values gives a KD of 22 nM. In a similar manner, van der Linden et al. 263

Kinning and Edwards (1999) compared the affinity of llama VHH fragments with monoclonal IgG antibodies. Similar affinities for the VHH fragments and whole antibodies were determined. Temperature stability and binding experiments were performed in the same study. 3.1.6. Nuclear transport. The RM has also been used to determine that calreticulum, CRT, a calcium binding protein involved in protein folding, is a receptor for nuclear export. Holaska et al. (2001) used a biotin surface modified by streptavidin to attach a biotinylated nuclear export signal (NES) peptide. It was found that CRT bound to NES only in the presence of RanGTP with a KD of 8.5 nM. 3.1.7. Structure. The motif of mainly four amino acids - Pro, Glu, Val, Lys (PEVK) within the elastic protein titin is thought to be the region conferring elasticity. Gutierrez-Cruz et al. (2001) immobilised a fragment containing 16 PEVK modules to a CMD surface. The interaction of actin and cloned nebulin fragments with the immobilized fragment was investigated. It was found that actin and some nebulin fragments bound with high affinity and that Ca2+/ calmodulin could prevent this binding. The authors suggested that the PEVK region may be involved in interfilament association with thin filaments in a 2+ . . . . . Ca /calmoduhn-dependent manner and thus may modulate tmn elasticity. -

Protein 4.1R (4.1R) in red blood cells plays a critical role in maintaining cell morphology and membrane mechanical properties. The elasticity and mechanical stability of the erythrocyte membrane are regulated by the interaction of 4.1R with membrane proteins such as glycophorin C (GPC) and p55. The binary interactions of GPC, 4.1R, and p55 were investigated with the RM using planar amino sensor surfaces (Nunomura et al. 2000a). Formation of the ternary complex, CPC-4.1R-p55, was also investigated using the binary complexes GPC4.1R, GPC-p55, and p55--4.1R. It was determined that sequences within a 30kDa domain of 4.1R constitute the binding interfaces for GPC and p55. Evidence was also found to imply that 4.1R modulates the interaction between p55 and GPC by increasing the affinity of p55 binding GPC. Nunomura et al. (2000b) have also investigated the interaction of 4.1R with calmodulin (CAM) using the RM. The study identified two distinct CaM binding sites in 4.1R encoded by two different exons. Exon 11 binds CaM with high affinity in the presence or absence of Ca 2§ Exon 9 binds CaM with high affinity only in the presence of Ca 2§ The presence of two distinct Ca 2§ dependent and independent CaM binding sites makes 4.1R unique as a CaM binding protein. 3.1.8. Miscellaneous. Zimmer et al. (2000) investigated the interaction of a new cz-tocopherol binding protein (cx-tocopherol-associated protein, TAP) with its ligand, cx-tocopherol. The authors loaded a biotinylated cuvette with avidin and attached a biotinylated cz-tocopherol derivative (in 10% DMSO). The

264

Resonant Mirror Biosensors interaction of recombinant TAP with this attached cz-tocopherol derivative was investigated using bovine serum albumin as a control surface. A Ko of 0.46 #M was determined, illustrating that the recombinant TAP recognises cz-tocopherol in a specific manner. Heparin affin regulatory peptide (HARP) is a polypeptide that binds heparin with a high affinity. Given this high affinity, it has been proposed that this polypeptide may also bind to heparin sulfate proteoglycans derived from the cell surface and extracellular matrix. In order to determine the kinetics of interaction between HARP and heparin and dermatan sulfate (DS), Vacherot et al. (1999) attached biotinylated heparin and biotinylated DS to a planar biotin surface loaded with streptavidin. The binding of differing concentrations of HARP to either ligand allowed the kinetic of the interactions to be investigated. The binding of HARP to heparin and DS was always homogenous; there was no evidence for the presence of more than one binding site for HARP in either of the glucosaminoglycans. The association rate constant for the HARP-heparin interaction (1.6 xl06 Mas a) was found to be twice as fast as that for the HARPDS interaction (0.68 x 10 6 M-isl). The dissociation was slightly slower from the heparin. As such, the affinity for HARP with heparin was higher than for the DS interaction (13 nM compared to 51 nM). Competition assays on the RM illustrated that the binding of HARP to either of the glucosaminoglycans recognises a distinct motif compared to the heparin binding growth factor, FGF2.

3.2. Protein-carbohydrate Ricin toxin is composed of two glycoprotein chains A (32 kDa) and B (34 kDa) linked by a disulfide bridge. The first step in ricin activity against eukaryotic cells is the binding of the B chain to cell-surface galactosides and subsequent transport of the A chain into the cell via endocytosis. Despeyroux et al. (2000) used the RM to investigate the heterogeniety of ricin toxin produced by differing varieties of Ricin communis in differing geographies. In these studies, galactose derivatives were used as ligands on sensor surfaces. The RM, combined with capillary isoelectric focusing electrophoresis (CIFE) and mass spectrometry, highlighted differences in both cultivars and geographical sources. It was proposed that the difference in most of the cultivars investigated was due to differences in glycosylation. 3.3. DNA Several resonant mirror papers have focused on interactions involving DNA. The high negative charge on DNA could potentially cause problems when trying to immobilise to CMD. Furthermore, many DNA binding proteins have a low pI thus increasing the likelihood of non-specific ionic interactions. While under some circumstances the CMD surface can be used (Torigoe et al. 2001), the surface of choice is often planar biotin. 265

Kinning and Edwards At physiological pH, the instability of pyridine motif triplex DNA limits its use as a control of gene expression. In order to improve their general stability, a new class of nucleic acid modification called bridged nucleic acid (BNA) has been developed. The RM was used to determine the kinetic properties of these modifications in terms of triplex formation. Torigoe et al. (2001) immobilised streptavidin to CMD and bound a biotinylated 23-mer oligonucleotide, Pur23T. Duplex formation was achieved by binding a complimentary 23-mer oligonucleotide, Pur23A. Upon duplex formation, a variety of triplex-forming oligonucleotides (TFO) were bound and their interaction kinetics determined. It was found that the BNA modification had a profound affect on the dissociation kinetics while the association kinetics seemed unaffected. As such the observed increase in KA was due to the slowing of the dissociation rate by up to 100-fold. These findings suggest that BNA modification improves the stability of triplex formation and as such may allow progress in therapeutic applications. The interaction of the repressor protein ModE with the operator/promotor region of ModABCD was investigated using RM (Grunden et al., 1999). In Escherichia coli, the production of molybdoenzymes requires the uptake of molybdate by the specific transporter coded by ModABCD. A biotinylated oligonucleotide complementary to the ModA operator was captured on a planar biotin surface loaded with streptavidin. The interaction of ModE repressor in the presence and absence of molybdate was investigated. Furthermore, the interaction of scrambled sequences with ModE was followed. The authors found the specific interaction had a KD of 0.3 nM in the presence of molybdate and 8 nM in the absence. The scrambled sequences gave no discernible response. The authors also investigated a mutant ModE protein. RM results demonstrated no significant effect of molybdate on association or dissociation of the mutant protein from the operator/promoter region. The binding of human Rad51 protein (hRAd51) and bacterial RecA protein, enzymes involved in DNA strand exchange, to single stranded DNA (ssDNA) was investigated using the RM (De Zutter et al., 2001). This binding was monitored in the presence and absence of cofactors ATP and ATP3(S. Regeneration was achieved using 3M MgClz. The study showed that hRAd51 had over a forty-fold higher association rate with ssDNA than RecA while the dissociation rates were equivalent. Marked differences in the influence of cofactors on the binding constants were also seen. hRAd51 was largely unaffected by ATP and ATP~S while, in the case of RecA, ATP increased the association by approximately 50-fold and reduced the dissociation rate dramatically. The isotherm plot yielded a sigmoidal curve for hRAd51 in the presence and absence of ATP while RecA only showed this sigmoidal relationship in the presence of ATP. The sigmoidal plot is indicative of cooperativity. The results supported the conclusion that, while both proteins are

266

Resonant Mirror Biosensors involved in DNA strand exchange, their mechanisms of interaction with DNA differ. Sassoon et al. (2001_ investigated the interaction of cmbl fragments with DNA to further characterise its DNA-binding motif within its High Mobility Group (HMG) box. Cmbl has previously been identified as a potential mismatclabinding protein using a band-shift assay (Fleck et al. 1998). In the recent work, the binding constants and stoichiometry for both homoduplex and heteroduplex DNA were determined using the RM. NeutrAvidin was attached to a planar biotin sensor surface and biotinylated DNA attached. A difference in binding properties of the different cmbl fragments was observed with A 41 giving the highest response. No binding of cmbl A 41 was observed to NeutrAvidin or to single stranded DNA, while binding to double stranded homoduplex and heteroduplex was observed. From the results, the authors confirmed that cmbl does interact with both homoduplexes and heteroduplexes but with a marked preference for mismatched DNA.

3.4. Cells The simple open well format of the cuvette used in the RM allows the use of cellular material while preventing blockages associated with small sample delivery tubing. As a result, RM has been used for a number of investigations using a variety of cells. The size of cells often precludes the use of CMD, as the majority of the cell .is kept out of the evanescent field. General overviews of some of the recent applications involving cells are discussed below. The effect of perflubron treatment of endothelial cells on the binding of activated neutrophils was investigated by Woods et al. (2000). Perflubron treatment has been shown to reduce the likelihood of damage resulting from interaction of activated neutrophils with endothelial cells of the lung. In addition to perflubron treatment, a subpopulation of endothelial cells was subsequently treated with cytokines to increase the surface expression of E-selectin and ICAM-1. This study showed that activated neutrophils bound more slowly to perflubron-treated endothelial cells, including those expressing E-selectin or ICAM-1. High molecular weight kininogen (HK) plays an essential role in the activation of coagulation when blood encounters a foreign substance. HK is cleaved by kallikrein into a two-chain disulphide linked molecule, HKa. It has been shown that HKa/HK binds specifically to neutrophils in the presence of Zn 2+ although the cell receptor for this is unclear. Sheng et al. (2000) showed that the purified integrin Mac-1 (CDllb/CD18) bound specifically to immobilized HKA with an association rate constant of 5.6 x 106 M-isI and a dissociation rate constant of 1.8 x 10.2 sq. They thus concluded that Mac-1 is a specific receptor for HKa interaction. To further illustrate this, the authors compared the binding of Mac-1 transfected HEK (human embryonic kidney) 293 cells with untransfected cells. 267

Kirming and Edwards The binding of Mac-1 transfected HEK cells to HKa was significantly greater than that of untransfected cells to HKa at equivalent cell numbers. The observed binding was also Zn 2+ dependent as the interaction in EDTA resulted in lower response values.

Helicobacter pylori colonizes the lower gastric mucin layer of the stomach and adheres to human gastric epithelial cell receptors. Gastric mucin glycoprotein fraction was immobilised to a sensor surface and the interaction of H. pylori investigated by the RM (Hirmo et al. 1999). Two strains of H. pylori were investigated and differences in response levels observed. In order to block differing sites on the mucin, binding experiments were performed in the presence of competitive glycoconjugates. Sialyl (~-2,3)lactose caused a 30% drop in response while preincubation with heparin completely eliminated binding. The authors proposed that, given its ability to observe binding of a variety of H. pylori strains and the lack of labelling requirements, the RM was therefore better at simulating conditions within the human stomach compared to more conventional assays. This group also illustrated the importance of sialic acid containing oligosaccharides as adherence ligands for H. pylori (Hirmo et al., 1998). Binding characteristics of different strains of H. pylori to immobilized sialyl(ot-2,3)lactose-PVA conjugates were investigated together with the effect of competing glycoconjugates. It was possible to distinguish between strains according to their ability to bind to the immobilised ligand; both hemagglutinating and poorly hemagglutinating strains were tested. The specificity of the interaction with the sialyl group was shown in competition experiments with free sialyl(t~-2,3)lactose. 3.5, Membranes The RM has been used to monitor lipid interactions. The deposition of the lipid layer can be performed in a variety of ways, as illustrated in the literature discussed below. Various methods of liposome deposition have been investigated by Fisher and Tj~irnhage (2000). The first method was direct capture of biotinylated liposomes onto a streptavidin-coated biotin-derivatized sensor surface. The second was by direct adsorption of liposomes onto an underivatised surface. The final method involved deposition of liposomes via pre-captured biotinylated lipid on a streptavidin-coated planar biotin sensor surface as per method 1. These lipids were then washed to ensure only biotinylated lipid was attached, and the exposed hydrophobic tails of the lipids acted as anchors for subsequent adsorption and fusion of liposomes. Binding of cholera toxin B showed that all of the methods produced active biomembranes. The fusion rate of the liposomes was dependant on the method of attachment with the biotinylated liposomes on the streptavidincoated sensor surface having the highest rate and those on to the biotinylated lipid the slowest. Subsequent AFM analysis of these surfaces showed larger 268

Resonant Mirror Biosensors structures indicative of non-fused liposomes. It was found that the biotinylated liposomes deposited on the streptavidin surface had the smoothest surface while those attached via the biotinylated lipid layer exhibited the roughest. In terms of overall stability, the surfaces produced via direct adsorption had the highest stability (over 40 days). Puu and Gustafson (1997) produced lipid bilayers on RM surfaces by adsorption of liposomes onto underivatized surfaces and obtained response vaues twice those obtained for a single lipid layer. Furthermore, the type and level of disaturated PC present influenced the stability of the layer. It was found thatwhen the major component was dipalmitoyl-PC, washing and drying could not remove the bilayer. More recently, Puu et al. (2000) adsorbed liposomes with a biotinylated membrane-spanning protein (biotinylated acetylcholine receptor) to underivitised sensor surfaces and monitored the binding of streptavidin. Puu (2001) also fused liposomes containing various glycolipids onto underivatised sensor surfaces and investigated the binding of protein toxins to the lipid-coated surfaces at pH 7.4 and pH 5.2. Using surfaces coated with different glycolipids at the two pH values, it was possible to identify the binding toxin. This method of toxin identification is simple and rapid, and the surface can be regenerated after binding. In a publication by Nagel et al. (1998b), liposomes containing phosphatidylinositol-(3,4,5)-trisphosphate were deposited on a planar underivatised surface. The interaction of this phosphorylated ligand with cytohesin-1 or mutant cytohesin-1 was investigated. The authors found that the mutant lacking the pleckstrin homology (PH) domain failed to bind to the phosphorylated ligand. This PH domain is thought to aid in membrane recruitment of proteins through their interactions with phosphorylated ligands. Nagel et al. (1998a) subsequently determined the kinetics of the binding of PH domains of cytohesin-1 when expressed as a (GST) fusion protein. It was discovered that phosphatidylinositol-(3,4,5)-trisphosphate binds to the PH domain and c domain together with high affinity (100 nM), whereas the isolated PH domain has a substantially lower affinity (2-3 mM). Surfactant protein A (SP-A) is one of the surfactant proteins that modulate surface tension in the lung to enable breathing to occur. The RM has been used to investigate the role of this calcium binding protein (Meyboom et al., 1997). SP-A was attached to an amino surface and the interaction of liposomes with this surface was monitored. Liposome binding was calcium dependent with an optimum around 20 /~M and highly co-operative. Bare amino surfaces or surfaces coated with heat-treated SP-A were unable to bind liposomes. The deposition of lipid monolayers does not require the preparation of liposomes. Both pure and binary mixtures can be used to form phospholipid monolayers on hydrophobic surfaces (Athanassopoulou et al.,1999). This method of monolayer 269

Kinning and Edwards formation uses small amounts of reagents and is rapid. The authors used the partially water-soluble brain ganglioside GM1 to illustrate the applicability of the method. Specific binding of cholera toxin B sub-units to the GM1 layer was observed with no uptake of bovine serum albumin.

3.6. Industrial applications This section describes the industrial applications of the RM within the biotechnology arena. The areas discussed include fermentation monitoring, ligand fishing and contamination monitoring.

3.6.1. Fermentation Monitoring.

Optical biosensors provide rapid response times and high specificity and sensitivity. As a consequence, they are highly suited to process monitoring. Tsoka et al. 1998 used the RM to rapidly analyse the interaction of virus-like particles (VLP) with antibodies. VLP are multimeric protein assemblies produced by the yeast transposon Ty-1 in Saccharomyces cerevisiae and may be of potential interest in diagnostics, vaccines and research reagents. A monoclonal antibody specific for a surface epitope on the VLP was attached to a planar amino surface. The interaction of VLP particles was then followed. Concentration-dependent binding profiles were seen. The RM was also used to monitor the production and subsequent purification of antibody fragments expressed during batch fermentation of recombinant Escherichia coli. Gill et al. (1998) attached hen egg lysozyme to a sensor surface and used it to monitor the interaction of fermentation samples containing D1.3 Fv. A control surface coated with turkey egg lysozyme was also employed. The RM could detect binding of D 1.3 Fv in a concentration-dependent manner within 10 seconds of sample addition. The speed of this analysis allowed the authors to propose its use in fermentation monitoring in order to optimise harvesting. Bracewell et al. (1998) investigated the use of the RM in affmity chromatography applications to determine the moment of product breakthrough. Modeling of these data allowed the moment of breakthrough to be predicted prior its occurrence and thus prevented wastage of product resulting from column overloading.

3.6.2. Ligand fishing. Catimel et al. (2000) utilised the large surface area of select RM sensor surfaces (14.5mmz) to perform micropreparative ligand fishing. Fractions containing A33 antigen from anion exchange HPLC or from A33transfected cell extracts were identified using the biosensor. Repeat binding cycles were performed during ligand fishing using an immobilized F(ab)'2 fragment or whole antibodies as recognition elements. Fractions containing A33 were eluted from the RM using either 10 mM NaOH (IgG capture) or 10 mM HC1 (F(ab)'2 capture). Sufficient material was eluted for silver stained gel SDSPAGE analysis which showed the material to be homogeneous. 270

Resonant Mirror Biosensors

3.6.3. Contamination monitoring. Many molecules of interest are often only partially soluble in aqueous media. As such, it is of interest to determine the effect of organic solvents such as methanol on interactions. Skladal (1999) investigated the influence of methanol on the binding of monoclonal antibody to atrazine immobilised on a planar amino surface. It was found that the highest affinity was determined at 10% methanol with the fastest association rate and the slowest dissociation rate. Finally, the author investigated the possibility of performing a competition assay for atrazine with the antibody in the presence of methanol. Once the system was optimised, the RM was used for the investigation of atrazine in soil samples. Results from RM analysis correlated well with those from GC and ELISA (Skladal et al., 1999). An interesting application of the RM is in the field of food quality monitoring. Rasooly and Rasooly (1999) developed an assay for staphylococcal enterotoxin A (SEA). The authors immobilised an anti-SEA antibody to the sensor surface and allowed standard or spiked food suspensions to interact. This direct analysis gave an assay time of around 4 minutes with no significant non-specific responses observed. In order to further increase the specific response values, a sandwich format was also employed using a second antibody. The sandwich format was used to monitor levels of SEA in foods such as hot dog extracts. Sensitivities of 10 to 100 ng SEA per g of total protein were achieved. Autoclaved samples failed to produce a positive response.

4. Advantages and Limitations The RM range of instruments have several advantages over more classical methods of interaction analysis. Perhaps the most important is the generic nature of the technique. RM instruments have been used to measure many different types of interactions, including DNA:DNA (Torigoe et al., 1997, 2001), protein:protein (Conrath et al., 2001; Spinelli et al., 2000), protein:DNA (De Zutter et al., 2001; Yang et al., 1998), protein:carbohydrate (Despeyroux et al., 2000; Feldman et al., 1998; Lee et al., 1998), lipid:protein (Nagel et al., 1998a; Yagisawa et al., 1998) and protein:cell (Hirmo et al., 1998; Morgan et al., 1998; Thakkar et al., 1998). The wide applicability of the instrumentation is due to the fact that an inherent property of the interaction is being measured, the refractive index, and no isotopic or fluorescent labelling is required. Immobilisation of one of the interactants may facilitate re-use of one of the interacting partners, saving on material. The ease of recovery from the cuvette of either unbound or bound material further saves on reagent costs. Recovery of bound material further expands the applicability of the system to that of ligand fishing. This technique can be used to screen complex mixtures, for example cell extracts, for molecules capable of binding to a target of interest immobilised on

271

Kinning and Edwards the surface. The large surface area, approximately 14.5 mm 2, allows recovery of sufficient material after a very few number of binding/recovery cycles for further analysis by a different method, such as mass spectrometry. Automated software available enables samples not showing any specific response over that of a control well to be rapidly passed over and only positive samples to be focused upon. Automatic recovery, with repeat cycles can be performed on positive samples. The real-time nature of the response enables affinity determinations to be resolved into on and off rates. Knowing the association and dissociation rates enables far more sophisticated targeting of protein engineering to be performed. When set up in flow configuration, very rapid experimental cycles of baseline/binding/regeneration can be performed. For example, Bracewell et al. (1998) achieved 30-second cycle times. This response time enabled saturation of an affinity column to be rapidly detected, enabling column loading to be stopped, thus reducing losses of the material being loaded. Subsequent detection of the 'active' fractions of eluting material was also achieved. Using an initial rate algorithm, Gill et al. (1998) were able to rapidly quantify the amount of a Fv fragment in a fermentation process, enabling the optimal time of harvest to be predicted. At present, the major limitation is the lack of sensitivity for the direct measurement of molecules of a few hundred Daltons. As the detection is influenced by the mass of the molecule, small molecules give rise to smaller responses. Direct measurement is possible; however, micromolar quantities are usually required. As most small molecules of pharmaceutical interest require a solvent in the buffer to maintain solubility, for example DMSO, this will. also gives rise to a measurable response. Typically these bulk responses are of considerably higher order of magnitude than the response due to the binding alone, making results unreliable. These bulk responses are mainly due to two factors: the relative high refractive index of the solvent relative to the buffer and the effect of the solvent on the swelling of the protein:dextran matrix. Any changes in the thickness of the matrix will result in more or less protein within the evanescent field, resulting in a bulk response. Great care must be taken in exactly matching any control wells to that of the test wells in order to correctly model the bulk responses. Performing competition assays, where the binding of a large protein-hapten conjugate is detected, can enhance sensitivity, but this requires making a conjugate for each molecule of interest. Although assay times can be shortened to about 30 seconds (see above), sample throughput may still prove problematic. At present, a true medium to high throughput optical biosensor system is not commercially available, although there are systems in development. Until true multi-channel, array-based systems are available, the numbers of samples that can be processed are low.

272

Resonant Mirror Biosensors 5. Future Potential

If the present limitations are taken into consideration, there are probably two main areas for development: sensitivity and throughput. Sensitivity improvements can be achieved in two ways: improved instrumentation and waveguide construction. Firstly, the signal-to-noise ratio can be improved, resulting in resolution of smaller responses. Software algorithms can be used to smooth noise from the baseline responses. The whole of the optical system could be enclosed in a vacuum, reducing any influence of convection currents in the light path. Improvements can be made to both the quality of the resonant layer and the material from which it is fabricated. Maximising the refractive index differential between the sensed layer and the waveguide is one way of improving the sensitivity. However, this requires careful choice of the deposition method used and the chemical stability of the resultant layer. Careful experimentation with the deposition method should result in uniform layers of high refractive index dielectric material; the more uniform the coating, the less loss due to light scattering. Alterations to the refractive index of the waveguide and/or the detection optics will enable both TE and TM modes to be measured, resulting in both better sensitivity and more information. Presently only the TE mode is measured, allowing only the refractive index to be measured. Measuring both TE and TM modes will enable both refractive index and thickness to be resolved. This should potentially enable conformation changes to be detected as well as binding kinetics. Increases in throughput will come from changing the diode array detection systems to CCD arrays. This will enable true multi-channel array sensors to be developed. Based on standard 96- and 384-well spacing, liquid handling will be simplified by already available robotic handling systems. This will lead to true high throughput sensing capability. Using sensors in conjunction with other technologies will also play an important part in the future of RM sensors. The coupling of the ability of the sensor to specifically detect and quantify binding partners of interest from crude extracts has already been demonstrated. The increase in interest in proteomics will only lead to more complete protein detection and identification systems, such as a system including RM to detect binders of interest, a liquid handling system to recover bound samples, and a mass-spectrometer for identification.

273

Kinning and Edwards Miniaturisation should lead to sensors being used as specific detection systems linked to LC systems. This should lead to better identification of eluted active fractions. In the field of downstream processing of recombinant proteins produced by fermentation, this could lead to significant improvements in yields. Linked to this on-line sensing is the use of remote sensors. With increases in sensitivity, it should be possible to set up systems, for example, to monitor the toxin levels in water supplies or effluents from manufacturing processes. For remote sensing purposes, RM could be used with a recognition element stable to the sample to which it is repeatedly exposed and to any regeneration conditions used to re-set the system. For use with fermentation, the RM and the sensing entity should be stable to sterilisation procedures. The advances in protein engineering and the development of artificial recognition elements should fulfil the requirements. Coupling of an array of artificial recognition elements to pattern recognition software should enable optical sensors to become ideal for all areas where rapid and specific detection of contaminants is required, for example, for protecting installations and manpower from chemical and biological warfare attacks.

6. References Athanassopoulou, N., R. J. Davies, P. R. Edwards, D. Yeung and C. H. Maule, 1999, B iochem. Soc. Trans. 27,340. Bracewell, D. G., A. Gill, M. Hoare, P. A. Lowe and C. H.Maule, 1998, Biosens. Bioelectron, 13, 847. Catimel, B., J. Weinstock, M. Nerrie, T. Domagala and E.C. Nice, 2000, J. Chromatogr. A. 869, 261. Charalambous, B. M., J. Evans, I. M. Feavers and M. C. Maiden, 1999, Clin. Diagn. Lab. Immunol. 6, 838. Charalambous, B. M. and I. A. Feavers, 2000, FEMS Microbiol. Lett. 191, 45. Conrath, K., M. Lauwereys, L. Wyns and S. Muyldermans, 2001, J. Biol. Chem. 276. Cush, R., J. M. Cronin, W. J. Stewart, C. H. Maule, J. Molloy and N.J. Goddard, 1993, Biosens. Bioelectron. 8, 347. Davies, R. J. and D. Pollard-Knight, 1993, Am. Biotechnol. Lab. 11, 52. De Zutter, J., A. Forget, K. Logan and K. Knight, 2001, Structure 9, 47. Despeyroux, D., N. Walker, M. Pearce, M. Fisher, M. McDonnell, S. C. Bailey, G. D. Griffiths and P. Watts, 2000, Anal. Biochem. 279, 23. Edwards, P. R., A. Gill, D. V. Pollard-Knight, M. Hoare, P. E. Buckle, P.A. Lowe and R. J. Leatherbarrow, 1995, Anal. B iochem. 231,210. Feldman, R. G., M. A. Breukels, S. David and G. T. Rijkers, 1998, Clin. Immunol. Immunopathol. 86, 161. Fisher, M. and T. Tj~irnhage, 2000, Biosens. Bioelectron. 15,463. Fleck, O., C. Kunz, C. Rudolph and J. Kohli, 1998, J. Biol. Chem. 273, 30398. 274

Resonant Mirror Biosensors Gill, A., D. G. Bracewell, C. H. Maule, P. A. Lowe and M. Hoare, 1998, J. Biotechnol. 65, 69. Glaser, R. W., 1993, Anal. Biochem. 213, 152. Griffith, M. E., A. Coulthart, S. Pemberton, A. J. George and C. D. Pusey, 2001, Clin. Exp. Immunol. 123, 170. Grunden, A. M., W. T, Self, M. Villain, J. E. B lalock and K. T. Shanmugam, 1999, J. Biol. Chem. 274, 24308. Gutierrez-Cruz, G., A. H. Van Heerden and K. Wang, 2001, J. Biol. Chem. 276, 7442. Hall, D. R. and D. J. Winzor, 1997, Anal. Biochem. 244, 152. Hirmo, S., E. Artursson, G. Puu, T. Wadstrom and B. Nilsson, 1998, Anal. Biochem. 257, 63. Hirmo, S., E. Artursson, G. Puu, T. Wadstrom and B. Nilsson, 1999, J. Microbiol. Methods, 37, 177. Holaska, J., B. Black, D. Love, J. Hanover, J. Leszyk and B. Paschal, 2001, J. Cell Biol. 152, 127. Jostock, T., J. Mullberg, S. Ozbek, R. Atreya, G. Blinn, N. Voltz, M. Fischer, M. F. Neurath and S. Rose-John, 2001, Eur. J. Biochem. 268, 160. Kobayashi, K., E. Matsuura, Q. Liu, J. Furukawa, K. Kaihara, J. Inagaki, T. Atsumi, N. Sakairi, T. Yasuda, D. R. Voelker and T. Koike, 2001, J. Lipid Res. 42, 697. Lee, J., T. Cairns, W. McKane, M. Rashid, A. J. George and D. Taube, 1998, Transplantation 66, 1117. Liedberg, B., C. Nylander and I. Lundstrom, 1983, Sens. Actuators 4, 299. Meyboom, A., D. Maretzki, P. A. Stevens and K. P. Hofmann, 1997, J. Biol. Chem. 272, 14600. Morgan, C. L., D. J. Newman, S. B. Cohen, P. Lowe and C. P. Price, 1998, Biosens. Bioelectron. 13, 1099. Nagel, W., P. Schilcher, L. Zeitlmann and W. Kolanus, 1998a, Mol. Biol. Cell 9, 1981. Nagel, W., L. Zeitlmann, P. Schilcher, C. Geiger, J. Kolanus and W. Kolanus, 1998b, J. Biol. Chem. 273, 14853. Nakra, P., V. Manivel, R. A. Vishwakarma and K. V. Rao, 2000, J. Immunol. 164, 5615. Nunomura, W., Y. Takakuwa, M. Parra, J. Conboy and Mohandas, N., 2000a, J. Biol. Chem. 275, 24540. Nunomura, W., Y. Takakuwa, M. Parra, J.G. Conboy and N. Mohandas, 2000b, J. Biol. Chem. 275, 6360. Parsons, S. F., G. Lee, F. A. Spring, T. N. Willig, L. L. Peters, J. A. Gimm, M. J. Tanner, N. Mohandas, D. J. Anstee and J. A. Chasis, 2001, Blood 97, 312. Puu, G., 2001, Anal. Chem. 73, 72. Puu, G., E. Artursson, I. Gustafson, M. Lundstrom, and J. Jass, 2000, Biosens. Bioelectron. 15, 31. Puu, G. and I. Gustafson, 1997, Biochim. Biophys. Acta 1327, 149. 275

Kinning and Edwards Rasooly, L. and A. Rasooly, 1999, Int. J. Food Microbiol. 49, 119. Rhodes, D., 2000, Immunol. Cell Biol. 78,474. Salek-Ardakani, S., J. R. Arrand, D. Shaw and M. Mackett, 2000, Blood 96, 1879. Sassoon, J., H. Lilie, U. Baumann and J. Kohli, 2001, J. Mol. Biol. 309, 1101. Schuck, P., 1996, Biophys. J. 70, 1230. Sheng, N., M. B. Fairbanks, R. L. Heinrikson, G. Canziani, I. M. Chaiken, D. M. Mosser, H. Zhang and R. W. Colman, 2000, Blood 95, 3788. Shenolikar, S., C. M. Minkoff, D. A. Steplock, C. Evangelista, M. Liu and E. J. Weinman, 2001, FEBS Lett. 489, 233. Skladal, P., 1999, Biosens. Bioelectron. 14, 257. Skladal, P., A. Deng and V. Kolar, 1999, Anal. Chim. Acta 299, 29. Spinelli, S., L. G. Frenken, P. Hermans, T. Verrips, K. Brown, M. Tegoni and C. Cambillau, 2000, Biochem. 39, 1217. Thakkar, H., P. A. Lowe, C. P. Price and D. J. Newman, 1998, Kidney Int. 54, 1197. Torigoe, H., R. Shimizume, A. Sarai and H. Shindo, 1997, Nucleic Acids Symp. Ser. 37, 267. Torigoe, H., Y. Hari, M. Sekiguchi, S. Obika and T. Imanishi, 2001, J. Biol. Chem. 276, 2354. Tsoka, S., A. Gill, J. L. Brookman and M. Hoare, 1998, J. Biotechnol. 63, 147. Tsukahara, F., T. Yoshioka and T. Muraki, 2000, Mol. Pharmacol. 58, 1257. V acherot, F., J. Delbe, M. Heroult, D. Barritault, D. G. Fernig and J. Courty, 1999, J. Biol. Chem. 274, 7741. van der Linden, R. H., L. G. Frenken, B. de Geus, M. M. Harmsen, R. C. Ruuls, W. Stok, L. de Ron, S. Wilson, P. Davis and C. T. Verrips, 1999, Biochim. Biophys. Acta 1431, 37. Woods, C. M., G. Neslund, E. Kornbrust and S. F. Flaim, 2000, Am. J. Physiol. Lung Cell Mol. Physiol. 278, L1008. Yagisawa, H., K. Sakuma, H. F. Paterson, R. Cheung, V. Allen, H. Hirata, Y. Watanabe, M. Hirata, R. L. Williams and M. Katan, 1998, J. Biol. Chem. 273, 417. Yang, M., H. L. Chan, W. Lam and W. F. Fong, 1998, Biochim. Biophys. Acta 1380, 329. Zimmer, S., A. Stocker, M. N. Sarbolouki, S. E. Spycher, J. Sassoon and A. Azzi, 2000, J. Biol. Chem. 275, 25672.

276

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 9

INTERFEROMETRIC BIOSENSORS

1DANIEL P. CAMPBELL,PH.D. AND 2CANDICEJ. MCCLOSKEY, PH.D.

~Georgia Tech Research Institute, Atlanta, GA 30332 USA 2Life University, School of Arts and Sciences, Marietta, GA 30060 USA

Interferometry is an optical method that compares differences experienced by two light beams traveling along similar paths. A bioconjugate reaction taking place within one of these beams provides the basis for a biosensor. The most common form chosen for the interferometer involves propagation of the light within planar waveguides, which are preferred for their long interaction length. Planar waveguides have evanescent fields sensitive to index of refraction changes in the volume immediately above the waveguide surface. These fields extend up to 5000 A above the surface. Placing a chemically sensitive film within this region concentrates the reaction within this field. Chemical or physical interactions change the index of refraction causing the propagating light speed, or phase, to change in a direction oppositeto that of the index change. To measure this change, a reference propagating beam, placed adjacent to a sensing beam, is optically combined with the sensing beam to create an interference pattern of alternating dark and light fringes. When chemical or physical changes occur in the sensing arm, the interference pattern will shift producing a sinusoidal output. Several interferometric sensing schemes are presented, and their sensitivities are compared. Sensitivity to a biochemical event is a function of the extent of evanescent field interaction with the bioconjugate reaction as well as the path length of this interaction.

I. Principle of Operation Interferometry is a technique used to measure one of three changes in a propagating light beam's path length, wavelength, or light speed along the path of propagation. The change is reflected as a change in the phase ~ of the light 277

Campbell and McCloskey where the phase is dependent on the path length, refractive index n, and wavelength ~,, as shown in Equation 1. (1)

r = 2 n L n/X

Of these three parameters, the change in refractive index is typically used in biosensor applications. When a bioconjugation event takes place, such as binding of one protein with another, a change in the refractive index of the complex occurs. If the propagating light beam passes through the volume where the binding event has taken place, there will be a change in the speed of the light due to the change in refractive index. To measure this change, a reference propagating beam is used. This reference beam is placed adjacent to the sensing beam, but it does not encounter the event. It is combined with the sensing beam to create an interference pattern of alternating dark and light fringes. Whenever a chemical or physical change occurs in the sensing arm, the interference pattern will shift, producing a sinusoidal output. The type of interferometer that is preferred for biosensing is the Mach-Zehnder interferometer, not the Michaelson interferometer familiar from FTIR. With the Michaelson interferometer, a reference interferogram is run prior to the sample interferogram. Fourier transform of the difference yields the spectrum of the sample. The Mach-Zehnder works better for biosensor applications because it is preferable for a sensor to use a monochromatic light source, have no moving mirrors as in FTIR, and measure only the real part of the index of refraction (not the imaginary part such as absorption). The comparative simplicity of the Mach-Zender mechanics and electronics allows for a very inexpensive and portable device. Lack of complexity in the hardware design may sacrifice discrimination capability; this is made up for by the choice of chemistry in both the sensing and reference arms. The simplest form for the interferometer places it on a planar waveguide, along with the chemistry arms.

2. History The study of optical interference and interferometry have their origins in experiments conducted by Thomas Young in 1803 (Young, 1803). The interference phenomena he explored helped to establish the wave nature of light. Since then, interferometry has typically dealt with through-space/volume measurements. If a bioconjugate reaction takes place in bulk solution, the product will be indistinguishable from the unreacted starting materials. It was not until the bioconjugate reaction was fixed to a surface could the idea of an interferometric biosensor proceed. In 1983, Lukosz and Tiefenthaler (1983a; Tiefenthaler and Lukosz, 1984a), while working on grating couplers for planar 278

Interferometric Biosensors waveguides, discovered that relative humidity changes affected their grating coupling experiments. The humidity changes affected coupling angles and efficiencies, due to the interaction of water vapor with the evanescent fields of these thin waveguide devices. Further experiments exploited these effects as a way to monitor gas and humidity changes (Thiefenthaler and Lukosz, 1984a; 1985). Subsequently, they measured bioconjugate interactions in grating coupler devices and then in planar waveguide interferometric biosensors (Nellen et al. 1988; Lukosz and Tiefenthaler, 1988, 1989; Lukosz, 1991).

3. Technology and State of the Art Optical interferometric biosensors borrow from several rapidly evolving technologies. The waveguides and integrated optical components come out of the optical telecommunication arena. The fabrication methods including photolithography, etching, and deposition are a product of the semiconductor industry. The lasers and detectors are products of the computer/entertainment world with the lasers being taken from CD-ROM/players and the detector arrays from webcams. The fluidic design is borrowed from MEMS research. With such rapid development in these areas, state-of-the-art for optical biosensors is a target moving as fast as the technologies that support it.

3.1. Waveguides The waveguide is the heart of the typical interferometric sensor. The waveguide provides the conduit for the light and the platform for the bioconjugation event. The length of the waveguide provides the sensitivity. The different optical configurations that have been explored all provide similar sensitivities for similar interaction lengths. The minimum components required are a laser source, the waveguide and a detector. Planar waveguides provide an optical platform for interferometry that finds kinship with fiber optics. Fiber optic waveguides have high refractive index cores, which direct the light along the fiber by total internal reflection (Chapter 1). Associated with the guided beam is an evanescent electric field, which extends outward from the fiber into the cladding layer (Chapter 2). All the field is confined to the fiber and the cladding, with none extending into the surrounding environment. The planar waveguide acts like an optical fiber that has been unfurled. The high refractive index guiding layer sits atop a lower refractive index substrate. The evanescent field extends above the waveguide layer and is used to interrogate changes in the refractive index of the environment.

279

Campbell and McCloskey

i

cover I

waveguide

Graded index waveguide

substrate

cover I waveguide

Step index waveguide

substrate

increasing refractive index

h._ v

Figure 1. Graded and step index waveguide profiles.

The planar configuration allows for facile addition of other optical components onto or into the waveguide or substrate. Gratings can be fashioned to couple light in and out, and mirrors, beamsplitters and modulators can be used to manipulate the optical beams. Waveguides can be composed of any optically transparent material such as glassy materials and polymers. The sensitivity of a planar waveguide is dictated by the refractive index of the waveguide, the thickness of the waveguide, and the refractive index of the substrate that contributes to the strength of the evanescent field above the waveguide. Sensitivity increases with increasing difference between the refractive index of the waveguide and that of the substrate. There are two major classes of planar waveguides - those with a graded index and those that are step-indexed (Figure 1). The graded index waveguide has a tapered range of indices of refraction. The outside surface has the highest index, and it tapers down to the substrate's index. This type of waveguide is produced using a substrate (usually glass) that has monovalent ions present, such as sodium, that are exchangeable with silver, cesium, potassium, lithium, thallium and other monovalent ions. The exchange is achieved by placing the substrate in a molten salt bath that contains the ion to be exchanged. The concentration of the ion in the bath, the choice of ion, and the 280

Interferometric Biosensors immersion times all determine the depth of the waveguide, the number of modes, and the difference in refractive index. Further modification can be accomplished by either applying an electric field while the waveguide material is in the molten salt bath or by an annealing process after the bath. Index of refraction changes near the surface of 0.003 to 0.1 have been reported for graded-index waveguides (Millar and Hutchins, 1978; Walker and Wilkinson, 1983; Gato and Srivastava, 1996). Step index waveguides on the other hand, show a step change in index of refraction. This is accomplished by depositing a high index material onto a lower index substrate. For polymer and sol-gel waveguides, the waveguide material can be spin-coated or dip-coated onto the substrate. For other glassy materials, processes borrowed from the semiconductor industry are used: chemical vapor deposition, evaporation, and sputtering. These give finer control of the deposition process and allow one to tailor the sensitivity of the waveguide. The best sensitivity is attained near cutoff, that is, the thickness of the waveguide at which guiding becomes possible. The closer to cutoff the deposited material is, the more evanescent field will exist above the waveguide. One can evaluate the sensitivity of various waveguide/substrate systems by running a salt solution or something similar over the waveguide surface and measuring the phase change interferometrically. As the concentration of the salt solution increases, so does the refractive index of the solution. The graded index waveguides are inherently less sensitive than step index waveguides because only a small perturbation of the index is possible using simple ion exchange. Step index waveguides are best for maximizing the difference between the refractive indices of the waveguide and substrate. For example, an ion exchange waveguide that was produced using a B K-7 glass substrate immersed in a molten salt bath containing 0.25 mole% AgNO3 in NaNO3 for 20 minutes at 325~ generated 0.22 n radians of phase shift for a 0.001 change in refractive index in the cover solution over a 1 cm pathlength. A step index waveguide that was made by depositing 1100/~ of Si3N4 through a chemical vapor deposition technique onto a fused silica substrate produced a 7.6 n radian phase shift for the same 0.001 change in refractive index above the waveguide for the same 1 cm pathlength. This 34-fold increase in interferometric sensitivity is due to the relative difference between the ion exchange waveguides, a few thousandths of a refractive index difference, versus the 0.4 difference between the Si3N4 waveguide and the fused silica substrate. Waveguiding can occur only if a critical refractive index difference and thickness are achieved. Waveguide behavior is understood in elaborate detail (Nishihara, et al., 1985a); for biosensor applications, it is only necessary to know that for light to be guided, the transverse resonance condition must be met.

281

Campbell and McCloskey The transverse resonance condition states that a guided beam must experience a 2rr phase shift in equivalent points of one cycle of propagation and reflections in the ray trace, translated along the waveguide. Equations 2 and 3 define this for the transverse electric field TE and the transverse magnetic field TM light, respectively. Each equation accounts for the phase shift due to transmission through the waveguide media, (2 k nf W cos 0), and the two reflections, one off the waveguide-substrate interface and the other off the waveguide-cover interface. 2 k n f W cos 0 - 2 tanl~(n 2 sin 20 - n s ~ nf cos 0 - 2 tan -1 f~n_f~sin 2 O- n~,~)~

L

1

= 2 rr m

(2)

nfcoso- j

2knfWc~ 2tan~In~-" l 2~-2si nnfcos0 n20-n~Z)l s /2 -2tanl~n~z(n~sin20-' =nc Lcosn02-~12] nf

2rim

(3)

where m = mode number (0,1 . . . ) = film refractive index nc = cover refractive index ns = substrate index refractive W = waveguide thickness in nm

nf

The ray trace and electric field distribution for the first two modes is shown in Figure 2. As the waveguide becomes thicker, or the index difference between waveguide and substrate increases, additional modes are able to propagate. Note that in Equations (2) and (3), an added multiplying factor appears for the TM mode. This results in the requirement for a thicker waveguide in order to support the TM mode. In turn, the TE mode is more buried in the waveguide as the thickness for TM guiding is achieved.

282

Interferometric Biosensors m=O

,e

m=l

.........

n c, Cover

~~/~, nf, W a v e g u i d e

w

Substrate, n s

nc

L

,,,

n_c

nf

nf _

ns

Figure 2. Optical ray and field distribution of first two guided modes.

3.2. Light coupling methods Light must be coupled into the waveguide to perform sensing duties. There are three commonly used methods for coupling the light into the waveguide: endfiring, prism coupling and grating coupling (Figure 3). Each has advantages and disadvantages that become obvious when designing an integrated sensor. End-firing is the simplest way to excite a guided mode in a waveguide. Light is fed into the waveguide from the edge face. The beam is either focused with a lens to the waveguide edge or is fed to this edge by use of a fiber optic. Maximum efficiency occurs when the beam profile closely matches the guided beam profile (Nishihara, 1985b). For the zeroth order mode typically used for sensing, a Gaussian beam supplies the match. The edge of the waveguide is either cleaved, as in the silicon substrates, or polished to a defect-free finish in order to maximize coupling and minimize scattering. With the high efficiencies possible, and the low light requirement for interferometric sensing, some loosening of these requirements is tolerable. End-firing works well when the waveguide is thick, such as in an ion-exchange waveguide. Alignment of the source and waveguide becomes problematic with the single mode, high refractive index, and step-index waveguides. With waveguide dimensions on the order of 283

Campbell and McCloskey

~

-

waveguide

optical fiber

b)

~:..._.._~

Figure 3. Optical waveguide coupling methods, a) end-firing focused beam and endfiring optical fiber; b) prism coupling; c) grating coupling.

0.1 micron, the position of the light to be coupled requires precise manipulation of the light source and waveguide. This is complicated further with channelized waveguides; these require additional positioning. The alignment problem can hinder the integration of a practical sensor, especially one in which the waveguide chip is to be changed frequently. Prism coupling offers high light coupling efficiency, but the prisms must be positioned on the topside of the waveguide. High refractive index prisms are used to excite the guided wave by phase-matching the incident wave with the guided mode, Prisms such as heavy flint glass (n = 2.009) and rutile (no = 2.584; ne = 2.872) are used. The prisms must be in intimate contact with the waveguide surface. Pressure and a scrupulously clean surface are necessary for the contact: Interference lines and/or a shiny patch of reflected light signify adequate contact. A slight bevel is polished into the leading edge of the prism in order to provide a thin air gap across which the light will jump into the waveguide. When the angle 0 from the incident light in the prism equals the propagation constant 13, where 13 = k nf sin 0 and k = 2rd~, and nf is the index of the waveguide, guiding will occur. Changing the angle allows one to address different modes if the waveguide structure will support them. However, the use of prisms complicates sensor fabrication. A screw mechanism is required to apply pressure to the prisms for coupling in and out. A rotation stage is used for adjusting the angle. In our lab, the prism setup is typically used for evaluation of waveguides and not for sensor testing. A major hurdle for prism coupling is the placement of a cell for aqueous testing. If placed in contact 284

Interferometric Biosensors with the waveguide, the gasket around the cell will usually decouple the guided light. Therefore the waveguide design needs some sort of buffering overlayer for the cell to sit on and yet permit the guided beam to travel beneath. More difficulties arise in the transition to and from the prism contact areas and cell mounting areas and the sensing regions. For this reason, few researchers employ the prism coupling method for coupling light into the waveguides. The third method is by means of a grating coupler fabricated with the waveguide structure. The grating is a periodic structure with alternating refractive indices that are rendered into the substrate or waveguide by embossing, etching or ion exchange. Using pressure to imprint a commercially available grating onto a deformable waveguide material such as a sol-gel or polymer produces an embossed grating. For example, the sol-gel SiO2-TiO2 is typically dip-coated onto a substrate and allowed to cure partially. A surface relief grating is pressed against the coating with a pressure of 50-100 lbs. for a few minutes (Lukosz and Tiefenthaler, 1983b; Hengerger and Lukosz 1986; Ramos et al., 1996). Upon release of the pressure, the die is removed and the sol-gel curing is completed thermally. Any shrinkage that occurs must be taken into account in the design of the procedure. Gratings from 0.25 to 0.85 micron, with efficiencies between 10% and 25%, have been formed using this method. Polymers can be used in a similar manner but they are less rigid, less inert and have larger dn/dt's than do sol-gels. Gratings are sealed by adding a protective film over them, such as SiO2, to make them immune to changes in the sensing environment. The SiO2 overlayer and the SiOJTiO2 waveguide materials provide the index contrast for the grating. Alternatively, the grating can be rendered with photoresist using a photomask. The mask can be generated in a number of ways. One method uses two overlapping laser beams. The patterning of the grating can be on the substrate or in the waveguide. The grating is etched with reactive ion etching or wet chemical etching (Hartman et al., 1998). Again, a SiO2 overlayer can be used to seal the grating from the environment. Either the substrate/waveguide or the waveguide/overlayer combination provides the index contrast. These gratings should have efficiencies similar to those of embossed gratings. The ionexchange gratings have much smaller contrast and therefore, poorer efficiencies. Gratings provide the advantage of having the light come in from the bottom of the waveguide thus allowing for placement of the test cell and related fluidics above. Use of gratings has less stringent alignment conditions than does the endfire approach, once the angle is set. Thus, waveguide chips can be routinely replaced, producing a "plug and play" sensor. Gratings, however, are not well suited for coupling into channelized waveguides since beam shape matching is more difficult as is the additional alignment for coupling into the channel. Gratings are wavelength-selective with angle, which can be advantageous if the 285

Campbell and McCloskey

Test Strip Gratings

Microscope Signal , Objective Beam Lens /

/

!111 _

II

Lase;

i111

Beam Splitter

.l

v

""

Reference

Waveguide

Reference Strip

Interference Pattern

Figure 4. Diagram of Mach-Zehnder interferometer.

source is not monochromatic. In addition, gratings of large widths can be used to excite several interferometers on a single chip, at once, with uniform coupling efficiency. 3.3. I n t e r f e r o m e t e r s

A waveguide has an evanescent field that is sensitive to changes occuring in the volume in the immediate vicinity of the waveguide surface. A chemically selective film in this region will provide the means for a guided light beam in the waveguide to interact with it. Chemical or physical interactions change the index of refraction, causing the propagating light speed, or phase, to change (Eq. 1). To measure this change, a reference propagating beam is optically combined with the sensing beam to create an interference pattern of alternating dark and light fringes. A schematic of this optical arrangement appears in Figure 4. When a chemical or physical change occurs in the sensing arm, the interference pattern will shift, producing a sinusoidal output. The phase shift is defined in terms of neff, the effective index change of the waveguide/sensing film/substrate combination. Over a length, L, this change accumulates. (I) -

2___g_~L (An~ff)

(4)

The detection level is limited by noise, both thermal and mechanical, from the waveguide system, the laser stability and the detector characteristics. Detection limits of 10-6, or better, change in refractive index, are typical. 286

Interferometric Biosensors Interferometric sensors hold the promise of direct label-less biodetection, the nulling of thermal and mechanical noise, optical cancellation of nonspecific binding and bulk index changes, optical integration for added stability, compact size and low cost. Several interferometric schemes have been investigated. Each configures the sensing and reference arms in different ways, such as with different beam-combining schemes. Each has advantages and disadvantages, with no configuration clearly eclipsing the others. The fiber optic schemes that require bulk optic/fiber optic hybridization appear to be somewhat cumbersome for practical sensor use. Sensors have evolved from interferometers with discrete and separate optical paths and those with side-by-side channels on a single surface, to those with stacked and colinear arrangements of channels. Waveguide interferometers are an outgrowth of the optical telecommunication industry that developed fiber optics for conduction of light and planar waveguides for integration of the optical circuitry to manipulate optical signals. Fiber optic interferometry initially used a combination of bulk- and fiber-optic components. As more fiber-optic components became available, the interferometer used all-fiber components. However, sensing was restricted to mainly temperature and pressure measurements. The various configurations that fiber optic interferometry can take are outlined in a review article by Kersey (1990). The fiber optic interferometer has been used in a biosensing application by a few investigators (Choquette and Locascio-Brown, 1994) as a temperature sensor to monitor the heat produced by an enzymatic reaction. Choquette loaded an antibody conjugate onto a fiber and placed a reference fiber adjacent to the sensing fiber. When the catalase-labeled antibody was reacted with H202, heat was produced and was measured by the interferometer. Fiber-optic interferometers continue to be developed and used for acoustic, gyroscopic, pressure and temperature measurements. For chemical sensors, however, most fiber optics are used for absorbance and fluorescence measurements. Most interferometric measurements are being pursued using planar waveguides. The planar configuration has the advantage of increased evanescent field, due to the possibility of thinner structures, and the ability to integrate various optical components onto or into the planar waveguide. A critical problem in waveguide interferometry is the design of the reference channel. Since interferometry measures the difference between the sensing and reference arms, the reference must be structured to null out as much of the background signal as possible. The background signal can be caused by thermal differences, mechanical changes, and bulk index effects. Other considerations include local variation in the bulk index of the sampling media and nonspecific reactions. The reference arm should be as close as possible to the sensing arm in structure, response, and distance, except for the specific chemical response, so that the reference serves as a true negative control.

287

Campbell and McCloskey The fiber-optic approach cannot compete with an integrated design since the reference arm is physically discrete from the sensing arm. Planar waveguide interferometry places both arms onto the same physical structure, with spacing as little as a millimeter or less apart. The close spacing essentially negates any thermal or mechanical perturbations. However, bulk refractive index changes can still be bothersome with the close arrangement and must be taken into account in the fluid dynamics of the cell: an ideal system would reflect changes in the bulk simultaneously in both the reference and sensing arms. Two common path interferometric designs are the polarimetric and the two-mode interferometer. These designs attempt to minimize the unwanted differential effects that are seen with two arms. The sensing and reference light paths are colinear, so that the light is passing through the same waveguide volume, thus canceling out the slight variations in thermal and mechanical effects that are seen in the side-by-side arrangement. However, in both designs, any bulk index of refraction changes that occur will be sensed differently by the two modes, or polarizations. In the side-by-side model, fluid flow is important in errant signal suppression; in the colinear design, changes in bulk index of refraction are sensed simultaneously but differently, due to a difference in the extensions of the evanescent fields of the two modes. Most interferometric designs, materials and fabrication processes originated with the optical telecommunication industry. Beam-splitters, Y-junctions and modulators have all been used to design a reference arm that maximizes detection. Ranganath and Wang (1977) have designed a waveguide that uses a Mach-Zehnder interferometric modulator; it is an electronic interaction rather than a chemical signal that changes the phase in one arm of the interferometer. This optical modulator buries the waveguides in order to shield them from environmental effects in the surroundings. Sensors, on the other hand, seek to maximize the effects from the surroundings as the means to alter the phase of the guided light. The arms of the interferometer serve as both interrogator and reference. If the reference arm is buffed beneath an inert film, and the sensing arm is in an exposed region, the sensing arm will be subjected to changes in index of refraction of the surroundings. The changes sensed can be due to a specific interaction, a nonspecific interaction or a change in the refractive index of the medium. Most commonly the change is some combination of all three. In order to detect a specific event such as a specific antibody-antigen binding, the reference is exposed to the medium so that it will be sensitive to any nonspecific or bulk refractive index changes. The design of the reference arm is crucial for maximum detection of the desired analyte. Channelized Mach-Zehnder interferometers have channels fabricated in or on a planar substrate. Y-junctions are used to split and to recombine the optical beams. There are two types of channels: ion-exchange (Helmers et al., 1996; Drapp et al., 1997; Luff et al., 1998) or deposited ribs (Fischer and Mtiller, 1992; 288

Interferometric Biosensors

J

a) IS .....

~

,

_ ~

b)

Ion exchange

Deposited rib

Figure 5. Diagrams of two types of channelized Mach-Zehnder interferometers.

Brosinger et al., 1997; Schipper et al, 1997; Stamm et al., 1998; Weisser et al., 1999) (Figure 5). The ion-exchange method begins with deposition of a patterned metal mask that has the channel areas open. Submersion in an ioncontaining melt diffuses the ions into the channels and creates the waveguide structure. The sensing area is then defined by a deposited overcoat. For ribbed channels, fabrication requires the deposition of a high-index film onto a lower index substrate. The channels are photolithographically rendered, and can be either etched or patterned using a lift-off process. Both methods require lengthy devices due to the fact that the divergence angle in a Y-junction has to be very small. When the angle is small, greater length is required to separate the channels sufficiently for chemical or biochemical functionalization. There is also a taper into the Y-junction; this angle must be designed so that the mode will propagate without converting to other lateral modes. The channel widths also must be set so as to not introduce other lateral modes. The channel edges should be smooth so as to minimize scattering, a problem seen mainly with deposited waveguides. For both types of channel devices, the sensing and reference windows are deposited as an overcoat on the waveguide structure. Light can be end-fired into the waveguide using either a lens to focus, or a fiber aligned with the edge 289

Campbell and McCloskey

I_

coupler

.\nput Figure 6. Mach-Zehnder with a three-waveguide output coupler.

(Nishihara et al., 1985b). V-groove technology for fiber alignment can be used if the waveguide is constructed on a silicon substrate (Fischer and MiJller 1992). The channelized Mach-Zehnder interferometer provides a means for sending a point light source down two or more channels and a way to combine two signals into one interferometric signal. However, it is difficult to design the phase position of the output signal to be near quadrature, to gain the greatest linear sensitivity. A deliberate index change in one of the device arms will allow determination of both phase locations as well as the min-max at the start of an experiment. The quadrature problem has been mitigated (Luff and Wilkinson, 1998) by using a three-waveguide coupler, again borrowing from the telecom industry. The coupler (Figure 6) is comprised of three adjacent waveguides: the two arms of the interferometer and a third, positioned between them. The output signals are offset in phase by 2rd3, providing at least one output near quadrature. The sum of all three outputs is a constant, thus the coupler allows for the monitoring of laser light stability or absorption changes in the interferometer. An interesting variation on the channelized Mach-Zehnder model, the Young interferometer (Brandenburg, 1997; Brandenburg et al., 2000) (Figure 7), cuts off the device after the sensing and reference arms without recombining, and allows the light to exit the waveguide. The divergence of the exiting beams overlaps at some distance from the waveguide and forms an interference pattern. Young's interferometer requires no additional optics and allows for complete monitoring of the fringe pattern. Using a single point detector the min-max of the fringe 290

Interferometric Biosensors

diverging output

interference pattern L ~~, ~--~

input Figure 7. Diagram of Young's interferometer.

pattern can be scanned by translation of the detector or by using a mirror. An array detector will analyze the fringes without translation. A disadvantage of Young's configuration is the distance from the waveguide output to the detector required for maximum fringe resolution. The period of the fringe is equal to ~,L/d, where L is the distance from the waveguide and d is the separation of the two channels. Since the fringe pattern exists in space, placement of the detector can be at any distance and will usually be dictated by detector capability. The spacing of the channels affects the size of the fringe pattern. The fringe pattern increases with channel proximity and will be dictated by the minimum distance needed for functionalizing the sensing and reference chemistries. Using a combination of photolithography and surface energy manipulation, this distance can be brought to the 100 gtm range. The non-channelized Mach-Zehnder interferometer places the beam division and combination off-chip. A beamsplitter and a lens are used instead of a Y-junction. This simplified model (see Figure 4) consists of the slab waveguide with the sensing and reference areas defined by the sensing chemistry and the overlying 291

Campbell and McCloskey films. The light is not launched by end-firing; rather, this system uses either prism coupling or a grating that is fabricated into the waveguide. The nonchannelized Mach-Zehnder interferometer typically uses a lens to combine the two beams. The beam splitter allows for minimum beam spacings of approximately 1 mm, so the fringe spacing is very small. The fringe pattern exists only at the convergence of the interferometric beams. A microscope objective is used to expand the pattern onto either a slit with a point detector or no slit and an array detector. Placing a mirror between the objective and the detector allows for mechanical manipulation of the fringe for both the measurement of the min-max and for setting the pattern at quadrature when using a point detector. One practical approach for determining min-max and setting quadrature (Heideman and Kooyman, 1994) incorporates a phase shifter on one of the interferometric arms. The phase shifter is a rotatable glass slide located externally to the waveguide. Heideman (Heideman et al., 1994, 1996) has placed an electrically addressable electrooptic material, ZnO2, on one of the interferometric arms on the waveguide. Application of a field alters the index of refraction, and therefore, the phase of the beam. This has the effect of shifting the interference pattern of the interferometer. The phase shift can be used to operate the device at its most sensitive and linear portion, quadrature. Alternatively, the voltage required to keep the device at quadrature can be the reading used to quantify any change in Neff due to a sensing event. The non-channelized Mach-Zehnder interferometer has the advantage of ease of fabrication of the waveguide, which allows for facile screening of various sensing chemistries. The major disadvantage is the requirement for multiple optical components to render a sensing scheme. Multiple components add additional mechanical noise, and only one sensing interferometer can be used at a time. The Young' s interferometer version of this sensor takes the two outputs from the waveguide and diffracts them through a slit (Hradetsky and Brandenburg 2000). In fact, if the natural divergence of the laser is used, the two output beams will interact a distance from the output and form an interference pattern with no need for any external output optics. With the colinear polarimetric or difference interferometer, mechanical and thermal disturbances are minimized (Stamm and Lukosz, 1993, 1994; Lukosz, 1995). In this configuration (Figure 8), the sensing and reference beams travel through the same volume with the waveguide. Sensing occurs due to the difference in extensions of the evanescent fields of the two polarizations, the transverse electric and the transverse magnetic. As examined earlier, the TE will propagate through in a thinner waveguide than will the TM mode. A thicker waveguide will confine the TE mode and allow the TM to propagate. The TM mode at this point has more evanescent field above the waveguide than does the TE. The difference accounts for different sensitivities to chemical or physical 292

Interferometric Biosensors X/2

Lens

TE~TM~ i" --_1.~

i ~

V .

.

.

.

.

plate

I [

Beam

Splitter

-! "d

.

~

k/4 plate

W2

.... I ~ ~ ._

D4Detectors D3

)

Wollaston Prisms

D1 D2 Detectors Figure 8. Diagram showing the guided wave ouput of a polarimetric (or difference) interferometer.

changes above the waveguide for the two modes. That the modes experience a different change in phase for the same effect is the source of the sensor's response. Both polarizations are coupled into the waveguide by rotating the laser approximately 45 ~ from normal. The approximation is noted because different coupling and propagation efficacies of the two polarizations require a slight biasing for equivalent signals. Both polarizations propagate down the waveguide with the TM mode experiencing a change in phase preferentially over the TE mode. Both modes are coupled out, then separated into different polarizations by using a beamsplitter, two Wollaston prisms, a ~,/2 and ~./4 plate, and four photodetectors to measure the phase difference (Lukosz et al., 1997). This array of optical components can be further simplified to either a Wollaston prism and polarizer, or a grating, lens and polarizer, with the resulting interference pattern imaged on a CCD array (Lukosz et al., 1997). The pattern appears orthogonal or off-axis, respectively, for the two schemes, allowing for the possibility of multiple sensors on one substrate and analysis by a single twodimensional array detector (Lukosz et al., 1997). In order to take full advantage of the polarimetric scheme, the waveguide needs to be designed so as to maximize the difference in sensitivity between the two modes. This is best accomplished by using thin single-mode step-index waveguides that are just slightly thicker than what is needed for the propagation of TM light. The pioneer researchers in this area employed sol gel-based TiO2293

Campbell and McCloskey SiO2 waveguides with refractive indices in the 1.8-2.0 range and an optimal thickness in the 2000 ~ range. Owing to the fact that the reference beam of the interferometer, the TE polarized beam, is not completely blind to the specific binding events and is therefore somewhat responsive, the magnitude of the response to the differences between TM and TE light is diminished. Depending on the waveguide system, this decrease in response can be as great as 50% less than that seen in the side-by-side Mach-Zehnder interferometer model. In addition, the TE and TM modes respond to different degrees to index change in the bulk media. Lack of a well-designed reference will hurt the sensitivity of the difference interferometer to specific binding events as well as the ability to null variations in bulk index changes. The gain in mechanical and thermal stability for this design may not be enough to compensate for these shortcomings unless bulk index changes can be minimized and change due to specific binding enhanced over any nonspecific binding. The polarimetric or difference interferometer cannot distinguish between a binding event and a change in bulk refractive index and/or temperature. However, one attempt to deal with this shortcoming uses two wavelengths at the same time in order to provide additional information to differentiate two effects, either binding and bulk index change, or binding and temperature (but not bulk index change and temperature) (Stamm et al., 1998). This design requires an additional laser; in this case, an Argon-ion 488 nm laser was added to the He-Ne operating at 633 nm. However, a semiconductor laser could also be used. But coupling the light in is limited to end-firing unless one is willing to deal with two input angles into a grating. Interestingly, it has been suggested that two difference interferometers of differing waveguide thicknesses operating at the same wavelength could also make this differentiation, largely because of the analogous sensitivity difference that the waveguides possess compared to the two wavelength and one waveguide thickness models (Lukosz 1995). The same accounting for both the bulk index and temperature changes in addition to a binding event is built into the side-by-side Mach-Zehnder device. The need for multiple optical components to analyze the polarimetric or difference interferometer's output is a disadvantage to compact sensor design. Recent work (Koster et al., 2000) has integrated the difference interferometer's polarization optics onto a waveguide chip (Figure 9). The device employs a polarization converter to convert some of the singly polarized TE input beam to TM mode. The two beams travel under the sensing area to another polarization converter that allows mixing, followed by decoupling of the two modes in a sequential manner. The device features detectors that are integrated in the supporting silicon substrate. Only the laser remains external.

294

Interferometric Biosensors

Side View of Output Area

Top View Polarization Converters

TE TM

,

/ TM

I

....

TM TE detector detector

TE

detector detector

Figure 9. Diagram of an integrated polarimetric (or difference) interferometer.

Another difference interferometric design (Hartman et al., 1988) uses differential evanescent fields from two separate modes with the same polarization. Although patented (Hartman, 1990), the device has never produced a biosensor, most probably due to the problems involved with the different coupling angles that are required for each of the two modes. The desirability of sampling for several analytes at once, or being able to analyze several samples sequentially without the need for regeneration or changing chips, can be realized with the multiple non-channelized Mach-Zehnder interferometer chip (Hartman et al., 1998). This design has thirteen interferometers on a single chip that measures 1 x 2 cm. A single laser is fanned out onto a broad grating, which launches light into all the interferometers at once. The waveguide consists of a deposited slab of SiaN4 on glass. The channel lengths are defined by using patterned thick SiO2. The sensingchemistry is placed within the channels. The channels are defined by the chemistry immobilized on the waveguide surface rather than by the beams themselves. The broad beams are combined, and then travel through a sequence consisting of a reflector, a beamsplitter, than another reflector. The resulting interference signal is coupled out using another broad grating, and an array detector analyzes the multiple interference patterns (Figure 10). The number of interferometers in this design is determined by the pixel pitch of the detector, along with the need for at least three pixels required to define the sine wave of the fringe. Recent drops in the cost of detectors with tighter pixel pitch have opened the door to increasing the number of interferometers possible on a single chip. The arrays in USB (Universal Serial Bus) cameras provide pixel pitches that are around 10-15 microns, which is equivalent to approximately fifty interferometers per square centimeter. 295

Campbell and McCloskey S~nefSernegnand Chemistry Strips ~

.q,p~ ~ =~

Detail of BeamCombiningOptics

Figure 10. Diagram of a chip with thirteen integrated non-channelized Mach-Zehnder interferometers.

3.4. Other interferometric methods A few other interferometric schemes have been investigated during the last decade, and include: the use of Fabry-Perot fringe generation with white light, Zeeman interferometry, a stacked Young's interferometer, and phase information from surface plasmon resonance. The Fabry-Perot approach has been explored at the end of a fiber and on the surface of a porous silicon chip (Gauglitz et al., 1993). Here, changes in thickness and index of refraction caused by binding of proteins, or by the swelling of a polymer at the end of a fiber, alter the pathlength of white light being sent down the fiber. The interference of the reflected light produces FabryPerot interference lines. A change in the reflection due to binding or swelling shifts these fringes. The amount of shift is indicative of the change. This reflecto-metric interference spectroscopy lacks sensitivity because the pathlength change is so small compared to the planar waveguide pathlength. The case of the reflection interference off a porous silicon substrate caused a stir when it was first presented (Lin et al., 1997), since it had unexpectedly high sensitivity. Porous silicon contains etched pores with very high surface area. White light shining on the porous area will reflect off both the top of the substrate and the bottom of the pores. The two reflections interfere to produce Fabry-Perot fringes. These fringes shift when the pores are functionalized with antibodies and then exposed to the conjugate antigen. Even though the pathlength difference between the two reflections is small and the pores only 1-5 microns in depth, the change is too large to account for mass loading alone but is 296

Interferometric Biosensors thought to arise from modification of carrier concentrations in the semiconductor due to the binding. Further analysis showed that the results were false; that the change seen was due to oxidation of the silicon pores (Dancil et al., 1999). However, the effect from reactions involving of the pore material itself have been exploited to detect fluorophosphate nerve agents from the etching of the silicon dioxide by the HF product of nerve agent hydrolysis (Sohn et al., 2000). Effects due to etching of the base material have also been utilized with planar waveguides, and to a greater degree because of the increased pathlength. Campbell (unpublished data) found it easy to monitor the etching of a Si3N4 Waveguide in weakly basic conditions, pH 8, to fewer than 2 , k ~ . Zeeman interferometry uses a Zeeman laser and a planar waveguide. This interferometer takes advantage of the two frequencies that are generated by the Zeeman laser; in this case, they are separated by a 250 kHz difference (Grace et al., 1997). As the two modes propagate through the waveguide, a phase difference accumulates between the modes as a result of a chemical change in the sensing layer. The 250 kHz beat frequency sine wave, generated when the two modes are combined, is measured at the output and compared to the reference sine wave from the laser. The phase difference is indicative of a change in the coating on the waveguide. Although this example is a chemical sensor, this technique could be used for measuring a bioconjugate reaction. As is the case of the polarimetric interferometer, this colinear scheme is also prone to sensing bulk index differentials. A stacked Young's interferometer has also appeared recently (Cross et al., 1999). Here, the waveguides are layered together, with a buffer medium separating the waveguides vertically. The diffracted light creates the interference pattern as it exits the waveguide. This elegant approach could be very useful in chemical sensing, where a buried reference is required. However, for biosensors, the buried reference prevents easy accounting of bulk index changes. Finally, Nikitin has used a unique approach that investigates the phase change information not usually measured in surface plasmon resonance (SPR) experiments (Kabashin and Nikitin, 1998; Grigorenko et al., 1999; Nikitin et al., 1999). The open path interferometer uses the metallized surface of an SPR prism as in a Mach-Zehnder setup with an expanded beam. Though not measuring the typical angular change, the reflected beam exiting the prism is combined with the reference beam and produces the interference. A sensitivity of 4 x 10-8 in refractive index change was estimated based on the measurement of different gases. Imaging of the surface is also possible. This approach could easily find application in combinatorial genomics or proteomics without the need for a fluorescent or colorimetric label.

297

Campbell and McCloskey 3.5. Surface functionalization

After a transducer is chosen, a bioconjugate receptor is immobilized on the transducing surface. A glassy surface such as Si3N4, TiOz, or SiO2, is populated with hydroxide moieties. In the case of Si3N4, simply cleaning the surface replaces amine groups with hydroxides. Protein attachment chemistry uses methods developed for affinity chromatography. Several books have been written on the subject (Hermanson 1996). Organosilane chemistry is used to attach reactive functionalities to the glass surface. These silanes usually terminate in amino, carboxaldehye, or sulfhydryl groups. A protein receptor will bind to an amino group after prior oxidation, or to a carboxaldehyde by reductive amination. Sulfhydryls will form disulfide bonds with cysteine. Any number of bifunctional linkers can also be used to form a long tether to the protein, thus increasing its mobility. Other schemes employ attachment of protein A to the surface, followed by binding to the Fc portion of an antibody. Avidin is also employed as a functionalized surface for biotinylated proteins and DNA, taking advantage of the strong binding constant of the avidin-biotin system. Subsequent backfilling with protein filler is used to minimize nonspecific binding. The reference arm of the interferometer, if the transducer uses one, is handled in a similar fashion. A dummy protein is attached, and the remaining surface is backfilled with a protein. The object is to make the reference equivalent to the sensing arm in all ways except the specific binding event that one is attempting to sense. Unfortunately, it is not that simple. Nonspecific binding of errant protein is typically quite prevalent, especially in clinical samples; therefore, the nulling out of the nonspecific signal is not easily accomplished. Much work has been done to minimize the nonspecific signal by using surfaces that resist the adsorption of proteins. Polyethylene glycols and their derivatives have garnered the most attention and appear to work relatively well. Various other surfaces with amide, phosphoamide, urea, and variations of polyethylene glycols have also been compared (Chapman et al., 2000). The ability to limit nonspecific binding sets the detection limit for the colinear polarimetric scheme, since both specific and nonspecific interaction look identical to the transducer. The side-by-side interferometer also relies on the minimization of nonspecific binding to set its detection limit. If nonspecific binding predominates, one could be seeing a small difference between two large changes. Separate analysis of both arms would at least quantify the bioreception signal-to-noise and allow a confidence level on the detection.

298

Interferometric Biosensors

3.6. Applications The literature presents different interferometric sensing schemes. With most schemes, the goal is to show the sensitivity and versatility of the interferometric design. Research has not focused on unique solutions to specific sensing needs, especially since the sensor's ability to detect specifically is no better than the chemistry on the transducer. With that in mind, most researchers have evaluated a specific design, usually with a generic bioconjugate system. Many have used an IgG/anti-IgG test system. DNA samples were of a generic nature with some 20 base pair matches (Schneider et al., 1997). Sometimes, whole cells like Salmonella have been used (Schneider et al., 1997; Seo et al., 1999). All have used direct label-free detection systems, taking advantage of the underlying strengths of the interferometer. In an attempt to gain additional sensitivity, a sandwich assay has been used. In this scheme, a second antibody is added after the first has reacted with the antigen. The second antibody might be linked to a microparticle such as gold or polystyrene. These microspheres provide amplification by adding large index of refraction changes (Schneider et al., 2000). Competitive assays using a waveguide inteferometer have centered on detecting atrazine, a herbicide, in drinking water, as a test case. An atrazine hapten is synthesized then covalently attached to the waveguide. The drinking water sample is introduced to the waveguide. Next a portion of the atrazine antibody is added to the solution. The dissolved atrazine competes with the bound atrazine hapten for antibodies. If there is no free atrazine in solution then there is much binding to the bound atrazine hapten, and therefore, a large signal. Conversely, if dissolved free atrazine concentrations are large, little antibody binds to the surface and a small signal is produced. This scheme detected down to 0.1 lxg/L or 0.1 ppb atrazine (Schipper and Bergevoet, 1997; Schipper and Rauchalles, 1998). Since the attached bioconjugate is not a protein, the surface can be regenerated without loss of activity and is ready for the next analysis. Indirect assays analogous to traditional ELISAs, with an interferometer serving as the reader, have been studied only briefly (Campbell and Hartman, 1998). The advantage of this approach is the reversability of the sensor for the next analysis; however, the ELISA carries additional steps. It is not necessary for a sensor to measure a color change; it can respond to some other enzymatic by-product such as ammonia from urea. In addition, other clinically relevant analytes such as ammonia and pH can be measured with an interferometer. In one case, the sensing films are thicker than the evanescent field, thus they block effects from nonspecific binding and bulk index changes. The thicker films provide enough reactive material to measure

299

Campbell and McCloskey ammonia to the sub-ppm levels, and pH to 0.01 units in aqueous solution (Campbell and Moore, 1998). Though most examples of biosensing have been in nonchallenging media, the interferometric scheme has been used for sensing analytes in blood, serum, and in the case of Salmonella, chicken carcass washings ( Keo et al., 1999).

4. Advantages and Limitations Many different interferometric and reflectrometric waveguides have been presented. Each was evaluated by exploring a bioconjugate binding reaction. It would be unfair to compare each directly. For the waveguide-based interferometers, the inherent sensitivities in the design prove them to be essentially equivalent. The reflectometric and fiber optic-based interferometers are far less sensitive due to their limitations: the small pathlength change in the reflectometric model and the minimal evanescent field in the fiber optic model. Interferometers can be prepared using different materials for both the waveguide and the substrate. The choice is ultimately limited by the investigator's available facilities. If all researchers could employ the same materials and thicknesses, there would be no distinction in detection limits, based on waveguide sensitivity and the change in n~ee. This would yield a universal limit of binding effects of 1 pg/mm 2 of coverage. What finally limits detection is physical: the detector scheme, the ability to suppress mechanical and thermal noise, bioconjugate binding constants, and fluidics. The researcher can control other aspects as well, such as the mechanics of sample introduction, diffusion and bulk index variation. The major advantage of an interferometer is that it can detect any chemical reaction, since all reactions and interactions produce a change in refractive index. Therefore, there is no need for labels, color change, or resonance effect to produce a response. The interferometer provides a universal platform for measuring any chemical event. But this advantage is also its major limitation. Being sensitive to any chemical event, the interferometer will respond to any change, be it the specific reaction, any nonspecific adsorption, temperature change, or background bulk index changes. Thus, the design of the sensing and reference chemistry must be specific and also must diminish any interference. The reference arm should respond to the desired event, as well as minimize any competing and background events that cause refractive index changes. The interferometer also requires exacting design to limit errant signals. The limitation of expense of lasers and detectors has evaporated over the years, and should provide little barrier to the development of interferometric sensors, as well as other optical sensor schemes. For the design of the waveguide itself, an array of thin film deposition systems is available at most institutions. If not, the 300

Interferometric Biosensors sol-gel waveguide is an inexpensive and adaptable alternative. The limiting factors for most optical devices are the electronics and software development needed for signal processing and data interpretation. It may be quite a leap for many to change from having a computer tethered to a sensor to having a complete self-contained sensor unit.

5. Potential for Improving Current Performance Except for the polarimetric scheme, which required four detectors, sensor configurations have employed one detector. Single point detection has held down sensor costs. But the inability to know one's position on the sine curve with a single point limits an exact quantification of the phase change. Scanning a fringe requires a mechanical device to shift the fringe to quadrature. Electrooptic phase shifters incorporated in the device will suppress any mechanical noise. However, having more points sampled along the fringe curve would further improve detection limits. Three points define a sine wave, but an array would allow even more points to be sampled. Multiple fringes are possible even with the small pixel pitches available. With the drop in price of detector arrays from around a thousand dollars to fewer than fifty for a USB camera, there is no reason to use anything else. Processing of a multiple fringed pattern shift has yielded phase resolution to less than 0.01 radians. The multiple fringe data analysis can also differentiate between a change in intensity versus a change in phase, while eliminating any laser fluctuations or sample absorbance changes. As detector resolution is increased, other sources of error will in turn need to be suppressed in order for the interferometer to reach its full potential. Controlling the temperature of the sample and of the transducer can minimize thermal noise. This is typically done with a Peltier device. Mechanical noise can be suppressed by integrating as much of the optical path as possible onto the waveguide. Gratings, reflectors, beam splitters and detectors have all been integrated onto one chip; only the laser remains external. It is only a matter of time before the laser is also incorporated onto the chip, at least in some hybrid form. Fluidics for sample introduction have yet to be engineered in order to reduce chemical noise. As the sample enters the cell above the waveguide, there occurs a change in refractive index. It may be a very slight change, but both arms of the interferometer must see it simultaneously or there will be a phase difference between them. Some researchers propose ignoring the initial data, and focus only on the subsequent phase changes from binding events. This could be a valid approach, unless one is automating the screening of numerous samples, where the flow may be constantly changing. Close proximity of the sensing arms will minimize the differential. The colinear polarimetric interferometer is uniquely qualified for lateral index variation but not for vertical variation. Fluidics and cell design need to be addressed in the future. 301

Campbell and McCloskey The rate-determining step in detection is the rate of diffusion of the analyte to the sensor surface. Large protein diffusion is slow and may require up to an hour for the sensor to reach equilibrium. Fluidic design could increase the delivery of the analyte to receptors on the surface. Presently, the two-dimensional sensor area of the interferometer does not fully exploit the evanescent field. The development of a third dimensional sensing matrix would take advantage of a larger share of the evanescent field and shorten detection times. The development of interferometric sensors appears to be leading along two divergent paths, each driven by application. One path leads toward integration of all optical components: to have the source, waveguide and detector all on a single chip. This approach adds value to the waveguide in that, economically, it requires long-term use. This type of sensor would find application as a long-term stand-alone used for in-line food monitoring and bioterrorism detection. The other path leads to a simple waveguide design with inexpensive minimal waveguides that can be discarded or recycled after a measurement is completed. Waveguides that have only a means to couple light in and out, along with the sensing chemistry, would f'md use in clinical applications where discarding is preferable to the possibility of contamination. The sensing chip would contain the source, beam combining optics, detector and processor. It would act as a "plug and play" component to an optical system. With clever chemists to functionalize a surface to detect the required compound to the exclusion of interferents, the current developments in fluidic design, and the advancement of signal processing and interpretation in ever cheaper and smaller packages, there will be a future for the sensing of biochemical reactions using "light speed."

6. Acknowledgements The authors would like to thank Sheree Collins of GTRI for generating the figures used in this chapter.

7. References

Brandenburg, A., 1997, Sens. Actuators B 38, 266. Brandenburg, A., R. Krauter, M. Kiinzel and H. Schulte, 2000, Appl. Opt. 39, 6396.

302

Interferometric Biosensors Brosinger, F., H. Freimuth, M. Lacher, W. Ehrfeld, E. Gedig, A. Katerkamp, F. Spencer, K. Cammann, 1997, Sens. Actuators B 44, 350. Campbell, D.P., N.F. Hartman, J.V. Suggs, J.L. Moore and J.M. Cobb, 1998, Proc. SPIE 3253, 20. Campbell, D.P., J.L. Moore, J.M. Cobb, N.F. Hartman, B.H. Schneider and M.G. Venugopul, 1998, Proc. SPIE 3540, 153. Chapman, R.G., E.Ostuni, S. Takayama, R.E. Holmlin, L. Yan and G.M. Whitesides, 2000, J. Am. Chem. Soc. 122, 8303. Choquette, S.J. and L. Locascio-Brown, 1994, Sens. Actuators B 22, 89. Cross, G., Y. Ren and N.J. Freeman, 1999, J. Appl. Phys. 86, 6483. Dancil, K-P.S., D.P. Greiner, and M.J. Sailor, 1999, J. Am. Chem. Soc. 121, 7925. Drapp, B., J. Piehler, A. Brecht, G. Granglitz, B.J. Luff, J.S. Wilkinson and J. Ingenhoff, 1997, Sens. Actuators B 38, 277. Fischer, K. and J. Muller, 1992, Sens. Actuators B 9, 209. Gato, L. and R. Srivastava, 1996, Opt. Commun. 123,483. Gauglitz, G., A. Brecht, G. Kraus, and W. Nahm, 1993, Sens. Actuators B 11, 21. Grace, K.M., K. Shrouf, S. Honkanen, P. Agr~is, P. Katila, M. Leppihalme, R.G. Johnston, X. Yang, B. Swanson, B. and N. Peyghambarian, 1997, Electronic Lett. 33, 1651. Grigorenko, A.N., P.I. Nikitin and A.V. Kabashin, 1999, Appl. Phys. Lett. 75, 3917. Hartman, N.F., D.P. Campbell and M. Gross, 1988a, Proc. IEEE-LEOS '88,298. Hartman, N., 1990, U.S. Patent No. 4,940,328. Hartman, N.F., J. Cobb and J.G. Edwards, 1998b, Proc. SPIE 3537, 302. Heideman, R.G., R.P.H. Kooyman and J. Greve, 1994, Biosen. Bioelectron. 9, 33. Heideman, R.G., G.J. Veldhuis, E.W.H. Jager and P.V. Lambeck, 1996, Sens. Actuators B 35, 23. Helmers, H., P. Greco, R. Rustad, R. Kherrat, G. Bouvier and P. Benech, 1996, Appl. Opt. 35,676. Hermanson, G.T., 1996, Bioconjugate Techniques, Academic Press, San Diego, 785 pp. Heuberger, K. and W. Lukosz, 1986, Appl. Opt. 25, 1499. Hradetzky, D. and A. Brandenburg, 2000, Europtrode V, 169. Kabashin, A.V. and P.I. Nikitin, 1998, Opt. Commun. 150, 5. Kersey, A.D., 1990, Proc. SPIE 1367, 2. Koster, T., N. Posthuma and P. Lambeck, 2000, Europtrode V, 179. Lin, V.S.-Y, K. Motesharei,K-P.S. Dancil, M.J. Sailor and M.R. Ghadiri, 1997, Science 278, 840. Luff, B.J., J.S. Wilkinson, J. Piehler, U. Hollenback, J. Ingenhoff and N. Fabricius, 1998, J. Lightwave Tech. 16, 583. Lukosz, W. and K. Tiefenthaler, 1983a, 2"a Eur. Conf. Integrated Optics, Florence, IEEE Conf. Proc. 227, 152. 303

Campbell and McCloskey Lukosz, W. and K. Tiefenthaler, 1983b, Optics Lett. 8, 537. Lukosz, W. and K. Tiefenthaler, 1988, Sens. Actuators 15,273. Lukosz, W. and K. Tiefenthaler, 1989, J. Opt. Soc Am. B 6, 209 Lukosz, W., 1995, Sens.Actuators B., 29, 37. Lukosz, W., C. Stamm, H.R. Moser,R. Ryf and J. Diabendorfer, J., 1997, Sens. Actuators B. 38, 316. Millar, C.A. and R.H. Hutchins, 1978, J. Phys. D; Appl. Phys. 11, 1567. Nellen, Ph.M., K. Tiefenthaler and W. Lukosz, 1988, Sens. Actuators 15,285. Nikitin, P.I., A.A. Beloglazov, V.E. Kochergin, M.V. Valeiko, and T.I. Ksrenevich, 1999, Sens. Actuators B, 54, 43 Nishihara, H., M. Haruna, T. Suhara, 1985a, Optical Integrated Circuits, McGraw-Hill, USA Chapter 2. Nishihara, H., M. Haruna, T. Suhara, 1985b, Optical Integrated Circuits, McGraw-Hill, USA, p.226. Ramos, B.L., S.J. Choquette and N.F. Fell, Jr., 1996, Anal. Chem. 68, 1245. Ranganath, T.R. and S. Wang, 1977, IEEE, J. Quantum Electron. QE-13, 290. Schipper, E.F., A.J.H. Bergevoet, R.P.H. Kooyman and J. Greve, 1997, Anal. Chim. Acta 341, 171. Schipper, E.F., A.M. Brugrnan, C. Dominguez, L.M. Lechuga, R.P.H. Kooyrnan and J. Greve, 1997, Sens. Actuators B 40, 147. Schipper, E.F., S. Rauchalles, R.P.H, Kooyman, B. Hock and J. Greve, 1998, Anal. Chem. 70, 1192. Schneider, B.H., J.G. Edwards and N.F. Hartman, 1997, Clin. Chem. 43, 1757. Schneider, B.H., E.L. Dickinson, M.D. Vack, J.V. Hoijer, and L.V., Howard, 2000, Biosens. Bioelectron. 15, 13. Seo, K.H., R.E. Brackett, N.F. Hartman, N.F. and D.P. Campbell, 1999, J. Food Protect. 62, 431. Sohn, H., S. l.Atant, M.J. Sailor, and W.C. Trogler, 2000, J. Am. Chem. Soc. 122, 5399. Stamm, Ch. and W. Lukosz, 1993, Sens. Actuators B 11, 177. Stamm, Ch.and W. Lukosz, 1994, Sens. Actuators B. 18, 183. Stamm, Ch., R. Dangel, and W. Lukosz, 1998, Opt. Commun. 1253, 347. Tiefenthaler, K. and W. Lukosz, 1984a, Optics Lett. 9, 137. Tiefenthaler, K. and W. Lukosz, 1984b, Proc. SPIE, 514, 215. Tiefenthaler, K. and W. Lukosz, 1985, Thin Solid Films 126, 205. Tiefenthaler, K. and W. Lukosz, 1989, J. Opt. Soc. Am. B 6, 209. Walker, R.G. and C.D.W. Wilkinson, 1983, Appl. Opt. 22, 1029. Weisser, M., G. Tovar, S. Mittler-Neher, W. Knoll, F. Brosinger, H. Greimuth, M. Lacher and W. Ehrfeld, 1999, Biosens. Bioelectron. 14, 409. Young, T., 1804, Phil. Trans.R. Soc. London 94, 1.

304

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 10

GENETIC ENGINEERING OF SIGNALING MOLECULES

AGATHA FELTUS, PH.D., AND SYLVIADAUNERT, PHARM.D., PH.D.

Departments of Chemistry and Pharmaceutical Sciences University of Kentucky Lexington, KY, USA

In order to expand the capabilities of biosensors, there is a need to develop new signaling molecules. This chapter focuses on molecules, produced through genetic engineering, that combine the recognition element with a signaling element (such as a fluorophore) in an effort to optimize the signal caused by the binding of the analyte to the recognition element. These systems, while not necessarily originally developed for an optical fiber, can be immobilized at the tip of the fiber either through chemical attachment or entrapment behind a membrane. Three different systems will be examined: fluorophore-labeled binding proteins, FRET-based systems, and bacteria-based sensors. These systems use optical signaling methods to reveal the binding event, taking advantage of molecular biological techniques to optimize the signal. The advantages and disadvantages of each system will be discussed, as well as the current state of the art of these biosensors.

1. Technical Concept In its simplest terms, a biosensor is a sensing system composed of a biological recognition element and a transducer. Under the strictest definition of the term, the transducer is responsible for converting the binding event into an electrical signal. Bacteria that fluoresce upon analyte binding and fluorophore-labeled binding proteins have been refered to as "reagentless biosensors" or "reagentless biosensing systems" even though they are used as assays, rather than 307

Feltus and Daunert immobilized in a sensor. There is no reason these sensing systems cannot be used as the recognition/signaling element of a fiber optic biosensor, however. These systems will be discussed in both contexts, i.e., both free in solution as assays and after adapation to a fiber optic probe. In order to avoid confusion, we will refer to the source of the optical signal as the signaling molecule.

I.I. Fluorophore-labeled binding proteins As the name implies, a fluorophore-labeled binding protein consists of two distinct moieties: the fluorophore and the binding protein. The fluorophore is covalently attached to the binding protein through the protein's amino acid side chains. Binding of the analyte to the binding protein causes a conformational change in the protein structure which may result in altered optical characteristics of the fluorophore (Figure 1). Depending upon the actual environmental change experienced by the fluorophore, this could result in an increase or decrease in the fluorescence intensity, a change in the emission wavelength, or a change in the lifetime of the fluorophore, depending upon the characteristic to be measured. These changes are generally caused by two separate phenomena: an increased or decreased polarity in the environment surrounding the fluorophore or rotational constraint of the fluorophore. For example, if the fluorophore moves from a position of high polarity (exposure to the buffer or the presence of local side chains from polar amino acids) to one of low polarity (a position inside the protein), the fluorescence will increase due to decreased quenching by the solvent molecules or dissolved oxygen in the solvent. Likewise, rotationally constraining the fluorophore's motion by trapping it inside the protein will increase fluorescence by removing frictional energy loss caused by fluorophore movement. These changes are difficult to predict beforehand and are often only revealed once the protein has actually been labeled. Having said this, even if the direction of fluorescence change cannot be predicted, knowledge of the protein structure serves as a good starting point for choosing the placement of the fluorophore. From the point of view of signal-tonoise ratio, it is most advantageous to have a single fluorophore placed in a location where a large environmental change can occur. Often, the most likely location for such a change is near the binding site. Therefore, most initial studies are conducted by labeling at a site near the binding site as determined from observations of the crystal structure or from mutational studies. Selective labeling of the binding protein is usually accomplished via labeling through cysteine residues using sulfhydryl-selective fluorophores. (For examples of commonly employed fluorophores, see Figure 2.) In order to create one-toone conjugates of fluorophore to protein, molecular biology is often necessary to create recombinant proteins with unique cysteine residues. Using recombinant

308

Genetic Engineering of Signalling Molecules

Figure 1. Schematic of a fluorescently labeled protein sensing system. The protein is labeled with an environmentally-sensitive fluorophore such that the binding of the analyte changes the conformation of the protein, altering the solvation of the fluorophore. a) In this example, amino acid 197 of phosphate binding protein (PBP) is located near the binding pocket and will undergo a change in environment as PBP closes around its ligand, phosphate. This can result in either b) an increase or c) a decrease in fluorescence upon ligand binding. In some cases, the emission wavelength of the fluorophore can also change.

DNA techniques, such as site-directed mutagenesis, all other cysteines in the protein are removed and other residues that will be the site of attachment are individually changed to cysteines. In doing so, care must be taken not to alter any amino acids necessary for the proper functioning of the protein, such as those residues involved in analyte binding or in oligomerization of the protein. This entire process will be examined in greater detail in Section 3.1. 309

Feltus and Daunert N(CH3)2

H2C-'----C~~C~

II

"

(Q-t 3(:1-t

0

2)2N~

O

H O~ N

O_ ~ IVD(X3

II 0

~

(OH3OH2)2N

)

O

O

HN~

N--C--CH21 1 ,,54 ~ / g , ~

8Doll

Figure 2. Structures of some cysteine-reactive environmentally sensitive fluorophores. The reaction with the protein takes place through the maleimide or iodoacetimide groups.

1.2. FRET-based systems Another method which has been used extensively to develop sensing systems, particularly in small volumes and inside living cells, is the use of FRET-based sensing systems (Giuliano and Post, 1995; Giuliano and Taylor, 1998) (Figure 3). Fluorescence resonance energy transfer (FRET) occurs when one fluorophore, a donor, nonradiatively transfers its energy to a second fluorophore, the acceptor. The acceptor then relaxes normally, producing light at its emission wavelength. In order for this to occur, there must be a significant overlap between the emission spectrum of the donor and the excitation spectrum of the acceptor. An important property of FRET is that the rate of energy transfer between the donor and the acceptor is proportional to the inverse sixth power of the distance between the two fluorophores (FRET ~ l/r6). For most pairs the F~3rster radius, 310

Genetic Engineering of Signalling Molecules

Figure 3. FRET-based sensing system for cAMP based on protein kinase A (PKA). This system consists of a cell line transfected with a vector coding for two fusion proteins: PKA regulatory subunit-BFP (blue fluorescent protein) and PKA catalytic subunit-GFP (green fluorescent protein). In the absence of cAMP, the regulatory and catalytic subunits associate, bringing the BFP and GFP moieties in close proximity and allowing FRET. The presence of cAMP dissociates the complex of regulatory and catalytic subunits, disrupting FRET. Adapted from Zaccolo et al., 2000.

the distance at which the efficiency of energy transfer is 50%, is between 20A and 50/~ (Lakowicz, 1983). This distance is comparable to the size of most proteins, which allows FRET to be used when the distance between the two fluorophores will be significantly changed by the binding event. This can occur if either both fluorophores are attached to the same protein molecule and binding of a ligand to the protein causes a conformational change that either shortens or lengthens the distance between them, or if the donor is attached to one of the binding molecules and the acceptor to the other. In the latter case, a donor fluorophore attached to one of the components can transfer its energy to an acceptor fluorophore attached to the other only while the two are closely associated. An example of this is given in Figure 3. FRET as a detection methodology has a number of advantages for biosenor applications. Because the system employs the excitation wavelength of the acceptor and the emission wavelength of the donor, the Stokes shift is more pronounced than for fluorescence, leading to a lower background. Another advantage of FRET is that the ratio of fluorescence intensities of the two 311

Feltus and Daunert

Figure 4. Schematic of a bacteria-based sensing system. The bacteria are transformed with a plasmid containing the reporter gene under the control of an analyte-sensitive promoter. In the presence of the analyte, the regulatory protein is released from the promoter region, allowing transcription of the reporter gene. The mRNA is then translated into protein, which can be assayed. The amount of protein produced is proportional to the amount of analyte present, although there is amplification at each step so that there are many more proteins present than reporter genes. Sometimes it is necessary to also place the gene for the regulatory protein on the plasmid as well as the reporter gene, as the native levels of reporter protein within the bacteria are insufficient for proper regulation of transcription.

fluorophores can be used; this ratiometric technique is more accurate than measuring just one fluorescence signal.

1.3. Bacteria-based sensing systems Amplification-based methods take advantage of the high turnover of substrates to produce a large number of product molecules. This is the basis of such techniques as PCR and RT-PCR. In these cases, DNA or RNA molecules are selectively amplified to quantify the numbers of their parent strands. Whole-cell sensing systems take this one step further by first producing DNA, which is then amplified again during the transcription to RNA, and finally amplified a third time by translation to protein. 312

Genetic Engineering of Signalling Molecules Table 1. Reporter proteins used in whole cell sensing systems. Detection method*

Protein

Gene

Catalyzed reaction

chloramphenicol acetyltransferase

Cat

acetylation of chloramphenicol

RI, FL

[3-galactosidase

LacZ

hydrolysis of 13-galactosides

CR, FL, EC, CL

firefly luciferase

Luc

luciferin + O2 + ATP oxyluciferin + AMP + PPi + h v

BL

bacterial luciferase

LuxAB

aequorin

AQ440

green fluorescent protein

GFP

FMNI-I2+ R-CHO+ 02 -~ FMN + HE0 + RCOOH + h v coelenterazine + 02 + Ca2+--~ coelenteramide + CO2 + h t, posttranslational formation of an internal chromophore

BL BL FL

* RI, radioisotope; FL, fluorescence; CR, colorimetric; EC, electrochemical; CL, chemiluminescence; BL, bioluminescence.

A typical whole-cell sensing system consists of an organism, generally a bacterium, that is transformed with a plasmid containing a reporter gene under the control of a promoter responsive to the analyte of interest (Daunert et al., 2000; Lewis et al., 1998; Ramanathan et al., 1997a). This plasmid may also contain genes that will produce the necessary accessory proteins for the promoter, such as the regulatory proteins. These additional genes are sometimes necessary, as the number of promoters on the plasmid may greatly outnumber the usual number of promoters; a larger number of regulatory proteins is necessary to regulate these plasmid-borne promoters. Once the cells are exposed to analyte, transcription of the reporter gene will begin (Figure 4). After transcription, the RNA molecules are translated into protein. Amplification occurs at each of these steps to produce many more protein molecules than there are reporter genes. If desired, an extra level of amplification can be achieved if the reporter protein is an enzyme that will turn over large numbers of substrate molecules. However, if this further amplification is not required, then a protein such as the green fluorescent protein (GFP) can be used. GFP does not require addition of an external substrate, as the protein itself emits green fluorescence upon excitation at 490 nm. Another way to obviate adding a substrate is to use the entire lux cassette, instead of just luxAB, to produce bacterial luciferase. In this way all the accessory proteins to produce the substrates necessary for bacterial luciferase activity are also transcribed (Manen et al., 1997).

313

Feltus and Daunert The sensitivity of these systems is determined by a number of factors. The response of the promoter must be taken into account, but the largest effect of the promoter/repressor protein is upon the selectivity of the system. The more controllable factor is the choice of reporter protein, since there is often only a limited choice of promoters for a given analyte. The ideal reporter protein will be easy to use, have an easily discernable signal over the background, and have a wide dynamic range and high sensitivity (Daunert et al., 2000). Examples of reporter proteins that have been used to develop whole-cell sensing systems are given in Table 1. Sensitivity can be a function of several factors, including the detection method, efficiency of expression, reporter protein turnover number (if the protein is an enzyme), and, if applicable, the endogeneous levels of the reporter protein. For this reason, bioluminescent reporter proteins are a popular choice because bioluminescence is not found in most organisms, and is a very sensitive method of detection.

2. History These three types of fluorescent signaling systems emerged from the need of researchers in the biological sciences to study protein response to the binding of various ligands. For example, bacteria-based sensing systems are the result of experiments on regulation of transcription at various promoters.

2.1. Fluorophore-labeled proteins and FRET-based systems These two systems share a common ancestry in studies of protein function. One way of examining the structural changes in proteins upon ligand binding, dimerization, or denaturation is by measuring in the native fluorescence of tryptophan residues. This approach since they might not be close can be used to measure binding only when the tryptophan is proximal to the active site. This limitation led to the use of fluorescent cofactors and substrates, such as flavin mononucleotide, to study changes occurring within the binding pocket. Later, proteins were labeled with extrinsic fluorophores. Such labeled proteins have been used for a number of applications, including microinjection into cells to study protein localization and solution studies of protein structural changes. Initially, biochemists used these fluorophore-labeled proteins to gain information about the alterations in size, shape, and binding properties of proteins. However, with the development of environmentally sensitive fluorophores and the ability to produce mutated recombinant proteins, the fluorophore-labeled sensing system as it stands today was born. Table 2 gives several examples of analytes that have been measured using these systems. Most of the currently-developed sensing systems of this type depend upon molecular biology to either create a unique site for fluorophore attachment, to translate the protein such that it incorporates nonnative fluorescent amino acids, or to fuse a GFP to the protein. 314

Genetic Engineering of Signalling Molecules Table 2. Examples of fluorophore-labeled protein sensing systems. In vitro/in vivo refers to whether the protein is used in situ after being produced by the cells or whether the proteins are expressed, isolated, and purified prior to use, and used as a sensing system. . . . Analyte

.

.

. . . . Protein* .

.

.

In vitro~in vivo .

Reference

.

P~

PBP

In vitro

Brune et al., 1994, 1998

fatty acids

I-FABP

In vitro

Richieri et al., 1992

maltose

MBP

In vitro

Gilardi et al., 1994

biotin

Streptavidin

In vitro

Murakami et al., 2000

Ca2+

CaM-YFP fusion

In vitro~in vivo

Baird et al., 1999

Ca 2+

CaM-EGFPM 13 fusion

in vitro~in vivo

Nakai et al., 2001

Ca z+

CaM

in vitro

Salins et al., 1998; Schauer-Vukasinovic et al., 1997

Co 2+, Zn 2+, Cu 1+

Carbonic anhydrase

in vitro

Thompson et al., 1998

glucose

GGBP

in vitro

Salins et al., 2001; Tolosa et al., 1999

Dattelbaum and Lakowicz, 2001 *Abbreviations: PBP, phosphate binding protein; I-FABP, intestinal fatty acid binding protein; MBP, maltose binding protein; CaM, calmodulin; YFP, yellow fluorescent protein; EGFP, enhanced green fluorescent protein; GGBP, galactose/glucose-binding protein; GlnBP, glutamine binding protein glutamine

GlnBP

in vitro

FRET-based systems can be considered as a subclass of the fluorophore-labeled proteins, different only because they depend upon the proteins being labeled with two fluorophores rather than one. Because FRET is a distance-dependent phenomenon, it was originally used to study assembly of multi-subunit protein complexes, such as ribosomes, or interaction between a protein and cellular membranes. In the 1990s, however, FRET-based systems started to be used for analytical purposes (Table 3). The most recent trend is to use GFP and its wavelength-shifted mutants as the donor and acceptor molecules. 315

Feltus and Daunert Table 3. FRET-based assays using labeled proteins.* Analyte

Protein

I)'onor

Acceptor

Reference

factor Xa

factor Xa site "

BFP

rsGfp

Mitra et ai.',' 1996

caspase-3

Caspase-3 site

BFP

GFP

Xu et., 1998

caspase-3

Caspase-3 site

CFP

YFP

Jones et al., 2000

Zn 2+

Lissamine

Ca 2+

zinc finger peptide aequorin

Aequorin

rhodamin e GFP

Godwin and Ber, 1996 Baubet et al., 2000

Ca 2+

CaM

BFP

GFP

Romoser et al., 1997

Ca 2+

Cam/M13

BFP/CFP

GFP/YFP

PKA

KID

B FP

GFP

Miyawaki et al., 1997 Nagai et al., 2000

CAMP

PKA

CAMP

PKA

fluorescei n BFP/CFP

rhodamin e GFP/YFP

Adams et al., 1991 Zaccolo et al., 2000

*Abbreviations: BFP, blue fluorescent protein; GFP, green fluorescent protein; CFP, cyan fluorescent protein; YFP, yellow fluorescent protein; CaM, calmodulin; PKA, protein kinase A; KID, kinase inducible domain.

Because these fluorophores are proteins themselves, plasmid constructs can be made that fuse the G F P to the sensing protein, allowing these proteins to be produced within a cell and used in situ as sensors. This is advantageous, since the analytes of interest are generally intracellular second messengers. Moreover, the need to microinject purified chemically-labeled proteins is avoided.

2.2. Bacteria-based sensing systems Sensing systems of this type trace their origin back to bioassays in which nutrient-deficient or antibiotic-resistant strains were plated on media containing various concentrations of the analyte and the surviving number of cells counted. From this methodology evolved the non-specific bacteria-based sensing systems. These bacteria constitutively express a reporter protein while alive, but as toxins begin tokill the bacteria, the protein is no longer produced, giving a lower signal. At the same time, the bacterial operons were discovered, and reporter genes were

316

Genetic Engineering of Signalling Molecules Table 4. Promoters used to develop whole-cell sensing systems i

Analyte/target response

Promoter

ii

References

antimonite/arsenite

Ars

Corbisier et al., 1993; Ramanathan et al.i 1997b, 1998; Scott et al., 1997; Tauriainen et al., 1997

copper

cupl

Corbisier et al., 1999; Shetty et al., 2000

cadmium

Cad

Tauriainen et al., 1998

lead

Pbr

Corbisier et al., 1999

mercury

Mer

Virta et al., 1995

linear alkanes

PatkB

Sticher et al., 1997

toluene

Pu

de Lorenzo et al., 1993; Ikariyama et al., 1997; Willardson et al., 1998

isopropylbenzene

ipbo/p

Selifonova and Eaton, 1996

naphthalene/salicylate

Pnah

Heitzer et al., 1994; King et al., 1990

chlorocatechols

Pc~c

Guan et al., 2000

L-arabinose

PBAD

Shetty et al., 1999 ~

[3-lactose

Plac

Daunert et al., 2000; Shrestha et al., 2000

DNA damage

recA, uvrA, umuC

Billinton et al., 1998; Ptitsyn et al., 1997; Rettberg et al., 1997, 1999; van der Lelie et al., 1997

placed under their control in order to study gene expression. The first of the bacteria-based sensing systems for specific analytes used metal- or toxinresistance promoters, presumably because these operons are generally carried on plasmids rather than within the bacterial genome and because expression is tightly regulated and induction occurs in response to the presence of the toxin or metal. Since then, other analytes have been targeted, including sugars (Table 4). Because fluorescent report proteins have been shown to work well, the newest trend is to use fluorescent reporter proteins, rather than bioluminescent ones, 317

Feltus and Daunert because fluorescent proteins require no addition of substrates. There is also a need to develop sensing strains that can respond to more than one analyte. This has been accomplished by using two separate promoters and reporter proteins.

3. State of the Art In the previous section we discussed the history of how the fluorescent signaling systems emerged. In this section, we Will focus on the more advanced forms of these systems, and describe particular sensing systems in detail to give the reader an idea of the full scope of the system, some considerations of how the system is designed, and the part that molecular biology plays in it.

3.1. Fluorophore-labeled binding proteins This particular mechanism has been exploited to develop a number of fluorescence assays for a variety of analytes (Table 2). Molecular biology is often employed in order to control the site of fluorophore attachment to give the greatest change upon ligand binding. In several cases, genetic engineering was employed to introduce a unique cysteine in the binding protein. This residue was then specifically labeled with a thiol-reactive environmentally-sensitive fluorescent probe (Brune et al., 1998; Gilardi et al., 1994; Salins et al., 1998; Schauer-Vukasinovic et al., 1997; Thompson et al., 1998). In the case of IFABP, this was unnecessary as acrylodan reacted only at lysine 27 and produced large changes in fluorescence upon addition of free fatty acids such as oleate, palmitate, and arachidonate (Richieri et al., 1992). Murakami et al. (2000) used a different approach, introducing a unique L-2-anthrylalanine into the amino acid sequence of streptavidin by an in vitro transcription method, thus creating a sensing system for biotin. The optimal site of fluorophore location can be determined through examination of the crystal structure of the protein or through NMR studies of the free and bound forms of the protein. If these have not been determined, then an educated guess can be made through other studies, such as mutagenesis to determine the location of the analyte binding site or through examination of a closely related protein. This protein engineering strategy allows for the specific attachment of a fluorophore to the protein at a site which undergoes a large change in its environment upon ligand binding. In some cases, it may be necessary to try several sites and fluorophores before the optimal response of the system can be established. The development of the sensing system for C a 2§ based on calmodulin (CAM) by Schauer-Vukashinovic et al. (1997) is a good example. The protein calmodulin binds four calcium ions located in pairs in each of two domains: an N-terminal domain and a C-terminal domain that are linked by a long helix (Babu et al., 318

Genetic Engineering of Signalling Molecules 1988; Kuboniwa et al., 1995). In the absence of calcium, the structure is more disordered. When calcium binds, two hydrophobic pockets open, one in each domain, for binding to proteins such as myosin light chain kinase (MLCK) or to drugs such as trifluoropiperazine and phenothiazine. Figure 5 shows a close-up view of the C-terminal binding pocket in both the presence and absence of calcium (Finn et al., 1993). Several mutants of CaM were produced with unique cysteine residues at positions 38, 81,109, and 113 (Schauer-Vukashinovic et al., 1997). The last three are seen in Figure 5 near the pocket; residue 38 is in the Nterminal domain. Several combinations of thiol-reactive fluorophores and labeling sites were examined. The best results were obtained with a CaM109MDCC conjugate (96% increase in fluorescence upon Ca 2+ binding). When the other sites were labeled with MDCC, the amount of increase was only 15%, 16%, and 28%, respectively. As seen in Figure 5, residues 81, 109, and 113 are located quite close to each other in the structure of CaM, but the three residues apparently have very different environmental changes upon Ca 2+ binding. There are also differences when the fluorophore at position 109 is exchanged for another. If, instead of MDCC, the related fluorophore CPM (Figure 2) is used, the amount of change is only 25 %. If fluorescein is used, then there is no change in signal upon Ca 2+ binding. The limit for detection of calcium using CaM109-MDCC is 2 x 10 -9 M C a 2+. A random labeling at lysine residues (of which CaM has 9) with fluorescein isothiocyanate shows a lower amount of change in fluorescence (23% increase upon Ca 2+ binding) and a higher detection limit (5 x 10.8 M) (Blair et al., 1994). The reason is that the change in fluorescence is highly dependent upon the location of the fluorophore within the protein. Nonspecific labeling with multiple fluorophores increases the background signal, giving a smaller relative increase in fluorescence upon calcium binding. This reduces the ability to detect lower levels of Ca 2§ Similar effects have been seen in the systems for maltose (Gilardi et al., 1994) and phosphate (Brune et al., 1994), indicating that the best detection limits are obtained when a unique fluorophore is properly positioned. An alternate strategy has recently been employed using GFP instead of a small organic fluorophores. These studies use circular permutations of GFP (cpGFP); the C-terminus is fused to the N-terminus. Baird et al. (1999) inserted CaM into a circular YFP at amino acid 145. The fluorescence of the YFP was retained while giving rise to a Ca2§ fusion protein with a detection limit of approximately 2 x 10 "6 M C a 2+. Nakai et al. (2001) used a similar construct in which GFP was circularized to produce a new N-terminus at residue 149 and Cterminus at residue 144. The calmodulin-binding peptide M13 was attached to the new N-terminus and CaM to the new C-terminus. The resulting protein showed an increase in fluorescence of up to 4.5-fold upon addition of Ca 2+ because the CaM moiety bound the M13 moiety, altering the conformation of GFP. The detection limit for this system was one order of magnitude better 319

Feltus and Daunert

Figure 5. Alterations in the C-terminal hydrophobic pocket of calmodulin upon calcium binding. Residue 109 is closer to the pocket than either residue 81 or 113. Labeling mutant calmodulins gives the most change with an MDCC-CaM109 conjugate. Labeling at 81 or 113 does not give as much fluorescence change upon calcium binding, presumably because the two residues are further from the hydrophobic pocket than amino acid 109. Adapted from Schauer-Vukasinovic et al. (1997).

(1 x 10"7 M) than the system developed by Baird et al., although the Nakai system has a narrower dynamic range. Both systems were also shown to be useful in detecting calcium fluxes in vivo (Baird et al., 1999; Nakai et al., 2001).

3.2. FRET-based systems FRET systems have been used to detect analytes and biological functions as varied as protease activity, ions, cyclic AMP, myosin II phosphorylation, and insulin-receptor signaling. As seen in Table 3, these assays can either be performed in vitro or in vivo by microinjection or transfection with genes to transcribe the sensing systems in situ. Molecular biology is used to create the GFP or other fusion proteins necessary for each sensor. For biosensing purposes, these labeled proteins can first be produced in vivo, then purified and immobilized at the tip of a fiber optic probe. It is also possible that the cells themselves could be immobilized, obviating the purification step. One of the most important contributions of molecular biology to these systems has been the creation of GFP mutants that can act as FRET pairs. Native Aequorea GFP absorbs blue light and emits green light. Its usual FRET donor is the blue fluorescent protein (BFP), a variant of GFP mutated in several residues 320

Genetic Engineering of Signalling Molecules

Figure 6. FRET-based sensing system for Ca2§ based on a BFP/GFP pair bridged with a MLCK CaM binding site. The FRET donor BFP is separated from the acceptor GFP by a CaM recognition sequence from myosin light chain kinase (MLCK). CaM can only bind this sequence in the presence of Ca2§ increasing the distance between the fluorophores and decreasing the amount of FRET. The system, therefore, responds to the amount of Ca2§ present. Adapted from Miyawaki et al. (1997).

in and around the chromophore of the protein; these changes shift the excitation to UV wavelengths and the emission to blue. This provides a spectral overlap with GFP, allowing FRET. The other FRET pair used, cyan fluorescent protein (CFP, donor) and yellow fluorescent protein (YFP, acceptor), is composed of two mutant GFPs created in the same way. In this pair, CFP absorbs in the blue region and emits in the blue-green region, overlapping with YFP's absorption spectrum. YFP then emits in the yellow. A discussion of the many different mutants of GFP can be found in a review paper by Tsien (1998). The assays of protease activity do not depend upon a binding event, but rather the physical separation of tethered fluorophores. In these systems, EGFP and BFP are connected by a short peptide sequence containing a cleavage site for the protease of interest. Before the protease acts at its site, EGFP and BFP are kept at a fixed distance from each other; cleaving the bond within the cut site causes the fluorophores to drift apart, thus disrupting the FRET. For trypsin, the amount of change in the fluorescence emission ratio was 4.6-fold and for factor Xa it was 3-fold (Mitra et al., 1996). Since EGFP, the linker, and BFP are all genetically encoded, an in vivo assay can be developed by transfecting cells with DNA to produce the sensor inside the cells. This was demonstrated by Xu et al. (1998) in their system for caspase-3. Activation of caspase-3 destroyed FRET between the GFPs. Such assays are of particular value in the high-throughput screening of apoptosis-inducing drugs, since caspase-3 is activated during apoptosis. Indeed, recently Jones et al. (2000) reported that this system could reliably identify apoptosis-inducing drugs, such as staurosporine, camptothecin, and etoposide. Cell signaling events, such as Zn 2§ or Ca 2§ release and cAMP accumulation, have also been found to be good targets for FRET-based systems. An in vitro system 321

Feltus and Daunert for Zn 2§ developed by Godwin and Berg (1996) uses a zinc finger peptide as the sensing element. Zinc fingers bind zinc tightly and have a great selectivity for Zn(II) over Co(II), Fe(II), and Ni(II). Godwin and Berg (1996) engineered a zinc finger with a lissamine donor at the N-terminus and a fluorescein acceptor at the C-terminus. Binding of Zn 2§ to the peptide brings together the two fluorophores, resulting in FRET. This system has the ability to detect Zn 2+ at levels of 5 x 10-7 M (Godwin and Berg, 1996). Ramoser et al. (1997) developed a sensing system for Ca ~+ by connecting two GFP variants, BFP and RGFP, with a peptide linker containing the calmodulin binding sequence from myosin light chain kinase. Binding of (Ca2+)4-CaM to the sensor increases the inter-fluorophore distance from ~25/~ to ~65A, effectively eliminating FRET (Figure 6). The change in the fluorescence emission ratio is dose-dependent for both Ca 2§ and (Ca2+)4-CaM and is shown to work well when microinjected into cells, as well as in vitro. A similar system was developed by Miyawaki et al. (1997) using BFP or CFP, CaM, CaM-binding peptide M13, and GFP. Binding of Ca 2+ causes the CaM moiety to wrap around the M13 peptide, decreasing the distance between the pairs and aligning them properly for FRET. This results in a 70% increase in the fluorescence emission ratio and a wide detection range of three orders of magnitude from 10 -7 tO 10 -4 M , with a detection limit of 2.5 x 10-8 M Ca 2§ By transfecting the DNA for this sensor into mammalian cells, it was found that a different pair of GFPs (CFP and YFP) worked better by improving the brightness (CFP fluoresces more intensely than BFP) and signal-to-noise ratio. However, the overall amount of change was only 1.5-fold for the CFP/BFP pair versus 1.8fold for the BFP/GFP pair, due to bleedthrough of CFP emission into the YFP spectrum. A bridged GFP chimera has also been developed for sensing of cAMP-related effects in cells. Nagai et al. (2000) created a sensor based on a bridge of kinaseinducible domain of CREB (cAMP response element binding protein). This domain is phosphorylated by cAMP-dependent protein kinase A (PKA), which results in a conformational change. BFP and RGFP were again used as the donor and acceptor, located at the two ends of the bridge. In vitro experiments with this system showed that the emission ratio increased from 0.68 to 0.83 when incubated with PKA and ATP. Transfection of the chimera into COS-7 cells showed an increase in fluorescence upon PKA activation while administration of PKA inhibitor H-89 significantly inhibited FRET within the cells. Cyclic AMP itself has been monitored in vivo using fluorescently-tagged PKA. PKA consists of regulatory and catalytic subunits that dissociate upon cAMP binding. This disrupts FRET between the donor on the regulatory subunit and the acceptor on the catalytic subunit (Figure 2). The original sensor developed by Adams et al. (1991) used a fluorescein/rhodamine pair. This sensor worked very well, but required expression and purification of the protein subunits, in vitro labeling, purification, and microinjection. Zaccolo et al. (2000) developed 322

Genetic Engineering of Signalling Molecules a completely in vivo system using BFP and GFP (Figure 2). In a population of transfected COS-7 cells treated with 10 #M isoproterenol, the emission ratio increased from 1.7 to 2.0, and was completely reversed by incubation with 10 #M propranolol. This reaction is almost instantaneous upon introduction of isoproterenol. These systems thus offer an extremely fast response to their analytes.

3.3. Bacteria-based sensing systems Bacteria-based sensing systems have been developed for a variety of analytes. As shown in Table 4, there are a number of promoters that have been used for either the specific sensing of a particular analyte, a family of compounds, or a stress response, such as starvation. Many are used to detect toxic substances, such as heavy metals, carcinogens, or organic pollutants. For example, sensing systems have been developed for arsenic/antimony (Corbisier et al., 1993; Ramanathan et al., 1997b, 1998; Scott et al., 1997; Tauriainen et al., 1997), copper (Corbisier et al., 1999; Shetty et al., 2000), cadmium/lead (Corbisier et al., 1999; Tauriainen et al., 1998), chromium (Peitzsch et al., 1998), aluminum (Guzzo et al., 1992), and mercury (Virta et al., 1995). The original purpose of these promoters is to produce proteins that either sequester the metal ions, transport them outside the cell, or enzymatically detoxify them (Brown et al., 1998; Nies, 1999). Experience has shown that these systems are extremely sensitive to very small amounts of the metal being present. In fact, in one case, by coupling the ars promoter with the gene coding for bacterial luciferase, Ramanathan et al. (1997b) found that a detection limit of 10~5 M arsenite could be obtained with high selectivity for antimonite and arsenite over other metals such as bismuth, cadmium, and cobalt. High selectivity was also seen in a sensing system for L-arabinose developed using the PBADpromoter and the gene for GFP developed by Shetty et al. (1999). In cases of low glucose levels, E. coli can use other sugars as an energy source. This system could detect 1 x 10 "7 M L-arabinose while it did not respond to other pentose sugars or their corresponding D-isomers. These bacteria were also immobilized at the tip of a fiber optic. A small sleeve was placed over the tip of the fiber optic, creating a small space in which the bacterial suspension was kept. The opening was covered with a dialysis membrane to prevent the bacteria from diffusing out of the sensing range of the fiber optic, but still allow the analyte to pass through and interact with the bacteria. The sensor had a detection limit one order of magnitude less sensitive than the solution-based system; this decrease in sensitivity was attributed to changes in the instrumental setup (i.e., a lowerpowered light source, decreased coupling efficiency, a less sensitive PMT, and an increased diffusion time). Another study using this system showed the probable direction that these systems will take in the future. A dual-detection system for L-arabinose and 13-lactose was developed by combining the arabinose system described above with a similar system to detect lactose (Daunert et al., 2000; 323

Feltus and Daunert Shrestha et al., 2000). The lactose system employed the gene for BFP, which emits in the blue region. Thus, two analytes could be measured at the same time by simultaneously monitoring the fluorescence emission at two different wavelengths. In order to survive in heavily polluted environments, certain organisms have also developed the capability of using organic pollutants as carbon sources. Promoters from operons metabolizing these environmental pollutants have been used to develop biosensing systems for the monitoring of the bioavailable amounts of chemicals such as alkanes (Sticher et al., 1997), benzene derivatives (de Lorenzo et al., 1993; Ikariyama et al., 1997; Selifonova and Eaton, 1996; Willardson et al., 1998), chlorocatechols (Guan et al., 2000), and PCBs (Layton et al., 1998). Likewise, sensing systems for carcinogens, such as the SOS-lux system, are capable of monitoring genotoxins by responding to actual DNA damage of the cda promoter by the environmental toxin by producing bacterial luciferase in a dose-dependent manner (Ptitsyn et al., 1997; Rettberg et al., 1997). It is important to note that this system responds not to the concentration of the carcinogen within the cell, but its activity. The SOS-/ux system also has advantages over the Ames test in that results are available within 1-2 h and kinetic effects of the toxin can be studied.

4. Advantages and Limitations Of the systems described in this chapter, the two with the fastest response times are the binding protein-based systems. Compared to FRET-based probes, fluorophore-labeled binding proteins usually have greater fluorescence changes, presumably because they depend upon a direct action upon the fluorophore rather than on the more indirect method of energy transfer. Also, because these systems usually are based on using single-chain proteins, they are capable of being covalently immobilized on a solid surface. This is more difficult in FRET-based systems because the two component proteins must be free to interact with or dissociate from one another. Despite these advantages, fluorophore-labeled binding proteins are more difficult to optimize, as several different immobilization sites and fluorophores must be tested. This represents a substantial amount of molecular biology, and can take quite a long time. The main reason for this is the difficulty in predicting the environmental change at a specific location on the protein. Although clues can be obtained from crystallographic and NMR structures, even small changes in location make a very large difference to the sensitivity of the system. It is much easier to predict whether the distance between two fluorophores will change upon the binding of the analyte.

324

Genetic Engineering of Signalling Molecules One of the main advantages of FRET-based sensing systems is that they employ a very small number of reagents. In fact, in some cases they use no additional reagents, as both the fluorophores (GFP, YFP, etc.) and the binding proteins can be genetically encoded and produced within the cells. Like the whole-cell based systems, they may be extremely cost-effective, since the transformed cells can be frozen and a new batch of sensitive cells regrown at any time. FRET as a detection methodology has additional advantages that make it attractive for use in biological systems. Because the system uses the excitation wavelength of the acceptor and the emission wavelength of the donor, the Stokes shift is more pronounced than for fluorescence, resulting in a lower background. Another advantage of FRET is that the ratio of fluorescence intensities can be used; this technique is more accurate than measuring a single fluorescence intensity. Ratiometric methods are also independent of path-length, accessible volume, and local concentration, points that become more important as we consider decreasing the assay volume (Giuliano and Taylor, 1998). Having said this, it is not always possible to develop a FPdET-based system due to the necessity of having some sort of change in distance occur between the fluorophores. Also, because of the association/dissociation of the protein components, FRET may, in some cases, be more susceptible to matrix effects if some component of the sample causes premature dissociation. The least susceptible system to matrix effects is probably the whole-cell based sensing system. In order for transcription activation to occur, the analyte must be taken up by the bacteria and then interact at the promoter to induce expression of the reporter gene. Not only is it unlikely that an interferent will mimic these steps, but the bacterial cell wall gives the bacteria a high tolerance to pH changes and to other environmental extremes. A high level of selectivity is also found in these systems for the same reasons; the interfering species must not only be able to enter the cell, but it must cause the proper conformational changes in the promoter's regulatory protein to initiate transcription. Another advantage is the improved sensitivity of the system due to the number of amplification steps. The amount of molecular biology required to develop a whole-cell sensing system is less than for either of the protein-based systems described above, as there is no need to mutate the reporter protein. In addition, the system is continuously renewable. If a new batch is needed, it is simply regrown; no purification is necessary. The main disadvantage of these systems is the long response times. Because the bacteria must be alive and growing, it may be necessary to incubate the ceils at 37~ for several hours in order to take up the analyte and produce a properly-folded reporter protein. Protein-based systems, on the other hand, bypass this step and require only minutes of incubation. Notwithstanding the more extended time requirement, the bacteria-based systems are limited only by the ability to find a promoter that responds to an analyte of interest. With proper choice of reporter gene and detection method, highly sensitive and selective systems can be developed to study not only the 325

Feltus and Daunert concentrations of various analytes, but, more importantly, to study their actual activities.

5. Potential for Expanding Current Capabilities The range of analytes that can be measured using the systems described in this chapter is limited only by the availability of recognition elements. For example, ' the selectivity of bacteria-based systems is controlled by the selectivity of the binding proteins regulating the promoter's activity. Not only is it possible that new promoters will be discovered, but that by mutation and selection of bacterial strains, new promoters can be created, as they have been naturally over the course of time by new stresses placed on microorganisms living in harsh or nutrient-depleted environments. New binding proteins as the basis of FRET- or fluorophore-labeled systems can be created through random synthesis or DNA shuffling. These can serve to either create binding proteins for new analytes or to increase the selectivity or binding affinity of known binding proteins. FRET-based systems depending upon two GFPs as the donor and acceptor molecules may also be improved through the creation of new GFP variants. In the past, there has been a focus on creating GFPs that are brighter, have higher quantum yields, are more stable, and have different absorption and emission wavelengths than the wild-type protein (Tsien, 1998). This approach also has the potential to expand whole-cell sensing systems into the multi-analyte area. Another area in which these systems can find use is in small-volume analyses. Many high-throughout screening (HTS) applications are beginning to take advantage of advances in microfluidics and microfabrication to shrink the size of assays. Since FRET-based systems are already performed in vivo and observed in single cells, they are already proven to be applicable to small volumes. Fluorophore-labeled binding proteins could also be of use, not only in small volumes and microfluidic platforms, where decreasing the number of aliquots will decrease the error, but also in single cells, where injection of a very limited number of assay components is necessary to prevent the cell from bursting. In time, ways will be found that obviate these microinjections, so that the sensing protein will be transcribed in situ within the cells, as FRET-based systems are today. This will further expand their use in biological systems.

6. References Adams, S.R., A.T. Harootunian, Y.J. Buechler, S.S. Taylor and R.Y. Tsien, 1991, Nature 349, 694. Babu, Y.S., C.E. Bugg and W.J. Cook, 1988, J. Mol. Biol. 204, 191.

326

Genetic Engineering of Signalling Molecules Baird, G.S., D.A. Zacharias and R.Y. Tsien, 1999, Proc. Natl. Acad. Sci. U. S. A. 96, 11241. Baubet, V., H. Le Mouellic, A.K. Campbell, E. Lucas-Meunier, P. Fossier and P. Brulet, 2000, Proc. Natl. Acad. Sci. U.S.A. 97, 7260. Billinton, N., M.G. Barker, C.E. Michel, A.W. Knight, W.D. Heyer, N.J. Goddard, P.R. Fielden and R.M. Walmsley, 1998, Biosens. Bioelectron. 13, 831. Blair, T.L., S.-T. Yang, T. Smith-Palmer and L.G. Bachas, 1994, Anal. Chem. 66, 300. Brown, N.L., J.R. Lloyd, K. Jakeman, J.L. Hobman, I. Bontidean, B. Mattiasson and E. Csoregi, 1998, B iochem. Soc. Trans. 26, 662. Brune, M., J.L. Hunter, J.E. Corrie and M.R. Webb, 1994, Biochemistry 33, 8262. Brune, M., J.L. Hunter, S.A. Howell, S.R. Martin, T.L. Hazlett, J.E. Corrie and M.R. Webb, 1998, Biochemistry 37, 10370. Corbisier, P., G. Ji, G. Nuyts, M. Mergeay and S. Silver, 1993, FEMS Microbiol. Lett. 110, 231. Corbisier, P., D. van der Lelie, B. Borremans, A. Provoost, V. de Lorenzo, N.L. Brown, J.R. Lloyd, J.L. Hobman, E. Csoregi, G. Johansson and B. Mattiasson, 1999, Anal. Chim. Acta 387, 235. Dattelbaum, J.D. and J.R. Lakowicz, 2001, Anal. Biochem. 291, 89. Daunert, S., G. Barrett, J.S. Feliciano, R.S. Shetty, S. Shrestha and W. SmithSpencer, 2000, Chem. Rev. 100, 2705. de Lorenzo, V., S. Fernandez, M. Herrero, U. Jakubzik and K.N. Timmis, 1993, Gene 130, 41. Finn, B.E., T. Drakenberg and S. Forsen, 1993, FEBS Lett. 336, 368. Gilardi, G., L.Q. Zhou, L. Hibbert and A.E. Cass, 1994, Anal. Chem. 66, 3840. Giuliano, K.A. and P.L. Post, 1995, Annu. Rev. Biophys. Biomol. Struct. 24, 405. Giuliano, K.A. and D.L. Taylor, 1998, Trends Biotechnol. 16, 135. Godwin, H.A. and J.M. Berg, 1996, J. Amer. Chem. Soc. 118, 6514. Guan, X., S. Ramanathan, J.P. Garris, R.S. Shetty, C.M. Ensor, L.G. Bachas and S. Daunert, 2000, Anal. Chem. 2423. Guzzo, J., A. Guzzo and M.S. DuBow, 1992, Toxicol. Lett. 64-65 Spec No, 687. Heitzer, A., K. Malachowsky, J.E. Thonnard, P.R. Bienkowski, D.C. White and G.S. Sayler, 1994, Appl. Environ. Microbiol. 60, 1487. Ikariyama, Y., S. Nishiguchi, T. Koyama, E. Kobatake, M. Aizawa, M. Tsuda and T. Nakazawa, 1997, Anal. Chem. 69, 2600. Jones, J., R. Heim, E. Hare, J. Stack and B.A. Pollok, 2000, J. Biomol. Screen. 5, 307. King, J.M.H., P.M. Digrazia, B. Applegate, R. Burlage, J. Sanseverino, P. Dunbar, F. Larimer and G.S. Sayler, 1990, Science 249, 778. Kuboniwa, H., N. Tjandra, S. Grzesiek, H. Ren, C.B. Klee and A. Bax, 1995, Nature Struct. Biol. 2, 768.

327

Feltus and Daunert Lakowicz, J.R., 1983, Principles of Fluorescence Spectroscopy: Energy Transfer, Eds. Academic Press, New York. Layton, A.C., M. Muccini, M.M. Ghosh and G.S. Sayler, 1998, Appl. Environ. Microbiol. 64, 5023. Lewis, J.C., A. Feltus, C.M. Ensor, S. Ramanathan and S. Daunert, 1998, Anal. Chem. 70, 579A. Manen, D., M. Pougeon, P. Damay and J. Geiselmann, 1997, Gene 186, 197. Mitra, R.D., C.M. Silva and D.C. Youvan, 1996, Gene 173, 13. Miyawaki, A., J. Llopis, R. Heim, J.M. McCaffery, J.A. Adams, M. Ikura and R.Y. Tsien, 1997, Nature 388, 882. Murakami, H., T. Hohsaka, Y. Ashizuka, K. Hashimoto and M. Sisido, 2000, Biomacromolecules 1, 118. Nagai, Y., M. Miyazaki, R. Aoki, T. Zama, S. Inouye, K. Hirose, M. fino and M. Hagiwara, 2000, Nature Biotechnol. 18, 313. Nakai, J., M. Ohkura and K. Imoto, 2001, Nature Biotechnol. 19, 137. Nies, D.H., 1999, Appl. Microbiol. Biotechnol. 51,730. Peitzsch, N., G. Eberz and D.H. Nies, 1998, Appl. Environ. Microbiol. 64, 453. Ptitsyn, L.R., G. Horneck, O. Komova, S. Kozubek, E.A. Krasavin, M. Bonev and P. Rettberg, 1997, Appl. Environ. Microbiol. 63, 4377. Ramanathan, S., M. Ensor and S. Daunert, 1997a, Trends Biotechnol. 15,500. Ramanathan, S., W. Shi, B.P. Rosen and S. Daunert, 1997b, Anal. Chem. 69, 3380. Ramanathan, S., W. Shi, B.P. Rosen and S. Daunert, 1998, Anal. Chim. Acta 369, 189. Rettberg, P., L.R. Ptitsyn, O. Komova, S. Kozubek, E. Krasavin, M. Bonev and G. Homeck, 1997, Mut. Res. 379, $206. Rettberg, P., C. Baumstark-Khan, K. Bandel, L.R. Ptitsyn and G. Horneck, 1999, Anal. Chim. Acta 387, 289. Richieri, G.V., R.T. Ogata and A.M. Kleinfeld, 1992, J. Biol. Chem. 267, 23495. Romoser, V.A., P.M. Hinkle and A. Persechini, 1997, J. Biol. Chem. 272, 13270. Salins, L.L.E., V. Schauer-Vukasinovic and S. Daunert, 1998, Proc. SPIE 3270, 16. Salins, L.L.E., R.A. Ware, C.M. Ensor and S. Daunert, 2001, Anal. Biochem. 294, 19. Schauer-Vukasinovic, V., L.C. Cullen and S. Daunert, 1997, J. Am. Chem. Soc. 119, 11102. Scott, D.L., S. Ramanathan, W. Shi, B.P. Rosen and S. Daunert, 1997, Anal. Chem. 69, 16. Selifonova, O.V. and R.W. Eaton, 1996, Appl. Envir. Microbiol. 62, 778. Shetty, R.S., Y. Liu, S. Ramanathan and S. Daunert, 2000, The Pittsburgh Conference on Analytical Chemistry Applied Spectrometry, New Orleans, LA. Shetty, R.S., S. Ramanathan, I.H. Badr, J.L. Wolford and S. Daunert, 1999, Anal. Chem. 71,763.

328

Genetic Engineering of Signalling Molecules Shrestha, S., R.S. Shetty, S. Ramanathan and S. Daunert, 2000, 219th Meeting of the American Chemical Society, San Francisco, CA, ANYL-047. Sticher, P., M.C. Jaspers, K. Stemmler, H. Harms, A.J. Zehnder and J.R. van der Meer, 1997, Appl. Environ. Microbiol. 63, 4053. Tauriainen, S., M. Karp, W. Chang and M. Virta, 1997, Appl. Environ. Microbiol. 63, 4456. Tauriainen, S., M. Karp, W. Chang and M. Virta, 1998, Biosens. Bioelectron. 13, 931. Thompson, R.B., B.P. Maliwal, V.L. Feliccia, C.A. Fierke and K. McCall, 1998, Anal. Chem. 70, 4717. Tolosa, L., I. Gryczynski, L.R. Eichhom, J.D. Dattelbaum, F.N. Castellano, G. Rao and J.R. Lakowicz, 1999, Anal. B iochem. 267, 114. Tsien, R.Y., 1998, Ann. Rev. Biochem. 67, 509. van der Lelie, D., L. Regniers, B. Borremans, A. Provoost and L. Verschaeve, 1997, Mutat. Res. 389, 279. Virta, M., J. Lampinen and M. Karp, 1995, Anal. Chem. 67, 667. Willardson, B.M., J.F. Wilkins, T.A. Rand, J.M. Schupp, K.K. Hill, P. Keim and P.J. Jackson, 1998, Appl. Environ. Microbiol. 64, 1006. Xu, X., A.L. Gerard, B.C. Huang, D.C. Anderson, D.G. Payan and Y. Luo, 1998, Nucleic Acids Res. 26, 2034. Zaccolo, M., F. De Giorgi, C.Y. Cho, L. Feng, T. Knapp, P.A. Negulescu, S.S. Taylor, R.Y. Tsien and T. Pozzan, 2000, Nature Cell Biol. 2, 25.

329

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 11

ARTIFICIAL RECEPTORS FOR CHEMOSENSORS

THOMAS W. BELL, PH.D. AND NICHOLAS M. HEXT, PH.D. Department of Chemistry, University of Nevada, Reno Reno, NV 89557-0020 USA

Chemosensors are molecules of abiotic origin that signal the presence of matter which can be used to measure the concentrations of analytes in solution. They consist of artificial receptors tailored to reversibly bind the analyte with sufficient affinity and selectivity, a chromophore or fluorophore, and a mechanism for communicating between binding and optical signaling. This chapter details chemosensor design considerations, gives historical background, and provides examples of chemosensors for neutral organic molecules and various anions. Chemosensors for biologically important analytes are particularly emphasized.

I. Technical Concept Sensors for solutes found in low concentration, as is typically the case for samples of biological or environmental origin, generally require binding or concentration of the analyte by the sensor for adequate sensitivity. Our ability to develop sensors for new analytes is often limited by the paucity of materials having adequate affinity, as well as selectivity, when the latter is needed to distinguish the analyte from interfering substances. Enzyme-based biosensors are restricted to the detection of naturally occurring substrates and cofactors. Major advances are being made in adapting biomolecules, such as antibodies and aptamers, for sensor applications, but artificial receptors have many potential advantages. Because they are created by enzymatic chemical reactions, biotic receptors are composed of a limited range of molecular subunits, including amino acids, nucleotides, and sugars. The analyte binding site is generally produced by 331

Bell and Hext

Guest v

Complex

Host

Figure 1. Cartoon showing binding ofan analyte (guest) by a chemosensor (host), producing a complex with altered optical properties, here an increase in fluorescence.

secondary interactions between subunits located along a linear chain. Thus, the critical ability of a biosensor to selectively bind the analyte can be destroyed by variations in ambient conditions, including pH, oxidizing agents, and heat, causing either chemical or thermal degradation, or denaturation. Abiotic receptors can be synthesized from chemically robust components and the binding site can consist of a cavity or cleft enforced by stable, covalent bonds. Their molecular architectures are limited only by the capabilities of synthetic organic chemistry, not by the range of substructures accepted as enzyme substrates. Hence, artificial receptors can be tailored for an unlimited variety of analytes. Their affinities, optical properties, solubilities, and other important characteristics can be adjusted to meet requisite sensor specifications.

1.1. Chemosensor design Chemical sensors are generally understood to be devices that transform chemical information into analytically useful signals (Hulanicki et al., 1991). The term chemosensor has been defined as a molecule of abiotic origin that signals the presence of matter or energy (Czamik, 1993a). Indeed, molecules can be thought of as miniscule devices that can be engineered, fabricated, and used to perform useful functions. Analyte binding can induce mechanical motion (conformational change) in molecules, leading some to term chemosensors operating in this way "molecular machines" (Shinkai et al., 2000; Pina et al., 2000), a category of molecular devices that currently is of intense interest (Balzani et al., 2000; Sauvage, 2001). Let's now examine how chemosensors work and what factors must be considered during chemosensor design.

A key requirement of chemosensor function is that analyte binding must occur reversibly (Czamik, 1993a). This allows analyte concentration to be measured at equilibrium by optical detection of either the chemosensor-bound species or the analyte-free chemosensor. It also permits continuous measurements to be made with dynamic optical response to changing analyte concentrations. Irreversible chemical reactions produce chemodosimeters (Czarnik, 1993a), which can 332

Artificial Receptors for Chemosensors measure cumulative amounts of reactants. Such reactions are useful in singlemeasurement applications, as with timed-response uses of disposable sensors. Here we deal only with the more broadly applicable phenomena of reversible association or reversible chemical reaction between the chemosensor and its analyte. A chemosensor consists of a molecule incorporating a binding site, a chromophore or fluorophore, and a mechanism for communication between the two (Czamik, 1994). Analyte binding thus produces a change in chemosensor optical properties (absorption or fluorescence), as illustrated in Figure 1. Expressed in terms of host/guest chemistry (Cram and Cram, 1978; Cram, 1988), the position of the equilibrium established between the host (chemosensor), its guest (analyte) and the complex (or reversible reaction product) is governed by the association constant (Ka) as described by Equations 1 and 2. H + G (host) (guest)

ga

[HG] = Ka[H][G]

HG (complex)

(1) (2)

Analyte affinity is a matter of primary importance in chemosensor design. Because abiotic hosts typically bind guests much more weakly than natural receptors, the usual challenge is to improve sensitivity by increasing binding strength. On the other hand, an analyte that is present in high concentration can saturate the chemosensor, driving Equation 1 so far to the right that fluctuations in guest concentration may not produce sufficiently large changes in the optical signal. Thus, a key performance characteristic of a chemosensor is its responsivity, or rate of change of optical signal as a function of analyte concentration. A good rule of thumb for chemosensor design is that the target association constant should be approximately the inverse of the median guest concentration for the concentration range of interest (Eq. 3). Assuming that the amount of guest in solution far exceeds the amount of host in the sensor, this ensures that the concentration of free host is comparable to that of the complex (Equation 4, derived from Eqs. 2 and 3). Ka = 1/[G]

(3)

[HG] = [H]

(4)

Thus, fluctuation of guest concentration in either direction from the median will cause significant changes in both the concentrations of [HG] and [H]. This approach is best if it is not known whether the optical signal of host or complex will be most responsive to complexation, but there are exceptions. For example, a chemosensor with a much smaller Ka than the inverse of the median guest concentration can be used if [HG] can be detected against a small background 333

Bell and Hext signal, as in the formation of a fluorescent complex from a nonemissive chemosensor. Measurement of complexation by UV-visible absorption and fluorescence spectroscopy has been treated in greater mathematical detail elsewhere in the literature (Schneider and Yatsimirsky, 2000). Another issue of great importance in chemosensor design is the choice of chromophore or fluorophore used to report analyte binding. In particular, the absorption wavelength must be compatible with the light-absorbing properties of the medium in which measurements are to be made and with the light source. For example, proteins absorb ultraviolet light, so optical chemosensors for analytes in biological fluids (e.g., blood) should have ~.maxvalues larger than ca. 350 nm. Fortunately, Stokes' Law ensures that fluorescent chemosensors will emit light (~em) at longer wavelengths than that used for excitation (~,ex), but practical considerations come into play here, too. For reasons of instrument configuration, sensor cost, and light scattering, it may be better to use optical filters to reduce excitation light reaching the detector, rather than to arrange the detector perpendicular to the incident light beam, as in conventional spectrofluorometers. Therefore, a large Stokes' shift is generally desired (e.g., ~em = ~ex > 50 nnl).

Keeping the issues of target analyte affinity and acceptable fluorophore or chromophore wavelengths in mind, the chemosensor designer should proceed to consider binding selectivity, optical signaling mechanism, and the method to be used for immobilizing or delivering the chemosensor. These considerations are discussed in the following sections.

1.2. Molecular recognition A chemosensor must recognize its target analyte, much in the manner of picking out a familiar face in a crowd. More importantly, it must respond quickly and specifically (e.g., Hi there, Fred!). This selection process can result from selective binding or selective response, but in the latter case interfering substances will competitively inhibit optical response to the desired analyte. Having said this, complete specificity for a single potential guest is not necessary because the chemosensor needs to pick out its guest only from the substances typically present in the analyte solution. For example, a chemosensor for measuring sodium or potassium in blood need not discriminate against transition metals, unless the test is intended for patients who are already deceased! How can the molecular structure of a chemosensor be engineered to specifically bind the target analyte? Factors affecting molecular recognition in designed host-guest complexes have been the subject of intense scrutiny over the last three decades (Cram and Cram, 1978; Cram, 1988; Lehn, 1988, 1990; Rebek, 1988, 1990; Schneider, 1991). While there is no reliable way to predict host-guest 334

Artificial Receptors for Chemosensors selectivity, research in this field and in the broader area of supramolecular chemistry (V/3gtle, 1991; Lehn, 1995; Steed and Atwood, 2000) has identified several intermolecular forces that play important roles. The fundamental electrostatic (ion-ion, ion-dipole and dipole-dipole), hydrogen bonding, and van der Waals interactions are of course important, but more subtle forces have also been examined, including n-stacking (Hunter and Sanders, 1990; Hunter, 1993), cation-n interaction (Ma and Dougherty, 1997), CH-rc interaction (Laatikainen et al., 1995; Cloninger and Whitlock, 1998) and solvophobic effects. How can these intermolecular forces be marshaled and controlled to effect molecular recognition? Cram put this task most succinctly in two simple terms: complementarity and preorganization. The principle of complementarity is that: "to complex, hosts must have binding sites which cooperatively contact and attract binding sites of guests without generating strong nonbonded repulsions" (Cram, 1988). Clearly, the number and type of binding sites in the host must match those in the guest to produce an optimally stable complex. This principle alone, however, is insufficient for the design of a host that must select between similar guests, such as alkali metal cations differing only in size. Here we need guidance from the principle of preorganization: "the more highly hosts and guests are organized for binding and low solvation prior to their complexation, the more stable will be their complexes" (Cram, 1986). Excellent selectivities can be achieved in the alkali metal series with highly preorganized hosts, justifying the description of preorganization as the "central determinant of binding power" (Cram, 1986; Reinhoudt, 1988). While host preorganization leads to stronger and more selective guest binding, it also increases rigidity. In chemosensors, rigidity can hinder access of the analyte to the binding site, slowing equilibration (Eq. 1) considerably. This can produce unacceptable delays in equilibrium measurements, simultaneously retarding kinetic measurements, as well. The rate of complex dissociation (kout) is the factor limiting equilibration time, but remember that the binding rate (k~) is proportional according to: kin =

Kako~

(5)

Even when Ka is large, kt~ may be too small for kinetic measurements to be made within the 1-2 minute time frame required in certain sensor applications. Therefore, chemosensor rigidity must be balanced with flexibility. Preorganization does not need to be completely sacrificed, as evidenced by the various degrees of complex stabilization attending the chelate, cryptate and macrocyclic effects (Hancock and Martell, 1988). Indeed, a certain degree of chemosensor flexibility may be desired in order to produce an optical response by an "induced fit" mechanism, which leads us to the next consideration in chemosensor design. 335

Bell and Hext

Intrinsic

h~

Extrinsic

\\1

//

Figure 2. Cartoon showing different optical responses of intrinsic and extrinsic chromophores or fluorophores in chemosensor.

1.3. Optical signaling We have considered the desirable binding and optical characteristics of chemosensors. What about the mechanism coupling the binding event with signal transduction via the chromophore or fluorophore? It makes sense that the binding site and the optical reporter should be structurally integrated as much as possible in order to maximize this communication. In this context, it is useful to draw a distinction between intrinsic and extrinsic fluorophores or chromophores (Bell et al., 1993; Lakowicz, 1999), as shown in Figure 2. Intrinsic optical reporters are structurally integrated with the analyte binding site to maximize the influence of the bound guest on the optical properties of the chemosensor. Here, chemosensors have a profound advantage over biosensors. It is much easier to build a chromophore or fluorophore into a chemosensor binding site during its synthesis than to modify or introduce an optical reporter into the active site of an enzyme or the recognition site of an antibody. Such modifications of biological molecules usually damage their molecular recognition capabilities, so the optical reporter must be conjugated extrinsically to their binding sites. During chemosensor synthesis, optical and other properties 336

Artificial Receptors for Chemosensors (e.g., pKa) of the chromophore or fluorophore can also be fine-tuned in order to optimize performance. Whether the optical reporter is intrinsic or extrinsic to the molecular recognition site, the molecular mechanism for optical response should be considered during chemosensor design. While the mechanism of many known sensors, especially fluorescent chemosensors and biosensors, may not be well understood, Table 1 lists many mechanisms that have been identified and incorporated into chemosensor design. Here an important distinction is made between guest binding effects on chromophores vs. fluorophores. Useful absorbance effects generally result from changes in molecular structure, including proton transfer, other chemical reactions, and isomerization. Fluorescence is much more sensitive to subtle changes in the geometry and electronic structure of the ground state, as well as the electronic excited state. It is uniquely responsive to physical processes affecting depopulation of the emissive excited state (Lakowicz, 1999), such as conformational restriction occurring upon analyte complexation (McFarland and Finney, 2001; Mello and Finney, 2001). As indicated in Table 1, fluorescent chemosensors can utilize several photophysical processes, in addition to all of the structural mechanisms available to chromophoric chemosensors. The structural changes listed in Table 1 for chromophore signaling generally change the polarity or degree of electronic delocalization (conjugation) within the host chromophore. T h e chromophore protonation state can change when a neutral host ionizes during binding of a cationic guest, or when binding drives proton transfer between host and guest or within the host (tautomerization). Guest binding can also change electron distribution in the ground state of the chromophore and the energy of the locally excited (LE) state. The resulting "polarization" mechanism (Table 1) operates in a manner that is similar to solvatochromism (Reichardt, 1979). The LE states of most chromophores are more polar than their ground states, so polar solvents stabilize them more than the ground state. The resulting decrease in the energy difference between ground and excited states causes the )~maxto shift to longer wavelength as solvent polarity increases (positive solvatochromic effect). A polar guest can, in principle, cause a bathochromic shift in the absorption of a chemosensor by the same mechanism, or a specific charge or dipole interaction could cause the opposite effect, a hypsochromic shift. Fluorescence wavelengths are much more sensitive to solvent polarity and are subject to other effects, as illustrated in Figure 3 (Lakowicz, 1999). Solvent (and host-guest) interactions do not have time to adjust immediately to photon absorption and concomitant electronic excitation, occurring on the time scale of 10~5 seconds. Specific interactions and general solvent relaxation involving partial orientation of solvent dipoles stabilize the polar excited state. Relaxation

337

Bell and Hext Table 1. Optical Response Mechanisms.

Proton transfer Tautomerism Skeletal isomerism

Chromophore

Host-guest reaction Polarization Solvent displacement Quenching by guest

Fluorophore

Internal charge transfer (ICT) Twisted internal charge transfer (TICT) Resonance energy transfer (RET) Photoinduced electron transfer (PET)

of polar solvents has a larger stabilizing effect on the excited state than solvation of the LE state, producing larger bathochromic shifts of fluorescence emission bands. Clearly, guest binding can strongly influence the energy of the emissive state of a chemosensor, either by displacing solvent or by introducing new electrostatic interactions in the complex. Moreover, many fluorophores can form an internal charge transfer (ICT) state, also shown in Figure 3, involving transfer of electron density from electron,donating to electron-accepting groups. Interactions of the solvent or the bound guest molecule with these groups will determine the energy of the ICT state and also determine which state has the lowest energy. The twisted intramolecular charge transfer (TICT) state (Rettig, 1994) provides even greater opportunities as a fluorescence sensing mechanism in chemosensors. Formation of the TICT state involves rotation of donor and acceptor groups of the fluorophore. As stated earlier (Section 1.1), a large difference between absorption and fluorescence wavelengths is desirable in chemosensor design, and TICT emission can produce Stokes' shifts in excess of 100 nm. Because formation of the TICT state requires conformational mobility, rigidification of the chemosensor upon guest complexation can profoundly influence the intensity of long-wavelength TICT fluorescence. Resonance energy transfer (RET), involving through-space jumping of electronic excitation energy from a donor fluorophore to an acceptor fluorophore or a quencher, has been described as "the most general and valuable phenomenon for

338

Artificial Receptors for Chemosensors LE state At

~V

~v r Specific interactionS------"

~~~

Solvent relaxation _ ,, (10 1~ s) , ......ICT state _

> OA

h~F

~h~;

k'~ h~'F

~|_n~F iii

(lo ~s s)

So-

~v

f

~V

Figure 3. Jabtofiski diagram showing effects of specific solvent-fluorophore interactions, general solvent relaxation and formation of internal charge transfer (ICT) states (adapted from Lakowicz, 1999).

fluorescence sensing" (Lakowicz, 1999). Because RET operates over macromolecular distances, it can be used to detect major conformational changes of nucleotides or association between biomolecules, as in immunoassays. RET is not sensitive to changes in donor-acceptor separation in the subnanometer range, so it has not been found to be useful for detecting changes in the conformation of synthetic chemosensors upon guest binding. On the other hand, photoinduced electron transfer (PET) is a very useful sensing mechanism in fluorescent chemosensors (Bissell et al., 1993; Czarnik, 1993a,b, 1994; Desvergne and Czarnik, 1997; Granda-Vald6s et al., 2000). PET quenching of fluorescence occurs when an electron-rich group, such as an amino or phenoxide group, donates an electron to the fluorophore excited state. PET is different from ICT in that the electron donor group is not in direct conjugation with the rr system of the fluorophore. Binding of an electron-deficient guest to the donor group increases fluorescence emission by stabilizing and decreasing the mobility of donor electrons. It is also possible that some chemosensors may operate by PET quenching involving electron transfer from the guest to the fluorophore upon complexation.

1.4. Immobilization Chemosensors, as molecular devices, do not need to be tied down and integrated with optics and electronics components of sensing instruments. When employed as free agents, they are more commonly termed indicators, reagents, or molecular 339

Bell and Hext probes. Indeed, fluorescent probes of the intracellular concentrations of metals and other analytes have proven to be extremely useful in biomedical research (Lakowicz, 1999). Also, when used to fabricate disposable or reusable sensing materials for sensing instruments, chemosensors may not need to be covalently attached to a surface or other substrate. Often they can be hydrophobically adsorbed to a nonpolar surface layer or dissolved in the plasticizer of a polymer film or membrane. Covalent immobilization of chemosensors on surfaces or in materials is often used to improve sensor stability and avoid migration of the chemosensor into the analyte solution. Immobilization is usually a late-stage activity in sensor development, but the convenience of adding functional groups or side chains as covalent linkage sites should be considered while planning chemosensor structure and synthesis. Incorporation of a tether that does not interfere with binding or optical response can also be used to apply the techniques of polymer supported synthesis and combinatorial chemistry that are now crucial to pharmaceutical development. It has been pointed out that both drug discovery and chemosensor development are host-guest research and that bead-based screening of combinatorial libraries should be easily accomplished with fluorescent chemosensors (Czarnik and Yoon, 1999).

2. History As artificial molecular recognition systems, chemosensors have their roots in coordination chemistry of metals, the lock-and-key model for enzyme action, and the biomolecular receptor. These concepts were introduced by Alfred Werner in 1893, Emil Fischer in 1894, and Paul Ehrlich in 1904, respectively. Studies in the first half of the 20 th century on hydrogen bonds, clathrates, inclusion compounds, and rr donor-acceptor complexes set the stage for the appearance in the second half of the century of the discipline alternately termed host-guest chemistry (Cram and Cram, 1978; Cram, 1988), supramolecular chemistry (Lehn, 1988, 1990, 1995; V6gtle, 1991; Steed and Atwood, 2000), or molecular recognition (Rebek, 1988, 1990). Macrocyclic ligands for transition metals (Hancock and Martell, 1988) played a special role in this drama, and organic chemists began to take notice in 1967 when Charles Pedersen discovered crown ethers capable of mimicking the alkali metal transport properties of ionophore antibiotics (Pedersen, 1984; Gokel, 1991; Bradshaw et al., 1996; Bradshaw and Izatt, 1997). The design and synthesis of host compounds was recognized as an established field of research by the award of the 1987 Nobel prize in chemistry to Pedersen, Cram and Lehn. Earlier studies in the field of supramolecular chemistry revolved around the complexation of metal cations. Incorporation of dyes and fluorophores into the structures of crown ethers gave chemosensors for alkali metal and alkaline earth 340

Artificial Receptors for Chemosensors metal ions (Takagi and Ueno, 1984; Lbhr and V6gtle, 1985). At about the same time, chelation-based fluorescent probes for intracellular calcium and other metals were developed (Tsien, 1993). Chromogenic reagents for alkali metals based on cryptands, cryptaspherands, and spherands have also been devised (Helgeson et al., 1989; Chapoteau et al., 1993; Dolman et al., 1996). Such reagents are of practical utility in the determination of sodium, potassium, and calcium in blood (Kumar et al., 1988; Chapoteau et al., 1993). More recently, there has been much interest in fluorescent chemosensors for transition metals, mainly involving polyamine ligands (Fabbrizzi et al., 1997a, 1998; Bargossi et al., 2000; Prodi et al., 2000). Supramolecular chemistry of neutral molecules and anions (Schreeder et al., 1996; Bianchi et al., 1997; Schmidtchen and Berger, 1997) has been explored more recently, and here is where many current challenges for chemosensors lie. Many organic analytes of interest in biological systems exist as neutral molecules or anions (e.g., carboxylates and phosphates). Such relatively complex guest molecules present problems in the design of both the molecular recognition and optical response functions of the chemosensor. Therefore, the examples in the following section of this chapter are drawn from recent work on artificial receptors and chemosensors for neutral organic molecules, as well as organic and inorganic anions.

3. State of the A r t - Chemosensors for Organic Analytes The examples of chemosensors in the following section show that a wide range of analytes can be detected by this approach. In selecting these examples, we have used the Czarnik (1994) definition of a chemosensor consisting of a molecule incorporating a binding site, a chromophore or fluorophore, and a mechanism for communicating between the two. This excludes reagent approaches involving irreversible formation of colored or fluorescent products (Davis et al., 1999; Lewis et al., 2000). Also excluded by this definition are receptor-based sensing strategies in which the analyte competes for binding and displaces a fluorophore from the receptor, though elegant work has been done in this area (Lavigne and Anslyn, 2001; Wiskur and Anslyn, 2001; Springsteen and Wang, 2001; Cabell et al., 2001). A key aspect of chemosensor architecture is the attachment or conjugation of a reporter chromophore or fluorophore to a synthetic molecule that is responsible for recognizing the analyte. Recently, there have been important advances in developing sensors for organic analytes based on molecularly imprinted polymers (Subrahmanyanet al., 2000; Appleton and Gibson, 2000), but these materials do not fit our chemosensor concept. Effective sensing materials (optodes) that are not molecular devices can also be prepared by incorporating ionophores and ionizable chromophores or fluorophores in permeable polymers 341

Bell and Hext (Bakker et al., 1997; Murkovic and Wolfbeis, 1997; Spichiger-Keller, 1997; Krause et al., 1999). Finally, cyclodextrins have been used extensively as the recognition component of fluorescent sensor molecules (Ueno, 1993; de Jong et al., 2000; Wang and Ueno, 2000; Narita et al., 2001), but these examples are excluded because cyclodextrins are natural, rather than designed or artifical receptors. As described in the previous section, chemosensors for metal cations have been studied for many years and have been extensively reviewed (Takagi and Ueno, 1984; L/3hr and V/Sgtle, 1985; Czarnik, 1993a,b; Bissel et al., 1993; Fabbrizzi and Pogi, 1995; Dolman et al., 1996; Desvergne and Czamik, 1997; de Silva et al., 1997; Kimura and Koike, 1998a; Fabbrizzi et al., 1998b, 2000; Granda-Vald6s et al., 2000; Yamauchi and Hayashita, 2000; Prodi et al., 2000; Baragossi et al., 2000). While innovative work on chemosensors for metals continues (e.g., Hayashita et al., 2000; Bronson et al., 2001; Baxter, 2001; Raker and Glass, 2001; McFarland and Finney, 2001; Mello and Finney, 2001), the relative maturity of this field of research can be seen from the use of chromogenic chemosensors in clinical chemistry (e.g., Kumar et al., 1988; Chapoteau et al., 1993) and application of fluorescent chemosensors as intracellular probes for metals (e.g., Tsien, 1993; Zalewski et al., 1994; Lakowicz, 1999). The following examples have been selected from the recent literature to illustrate the chemosensor approach to detecting neutral and anionic organic analytes, as well as inorganic anions. Analytes of biological interest, such as carbohydrates, phosphates and blood metabolites, are highlighted. In most cases, performance characteristics including sensitivity, selectivity, and absorption or emission wavelengths must be improved to enable practical application. Nevertheless, these challenges can be met by tuning chemosensor structure by means of the power of organic synthesis.

3.1. Chemosensors for carbohydrates Development of chemosensors for biologically important carbohydrates has become a target for many research groups over the last decade. In general, there have been two approaches to the problem. The first approach is based on a rapid and reversible covalent bond formation process (James et al., 1996a,b; Shinkai and Takeuchi, 1996; Shinkai et al., 2000). The second approach utilizes hydrogen-bonding interactions to recognize the carbohydrate (Davis and Wareham, 1999). Monoboronic acids covalently interact with saccharides in aqueous solutions by formation of the corresponding boronate ester, as shown in Figure 4 with ethylene glycol representing the 1,2-diol unit of saccharides. With phenylboronic acid, this only occurs under basic conditions because esterification is favored when the hydroxyboronate anion is produced. However, if the aromatic ring is 342

Artificial Receptors for Chemosensors

OH

HO

O~

OH

B\

B' ] _

OH

\O I

OH e

/ \

e OH

HO

\ OH

OH

~

e O'~ -

OH

Figure 4. Reversible reaction of phenylboronic acid with 1,2-diols. made sufficiently electron-deficient (lowering the pKa of the boronic acid), hydroxyboronate anion formation can occur at neutral pH. Alternatively, an ortho-anfinomethyl substituent can be attached to the ring, resulting in a Lewis acid-base, boron-nitrogen interaction which significantly lowers the pKa of the boronic acid (Wulff, 1982). The order of affinity of arylboronic acids for monosaccharides is D-fructose > Darabinose > D-mannose > D-glucose. However, for many biological applications a greater degree of selectivity and sensitivity is required, as well as a different specificity. O n e way this can be obtained is by employing suitably designed diboronic acids, with the sugar bridging the space between the two boronic acid groups. This has led to receptors that can sense glucose, for example, with good selectivity and sensitivity. Two glucose chemosensors that have been developed by the groups of Shinkai and Norrild are given as examples, as they are considered by these authors to be significant advances in the field. Anthracene receptors 1 (James et al., 1994) and 2 (Eggert et al., 1999) both give optical responses on complexing monosaccharides at neutral pH (Figure 5). In unbound receptor 1, a PET process quenches the fluorescence of the anthracene moiety by electron transfer from the amino group to the anthracene excited state. When boronic acids form cyclic boronate esters, the Lewis acidity of the boronic acid is enhanced (Lorand and Edwards, 1959). Thus for receptor 1, monosaccharide binding increases the boron-nitrogen interaction, resulting in suppression of the PET process, leading to enhanced fluorescence. For receptor 2, fluorescence of the anthracene moiety is enhanced by saccharide binding, but the reason is unclear. Receptor 1 affords greater increase in fluorescence (7-fold 343

Bell and Hext

H;C

)2 N/CH3

HOOH ~B~ .CH3

1[

B(OH)2

0

HOOH H3C. ~E]/

B;, jCH3

H3C% .B

1 9 D-glucose

HO .OH "B'~.OH

HQ.../OH HO--Be

,OH -BeI + glucose

,.-

2 9 D-glucose

Figure 5. Two receptors for glucose, and structures of their predicted complexes

2-fold) and greater selectivity for binding of glucose over other monosaccharides, when compared to receptor 2. At pH 7.77 (33.3 % methanol buffer), the stability constants for 1 are: D-glucose (logKa = 3.6); D-allose (logKa = 2.8); D-fructose (logKa - 2.2); D-galactose (logK~ - 2.2). At pH 7.4 the only stability constant reported for 2 is for D-glucose (logK~ = 3.4). However, receptor 2 has the advantage of being water soluble, which is required for many applications. In fact, it has been predicted that a receptor similar to 2 could be used to construct a blood glucose sensor. As shown in Figure 5, the proposed structures of the complexes of these glucose receptors involve the pyranose form for 1 (James et al., 1994) and the furanose form for 2 (Eggert et al., 1999). Norrild has proposed that the optical responses of both receptors involve binding glucose in the furanose form (Bielecki et al., 1999), and there is additional evidence to support this case (Cooper and James, 1998). vs

A "sugar tweezer" that was designed from a boronic-acid-appended ~xoxobis[porphinatoiron(III)] (3, Figure 6) was shown to have high selectivity, as well as the highest known association constant for glucose (Takeuchi et al., 1996a, 1998). Unfortunately, the absorption spectrum of 3 is not significantly affected by saccharide addition, so an optical response would have to be introduced to enable its use as a sensor. The binding of saccharides was 344

Artificial Receptors for Chemosensors

r 8

--

B(OH)3 e

N. ":"' '~/"

(HO

(HO)3B e

3

4

Figure 6. Boronic acid receptors selective for D-glucose (3) and for D-lactulose (4).

observed by circular dichroism spectroscopy, a technique that would be difficult to adapt to a sensor. Also, it would be necessary to lower the pK, of the boronic acid groups in 3 so studies could be performed at physiological pH. Alteration of the distance between two boronic acid groups has also led to receptors for small saccharides (James et al., 1997), as well as di- and trisaccharides. An example of a receptor which selectivity binds D-lactulose in methanol is tetraarylporphyrin 4, also shown in Figure 6 (Kijima et al., 1998). Receptors which discriminate between the two enantiomers of a sugar have also been synthesized, though overall the selectivity between two different sugars is not great (Takeuchi et al., 1997a; Mizuno et al., 1999, 2000). Most carbohydrate receptors based on the hydrogen-bonding approach have been studied by either NMR spectroscopy or circular dichroism (Davis and Wareham, 1999). As sugar binding was apparently not amenable to study by either UVvisible or fluorescence spectroscopy, these receptors will not be discussed in this chapter. However, in the case of four receptors, two of which are shown in Figure 7, the binding of carbohydrates was monitored by UV-visible spectroscopy (Rusin and Kr~d, 1999; Kr~il et al., 2001), and for one of these (receptor 5) fluorescence spectroscopy was used as well. These receptors show good affinity for some saccharides, even in polar solvents, though overall the selectivity is poor. In DMSO (containing 5% methanol) as measured by fluorescence spectroscopy, some association constants for 5 are: o-fructose (log/<, = 1.9); D-a-lactose (logKa- 3.4); D-trehalose (logKa = 3.1); D-maltotriose (logK, = 2.8). 345

Bell and Hext

I

I 6

Y = (CH2)6 X = CO(CH2)4CO

Figure 7. Hydrogen-bonding receptors for disaccharides (S) and trisaccharides (6).

However, Rusin and Knil (1999) reported a slight variance in the same association constants as measured by UV-visible spectroscopy. Also, the selectivity in water (5 % methanol) was different for D-fructose (logKa = 2.3) and D-o~-lactose (logKa = 2.4). In water (containing 5% methanol), some association constants for 6 are: D-glucose (logKa = 3.1); D-lactose (logKa = 4.5); Dmaltotriose (logKa = 4.7). Overall, receptor 6 shows some selectivity for the trisaccharide maltotriose, relative to monosaccharides. There have been a few reports of receptors designed for sensing carbohydratelike molecules (Takeuchi et al., 1996b, 1997b; Yamamoto et al., 1998). An interesting example is receptor 7 (Figure 8), designed for sensing D-glucosamine hydrochloride (Cooper and James, 1997). The system acts like an AND logic gate in that the anthracene fluorophore is turned off by both amino groups due to PET quenching. For the molecule to become fluorescent, two events are required: ammonium binding by the azacrown ether and diol binding by the boronic acid. This can be achieved in the binding of D-glucosamine but not of Dglucose.

3.2. Chemosensors for other neutral organic molecules Chemosensors for neutral organic molecules other than carbohydrates can be roughly divided into two types" a molecule that reversibly reacts with a particular functional group or one whereby the receptor has been tailored to selectively detect a specific compound. An approach to the detection of alcohols (Mohr and Spichiger-Keller, 1997; Mohr et al., 1998a) and amines (Mohr et al., 1998b) is based on their reversible reaction with an electron deficient carbonyl functionality that is part of a dye 346

Artificial Receptors for Chemosensors

(HO)2B~

Figure 8. A boronic acid receptor for D-glucosamine.

N•CsNr

Fa

F3C

8

8a

Figure 9. A reversible reaction-based sensor for alcohols and amines.

molecule, as shown in Figure 9. The dye molecule is incorporated into a lipophilic membrane which is then exposed to aqueous substrate solutions. Using dye 8 for the sensing 'of alcohols, a catalyst is required to shorten the response time (unlike amines where it was not necessary), and it was found that the response varied with the lipophilicity of the alcohol. Formation of the hemiacetal (Sa) is monitored by a hypsochromic shift in the UV-visible absorption spectrum and a decrease in the fluorescence emission spectrum of 8. The sensitivity of a membrane incorporating 8 is lower than that of enzymatic sensors and can not be used in clinical applications. However, it is suitable for monitoring ethanol content in alcoholic beverages. Membrane incorporation of a perylene-based dye that can react reversibly with aldehydes and ketones has also been used for the optical sensing of aqueous propanal at low pH (Mohr et al., 2000). 347

Bell and Hext

O

O

t-Bu

N

t-Bu t-Bu

0 R R = CH2

0 R

0 R

t-Bu

0 R

10

1

He 9

Figure 10. Two calixarene-based amine sensors.

Another approach to sensing amines uses derivatized calixarenes (Kubo, 1998, 1999; Kubo et al., 1998). The amine deprotonates a phenol moiety on the receptor, and the resulting anion binds the concomitant ammonium species. A calix[8]arene, for example, gives a different optical response in DMSO with simple monoamines when compared to diamines (Chawla and Srinivas, 1994). A calix[4]arene was found to sense butylamines, the selectivity depending on the shape of the amine (Kubo et al., 1995). This calixarene was then modified to afford 9, as shown in Figure 10, which can visually discriminate between enantiomers of phenylglycinol (Ka = 66 M -1) and phenylalaninol (Ka = 159 M "1) in ethanol, where binding was only observed for one enantiomer (Kubo et al., 1996, 1997). However, it should be noted that a hydrogen-bonding interaction between the substrate hydroxyl and receptor was required, as no discrimination/binding was observed for the enantiomers of 1-phenylethylamine. Calix[4]arene 10 (Figure 10) though does show some selectivity for (R)-Iphenylethylamine (Grady et al., 1996). Also, increased discrimination with this receptor is observed for phenylglycinol (Grady et al., 1998). Other compounds that sense amines are certain substituted zinc porphyrins (Sen and Suslick, 2000). These compounds have bulky groups appended to a tetraphenylporphyrin ring that offer some shape-selectivity discrimination. For example, 11, with one siloxy group missing (Figure 11), shows a wide range of sensitivity for various cyclic secondary amines. This shape-selectivity adds a 348

Artificial Receptors for Chemosensors

R

NH2 R

R

12

CH3 R1 = OH

R = -O-~i-C(CH3)3 OH3

Figure 11. A chemosensor for amines (11) and aminoacids (12).

second layer of molecular recognition to previously reported work. This involves an array of various metalloporphyrins, derived from tetraphenylporphine bound to different metals, that were immobilized on reverse phase silica gel plates (Rakow and Suslick, 2000) and used to visually identify a range of ligating vapors, e.g. ethanol, pyridine, hexylamine, triethylphosphite. On exposure to a nitrogen stream containing the substrate, each metal binds the compound to varying degrees by axial ligation. The combined responses of the metalloporphyrins were shown to be unique for various functional groups. A fluorescent sensor 12 (shown in Figure 11) has been developed which gives varying responses to aminoacids in MeOH:water (3:2) at pH 9.5 based on the length of the carbon chain between the two functionalities (de Silva et al., 1996). For example, some binding constants are: 5-aminopentanoic acid (Ka = 84 M'I); 3-aminopropanoic acid (Ka = 17 M~). The ammonium moiety binds to the azacrown, turning off a PET process so a fluorescence enhancement is observed. The carboxylate hydrogen bonds/ion pairs with the guanidinium moiety, but this has no significant effect on fluorescence. However, similar effects can be seen with simple propylamine (Ka = 79 M4), as well as sodium and potassium. Thus, to improve 12 as a sensor for aminoacids, the guanidinium moiety should be replaced by an alternative PET-active carboxylate binding group to produce an AND gate. An example of a chromogenic chemosensor for dicarboxylic acids is 13 (see Figure 12) which showed some substrate dependency in chloroform (Goodman et al., 1995). Some binding constants, as measured by UV-visible spectroscopy, are: 349

Bell and Hext

••N

N,,,H I-L,,~r~ o o

O

O--(X)-B~:).~O--(X)--O~OBn

.y

~

14

N,,.H

13 H / ~

(+)- (R,R)-14 X = CH2(P-CsH4)CH2

Figure 12. Chemosensors 13 and 14 for dicarboxylic acids.

glutaric acid (logKa = 4.9) and N-benzyloxycarbonylglutamic acid (logKa = 4.6). It is proposed that bond rotations in the receptor to favorably orient the carboxylic acid binding sites causes the chromogenic effect. In theory the receptor can bind two diacid molecules, but it is predicted that the major changes in the receptor occur on binding the first molecule. Fluorescent chemosensor 14 (Figure 12) was shown to bind N-(benzyloxy)carbonyl-protected aspartic (logKa = 4.8) and glutamic acid (logKa = 4.7) in dichloromethane (Lustenberger et al., 1998). The cause of the fluorescence change on binding was not discussed. Also, a-negative cooperativity effect was observed for the binding of a second substrate. Both receptors have a reasonable affinity for the selected diacids in the solvents studied, but it is questionable whether either system would work in aqueous media. Also of interest is a receptor that has been shown to give both chromogenic and fluorescence responses to tx,o~icarboxylate anions in DMSO, with some selectivity depending on the chain length being observed (Mei and Wu, 2001). D'Souza has reported the use of suitably appended porphyrins to sense quinones and hydroquinones (D'Souza, 1996; Deviprasad et al., 1998; D'Souza and Deviprasad, 2001). For example, quinone-appended porphyrin 15 (see Figure 13) was shown to give a fluorescence response to hydroquinone (D'Souza et al., 1997). The free receptor is weakly fluorescent, and this is believed to be due to a PET process involving the porphyrin and the appended quinone. The excitedstate electron transfer is inhibited when hydroquinone binds to the quinone, thereby enhancing the fluorescence emission of the porphyrin. The binding constant calculated from 1H NMR data was 13.1 M 1 in CDC13. 350

Artificial Receptors for Chemosensors

H-O 0

15. hydroquinone

15

Figure 13. A porphyrin chemosensor for hydroquinone (15) and proposed structure of its complex.

~N~

~~ON~ 0

Nicotine

oo4->

NHCO_

Ootinine

NHCO

v "/

NH

yHO-

NH CONH

NHCO ---

CONH

NHCO ~

~

17

16

Q = COC6H4N = NC6H4NMe2 F = (CH2)2NH- SO2CloH6NMe2 (Me, methyl; Ph, phenyi)

Figure 14. Chemosensor for nicotine (16) and some tripeptides (17).

351

CO2F

Bell and Hext CN

H

H

I

CN

I

N'- H

CN

H

H

CN

H~ N .,~ urea U

J

U

J 18

18 9 urea

NO2

NO:,

e

\"

-

o\ 02

N H

O2N"

H

~

"N

H

I:'I

creatinine r

19

19 9 creatinine

Figure 15. Chemosensors for urea (18) and creatinine (19), and proposed complexes formed on binding.

Compound 16, as shown in Figure 14, was observed by UV-visible and fluorescence spectroscopy to bind nicotine and cotinine in toluene, logKa = 5.66 and 4.63, respectively (Deviprasad and D'Souza, 2000). For nicotine, the pyridyl nitrogen axially ligates to the zinc ion and the pyrrolidine nitrogen hydrogen bonds to the carboxylic acid. This results in a bathochromic shift of the Soret and visible bands of 16 and a decrease in intensity of the fluorescence emission bands of the zinc porphyrin at 605 and 650 nm ()~ = 420 nm). An alternative approach to chemosensors has been taken by Still, whereby a potential receptor is synthesized and then tested against a bead-supported tripeptide library (Lowik et al., 1998; Ryan and Still, 1999; Iorio and Still, 1999; Ryan et al., 2000). By having a chromogenic moiety attached to the receptor, one is then able to find a receptor that binds specific tripeptide sequences. As such, the receptors are not chemosensors because no signalling mechanism for binding exists. However, this can be overcome by attaching a fluorophore (F) and a quencher (Q) to the receptor (Chen et al., 1998; Rothman and Still, 1999). For example, receptor 17 (Figure 14) bound only two sequences of a solid-phase binding screen consisting of 3375 N-acetylated side chain-protected tripeptides. Receptor 17 was then shown to bind one of these tripeptides (as their N-acetyl, C-n-propylamides) in chloroform. The free receptor is weakly fluorescent, 352

Artificial Receptors for Chemosensors indicative of highly effective fluorescence quenching of F by Q. Binding of the tripeptide increases the F-Q separation resulting in enhanced fluorescence. A binding constant of 2.6 x 105 M -1 was observed. Two receptors tailored to bind specific analytes, 18 and 19, are shown in Figure 15. Receptor 18 was designed to bind urea via six hydrogen bonds (Bell and Hou, 1997). Binding was observed in DMSO as evident by a 16 nm bathochromic shift in the UV-visible absorption spectrum of the receptor, and from which a stability constant of 1.4 x 104 M 1 was calculated. Receptor 19 was designed to bind creatinine via four hydrogen bonds (Bell et al., 1995). On binding creatinine, by extracting the analyte from buffered water into a chloroform solution of the receptor, a proton shift from the phenol OH to a naphthyridine nitrogen atom occurs. This causes a large bathochromic shift in the UV-visible absorption spectrum of the receptor, and from which a Ka of 2 x 106 M ~ can be calculated. 3.3. Chemosensors for anions

Sensing schemes for halide anions have been devised utilizing nonselective quenching of electronic excited states of N-alkyl heterocyclic fluorophores (e.g., Parker et al., 1997; Geddes, 2000; Geddes et al., 2001) or of transition metal complexes (e.g., Huber et al., 1998). Lacking any molecular recognition unit, these compounds do not fit our chemosensor definition. A recently reported approach to fluoride detection based on dye deprotonation in nonpolar media also fails to meet this criterion (Lee et al., 2001b). Polymers containing lipophilic cations and polarity sensitive dyes have also been used to develop sensor membranes for nitrate, nitrite and other inorganic anions (Mohr and Wolfbeis, 1995, 1996a,b; Mohr et al., 1997). Also not considered here are sensor approaches to citrate (Metzger and Anslyn, 1998), inositol triphosphate (Niikura and Anslyn, 1999) and other organic anions based on displacement of untethered anionic fluorophores from artificial receptors. Beginning with the work of Czarnik (1994), polyamines have proven to be useful for the design of fluorescent chemosensors for anions (see reviews by Fabbrizzi et al., 1998b, 2000). Aliphatic polyamines are protonated at neutral pH in water, and initial studies utilized ion pairing and hydrogen bonding to bind phosphate and other complex anions. Some examples of Fabbrizzi's approach employing polyamine chelates to bind zinc(R), which in turn coordinates anionic ligands, are shown in Figure 16. The zinc(H) complex of the anthracene-bearing tren ligand 20 (DeSantis et al., 1996) binds carboxylates in ethanol with K, values in the range of 105 M ~. Anthracene fluorescence is quenched by pdimethylaminobenzoate and by m-nitrobenzoate, but not by unsubstituted benzoate, suggesting a host-guest PET process. Chemosensor 21 (Fabbrizzi et al., 1999) consists of a tren platform having an N,N-dimethylaniline unit attached

353

Bell and Hext

,NH2

? H2

N

NH

20

/N\

N~HN ~ N

H~

2Zn2+

Figure 16. Fluorescent chemosensors for anions acting as ligands for polyamine-chelated zinc(II).

to each arm (Figure 16). Irradiation of the zinc(H) complex of 21 at 300 nm produces a TICT emission at 360 nm, which is quenched by any aromatic carboxylate, including benzoate. In the case of both chemosensors, aliphatic carboxylates, such as acetate, competitively bind with weaker Ka values, but do not quench fluorescence. Another chemosensor in which two zinc(H) ions are chelated by two tren units attached to anthracene (Fabbrizzi et al., 1997b) binds the imidazole moiety of histidine with fluorescence quenching. A limitation of this system is that basic conditions (pH 9.6) are required to deprotonate the imidazole ring upon binding. Polyamine 22 (Figure 16) is a macrobicyclic version in which the tren units are also linked by p-xylylene bridges (Fabbrizzi et al., 1998a). The complex of 22 with two zinc(H) ions is stable in water over a wide pH range and selectively binds small anions, such as azide (N3", logKa = 5.8). Formation of the complex with N3- quenches anthracene fluorescence, without interference by NO3, HCO3, SO42, CI or Br, even when present in 19-fold excess. Cyanate (NCO) binds somewhat more strongly to the cavity (logKa = 6.5), but does not quench fluorescence. This suggests a PET quenching mechanism, as N C O is a weaker reducing agent than N3. 354

Artificial Receptors for Chemosensors Neutral hydrogen bond donor groups, such as urea and thiourea, have been found useful for binding anions in nonpolar media with some degree of selectivity (Schmidtchen and Berger, 1997; Lee and Hong, 2000). Kato et al. (2001) have found that N,N'-bis(p-nitrophenyl)thiourea (23, Figure 17) acts as a chromogenic reagent for acetate in 99"1 CH3CN/I"I20 (Ka = 3 x 105 MI). A bathochromic shift of the longest wavelength absorption band from 343 to 392 nm appears to be an effect of chromophore polarization in the formation of the hydrogen bonded complex. No chromogenic response was observed with Br, I, SCN, NO3", HzSO4-or C104-. A weak response was observed with H2PO4, which may bind, but is less basic than acetate. The chromogenic response and acetate affinity of 23 are greater than reported earlier for a mono-p-nitrophenyl thiourea (Nishizawa et al., 1998a). Two thiourea units flanking a p-nitrophenylazo substituted phenol produce a chemosensor (24, Figure 17) that binds acetate, H2PO4 and fluoride in CHC13, producing an intense absorption at ca. 530 nm (Lee et al., 2001a). Hydrogenbonded recognition is clearly involved for acetate and H2PO4", but fluoride appears to act simply as a base to form the anion of 24. Also shown in Figure 17 is 1,2-diaminoanthraquinone (25), which is one of many commercially available compounds found by Miyaji and Sessler (2001) to act as chromogenic sensors showing some selectivity for more basic anions in nonpolar solvents. The authors proposed the formation of a host/guest charge transfer complex, but chromophore polarization may cause the observed bathochromic shifts, as proposed for 23. Hydrogen bond donor groups, including urea and thiourea, have also been used to produce fluorescent chemosensors for anions, and some examples are shown in Figure 18. Compound 26 (Hennrich et al., 2001) beating two sidearms containing both thiourea and guanidine subunits, as well as a one-arm analog, binds carbonate and bicarbonate in methanol with strong enhancement of the naphthalene-like emission bands at 387 and 392 nm. No response to CI, Br, I-, C104" or NO3" was observed, but HPO4 2" caused a weak fluorescence enhancement. Primary binding of carbonate to the guanidine subunit was proposed and the increase in fluorescence intensity was ascribed to rigidification. Tripodal naphthylurea chemosensor 27 shows weak fluorescence (Ln~x ca. 380 nm) in DMF due to PET quenching by electron transfer from the tertiary amino group (Xie et al., 1999). Acidic, tetrahedral anions, such as H2PO42 (Ka = 40,700 M 1) and HSO3 (Ka = 2,750 M l ) bind with fluorescence enhancement, apparently due to proton transfer from guest to host within the complex. Both 26 and 27 require excitation at a relatively short wavelength (310 nm).

355

Bell and Hext

02N

S

/NO2

0H3CO2e _

I H

02N

S

I H

I H

I H

i i

23

NO2

.

i i

%0

Bu I

S , ~ N-,,.H

I

CH3

-.N~H 0 2 ~ N ~

~~]0

OH

NH2

NH2

'-,,..N~H 0

S.~N / I Bu

25

24

Figure 17. Hydrogen bonding chromogenic chemosensors.

Pyrene-attached thiourea 28 (Nishizawa et al., 1998b) binds acetate in acetone (Ka = 7,000 M ]) with some selectivity over H2PO4 (Ka = 5,200 M 1) and CI (Ka = 1,000 M-l). These guests quench pyrene emission (396 nm) with similar efficiencies, apparently by increasing PET quenching involving electron transfer from the thiourea unit. The similarly positioned thiouronium unit in 29 apparently acts as an electron acceptor to quench naphthalene fluorescence (Kubo et al., 2000). Hence, acetate binding in acetonitrile (Ka > 106 M "1) causes an enhancement in fluorescence intensity (X~m= 336 nm). Also in acetonitrile, bis(2-hydroxynaphthalimides) 30 and 31 show fluorescence enhancement upon binding acetate, HzPO4 or 1:r (Yoshida et al., 2000), but negligible response to CI' Br, I or C104-. The response mechanism apparently involves deprotonation of one naphthol unit. Again, fluorescent chemosensors 28-31 require UV excitation (e.g., 314 nm for 28).

356

Artificial Receptors for Chemosensors 0

NH 0

0

HN

0

26

H~'N O / ~/H N/ H

CH3 I s , ~ N~H

~

H'N~--~O HN

N~.H

27

28

,.)

H N~R ====/ \

s

29

OH 0 OH 0

R = 1-naphthyl 31 R = CH2[CH2OCH2]CH2

30

Figure 18. Hydrogen bonding fluorescent chemosensors.

Ion pairing with anions can affect the electronic properties of cationic metal complexes, and Beer (1998) has demonstrated that pendant hydrogen bond donors can produce electrochemical and fluorescence sensors having a degree of selectivity (Beer et al., 1996). Of particular significance is the calix[4]arenebridged ruthenium(H) tris(bipyridyl) complex 32 (Figure 19), which binds H2PO4 (Ka = 2.8 x 104 M -I) in DMSO selectively over CI (1.6 x 103 M 1) or Br (3.6 x 10z M I) (Szemes et al., 1996). The metal to ligand charge transfer emission is shifted by H2PO4 from 640 to 622 nm and the fluorescence quantum yield is increased, apparently as a consequence of rigidification. Also shown in Figure 19 is a unique chemosensor (33) that combines metal ligand coordination with recognition by hydrogen bond donor sites (Cabell et al., 2001). Prior 357

Bell and Hext

t-Bu

t-Bu

t-Bu t-Bu 5"

0

0 0 H H

( " 0

N--H

I

O

)

H--

0

~ 0

cu(ll)

H\N HH

_

H--N

H\ H'

33

32

Figure 19. Fluorescent chemosensors consisting of metal complexes with hydrogen bond donor sites for anion recognition.

complexation of Cu(II) causes metal-induced quenching of the 1,10phenanthroline fluorescence centered at 365 nm. Binding of citrate partially restores fluorescence and model studies were done to show that this effect is not caused by stripping of Cu(II) from the phenanthroline moiety. The fluorosensor shows linear response to aqueous citrate concentrations in the #M range and can be used to measure citrate in commercial beverages. Macrocyclic polypyrroles, such as expanded porphyrins (Sessler et al., 1999) and calixpyrroles (Miyaji et al., 1999; Anzenbacher et al., 2000a,b), bind anions and have been investigated as optical chemosensors by the Sessler group. For example, enhancement of the long wavelength (ca. 680 nm, ~x = 450 nm) emission of sapphyrin 34 (Figure 20) can be used for aqueous phosphate measurement in the mM range (Sessler et al., 1999). Fluorinated calix[4]pyrrole 35 (Figure 20) binds H2PO4 more strongly than its nonfluorinated analog, and fluorinated dipyrrol-2-ylquinoxaline 36 can be used as a chromogenic or fluorescence quenching sensor for F-, CI and H2POs in dichloromethane (Anzenbacher et al., 2000b). Perhaps of greater significance are a series of "second generation" calixpyrrole chemosensors (37, Anzenbacher et al., 2000a) that display higher anion affinities than previous systems bearing anthracene 358

Artificial Receptors for Chemosensors

E

F

F F

F

~"~'/"'~ ~

NH

F 0

OH

36

OH

35

i--t,,,NtFI

OH OH 34

FI = fluorophore 37

Figure 20. Polypyrrole receptors and chemosensors for anions.

fluorophores (Miyaji et al., 1999). Calix[4]pyrroles of type 37 with sulfonamidelinked dansyl, lissamine-rhodamine B, or thiourea-linked fluorescein fluorophores (F1) showed fluorescence quenching with F , CI, H2PO4 and HP2073. They all show high selectivity for phosphate vs. chloride, and the thiourea-linked fluorescein system binds phosphate with high affinity in 96:4 acetonitrile/water at pH 7. Adenosine 5'-triphosphate (ATP) is an anion of biological relevance, and several approaches have been used to devise optical chemosensors for measuring ATP concentrations in water. Examples are a linear polyamine attached to anthracene (Albelda et al., 1999), cyclam appended to a ruthenium(H) bis(terpyridyl) luminophore (Padilla-Tosta et al., 2000) and the corresponding Cu(II)/cyclam complex (Padilla-Tosta et al., 2001). Nishizawa et al. (1999) have used pyrophosphate (P2074) complexation of a simple guanidinium-substituted pyrene to influence the intensity of excimer emission, and this approach could be applied to ATP. Another interesting approach involves the displacement of an anionic fluorophore tethered to a cationic site similar to that in 33 (Figure 19). Combinatorial chemistry was used to vary the structure of the peptide tether in 359

Bell and Hext order to enhance ATP binding in water (Ka = 3.4 x 103 M-a; Schneider et al., 2000). It is clear that efforts will continue to improve affinity, selectivity and optical response in chemosensors for this important analyte.

4. Advantages and Limitations of Chemosensors When compared to biosensors, chemosensors potentially offer many advantages. A receptor can be tailored for non-biological analytes, examples being 21 and 22, (Figure 16) which have been shown to sense aromatic carboxylates and azide respectively (Fabbrizzi et al., 1998a, 1999). As stated in Section 1, chemosensors are generally more robust than biosensors toward chemical and thermal degradation. They are stable indefinitely and can operate under diverse conditions. Properties of the receptor, including affinity, selectivity, optical response, and absorption and emission wavelengths, can be tuned for the application desired. For example, Shinkai and co-workers (James et al., 1994) prepared a sensor that can detect glucose at physiological levels (in blood, 0.3 1.0 mM), which is often the primary aim of many researchers. However, lowering selectivity and sensitivity by altering the structure of the receptor gave a chemosensor that can detect the overall saccharide concentration at a higher level (ca. 10 mM, Linnane et al., 1995). This may be of use when confectionery is formulated in industry. Modifying the receptor also can allow the chemist to alter solubility, kinetics of binding, and provide a site to attach a tether required for immobilization. Incorporation of an intrinsic chromophore or fluorophore provides a sensor with good communication of the binding event. As stated previously in Section 1.3, modification of a biosensor so that the binding site includes an optical reporter usually damages its molecular recognition capabilities. Finally, a chemosensor can work in a non-aqueous environment, even the gas phase, in principle. The main limitation of chemosensors is that they often do not possess sufficient sensitivity or selectivity for the application desired. The design of artificial receptors having high affinity and selectivity toward the target analyte is not reliable and requires a process of trial and error. Synthesis of each target chemosensor is also a tedious process.

5. Potential for Expanding Current Capabilities As stated in the previous section, affinities and selectivities of chemosensors (particularly in water) generally need to be improved, especially for biological applications. When combined with the fact that carefully designing and synthesizing a target receptor may not produce a potential chemosensor for the target analyte, it is clear that a better process for predicting the performance of designed chemosensors is desirable. Many aspects of chemosensor design might 360

Artificial Receptors for Chemosensors be addressed through computer modeling. The ideal situation would be to input the analyte of interest, along with characteristics required for the chemosensor, and the program would then design an optimum compound. Obviously, this level of predictability will not be achieved in the near future. However, the development of computer programs to screen various hypothetical receptors at a higher level of theory than simple molecular mechanics (Burkert and Allinger, 1982) may be feasible in the near future. Another possible solution would be to employ the power of combinatorial chemistry, which has revolutionized drug discovery (Seneci, 2000). Then a vast array of receptors could be generated and tested against the analyte of interest (Leipert et al., 1999). Fortunately, a lot of progress is being made in the synthesis of non-peptide molecules using the combinatorial approach, but the range of molecular structures that can be synthesized on solid supports need to be expanded. As noted in Section 3.2, reverse logic employing this technology was used to design chemosensor 17 for a tripeptide sequence (Chen et al., 1998). The approach of Suslick (Rakow and Suslick, 2000), in which the response of an array of chemosensors was shown to be unique for a chemical class of compounds, may also prove useful. This strategy may overcome problems with poor selectivities of individual chemosensors, but affinities available from current technology would have to be improved. Ultimately, this leads to the concept of a system similar to the "artificial tongue" (Lavigne and Anslyn, 2001) or nose (Dickinson et al., 1996). Currently, a lot of interest is being shown in the development of sensors to detect the vapors emitted from landmines (Yang and Swager, 1998), and this technology may prove adaptable for other classes of analytes. Significant progress in the employment of chemosensors in research and industrial settings has occurred, particularly in the area of detecting metal cations (Section 2). Recently the detection of neutral organic molecules and anions has been explored, as described in this chapter. Of particular current interest is the development of a chemosensor for monitoring blood glucose levels in diabetes patients. Two target applications are in vivo monitoring of glucose, which would allow automatic injection of insulin from a small portable device carried on the person (type I diabetes), and regular checking of blood/urine in vitro in type II diabetes patients. Note, as the level of glucose in the two cases are different (0.31.0 vs 10 mM), two different chemosensors are desired (James et al., 1994; Linnane et al., 1995). The examples in Section 3 of this chapter illustrate the range of analyte structures and chemosensor designs that have already been addressed with the artificial receptor approach. Overall, the sensitivities and selectivities of these chemosensors are weaker than biosensors, when available for the same analytes. On the other hand, the wide variability of the molecular structures of artificial 361

Bell and Hext receptors and of the mechanisms for optical response indicate that artificial chemosensors can effectively compete with biosensors for many applications. By and large, the work published to date in this field emanates from academic laboratories and focuses on conceptual, rather than practical, advances. In particular, more attention needs to be directed to the problems of water compatibility, absorption and emission wavelengths, and methods for immobilization to produce practical sensors. The design of optical chemosensors is still a relatively new concept, and much work lies ahead in the application of these molecules to detection and quantitation of analytes in biomedical, environmental, and industrial fields of interest.

6. References

Albelda, M. T., M.A. Bernardo, E. Garcfa-Espafia, M.L. Godino-Salido, S.V. Luis, M.J. Melo, F. Pina and C. Soriano, 1999, J. Chem. Soc. Perkin Trans. 2, 2545. Anzenbacher, P., Jr., K. Jurs~owi and J.L. Sessler, 2000a, J. Am. Chem. Soc. 122, 9350. Anzenbacher, P., Jr., A.C. Try, H. Miyaji, K. Jursikowi, V.M. Lynch, M. Marquez and J.L. Sessler, 2000b, J. Am. Chem. Soc. 122, 10268. Appleton, B. and T.D. Gibson, 2000, Sens. Actuators B 65, 302. Bakker, E., P. Btihlmann and E. Pretsch, 1997, Chem. Rev. 97, 3083. Balzani, V., A. Credi, F.M. Raymo and J.F. Stoddart, 2000, Angew. Chem. Int. Ed. 39, 3348. Bargossi, C., M.C. Fiorini, M. Montalti, L. Prodi and N. Zaccheroni, 2000, Coord. Chem. Rev. 208, 17. Baxter, P. N. W., 2001, J. Org. Chem. 66, 4170. Beer, P. D., 1998, In Molecular Recognition and Inclusion, Ed., A.W. Coleman, Kluwer, Dordrecht, p. 97. Beer, P. D., S.W. Dent, N.C. Fletcher and T.J. Wear, 1996, Polyhedron 15, 2983. Bell, T. W. and Z. Hou, 1997, Angew. Chem. Int. Ed. 36, 1536. Bell, T. W., D.L. Beckles, P.J. Cragg, J. Liu, J. Maioriello, A.T. Papoulis and V.J. Santora, 1993, In Fluorescent Chemosensors for Ion and Molecule Recognition, Chap. 7, Ed., A.W. Czamik, ACS Symposium Series 538, American Chemical Society, Washington, DC, pp. 85-103. Bell, T. W., Z. Hou, Y. Luo, M.G.B. Drew, E. Chapoteau, B.P. Czech and A. Kumar, 1995, Science 269, 671. Bianchi, A., K. Bowman-James and E. Garcfa-Espafia, Eds., 1997, Supramolecular Chemistry of Anions, John Wiley Sons, New York. Bielecki, M., H. Eggert and J.C. Norrild, 1999, J. Chem. Soc. Perkin Trans. 2, 449. Bissell, R. A., A.P. de Silva, H.Q.N. Gunaratne, P.L.M. Lynch, G.E.M. Maguire, C.P. McCoy and K.R.A.S. Sandanayake, 1993, In Topics in Current Chemistry, Vol. 168, Ed., J. Mattay, Springer-Verlag, Berlin, p. 223. 362

Artificial Receptors for Chemosensors Bradshaw, J. S. and R.M. Izatt, 1997, Acc. Chem. Res. 30, 338. Bradshaw, J. S., R.M. Izatt, A.V. Bordunov, C.Y. Zhu and J.K. Hathaway, 1996, In Comprehensive Supramolecular Chemistry, Vol. 1, Pergamon, p. 35. Bronson, R. T., J.S. Bradshaw, P.B. Savage, S. Fuangswasdi, S.C. Lee, K.E. Krakowiak and R.M. Izatt, 2001, J. Org. Chem. 66, 4752. Burkert, U. and N.L. Allinger, 1982, Molecular Mechanics, ACS Monograph 177, American Chemical Society, Washington, DC, 339 pp. Cabell, L. A., M.D. Best, J.J. Lavigne, S.E. Schneider, D.M. Perreault, M.-K. Monahan and E.V. Anslyn, 2001, J. Chem. Soc. Perkin Trans. 2, 315. Chapoteau, E., B.P. Czech, W. Zazulak and A. Kumar, 1993, Clin. Chem. 39, 1820. Chawla, H.M. and K. Srinivas, 1994, J. Chem. Soc. Chem. Commun. 2593. Chen, C.-T., H. Wagner and W.C. Still, 1998, Science 279, 851. Cloninger, M. J. and H.W. Whitlock, 1998, J. Org. Chem. 63, 6153. Cooper, C. R. and T.D. James, 1997, J. Chem. Soc. Chem. Commun. 1419. Cooper, C. R. and T.D. James, 1998, Chem. Lett., 883. Cram, D. J. and J.M. Cram, 1978, Acc. Chem. Res. 11, 8. Cram, D. J., 1986, Angew. Chem. Int. Ed. Engl., 25, 1039. Cram, D. J., 1988, Science, 240, 760. Czarnik, A. W., Ed., 1993a, Fluorescent Chemosensors for Ion and Molecule Recognition, ACS Symposium Series 538, American Chemical Society, Washington, DC. Czarnik, A. W., 1993b, In Advances in Supramolecular Chemistry, Vol. 3, Ed., G.W. Gokel, JAI Press, Greenwich, Connecticut, p. 131. Czarnik, A. W., 1994, Acc. Chem. Res., 27, 302. Czarnik, A. W. and J. Yoon, 1999, In Supramolecular Technology, Chap. 4, Ed., D.N. Reinhoudt, Wiley Sons, Chichester. Davis, A. P. and R.S. Wareham, 1999, Angew. Chem. Int. Ed. 38, 2978. de J ong, M. R., J.F.J. Engbersen, J. Huskens and D.N. Reinhoudt, 2000, Chem. Eur. J. 6, 4034. De Santis, G., L. Fabbrizzi, M. Licchelli, A. Poggi and A. Taglietti, 1996, Angew. Chem. Int. Ed. Engl. 35, 202. de Silva, A. P., H.Q.N. Gunaratne, C. McVeigh, G.E.M. Maguire, P.R.S. Maxwell and E. O'Hanlon, 1996, J. Chem. Soc. Chem. Commun. 2191. de Silva, A. P., H.Q.N. Gunaratne, T. Funnlaugsson, A.J.M. Huxley, C.P. McCoy, J.T. Rademacher and T.E. Rice, 1997, Chem. Rev. 97, 1515. Desvergne, J. P. and A.W. Czamik, Eds., 1997, Chemosensors of Ion and Molecule Recognition, NATO ASI Series C: Mathematical and Physical Sciences, Vol. 492, Kluwer, Dordrecht. Deviprasad, G. R. and F. D'Souza, 2000, Chem. Commun. 1915. Deviprasad, G. R., B. Keshavan and F. D'Souza, 1998, J. Chem. Soc. Perkin Trans. 1, 3133. Dickinson, T. A., J. White, J.S. Kauer and D.R. Walt, 1996, Nature, 382, 697. Dolman, M., A.J. Mason, K.R.A.S. Sandanayake, A. Sheridan, A.F. Sholl and I.O. Sutherland, 1996, Analyst 121, 1775. 363

Bell and Hext D'Souza, F., 1996, J. Am. Chem. Soc. 118, 923. D'Souza, F. and G.R. Deviprasad, 2001, J. Org. Chem. 66, 4601. D'Souza, F., G.R. Deviprasad and Y.-Y. Hsieh, 1997, J. Chem. Soc. Chem. Commun. 533. Eggert, H., J. Frederiksen, C. Morin and J.C. Norrild, 1999, J. Org. Chem. 64, 3846. Fabbrizzi, L. and A. Poggi, 1995, Chem. Soc. Rev. 24, 197. Fabbrizzi, L., G. Francese, M. Licchelli, P. Pallavicini, A. Perotti, A. Poggi, D. Sacchi and A. Taglietti, 1997a, In Chemosensors of Ion and Molecule Recognition, Eds., J.P. Desvergne and A.W. Czarnik, Kluwer, The Netherlands, p. 75. Fabbrizzi, L., G. Francese, M. Licchelli, A. Perotti and A. Taglietti, 1997b, Chem. Commun. 581. Fabbrizzi, L., I. Faravelli, G. Francese, M. Licchelli, A. Perotti and A. Taglietti, 1998a, Chem. Commun. 971. Fabbrizzi, L., M. Licchelli, L. Parodi, A. Poggi and A. Taglietti, 1998b, J. Fluoresc. 8, 263. Fabbrizzi, L., M. Licchelli, L. Parodi, A. Poggi and A. Taglietti, 1999, Eur. J. Inorg. Chem. 35. Fabbrizzi, L., M. Licchelli, G. Rabaioli and A. Taglietti, 2000, Coord. Chem. Rev. 205, 85. Geddes, C. D., 2000, J. Photochem. Photobiol. A Chem. 137, 145. Geddes, C. D., K. Apperson, J. Karolin and D.J.S. Birch, 2001, Anal. Biochem. 293, 60. Gokel, G. W., 1991, Crown Ethers and Cryptands, The Royal Society of Chemistry, Cambridge. Goodman, M. S., A.D. Hamilton and J. Weiss, 1995, J. Am. Chem. Soc. 117, 8447. Grady, T., S.J. Harris, M.R. Smyth and D. Diamond, 1996, Anal. Chem. 68, 3775. Grady, T., T. Joyce, M.R. Smyth, S.J. Harris and D. Diamond, 1998, Anal. Commun. 35, 123. Granda-Vald6s, M., R. Badfa, G. Pina-Luis, M.E. Dfaz-Garcfa, 2000, Qufmica Analftica 19 [Suppl. 1], 38. Hancock, R. D. and A.E. Martell, 1988, Comments Inorg. Chem. 6, 237. Hayashita, T., S. Taniguchi, Y. Tanamura, T. Uchida, S. Nishizawa, N. Teramae, Y.S. Jin, J.C. Lee and R.A. Bartsch, 2000, J. Chem. Soc. Perkin Trans. 2, 1003. Helgeson, R. C., B.P. Czech, E. Chapoteau, C.R. Gebauer, A. Kumar and D.J. Cram, 1989, J. Am. Chem. Soc. 111, 6339. Hennrich, G., H. Sonnenschein and U. Resch-Genger, 2001, Tetrahedron Lett. 42, 2805. Huber, C., T. Werner, C. Krause, I. Klimant, O.S. Wolfbeis, 1998, Anal. Chim. Acta 364, 143. Hulanicki, A., S. Glab and F. Ingman, 1991, Pure Appl. Chem. 63, 1247. 364

Artificial Receptors for Chemosensors Hunter, C. A., 1993, Angew. Chem. Int. Ed. Engl. 32, 1584. Hunter, C. A. and J.K.M. Sanders, 1990, J. Am. Chem. Soc. 112, 5525. Iorio, E. J. and W.C. Still, 1999, Bioorg. Med. Chem. Lett. 9, 2145. James, T. D., K.R.A.S. Sandanayake and S. Shinkai, 1994, Angew. Chem. Int. Ed. Engl. 33, 2207. James, T. D., K.R.A.S. Sandanayake and S. Shinkai, 1996a, Angew. Chem. Int. Ed. 35, 1910. James, T. D., P. Linnane and S. Shinkai, 1996b, J. Chem. Soc. Chem. Commun. 281. James, T. D., H. Shinmori and S. Shinkai, 1997, J. Chem. Soc. Chem. Commun. 71. Kato, R., S. Nishizawa, T. Hayashita and N. Teramae, 2001, Tetrahedron Lett. 42, 5053. Kijima, H., M. Takeuchi and S. Shinkai, 1998, Chem. Lett. 781. Kimura, E. and T. Koike, 1998a, Chem Soc. Rev. 27, 179. Kimura, E. and T. Koike, 1998b, Chem. Commun. 1495. Kr~il, V., O. Rusin and F.P. Schmidtchen, 2001, Org. Lett. 3, 873. Krause, C., T. Werner, C. Huber, O.S. Wolfbeis and M.J.P. Leiner, 1999, Anal. Chem. 71, 1544. Kubo, Y., 1998, J. Inclusion Phenom. Mol. Recognit. Chem. 32, 235. Kubo, Y., 1999, Synlett 161. Kubo, Y., S. Maruyama, N. Ohhara, M. Nakamura and S. Tokita, 1995, J. Chem. Soc. Chem. Commun. 1727. Kubo, Y., S. Maeda, S. Tokita and M. Kubo, 1996, Nature 382, 522. Kubo, Y., S. Maeda, S. Tokita and M. Kubo, 1997, Enantiomer 2, 287. Kubo, Y., N. Hirota, S. Maeda and S. Tokita, 1998, Anal. Sci. 14, 183. Kubo, Y., M. Tsukahara, S. Ishihara and S. Tokita, 2000, J. Chem. Soc. Chem. Commun. 653. Kumar, A., E. Chapoteau, B.P. Czech, C.R. Gebauer, M.Z. Chimenti, O. Raimondo, 1988, Clin. Chem. 34, 1709. Laatikainen, R., J. Ratilainen, R. Sebastian and H. Santa, 1995, J. Am. Chem. Soc. 117, 11006. Lakowicz, J. R., 1999, Principles of Fluorescence Spectroscopy, 2nd ed., Kluwer Academic / Plenum, New York. Lavigne, J. J. and E.V. Anslyn, 2001, Angew. Chem. Int. Ed. 40, 3118. Lee, D. H., K.H. Lee and J.-I. Hong, 2001a, Org Lett. 3, 5. Lee, K. H. and J.-I. Hong, 2000, Tetrahedron Lett. 41, 6083. Lee, K. H., H.-Y. Lee, D.H. Lee and J.-I. Hong, 2001b, Tetrahedron Lett. 42, 5447. Lehn, J.-M., 1995, Supramolecular Chemistry: Concepts and Perspectives, VCH, Weinheim. Lehn, J.-M., 1988, Angew. Chem. Int. Ed. Engl. 27, 89. Lehn, J.-M., 1990, Angew. Chem. Int. Ed. Engl. 29, 1304. Leipert, D., J. Mack, R. Ttinnemann and G. Jung, 1999, Comb. Chem. 335. 365

Bell and Hext Lewis, P. T., C.J. Davis, L.A. Cabell, M. He, M.W. Read, M.E. McCarroll and R.M. Strongin, 2000, Org. Lett. 2, 589. Linnane, P., T.D. James, S. Imazu and S. Shinkai, 1995, Tetrahedron Lett. 36, 8833. L/3hr, H.-G. and F. V/3gtle, 1985, Acc. Chem. Res. 18, 65. Lorand, J. P. and J.O. Edwards, 1959, J. Org. Chem. 24, 769. L/3wik, D. W. P. M., M.D. Weingarten, M. Broekema, A.J. Brouwer, W.C. Still and R.M.J. Liskamp, 1998, Angew. Chem. Int. Ed. 37, 1846. Lustenberger, P., R. Welti and F. Diederich, 1998, Helv. Chim. Acta 81, 2190. Ma, J. C. and D.A. Dougherty, 1997, Chem. Rev. 97, 1303. McFarland, S. A. and N.S. Finney, 2001, J. Am. Chem. Soc. 123, 1260. Mei, M. and S. Wu, 2001, New J. Chem. 25, 471. Mello, J. V. and N.S. Finney, 2001, Angew. Chem. Int. Ed. 40, 1536. Metzger, A. and E.V. Anslyn, 1998, Angew. Chem. Int. Ed. 37, 649. Miyaji, H. and J.L. Sessler, 2001 Angew. Chem. Int. Ed. 40, 154. Miyaji, H., P. Anzenbacher, Jr., J.L. Sessler, E.R. Bleasdale and P.A. Gale, 1999, Chem. Commun. 1723. Mizuno, T., M. Takeuchi and S. Shinkai, 1999, Tetrahedron 55, 9455. Mizuno, T., M. Yamamoto, M. Takeuchi and S. Shinkai, 2000, Tetrahedron 56, 6193. Mohr, G. J. and O.S. Wolfbeis, 1995, Anal. Chim. Acta 316, 239-246. Mohr, G. J. and O.S. Wolfbeis, 1996a, Sens. Actuators B 37, 103. Mohr, G. J. and O.S. Wolfbeis, 1996b, Analyst 121, 1489. Mohr, G. J. and U.E. Spichiger-Keller, 1997, Anal. Chim. Acta 351, 189. Mohr, G. J., D. Citterio and U.E. Spichiger-Keller, 1998a, Sens. Actuators B 49, 226. Mohr, G. J., C. Demuth and U.E. Spichiger-Keller, 1998b, Anal. Chem. 70, 3868. Mohr, G. J., I. Murkovic, F. Lehmann, C. Haider and O.S. Wolfbeis, 1997, Sens. Actuators B 38-39, 239. Mohr, G. J., U.E. Spichiger-Keller, W. Jona and H. Langhals, 2000, Anal. Chem. 72, 1084. Murkovic, I. and O.S. Wolfbeis, 1997, Sens. Actuators B 38-39, 246. Narita, M., S. Mima, N. Ogawa and F. Hamada, 2001, Anal. Sci. 17, 379. Niikura, K. and E.V. Anslyn, 1999, J. Chem. Soc. Perkin Trans. 2, 2769. Nishizawa, S., R. Kato, T. Hayashita and N. Teramae, 1998a, Anal. Sci. 14, 595. Nishizawa, S., H. Kaneda, T. Uchida and N. Teramae, 1998b, J. Chem. Soc. Perkin Trans. 2, 2325. Nishizawa, S., Y. Kato and N. Teramae, 1999, J. Am. Chem. Soc. 121, 9463. Padilla-Tosta, M. E., J.M. Lloris, R. Martfnez-M~ifiez, T. Pardo, J. Soto, A. Benito and M.D. Marcos, 2000, Inorg. Chem. Commun. 3, 45. Padilla-Tosta, M. E., J.M. Lloris, R. Martfnez-M~ifiez, T. Pardo, F. Sancen6n, J. Soto and M.D. Marcos, 2001, Eur. J. Inorg. Chem. 1221. Parker, D., K. Senanayake and J.A.G. Williams, 1997, Chem. Commun. 1777. Pedersen, C. J., 1984, Science 241,536. 366

Artificial Receptors for Chemosensors Pina, F., M.A. Bemardo and E. Garcia-Espafia, 2000, Eur. J. Inorg. Chem. 2143. Prodi, L., F. Bolletta, M. Montalti and N. Zaccheroni, 2000, Coord. Chem. Rev. 2O5, 59. Raker, J. and T.E. Glass, 2001, J. Org. Chem. 66, 6505. Rakow, N. A. and K.S. Suslick, 2000, Nature 406, 710. Rebek, J., Jr., 1988, In Topics in Current Chemistry, Vol. 149, Ed., E. Weber, Springer-Verlag, Berlin, p. 189. Rebek, J., Jr., 1990, Acc. Chem. Res. 23, 399. Reichardt, C., 1979, In Monographs in Modem Chemistry, Vol. 3,Verlag Chemie, Weinheim, Chap. 6. Reinhoudt, D. N. and P.J. Dijkstra, 1988, Pure Appl. Chem. 60, 477. Rettig, W., 1994, In Topics in Current Chemistry, vol. 169, Ed., J. Mattay, Springer-Verlag, Berlin, p. 253. Rothman, J. H. and W.C. Still, 1999, Bioorg. Med. Chem. Lett. 9, 509. Rusin, O. and V. Kr~il, 1999, J. Chem. Soc. Chem. Commun. 2367. Ryan, K. and W.C. Still, 1999, Bioorg. Med. Chem. Lett. 9, 2673. Ryan, K., L.J. Gershell and W.C. Still, 2000, Tetrahedron 56, 3309. Sauvage, J.-P., Ed., 2001, Molecular Machines and Motors, Springer-Verlag, Berlin. Scheerder, J., J.F.J. Engbersen and D.N. Reinhoudt, 1996, Rec. Trav. Chim. Pays-Bas 115,307. Schmidtchen, F. P. and M. Berger, 1997, Chem. Rev. 97, 1609. Schneider, H.-J., 1991, Angew. Chem. Int. Ed. Engl. 30, 1417. Schneider, H.-J. and A. Yatsimirsky, 2000, Principles and Methods in Supramolecular Chemistry, John Wiley Sons, Chichester. Schneider, S. E., S.N. O'Neil and E.V. Anslyn, 2000, J. Am. Chem. Soc. 122, 542. Sen, A. and K.S. Suslick, 2000, J. Am. Chem. Soc. 122, 11565. Seneci, P., 2000, Solid-Phase Synthesis and Combinatorial Technologies, WileyInterscience, New York, 637 pp. Sessler, J. L., N.A. Tvermoes, J. Davis, P. Anzenbacher, Jr., K. Jursl'kov~i, W. Sato, D. Seidel, V. Lynch, C.B. Black, A. Try, B. Andrioletti, G. Hemmi, T.D. Mody, D.J. Magda and V. Kr~il, 1999, Pure Appl. Chem. 71, 2009. Shinkai, S. and M. Takeuchi, 1996, Trends Anal. Chem. 15,418. Shinkai, S., M. Takeuchi and A. Ikeda, 2000, In Polymer Sensors and Actuators, Eds. Osada, DeRossi and Danilo, Springer, Berlin, Chap. 6. Spichiger-Keller, U.E., 1997, Sens. Actuators B 38-39, 68. Springsteen, G. and B. Wang, 2001, Chem. Commun. 1608. Steed, J. W. and J.L. Atwood, 2000, Supramolecular Chemistry, John Wiley Sons, Chichester. Subrahmanyam, S., S.A. Piletsky, E.V. Piletska, B. Chen, R. Day and A.P.F. Turner, 2000, Adv. Mater. 12, 722. Szemes, F., D. Hesek, Z. Chert, S.W. Dent, M.G.B. Drew, A.J. Goulden, A.R. Graydon, A. Grieve, R.J. Mortimer, T. Wear, J.S. Weightman and P.D. Beer, 1996, Inorg. Chem. 35, 5868. 367

Bell and Hext Takagi, M. and K. Ueno, 1984, In Topics in Current Chemistry, Vol. 121: Host Guest Complex Chemistry m, Eds., F. Voegtle and E. Weber, SpringerVerlag, Berlin, p. 39. Takeuchi, M., T. Imada and S. Shinkai, 1996a, J. Am. Chem. Soc. 118, 10658. Takeuchi, M., M. Taguchi, H. Shinmori and S. Shinkai, 1996b, Bull. Chem. Soc. Jpn. 69, 2613. Takeuchi, M., S. Yoda, T. Imada and S. Shinkai, 1997a, Tetrahedron 53, 8335. Takeuchi, M., M. Yamamoto and S. Shinkai, 1997b, J. Chem. Soc. Chem. Commun. 1731. Takeuchi, M., T. Imada and S. Shinkai, 1998, Bull. Chem. Soc. Jpn. 71, 1117. Tsien, R. Y., 1993, In Fluorescent Chemosensors for Ion and Molecule Recognition, Ed., A . W . Czamik, American Chemical Society, Washington, DC, Chap. 9. Ueno, A., 1993, In Fluorescent Chemosensors for Ion and Molecule Recognition, Ed., A.W. Czarnik, American Chemical Society, Washington, DC, Chap. 6. Vt~gtle, F., 1991, Supramolecular Chemistry: An Introduction, John Wiley Sons, Chichester. Wang, J. and A. Ueno, 2000, Macromol. Rapid Commun. 21,887. Wiskur, S. L. and E.V. Anslyn, 2001, J. Am. Chem. Soc. 123, 10109. Wulff, G., 1982, Pure Appl. Chem. 54, 2093. Xie, H., S. Yi, X. Yang and S. Wu, 1999, New J. Chem. 23, 1105. Yamamoto, M., M. Takeuchi and S. Shinkai, 1998, Tetrahedron 54, 3125. Yamauchi, A. and T. Hayashita, 2000, Bunsen Kagaku 49, 75 (in Japanese). Yang, J.-S. and T.M. Swager, 1998, J. Am. Chem. Soc. 120, 5321. Yoshida, H., K. Saigo and K. Hiratani, 2000, Chem. Lett., 116. Zalewski, P. D., I.J. Forbes, R.F. Seamark, R. Borlinghaus, W.H. Betts, S.F. Lincoln and A.D. Ward, 1994, Chem. Biol. 1,153.

368

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 12

NUCLEIC ACIDS FOR REAGENTLESS BIOSENSORS

MANJULA RAJENDRAN AND ANDREW D. ELLINGTON, PH.D.

Department of Chemistry and Biochemistry University of Texas at Austin, Austin, TX 78712 USA

In vitro selection has yielded a range of nucleic acid binding species (aptamers) and catalysts (ribozymes) whose ligand-binding affinities and activation parameters rival those of proteins. Precisely because functional nucleic acids can be engineered based largely on an appreciation of their secondary structures and chemically synthesized in bulk, it has proven remarkably easy to incorporate aptamers and ribozymes into reagentless biosensors that directly transduce ligand recognition to optical signals. Aptamers undergo conformational changes upon interaction with their cognate ligands, and, by appending fluorophores to aptamers, it has proven possible to generate 'signaling aptamers.' Ribozymes can act on fluorescent substrates to generate fluorescent signals; any molecules or processes that affect ribozyme catalysis can therefore be reported. Aptamers can also be appended to ribozymes to generate aptazymes, or effector-activated ribozymes, that transduce molecular recognition to ribozyme catalysis. Aptazymes have proven remarkably plastic and can be activated by metals, small organic molecules,~peptides and proteins. By simply cleaving apart or adjoining fluorophores and quenchers on oligonucleotide substrates, one can potentially construct reagentless ribozyme or aptazyme chips that could simultaneously report the concentration of multiple different analytes. Prototypes of such chips have now been made. While reagentless nucleic acid biosensors may ultimately prove less sensitive or robust than reagentless protein biosensors, it is nonetheless likely that nucleic acid biosensors will prove much more amenable to generation by high-throughput selection methods, and thus may be the best vehicle for developing chips that can acquire organismal proteomes and metabolomes.

369

Rajendran and Ellington 1. The Concept of a Reagentless Biosensor Reagentless biosensors are sensors which can detect a target analyte in a homogenous format in solution; that is, without the addition of reagents other than the sample. A reagentless biosensor has several advantages over conventional diagnostic assays. The first and most obvious advantage of a reagentless biosensor is its simplicity and ease of use. Reagentless biosensors should prove to be intrinsically practical in a variety of settings, from the detection of hazardous biological agents in the field to monitoring blood glucose levels in a hospital setting to providing real-time readouts of cellular states in a research lab. Second, since the detection process does not require the addition of reagents, the system remains unperturbed throughout the course of the assay. This is of particular importance for the in vivo detection and quantitation of analytes. Third, because reagentless biosensors directly transduce the sensitivity and specificity of biomolecular detection to the production of a signal, they can potentially be integrated with any number of detection platforms. Finally, as a consequence of their ease of use, utility, and adaptability, it is expected that reagentless biosensors may prove to be significantly more cost-effective than conventional diagnostic assays. As alluded to above, reagentless biosensors can be constructed by integrating a signaling or reporter component with a biological macromolecule, allowing the biomolecule to directly transduce the molecular recognition event into a detectable signal (e.g., an optical or electrochemical signal). The simplest reagentless optical biosensors can be envisioned as fluorescentlylabeled macromolecules. For example, many proteins undergo ligand-induced conformational changes. Fluorophores incorporated at specific sites on such molecules can be used to sensitively transduce molecular recognition events; as an example, the maltose binding protein has been derivatized with fluorescent dyes, and the hinge-bending motion that occurs upon interaction with maltose can be used to quantitate maltose levels in solution (Hellinga and Marvin, 1998; Marvin et al., 1997). The reason such sensors work is that fluorophores are frequently sensitive to minute changes in their chemical environments, and such changes are reported as changes in fluorescence intensity, wavelength, or anisotropy. Alternately, changes in fluorescence-resonance energy transfer between pairs of fluorophores, or between fluorophores and quenchers, can be used to detect binding. Similarly, reagentless electrochemical biosensors have been made by electrostatic self-deposition of redox polyelectrolyte mediators and enzymes (Leech and Daigle, 1998; Mulchandani and Pan, 1999; Narvaez et al., 2000). The enzymes turnover analytes (substrates) and produce electrons that are in turn coupled to electrodes via the adjacent mediators. Such integrated molecular 370

Nucleic

Acids for Reagentless Biosensors

sensors are not only inherently more practical, but have enhanced sensitivity and speed of response.

2. H i s t o r y - In Vitro Selection

Aptamers are single stranded nucleic acids (RNA, ssDNA, modified RNA or ssDNA), capable of binding tightly and specifically to their targets. They are isolated from combinatorial oligonucleotide libraries by a process known as in vitro selection. In vitro selection mimics the process of natural evolution inthat a pool of nucleic acids are sieved for a desired functional property, such as the ability to bind to a target or catalyze a reaction. Once functional species have been isolated, they are preferentially amplified via conventional molecular biology techniques, such as reverse transcription, polymerase chain reaction, and in vitro transcription. Over multiple rounds of selection and amplification, quite large populations (> 1013 different sequences) can be sieved and those few, "fittest" nucleic acid species isolated. The in vitro evolution of nucleic acids was first reported by Sol Spiegelman in the 1960's. Spiegelman and his co-workers established a cell-free system in which the genomic RNA of bacteriophage Q/3 could be evolved. For example, in a serial dilution experiment using purified Qfl replicase the genomic RNA was evolved to replicate more quickly (Mills et al., 1967; Spiegelman, 1971). While other phenotypes, such as resistance to ethidium bromide (Spiegelman , 1971) were also evolved, these in vitro selection methods may have ultimately been limited in scope because the diversity of the RNA population was dictated solely by the error rate of the polymerase. More practical approaches to in vitro selection became possible in the 1990's with the advent of solid phase DNA synthesis, and the invention of the polymerase chain reaction (PCR). Using solid phase DNA synthesis, synthetic oligonucleotides with randomized regions could be generated and large nucleic acid libraries (up to 10~5 species) that contained extremely diverse sequences could be obtained. In 1989, Kevin Struhl and co-workers reported the identification of double-stranded DNA binding species from a random library (Oliphant et al., 1989). In 1990, two separate groups reported the in vitro selection of RNA binding species. Tuerk and Gold (1990) started with a library derived from the natural RNA substrate of T4 DNA polymerase and returned both wild-type and non-wild-type winners. Ellington and Szostak (1990) evolved nucleic acid ligands for targets with no previously known nucleic acid affinity, starting from a library with 100 randomized positions. The first selection of a ribozyme from a randomized population was reported by the Joyce group in the same year (Robertson and Joyce, 1990); they evolved a natural ribozyme, the Tetrahymena self-splicing intron, to carry out a novel DNA cleavage reaction. 371

Rajendran and Ellington

Figure 1. Scheme for in vitro selection. A chemically synthesized, single-stranded DNA pool is PCR amplified and transcribed in vitro to generate an RNA pool. Target molecules are incubated with the pool. Those sequences and shapes that can interact with the target are sieved from the population by one of a variety of methods (affinity chromatography, filter capture). The captured sequences are eluted, amplified, and subjected to additional rounds of selection and amplification. 372

Nucleic Acids for Reagentless Biose.nsors

-[I

......! !!

! I i i 1~

-.-/ Molecular Beacon

Target

Hybrid

Figure 2. Molecular Beacon. A stem-loop structure contains a fluorophore (red) and a quencher (blue). Sequence-specific interactions between the loop and a target nucleic acid results in the formation of an extended helix that pries apart the fluorophore and quencher, resulting in a target-specific increase in fluorescent signal. Contemporary in vitro selection experiments begin with the chemical synthesis of single-stranded DNA libraries. The inclusion of constant sequence regions flanking the random sequence core allows the amplification of such libraries via the polymerase chain reaction, and the conversion of libraries to different chemical forms, for example the conversion of a double-stranded DNA template to single-stranded RNA via in vitro transcription. DNA or RNA libraries can be sieved for a variety of functions, the simplest of which is binding. By passing an RNA library over an affinity column or by collecting protein:RNA complexes via filtration on modified cellulose filters, binding species can be sieved from nonbinding species. Eluted binding species can then be amplified by a combination of reverse transcription, PCR, and in vitro transcription. Multiple cycles of selection and amplification generally result in the purification of those binding species (also known as 'aptamers') that have the highest affinity and specificity for a given target. Aptamers possess many properties, which make them potential candidates for biosensor applications (for recent reviews, see Brody and Gold, 2000; Famulok et al., 2000; Hesselberth et al., 2000a; Jayasena, 1999; Wilson and Szostak, 1999). They are the only reagents that rival antibodies in their universal molecular recognition properties and have been selected against an amazingly wide range of targets, ranging from metal ions (Ciesiolka and Yams, 1996; Kawakami et al., 2000) to complex cellular structures such as the ribosome (Ringquist et al., 1995) and even to whole cells (Morris et al., 1998). In fact, aptamers have been identified which bind to viral particles (Pan et al., 1995) and live pathogenic protozoa (Homann and Goringer, 1999). Also, since aptamers are selected in vitro, they can potentially be raised against pathogens, toxins (e.g., ricin, Hesselberth et al., 2000b), and purportedly to biological warfare agents, targets which frequently prove problematic for in vivo immunization procedures. In addition to their comprehensive binding properties, aptamers have high binding affinities and also remarkable specificities. The binding affinities of aptamers are typically in the nanomolar to picomolar range for protein targets, and in the micromolar to nanomolar range for small organic targets. Aptamers 373

Rajendran and Ellington can discriminate between targets on the basis of subtle differences such as single amino acid changes in protein targets (Conrad et al., 1994; Hirao et al., 1998), or the presence or absence of a methyl (Hailer and Samow, 1997; Jenison et al., 1994) or a hydroxyl group (Mannironi et al., 1997; Sassanfar and Szostak, 1993) in small organic targets. Finally, since aptamers can frequently be minimized to relatively small (30-50 nucleotide) oligonucleotides, they can be chemically synthesized in bulk and modified during chemical synthesis for conjugation or sensor function. The primary limitation on the use of aptamers as recognition and/or transduction elements in biosensors has been the perception that they are unstable and highly susceptible to degradation in biological media. However, the incorporation of modified nucleotides either pre- or post-selection can protect aptamers from nuclease degradation (Eaton and Pieken, 1995; Green et al., 1995; Jellinek et al., 1995; Lin et al., 1994; Pagratis et al., 1997), and the conjugation of aptamers to supermolecular carriers such as PEG or liposomes (Tucker et al., 1999; Willis et al., 1998) can greatly increase their stability and retention in biological fluids. Nucleic acid pools can also be sieved for catalytic function (for a review, see Wilson and Szostak, 1999). For example, nucleic acid pools can be immobilized on columns, and catalytically active species will cleave themselves from the column and can be collected in the eluate (Breaker and Joyce, 1994). Conversely, ribozyme ligases can be selected from random sequence pools based on their ability to append particular sequences to themselves that can be captured on affinity columns and used as primer-binding sites for PCR amplification (Bartel and Szostak, 1993). A variety of other ribozymes have been selected based on variations on these two themes" cleave away or add to. Alkyl transferase ribozymes have been selected that can add an activated biotin to themselves (Wilson and Szostak, 1995), tRNA synthetase-like ribozymes have been selected that can aminoacylate themselves (Illangasekare et al., 1995), and amide synthases have been selected by selectively modifying the 5' end of a pool with an amine and then identifying those ribozymes that can conjugate an activated carboxylate to themselves (Lohse and Szostak, 1996). It has also proven possible to meld aptamer and catalytic selections. Just as catalytic antibodies can be selected by identifying antibody variants that bind transition state analogues of a given reaction, aptamers that bind transition state analogues have proven to have catalytic activity. Peter Schultz and coworkers initially isolated ribozymes that could catalyze the isomerisation of a biphenyl compound to its diastereomer by using a transition state analogue as a target (Prudent et al., 1994). Similarly, using N-alkylated porphyrin transition state analogues, both catalytic RNA (Conn et al., 1996) and catalytic DNA (Li and Sen, 1996) have been selected that catalyze porphyrin metalation. More recently, an aptamer having cholesterol esterase activity was isolated by in vitro selection of RNA using a phosphate ester transition-state analogue of cholesterol ester hydrolysis as a target (Chun et al., 1999). 374

Nucleic Acids for Reagentless Biosensors 3. State of the Art 3.1. Signaling aptamers

Nucleic acid biosensors that can directly transduce the molecular recognition of other nucleic acids into optical signals have previously been described. K_ramer and his co-workers originally designed 'molecular beacons' (Figure 2) that juxtaposed fluorophores and quenchers in a stem structure (Tyagi et al., 1998; Tyagi and Kramer, 1996). A loop capping the stem was complementary to some RNA or DNA target; upon interaction with the target, the formation of a new, stable helical structure resulted in the original stem being pried apart, which in turn freed the fluorophore from the adjacent quencher, resulting in a strong optical signal. Molecular beacons have now become mainstays in the diagnostics industry, and have been adapted to a variety of applications, including allele discrimination in real-time PCR assays of genomic DNA (Tyagi et al., 1998), detection of target genes (for example, detection of drug resistance in Mycobacterium tuberculosis (Piatek et al., 1998)), as sensitive DNA biosensors (Liu et al., 2000; Liu and Tan, 1999), for studying protein-DNA interactions (Bar-Ziv and Libchaber, 2001; Fang et al., 2000; Tan et al., 2000), and in antisense research for the real time detection of DNA-RNA hybridization in living cells (Sokol et al., 1998). While sequence recognition alone can potentiate a wide range of diagnostic and other applications, the utility of nucleic acids as reagentless biosensors would be greatly expanded if they could also signal the presence of non-nucleic acid analytes, such as proteins or small organics. To this end, there are several different schemes that can be imagined for converting aptamers to biosensors or 'signaling aptamers' (Figure 3). Each of these models has been realized in practice, and will be described in turn. Just like the reagentless protein biosensors described above, aptamers can be adapted to signal the presence of non-nucleic acid analytes. Structural studies have shown that aptamers frequently undergo small but significant conformational changes or reorganizations upon binding their cognate ligands (Hermann and Patel, 2000; Patel and Suri, 2000). By incorporating a fluorophore into a conformationally labile region of an aptamer, the binding event can lead to a change in the chemical environment of the fluorophore and hence to a change in fluorescence intensity (Figure 3a). Based on the known three-dimensional structures of anti-adenosine RNA and DNA aptamers (Dieckmann et al., 1996; Jiang et al., 1996; Lin and Patel, 1997), fluorescent dyes were introduced in the proximity of the adenosine binding site. The resultant signaling aptamers not

375

Rajendran and Ellington

Figure 3. Strategies for signaling aptamers. (a) Exploitation of small conformational changes. (b) Exploitation of larger secondary structural conformational changes. (c) Exploitation of tertiary structural conformational changes. (d) Exploitation of quartemary structural conformational changes. In the case of (a), the conformational change may be inherent to the aptamer. In (b) - (d), the conformational changes have been engineered into the aptamer by altering its secondary, tertiary, and / or quarternary structure.

only showed an ATP-dependent increase in fluorescence, they were also selective for ATP relative to other nucleotides and could track ATP concentrations in solution (Jhaveri et al., 2000a). Interestingly, while the sites of fluorophore insertion were chosen to interfere with ATP binding as little as possible, the apparent Kd of the designed signaling aptamers (-30 lxM for the DNA signaling aptamer a n d - 300 ~M for the RNA signaling aptamer) was much higher than that of the parental aptamers (Huizenga and Szostak, 1995; Sassanfar and Szostak, 1993) (~6 ~M for the anti-adenosine DNA aptamer and 6-8 ~tM for the anti-adenosine RNA aptamer). This may indicate that the design process must be greatly refined, or it may be that there is an inherent loss of binding affinity during conformational transduction. While the conformational changes that the anti-adenosine aptamers underwent were relatively small, much larger conformational changes are possible. For example, the ligand-dependent organization of aptamer secondary structure can be envisaged. A secondary structure could be poised so that it would be largely unstructured in the absence of analyte, but substantially stabilized upon analyte binding (Figure 3b). An anti-cocaine DNA aptamer has been converted into a signaling aptamer using this strategy (Stojanovic et al., 2001). One of the stems 376

Nucleic Acids for Reagentless Biosensors of a three-way junction that constituted the cocaine binding region was destabilized by truncation. At the same time, the stem was labeled with a 5' fluorophore and a 3' quencher. Ligand binding stabilizes the engineered stem and the aptamer goes from a fluorescent, unliganded form to a quenched, ligandbound form. The signaling aptamer was not only able to measure cocaine concentration in the concentration range from 10 r to 2.5 mM, but was robust enough to report cocaine concentrations in serum. The aptamer appeared selective for cocaine relative to cocaine derivatives, such as benzoyl ecgonine, but it was possible that the observed selectivity was for hydrophobicity rather than for a defined chemical structure. Signaling aptamers that rely on tertiary structural transitions can also be engineered. In this instance, an extant secondary structural element is not merely destabilized, but instead an entirely new conformation is pre-engineered into the aptamer (Figure 3c). A short, anti-thrombin aptamer that was known to form a quadruplex structure (Bock et al., 1992; Macaya et al., 1993; Schultze et al., 1994) served as the starting point for the design of a simple 'aptamer beacon.' Sequences were added to the 5' terminus of the anti-thrombin aptamer that were complementary to critical residues within the thrombin-binding structure (Hamaguchi et al., 2001). In consequence, the dominant structure in solution was not a quadruplex, but rather was a hairpin stem. A fluor was added to the 5' end of the hairpin, while a quencher was added to the 3' end. The addition of thrombin shifts the equilibrium from the quenched stem-loop form to the thrombin-bound form, resulting in dequenching of the fluorophore and the creation of an optical signal. The thrombin 'beacon' could detect thrombin concentrations as low as 5nM and was able to discriminate against other serine proteases, such as factors IX and Xa. An advantage of this tertiary structural rearrangement strategy as opposed to the secondary structural rearrangement strategy related above is that the signal is a dequenched increase in fluorescence, rather than a quenched decrease in fluorescence. While there are numerous analytes in a complex mixture or biological sample that might lead inadvertently to fluorescence quenching, there should be relatively few compounds other than the target analyte that should lead to an increase in fluorescence intensity. Finally, aptamer quaternary structure can also be engineered to yield analytedependent changes in optical signals (Figure 3d). In this case, aptamers are split into separate pieces that can self-assemble in the presence of a cognate ligand. Again, fluorophores are used to label each of the aptamer pieces. In the absence of the target ligand, the two oligomers exist as individual units in solution, but target binding brings the oligomers together and leads to ternary complex stabilization, ultimately resulting in a quenched optical signal (or in a fluorescence resonance energy transfer signal). Anti-cocaine and anti-rATP aptamers have been converted into signaling aptamers using this strategy (Stojanovic et al., 2000). The signaling aptamers were again not only reproducibly sensitive, but also selective for their cognate ligands. Within the 377

Rajendran and Ellington concentration ranges of 10 #M to lmM ATP and 10-150 pM cocaine, it proved possible to simultaneously report the concentrations of the two analytes using the two signaling aptamers via fluorescence changes. In a similar set of experiments, an aptamer that binds the Tat protein of HIV has been converted into a signaling aptamer by this method. In this case, one of the two oligomers was designed to be a molecular beacon construct, which opens up on ligand binding to generate a signal. The obvious advantage of this method is again that it allows a 'positive' optical signal to be generated, rather than looking for a 'negative' signal against a highly fluorescent background. The anti-Tat aptamer had an extremely low Kd for Tat (---120 pM), and the adapted biosensor could quantitate Tat samples as low as 100 nM (Yamamoto et al., 2000). In all the above examples, the generation of signaling aptamers required a prior knowledge of the secondary or even tertiary structure of the aptamer; they were all 'designed' signaling aptamers. Obviously, the need to understand the detailed structure of an aptamer may limit the applicability of these methods, especially in the development of large-scale sensor arrays (see also Section 3.4). Therefore, it is important to determine whether other methods might be developed that would directly couple aptamer selection to signal transduction. In other words, can methods be devised for the direct selection of signaling aptamers? As a first attempt, we have incorporated modified, fluorescent nucleotides directly into selection experiments (Jhaveri et al., 2000b). A single stranded DNA pool was synthesized that contained largely A, G, and C, and only a small fraction of T. In vitro transcription was used to generate an RNA pool that completely incorporated a uridine analogue, fluorescein UTP, at all positions that would have normally contained uridine. The random region in the pool was skewed and the uridine ratio was kept low to avoid an intrinsic background which could mask signaling by selected aptamers. Aptamers were isolated that could bind to the target analyte ATP. Several different families emerged from the selection, and all contained a relatively small number of (or even no) uridine residues. Individual families were then screened for their ability to signal the presence of ATP by a ligand-dependent change in fluorescence intensity. One family showed excellent signaling abilities; the best signaling aptamer contained only one uridine, could sense ATP concentrations as low as 25 txM, was selective for ATP relative to other nucleotides, and was stable enough in complex mixtures to quantitate ATP. The presence of signaling aptamers in the selected population suggested that fluors may be present during the selection of aptamers, rather than added later. 3.2. Nucleic acid catalysts as biosensors

Aptamers function as reagentless biosensors because their ability to signal is embedded within the receptor itself. This concept can also be carried over to catalytic nucleic acids (ribozymes) by simply embedding the ability to signal 378

Nucleic Acids for Reagentless Biosensors within the catalyst, its substrate, or some aspect of the catalytic mechanism. For example, John Burke and his co-workers have generated variants of the hairpin ribozyme that cleave an RNA substrate containing both a fluorophore and a quencher (Vitiello et al., 2000). In the presence of the ribozyme, a fluorescent signal is produced and the kinetics of the ribozyme can be readily followed. Michael Famulok and his co-workers have developed a similar signaling system for the hammerhead ribozyme (Jenne et al., 2001), and Krupp and his co-workers have followed the kinetics of the Group I self-splicing ribozyme by monitoring the ligation-mediated release of a fluorescent dye and concomitant changes in fluorescence polarization (Singh et al., 2000). In each instance, the system is homogenous and tracks ribozyme activity. However, since ribozyme activity is itself dependent upon a number of cofactors, notably metals, these reactions can also be viewed as biosensors for any of the reaction components that lead to catalysis. In this respect, it is interesting to note that the metal-dependence of ribozymes and deoxyribozymes can be altered or de n o v o engineered, seemingly at will. For example, it has long been known that yeast tRNA (Phe) undergoes site-specific cleavage in the presence of lead ions. Pan and Uhlenbeck (1992) exploited this property to select for variants that were even better 'leadzymes'. One variant looked completely unlike the original tRNA, yet still used lead hydroxide to cleave the phosphodiester bond, generating a 2', 3' cyclic phosphate, which was in turn hydrolyzed (Pan et al., 1994). The structure of this leadzyme has now been solved (Wedekind and McKay, 1999), and it appears as though lead coordinates to the 2' hydroxyl of the scissile residue, a mechanism that has previously been seen for protein enzymes. Breaker and Joyce (1994) were able to select a lead-dependent deoxyribozyme from a random sequence pool that could cleave a substrate with a single ribotide. The same technique was further generalized to the selection of deoxyribozymes that were dependent on other ions, such as magnesium, manganese, and zinc (Breaker and Joyce, 1995). Li et al. (2000) repeated Breaker and Joyce's experiments in order to better describe zinc-binding and catalytic motifs, but obtained sequences similar to those that had already been found by Breaker and Joyce. Li and Lu (2000) have taken advantage of the ability to track ribozyme kinetics using fluorescent reporters by developing a lead-sensing ribozyme biosensor. Substrate cleavage could be monitored in real-time using kinetic fluorescence spectroscopy, and there was a modest specificity for lead (80-fold relative to other divalent ions). Amazingly, the chemistry of nucleic acid catalysts can also be extended beyond simple metal-dependence. Since most catalysis is pH dependent, it is not surprising that ribozymes can be evolved to respond to different pH optima (Jayasena and Gold, 1997). Other variables that affect catalysis can also be probed by selection. Breaker and his co-workers have selected a deoxyribozyme that relies upon copper and peroxide to cleave a DNA substrate (Carmi et al., 1996). The radical cleavage induced by the 'DNAzyme' is mechanistically 379

Rajendran and Ellington identical to nucleic acid scission mediated by other radical-generating reagents, such as iron:EDTA complexes. However, the deoxyribozyme locally produces radicals and directs their attack at a restricted set of sites on the substrate. Similarly, Roth and Breaker (1998) have shown that a ribose moiety embedded within a DNA strand can be cleaved by a deoxyribozyme selected in the presence of histidine. The imidazole ring apparently functions as a general base. As a biosensor, the deoxyribozyme is surprisingly specific, eschewing a variety of histidine analogues, including such closely related compounds as D-histidine and 3-methyl-L-histidine. Ribozymes can also report on the presence of inhibitors. Most recently, Famulok and co-workers (Jenne et al., 2001) have used the fluorescence-based assay they originally developed to screen hammerhead ribozymes for novel inhibitors. Such technologies could potentially even be used to identify inhibitors of other RNA sequences that were appended to or activated by ribozymes.

3.3.

Aptazymes

Another variable that clearly impinges on catalytic mechanism is ribozyme structure. While ribozyme structures can be subtly altered by metals or pH or other conditions, more global alterations can be induced by changes in basepairing. Lizardi engineered an allosteric hammerhead ribozyme that initially folded into an inactive conformation which was in turn relieved by the addition of an oligonucleotide effector (Porta and Lizardi, 1995). This strategy is most similar to that shown in Figure 3c for signaling aptamers. However, allosteric activation was only about 10-fold. More recently, Taira and coworkers have generated novel allosteric ribozymes, called maxizymes, which form active quaternary structures following oligonucleotide recognition (Kuwabara et al., 2000a, 2000b; Warashina et al., 2000). This strategy is most similar to that shown in Figure 3d. A hammerhead ribozyme dimer was broken into two pieces such that both ends present 'arms' that can hybridize to a specific nucleic acid sequence. Once the ribozyme is brought together by hybridization of the substrate-binding 'arms' at one end to a mRNA sequence, a second mRNA or another portion of the same mRNA can hybridize to the substrate-binding 'arms' at the other end, leading to cleavage at both sites. Alternatively, a hammerhead heterodimer can be dissected such that one end of stem-loop II serves as a binding (but not cleavage) site for a particular sequence, while the other end still cleaves a desired target. In this way, the maxizyme can serve as a biosensor. It has also proven possible to construct ribozymes that are structurally responsive to effectors other than oligonucleotides. As we have seen in earlier sections, functional nucleic acids undergo conformational changes upon interactions with their cognate ligands. It therefore seemed reasonable to suppose that by appending nucleic acid aptamers to nucleic acid catalysts (ribozymes), it might 380

Nucleic Acids for Reagentless Biosensors prove possible to alter the conformation, and hence the catalytic activity, of a nucleic acid catalyst in a ligand-dependent fashion. Ron Breaker and his coworkers were the first to attempt this feat by swapping an anti-adenosine aptamer with a stem of the hammerhead ribozyme whose sequence was known to be relatively unimportant for catalysis, yet was juxtaposed with the catalytic core (Tang and Breaker, 1997) (Figure 4a). The activity of the resultant chimeric 'aptazyme' was in fact modulated by ATP. While there were various mechanisms by which it could be imagined that the conformational change of the aptamer regulated the activity of the ribozyme, one likely hypothesis was that the joining region between the aptamer and the ribozyme strongly affected the structure of the catalytic core. In order to further probe this hypothesis, Soukup and Breaker (1999) randomized the joining region and selected for ribozymes that were either activated or inhibited in the presence of FMN (Figure 4b). The generality of the selection procedure (Figure 3) was such that both types of aptazymes could be derived from the same random sequence pool. In one instance, ribozymes that reacted prior to the addition of FMN were removed from the population, then the FMN-dependent ribozymes were harvested; in the other instance, FMN was initially added, those ribozymes that reacted were removed, and the remaining catalytically active population was then harvested. The best aptazymes that were selected by these procedures had activities that were modulated by hundreds of fold in the presence of FMN. Moreover, it was found that the 'communication modules' that emerged from the randomized joining regions could actually intermediate between different ligandbinding domains and the hammerhead ribozyme (Figure 4c). For example, when an anti-theophylline aptamer was joined to the hammerhead ribozyme via a communication module originally selected for its ability to transduce FMNbinding to catalysis, the cleavage activity of the hammerhead was again activated, only this time by theophylline. Koizumi et al. (1999) went beyond the modular joining of aptamers and catalysts by randomizing the entire allosteric domain (Figure 4d). A random sequence pool of 25 residues in length was appended to the catalytic core of the ribozyme via the same stem that had previously proven useful for the addition of aptamers, and a selection for ribozymes whose activities were modulated by cyclic mononucleotides was initiated. All four cyclic mononucleotides were included in the selection, and after 16 rounds of selection and amplification the population was observed to be somewhat dependent upon cGMP for activity. This effector was then removed from the population and the selection was continued. Ultimately, ribozymes were found that were dependent upon three of the four cyclic mononucleotides; no cUMP-specific ribozymes were ever discovered. Individual aptazymes could be activated up to 5,000-fold by their cognate effectors.

381

Rajendran and Ellington

Figure 4. Strategies for adapting the hammerhead ribozyme to be an allosteric enzyme. Top: Sequence and structure of the hammerhead ribozyme. (a) Appending an aptamer to the hammerhead ribozyme in place of a non-essential stem. (b) Randomization of the joining region between the catalytic core of the hammerhead ribozyme and the appended aptamer. Selection is for inhibition or activation by the aptamer's cognate ligand, and typically involves coupled negative and positive selection steps. (c) Changing the effector-binding domain on an allosteric ribozyme. The pre-identification of a 'communication module' by the method described in (b) sometimes assists in this process. (d) Complete randomization of the effector-binding domain and selection for inhibition or activation of the ribozyme. 382

Nucleic Acids for Reagentless Biosensors Interestingly, the selected allosteric domains have many of the same characteristics as selected binding domains. A designed, theophylline-dependent hammerhead ribozyme was partially randomized, and variants that could be activated by other effectors were selected (Soukup et al., 2000). A single mutation altered the specificity of the aptamer from theophylline to 3methylxanthine, which differs from theophylline by the absence of a single methyl group. In fact, selected allosteric domains can be detached from ribozymes and function as aptamers in their own right (Soukup et al., 2001). The cNMP binding domains originally selected by Koizumi et al. (1999b) were probed by random mutagenesis and re-selection (Koizumi et al., 1999a). Based on the relative degree of sequence conservation and variation at different positions, it proved possible to predict which residues were involved in effector binding and which were more likely involved in catalysis. The predicted allosteric domains were separately synthesized, and largely retained the ligandbinding properties that they exhibited within the aptazymes. While the hammerhead ribozyme has obviously proven to be an excellent platform for the design and selection of aptazymes, it was originally unclear whether this ribozyme and the design principles that were built around it were unique. Our lab therefore undertook similar experiments with a ribozyme ligase (L1) that had been selected from a random sequence pool (Robertson and Ellington, 1999). The L1 ligase initially (and fortuitously) proved to be highly dependent (10,000-fold activation) on an oligonucleotide effector that was present during the selection, and thus it seemed likely that it might be adapted to other types of effector-dependence as well. To this end, anti-adenosine, antitheophylline, and anti-flavin aptamers were adjoined to the ligase in place of a stem structure that was relatively unimportant for catalysis, yet was adjacent to the catalytic core (Robertson and Ellington, 2000). As with the hammerhead ribozyme, these L1 ligase chimeras proved to be ATP- and theophyllinedependent. However, flavin-dependence was initially minimal, but was readily optimized by randomization and selection of the 'communication module' connecting the aptamer and ribozyme. We were most interested in generating aptazymes that were protein-dependent. However, the design principles that had previously proven effective in identifying aptazymes that were modulated by small organic effectors did not readily yield protein-dependent aptazymes. Anti-protein aptamers directly conjoined to the L1 ligase did not impart protein-dependence, nor did randomization and selection of the communication module, nor did randomization of the entire allosteric domain followed by a coupled negative and positive selection for protein-dependence (Figure 5). However, when both the allosteric domain and a portion of the catalytic domain were randomized and protein- dependence was selected for, it proved possible to identify aptazymes that were highly dependent on their cognate proteins (Robertson and Ellington, 383

Rajendran and Ellington 2001). For example, one aptazyme was isolated that was 75,000-fold dependent on tyrosyl tRNA synthetase from Neurospora mitochondria (Cytl 8), and another was found to be 3,500-fold dependent on hen egg white lysozyme. The proteindependent aptazymes had many of the characteristics previously observed for protein-dependent aptamers, in that they could readily distinguish between cognate and non-cognate proteins, including between the native and denatured states of the same protein. Given that the design principles originally elaborated by Breaker and his coworkers seemed to be generalizable, it is easy to imagine the development of a wide variety of aptazymes with a wide variety of catalytic functionalities. For example, by mounting two effector domains on the same ribozyme it should be possible to make aptazymes that are doubly-dependent on their effectors and that function as molecular 'and' gates (Figure 6). This concept was originally put into practice by Breaker and his co-workers (Jose et al., 2001), who mounted two aptamers on the hammerhead ribozyme in series. The resultant aptazyme was in fact dependent upon both ligands for full activity, and appeared to exhibit cooperative interactions between the effector-binding domains. We have similarly mounted two aptamers on the L1 ligase, only in this instance the aptamers were appended in parallel, rather than in series. Nonetheless, the resultant aptazyme was again dependent upon both ligands for full activity (Michael Robertson, personal communication). 3.4. Nucleic acid biosensor chips In the initial discussion of reagentless biosensors, it was clear that the utility of such sensors was in their ability to be adapted to multiple types of detection platforms. For reagentless biosensors based on nucleic acids, the ease of in vitro selection procedures offers the possibility that multiple different aptamers, ribozymes or aptazymes could be selected and modularly adopted to a single type of detector. Ultimately, it may be possible to develop chip arrays of nucleic acid biosensors that would be suitable for 'large' biological problems such as the acquisition of information about whole proteomes, metabolomes, or environmental dispositions of organisms and compounds (Brody et al., 1999).

384

Nucleic Acids for Reagentless Biosensors

,

/

/( ,1Incubation t-l,ncuua,

Affinity ~ ----------capture Eluani

"~

)/'(+) Incubation//

1. Reversetranscription

21S:lgeCtleVeatiqCRcR 4. Transcription

~ ~ /

p /

I

! ~o

_I

Figure 5. In vitro selection of protein-dependent aptazyme ligases. This scheme is similar to that shown for the isolation of effector-dependent hammerhead ribozymes (Figure 3d). However, in this instance, ligation of a substrate oligonucleotide to a random sequence population allows the capture of active catalysts via two mechanisms: binding to an oligonucleotide affinity column, and preferential amplification via PCR. It is interesting to note that to achieve the selection of protein-dependent aptazymes both the effector-binding domain and the catalytic core of the ribozyme had to be randomized.

385

Rajendran and Ellington

Figure 6. Different 'dual effector' aptazymes. (a) A dual effector hammerhead aptazyme designed by Breaker and co-workers. The two aptamers are joined in series. Stabilization of one ligand-binding domain results in stabilization of the adjacent domain, and overall stabilization of this structure results in stabilization of the catalytic core. (b) A dual effector ligase aptazyme designed by Michael Robertson and co-workers. The two aptamers are mounted in parallel on different stems of the L1 ligase. The responsivity profile as a function of effector concentration is shown at right. Maximal activity of the nucleic acid 'and gate' is only seen when both effectors are present at high concentrations. In the presence of single effectors, activation is minimal. Initial work on adapting aptamers to optical biosensors was modeled after antibody diagnostics. Aptamers have been used in a sandwich ELISA-!ike format (called ELONA, enzyme-linked oligonucleotide assay) to quantitate human vascular endothelial growth factor (VEGF) in serum samples (Drolet et al., 1996). The ELONA assay was found to be reproducible, could be used to measure VEGF concentrations as low as 25 pg/ml, and had a dynamic range of over three orders of magnitude. Capillary electrophoresis with laser-induced fluorescence detection (CE-LIF) has been used to sensitively detect IgE in solution using fluorescently labeled anti-IgE aptamer as a selective fluorescent probe (German et al., 1998). The method was highly sensitive, with a mass 386

Nucleic Acids for Reagentless Biosensors detection limit of 37 zmol of IgE, and a dynamic range of 105. Aptamers have also been used as probes in flow cytometry: a fluoresceinated anti-human neutrophil elastase (HNE) aptamer has been compared with an anti-HNE antibody in detecting HNE coated on beads (Davis et al., 1996). Similarly, aptamers against human CD4 have been conjugated to different fluorophores and used to stain human CD4 expressed on cells by flow cytometry (Davis et al., 1998). These somewhat simplistic antibody substitutions have been followed up by more sophisticated attempts to adapt compact, easily labeled aptamers to more system-specific analytic detection techniques and instrumentation. A fiber-optic microarray biosensor has been developed for thrombin by immobilizing antithrombin DNA aptamer at the distal tip of an imaging fiber coupled to a modified epifluorescence microscope system (Lee and Walt, 2000). The system has a detection limit of 1 nM and was used to measure thrombin concentrations in the range from nanomolar to low micromolar. Other examples include an aptamer biosensor for L-adenosine based on total internal reflection fluorescence detection which Could detect L-adenosine in the submicromolar range (Kleinjung et al., 1998) and one for thrombin based on evanescent wave-induced fluorescence detection (Potyrailo et al., 1998). The thrombin sensor had a dynamic range of three orders of magnitude, and was highly sensitive with a detection limit of 0.7 amol in a 140 pL volume. While these systems and adaptations have variable potential for the eventual development of chip arrays, the very fact that aptamers are readily adapted across such diverse platforms bodes well for the eventual use of aptamers in virtually any detection modality that comes to the fore. The fact that aptazymes directly transduce molecular recognition into catalysis may allow them to function as reagentless biosensors in simple but robust chip arrays. As an example, Seetharaman et al. (2001) immobilized various radioactive hammerhead aptazymes on a gold surface via a 5' phosphorothioate moiety. Following the addition of effectors or effector mixtures, appropriatelyactivated hammerheads cleaved themselves away from the surface and could be detected by simply transferring the supernatant to a new microtitre plate. Most remarkably, a cAMP-sensing aptazyme could be used to accurately quantitate the amount of cAMP in the culture medium of various E. coli strains that had alterations to adenosine metabolism. We have developed similar chips based on the L1 ligase. In this manifestation, however, activated aptazyme ligases conjugate themselves to substrates and coimmobilize radiolabels onto a surface (Figure 7). One of the advantages of aptazymes for chip-based applications is that they are already known to recognize a chemically diverse set of substrates. The ligase chip simultaneously detects the presence of oligonucleotide, small organic molecules, peptides, and proteins, a feat that would not be possible in a conventional ELISA assay. 387

Rajendran and EUington

Figure 7. Aptazyme ligase chip. Radiolabeled aptazymes derived from the L1 ligase (see ligand specificities at left) were incubated with biotinylated substrates. Following reaction, the mixtures were incubated with a streptavidin microtitre plate and unbound material was washed away. While this is not a reagentless application, the figure above does show the ability of aptazymes to specifically detect small molecules (FMN, ATP, theophylline), oligonucleotides (18.90, the effector for the original L1 ligase, see Figure 6, sequence in italics), peptides (the arginine-rich motif of Rev), and proteins (Lys = lysozyme). '-' is the no effector control, and '+' is all effectors mixed together.

4. Advantages and Limitations While the use of signaling aptamers and aptazymes in biosensors holds great potential, it is nonetheless true that nucleic acids are generally not as effective as proteins both as binding reagents and as catalysts. While aptamers typically bind their ligands in the nanomolar to micromolar range, antibodies can generally bind ligands well into the femtomolar range. Similarly, the few protein-sensing aptazymes developed so far can sense proteins into the nanomolar range, whereas ELISA assays can sense as few as 108 molecules in a sample (Rogers, 2000). 388

Nucleic Acids for Reagentless Biosensors Thus, an important question is whether the seemingly superior recognition elements, proteins, can function in a reagentless system in much the same way that aptamers and aptazymes can. Reagentless optical biosensors based on fluorescently labeled proteins are known (Brennan, 1999; Hellinga and Marvin, 1998). Many bacterial periplasmic binding proteins such as the phosphate binding protein (Brune et al., 1994), maltose binding protein (Gilardi et al., 1994; Marvin et al., 1997), glucose/galactose binding protein (Marvin and Hellinga, 1998), and glutamine binding protein (Dattelbaum and Lakowicz, 2001), have been converted into receptors for reagentless biosensors by the incorporation of single fluorophores that report ligand-dependent conformational changes. Such protein-based biosensors are usually generated by incorporating fluorophores (via covalent coupling to site specific single point cysteine mutations) either in close proximity to the ligand binding site, so that they can directly report ligand binding, or in distal regions of the protein such that the fluor can give an indirect readout based on domain movements in proteins which involve an allosteric coupling mechanism. This strategy is much the same as we have described above for signaling aptamers. As an example, when reporter fluors were incorporated into different positions in E. coli glucose binding protein, two of the variants were found to signal well

(Marvin and Hellinga, 1998). One of the variants had the fluorescent label in an allosterically-linked site and showed a 2-fold decrease in fluorescence upon ligand binding with minimal effect on the sugar binding constant [Ka (glc) !ncreased by a factor of 2, and Ka (gal) by a factor of--1.5]. A variant with the reporter fluor in the ligand-binding pocket showed a 4-fold increase in fluorescence, but with a much larger associated loss of binding [K4 (glc) increased--100 fold, and Ka (gal)--500 ~fold]. While these glucose binding proteins can potentially be used t o measure glucose concentrations in the micromolar range, a composite maltose biosensor obtained by mixing four similarly engineered maltose binding proteins measured maltose concentrations over a range of 0.1 jam - 20mM with an accuracy of 5% (Marvin et al., 1997). The best maltose binding protein had a fluorophore incorporated in an allosteric site and showed a greater than 4-fold increase in fluorescence. The remaining three were made by mutating residues in the binding pocket known to interact with maltose; this decreased the affinity for maltose without effecting signaling ability. These latter results suggest that molecular recognition and allosteric signal transduction can be independently-manipulated in protein-based biosensors. Therefore, it should be possible to change the binding specificity of a protein without affecting signaling. Hellinga's group recently converted a maltose sensor into a zinc biosensor by changing the specificity of the maltose binding protein using a rational design strategy (Marvin and Hellinga, 2001). By 389

Rajendran and Ellington

employing an iterative progressive design strategy, they were able to increase the zinc affinity of the maltose binding protein (one of their final constructs had a Ka -- 350 nM for zinc); the mutant also showed a greater than 17-fold increase in fluorescence upon zinc binding, indicating an alteration of the initial structure. Protein biosensors based on other metalloproteins such as zinc finger peptides (Godwin and Berg, 1996; Walkup and Imperiali, 1996; Walkup and Imperiali, 1997) and carbonic anhydrase (Elbaum et al., 1996) have also been made in a similar manner by incorporating environmentally sensitive fluors. Alternately, the intrinsic fluorescence of reporter proteins, such as the green fluorescence protein (GFP), has been exploited through the introduction of engineered analyte-binding sites. Calcium biosensors have been made by fusing two variants of GFP with calmodulin, and calmodulin-binding peptide. The introduction of Ca 2+ induced the binding of calmodulin around the calmodulinbinding peptide, ultimately resulting in a fluorescence energy transfer between the flanking GFPs (Miyawaki et al., 1997). This construct showed a 70% increase in the ratio of ultraviolet-excited emissions at 510 and 445 nm (1.8-fold change in the maximal ratio), and with the introduction of mutations into calmodulin, was able to report calcium concentrations over a very wide range from 10.8 to 10.2 M. In another study, a GFP-based calcium sensor was made by again joining together two GFP variants using the calmodulin-binding peptide (Romoser et al., 1997). This sensor showed a six-fold decrease in the ratio of emissions at 505 nm:440 nm and was able to report calcium concentrations from 50 nM to 1 gM. Peter Schultz and coworkers have developed a totally different strategy for modulating protein-protein and protein-nucleic acid interactions using small molecules (Guo et al., 2000). The method is based on first creating a cavity at a protein-protein interface such that it results in a reduction in binding affinity, and then screening libraries of small molecules to identify ligands that can bind the cavity and restore the interaction. This method is similar to one of the strategies for aptazyme construction (Figure 4d). However, rather than adapting the randomized allosteric domain to a particular effector, a random set of effectors is adapted to a particular allosteric domain. Two amino acids at the interface between human growth hormone (hGH) and the hGH receptor were mutated to glycines, creating a cavity at the interface and decreasing the binding affinity between the two by a factor of 106. By screening a library of indole derivatives, a ligand was identified which increased the affinity of mutated hGH for its mutated receptor more than 1000-fold.

5. Potential for Improving Performance

Although proteins with their wider array of functional groups may be inherently better biopolymer binding species and catalysts than nucleic acids, they have so 390

Nucleic Acids for Reagentless Biosensors far not proven to be much better for reagentless biosensors. For example, initially designed signaling aptamers showed approximately 25-45% increase in fluorescence upon interaction with ATP, while selected signaling aptamers showed an 80% increase (Jhaveri et al., 2000b); these results are similar to those that have been exhibited by the designed protein based biosensors described above. The anti-thrombin aptamer formulated as a signaling aptamer showed a sensitivity of 5 nM in solution (Hamaguchi et al., 2001), while the same aptamer mounted on a glass platform also showed a sensitivity of 5 nM (the biosensor could detect 0.7 amol of thrombin in a 140 pL volume) (Potyrailo et al., 1998). The values are the same or perhaps even better than those exhibited by the reagentless protein biosensors we have examined. Finally, we have seen that aptazymes can be activated over a thousand-fold by their cognate effectors, results that are as good as the engineered allostery that Schultz and co-workers achieved with hGH and its receptor (Guo et al., 2000). Thus, at the current time, nucleic acid and protein biosensors seem to be equivalant with regards to function, and the relative newness of aptamers, ribozymes, and aptazymes in the sensor arena means that virtually any new experiment can potentially result in greatly improved performance. Overall, nucleic acids for use in reagentless biosensors are essentially only as good as three inherent parameters: binding affinity, binding specificity, and signaling relative to background. Binding affinity can potentially be improved by augmenting the chemically simple complement of canonical nucleotides with modified nucleotides. For example, would the inclusion of a uridine residue that contained a branched, hydrophobic (isoleucine-like) group at the 5 position enable the selection of structures that could better recognize more hydrophobic epitopes and small molecules? The binding specificities exhibited by aptamers and aptazymes have so far been quite good, and can likely be greatly improved by a continued focus on negative selection experiments with highly related targets. One of the great advantages of nucleic acid selection relative to in vivo immunization is the ability to discretely and exactly control selection stringencies and parameters in order to deliver up molecules with just the right functional properties. As indicated above, it is the activation parameters of nucleic acid receptors in reagentless biosensors that truly shine relative to that of protein counter parts, and it is likely that sly manipulations of nucleic acid conformational changes by design or selection will continue to tweak these numbers upwards for sometime to come. Paradoxically, though, increases in activation seem to come at a cost in sensitivity: signaling aptamers show higher apparent Ka's for their ligands than do their parental aptamers; aptazymes show higher K4's for their effectors than do corresponding aptamers. In both instances, it is not unreasonable to suspect that ligand-binding energy is transduced into conformational changes, and that the greater the conformational change, the more ligand-binding energy must be diverted to that conformational change. In turn, the more ligand-binding energy that is diverted to a conformational change, the lower the intrinsic affinity of a given signaling aptamer or aptazyme will be for 391

Rajendran and Ellington its cognate ligand. The question thus becomes whether there is an experimental resolution of this seeming thermodynamic paradox; is it possible to design o r (more likely) select reagentless nucleic acid biosensors that simultaneously have both higher affinities and higher activation parameters? The answers to this question will likely become available shortly, but in the interim Ron Breaker has found that it is indeed possible to select for increasingly lower apparent K4'sfor cyclic nucleotide activated hammerhead aptazymes without loss of activation (Koizumi et al., 1999b). The greatest advantage and potential that aptamers and aptazymes currently have relative to protein reagents is the prospect for the high-throughput generation of multiple different receptors against multiple different targets. The ability to generate large numbers of nucleic acid recognition elements at will should make it possible to plumb the constitution and functionality of organismal proteomes and metabolomes. We have recently developed automated methods for the selection of both aptamers and nucleic acid enzymes (Cox and Ellington, 2001; Cox et al., 1998).

6. References

Barrel, D. P. and J. W. Szostak, 1993, Science 261, 1411. Bar-Ziv, R. and A. Libchaber, 2001, Proc. Natl. Acad. Sci. USA 98, 9068. Bock, L. C., L. C. Griffin, J. A. Latham, E. H. Vermaas and J. J. Toole, 1992, Nature 355, 564. Breaker, R. R. and G. F. Joyce, 1994, Chem. Biol. 1,223. Breaker, R. R. and G. F. Joyce, 1995, Chem. Biol. 2, 655. Brennan, J. D., 1999, J. Fluoresc. 9, 295. Brody, E. N. and L. Gold, 2000, Rev. Mol. Biotechnol. 74, 5. Brody, E. N., M. C. Willis, J. D. Smith, S. Jayasena, D. Zichi and L. Gold, 1999, Mol. Diagn. 4, 381. Brune, M., J. L. Hunter, J. E. Corrie and M. R. Webb, 1994, Biochem. 33, 8262. Carmi, N., L. A. Shultz and R. R. Breaker, 1996, Chem. Biol, 3, 1039. Chun, S. M., S. Jeong, J. M. Kim, B. O. Chong, Y. K. Park, H. Park and J. Yu, 1999, J. Am. Chem. Soc. 121, 10844. Ciesiolka, J. and M. Yams, 1996, RNA 2, 785. Conn, M. M., J. R. Prudent and P. G. Schultz, 1996, J. Am. Chem. Soc. 118, 7012. Conrad, R., L. M. Keranen, A. D. Ellington and A. C. Newton, 1994, J. Biol. Chem. 269, 32051. Cox, J. C. and A. D. Ellington, 2001, Bioorg. Med. Chem. 9, 2525. Cox, J. C., P. Rudolph and A. D. Ellington, 1998, Biotechnol. Prog. 14, 845. Dattelbaum, J. D. and J. R. Lakowicz, 2001, Anal. Biochem. 291, 89. Davis, K. A., B. Abrams, Y. Lin and S. D. Jayasena, 1996, Nucleic Acids Res. 24, 702. 392

Nucleic Acids for Reagentless Biosensors Davis, K. A., Y. Lin, B. Abrams and S. D. Jayasena, 1998, Nucleic Acids Res. 26,3915. Dieckmann, T., E. Suzuki, G. K. Nakamura and J. Feigon, 1996, RNA 2, 628. Drolet, D. W., L. Moon-McDermott and T. S. Romig, 1996, Nature Biotechnol. 14, 1021. Eaton, B. E. and W. A. Pieken, 1995, Annu. Rev. Biochem. 64, 837. Elbaum, D., S. K. Nair, M. W. Patchan, R. B. Thompson and D. W. Chri.stianson, 1996, J. Am. Chem. Soc. 118, 8381. Ellington, A. D. and J. W. Szostak, 1990, Nature 346, 818. Famulok, M., G. Mayer and M. Blind, 2000, Acc. Chem. Res. 33, 591. Fang, X., J. J. Li and W. Tan, 2000, Anal. Chem. 72, 3280. German, I., D. D. Buchanan and R. T. Kennedy, 1998, Anal. Chem. 70, 4540. Gilardi, G., L. Q. Zhou, L. Hibbert and A. E. Cass, 1994, Anal. Chem. 66, 3840. Godwin, H. A. and J. M. Berg, 1996, J. Am. Chem. Soc. 118, 6514. Green, L. S., D. Jellinek, C. Bell, U A. Beebe, B. D. Feistner, S. C. Gill, F. M. Jucker and N. Janjic, 1995, Chem. Biol. 2, 683. Guo, Z., D. Zhou and P. G. Schultz, 2000, Science 288, 2042. Haller, A. A. and P. Sarnow, 1997, Proc. Natl. Acad. Sci. USA 94, 8521. Hamaguchi, N., A. Ellington and M. Stanton, 2001, Anal. Biochem. 294, 126. Hellinga, H. W. and J. S. Marvin, 1998, Trends Biotechnol. 16, 183. Hermann, T. and D. J. Patel, 2000, Science 287, 820. Hesselberth, J., M. P. Robertson, S. Jhaveri and A. D. Ellington, 2000a, Rev. Mol. Biotechnol. 74, 15. Hesselberth, J. R., D. Miller, J. Robertus and A. D. Ellington, 2000b, J. Biol. Chem. 275, 4937. Hirao, I., M. Spingola, D. Peabody and A. D. Ellington, 1998, Mol. Divers. 4, 75. Homann, M. and H. U. Goringer, 1999, Nucleic Acids Res. 27, 2006. Huizenga, D. E. and J. W. Szostak, 1995, Biochem. 34, 656. Illangasekare, M., G. Sanchez, T. Nickles and M. Yarus, 1995, Science 267, 643. Jayasena, S. D., 1999, Clin. Chem. 45, 1628. Jayasena, V. K. and L. Gold, 1997, Proc. Natl. Acad. Sci. USA 94, 10612. Jellinek, D., L. S. Green, C. Bell, C. K. Lynott, N. Gill, C. Vargeese, G. Kirschenheuter, D. P. McGee, P. Abesinghe, W. A. Pieken, 1995, Biochem. 34, 11363. Jenison, R. D., S. C. Gill, A. Pardi and B. Polisky, 1994, Science 263, 1425. Jenne, A., J. S. Hartig, N. Piganeau, A. Tauer, D. A. Samarsky, M. R. Green, J. Davies and M. Famulok, 2001, Nature Biotechnol. 19, 56. Jhaveri, S., R. Kirby, R. Conrad, E. J. Maglott, M. Bowser, R. T. Kennedy, G. Glick and A. D. Ellington, 2000a, J. Am. Chem. Soc. 122, 2469. Jhaveri, S., M. Rajendran and A. D. Ellington, 2000b, Nature Biotechnol. 18, 1293. Jiang, F., R. A. Kumar, R. A. Jones and D. J. Patel, 1996, Nature 382, 183. Jose, A. M., G. A. Soukup and R. R. Breaker, 2001, Nucleic Acids Res. 29, 1631. Kawakami, J., H. Irnanaka, Y. Yokota and N. Sugimoto, 2000, J. Inorg. Biochem. 82, 197. 393

Rajendran and Ellington Kleinjung, F., S. Klussmann, V. A. Erdmann, F. W. Scheller, J. P. Fuerste and F. F. Bier, 1998, Anal. Chem. 70, 328. Koizumi, M., J. N. Kerr, G. A. Soukup and R. R. Breaker, 1999a, Nucleic Acids Symp. Ser. 275. Koizumi, M., G. A. Soukup, J. N. Kerr and R. R. Breaker, 1999b, Nature Struct. Biol. 6, 1062. Kuwabara, T., M. Warashina and K. Taira, 2000a, Trends Biotechnol. 18,462. Kuwabara, T., M. Warashina and K. Taira, 2000b, Curr. Opin. Chem. Biol. 4, 669. Lee, M. and D. R. Walt, 2000, Anal. Biochem. 282, 142. Leech, D. and F. Daigle, 1998, Analyst 123, 1971. Li, J. and Y. Lu, 2000, J. Am. Chem. Soc. 122, 10466. Li, J., W. Zheng, A. H. Kwon and Y. Lu, 2000, Nucleic Acids Res. 28, 481. Li, Y. and D. Sen, 1996, Nature Struct. Biol. 3,743. Lin, C. H. and D. J. Patel, 1997, Chem. Biol. 4, 817. Lin, Y., Q. Qiu, S. C. Gill and S. D. Jayasena, 1994, Nucleic Acids Res. 22, 5229. Liu, X., W. Farmerie, S. Schuster and W. Tan, 2000, Anal. Biochem. 283, 56. Liu, X. and W. Tan, 1999, Anal. Chem. 71, 5054. Lohse, P. A. and J. W. Szostak, 1996, Nature 381,442. Macaya, R. F., P. Schultze, F. W. Smith, J. A. Roe and J. Feigon, 1993, Proc. Natl. Acad. Sci. USA 90, 3745. Mannironi, C., A. Di Nardo, P. Fruscoloni and G. P. Tocchini-Valentini, 1997, Biochem. 36, 9726. Marvin, J. S., E. E. Corcoran, N. A. Hattangadi, J. V. Zhang, S. A. Gere and H. W. Hellinga, 1997, Proc. Natl. Acad. Sci. USA 94, 4366. Marvin, J. S. and H. W. Hellinga, 1998, J. Am. Chem. Soc. 120, 7. Marvin, J. S. and H. W. Hellinga, 2001, Proc. Natl. Acad. Sci. USA 98, 4955. Mills, D. R., R. L. Peterson and S. Spiegelman, 1967, Proc Natl Acad Sci USA 58,217. Miyawaki, A., J. Llopis, R. Heim, J. M. McCaffery, J. A. Adams, M. II~ra and R. Y. Tsien, 1997, Nature 388, 882. Morris, K. N., K. B. Jensen, C. M. Julin, M. Weil and L. Gold, 1998, Proc. Natl. Acad. Sci. USA 95, 2902. Mulchandani, A. and S. Pan, 1999, Anal. Biochem. 267, 141. Narvaez, A., G. Suarez, I. C. Popescu, I. Katakis and E. Dominguez, 2000, Biosens. Bioelectron. 15, 43. Oliphant, A. R., C. J. Brandl and K. Struhl, 1989, Mol. Cell Biol. 9, 2944. Pagratis, N. C., C. Bell, Y. F. Chang, S. Jennings, T. Fitzwater, D. Jellinek and C. Dang, 1997, Nature Biotechnol. 15, 68. Pan, T., B. Dichtl and O. C. Uhlenbeck, 1994, Biochem. 33, 9561. Pan, T. and O. C. Uhlenbeck, 1992, Biochem. 31, 3887. Pan, W., R. C. Craven, Q. Qiu, C. B. Wilson, J. W. Wills, S. Golovine and J. F. Wang, 1995, Proc. Natl. Acad. Sci. USA 92, 11509. Patel, D. J. and A. K. Suri, 2000, Rev. Mol. Biotechnol. 74, 39. 394

Nucleic Acids for Reagentless Biosensors Piatek, A. S., S. Tyagi, A. C. Pol, A. Telenti, L. P. Miller, F. R. Kramer and D. Alland, 1998, Nature Biotechnol. 16, 359. Porta, H. and P. M. Lizardi, 1995, Biotechnology (N Y), 13, 161. Potyrailo, R. A., R. C. Conrad, A. D. Ellington and G. M. Hieftje, 1998, Anal. Chem. 70, 3419. Prudent, J. R., T. Uno and P. G. Schultz, 1994, Science 264, 1924. Ringquist, S., T. Jones, E. E. Snyder, T. Gibson, I. Boni and L. Gold, 1995, Biochem. 34, 3640. Robertson, D. L. and G. F. Joyce, 1990, Nature 344, 467. Robertson, M. P. and A. D. Ellington, 1999, Nature Biotechnol. 17, 62. Robertson, M. P. and A. D. Ellington, 2000, Nucleic Acids Res. 28, 1751. Robertson, M. P. and A. D. Ellington, 2001, Nature Biotechnol. 19, 650. Rogers, K. R., 2000, Mol. Biotechnol, 14, 109. Romoser, V. A., P. M. Hinkle and A. Persechini, 1997, J. Biol. Chem. 272, 13270. Roth, A. and R. R. Breaker, 1998, Proc. Natl. Acad. Sci. USA 95, 6027. Sassanfar, M. and J. W. Szostak, 1993, Nature 364, 550. Schultze, P., R. F. Macaya and J. Feigon, 1994, J. Mol. Biol. 235, 1532. Seetharaman, S., M. Zivarts, N. Sudarsan and R. R. Breaker, 2001, Nature Biotechnol. 19, 336. Singh, K. K., T. Rucker, A. Hanne, R. Parwaresch and G. Krupp, 2000, Biotechniques 29, 344. Sokol, D. L., X. Zhang, P. Lu and A. M. Gewirtz, 1998, Proc. Natl. Acad. Sci. USA 95, 11538. Soukup, G. A. and R. R. Breaker, 1999, Proc. Natl. Acad. Sci. USA 96, 3584. Soukup, G. A., E. C. DeRose, M. Koizumi and R. R. Breaker, 2001, RNA 7, 524. Soukup, G. A., G. A. Emilsson and R. R. Breaker, 2000, J. Mol. Biol, 298, 623. Spiegelman, S., 1971, Q Rev Biophys 4, 213. Stojanovic, M. N., P. de Prada and D. W. Landry, 2000, J. Am. Chem. Soc. 122, 11547. Stojanovic, M. N., P. de Prada and D. W. Landry, 2001, J. Am. Chem. Soc. 123, 4928. Tan, W., X. Fang, J. Li and X. Liu, 2000, Chemistry 6, 1107. Tang, J. and R. R. Breaker, 1997, Chem. Biol. 4, 453. Tucker, C. E., L. S. Chen, M. B. Judkins, J. A. Farmer, S. C. Gill and D. W. Drolet, 1999, J. Chromatogr. B Biomed. Sci. Appl. 732, 203. Tuerk, C. and L. Gold, 1990, Science 249, 505. Tyagi, S., D. P. Bratu and F. R. Kramer, 1998, Nature Biotechnol. 16, 49. Tyagi, S. and F. R. Kramer, 1996, Nature Biotechnol. 14, 303. Vitiello, D., D. B. Pecchia and J. M. Burke, 2000, RNA 6, 628. Walkup, G. K. and B. Imperiali, 1996, J. Am. Chem. Soc. 118, 3053. Walkup, G. K. and B. Imperiali, 1997, J. Am. Chem. Soc. 119, 3443. Warashina, M., T. Kuwabara and K. Taira, 2000, Structure Fold Des. 8, R207. Wedekind, J. E. and D. B. McKay, 1999, Nature Struct. Biol. 6, 261. 395

Rajendran and EUington Willis, M. C., B. D. Collins, T. Zhang, L. S. Green, D. P. Sebesta, C. Bell, E. Kellogg, S. C. Gill, A. Magal!anez, S. Knauer, R. A. Bendele, P. S. Gill, N. Janjic and B. Collins, 1998, Bioconjug. Chem. 9, 573. Wilson, C. and J. W. Szostak, 1995, Nature 374, 777. Wilson, D. S. and J. W. Szostak, 1999, Annu. Rev. Biochem. 68, 611. Yamamoto, R.~ T. Baba and P. K. Kumar, 2000, Genes Cells 5,389.

396

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 13

NEW MATERIALS BASED ON IMPRINTED POLYMERS AND THEIR APPLICATION IN OPTICAL SENSORS

SERGEY A. PILETSKY, PH.D* AND ANTHONY P.F. TURNER, PH.D., D.Sc.

Institute of BioScience and Technology, Cranfield University, Silsoe, Bedfordshire, MK45 4DT, UK.

Molecular imprinting is the process of template-induced formation of specific recognition sites (binding or catalytic) in a material where the template directs the positioning and orientation of the material's structural components by a self-assembling mechanism. Synthetic receptors prepared using molecular imprinting possess a unique combination of properties, such as high affinity, specificity, low price and robustness, which make them an attractive alternative to natural receptors, enzymes and antibodies used in biosensors. This review gives a brief overview of the technology with specific emphasis on the mechanisms underlying the ability of imprinted polymers to perform highly selective functions such as recognition and transformation of a binding event into a detectable optical signal. The problems associated with the application of molecularly imprinted polymers (MIPs) in sensors are highlighted. Possible solutions to these problems are discussed and recommendations made about where commercial application of imprinted sensors seems most feasible in the near future.

1. Molecular Imprinting The molecular imprinting approach exploits the formation of a complex between a template molecule and functional monomers, which is fixed by copolymerisation with cross-linker into a growing polymer network (Figure 1). Following removal of the template, binding sites are left in the polymers, which have the shape and orientation of functional groups, complementary to those of the template molecule (Wulff, 1995; Mayes and Mosbach, 1997). 397

Piletsky and Turner

Figure 1. Scheme of molecularly imprinted polymerization.

1.1. Different formats used for design of imprinted materials The typical recipe for MIP preparation includes mixing together target compound -template with corresponding functional monomer (most f r e q u e n t l y methacrylic acid) and cross-linker (e.g., ethylene glycol dimethacrylate) in appropriate solvent (chloroform, acetonitrile) and polymeristion of this mixture using UV or chemical initiation (O'Shannessy et al., 1989). The template can be extracted from the polymer by washing or by electrophoresis (Piletsky et al., 1992a). Subsequent polymer grinding and washing yields the polymer particles with receptor sites on the accessible surface. Other formats of molecular imprinting include: 1. Polycondensation of silica acid in the presence of template (Katz and Davis, 2000); 2. Electropolymerisation (Malitesta et al., 1999); 3. Formation of 2-dimensional templated monolayers onto a SIO2, metal oxide or gold surface (Starodub et al., 1992; Mirsky et al., 1999); 4. Grafting of imprinted polymers to the inert solid surface (Dhal et al., 1995; Piletsky et al., 2000a) (Figure 2); 5. Templating of a pre-formed polymer structure by precipitation or crosslinking in the presence of template (Braco et al., 1990; Peissker and Fischer, 1999); 6. Formation of imprinted poly/oligomers (e.g. peptides) in the presence of template (Giraudi et al., 2000; Piletska et al., 2000). 398

Materials Based on Imprinted Polymers

Figure 2. Scheme of MIP synthesis via surface photografting onto porous polymeric substrate (Piletsky et al., 2000a) Historically, formation of imprinted silica gels was the first example of molecular imprinting (Polyakov, 1931). Despite the fact that this specific technique reached its peak in the sixties and is now in the process of gradual decline, due to limited flexibility of the method, it still remains the most popular choice for the preparation of specific zeolites (Dong et al., 2000). Electropolymerisation faces the same type of problem as silica imprinting due to limited number of polymerisable functional monomers available, which are selected mainly from the group of aniline, phenol, pyrrole and thiophene (Panasyuk et al., 1999; B lanchard et al., 2000). Electropolymerisation retains its attraction, however, because it provides a means for precise deposition of a sensitive layer on an electrode surface, which is extremely important for microand multisensor production. Electropolymerised MIPs have been used almost exclusively in potentiometric (Boyle et al., 1989; u et al., 1990) and amperometric sensors (Piletsky et al., 1994a). Two-dimensional MIPs or imprinted monolayes were developed and used in optical sensors by Andersson and co-authors (1988). They used Tabushi's method (Tabushi et al., 1987) to ~immobilise octadecylchlorosilane in the presence of inert template hosts (n-hexadecane) onto a silicon oxide surface. 399

Piletsky and Turner After the extraction of hosts, vitamin K1 was detected by ellipsometry. Expanding this method for the preparation of monolayers, imprinted with watersoluble templates, we developed materials selective for amino and nucleic acids (Piletsky and Starodub, 1992b). This approach involves two steps" first, adsorption of the template on the surface of SiO2 or metal oxide; and second, treatment of the surface with adsorbed template by trimethyl chlorosilane from the gas phase. In another similar approach, a gold surface was imprinted with a cholesterol-specific monolayer using co-adsorption of the template with hexadecylmercaptane (Piletsky et al., 1999a). Despite some advantages, such as fast sensor response and easy preparation, these systems, however, suffer from lack of stability. The lateral mobility of the components of imprinted monolayer is responsible for steady decrease in the specificity of imprinted cavities. An essential improvement of the sensor stability was achieved by co-immobilisation of the template in imprinted layer. Using a new approach called "spread bar architecture design," it was possible to develop stable monolayers, consisting of template - thiobarbituric acid - and functional monomer- hexadecylmercaptane. A depression in the hexadecylmercaptane layer formed by the template was able to accommodate barbituric acid, changing electrode capacitance in the binding proces (Mirsky et al., 1999). This two-dimensional format for MIP design is particularly attractive for evanescence-wave sensing, e.g. surface plasmon resonance. Several reports on the preparation of MIPs by surface grafting have appeared, where a thin imprinted layer, most frequently a monolayer, is formed on a solid support (Dhal et al., 1995; Lele et al., 1999; Piletsky et al., 2000a). Grafting can be performed using chemical, UV or plasma initiation (Shi et al., 1999; Piletsky et al., 2000b). The advantage of this approach lies in the possibility of modifying very inert surface (polystyrene, polypropylene, etc.) with specific polymers. The additional attraction of electropolymerisation and grafting methods is their convenient format, which does not require an additional processing step. MIP synthesis and immobilisation is performed as a one-step procedure, directed by applied potential or by exposing the monomer mixture-coated detector to UV light (Figure 2). A further approach, frequently called "bioimprinting", involves precipitation or cross-linking of biological molecules (proteins) in the presence of template (Braco et al., 1990; Peissker and Fischer, 1999). The conformation adopted by interacting biopolymer around the template remains fixed after template extraction with an appropriate solvent. Although the authors are unaware of any examples where bioimprinting has been used for sensor design, this technique could potentially be useful for introducing either additional recognition sites into enzymes or catalytic sites into antibodies. These chimeric molecules might possess the combined characteristics of antibodies and enzymes and, in this way, be useful for the development of new, label-free types of assays and sensors. 400

Materials Based on Imprinted Polymers The last format of molecular imprinting is template-directed synthesis. This process includes the formation of a new substance by a chemical modification of the substrate, or by the coupling of two or more molecules, in the presence of a template to serve as a pattern for the formation of a new structure. The most well known example of this process is gene replication. An important issue is that the synthesised molecule always has a structure, complementary to that of template, which can be exploited for the synthesis of biospecific ligands or to obtain information about the structure and properties of the template molecule. This approach is actively pursued in molecular biology (gene sequencing) and in DNA sensors where complementary DNA or RNA chains are synthesised using transcription facilitated by enzymes such as DNA polymerase or reverse transcriptase. Unfortunately, a similar technique does not exist for the analysis of molecules other than DNA, such as proteins and polysaccharides. It is possible, however, to produce a complementary ligand for a target molecule using a synthetic approach (Giraudi et al., 2000; Piletska et al., 2000). The method involves the formation of oligomers, e.g., peptides, in the presence of template. Prior to the initiation of polymerisation, and during polymerisation, the monomers, which could be amino acids or nucleotides, spatially distribute themselves around the template molecules in accordance with the size, polarity and functionality of the template. The monomers are polymerised into linear, water-soluble oligomers specific for the template. The advantage of this approach is the possibility of obtaining watersoluble ligands, which can be treated in the same manner as antibodies and other natural receptors. Not withstanding the new methods detailed above, traditional bulk polymerisation remains the most popular choice for the preparation of molecularly imprinted polymers for theoretical study and practical application in separation and sensing.

1.2. Mechanism of template recognition by imprinted polymer Three major factors determine the recognition process" the quantity of the functional groups participating in the interaction, their correct arrangement within the cavity, and the shape of the cavity itself. The types of interactions explored in molecular imprinting include reversible covalent bonds (Wulff and Haarer, 1991), electrostatic interactions (ionic and hydrogen bonds) (Piletsky et al., 1990a; Nicholls et al., 1995), van der Waals (Dickert et al., 1998), hydrophobic interactions (Yu et al., 1997), and metal chelation (Matsui et al., 1996) (Figure 3). The shape of the cavity alone can provide specificity (Yoshizako et al., 1998) although the specificity is substantially better when the

401

Piletsky and Turner

OH) 2 +

~

+ H20

(a)

H

coo. § R.H ~

oO'U..~" .rid ~;H

(b)

+ ~N.,.R

0

~H__R

Figure 3. Different types of interactions explored in molecular imprinting: (a) reversible covalent bond formation; (b) electrostatic interactions; (c) metal chelation.

template interacts with one or more properly oriented functional monomers (Ramstrom et al., 1993). The required strength of monomer-template interaction varies depending on the size and the structure of the template. For a small template molecule, the presence of strong interactions, preferably ionic and/or hydrogen bonds, is critically important. For a large molecule such as a protein or nucleic acid, successful results can be achieved with a combination of multiple weak interactions (Hjerten et al., 1997). The choice of solvent depends on the type of interaction. Thus if template recognition depends on hydrogen bond formation, better results can be achieved if both polymer synthesis and re-binding takes place in a hydrophobic solvent, where hydrogen bonds are stronger (Andersson, 1996; Yu and Mosbach, 2000). The equilibrium dissociation constants (Kd) for the binding of ligands to their corresponding polymers have been estimated by Scatchard plot analysis of binding data. Mostly, non-linear plots were obtained because of multiple Kd values, varying in range in the majority of cases from micromolar to nanomolar. In a similar way to polyclonal antibodies, imprinted polymers contain a heterogeneous population of binding sites (Wulff, 1995). One of the important components of the recognition mechanism observed in MIP systems is the conformational change in the polymer induced by template interaction (Piletsky et al., 1992a; Watanabe et al., 1998; Wolfbeis et al., 1998). Depending on experimental conditions (solvent, temperature, and types of the monomer-template and monomer-monomer interactions), the polymer matrix can 402

Materials Based on Imprinted Polymers Table 1. Comparison of natural antibodies and receptors with MIPs. ..

..,.

,,,

Property

,

Natural Biomolecules .,

Stai~iiity

Low

Cost

High

Integrati& into multisensor unit

Compatibility with micromaching technology/.miniaturisation Spectrum of analytes

i

|,|

,

n

1=,.

i

-

MIPs

,

' Difficult due to integration OiY" natural biomolecules in multisensor unit is difficult due to different operational requirements of these molecules (pH, ionic strength, temperature, subs tsr,ate) Poor

Stable at low/high pHs, pressure and temperature . Inexpensive and easy preparation Flexible MIP design allows preparation of MIPs against many combinations of analytes Fully Compatible Practicaliy unlimited

Limited

i

ii

shrink or swell in the presence of template. The mechanism is similar to "induced fit" observed for natural enzymes and receptors (Koshland, 1995; Agmon, 2000). The importance of this effect for sensor technology lies in the possibility of use for measuring template concentration (Piletsky et al., 1998). An additional factor contributing to MIP recognition properties is the presence of nanopores in the polymer structure with specificity for the template molecules (Piletsky et al., 1990b; MathewKrotz and Shea, 1996). Membranes prepared by molecular imprinting possess selective permeability for the imprinted species and can be used for purification of desirable analytes or removal of potential interfering compounds. MIPs are capable of recognising small variations in the structure of the template and the specificity of imprinted polymers under optimised conditions is often equal to or even superior to that of natural enzymes and receptors (Andersson et al., 1995). However, quite often MIPs demonstrate a high level of non-specific binding. Although they quite often referred to in this way, it would be a mistake to see imprinted polymers as "plastic" antibodies or receptors. They are different materials with their own advantages and disadvantages and thus should be considered as additive, complementary systems rather than substitutes (Table 1).

403

Piletsky and Turner 1.3. "Pluses" and "minuses" in MIP technology and their comparison with natural enzymes and receptors

Being purely synthetic materials, it is natural that the imprinted polymers have a much higher stability than enzymes and receptors. The reason for this lies, first of all, in the high level of cross-linking, which provides adequate protection for binding sites created in the polymer by imprinting. Imprinted polymers can withstand harsh treatment with acidic and basic solutions or with organic solvent. They are stable under both high and low pressure, and, as well as at extreme temperatures (Kriz and Mosbach, 1995; Svenson and Nicholls, 2001). Imprinting polymerisation is a very inexpensive procedure for the development of artificial receptors. In the majority of cases, the price of a MIP depends almost entirely on the price of the template used. Furthermore, if the templates themselves are expensive, it maybe possible to recover the template and use it again. Alternatively, inexpensive template analogues can be used for the preparation of MIPs. Generally speaking, MIP preparation is three-to-four orders less expensive than production of the equivalent natural receptor, and this makes the technology very competitive. The possibility of using MIPs in organic solvents opens new areas of application such as biomimetic sensing and catalysis in chemical and pharmaceutical manufacturing. Quality control and on-line monitoring of manufacturing processes are particularly attractive. One of the most challenging problems associated with development of multisensors is related to the significant differences in the performance of natural enzymes and receptors. These biological materials all have different stability, activity and sensitivity; in many cases, they require different substrates and buffers with different ionic strengths and pHs. Due to such factors, the integration of naturally occurring bio-molecules in one single unit maybe problematic. Since a MIP's design is flexible and variety of monomers are available for their preparation, it is possible to develop a set of polymers specific for a range of templates which will have almost identical operational requirements (solvent, temperature, pH, etc.). An additional benefit comes from the possibility of processing MIPs in the same way as traditional photoresist materials. MIPs can be immobilised at precise spots on the detector surface using masks and photopolymerisation. The compatibility of MIPs with micromachining technology makes MIP-based multisensors feasible. Last, but not least, is the ability to develop MIPs for practically any type of compound. Examples of templates producing MIPs successful include inorganic ions, drugs, nucleic acids, proteins and even cells (Table 2). Although antibodies 404

Materials Based on Imprinted Polymers Table 2. Examples of templates used in molecular imprinting. i|

Ill

i

i

Template

Application

Reference

Amino acids and derivatives

Separation, sensors

Kempe and Mosbach, 1995; Vidiasankar et al., 1997; Piletsky et al., 1998

Aniline, phenol, derivatives

Sensing

Vinokurov and Grigoreva, 1990; Morita et al., 1997

Drugs

Separation, sensing

Levi et al., 1997; Wang et al., 1997; Mirsky et al., 1999; Andersson, 2000

Flavanoids

Sensing

SuS.rez-Rodriguez and Diaz-Garcia, 2000

Herbicides

Separation, sensing

Kroger et al., 1999; Sergeeva et al., 1999, 2001

Inorganic ions

Separation and sensing

Hutchins and Bachas, 1995; Yoshida et al., 2000; Kimaro et al., 2001.

Microorganisms

Recognition

Alexander and Vulfson, 1997; Dickert et al., 2001

Nucleic acids and derivatives

Separation, sensing

Piletsky et al., 1990a, 1990b; MathewKrotz and Shea, 1996

Polynuclear aromatic hydrocarbons

Sensing

Dickert et al., 1998

Proteins

Separation, recognition

Hjerten et al., 1997; Shi et al., 1999

Steroids

Separation, detection

Hishiya et al., 1999; Rachkov et al., 2000

Sugars, sugar derivatives

Separation, sensing

Wulff and Haarer, 1991; Piletsky et al., 1998

Toxins and narcotics

Separation, sensing

Kriz and Mosbach, 1995; Matshui et al., 1996; Takeuchi et al., 2001.

Volatile compounds

Sensing

Ji et al., 200; Dickert et al., 2001.

i

405

Piletsky and Turner can also be prepared for a broad range of analytes, they have two disadvantages when compared to MIPs. Firstly, small compounds often have to be derivatised in order to generate the antibodies. This necessitates an additional synthetic step, which can sometimes drastically change the recognition characteristics. Secondly, flexibility in antibody preparation is limited to twenty naturally occurring amino acids. In the case of MIPs, the large number of synthetic monomers available make it possible to engineer binding sites with a variety and flexibility unmatched by nature. As any other technology, molecular imprinting has shortcomings. Among them are: (i) absence of a general technology for MIP design; (ii) poor performance of MIPs in aqueous environments; (iii) high level of non-specific binding which produces too low a signal-to-noise ratio in sensors; (iv) poor processability of MIPs; and (v) difficulty in transforming binding events into electrical signals. Several attempts have been made in the past to develop a general procedure for the rational design of imprinted polymers with predictable properties (Nicholls, 1995; Whitcombe et al., 1998; Takeuchi et al., 1999; Lanza and Sellergren., 1999). In the best examples, workers have produced rules or hints, indicating how MIPs should be made in order to possess a certain level of specificity. The most important conclusion is that the stability of the monomer-template complex formed during polymerisation determines the affinity of the resulting polymer. Thus it is known that polymerisation should be performed in a hydrophobic solvent in order to produce a material able to interact with template through electrostatic interactions. At the same time, the choice of the monomer, solvent and polymerisation conditions generally depends on common knowledge, one's personal experience, or available information describing the behaviour of the similar systems. Recently we developed a method that is believed to be a general solution for MIP design (Piletsky et al., 2000c, 2001). The method involves computational screening of a virtual library of functional monomers against a target molecule. The monomers giving the best score in virtual binding experiments are then brought into the contact with template and left to equilibrate. The composition of the monomer shell surrounding the template after equilibration provides the information on the type and quantity of monomers, and should be used for polymer design. Commercially available software permits calculations to be performed using different dielectric constants, reflecting the polarity of the environment (solvent) where the polymers are prepared and used. Polymers designed using this computational approach have proved to have excellent affinity and specificity for the target compound, surpassing those of polyclonal antibodies (Table 3). The possibility of tailoring MIPs for specific target analytes and specific operational conditions is very attractive since it permits polymers to be 406

Materials Based on Imprinted Polymers Table 3. Affinity and sensitivity range of computationally designed molecularly imprinted polymer in comparison with antibodies for the template-- microcystine-LR. Receptor

Ka, (nM)

Sensitivity range (/xg 1~)

Computational MIP

0.3 + 0.08

0.1-100

Monoclonal antibody

0.03 _+0.004

0.025-5

Polyclonal antibody

0.5 +_0.07

0.05-10

i

iii

developed with optimised characteristics and shortens the time needed for design, preparation and testing of the polymers. The computer simulation and molecular modelling approaches could also help to solve a second major problem associated with MIPs - their poor performance in aqueous environments. The majority of monomers used so far in polymer design form hydrogen and ionic bonds in the process of template recognition. These interactions are less effective in polar solvents and as a result, the use of such MIPs is restricted mainly to hydrophobic solvents such as chloroform, toluene, and acetonitrile. Although MIPs capable of forming hydrophobic and van der Waals interactions with the template under aqueous conditions have been developed (Dickert et al., 1999), the design of such polymers is much more difficult than the design of MIPs which exploit electrostatic interactions. The reason lies in the complex nature of factors contributing to hydrophobic and van der Waals interactions. Computer simulation and molecular modeling can, in principle, solve this problem and help to select the monomers ideally suited for the recognition of the template in water. Typically the ratio of functional monomer:template used in molecular imprinting is 4:1 to 10:1. Therefore, the resulting polymer contains large amounts of monomers outside of the specific binding sites; these are capable of non-specific interaction with the molecules other than the template. Additionally the crosslinker itself can interact with variety of analytes in aqueous media. A combination of these factors, together with the large surface area (80-200 m2/g) of the polymer, is responsible for a high level of non-specific binding, which hinders the development of MIP-based affinity materials and sensors. It is possible to overcome this problem however, by further optimisation of the polymerisation procedure and by rational selection of monomers capable of forming stoichiometric complexes with templates (Lubke et al., 2000).

407

Piletsky and Turner

Figure 4. Three principal types of MIP sensors: (a) Affinity sensor, where response is produced by accumulation of template on MIP surface; (b) Receptor sensor where response is generated by changes in polymer characteristics, induced by its inl:eraction with template; (c) Enzyme-mimicking sensor responding to the change in the environment induced by MIP-mediated catalytic reaction.

Detection of binding can be achieved with the help of optical devices, if the template has, for example, fluorescent properties. At least three general characteristics of MIPs can be used for the design of MIP-based sensors (Figure 4)' First, by substituting MIPs for the antibodies in immunosensors (affinity sensors); Second, by exploring of the receptor properties of the imprinted polymers (receptor sensors); Third, by combining MIPs possessing catalytic properties with traditional electrochemical or optical transducers (catalytic sensors). The majority of biosensors produced to date use enzymes as biorecognition element (Turner, 1999). The reason for this lies in the amplification effect achieved as result of multiple turnover of catalytic processes. Many of the unique characteristics of enzymes are connected with their polymeric nature and this fact attracts attention to the methods of development of the MIP catalysts (Srikovsky et al., 2000). The application of catalytically-active MIPs for sensor development seems to be promising and attractive as the most direct way of achieving the replacement of current biosensors by more stable devices. Nevertheless, no practical examples exist of the integration of MIPs that mimic natural enzymes into sensors. Although essential progress has been made in MIP catalysis, imprinted polymers still have properties inferior to natural enzymes 408

Materials Based on Imprinted Polymers e.g., much lower activity and turnover. Success in the preparation of more effective MIPs-based catalysts, with high turnover and reacting with watersoluble and practically important analytes, will change this situation. Due to the very limited information available on the development of MIP-based catalytic sensors, this review will concentrate on the development of the two remaining types of sensors: affinity and receptor-min'ticking devices. Integration of imprinted polymers with detectors remains a difficult issue. Despite significant improvement in MIP technology, the processability of these materials remains challenging. The high level of cross-linking, necessary for maintaining the polymer's specificity, makes them extremely hard, solid, and fragile materials. One solution to this problem is the use of plasticisers, such as oligourethane acrylates (Sergeeva et al., 1999). As a result, polymers can be made in the form of thin and stable membranes which can be used directly in sensors. Alternatively, imprinted polymers can be grafted (Mirsky et al., 1999) or electropolymerised (Boyle et al., 1989) onto the detector surface.

2. Development of MIP-based Optical Sensors The majority of published papers related to MIP sensors deal with electrochemical or piezoelectric devices. We believe, however, that optical and, in particular, fluorescent sensors will play a major role in the future. This confidence is based on the great flexibility which fluorescent detection offers to MIP technology. The account below highlights current achievements and prospects for the development and commercialisation of MIP sensors, which use optical detection for template recognition.

2.1. Affinity sensors The most common type of MIP sensor is the affinity, immunosensor-type device. The detection principle here is based on the measurement of the concentration of template adsorbed by MIP immobilised on the detector surface. The first example of this type of device was the development of a two-dimensional MIP sensor for vitamin K1 (Andersson et al., 1988). Ellipsometry was used for the measurement of template concentration. Although this work was very preliminary and suffered from lack of appropriate controls, it demonstrated the possibility for direct detection of a template adsorbed by an imprinted mono!ayer. Steinke and co-authors, proposed an interesting variant of an optical sensor device based on MIPs (Steinke et al., 1996). The completely transparent imprinted polymer prepared for their experiments had anisotropic properties and provided a particular orientation of bound template molecules. The polymers therefore showed a pronounced dichroism in UV light, which enabled specific

409

Figure 5. Selectivity pattern of pyrene detection by fluorescence, using polyurethanes imprinted with polyaromatic hydrocarbons of different sizes (Dickert and Tortschanoff, 1999).

binding to be recognised. This work could be applicable in particular for the detection of optical isomers. An optical sensor specific for the fluorescent substance dansyl-L-phenylalanine was developed using a dansyl-L-phenylalanine-imprinted polymer and a fibreoptic sensing device (Kriz et al., 1995). Accumulation of fluorescent template in the polymer matrix resulted in an increase in fluorescence that could be used to detect 10 mg/1 of substrate within 4 h. In another example, fluorescent polycyclic aromatic hydrocarbons were selectively enriched and detected using optical sensors based on imprinted polyurethanes (Figure 5) (Dickert and Tortschanoff, 1999). A problem associated with broadening the scope of this method, is the limited quantity of fluorescent substances, which are practically important and can be used as templates in the preparation of MIPs for sensor technology. To overcome this problem, sensors can be developed that operate in a competitive 410

Materials Based on Imprinted Polymers

0.12 0.1

< J c3

0.08

Triazine

/

~

0.06

e

r~

O

0.04 0.02

Simazine

ill

_

1.E-10

I

t,..x

1.E-08

1.E-06

1.E-04

Log C, M Figure 6. Displacement of fluorescein-triazine from triazine-imprinted polymer by analytes (Piletsky et al., 1997).

mode. The important question was whether the binding sites in imprinted polymers are capable of recognising template molecules that are labeled with a fluorescent dye or enzyme. Successful demonstration of this possibility was performed for polymer imprinted with triazine (Piletsky et al., 1997). The competition between fluorescein-labeled and unlabeled template was used to measure 10s - 10.5 M concentrations of free template dissolved in ethanol (Figure 6). The polymer was able to discriminate the template from other triazines (e.g., atrazine) and triazinone (simazine). Later, competitive assays withenzyme-labeled templates were developed for epinephrine (Piletsky et al., 2000b) and 2,4-dichlorophenoxyacetic acid (Surugiu et al., 2001). The displacement format has been used for the development of an MIP sensor for chloramphenicol (CA) (Levi et al., 1997). The sensor included an HPLC column with CA-specific MIPs. A constant flow of dye-labeled CA (chloramphenicolMethyl Red) at a concentration of 0.5 ~tg/ml was run through the column under equilibrium conditions. When analyte containing free CA was injected, it displaced the adsorbed conjugate, giving a peak with an area proportional to CA concentration (Figure 7). Successful analysis of chloramphenicol was achieved in model and real samples (blood serum). 411

Piletsky and Turner

_

_

/

<

/

/

/ t~

/

_

0

_

_

/

~ I

i

!

i

i

i

i

5

10

15

20

25

30

35

Chloramphenicol, mg/I

Figure 7. Displacementof chloramphenicol-Methyl Red from a chloramphenicolimprinted polymer upon injection of template (Levi et al., 1997). Unfortunately, not all analytes can be easily modified with dyes and the modification itself can change the affinity of analyte. Recently, the displacement of non-specific dyes from a MIP has been used for the detection and quantification of ligand-polymer binding events (Piletsky et al., 1999b). Rhodamine B solution was passed through an HPLC column with L-Phe-amidespecific MIP. When template was injected, part of the dye was competitively replaced by the analyte from the MIP. This displacement peak was three times higher for the template than for the opposite enantiomer. This approach can be considered as general, suited for different kinds of templates, dyes, and polymers. A similar displacement principle was used also in combination with electrochemical measurements for template detection (Kroger et al., 1999). It is proposed that the displacement of non-specific indicator molecules from a set or array of MIPs could be used to develop multisensors. The affinity sensors described above are able to detect templates that possess a specific property such as optical absorbance, fluorescence, or electrochemical activity. Direct detection of "inert" templates can be realised in receptor sensors. 2.2. Receptor sensors based on MIPs

Two approaches exist for the development of receptor-like MIP sensors. One is 412

Materials Based on Imprinted Polymers

0.97 0.92 <

0.87 0.82 0.77

~ 0.72 0

~

0.67

/

Glucose/mannose

0.62 0.57 0

1

2

3

4

5

6

7

8

9

10

Sugar, uM Figure 8. Influence of sugars on the formation of fluorescent isoindole complex between OPA reagent and amino-functionality in the polymer imprinted with sialic acid (Piletsky et al., 1996).

connected with an MIP's ability to change conformation upon binding with template, leading to change in a measurable property, such as conductivity, permeability, or surface potential (Piletsky et al., 1998). A second principle is based on the ability of a functional monomer to change its property upon interaction with template, most frequently, fluorescence (Rathbone et al., 2000). Receptor properties of imprinted materials were first reported in 1992 (Piletsky et al., 1992a). It was shown that templates such as amino acids, nucleic acids and cholesterol increase the transport of ions passing through the imprinted membranes (Piletsky et al., 1994b; Piletsky et al., 1998). This so-called "gate effect" has been used for quantification of the concentration of templates. Most often, MIP-based receptor sensors measure the change in membrane electroconductivity, induced by specific interaction of MIPs with template molecules. Sensors specific for L-phenylalanine, cholesterol, sialic acid and atrazine have shown high selectivity and sensitivity at the micromolar and even nanomolar range (Sergeeva et al., 1999). The "gate effect" also can be probed using optical detection (Piletsky et al., 1996). In this work, imprinted polymer based on allylamine was imprinted with sialic acid. When polymer suspension was brought into the contact with OPA reagent, a mixture of o-phtaleic dialdehyde and mercaptoethanol, a fluorescent complex was formed. The kinetics of complex formation depended on the presence of the template, sialic 413

Piletsky and Turner

E 4.5

3.5

2.0

9

I

480

E

I,

,

.

I

s2o 540 wavelength [nm]

..-

4.s

4.O

,

500

,,

I

,

I

sso

56o

"....,.,.

/

3.5 3.0 2.5 2.0 1.5 "., .... ,,,,,

1.0 9

,=80

=

~o

m

I

~o

m

I

54o

I

~

9

I

,~o

9

I

600

wavelength [nm]

Figure 9. Relative emission spectra of polymers containing fluorescein (MIP-F), eosin (MIP-E) and their mixture (MIP-E,F) (top); and spectral changes of donor and acceptor emission in MIP-E,F induced by 10 ~M concentration of L-phenylalanineamide (bottom) Wolfbeis et al., 1997).

acid, which modulated the diffusion of soluble components to the reactive sites (Figure 8). The polymer was able to discriminate sialic acid from other sugars such as glucose and mannose. Potentially, it should be possible to design a sensor where the "gate effect" would be used for direct monitoring of conformational changes in imprinted polymers. Imprinted polymer can be labeled with two different chromophores, one being the donor, the other the acceptor. The light energy adsorbed by the chromophore may be dissipated nonradiatively via a mechanism known as fluorescence resonance energy (F/3rster) transfer (FRET), which is sensitive to intramolecular and intermolecular interactions (Lakowicz et al., 1993). FRET occurs through induction of a dipole oscillation in the unexcited acceptor by the excited-state 414

Materials Based on Imprinted Polymers

CH3 I

0.01

~ NC ' H3

Phe no I b lue N

0.008 o 0.006 e~ 0

0.004 0.002

I

0

-

I

I

I

I

I

2

4

6

8

10

C solvent, o/oo Figure 10. Absorbance vs solvent concentration in air for a polymer with polymerimmobilised phenol blue exposed to tetrahydrofuran (e), chloroform (m) and ethanol (,) (Dickert et al., 2000).

donor chromophore. The rate of energy transfer between the chromophores is a function of the sixth power of the distance between the donor and acceptor. Because of the strong distance dependence of energy transfer, monitoring the fluorescent intensity in a system of labeled polymers can be used to quantify the concentration of a template. In our experiments, imprinted polymers specific for L-phenylalanineamide were labeled with fluorescein (donor), eosin (acceptor) and their mixture (Wolfbeis et al., 1997). The emission spectra of the resulting polymers are shown in Figure 9, top. It can be seen that there is strong FRET occurring in the polymer labeled with both chromophores. The added template decreased the distance between donor and acceptor as a result of the polymer shrinking, which led to an increase in fluorescent emission (Figure 9, bottom). The sensitivity of the system can be further improved using donor-acceptor pairs with different Fi3rster distances. An interesting approach for the design of signalling polymers and their use in sensors was proposed by Cooper and co-authors (1997). They used an environment sensitive functional monomer (I) integrated into a cross-linked matrix, which was able to change its fluorescent properties in the presence of compounds with proton donor properties. 415

Piletsky and Turner

H3C, N~CH3

oH

N

I NX> OH N

OSCH2 O I

|

CH2

OH

0H

|

O

Figure 11. Schematic representation of polymer binding site for cAMP with signalling trans-4-[p-(N,N-dimethylamino)styryl]-N-vinylbenzylpyridinium monomer (Turkewitsch et al., 1998).

CMe

Me

CH2--CHCOO

OOCCH=CH2

Strong quenching of fluorescent emission induced by hydrogen bonding might be used for recognition of polar templates. Similarly betaine dyes with strong hypsochromic effect for protic solvents were used for gas phase analysis (Figure 10) (Dickert et al., 2000). Environmentally-sensitive dye has been used in design of a fluorescent sensor for cAMP detection (Turkewitsch et al., 1998). In this case a fluorescent dye, trans4-[p-(N,N-dimethylamino)styryl]-N-vinylbenzylpyridinium chloride, was copolymerised with cross-linker and template. The resulting polymer displayed two functions simultaneously: template recognition and sensing (Figure 11). 416

Materials Based on Imprinted Polymers A new type of proximity scintillation assay has been developed for (S)propranolol (Ye and Mosbach, 2001). A scintillation monomer, 4hydroxymethyl-2,5-diphenyloxazole acrylate (II), has been covalently incorporated into MIP microparticles during the imprinting reaction. This monomer is capable of transforming 13-radiation from the bound tritium-labeled template into a fluorescent signal. The small size of the particles (0.6-1 ~tm) guarantees that the reporter group, randomly distributed throughout the polymer matrix, is located in close proximity to the MIP binding site for signal generation.

.CH2OCOCH==CH2 II

A reverse scenario can be used, in principle, for the quantification of the concentration of environment-sensitive templates (Matsui et al., 2000). Fluorescent spectra of the cinchona alkaloids exhibit a characteristic shift through binding to these polymer particles, containing acidic m o n o m e r - 2(trifluoromethyl)acrylic acid (TFMAA). The authors demonstrated the possibility of using TFMAA-based imprinted polymers as polymer reagents for analysis of the cinchona alkaloid bound to the polymers without bound/free separation (Figure 12).

0

0

III

Fluorescent polymers containing monomer (III) were imprinted with various N abenzylidene pyridine-2-carboxamidrazones (Rathbone and Ge, 2001). Dramatic quenching of fluorescence approaching background levels was observed for most

417

Piletsky and Turner | .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

,,

.

.=,

1

..g

5:000

h

4O0O

po

:C o C~

30.00

(-)-cinchonidine

0

=

2:000

(9

> r

1000

340

390

440

490

540

5.90

Wav etength (rim) Figure 12. Fluorescence spectra of MIP suspension incubated with cinchonidine in chloroform/acetonitrile at concentrations of (mM): a) 0; b) 0.005; c) 0.02; d) 0.05; e) 0.10; f) 0.15; g) 0.25; h) 0.50. (~,ex=330nm) (Matsui et al., 2000).

cases where the "empty" MIP was re-exposed to its template. The authors claim that this approach is suitable for high throughput screening. Jenkins and coauthors developed a very sensitive lanthanide-based luminescent sensor for sarin and soman with a detection limit of 7 ppt (Jenkins et al., 1998). The sensor functions by selectively and reversibly binding the phosphonate hydrolysis product of this agent to a MIP containing a coordinatively bound Eu 3§ ion. This binding leads to the appearance of a narrow luminescence band in the 610-nm region of the Eu 3§ spectrum, which can be monitored using a miniature spectrometer. A high degree of selectivity is obtained by combining both chemical and spectroscopic selectivities. Very promising combinations of group-specific fluorescent reporters with template-specific MIPs were reported for sugars (Wang et al., 1999), carboxylic 418

Materials Based on Imprinted Polymers acids (Zhang et al., 2001), and primary amines (Subrahmanyam et al., 2000). In the first two cases, the anthracene reporting group was modified in order to introduce polymerisable and recognition functionalities. The interactions of boronic acid (IV) with cis-dioles and guanidine (V) with carboxylic acids is non-specific by its nature. Nevertheless they could be made specific by incorporating these monomers into specific binding sites created by imprinting.

0

\

//

_ / N\

B(OH)2 IV

Similarly non-specific interaction between thioacetale and primary amines, which leads to formation of fluorescent isoindole complex, was made specific for creatine by imprinting in the presence of methylated analogue of the template (Subrahmanyam et al., 2000) (Figure 13).

/

/~

H

H2N

u

Diode lasers are instruments of choice for the development of fluorescent sensors: their light flux is coherent and allows better integration with waveguides, and they are inexpensive and small. However, diode lasers necessitate polymerisable fluorescent markers with long wave adsorption and emission, which at present are scarce. A further problem associated with the development of MIP-based optical sensors is light scatter due to heterogeneity in polymer structure.

419

Piletsky and Turner

S

~

S OH H

+

H2N-R

O

Figure 13. Complex formation between polymerisable thioacetale and primary amine, and recognition of analytes by Blank polymer and polymer imprinted with methylated creatine analogue (Subrahmanyam et al., 2000).

One way to overcome this problem is optimisation of the polymerisation conditions (first of all by choice of solvent and polymerisation temperature) which will lead to synthesis of optically transparent and homogeneous materials. Another way is to measure the decay time of luminescence, rather than its intensity. This approach is highly advantageous because measurements of decay time are less affected by light scattering, analyte concentration and detector sensitivity.

420

Materials Based on Imprinted Polymers 3. Market Potential of MIP Sensors and Future Prospects

Three particular properties make commercial application of MIP sensors attractive: (i) polymers are highly stable and can be autoclaved; (ii) they are fully compatible with microfabrication technology, and (iii) the low cost of the materials and easy processes of polymer preparation in comparison with natural and other artificial receptor systems. The most promising areas of MIP sensor applications are: 1. chemical and pharmaceutical manufacturing: using MIP sensors in extreme conditions (high and low pH's, toxic solvents and high temperature, pressure, and radiation); 2. medicine and pharmaceuticals: application of MIPs mimicking natural receptors for drug screening and for in vivo monitoring; 3. environment: remote sensing, continuous emissions sensors and pointsource monitors; 4. defence: rapid detection of chemical and biological warfare agents under battlefield and civil conditions; 5. deep ocean and space exploration: sensors for analysis of extreme environments. Several key problems associated with MIP development need to be addressed, however, before the successful commercialisation can commence. The issues include: 1. development and validation of a general protocol for MIP design; 2. development of MIPs capable of effective functioning in water; 3. the need for a substantial increase in polymer affinity and improvement of the ratio between specific and non-specific binding; 4. development of effective immobilisation protocols. With further progress in polymer science and engineering we can expect to see the appearance of a new generation of MIP sensors which will gradually replace traditional biosensors and chemical sensors in many areas of biotechnology and pharmacology, environmental, clinical and food analysis.

4. References Agmon, N., 2000, J. Phys. Chem. B 104, 7830. Alexander, C. and E. N. Vulfson, 1997, Adv. Mater. 9, 751. Andersson, L. I., R. Muller, G. Vlatakis and K. Mosbach, 1995, Proc. Natl. Acad. Sci USA, 92, 4788. Andersson, L. I., 1996, Anal. Chem. 68, 111. Andersson, L. I., C. F. Mandenius and K. Mosbach, 1988, Tetrahydron Lett. 29, 5437. 421

Piletsky and Turner Andersson, L. I., 2000, J. Chromatogr. B 739, 163. B lanchard, P., L. Huchet, E. Levillain and J. Roncali, 2000, Electrochem. Commun. 2, 1. Boyle, A., E. M. Genies and M. Lapkowski, 1989, Synth. Metals. 28, C769. Braco, L., K. Dabulis and A. M. Klibanov, 1990, Proc. Natl. Acad. Sci. USA 87, 274. Cooper, M. E., B. P. Hoag and D. L. Gin, 1997, Polym. Prepr. 38, 209. Dhal, P. K., S. Vidyasankar and F. H. Arnold, 1995, Chem. Mater. 7, 154. DickeR, F. L., H. Besenbock and M. Tortschanoff, 1998, Adv. Mater. 10, 149. Dickert, F. and M. Tortschanoff, 1999, Anal. Chem. 71, 4559. Dickert, F., U. Geiger, P. Lieberzeit and U. Reumer, 2000, Sens. Actuators B 70, 263. Dickert, F. L., O. Hayden and K. P. Halikias, 2001, Analyst 126, 766. Dong, J., Y. S. Lin, M. Z.-C. Hu, R. A. Peascoe and E. A. Payzant, 2000, Micropor. Mesopor. Mat. 34, 241. Giraudi, G., C. Giovannoli, C. Tozzi, C. Baggiani and L. Anfossi, 2000, Chem. Commun. 1135. Hishiya, T, M. Shibata, M. Kakazu, H. Asanuma and M. Komiyama, 1999, Macromolecules 32, 2265. Hjerten, S., J. L. Liao, K. Nakazato, Y. Wang, G. Zamaratskaia and H. X. Zhang, 1997, Chromatograph. 44, 227. Hutchins, R. S. and G. Bachas, 1995, Anal. Chem. 67, 1654. Jenkins, A. L., O. M. Uy and G. M. Murray, 1998, Anal. Chem. 71,373. Ji, H. S., S. McNiven, K. H. Lee, T. Saito, K. Ikebukuro and I. Karube, 2000, Biosens. Bioelectron. 15,403. Katz, A. and M. E. Davis, 2000, Nature 403,286. Kempe, M. and K. Mosbach, 1995, J. Chromatogr. A, 691, 317. Kimaro, A., L. A. Kelly and G. M. Murray, 2001, Chem. Commun. 1282. Koshland, D. E., 1995, Angew. Chem. Int. Ed. 33, 2375. Kriz, D. and K. Mosbach, 1995, Anal. Chim. Acta 300, 71. Kriz, D., O. Ramstrom, A. Svensson and K. Mosbach, 1995, Anal. Chem. 67, 2142. Kroger, S., A. P. F. Tutner, K. Mosbach and K. Haupt, 1999, Anal. Chem. 71, 3698. Lakowicz, J. R., W. Wiczk, I. Gryczynsky, M. Fishman and M. L. Johnson, 1993, Macromolecules 26, 349. Lanza F. and B. Sellergren, 1999, Anal. Chem. 71, 2092. Lele, B. S., M. G. Kulkarni and R. A. Mashelkar, 1999, React. Functional Polym. 39, 37. Levi, R., S. McNiven, S. A. Piletsky, S.-H. Cheong, K.Yano and I. Karube, 1997, Anal. Chem. 69, 2017. Lubke, C., M. Lubke, M. J. Whitcombe and E. N. Vulfson, 2000, Macromolecules, 33, 5098. Malitesta, C., I. Losito and P. G. Zambonin, 1999, Anal. Chem. 71, 1366. MathewKrotz, J. and K. J. Shea, 1996, J. Am. Chem. Soc. 118, 8154. 422

Materials Based on Imprinted Polymers

Matsui, J., I. A. Nicholls, T. Takeuchi, K. Mosbach and I. Karube, 1996, Anal. Chim Acta 335, 71. Matsui J., H. Kubo and T. Takeuchi, 2000, Anal. Chem. 72, 3286. Mayes, A. G. and K. Mosbach, 1997, TrAC 16, 321. Mirsky, V. M, T. Hirsch, S. A. Piletsky and O. S. Wolfbeis, 1999, Angew. Chemie, Intern. Ed. 38/8, 1108. Morita, M., O. Niwa and T. Horiuchi, 1997, Electrochim. Acta 42, 3177. Nicholls, I. A., 1995, Chem. Lett. 1035. Nicholls, I. A., O. Ramstrom and K. Mosbach, 1995, J. Chromatogr. A, 691,349. O'Shannessy, D. J., B. Ekberg and K. Mosbach, 1989, Anal. Biochem. 177, 144. Panasyuk, T. L., V. M. Mirsky, S. A. Piletsky and O. S. Wolfbeis, 1999, Anal. Chem. 71, 4609. Peissker, F. and L. Fischer, 1999, Bioorg. Med. Chem. 7, 2231. Piletska E. V., S. A. Piletsky, S. Subrahmanyam, I. A. Nicholls and A. P. F. Turner, 2000, Proc. 1st Int. Workshop on Molecular Imprinting, Cardiff, UK, 2000, 87. Piletsky, S. A., D. M. Fedoryak and V. P. Kukhar, 1990a, Dokl. Acad. Sci. Ukraine B. 4, 53 (in Russian). Piletsky, S. A., I. Ya. Dubey, D. M. Fedoryak and V. P. Kukhar, 1990b, Biopolym. Cell 6, 55 (in Russian). Piletsky, S. A., I. A. Butovich and V. P. Kukhar, 1992a, Zh. Anal. Khim. 47, 1681 (in Russian). Piletsky, S. A. and N. F. Starodub, 1992b, Zh. Anal. Khim. 47, 623 (in Russian). Piletsky, S. A., Ya. I. Kuris', A. E. Rachkov and A. V. Erskaya, 1994a, Russ. J. Electrochem. 30, 1090 (in Russian). Piletsky, S. A., Yu. P. Parhometz, T. L. Panasyuk and A. V. El' skaya, 1994, Sens. Actuators B. 18/19, 629. Piletsky, S. A., E. V. Piletska, K. Yano, A. Kugimiya, A. V. Elgersma, R. Levi, U. Kahlow, T. Takeuchi, I. Karube, T. L. Panasyuk and A. V. El' skaya, 1996, Anal. Lett. 29, 157. Piletsky, S. A., E. V. Piletska, A. V. El' skaya, R. Levi, K. Yano and I. Karube, 1997, Anal. Lett. 30, 445. Piletsky, S. A., E. V. Piletskaya, T. L. Panasyuk, A. V. El' skaya, R. Levi, I. Karube and G. Wulff, 1998, Macromolecules 31, 2137. Piletsky, S. A., E. V. Piletskaya, T. A. Sergeeva, T. L. Panasyuk and A. V. El' skaya, 1999a, Sens. Actuators B 60, 216. Piletsky, S. A., E. Terpetschnig, H. S. Andersson, I. A. Nicholls and O. S. Wolfbeis, 1999b, Fresenius J. Anal. Chem. 364, 512. Piletsky, S. A., H. Matuschewski, U. Schedler, A. Wilpert, E. V. Piletska, T. A. Thiele and M. Ulbricht, 2000a, Macromolecules 33, 3092. Piletsky, S. A. E. V. Piletska, B. Chen, K. Karim, D. Weston, G. Barrett, P. Lowe and A. P. F. Turner, 2000b, Anal. Chem. 72, 4381. Piletsky, S. A., R. M. Day, B. Chen, S. Subrahmanyam, O. Piletska and A. P. F. Turner, 2000c, UK patent application 0001513.1. 423

Piletsky and Turner Piletsky, S. A., K. Karim, E. V. Piletska, C. J. Day, K. W. Freebairn, C. Legge and A. P. F. Turner, 2001, Analyst, in press. Polyakov, M. V., 1931, Zhur. Fiz. Khim. 2, 799 (in Russian). Rachkov, A., S. McNiven, A. V. Erskaya, K. Yano and I. Karube, 2000, Anal. Chim. Acta 405, 23. Ramstrom, O., L. I. Andersson and K. Mosbach, 1993, J. Org. Chem. 58, 7562. Rathbone, D. L., D. Su, Y. Wang and D. C. B illington, 2000, Tetrahedron Lett. 41, 123. Rathbone, D. L. and Y. Ge, 2001, Anal. Chim. Acta 435,129. Sergeyeva, T. A., S. A. Piletsky, A. A. Brovko, E. A. Slinchenko, L. M. Sergeeva and A. V. Erskaya, 1999, Anal. Chim. Acta 392, 105. Sergeyeva, T. A., H. Matuschewski, S. A. Piletsky, J. Bendig, U. Schedler and M. Ulbricht, 2001, J. Chromatogr. A 907, 89. Shi, H. Q., W. B. Tsai, M. D. Garrison, S. Ferrari and B. D. Ratner, 1999, Nature 398, 593. Starodub, N. F., S. A. Piletsky, N. V. Lavryk and E. V. El' skaya, 1992, Sens. Actuators B 13-14, 708. Steinke, J. H. G., I. R. Dunkin and D. C. Sherrington, 1996, Macromolecules 29, 407. Strikovsky, A. G., D. Kasper, M. Grtin, B. S. Green, J. Hradil and G. Wulff, 2000, J. Am. Chem. Soc. 122, 6295. Su~irez-Rodrfguez, J. L. and M. E. Dfaz-Garcia, 2000, Anal. Chim. Acta, 405, 67. Subrahmanyam, S., S. A. Piletsky, E. V. Piletska, B. Chen, R. Day and A. P. F. Turner, 2000, Adv. Mater. 12, 722. Surugiu, I., B. Danielsson, L. Ye, K. Mosbach and K. Haupt, 2001, Anal. Chem. 73,487. Svenson, J. and I. A. Nicholls, 2001, Anal. Chim. Acta 435, 19. Tabushi, I., K. Kurihara, K. Naka, K. Yamamura and H. Hatakeyama, 1987, Tetrahedron Lett. 28, 4299. Takeuchi, T., D. Fukuma and J. Matsui, 1999, Anal. Chem. 71,285. Takeuchi, T., A. Seko, J. Matsui and T. Mukawa, 2001, Instrum. Sci. Technol. 29, 1. Turner, A. P. F., 1999, Biosensor: McGraw-Hill Yearbook of Science and Technology, McGraw-Hill, New York. Vidiasankar, S., M. Ru and F. H. Arnold, 1997, J. Chromatogr. A, 775, 51. Vinokurov, I. A. and M. A. Grigoreva, 1990, Zh. Anal. Khim. 45, 1009 (in Russian). Wang, H. Y., T. Kobayashi, T. Fukaya and N. Fujii, 1997, Langmuir 13, 5396. Wang, W., S. Gao and B. Wang, 1999, Org. Lett. 1, 1209. Watanabe, M., T. Akahoshi, Y. Tabata and D. Nakayama, 1998, J. Am. Chem. Soc. 120, 5577. Whitcombe, M. J., L. Martin and E. N. Vulfson, 1998, Chromatogr. 47, 457. Wolfbeis, O. S., E. Terpetschnig, S. A. Piletsky and E. Pringsheim, 1998, In Iimprinted Polymers: Applied Fluorescence in Chemistry, Biology and

424

Materials Based on Imprinted Polymers Medicine, Eds. W. Rettig, B. Strehmel, S. Schrader, H. Seifert, Springer, Berlin-Heidelberg. Wulff, G. and J. Haarer, 1991, Makromol. Chem. 192, 1329. Wulff, G., 1995, Angew. Chem. Int. Ed. Engl. 34, 1812. Ye, L. and K. Mosbach, 2001, J. Am. Chem. Soc. 123, 2901. Yoshida, M., Y. Hatate, K. Uezu, M. Goto, S. Furusaki, 2000, J. Polym. Sci. Pol. Chem. 38, 689. Yoshizako, K., K. Hosoya, Y. Iwakoshi, K. Kimata, N. Tanaka, 1998, Anal. Chem. 70, 386. Yu, C., O. Ramstrom, K. Mosbach, 1997, Anal. Lett. 30, 2123. Yu, C., K. Mosbach, 2000, J. Chromatogr. A 888, 63. Zhang, H., W. Verboom, D. N. Reinhoudt, 2001, Tetrahedr. Lett. 42, 4413.

425

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 13

NEW MATERIALS BASED ON IMPRINTED POLYMERS AND THEIR APPLICATION IN OPTICAL SENSORS

SERGEY A. PILETSKY, PH.D* AND ANTHONY P.F. TURNER, PH.D., D.Sc.

Institute of BioScience and Technology, Cranfield University, Silsoe, Bedfordshire, MK45 4DT, UK.

Molecular imprinting is the process of template-induced formation of specific recognition sites (binding or catalytic) in a material where the template directs the positioning and orientation of the material's structural components by a self-assembling mechanism. Synthetic receptors prepared using molecular imprinting possess a unique combination of properties, such as high affinity, specificity, low price and robustness, which make them an attractive alternative to natural receptors, enzymes and antibodies used in biosensors. This review gives a brief overview of the technology with specific emphasis on the mechanisms underlying the ability of imprinted polymers to perform highly selective functions such as recognition and transformation of a binding event into a detectable optical signal. The problems associated with the application of molecularly imprinted polymers (MIPs) in sensors are highlighted. Possible solutions to these problems are discussed and recommendations made about where commercial application of imprinted sensors seems most feasible in the near future.

1. Molecular Imprinting The molecular imprinting approach exploits the formation of a complex between a template molecule and functional monomers, which is fixed by copolymerisation with cross-linker into a growing polymer network (Figure 1). Following removal of the template, binding sites are left in the polymers, which have the shape and orientation of functional groups, complementary to those of the template molecule (Wulff, 1995; Mayes and Mosbach, 1997). 397

Piletsky and Turner

Figure 1. Scheme of molecularly imprinted polymerization.

1.1. Different formats used for design of imprinted materials The typical recipe for MIP preparation includes mixing together target compound -template with corresponding functional monomer (most f r e q u e n t l y methacrylic acid) and cross-linker (e.g., ethylene glycol dimethacrylate) in appropriate solvent (chloroform, acetonitrile) and polymeristion of this mixture using UV or chemical initiation (O'Shannessy et al., 1989). The template can be extracted from the polymer by washing or by electrophoresis (Piletsky et al., 1992a). Subsequent polymer grinding and washing yields the polymer particles with receptor sites on the accessible surface. Other formats of molecular imprinting include: 1. Polycondensation of silica acid in the presence of template (Katz and Davis, 2000); 2. Electropolymerisation (Malitesta et al., 1999); 3. Formation of 2-dimensional templated monolayers onto a SIO2, metal oxide or gold surface (Starodub et al., 1992; Mirsky et al., 1999); 4. Grafting of imprinted polymers to the inert solid surface (Dhal et al., 1995; Piletsky et al., 2000a) (Figure 2); 5. Templating of a pre-formed polymer structure by precipitation or crosslinking in the presence of template (Braco et al., 1990; Peissker and Fischer, 1999); 6. Formation of imprinted poly/oligomers (e.g. peptides) in the presence of template (Giraudi et al., 2000; Piletska et al., 2000). 398

Materials Based on Imprinted Polymers

Figure 2. Scheme of MIP synthesis via surface photografting onto porous polymeric substrate (Piletsky et al., 2000a) Historically, formation of imprinted silica gels was the first example of molecular imprinting (Polyakov, 1931). Despite the fact that this specific technique reached its peak in the sixties and is now in the process of gradual decline, due to limited flexibility of the method, it still remains the most popular choice for the preparation of specific zeolites (Dong et al., 2000). Electropolymerisation faces the same type of problem as silica imprinting due to limited number of polymerisable functional monomers available, which are selected mainly from the group of aniline, phenol, pyrrole and thiophene (Panasyuk et al., 1999; B lanchard et al., 2000). Electropolymerisation retains its attraction, however, because it provides a means for precise deposition of a sensitive layer on an electrode surface, which is extremely important for microand multisensor production. Electropolymerised MIPs have been used almost exclusively in potentiometric (Boyle et al., 1989; u et al., 1990) and amperometric sensors (Piletsky et al., 1994a). Two-dimensional MIPs or imprinted monolayes were developed and used in optical sensors by Andersson and co-authors (1988). They used Tabushi's method (Tabushi et al., 1987) to ~immobilise octadecylchlorosilane in the presence of inert template hosts (n-hexadecane) onto a silicon oxide surface. 399

Piletsky and Turner After the extraction of hosts, vitamin K1 was detected by ellipsometry. Expanding this method for the preparation of monolayers, imprinted with watersoluble templates, we developed materials selective for amino and nucleic acids (Piletsky and Starodub, 1992b). This approach involves two steps" first, adsorption of the template on the surface of SiO2 or metal oxide; and second, treatment of the surface with adsorbed template by trimethyl chlorosilane from the gas phase. In another similar approach, a gold surface was imprinted with a cholesterol-specific monolayer using co-adsorption of the template with hexadecylmercaptane (Piletsky et al., 1999a). Despite some advantages, such as fast sensor response and easy preparation, these systems, however, suffer from lack of stability. The lateral mobility of the components of imprinted monolayer is responsible for steady decrease in the specificity of imprinted cavities. An essential improvement of the sensor stability was achieved by co-immobilisation of the template in imprinted layer. Using a new approach called "spread bar architecture design," it was possible to develop stable monolayers, consisting of template - thiobarbituric acid - and functional monomer- hexadecylmercaptane. A depression in the hexadecylmercaptane layer formed by the template was able to accommodate barbituric acid, changing electrode capacitance in the binding proces (Mirsky et al., 1999). This two-dimensional format for MIP design is particularly attractive for evanescence-wave sensing, e.g. surface plasmon resonance. Several reports on the preparation of MIPs by surface grafting have appeared, where a thin imprinted layer, most frequently a monolayer, is formed on a solid support (Dhal et al., 1995; Lele et al., 1999; Piletsky et al., 2000a). Grafting can be performed using chemical, UV or plasma initiation (Shi et al., 1999; Piletsky et al., 2000b). The advantage of this approach lies in the possibility of modifying very inert surface (polystyrene, polypropylene, etc.) with specific polymers. The additional attraction of electropolymerisation and grafting methods is their convenient format, which does not require an additional processing step. MIP synthesis and immobilisation is performed as a one-step procedure, directed by applied potential or by exposing the monomer mixture-coated detector to UV light (Figure 2). A further approach, frequently called "bioimprinting", involves precipitation or cross-linking of biological molecules (proteins) in the presence of template (Braco et al., 1990; Peissker and Fischer, 1999). The conformation adopted by interacting biopolymer around the template remains fixed after template extraction with an appropriate solvent. Although the authors are unaware of any examples where bioimprinting has been used for sensor design, this technique could potentially be useful for introducing either additional recognition sites into enzymes or catalytic sites into antibodies. These chimeric molecules might possess the combined characteristics of antibodies and enzymes and, in this way, be useful for the development of new, label-free types of assays and sensors. 400

Materials Based on Imprinted Polymers The last format of molecular imprinting is template-directed synthesis. This process includes the formation of a new substance by a chemical modification of the substrate, or by the coupling of two or more molecules, in the presence of a template to serve as a pattern for the formation of a new structure. The most well known example of this process is gene replication. An important issue is that the synthesised molecule always has a structure, complementary to that of template, which can be exploited for the synthesis of biospecific ligands or to obtain information about the structure and properties of the template molecule. This approach is actively pursued in molecular biology (gene sequencing) and in DNA sensors where complementary DNA or RNA chains are synthesised using transcription facilitated by enzymes such as DNA polymerase or reverse transcriptase. Unfortunately, a similar technique does not exist for the analysis of molecules other than DNA, such as proteins and polysaccharides. It is possible, however, to produce a complementary ligand for a target molecule using a synthetic approach (Giraudi et al., 2000; Piletska et al., 2000). The method involves the formation of oligomers, e.g., peptides, in the presence of template. Prior to the initiation of polymerisation, and during polymerisation, the monomers, which could be amino acids or nucleotides, spatially distribute themselves around the template molecules in accordance with the size, polarity and functionality of the template. The monomers are polymerised into linear, water-soluble oligomers specific for the template. The advantage of this approach is the possibility of obtaining watersoluble ligands, which can be treated in the same manner as antibodies and other natural receptors. Not withstanding the new methods detailed above, traditional bulk polymerisation remains the most popular choice for the preparation of molecularly imprinted polymers for theoretical study and practical application in separation and sensing.

1.2. Mechanism of template recognition by imprinted polymer Three major factors determine the recognition process" the quantity of the functional groups participating in the interaction, their correct arrangement within the cavity, and the shape of the cavity itself. The types of interactions explored in molecular imprinting include reversible covalent bonds (Wulff and Haarer, 1991), electrostatic interactions (ionic and hydrogen bonds) (Piletsky et al., 1990a; Nicholls et al., 1995), van der Waals (Dickert et al., 1998), hydrophobic interactions (Yu et al., 1997), and metal chelation (Matsui et al., 1996) (Figure 3). The shape of the cavity alone can provide specificity (Yoshizako et al., 1998) although the specificity is substantially better when the

401

Piletsky and Turner

OH) 2 +

~

+ H20

(a)

H

coo. § R.H ~

oO'U..~" .rid ~;H

(b)

+ ~N.,.R

0

~H__R

Figure 3. Different types of interactions explored in molecular imprinting: (a) reversible covalent bond formation; (b) electrostatic interactions; (c) metal chelation.

template interacts with one or more properly oriented functional monomers (Ramstrom et al., 1993). The required strength of monomer-template interaction varies depending on the size and the structure of the template. For a small template molecule, the presence of strong interactions, preferably ionic and/or hydrogen bonds, is critically important. For a large molecule such as a protein or nucleic acid, successful results can be achieved with a combination of multiple weak interactions (Hjerten et al., 1997). The choice of solvent depends on the type of interaction. Thus if template recognition depends on hydrogen bond formation, better results can be achieved if both polymer synthesis and re-binding takes place in a hydrophobic solvent, where hydrogen bonds are stronger (Andersson, 1996; Yu and Mosbach, 2000). The equilibrium dissociation constants (Kd) for the binding of ligands to their corresponding polymers have been estimated by Scatchard plot analysis of binding data. Mostly, non-linear plots were obtained because of multiple Kd values, varying in range in the majority of cases from micromolar to nanomolar. In a similar way to polyclonal antibodies, imprinted polymers contain a heterogeneous population of binding sites (Wulff, 1995). One of the important components of the recognition mechanism observed in MIP systems is the conformational change in the polymer induced by template interaction (Piletsky et al., 1992a; Watanabe et al., 1998; Wolfbeis et al., 1998). Depending on experimental conditions (solvent, temperature, and types of the monomer-template and monomer-monomer interactions), the polymer matrix can 402

Materials Based on Imprinted Polymers Table 1. Comparison of natural antibodies and receptors with MIPs. ..

..,.

,,,

Property

,

Natural Biomolecules .,

Stai~iiity

Low

Cost

High

Integrati& into multisensor unit

Compatibility with micromaching technology/.miniaturisation Spectrum of analytes

i

|,|

,

n

1=,.

i

-

MIPs

,

' Difficult due to integration OiY" natural biomolecules in multisensor unit is difficult due to different operational requirements of these molecules (pH, ionic strength, temperature, subs tsr,ate) Poor

Stable at low/high pHs, pressure and temperature . Inexpensive and easy preparation Flexible MIP design allows preparation of MIPs against many combinations of analytes Fully Compatible Practicaliy unlimited

Limited

i

ii

shrink or swell in the presence of template. The mechanism is similar to "induced fit" observed for natural enzymes and receptors (Koshland, 1995; Agmon, 2000). The importance of this effect for sensor technology lies in the possibility of use for measuring template concentration (Piletsky et al., 1998). An additional factor contributing to MIP recognition properties is the presence of nanopores in the polymer structure with specificity for the template molecules (Piletsky et al., 1990b; MathewKrotz and Shea, 1996). Membranes prepared by molecular imprinting possess selective permeability for the imprinted species and can be used for purification of desirable analytes or removal of potential interfering compounds. MIPs are capable of recognising small variations in the structure of the template and the specificity of imprinted polymers under optimised conditions is often equal to or even superior to that of natural enzymes and receptors (Andersson et al., 1995). However, quite often MIPs demonstrate a high level of non-specific binding. Although they quite often referred to in this way, it would be a mistake to see imprinted polymers as "plastic" antibodies or receptors. They are different materials with their own advantages and disadvantages and thus should be considered as additive, complementary systems rather than substitutes (Table 1).

403

Piletsky and Turner 1.3. "Pluses" and "minuses" in MIP technology and their comparison with natural enzymes and receptors

Being purely synthetic materials, it is natural that the imprinted polymers have a much higher stability than enzymes and receptors. The reason for this lies, first of all, in the high level of cross-linking, which provides adequate protection for binding sites created in the polymer by imprinting. Imprinted polymers can withstand harsh treatment with acidic and basic solutions or with organic solvent. They are stable under both high and low pressure, and, as well as at extreme temperatures (Kriz and Mosbach, 1995; Svenson and Nicholls, 2001). Imprinting polymerisation is a very inexpensive procedure for the development of artificial receptors. In the majority of cases, the price of a MIP depends almost entirely on the price of the template used. Furthermore, if the templates themselves are expensive, it maybe possible to recover the template and use it again. Alternatively, inexpensive template analogues can be used for the preparation of MIPs. Generally speaking, MIP preparation is three-to-four orders less expensive than production of the equivalent natural receptor, and this makes the technology very competitive. The possibility of using MIPs in organic solvents opens new areas of application such as biomimetic sensing and catalysis in chemical and pharmaceutical manufacturing. Quality control and on-line monitoring of manufacturing processes are particularly attractive. One of the most challenging problems associated with development of multisensors is related to the significant differences in the performance of natural enzymes and receptors. These biological materials all have different stability, activity and sensitivity; in many cases, they require different substrates and buffers with different ionic strengths and pHs. Due to such factors, the integration of naturally occurring bio-molecules in one single unit maybe problematic. Since a MIP's design is flexible and variety of monomers are available for their preparation, it is possible to develop a set of polymers specific for a range of templates which will have almost identical operational requirements (solvent, temperature, pH, etc.). An additional benefit comes from the possibility of processing MIPs in the same way as traditional photoresist materials. MIPs can be immobilised at precise spots on the detector surface using masks and photopolymerisation. The compatibility of MIPs with micromachining technology makes MIP-based multisensors feasible. Last, but not least, is the ability to develop MIPs for practically any type of compound. Examples of templates producing MIPs successful include inorganic ions, drugs, nucleic acids, proteins and even cells (Table 2). Although antibodies 404

Materials Based on Imprinted Polymers Table 2. Examples of templates used in molecular imprinting. i|

Ill

i

i

Template

Application

Reference

Amino acids and derivatives

Separation, sensors

Kempe and Mosbach, 1995; Vidiasankar et al., 1997; Piletsky et al., 1998

Aniline, phenol, derivatives

Sensing

Vinokurov and Grigoreva, 1990; Morita et al., 1997

Drugs

Separation, sensing

Levi et al., 1997; Wang et al., 1997; Mirsky et al., 1999; Andersson, 2000

Flavanoids

Sensing

SuS.rez-Rodriguez and Diaz-Garcia, 2000

Herbicides

Separation, sensing

Kroger et al., 1999; Sergeeva et al., 1999, 2001

Inorganic ions

Separation and sensing

Hutchins and Bachas, 1995; Yoshida et al., 2000; Kimaro et al., 2001.

Microorganisms

Recognition

Alexander and Vulfson, 1997; Dickert et al., 2001

Nucleic acids and derivatives

Separation, sensing

Piletsky et al., 1990a, 1990b; MathewKrotz and Shea, 1996

Polynuclear aromatic hydrocarbons

Sensing

Dickert et al., 1998

Proteins

Separation, recognition

Hjerten et al., 1997; Shi et al., 1999

Steroids

Separation, detection

Hishiya et al., 1999; Rachkov et al., 2000

Sugars, sugar derivatives

Separation, sensing

Wulff and Haarer, 1991; Piletsky et al., 1998

Toxins and narcotics

Separation, sensing

Kriz and Mosbach, 1995; Matshui et al., 1996; Takeuchi et al., 2001.

Volatile compounds

Sensing

Ji et al., 200; Dickert et al., 2001.

i

405

Piletsky and Turner can also be prepared for a broad range of analytes, they have two disadvantages when compared to MIPs. Firstly, small compounds often have to be derivatised in order to generate the antibodies. This necessitates an additional synthetic step, which can sometimes drastically change the recognition characteristics. Secondly, flexibility in antibody preparation is limited to twenty naturally occurring amino acids. In the case of MIPs, the large number of synthetic monomers available make it possible to engineer binding sites with a variety and flexibility unmatched by nature. As any other technology, molecular imprinting has shortcomings. Among them are: (i) absence of a general technology for MIP design; (ii) poor performance of MIPs in aqueous environments; (iii) high level of non-specific binding which produces too low a signal-to-noise ratio in sensors; (iv) poor processability of MIPs; and (v) difficulty in transforming binding events into electrical signals. Several attempts have been made in the past to develop a general procedure for the rational design of imprinted polymers with predictable properties (Nicholls, 1995; Whitcombe et al., 1998; Takeuchi et al., 1999; Lanza and Sellergren., 1999). In the best examples, workers have produced rules or hints, indicating how MIPs should be made in order to possess a certain level of specificity. The most important conclusion is that the stability of the monomer-template complex formed during polymerisation determines the affinity of the resulting polymer. Thus it is known that polymerisation should be performed in a hydrophobic solvent in order to produce a material able to interact with template through electrostatic interactions. At the same time, the choice of the monomer, solvent and polymerisation conditions generally depends on common knowledge, one's personal experience, or available information describing the behaviour of the similar systems. Recently we developed a method that is believed to be a general solution for MIP design (Piletsky et al., 2000c, 2001). The method involves computational screening of a virtual library of functional monomers against a target molecule. The monomers giving the best score in virtual binding experiments are then brought into the contact with template and left to equilibrate. The composition of the monomer shell surrounding the template after equilibration provides the information on the type and quantity of monomers, and should be used for polymer design. Commercially available software permits calculations to be performed using different dielectric constants, reflecting the polarity of the environment (solvent) where the polymers are prepared and used. Polymers designed using this computational approach have proved to have excellent affinity and specificity for the target compound, surpassing those of polyclonal antibodies (Table 3). The possibility of tailoring MIPs for specific target analytes and specific operational conditions is very attractive since it permits polymers to be 406

Materials Based on Imprinted Polymers Table 3. Affinity and sensitivity range of computationally designed molecularly imprinted polymer in comparison with antibodies for the template-- microcystine-LR. Receptor

Ka, (nM)

Sensitivity range (/xg 1~)

Computational MIP

0.3 + 0.08

0.1-100

Monoclonal antibody

0.03 _+0.004

0.025-5

Polyclonal antibody

0.5 +_0.07

0.05-10

i

iii

developed with optimised characteristics and shortens the time needed for design, preparation and testing of the polymers. The computer simulation and molecular modelling approaches could also help to solve a second major problem associated with MIPs - their poor performance in aqueous environments. The majority of monomers used so far in polymer design form hydrogen and ionic bonds in the process of template recognition. These interactions are less effective in polar solvents and as a result, the use of such MIPs is restricted mainly to hydrophobic solvents such as chloroform, toluene, and acetonitrile. Although MIPs capable of forming hydrophobic and van der Waals interactions with the template under aqueous conditions have been developed (Dickert et al., 1999), the design of such polymers is much more difficult than the design of MIPs which exploit electrostatic interactions. The reason lies in the complex nature of factors contributing to hydrophobic and van der Waals interactions. Computer simulation and molecular modeling can, in principle, solve this problem and help to select the monomers ideally suited for the recognition of the template in water. Typically the ratio of functional monomer:template used in molecular imprinting is 4:1 to 10:1. Therefore, the resulting polymer contains large amounts of monomers outside of the specific binding sites; these are capable of non-specific interaction with the molecules other than the template. Additionally the crosslinker itself can interact with variety of analytes in aqueous media. A combination of these factors, together with the large surface area (80-200 m2/g) of the polymer, is responsible for a high level of non-specific binding, which hinders the development of MIP-based affinity materials and sensors. It is possible to overcome this problem however, by further optimisation of the polymerisation procedure and by rational selection of monomers capable of forming stoichiometric complexes with templates (Lubke et al., 2000).

407

Piletsky and Turner

Figure 4. Three principal types of MIP sensors: (a) Affinity sensor, where response is produced by accumulation of template on MIP surface; (b) Receptor sensor where response is generated by changes in polymer characteristics, induced by its inl:eraction with template; (c) Enzyme-mimicking sensor responding to the change in the environment induced by MIP-mediated catalytic reaction.

Detection of binding can be achieved with the help of optical devices, if the template has, for example, fluorescent properties. At least three general characteristics of MIPs can be used for the design of MIP-based sensors (Figure 4)' First, by substituting MIPs for the antibodies in immunosensors (affinity sensors); Second, by exploring of the receptor properties of the imprinted polymers (receptor sensors); Third, by combining MIPs possessing catalytic properties with traditional electrochemical or optical transducers (catalytic sensors). The majority of biosensors produced to date use enzymes as biorecognition element (Turner, 1999). The reason for this lies in the amplification effect achieved as result of multiple turnover of catalytic processes. Many of the unique characteristics of enzymes are connected with their polymeric nature and this fact attracts attention to the methods of development of the MIP catalysts (Srikovsky et al., 2000). The application of catalytically-active MIPs for sensor development seems to be promising and attractive as the most direct way of achieving the replacement of current biosensors by more stable devices. Nevertheless, no practical examples exist of the integration of MIPs that mimic natural enzymes into sensors. Although essential progress has been made in MIP catalysis, imprinted polymers still have properties inferior to natural enzymes 408

Materials Based on Imprinted Polymers e.g., much lower activity and turnover. Success in the preparation of more effective MIPs-based catalysts, with high turnover and reacting with watersoluble and practically important analytes, will change this situation. Due to the very limited information available on the development of MIP-based catalytic sensors, this review will concentrate on the development of the two remaining types of sensors: affinity and receptor-min'ticking devices. Integration of imprinted polymers with detectors remains a difficult issue. Despite significant improvement in MIP technology, the processability of these materials remains challenging. The high level of cross-linking, necessary for maintaining the polymer's specificity, makes them extremely hard, solid, and fragile materials. One solution to this problem is the use of plasticisers, such as oligourethane acrylates (Sergeeva et al., 1999). As a result, polymers can be made in the form of thin and stable membranes which can be used directly in sensors. Alternatively, imprinted polymers can be grafted (Mirsky et al., 1999) or electropolymerised (Boyle et al., 1989) onto the detector surface.

2. Development of MIP-based Optical Sensors The majority of published papers related to MIP sensors deal with electrochemical or piezoelectric devices. We believe, however, that optical and, in particular, fluorescent sensors will play a major role in the future. This confidence is based on the great flexibility which fluorescent detection offers to MIP technology. The account below highlights current achievements and prospects for the development and commercialisation of MIP sensors, which use optical detection for template recognition.

2.1. Affinity sensors The most common type of MIP sensor is the affinity, immunosensor-type device. The detection principle here is based on the measurement of the concentration of template adsorbed by MIP immobilised on the detector surface. The first example of this type of device was the development of a two-dimensional MIP sensor for vitamin K1 (Andersson et al., 1988). Ellipsometry was used for the measurement of template concentration. Although this work was very preliminary and suffered from lack of appropriate controls, it demonstrated the possibility for direct detection of a template adsorbed by an imprinted mono!ayer. Steinke and co-authors, proposed an interesting variant of an optical sensor device based on MIPs (Steinke et al., 1996). The completely transparent imprinted polymer prepared for their experiments had anisotropic properties and provided a particular orientation of bound template molecules. The polymers therefore showed a pronounced dichroism in UV light, which enabled specific

409

Figure 5. Selectivity pattern of pyrene detection by fluorescence, using polyurethanes imprinted with polyaromatic hydrocarbons of different sizes (Dickert and Tortschanoff, 1999).

binding to be recognised. This work could be applicable in particular for the detection of optical isomers. An optical sensor specific for the fluorescent substance dansyl-L-phenylalanine was developed using a dansyl-L-phenylalanine-imprinted polymer and a fibreoptic sensing device (Kriz et al., 1995). Accumulation of fluorescent template in the polymer matrix resulted in an increase in fluorescence that could be used to detect 10 mg/1 of substrate within 4 h. In another example, fluorescent polycyclic aromatic hydrocarbons were selectively enriched and detected using optical sensors based on imprinted polyurethanes (Figure 5) (Dickert and Tortschanoff, 1999). A problem associated with broadening the scope of this method, is the limited quantity of fluorescent substances, which are practically important and can be used as templates in the preparation of MIPs for sensor technology. To overcome this problem, sensors can be developed that operate in a competitive 410

Materials Based on Imprinted Polymers

0.12 0.1

< J c3

0.08

Triazine

/

~

0.06

e

r~

O

0.04 0.02

Simazine

ill

_

1.E-10

I

t,..x

1.E-08

1.E-06

1.E-04

Log C, M Figure 6. Displacement of fluorescein-triazine from triazine-imprinted polymer by analytes (Piletsky et al., 1997).

mode. The important question was whether the binding sites in imprinted polymers are capable of recognising template molecules that are labeled with a fluorescent dye or enzyme. Successful demonstration of this possibility was performed for polymer imprinted with triazine (Piletsky et al., 1997). The competition between fluorescein-labeled and unlabeled template was used to measure 10s - 10.5 M concentrations of free template dissolved in ethanol (Figure 6). The polymer was able to discriminate the template from other triazines (e.g., atrazine) and triazinone (simazine). Later, competitive assays withenzyme-labeled templates were developed for epinephrine (Piletsky et al., 2000b) and 2,4-dichlorophenoxyacetic acid (Surugiu et al., 2001). The displacement format has been used for the development of an MIP sensor for chloramphenicol (CA) (Levi et al., 1997). The sensor included an HPLC column with CA-specific MIPs. A constant flow of dye-labeled CA (chloramphenicolMethyl Red) at a concentration of 0.5 ~tg/ml was run through the column under equilibrium conditions. When analyte containing free CA was injected, it displaced the adsorbed conjugate, giving a peak with an area proportional to CA concentration (Figure 7). Successful analysis of chloramphenicol was achieved in model and real samples (blood serum). 411

Piletsky and Turner

_

_

/

<

/

/

/ t~

/

_

0

_

_

/

~ I

i

!

i

i

i

i

5

10

15

20

25

30

35

Chloramphenicol, mg/I

Figure 7. Displacementof chloramphenicol-Methyl Red from a chloramphenicolimprinted polymer upon injection of template (Levi et al., 1997). Unfortunately, not all analytes can be easily modified with dyes and the modification itself can change the affinity of analyte. Recently, the displacement of non-specific dyes from a MIP has been used for the detection and quantification of ligand-polymer binding events (Piletsky et al., 1999b). Rhodamine B solution was passed through an HPLC column with L-Phe-amidespecific MIP. When template was injected, part of the dye was competitively replaced by the analyte from the MIP. This displacement peak was three times higher for the template than for the opposite enantiomer. This approach can be considered as general, suited for different kinds of templates, dyes, and polymers. A similar displacement principle was used also in combination with electrochemical measurements for template detection (Kroger et al., 1999). It is proposed that the displacement of non-specific indicator molecules from a set or array of MIPs could be used to develop multisensors. The affinity sensors described above are able to detect templates that possess a specific property such as optical absorbance, fluorescence, or electrochemical activity. Direct detection of "inert" templates can be realised in receptor sensors. 2.2. Receptor sensors based on MIPs

Two approaches exist for the development of receptor-like MIP sensors. One is 412

Materials Based on Imprinted Polymers

0.97 0.92 <

0.87 0.82 0.77

~ 0.72 0

~

0.67

/

Glucose/mannose

0.62 0.57 0

1

2

3

4

5

6

7

8

9

10

Sugar, uM Figure 8. Influence of sugars on the formation of fluorescent isoindole complex between OPA reagent and amino-functionality in the polymer imprinted with sialic acid (Piletsky et al., 1996).

connected with an MIP's ability to change conformation upon binding with template, leading to change in a measurable property, such as conductivity, permeability, or surface potential (Piletsky et al., 1998). A second principle is based on the ability of a functional monomer to change its property upon interaction with template, most frequently, fluorescence (Rathbone et al., 2000). Receptor properties of imprinted materials were first reported in 1992 (Piletsky et al., 1992a). It was shown that templates such as amino acids, nucleic acids and cholesterol increase the transport of ions passing through the imprinted membranes (Piletsky et al., 1994b; Piletsky et al., 1998). This so-called "gate effect" has been used for quantification of the concentration of templates. Most often, MIP-based receptor sensors measure the change in membrane electroconductivity, induced by specific interaction of MIPs with template molecules. Sensors specific for L-phenylalanine, cholesterol, sialic acid and atrazine have shown high selectivity and sensitivity at the micromolar and even nanomolar range (Sergeeva et al., 1999). The "gate effect" also can be probed using optical detection (Piletsky et al., 1996). In this work, imprinted polymer based on allylamine was imprinted with sialic acid. When polymer suspension was brought into the contact with OPA reagent, a mixture of o-phtaleic dialdehyde and mercaptoethanol, a fluorescent complex was formed. The kinetics of complex formation depended on the presence of the template, sialic 413

Piletsky and Turner

E 4.5

3.5

2.0

9

I

480

E

I,

,

.

I

s2o 540 wavelength [nm]

..-

4.s

4.O

,

500

,,

I

,

I

sso

56o

"....,.,.

/

3.5 3.0 2.5 2.0 1.5 "., .... ,,,,,

1.0 9

,=80

=

~o

m

I

~o

m

I

54o

I

~

9

I

,~o

9

I

600

wavelength [nm]

Figure 9. Relative emission spectra of polymers containing fluorescein (MIP-F), eosin (MIP-E) and their mixture (MIP-E,F) (top); and spectral changes of donor and acceptor emission in MIP-E,F induced by 10 ~M concentration of L-phenylalanineamide (bottom) Wolfbeis et al., 1997).

acid, which modulated the diffusion of soluble components to the reactive sites (Figure 8). The polymer was able to discriminate sialic acid from other sugars such as glucose and mannose. Potentially, it should be possible to design a sensor where the "gate effect" would be used for direct monitoring of conformational changes in imprinted polymers. Imprinted polymer can be labeled with two different chromophores, one being the donor, the other the acceptor. The light energy adsorbed by the chromophore may be dissipated nonradiatively via a mechanism known as fluorescence resonance energy (F/3rster) transfer (FRET), which is sensitive to intramolecular and intermolecular interactions (Lakowicz et al., 1993). FRET occurs through induction of a dipole oscillation in the unexcited acceptor by the excited-state 414

Materials Based on Imprinted Polymers

CH3 I

0.01

~ NC ' H3

Phe no I b lue N

0.008 o 0.006 e~ 0

0.004 0.002

I

0

-

I

I

I

I

I

2

4

6

8

10

C solvent, o/oo Figure 10. Absorbance vs solvent concentration in air for a polymer with polymerimmobilised phenol blue exposed to tetrahydrofuran (e), chloroform (m) and ethanol (,) (Dickert et al., 2000).

donor chromophore. The rate of energy transfer between the chromophores is a function of the sixth power of the distance between the donor and acceptor. Because of the strong distance dependence of energy transfer, monitoring the fluorescent intensity in a system of labeled polymers can be used to quantify the concentration of a template. In our experiments, imprinted polymers specific for L-phenylalanineamide were labeled with fluorescein (donor), eosin (acceptor) and their mixture (Wolfbeis et al., 1997). The emission spectra of the resulting polymers are shown in Figure 9, top. It can be seen that there is strong FRET occurring in the polymer labeled with both chromophores. The added template decreased the distance between donor and acceptor as a result of the polymer shrinking, which led to an increase in fluorescent emission (Figure 9, bottom). The sensitivity of the system can be further improved using donor-acceptor pairs with different Fi3rster distances. An interesting approach for the design of signalling polymers and their use in sensors was proposed by Cooper and co-authors (1997). They used an environment sensitive functional monomer (I) integrated into a cross-linked matrix, which was able to change its fluorescent properties in the presence of compounds with proton donor properties. 415

Piletsky and Turner

H3C, N~CH3

oH

N

I NX> OH N

OSCH2 O I

|

CH2

OH

0H

|

O

Figure 11. Schematic representation of polymer binding site for cAMP with signalling trans-4-[p-(N,N-dimethylamino)styryl]-N-vinylbenzylpyridinium monomer (Turkewitsch et al., 1998).

CMe

Me

CH2--CHCOO

OOCCH=CH2

Strong quenching of fluorescent emission induced by hydrogen bonding might be used for recognition of polar templates. Similarly betaine dyes with strong hypsochromic effect for protic solvents were used for gas phase analysis (Figure 10) (Dickert et al., 2000). Environmentally-sensitive dye has been used in design of a fluorescent sensor for cAMP detection (Turkewitsch et al., 1998). In this case a fluorescent dye, trans4-[p-(N,N-dimethylamino)styryl]-N-vinylbenzylpyridinium chloride, was copolymerised with cross-linker and template. The resulting polymer displayed two functions simultaneously: template recognition and sensing (Figure 11). 416

Materials Based on Imprinted Polymers A new type of proximity scintillation assay has been developed for (S)propranolol (Ye and Mosbach, 2001). A scintillation monomer, 4hydroxymethyl-2,5-diphenyloxazole acrylate (II), has been covalently incorporated into MIP microparticles during the imprinting reaction. This monomer is capable of transforming 13-radiation from the bound tritium-labeled template into a fluorescent signal. The small size of the particles (0.6-1 ~tm) guarantees that the reporter group, randomly distributed throughout the polymer matrix, is located in close proximity to the MIP binding site for signal generation.

.CH2OCOCH==CH2 II

A reverse scenario can be used, in principle, for the quantification of the concentration of environment-sensitive templates (Matsui et al., 2000). Fluorescent spectra of the cinchona alkaloids exhibit a characteristic shift through binding to these polymer particles, containing acidic m o n o m e r - 2(trifluoromethyl)acrylic acid (TFMAA). The authors demonstrated the possibility of using TFMAA-based imprinted polymers as polymer reagents for analysis of the cinchona alkaloid bound to the polymers without bound/free separation (Figure 12).

0

0

III

Fluorescent polymers containing monomer (III) were imprinted with various N abenzylidene pyridine-2-carboxamidrazones (Rathbone and Ge, 2001). Dramatic quenching of fluorescence approaching background levels was observed for most

417

Piletsky and Turner | .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

,,

.

.=,

1

..g

5:000

h

4O0O

po

:C o C~

30.00

(-)-cinchonidine

0

=

2:000

(9

> r

1000

340

390

440

490

540

5.90

Wav etength (rim) Figure 12. Fluorescence spectra of MIP suspension incubated with cinchonidine in chloroform/acetonitrile at concentrations of (mM): a) 0; b) 0.005; c) 0.02; d) 0.05; e) 0.10; f) 0.15; g) 0.25; h) 0.50. (~,ex=330nm) (Matsui et al., 2000).

cases where the "empty" MIP was re-exposed to its template. The authors claim that this approach is suitable for high throughput screening. Jenkins and coauthors developed a very sensitive lanthanide-based luminescent sensor for sarin and soman with a detection limit of 7 ppt (Jenkins et al., 1998). The sensor functions by selectively and reversibly binding the phosphonate hydrolysis product of this agent to a MIP containing a coordinatively bound Eu 3§ ion. This binding leads to the appearance of a narrow luminescence band in the 610-nm region of the Eu 3§ spectrum, which can be monitored using a miniature spectrometer. A high degree of selectivity is obtained by combining both chemical and spectroscopic selectivities. Very promising combinations of group-specific fluorescent reporters with template-specific MIPs were reported for sugars (Wang et al., 1999), carboxylic 418

Materials Based on Imprinted Polymers acids (Zhang et al., 2001), and primary amines (Subrahmanyam et al., 2000). In the first two cases, the anthracene reporting group was modified in order to introduce polymerisable and recognition functionalities. The interactions of boronic acid (IV) with cis-dioles and guanidine (V) with carboxylic acids is non-specific by its nature. Nevertheless they could be made specific by incorporating these monomers into specific binding sites created by imprinting.

0

\

//

_ / N\

B(OH)2 IV

Similarly non-specific interaction between thioacetale and primary amines, which leads to formation of fluorescent isoindole complex, was made specific for creatine by imprinting in the presence of methylated analogue of the template (Subrahmanyam et al., 2000) (Figure 13).

/

/~

H

H2N

u

Diode lasers are instruments of choice for the development of fluorescent sensors: their light flux is coherent and allows better integration with waveguides, and they are inexpensive and small. However, diode lasers necessitate polymerisable fluorescent markers with long wave adsorption and emission, which at present are scarce. A further problem associated with the development of MIP-based optical sensors is light scatter due to heterogeneity in polymer structure.

419

Piletsky and Turner

S

~

S OH H

+

H2N-R

O

Figure 13. Complex formation between polymerisable thioacetale and primary amine, and recognition of analytes by Blank polymer and polymer imprinted with methylated creatine analogue (Subrahmanyam et al., 2000).

One way to overcome this problem is optimisation of the polymerisation conditions (first of all by choice of solvent and polymerisation temperature) which will lead to synthesis of optically transparent and homogeneous materials. Another way is to measure the decay time of luminescence, rather than its intensity. This approach is highly advantageous because measurements of decay time are less affected by light scattering, analyte concentration and detector sensitivity.

420

Materials Based on Imprinted Polymers 3. Market Potential of MIP Sensors and Future Prospects

Three particular properties make commercial application of MIP sensors attractive: (i) polymers are highly stable and can be autoclaved; (ii) they are fully compatible with microfabrication technology, and (iii) the low cost of the materials and easy processes of polymer preparation in comparison with natural and other artificial receptor systems. The most promising areas of MIP sensor applications are: 1. chemical and pharmaceutical manufacturing: using MIP sensors in extreme conditions (high and low pH's, toxic solvents and high temperature, pressure, and radiation); 2. medicine and pharmaceuticals: application of MIPs mimicking natural receptors for drug screening and for in vivo monitoring; 3. environment: remote sensing, continuous emissions sensors and pointsource monitors; 4. defence: rapid detection of chemical and biological warfare agents under battlefield and civil conditions; 5. deep ocean and space exploration: sensors for analysis of extreme environments. Several key problems associated with MIP development need to be addressed, however, before the successful commercialisation can commence. The issues include: 1. development and validation of a general protocol for MIP design; 2. development of MIPs capable of effective functioning in water; 3. the need for a substantial increase in polymer affinity and improvement of the ratio between specific and non-specific binding; 4. development of effective immobilisation protocols. With further progress in polymer science and engineering we can expect to see the appearance of a new generation of MIP sensors which will gradually replace traditional biosensors and chemical sensors in many areas of biotechnology and pharmacology, environmental, clinical and food analysis.

4. References Agmon, N., 2000, J. Phys. Chem. B 104, 7830. Alexander, C. and E. N. Vulfson, 1997, Adv. Mater. 9, 751. Andersson, L. I., R. Muller, G. Vlatakis and K. Mosbach, 1995, Proc. Natl. Acad. Sci USA, 92, 4788. Andersson, L. I., 1996, Anal. Chem. 68, 111. Andersson, L. I., C. F. Mandenius and K. Mosbach, 1988, Tetrahydron Lett. 29, 5437. 421

Piletsky and Turner Andersson, L. I., 2000, J. Chromatogr. B 739, 163. B lanchard, P., L. Huchet, E. Levillain and J. Roncali, 2000, Electrochem. Commun. 2, 1. Boyle, A., E. M. Genies and M. Lapkowski, 1989, Synth. Metals. 28, C769. Braco, L., K. Dabulis and A. M. Klibanov, 1990, Proc. Natl. Acad. Sci. USA 87, 274. Cooper, M. E., B. P. Hoag and D. L. Gin, 1997, Polym. Prepr. 38, 209. Dhal, P. K., S. Vidyasankar and F. H. Arnold, 1995, Chem. Mater. 7, 154. DickeR, F. L., H. Besenbock and M. Tortschanoff, 1998, Adv. Mater. 10, 149. Dickert, F. and M. Tortschanoff, 1999, Anal. Chem. 71, 4559. Dickert, F., U. Geiger, P. Lieberzeit and U. Reumer, 2000, Sens. Actuators B 70, 263. Dickert, F. L., O. Hayden and K. P. Halikias, 2001, Analyst 126, 766. Dong, J., Y. S. Lin, M. Z.-C. Hu, R. A. Peascoe and E. A. Payzant, 2000, Micropor. Mesopor. Mat. 34, 241. Giraudi, G., C. Giovannoli, C. Tozzi, C. Baggiani and L. Anfossi, 2000, Chem. Commun. 1135. Hishiya, T, M. Shibata, M. Kakazu, H. Asanuma and M. Komiyama, 1999, Macromolecules 32, 2265. Hjerten, S., J. L. Liao, K. Nakazato, Y. Wang, G. Zamaratskaia and H. X. Zhang, 1997, Chromatograph. 44, 227. Hutchins, R. S. and G. Bachas, 1995, Anal. Chem. 67, 1654. Jenkins, A. L., O. M. Uy and G. M. Murray, 1998, Anal. Chem. 71,373. Ji, H. S., S. McNiven, K. H. Lee, T. Saito, K. Ikebukuro and I. Karube, 2000, Biosens. Bioelectron. 15,403. Katz, A. and M. E. Davis, 2000, Nature 403,286. Kempe, M. and K. Mosbach, 1995, J. Chromatogr. A, 691, 317. Kimaro, A., L. A. Kelly and G. M. Murray, 2001, Chem. Commun. 1282. Koshland, D. E., 1995, Angew. Chem. Int. Ed. 33, 2375. Kriz, D. and K. Mosbach, 1995, Anal. Chim. Acta 300, 71. Kriz, D., O. Ramstrom, A. Svensson and K. Mosbach, 1995, Anal. Chem. 67, 2142. Kroger, S., A. P. F. Tutner, K. Mosbach and K. Haupt, 1999, Anal. Chem. 71, 3698. Lakowicz, J. R., W. Wiczk, I. Gryczynsky, M. Fishman and M. L. Johnson, 1993, Macromolecules 26, 349. Lanza F. and B. Sellergren, 1999, Anal. Chem. 71, 2092. Lele, B. S., M. G. Kulkarni and R. A. Mashelkar, 1999, React. Functional Polym. 39, 37. Levi, R., S. McNiven, S. A. Piletsky, S.-H. Cheong, K.Yano and I. Karube, 1997, Anal. Chem. 69, 2017. Lubke, C., M. Lubke, M. J. Whitcombe and E. N. Vulfson, 2000, Macromolecules, 33, 5098. Malitesta, C., I. Losito and P. G. Zambonin, 1999, Anal. Chem. 71, 1366. MathewKrotz, J. and K. J. Shea, 1996, J. Am. Chem. Soc. 118, 8154. 422

Materials Based on Imprinted Polymers

Matsui, J., I. A. Nicholls, T. Takeuchi, K. Mosbach and I. Karube, 1996, Anal. Chim Acta 335, 71. Matsui J., H. Kubo and T. Takeuchi, 2000, Anal. Chem. 72, 3286. Mayes, A. G. and K. Mosbach, 1997, TrAC 16, 321. Mirsky, V. M, T. Hirsch, S. A. Piletsky and O. S. Wolfbeis, 1999, Angew. Chemie, Intern. Ed. 38/8, 1108. Morita, M., O. Niwa and T. Horiuchi, 1997, Electrochim. Acta 42, 3177. Nicholls, I. A., 1995, Chem. Lett. 1035. Nicholls, I. A., O. Ramstrom and K. Mosbach, 1995, J. Chromatogr. A, 691,349. O'Shannessy, D. J., B. Ekberg and K. Mosbach, 1989, Anal. Biochem. 177, 144. Panasyuk, T. L., V. M. Mirsky, S. A. Piletsky and O. S. Wolfbeis, 1999, Anal. Chem. 71, 4609. Peissker, F. and L. Fischer, 1999, Bioorg. Med. Chem. 7, 2231. Piletska E. V., S. A. Piletsky, S. Subrahmanyam, I. A. Nicholls and A. P. F. Turner, 2000, Proc. 1st Int. Workshop on Molecular Imprinting, Cardiff, UK, 2000, 87. Piletsky, S. A., D. M. Fedoryak and V. P. Kukhar, 1990a, Dokl. Acad. Sci. Ukraine B. 4, 53 (in Russian). Piletsky, S. A., I. Ya. Dubey, D. M. Fedoryak and V. P. Kukhar, 1990b, Biopolym. Cell 6, 55 (in Russian). Piletsky, S. A., I. A. Butovich and V. P. Kukhar, 1992a, Zh. Anal. Khim. 47, 1681 (in Russian). Piletsky, S. A. and N. F. Starodub, 1992b, Zh. Anal. Khim. 47, 623 (in Russian). Piletsky, S. A., Ya. I. Kuris', A. E. Rachkov and A. V. Erskaya, 1994a, Russ. J. Electrochem. 30, 1090 (in Russian). Piletsky, S. A., Yu. P. Parhometz, T. L. Panasyuk and A. V. El' skaya, 1994, Sens. Actuators B. 18/19, 629. Piletsky, S. A., E. V. Piletska, K. Yano, A. Kugimiya, A. V. Elgersma, R. Levi, U. Kahlow, T. Takeuchi, I. Karube, T. L. Panasyuk and A. V. El' skaya, 1996, Anal. Lett. 29, 157. Piletsky, S. A., E. V. Piletska, A. V. El' skaya, R. Levi, K. Yano and I. Karube, 1997, Anal. Lett. 30, 445. Piletsky, S. A., E. V. Piletskaya, T. L. Panasyuk, A. V. El' skaya, R. Levi, I. Karube and G. Wulff, 1998, Macromolecules 31, 2137. Piletsky, S. A., E. V. Piletskaya, T. A. Sergeeva, T. L. Panasyuk and A. V. El' skaya, 1999a, Sens. Actuators B 60, 216. Piletsky, S. A., E. Terpetschnig, H. S. Andersson, I. A. Nicholls and O. S. Wolfbeis, 1999b, Fresenius J. Anal. Chem. 364, 512. Piletsky, S. A., H. Matuschewski, U. Schedler, A. Wilpert, E. V. Piletska, T. A. Thiele and M. Ulbricht, 2000a, Macromolecules 33, 3092. Piletsky, S. A. E. V. Piletska, B. Chen, K. Karim, D. Weston, G. Barrett, P. Lowe and A. P. F. Turner, 2000b, Anal. Chem. 72, 4381. Piletsky, S. A., R. M. Day, B. Chen, S. Subrahmanyam, O. Piletska and A. P. F. Turner, 2000c, UK patent application 0001513.1. 423

Piletsky and Turner Piletsky, S. A., K. Karim, E. V. Piletska, C. J. Day, K. W. Freebairn, C. Legge and A. P. F. Turner, 2001, Analyst, in press. Polyakov, M. V., 1931, Zhur. Fiz. Khim. 2, 799 (in Russian). Rachkov, A., S. McNiven, A. V. Erskaya, K. Yano and I. Karube, 2000, Anal. Chim. Acta 405, 23. Ramstrom, O., L. I. Andersson and K. Mosbach, 1993, J. Org. Chem. 58, 7562. Rathbone, D. L., D. Su, Y. Wang and D. C. B illington, 2000, Tetrahedron Lett. 41, 123. Rathbone, D. L. and Y. Ge, 2001, Anal. Chim. Acta 435,129. Sergeyeva, T. A., S. A. Piletsky, A. A. Brovko, E. A. Slinchenko, L. M. Sergeeva and A. V. Erskaya, 1999, Anal. Chim. Acta 392, 105. Sergeyeva, T. A., H. Matuschewski, S. A. Piletsky, J. Bendig, U. Schedler and M. Ulbricht, 2001, J. Chromatogr. A 907, 89. Shi, H. Q., W. B. Tsai, M. D. Garrison, S. Ferrari and B. D. Ratner, 1999, Nature 398, 593. Starodub, N. F., S. A. Piletsky, N. V. Lavryk and E. V. El' skaya, 1992, Sens. Actuators B 13-14, 708. Steinke, J. H. G., I. R. Dunkin and D. C. Sherrington, 1996, Macromolecules 29, 407. Strikovsky, A. G., D. Kasper, M. Grtin, B. S. Green, J. Hradil and G. Wulff, 2000, J. Am. Chem. Soc. 122, 6295. Su~irez-Rodrfguez, J. L. and M. E. Dfaz-Garcia, 2000, Anal. Chim. Acta, 405, 67. Subrahmanyam, S., S. A. Piletsky, E. V. Piletska, B. Chen, R. Day and A. P. F. Turner, 2000, Adv. Mater. 12, 722. Surugiu, I., B. Danielsson, L. Ye, K. Mosbach and K. Haupt, 2001, Anal. Chem. 73,487. Svenson, J. and I. A. Nicholls, 2001, Anal. Chim. Acta 435, 19. Tabushi, I., K. Kurihara, K. Naka, K. Yamamura and H. Hatakeyama, 1987, Tetrahedron Lett. 28, 4299. Takeuchi, T., D. Fukuma and J. Matsui, 1999, Anal. Chem. 71,285. Takeuchi, T., A. Seko, J. Matsui and T. Mukawa, 2001, Instrum. Sci. Technol. 29, 1. Turner, A. P. F., 1999, Biosensor: McGraw-Hill Yearbook of Science and Technology, McGraw-Hill, New York. Vidiasankar, S., M. Ru and F. H. Arnold, 1997, J. Chromatogr. A, 775, 51. Vinokurov, I. A. and M. A. Grigoreva, 1990, Zh. Anal. Khim. 45, 1009 (in Russian). Wang, H. Y., T. Kobayashi, T. Fukaya and N. Fujii, 1997, Langmuir 13, 5396. Wang, W., S. Gao and B. Wang, 1999, Org. Lett. 1, 1209. Watanabe, M., T. Akahoshi, Y. Tabata and D. Nakayama, 1998, J. Am. Chem. Soc. 120, 5577. Whitcombe, M. J., L. Martin and E. N. Vulfson, 1998, Chromatogr. 47, 457. Wolfbeis, O. S., E. Terpetschnig, S. A. Piletsky and E. Pringsheim, 1998, In Iimprinted Polymers: Applied Fluorescence in Chemistry, Biology and

424

Materials Based on Imprinted Polymers Medicine, Eds. W. Rettig, B. Strehmel, S. Schrader, H. Seifert, Springer, Berlin-Heidelberg. Wulff, G. and J. Haarer, 1991, Makromol. Chem. 192, 1329. Wulff, G., 1995, Angew. Chem. Int. Ed. Engl. 34, 1812. Ye, L. and K. Mosbach, 2001, J. Am. Chem. Soc. 123, 2901. Yoshida, M., Y. Hatate, K. Uezu, M. Goto, S. Furusaki, 2000, J. Polym. Sci. Pol. Chem. 38, 689. Yoshizako, K., K. Hosoya, Y. Iwakoshi, K. Kimata, N. Tanaka, 1998, Anal. Chem. 70, 386. Yu, C., O. Ramstrom, K. Mosbach, 1997, Anal. Lett. 30, 2123. Yu, C., K. Mosbach, 2000, J. Chromatogr. A 888, 63. Zhang, H., W. Verboom, D. N. Reinhoudt, 2001, Tetrahedr. Lett. 42, 4413.

425

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 14

OPTICALLY BASED SOL-GEL BIOSENSOR MATERIALS

JENNA L. RICKUS 1, BRUCE DUNN, PH.D. 2, AND JEFFREY I. ZINK, PH.D. 3 Xlnterdepartmental Program for Neuroscience, Neuroengineering Program 2Department of Materials Science and Engineering 3Department of Chemistry And Biochemistry University Of California, Los Angeles USA

The sol-gel process is a chemical technique for synthesizing a silicate matrix around a biomolecule that can function as the recognition and signaling element for a sensor. Within the past decade, biologically doped sol-gel glasses have surfaced as having great potential in optical biosensor applications. The materials are transparent in the UV and visible spectra allowing for transmission of optical signals. The glass is porous such that small analyte molecules can diffuse through the matrix and reach the large biomolecule that is physically trapped. Biological molecules including heme proteins, enzymes, and antibodies can remain active within the porous sol-gel glass. The flexibility of the method has allowed the encapsulation of a wide range of biomolecules and cells, resulting in sensor materials able to detect small molecules in both gases and in liquids.

1. Technical Concept 1.1. Introduction Solid state optical biosensors must fulfill stringent criteria that often are mutually incompatible. The first obvious criterion is that the material must allow the optical signal to pass in and out, i.e., it must be transparent in the desired wavelength region that may range from the near UV to the near IR. The material must be chemically stable and inert and must be able to function in both liquids and gases. Ideally, it should be compatible with existing optical technologies such as fiber optics and waveguides. This same material must not interfere with 427

Figure 1. Large protein molecules are trapped within the pores of the glass matrix. The small analyte can diffuse into the pore and bind to the protein.

the function of the recognition and signaling molecules. Furthermore, in the operating environment, the material must allow the analyte access to the transduction molecules without loss of the latter's functionality. The usual approach for solid state optical biosensors is to immobilize the recognition and signaling biomolecules (commonly proteins that transduce the recognition event into an optical signal) on solid supports by physical adsorption, covalent attachment, or physical entrapment (Shuler and Kargi, 1992). However, the development of a support material that meets all of the above criteria has been problematic. Sol-gel silica materials that have been developed over the past decade meet most of these criteria. Sol-gel encapsulation methods are a new and promising alternative to the more traditional immobilization techniques. It is a physical encapsulation process that eliminates desorption issues and does not require covalent modifications of the biomolecule. The solid matrix is transparent in the ultraviolet and visible spectral regions allowing for spectroscopic analysis (Avnir et al., 1994; Dave et al., 1994). Silica is a mechanically, thermally, and chemically stable material. Its hydrophilic properties and high porosity provide water-filled pores to house and stabilize signaling and transduction biomolecules. A key feature of the sol-gel material is its nanodimensional porosity. Although relatively large biomolecules can be immobilized within the inorganic network, small ions and molecules are able to diffuse into and out of the network. In this way it is possible to use the biomolecule to respond to chemical changes in its environment. Figure 1 illustrates the geometric properties of the matrix. One of the major advantages of the sol-gel encapsulation strategy is its versatility. Since the report of a sol-gel procedure for protein encapsulation in transparent silica in 1992 (Ellerby et al., 1992), there has been steady growth of bio-doped sol-gel materials. Table 1 presents some examples. With only minor 428

Optically Based Sol-Gels procedural variations, many biomolecules including enzymes and antibodies can be encapsulated. The matrix most likely forms around large molecules, thus stretching the upper limits on dopant size (Dave et al., 1997; Dunn and Zink, 1997). Large globular proteins and even cells have been successfully encapsulated (Chia et al., 2000; Pope et al., 1997; Rickus et al., 2001). Under proper conditions, the function of the biological dopant is maintained. Enzymes retain their catalytic ability, antibodies retain their binding affinity, and cells remain viable. In addition to their chemical and physical properties, sol-gel materials have a number of advantages for sensor development. Encapsulation procedures are fast, simple, and flexible. Flexibility of sensor form is provided by the variety of geometries that can be formed including powders, bulk materials, thin films and fibers. Despite the proliferation of bio-doped sol-gel materials with biosensor potential, only a few functional sensors have been realized. Most of the studies have investigated dosimeters rather than continuous sensors. In spite of the important distinction between continuous sensors and dosimeters, most of the reports in the sol-gel literature loosely use the term sensor for both. In this review, we will generically use the term 'sensor' even though the reported material was functioning as a dosimeter. This chapter will first describe the fundamentals of sol-gel encapsulation and the history of these materials. Next, we will review the effects of sol-gel encapsulation on biomolecules. Then the current state of the art will be reviewed using specific examples of sol-gel sensors and potential applications. We conclude with a discussion of the advantages and limitations of sol-gel optical biosensors.

1.2. Sol-gel chemistry The sol-gel process is a chemical synthesis technique for producing amorphous inorganic solids. A colloidal sol forms from the hydrolysis and polycondensation of a metaUo-organic precursor. Metal alkoxides are common precursors because they are easy to hydrolyze (Brinker and Scherer, 1990). Tetramethoxysilane (TMOS) and tetraethoxysilane (TEOS) are typically used because of the ability to carry out well-controlled hydrolysis and condensation reactions with these precursors (Dave et al., 1994). During biomolecule encapsulation, the silica glass matrix forms around the biomolecule from a silicon alkoxide precursor.

429

Rickus, Dunn, and Zink Table 1. Examples of Sol-Gel Encapsulated Biomolecules and Cells Biomoiecule/Cell Heine Proteins Cytochrome c Cytochrome cdl Hemoglobin Myogl0bin Enzymes Alkaline phosphatase Cholinesterase i

Nitrite reductase Glucose oxidase Glutamate dehydrogenase Horseradish peroxidase Lipase Oxalate oxidase Superoxide dismutase Tr)/psin Antibodies Anti-antrazine Anti -D dimer Anti -fluorescein Anti -TNT Other Proteins Bacteriorhodopsin Cells Bacteria Pancreatic islets Parasitic protozoa Plant cells Yeast

.... i

R'eference'

mm

-

ill

9

Ellerby et al., 1992; Wu et al., 1994 Ferretti et al., 2000 Wu et al., 1994; Khan et al., 2000 Chung et aL, 1995; Wu et al., 1994 Braun et al., 1990 Akbarian et al., 1997; Diaz and Peinado, 1997 Ferretti et al., 2000 Chen et al., 1998; Yamanaka et al., 1992 Husing et al., 1999; Rickus et al., 2001 Wu et al., 1994; Diaz et al., 1998 Reetz et al., 1996 Yamanaka et al., 1996 Ellerby et al., 1992 Braun et al., 1992 Bronshtein et al., 1997; Turniansky et al., 1996 Grant and Glass, 1999 Jordan et al., 1996; Wang et al., 1993 Lan et al., 2000 Weetall, 1996; Wu et al., 1993 Fennouh et al., 1999; Livage et al., 1996 Pope et al., 1997 Barreau et al., 1994;Livage et al., 1996 Campostrini et al., 1996 AI-Sara) et al., 1999; Chia et al., 2000

Hydrolysis is the first reaction to occur (Figure 2, top). The result is the formation of silanol (Si - OH) groups by reaction of the alkoxy (- OR, where R is - C H 3 , - C H 2 C H 3 , etc.) with water. Alcohol is released into solution. The hydrolysis is usually acid-catalyzed rather than base-catalyzed because of the preferred matrix structure. Gels formed under acidic conditions tend to form continuous, transparent polymeric structures that can be made into various

430

Optically Based Sol-Gels Hy drolysis OCHs

I

H+

C H 3 0 - - Si--OCH3 + 4H20

~

OH

I

HO--Si--OH

+ 4HOCH3

Water Condensation

OH

OH

l

OH

OH

I

I

I

HO--Si--OH

+

HO--Si--OH

~-

--~

H O - - S i - - O - - S i - - O H + H2C OH

OH

Alcohol Condelasation OH

OClls

OCH3

I

I

I

HO--Si--OH

I

OH

+

CH30--Si--OCH34-'---~

I

OCHs

OH

I

H3CO--Si--O---Si--OH + HOCH

I

OCHa

I

OH

Figure 2: The hydrolysis and polycondensation reactions for the production of the SiO2 matrix using TMOS as the precursor metal alkoxide. The condensation reactions release water and methanol into solution as the matrix forms.

optical components including lenses (Brinker and Scherer, 1990; Hench and West, 1990). Hydrolysis is followed by condensation of the silanol (Si -- OH) groups to release water (Figure 2, middle) and condensation of the silanol and alkoxy groups to release an alcohol and form siloxane (Si - O - Si) groups (Figure 2, bottom). As the condensation reactions occur, a three dimensional network of silica forms. The structural character of the final matrix, including pore size distribution, depends on the relative rates of the three reactions (hydrolysis, water condensation and alcohol condensation) (Brinker and Scherer, 1990). The gelation point is defined as the instant that the silica rrmtrix forms a continuous solid throughout its container. The material is now composed of two distinct phases, amorphous silica particles (5 - 10 nm in diameter) with an interstitial liquid phase. The gelation time can vary from a few seconds to many days depending on the synthesis conditions used. After gelation, the sol-gel 431

Rickus, Dunn, and Zink material remains a dynamic structure; condensation reactions continue to occur as long as remaining hydroxy and alkoxy groups are close enough to react with one another. Anytime after gelation, the gel can be dried under ambient conditions to form a xerogel. The pores of the silica matrix collapse as solvent is removed during drying. As a result, xerogels typically shrink to - 10-15% of their original volume and have a pore size of <100 angstroms. Alternatively gels can be kept hydrated with water or buffer and are referred to as aged gels. Protein-doped gels are often used in the form of wet or aged gels because the pore collapse and solvent evaporation that occurs during the drying process can be detrimental to proteins. Most biosensors are intended for aqueous environments and need never be dried. However, some proteins can withstand the drying process. The drying of cytochrome c-doped monoliths to form xerogels, for example, does not affect the integrity of the protein (Dave et al., 1997).

1.3. Dopant molecules The first sol-gel sensor materials were based on the addition of organic dye molecules as dopants (Pouxviel et al., 1989). Unlike proteins, these molecules can withstand the rougher conditions, e.g., extreme acidity and alcohol, of traditional sol-gel methods. Sol-gel materials doped with organic molecules whose spectroscopic properties are sensitive to changes in pH or ion concentrations have been used as chemical sensors. Many proteins, including enzymes and antibodies, can recognize other proteins and small molecules with a high specificity. This feature makes them popular participants in biosensor schemes. Enzymes are a natural choice for biosensing because they provide both recognition (a high specificity binding site) and a reaction product that can be detected. Many enzymatic reactions, for example, produce a colored or fluorescent product. The kinetics and thermodynamics of the reaction are a function of the reactant concentration. Antibodies also provide a binding site with excellent specificity, but the recognition event is often difficult to observe. However, antibodies have the advantage that they can be produced to recognize almost any small molecule or peptide.

1.4. Diffusion and response times As depicted in Figure 3, sol-gel materials can be categorized into three morphological forms: monoliths, thin films, and fibers. Monoliths are bulk materials with dimensions greater than 1 mm. They take on the shape of their container, and can be cast into virtually any geometry. Thin films can be formed on planar substrates or on cylindrical fibers using dip-coating methods (Brinker et al., 1991; Dislich, 1988). A high quality thin film typically has a uniform

432

Optically Based Sol-Gels

Figure 3. Sol-gel sensor materials take on three of the possible forms: monoliths, thin films on planar substrates, and thin films on optical fibers. A. Monoliths are bulk materials with all dimensions at least 1 mm. The gel retains the shape of the container. B. Liquid sol is dip coated onto a planar substrate to form a thin film. The thin films are generally a few hundred nanometers in thickness. C. A thin film forms on an optical fiber during dip coating. Again, the thin films are typically a few hundred nanometers in thickness. Optical fibers generally range from a few microns to a few hundred microns in diameter.

thickness of a few thousand angstroms. Both the withdrawal speed and the sol viscosity can be used to control the thickness of the film. Most research with doped sol-gel materials begins with the production of monolithic samples. Monolith formation is a well-defined process consisting of controlled hydrolysis, condensation, aging, and drying stages. In contrast, the structural formation of thin films is much more complex. Solvent evaporation, gravitational draining, and film collapse complicate the gelation and aging processes, which typically occur within a thirty-second time period (Nishida et 433

Rickus, Durra, and Zhak al., 1995). As a result, the microstructures of thin films can be very different from those of monoliths. The observed rates of enzyme reactions and analyte binding in monoliths are often slow, taking minutes to reach equilibrium or completion (Akbarian et al., 1997; Ferretti et al., 2000). These reduced rates in bulk materials compared to those in solution are strongly influenced by diffusion limitations. Small analyte molecules have a tortuous path through the silica matrix. The use of thin films rather than thick monoliths reduces the diffusion lengths and decreases response times commensurately. While dip coating can result in active protein-doped thin films, the task of producing a quality film while maintaining the protein's integrity can become a difficult task particularly for large or delicate proteins (Dave et al., 1997). A reasonable gelation time is required to maintain a low sol viscosity during dipping. A changing viscosity can disrupt the homogeneity of the film along the dipping axis. However, the factors that increase gelation time, high alcohol and low pH, can be detrimental to protein stability. These hurdles can be overcome, however, with optimization of the synthesis conditions for each particular protein. The production of active thin films has enabled the development of sol-gel fiber optic sensors. Silica fibers are dip coated with protein-doped sol so that a thin film forms along the tip of the fiber. Optical fibers are a convenient means of guiding light to and from the sensing region. They vary in diameter from a few microns to hundreds of microns. Light can travel long distances with little loss because it is confined to the core of the optical fiber. Because of this, however, a thin film on the outer cladding of a fiber would experience almost no excitation light from the fiber. The sensing region of the fiber must therefore be stripped of its cladding, so that light can escape from the fiber and into the protein-doped thin film.

1.5. Sensor output Optical signal detection is possible due to the transparency of the sol-gel materials. Existing sensors use several different types of optical output. Much of the reported work has utilized optical absorption spectroscopy by monitoring changes in the absorbance or the peak wavelength of detection molecules. In addition, the time-dependence of these changes can be correlated with analyte concentration. Signal detection based on changes in fluorescence emission intensity and wavelength is a commonly employed method and is especially convenient for fiber optic sensors. Fluorescence lifetime measurements have also been used and have the advantage of being independent of indicator concentration, photobleaching, and excitation intensity (McCulloch and Uttamchandani, 1999). In addition to changes in the optical properties of the 434

Optically Based Sol-Gels reporter molecule, changes in the physical properties of the material itself can be monitored for sensing (Ben-David et al., 1997). Surface plasmon resonance (SPR), for example, reports refractive index changes, by measuring the reflected light off a surface at varying angles. It has been suggested that changes in the physical properties of a surface due to antigen binding to a layer of antibodies, for example, could be detected using SPR (Collino et al., 1994; Kambhampati et al., 2001).

1.6. Applications The potential applications for sol-gel optical biosensors cover a broad range of analytes and biological environments. Existing prototype sensors measure small molecules in both gas and liquid samples. Table 2 provides some examples from the literature. While existing sensors are designed to measure a particular analyte, most are not designed to measure that analyte in a specific biological environment. Most studies provide proof of concept in a static buffer solution. The chemical signals that are targets of biosensors are often dynamic, localized, and in a complex environment. To design a biosensor well, one must consider the temporal and spatial distribution of the analyte. In addition, the character of the biological environment, including pH, physical properties, and chemical composition will affect the sensor's performance. Molecules that compete with the analyte or interfere with the optical signal, for example, may be present in different biological environments and affect the background signal and sensitivity of the sensor. Future sensors will need to address the properties of the analyte signal for a particular application.

2. History Sol-gel silica was first produced and studied around 1850, but the field did not gain momentum until the 1930's (Brinker and Scherer, 1990). The first sol-gel dopants were organic molecules (Dickey, 1949). The luminescent properties of encapsulated organic dyes were then explored in the 1980's (Avnir et al., 1985; Reisfeld et al., 1988). The organic dopant was present during both the hydrolysis and condensation reactions and was therefore exposed to extreme pH and alcohol concentrations. Early attempts to encapsulate proteins in sol-gel silica resulted in protein precipitation and low activities. In 1990, Braun and Avnir encapsulated alkaline phosphatase under base-catalyzed conditions, but the resulting powder was opaque due to protein precipitation and had a low enzymatic activity (Braun et al., 1990).

435

Rickus, Duma, and Zink Table 2. Examples of Sol-Gel Optical Sensors. The sensors use a variety of sensing molecules, analytes of interest, forms, and detection methods. Dopant

Analyte L''

Sensor ........ Deiection Type Method

Reference

~-Naphtholphthalein

pH

fiber cladding

Absorbance

Ben-David et al., 1997

Tris(1,10phenanthroline) Ruthenium(II) Ru(II) 4,7-diphenyl

0 2

coated fiber

Fluorescence quenching

02

doped fiber

Fluorescence quenching

McCulloch and Uttamchandani, 1999 Krihak and Shahriari, 1996

pH

coated tapered fiber monolith

1,10-phenanthroline Seminaphthorhodamine- 1 carboxylate (SNARF- 1C) Anti-trinitrotoluene antibodies

TNT

Ratiometric fluorescence emission Competitive, displacement fluorescent immunoassay m o n o l i t h , Chemifiber luminescence

Grant and Glass, 1997 Lan et al., 2000

Horseradish peroxidase

H202

Anti-D-dimer antibodies

D-Dimer

fiber

Fluorescence quenching

Grant and Glass, 1999

Myoglobin

O2

monolith

Absorbance

Chung et al., 1995

Glucose oxidase

Glucose

monolith

Absorbance of dye product

Yamanaka et al., 1992

Glutamate dehydrogenase

Glutamate monolith

Fluorescence intensity

Rickus et al., 2001

Oxalate oxidase peroxidase

Oxalate

Absorbance of dye product

Yamanaka et al., 1996

monolith

Diaz et al., 1998

In 1992, we reported a modified sol-gel procedure and prepared transparent monoliths of biomolecule-doped materials, suitable for optical characterization (Ellerby et al., 1992). These materials retained their spectroscopic properties and biological function. Two significant modifications to the protocol were made. Although TMOS is immiscible in water, rather than add methanol as a cosolvent, the TMOS:water mixture was sonicated before the addition of the protein dopant. Second, we developed a two-step procedure in which the pH of the hydrolyzed TMOS was brought up towards physiological pH before the addition of buffered protein solution. With this procedure the protein dopant is exposed to lower alcohol and acid concentrations. 436

Optically Based Sol-Gels Since this initial work, numerous proteins and even living cells have been successfully encapsulated in sol-gel silica. These materials have demonstrated potential not only as biosensors but also as bioreactors, drug release agents, and even artificial organs (Bottcher, 2000; Pope et al., 1997).

3. State of the Art 3.1. Protein-doped sol-gel materials The earliest examples of protein sol-gel encapsulation defined success as a retention of function, i.e., enzyme activity or substrate binding ability (Braun et al., 1990; Ellerby et al., 1992; Wang et al., 1993). On a more fundamental level however, are questions of conformation, rotational mobility, stability and interaction with the silica matrix (Brennan, 1999). The early studies did give some clues as to the answers of these questions. For example, the blue-green absorption band of copper-zinc superoxide dismutase is sensitive to changes in protein conformation. This band remained unchanged after encapsulation, indicating that the protein within the pores was in a similar state as in solution (Ellerby et al., 1992). In the case of enzymes, catalytic activity often depends on conformational changes. Encapsulated alkaline phosphatase retained about 30% of its enzymatic activity, hinting that at least a portion of the protein must have some freedom to move within the encapsulated pores (Braun et al., 1990). 3.1.1. Kinetics. The kinetics of the binding and transduction event plays a key role in how the protein-doped material performs as a biosensor. Most significantly, it will determine the temporal resolution of the dynamic sensors, i.e., the time it takes to make a measurement and the fastest signal change the sensor can detect.

Several groups have used Michaelis-Menten methods to compare the kinetics of encapsulated enzymes with the kinetics of enzymes in solution. Initial reaction velocities at varying reactant substrate concentrations are fit to the MichaelisMenten equation (Equation 1).

Vmax[S] Vo= KM + [S]

Wo IS]

=

W lllaX

-"

KM =

initial velocity substrate concentration maximum velocity (1) Michaelis-Menten constant

The equation results in two parameters that characterize the kinetics of the reaction (Lehninger et al., 1993). Vm~xis the maximum initial velocity of the reaction that occurs at saturating substrate concentration. The Michaelis-Menten

437

Rickus, Dunn, and Zink Table 3. Effects of encapsulation on Michaelis-Menten kinetic parameters. Enzyme

Alkaline phosphatase a

- ]

Solution . . . . KM (mM) Vmax 0.08+ 0.01

Glutamate dehydrogenase b ~ 0.397 G-6-P 03 dehydrogenaseb 2.17xl Oxalate oxidase/ 0.11 peroxidase b Glucose oxidase b 28+3 aPowder, two populations bMonolith

~ 0.2 min.1 0.746 min.1

Sol-Gel ' [ /~eference KM (raM) Vmax 0.065+0.001 0.08+0.01 U/mL Braun et al., 7.0+0.3 0.59+0.01 1990 U/mL 4.65x10 3 Husing et al., 0.397 min.1 1999 Dave et al., 7 . 8 4 x 1 0 3 0.271 min1 1996

187x10"4 min.1

0.41

9.4x10-4 min.1

Yamanaka et al., 1996

251+25 s1

50+_50

250+80 s1

Yamanaka et al., 1992

2.1+0.1 U/mL

I

ii

|

J

constant, KM, is the substrate concentration at which the initial reaction velocity is half of the maximum possible velocity. Table 3 summarizes the findings of selected studies comparing the KM and Vmax for proteins in solution and encapsulated within sol-gel materials. The catalytic and binding rates are often reduced for encapsulated proteins compared to those in solution (Braun et al., 1990; Husing et al., 1999; Rickus et al., 2001; Wang et al., 1993). Encapsulation sometimes results in a change in the KM of the reaction. Although KM is a compilation of multiple kinetic constants, it generally reflects the affinity of the enzyme for the reaction substrate (Lehninger et al., 1993). The increase in KM observed for glucose oxidase after encapsulation, for example, reflects a weakening of the enzyme's affinity for glucose (Yamanaka et al., 1992). The effect of encapsulation on the KM of different enzymes may depend upon the substrate-enzyme interaction and the effects of the matrix on the enzyme conformation. In addition, encapsulation often results in a reduction in V~x. Most of the kinetic work has been done in bulk materials, and the rate reduction is usually considered to be due to diffusion limitations. While diffusion surely plays a role in monoliths by reducing access of substrate to the enzyme, several other factors may contribute: the presence of the silica matrix may change the protein's microenvironment, mobility, and stability. All of these changes will affect the rates of enzymatic reactions and analyte binding. One cannot easily sort out these effects by considering reaction rates alone. The next three sections will address the topics separately. 438

Optically Based Sol-Gels

3.1.2. Microenvironment within the Pores - p H . When comparing the kinetics of proteins in solution to those in encapsulated gels, it is important to understand that the characteristics of the solution within the pores may be different than the characteristics of the bulk solution. The pH within the pores may differ from the bulk solution by as much as 1 pH unit (Dunn and Zink, 1997). This difference in pH may have significant effects on the kinetics and mechanism of enzymatic reactions. Comparing the reaction kinetics of an enzyme in solution with an encapsulated enzyme while controlling the bulk solution pH may not be appropriate. The bulk solution pH may not reflect the pore pH, and it is the pore pH that affects the encapsulated enzyme directly. Interestingly, Braun and Avnir reported an apparent shift in optimal pH of 1 unit for alkaline phosphatase (Braun et al., 1990). It is possible that while the optimal bulk solution pH shifted by one unit, the optimal pore pH did not. On the other hand, the optimal pH of enzymes is known to be a function of both the enzyme stability and the ionization state of side groups involved in the reaction (Piszkiewicz, 1977). Sol-gel encapsulation has been shown to affect protein stability (Chen et al., 1998; Eggers and Valentine, 2001). This effect may also influence the observed shift in optimal pH. A better understanding of the pore environment and the optimal environment for the protein dopant of interest will allow for better design of protein doped sol-gel materials.

3.1.3. Rotational and intramolecular mobility of encapsulated molecules. The ability of proteins to rotate and to change conformation is essential to their function as sensing molecules. It is therefore important to understand how confinement affects mobility. Many enzymes are functional in sol-gel matrices, so at least some portion of the encapsulated enzyme must be able to undergo rotational and conformational changes. Understanding the impact of encapsulation on mobility, however, will allow for improved designer biosensor materials. Small molecular probes are convenient reporters of molecular mobility within the pores. Static and time-resolved fluorescence polarization methods provide direct measurements of the molecular rotation of a fluorescent probe and the correlating microviscosity of the medium (Dunn and Zink, 1997). These methods monitor the change in orientation of polarization of fluorescence emission by measuring the intensity of emission both parallel and perpendicular to the excitation path. The change in emission polarization direction reflects the reorientation movements of the fluorophore.

439

Rickus, Dunn, and Zink

"///////,

:,'

_~/s.,o.,...,,~/v/ff,,. ..'~,~,~,~~/////~/ Solvent

Figure 4. The four possible locations for dopant molecules in a sol-gel matrix: the pore interior, the pore-matrix interface, the matrix itself, and confining narrow passageways in the matrix. Reproduced from Dunn and Zink (1997) with permission from the American Chemical Society.

Results of fluorescence polarization studies have uncovered multiple sol-gel domains with differing microviscosities (Gottfried et al., 1999; Narang et al., 1994a, 1994b). It is likely that these domains correspond to the different possible locations within the sol-gel material: the pore interior, the pore-matrix interface, the matrix itself, and the confining narrow passageways within the matrix (Dunn and Zink, 1997). Figure 4 illustrates these four domains. Molecules within the interior of the pore are likely to experience conditions similar to that in bulk solution. Molecules at the interface and in constraining regions feel the impact of the matrix. A better understanding of the pore environment and the optimal environment for the protein dopant of interest will allow for better design of protein-doped sol-gel materials. The effects of encapsulation on rotaiional mobility seem to depend on the size of the dopant molecule, preparation methods, and aging time (Gottfried et al., 1999). All three of these factors change the relative size ratio of the dopant molecule to the pore. The higher this ratio, the greater the likelihood a dopant molecule will interact with the silica wall. Proteins are much larger than the probe molecules used for the fluorescence polarization studies. In fact, it seems that the presence of a protein dopant actually shapes the pore that encapsulates it (Dunn and Zink, 1997). The silica matrix may form around the protein indicating a close proximity of the pore wall to the protein (Dave et al., 1997; Dunn et al., 1998). Only a thin layer of water is necessary for enzymes to be active (Klibanov, 1986). The silica matrix can interact with the protein directly through ionic interactions or hydrogen bonding, for example. In addition, the matrix can indirectly interact with the protein by changing the properties of the surrounding water molecules 440

Optically Based Sol-Gels (Eggers and Valentine, 2001). These changes all potentially affect the conformation of the protein as well as its rotational mobility. Several observations have led to the idea that the conformational changes of a protein may in some cases be restricted or directed by the silica matrix (Lan et al., 1999). Encapsulated hemoglobin provides one example. The binding of O2 to the oxygen-transport protein hemoglobin is cooperative in solution and in the blood stream (Stryer, 1988). Hemoglobin possesses four O2-binding heme groups. The binding of 02 to one heme group increases the affinity of the other sites for 02 by inducing a conformational change in the protein. When encapsulated in sol-gel however, hemoglobin binds 02 non-cooperatively (Shibayama and Saigo, 1995). The interpretation of this result is that hemoglobin is "frozen" in its original conformation even after binding 02. The silica matrix seems to stabilize the original conformation or prevents the movement to the second conformation. Whatever the mechanism, an oxygen sensor based on this particular sol-gel material will not have the same oxygen calibration curve that it has in solutionl

3.1.4. Stability of Encapsulated Proteins. Understanding the stability of encapsulated proteins is essential for the development of quality sol-gel biosensors. Changes in protein activity over time result in shifts in calibration curves and determine the lifetime of the sensor. In general, encapsulation seems to benefit protein stability, allowing for sensor materials to be functional for several weeks or even months. Encapsulation within silica pores has been shown to protect many proteins from chemical and thermal destabilization (Dave et al., 1997; Heller and Heller, 1998; Eggers and Valentine, 2001). Limitation of conformational changes may be responsible for the increase in stability by reducing the likelihood of irreversible intramolecular collisions of hydrophobic regions (Chen et al., 1998). On the other hand, the matrix may simply provide a physical barrier that prevents aggregation due to intermolecular association (Eggers and Valentine, 2001). Aggregation is a major factor in protein instability (Wang, 1999). Time or destabilizing conditions such as temperature can lead to structural changes in proteins, such as exposure of hydrophobic regions for example, that may cause a protein to aggregate. Aggregated protein tends to precipitate and lose activity. On the other hand, some enzymes are destabilized by encapsulation. Heller and Heller have used this difference as a basis for understanding the interactions between the silica pore walls and the protein (Chen et al., 1998; Heller and Heller, 1998). Silica has an isoelectric point of 2.1, and therefore the pore surfaces are negatively charged at physiological pH. Electrostatic interactions between the negatively charged silica matrix and positive charges on the protein may play a role in this destabilization. Whether or not a protein is stabilized or

441

Rickus, Durra, and Zink destabilized by encapsulation may be partially determined by the charge profile of its surface and active sites (Heller and Heller, 1998). Despite all of these questions, protein-doped sol-gel materials are active and functional as biosensor materials. Molecules including large enzymes have enough mobility to maintain function as defined by substrate binding and catalytic activity. Current understanding of the protein mobility, stability, microenvironment and kinetic behavior is incomplete. Further research on the subtleties of encapsulation will allow for maximized function and more tailored materials in the future.

3.2. Fiber optic pH sensor- In vivo monitoring of stroke patients Grant and Glass (1997) developed a fiber optic pH sensor as one of a family of sensors intended to monitor changes in blood chemistry at the site of occlusion in stroke patients. A sol-gel silica thin film doped with seminaphthorhodamine-1 carboxylate (SNARF-1C) covered the tip of an optical fiber. The acid form of SNARF-1C has a peak fluorescence of 580 nm while the base form has a peak fluorescence of 640 nm. The ratio of the two peak intensities varies linearly with pH. The fiber had a diameter of 125 ~xm and the tip tapered down to 50 ~tm at the end. The tip was etched and tapered in order to increase the fluorescence capture from SNARF-1C (Anderson et al., 1994a, 1994b). Blood samples of varying pH were used to evaluate the sensor's performance. The sensor displayed a response time of 15 seconds and showed good correlation with standard pH electrode measurements of blood samples. Four identical sensors were produced. Repeatable measurements were observed for each sensor over a three-day period. Large variability, however, was observed between the four sensors indicating that every sensor required individual calibration. This dipped fiber, however, is a good example of a sensor designed well for a particular application. It covers the detection range of interest, pH 6.8 - 8.0 and SNARF-1C has a pKa of 7.4. The optical fiber is MRI compatible, which is an important consideration for implantation into stroke patients. Finally, the optical signal allows for combination with other adjacent sensors without interference. Further characterization of the sensor is required before implantation into patients can be considered. Questions of biocompatibility, stability, and SNARF-1C leaching still remain.

3.3. Micro-optrode pH sensor - Intracellular recordings Many of the chemical signals of interest in biology are spatially localized to small volumes. Some biological applications will require miniaturization of the biosensor. Most of the existing optical fiber sensors, for example, are unable to measure intracellular signals because the fiber diameter (~ 100 ~tm) is larger than 442

Optically Based Sol-Gels

T j

-

9

[I

Co~--r

%: A i r , ~

Figure 5. Myoglobin-doped monoliths measure the dissolved oxygen concentration in sample solutions. The rate of absorbance change at 418 (circles), 431.5 (diamonds), and 436 nm (squares) varies with the dissolved oxygen concentrations. Reproduced from Chung et al. (1995) with permission from the American Chemical Society. the cell (--10 ~m). As pH sensors have led the way in optical sol-gel sensors, they were also the first to be miniaturized. Current pH sensors are now small enough for intracellular measurements. McCulloch and Uttamchandani (1997) have combined the advantages of noncovalent sol-gel encapsulation with miniaturization techniques pioneered by Tan et al. (1992). Tan combined microfabrication techniques and near field optics to achieve a 1000-fold miniaturization of an optical pH sensor. The sensing tip was reduced in size by pulling the fiber into a taper such that the final tip diameter was less than one micron. The fiber was coated with aluminum, leaving only a sub-micron aperture at the tip. McCulloch and Uttamchandani (1997) combined these techniques with sol-gel immobilization methods. A sol-gel film doped with the pH indicator fluorescein was deposited on the tip of the tapered, aluminumcoated optical fiber. With this approach, the resulting pH sensor was small enough to measure the intracellular pH of embryonic fibroblasts. 3.4. Myoglobin-based oxygen sensor Myoglobin (Mb) is a heme-containing protein used for oxygen storage and transport in vertebrates (Stryer, 1988). The absorption spectrum of myoglobin changes depending on the oxidation state of the bound iron heme and the presence of bound oxygen. These changes can be exploited to probe the dissolved oxygen content of the solution surrounding the protein. We encapsulated Mb in silica monoliths to evaluate its potential as an oxygen sensor (Chung et al., 1995). Mb-doped sol was gelled, aged, and then reduced with sodium dithionite; the iron heme must be reduced from the Fe § state to the Fe § state before it can bind oxygen. The absorption spectra for metMb (Fe § state), 443

Rickus, Durra, and Zink deoxyMb (Fe § state, no 02), and oxyMb (Fe +2 state, 02 bound) are similar to those seen in solution (Ellerby et al., 1992; Lan et al., 1995). As oxygen binds to Mb, a gradual decrease in absorbance at 432 nm is seen. This decrease can b e used to determine dissolved oxygen concentration. As is shown in Figure 5, the rate of change at 432 nm and at other wavelengths was linearly proportional to dissolved oxygen concentration of the solution (from 25% to 100% air saturation). It is interesting to note that the correlation between oxygenation rate and dissolved oxygen concentration is likely to be a result of oxygen transport through the matrix. For Mb in solution, oxygen binding occurs within milliseconds. In the gels, oxygen binding occurred at steady rate over a period of 5 to 30 minutes until over 70% of Mb was oxygenated, indicating that the concentration of dissolved oxygen was changing very little during this time and that the oxygenation process was diffusion limited. This example highlights the importance of understanding transport processes in sol-gel matrices for sensor applications. 3.5. Antibody-based sensors

Antibodies are proteins designed to recognize specific molecules and are therefore attractive biomolecules for sensor applications. As illustrated in Figure 6, there are three basic detection schemes for antibody-based sensors: the competitive assay, the displacement assay, and fluorescence quenching. Antibody-doped sol-gel sensors have utilized all three of these approaches (Grant and Glass, 1999; Lan et al., 2000; Wang et al., 1993). Our laboratories used both the competitive and displacement assays for the detection of trinitrotoluene (TNT) using anti-TNT antibody-doped monoliths (Lan et al., 2000). In the competitive assay, a known quantity of fluorescently labeled TNT was added to the sample solution and competed with unlabeled TNT for binding sites on the encapsulated antibodies. The number of bound fluorescent antigen molecules decreased with the concentration of unlabeled TNT in the sample. As seen in Figure 7A, the fluorescent signal in the presence of the sample relative to the baseline signal (B/Bo) decreased with the log of the sample TNT concentration. The disadvantage of the competitive assay is that a washing step is required to remove the unbound and labeled antigen. The displacement assay provides an alternative to the competitive assay when a washing step is inconvenient. In the displacement assay, fluorescently labeled TNT was pre-bound to the antibody. When exposed to the sample solution, the unlabeled TNT displaced some of the bound fluorescent analyte. As shown in Figure 7B, the magnitude of the displacement was a function of the sample TNT concentration.

444

Optically Based Sol-Gels

Figure 6. Antibody-based sol-gel sensors utilize one of three detection schemes. A. Competitive assay. I. Antibody is encapsulated in a gel. II. The sol-gel sensor is immersed in a sample containing an unknown analyte concentration and a known fluorescently labeled analyte solution. III. Excess analyte is washed from the gel. IV. The fluorescence emission from the remaining bound analyte is measured. B. Displacement assay. I. Antibody is encapsulated in a gel with pre-bound fluorescentlylabeled analyte. II. The gel is immersed in sample containing an unknown analyte concentration. III. The gel is removed from the solution and the fluorescence emission from the undisplaced analyte is measured. C. Fluorescence quenching. I. Fluorescenfly labeled antibody is encapsulated in a gel. II. The gel is immersed in the sample. Bound analyte quenches the fluorescence from the antibody tag.

445

Rickus, Dunn, and Zink

A.

1.0

.

B. so1.0r Solution

O. 8

:~

2.o

i

~

t.E;.

i

0.6

g

. 0.4

0.2

0.0

9 Unlabeled

10 TNT

100

0

1

2

:3

4

~

I~

(ppm)

Figure 7. TNT detection based on sol-gel encapsulation of anti-TNT antibodies. A. Competitive assay. The fluorescence signal ratio varies as a function of TNT concentration for antibody in solution and in the gel. B. Displacement assay. The fluorescence signal decreases with TNT concentration. The range of the calibration varies with TNT antibody concentration (Lan et al., 2000). Reproduced with permission from the American Chemical Society.

In the fluorescence quenching scheme, a fluorescently tagged antibody is encapsulated within the sol-gel. The binding of the antigen to its antibody quenches the fluorescence from the antibody's tag. The fluorescence quenching detection scheme was used in a fiber optic sensor for the detection of D-dimer (Grant and Glass, 1999). D-dimer is a peptide fragment formed from the breakdown of the fibrin clots found at the site of vascular occlusion in thrombolytic stroke patients. This sensor was one of a family of sensors created for the diagnosis and treatment of stroke patients. The fiber tip of the D-dimer sensor formed a sloping taper ranging from a starting diameter of 125 ~tm down to 20 ~m at the tip. This tapered portion of the fiber was dip-coated into a silica sol solution doped with fluorescein-labeled anti-Ddirner antibodies. A thin film of less than l~tm in thickness was formed on the fiber. The fiber sensors were introduced into solutions of varying D-dimer concentrations. As seen in Figure 8, the intensity decreased linearly with the log of D-dimer concentration within the range of current diagnostic tests, 0.5 ~tg/mL to 6 ~tg/mL. A similar calibration curve was seen for measurements taken in solutions of phosphate buffered saline (PBS), plasma, and human whole blood. While this sensor exhibited wonderful potential as a practical biosensor, it also highlighted the importance of understanding the material's biocompatibility. In contrast to sensors in PBS, sensors in whole blood samples were functional for only about 5 hours. Longer measurement times resulted in a fouling of the film by various blood components; a visible "cocoon" wrapped the fiber tip. This result is an example of performance in buffer solutions failing to predict performance in a biological environment and further demonstrates the importance of using biological samples to test sensor prototypes. 446

Optically Based Sol-Gels

0.$7 0.566 0.S6 0,SSS O.SS 0.MS ~S4

0.S3

nl

.

| 1

L o g ot .,~.gen Cor~contr~on ( ~ g ~ )

Figure 8. An optical fiber coated with sol-gel containing anti-D-Dimer antibodies measures D-dimer concentration. The fluorescence intensity from the fluorescently labeled antibodies varies as a function of D-dimer concentration. Reproduced from Grant and Glass (1999) with permission.

3.6. Enzyme-based sensors

Enzymes are proteins that catalyze a specific reaction usually by binding to the reactant(s) and bringing these molecules into a favorable position relative to the other reactant molecules or side groups. Because the binding sites on the enzyme are specific, enzymes, like antibodies, are another common choice for use in biosensors. In addition to providing specificity, some enzymes can transduce the binding of analyte into a measurable signal (production of light, a colored product, or a fluorescent product). A number of approaches are available for determining the analyte concentration from the enzyme reaction output. According to the laws of mass action, the rate of the enzymatic reaction is a function of the reactant concentration. The rates of the reaction can therefore be used to calculate the starting concentration of the reactant. The specifics of the calibration depend upon the stoichiometry of the reaction and the source of the output signal. For enzymatic reactions that produce a fluorescent product, a calibration curve of the initial reaction rate as a function of reactant concentration is commonly developed. Use of initial rate data is a convenient kinetic tool because it eliminates the complication of product accumulation for reversible reactions. Using this calibration method, we developed enzyme-based sensor materials for detection of the neurotransmitter, glutamate. As an inter-neuronal signaling molecule, glutamate plays an important role throughout the normal and diseased brain. A practical in vivo glutamate sensor could make a significant impact on the study of neurological disorders including Parkinson's Disease and epilepsy. 447

Rickus, Duma, and Zink 2500

2000

,~

Inject Glutamate

1500

o,,-i

7

1 mM Glutamate

I ......

0 mM Glutamate

lo0o

500 x 9

o

~o

200

9

9

. . . . 9 w.m . . . . w.=.w

a~o

,00

do

600

Time (Seconds) Figure 9. An optical fiber coated with GDH-doped sol-gel responds to a 1 mM glutamate solution, but shows no response to a glutamate-free buffer solution. The fluorescence intensity increases as the fluorescent reaction product, NADH, is produced (unpublished results).

We encapsulated the enzyme glutamate dehydrogenase (GDH) in silica sol-gel monoliths (Rickus et al., 2001; Husing et al., 1999). As shown in Equation 2, GDH catalyses the oxidative deamination of glutamate to (x-ketoglutarate and the concomitant reduction of NAD + to NADH. NADH fluoresces (~ 460nm) when excited with UV light (-- 340nm) while its oxidized form, NAD +, exhibits no fluorescence (Sharma and Schulman, 1999). GDH-doped monoliths measured glutamate concentrations in the presence of saturating NAD § An initial rate calibration curve showed that the rate of change of NADH fluorescence intensity was linearly proportional to the log of the glutamate concentration. The monoliths were active for at least 6 weeks and no detectable GDH leached from the gel after extensive soaking. GDH Glutamate+ H 2 0 + N A D + (_._) c t - k e t o g l u t a r t t e + N H 3 + N A D H + H § (2)

We have extended the above research to the detection of glutamate using optical fibers. A 200 txm diameter optical fiber coated with GDH-doped sol-gel material was able to detect glutamate as well (unpublished results). Figure 9 displays the response of the coated fiber to a solution of 1 mM glutamate. To convert this sensor into one capable of measuring dynamic in vivo signals, it is necessary to develop a dynamic calibration curve not based on initial rate assumptions. Initial rate calibration curves are useful only for dosimeters in a closed system. The calibration assumes the analyte concentration is static 448

Optically Based Sol-Gels

(consumed only by the enzyme reaction itself) for the period of time required to take a measurement. In reality, many interesting biological signals are dynamic; they change rapidly with time. The extracellular glutamate concentrations that we are interested in measuring, for example, can change on a timescale of seconds to milliseconds.

3.7. Encapsulation of cells The sol-gel encapsulation procedure is mild enough to even encapsulate viable cells. Yeast, bacteria and mammalian cells have all been encapsulated in sol-gel silica with a procedure similar to that used for protein encapsulation (A1-Saraj et al., 1999; Chia et al., 2000; Fenn0uh et al., 1999; Pope et al., 1997). Encapsulated bacteria retained [3-galactosidase activity in wet and dry gels even after one week of aging (Fennouh et al., 1999). The wet gels displayed activity curves similar to those of aqueous cell suspensions. Pope et al. (1997) encapsulated murine pancreatic islets for treatment of diabetes mellitus. Encapsulated islets continued to secrete insulin at 3 weeks after encapsulation. Encapsulated islets implanted intra-peritoneally into diabetic mice eliminated diabetic measures (glucosuria and high blood sugar and insulin levels), and in some animals for up to 11 weeks. This study demonstrates that sol-gel encapsulated mammalian cells can remain viable and maintain a normal phenotype in both in vitro and in vivo environments. These studies suggest the possibility of using encapsulated cells as biosensors.

4. Advantages and Limitations 4.1. Advantages Sol-gel methods of biomolecule immobilization provide a number of advantages over competing methods. First of all, the sol-gel method provides a number of procedural advantages. The encapsulation procedure is fast, inexpensive, and flexible. Because the immobilization is based on physical encapsulation rather than the formation of specific covalent interactions, the approach is adaptable to many biomolecules. Minor materials synthesis optimization allows for the encapsulation of most proteins in monoliths. The materials are not only compatible with a number of biomolecule dopants, they are also well suited for electrochemical detection as well as optical detection. A number of amperometric sensors have been created with protein-doped sol-gel materials (Sampath and Lev, 1997; Wang et al., 1998a; Wang et al., 1998b). In addition to providing design flexibility, the physical nature of the immobilization may provide improved sensor performance in biological applications. The sol-gel matrix acts as a barrier between the sample tissue/fluid

449

Rick-us, Duma, and Zink and the sensing biomolecule. By limiting the exposure of the host to a foreign protein, the potential for an immune response to the sensor may be reduced. The silica matrix barrier not only protects the host tissue from the sensing molecule, but also limits the exposure of the sensing molecule to degradative processes in the host tissue. Proteases, for example, are enzymes that cleave unwanted proteins and are present in all cells and tissues (Lehninger et al., 1993). Because the pores of the sol-gel matrix are designed to prevent macromolecule exchange, proteases may not have access to the encapsulated protein. In contrast, covalently immobilized proteins that do not have a physical protection layer are likely to be inactivated by proteolytic cleavage. Encapsulation of proteins in porous silica can increase the stability of the protein (Chen et al., 1998; Eggers and Valentine, 2001). Protein stabilization can have dramatic effects on the shelf life of sensor materials. Sol-gel sensors are therefore more rugged than sensors with covalently linked enzymes. Because silica encapsulated proteins can withstand wider ranges of pH, temperature, and alcohol concentrations, sol-gel sensors can be designed for more demanding sensing conditions. Because encapsulated proteins maintain their integrity and do not aggregate in sol-gels, certain types of analyses are available that would not otherwise be possible in solution. This property has been used to advantage for purposes of observing protein behavior. For example, quantitative electron paramagnetic resonance (EPR) of cytochrome c complexes can now be performed at room temperature when the system is encapsulated in a sol-gel matrix (Lin et al., 1996). Previously, EPR studies of the system in solution required temperatures below -100~ It is thought that the "freezing" of the complex by the sol-gel matrix is responsible for the difference. In another example, sol-gel encapsulation provided a means to observe the unfolding and refolding of carbonmonoxymyoglobin (COMb)(Samuni et al., 2000). In solution, the short lifetimes and wide distribution of the folding intermediates can limit studies of protein unfolding. Samuni et al. used sol-gel encapsulation to slow down the kinetics of unfolding and refolding of COMb, so that nonequilibrium and novel intermediates could be probed. In the future, it is possible that the confining properties of the silica cage could be again used to reveal sensing schemes not otherwise possible in solution.

4.2. Current technology hurdles 4.2.1. Size limitations on sensing molecules, analytes, and co-factors. In some instances, it may be necessary to design the pore size for the diffusion of large analytes. Large molecules in surrounding fluids can be excluded from the pores just as the large sensing molecules are imprisoned within them. The prevention of interaction between these two groups of molecules may eliminate some 450

Optically Based Sol-Gels proteins and peptides as potential analytes if the pore size is not appropriate for the application. On the other hand, the matrix porosity may need to be tailored to retain small molecules. Small molecules are free to exchange between the surrounding environment and the pores. The presence of small molecules in the pores may be required for sensor function in some cases. For example, many enzymes require small molecular co-factors such as NAD(H). The co-factor must be maintained at saturating levels so that the enzyme reaction does not become rate limited by the co-factor, rather than the analyte. In this case, strategies for maintaining high levels of small molecules in the pores over long periods of time are required for practical sensor function.

4.2.2. Effects of Biological Environment on Enzyme Kinetics. In the cell, small metabolites and co-factors regulate the activity of certain enzymes (Lehninger et al., 1993). These enzymes, when used in a biosensor, will be under the same regulation if modulatory molecules are present in the biological sample. For example, GTP and ADP can activate and inhibit, respectively, the enzyme glutamate dehydrogenase after encapsulation in monoliths (Husing et al., 1999). The silica matrix provides no protection from these small allosteric modulators. Their presence can have significant effects on the calibration of a sensor that is based on enzyme kinetics. The access of small molecules to the enzyme is an important consideration when designing a biosensor for measurements in a particular biological tissue. 4.2.3. Continuous sensing. As we pointed out in the introduction, most of the current sensors operate as dosimeters. The challenge is to design the sensors that have continuous monitoring capabilities.

5. Future Directions 5.1. Multianalyte sensors The future of sol-gel biosensors is likely to include the development of multianalyte sensors via the encapsulation of multiple sensing biomolecules (MacCraith et al., 1995). Spatial location or wavelength of the optical signal can distinguish between different analytes. Rowe-Taitt and Ligler created a patterned array of covalently immobilized antibodies on a planar waveguide (Rowe-Taitt et al., 2000). Each antibody was specific for a different bacterial analyte or toxin. A CCD camera imaged the sensor's response to different analyte compositions. The identities of the analytes present were determined by the spatial location of the signals. Similar sensor strategies using sol-gel immobilization methods could be used to develop multianalyte sensors that benefit from the advantages of solgel encapsulation over covalent attachment. 451

Rickus, Durra, and Zink The co-encapsulation of multiple sensor molecules whose optical signals can be resolved according to their emission wavelength is another likely sensor advancement. This strategy would allow for the detection of multiple analytes within a single fiber optic sensor for example.

5.2. Designing the microenvironment As we understand more about the interactions between the matrix and the dopant biomolecule, we can begin to tailor the pore microenvironment. Modification of the matrix itself or the addition of other dopant molecules can change the hydrophobicity and charge profiles within the pores. Changes in the local environment around the dopant molecule can have profound effects on its function and stability. The co-entrapment of the surfactant cetyltrimethylammonium bromide (CTAB) results in large shifts in the pKi's of several pH-sensitive dyes (Rottman et al., 1999). The pH indicator alizarin is sensitive to changes in either acidic or basic ranges depending on the presence or absence of CTAB as a co-dopant in the silica pores (Rottman et al., 1999). In another example, polycation polymers were used to stabilize the enzymes lactate oxidase and glycolate oxidase against thermal degradation (Chen et al., 1998). The enzymes' half-lives were increased over 100-fold at temperatures of 60~ and higher. Future work is likely to include more examples of microenvironment modifications to improve and alter function and stability of doped sol-gel biomaterials.

5.3. Biocompatibility The biocompatibility of protein-doped sol-gels as sensor materials is currently unexplored. When designing biosensor materials, it is important to consider that the sensor can induce changes in the host tissue and the host tissue can induce changes in the sensor. Both types of interactions have consequences for the ability of a biosensor to take meaningful measurements in biological samples. Sol-gel sensor materials must be designed to elicit little or no host response. Most of the understanding of the interaction between the body and sol-gel oxides comes from bone graft work. For many years it has been known that certain "bioactive" glasses can form a bond with bone tissue (Hench et al., 1971). Unfortunately, the goal of bioactive sol-gel glasses for bone replacement is very different than the goal of sol-gel glasses for biosensors. It is the ability of the glass to break down, attract proteins, and form a hydroxy-carbonate apatite layer that facilitates the bone binding process (Zhong and Greenspan, 2000). In a biosensor these events could lead to a loss of function due to dopant loss and pore fouling. Modifications to the material's surface may be necessary to reduce reaction with surrounding tissue.

452

Optically Based Sol-Gels The challenge of designing biocompatible sensors is highlighted in a recent review of non-sol-gel amperometric sensors designed to monitor blood glucose levels in diabetic patients (Gerritsen et al., 1998). Most of these sensors performed well in test solutions, but failed shortly after being implanted. To avoid a similar fate for doped sol-gel sensors, a solid understanding of the interactions between the dopant, the silica matrix, and the surrounding host environment will be required. 5.4. General observations In this review we have shown how sol-gel materials have explosively grown from a laboratory curiosity into the basis for a large number of optical biosensors. As we have summarized in Tables 1 and 2, a broad range of biomaterials and biosensors are being explored. These tables underscore the unprecedented flexibility of the sol-gel process, with dopants ranging from small molecules to cells. The rapid emergence of this field can be attributed to the interdisciplinary efforts among materials scientists, chemists, and biochemists. The future holds many opportunities for advancements. Optically based sol-gel materials are poised to make a major impact in the biosensor field.

6. Acknowledgements The authors greatly appreciate the support of the research through the following grants: the NSF IGERT Program (9972802), the NSF DMR Program (0103952), and the NIH Morris K. Udall Center of Excellence for Parkinson's Disease Research (P50NS38369). The authors appreciate valuable discussions with Prof. Allan J. Tobin, Dr. Esther Lan, and Dr. Sofie Kleppner.

7. References Akbarian, F., A. Lin, B.S. Dunn, J.S. Valentine and J.I. Zink, 1997, J. Sol-Gel Sci. Technol. 8, 1037. A1-Saraj, M. and M.S. Abdel-Latif, I. E1-Nahal and R. Baraka, 1999, J. NonCryst. Solids 248, 137. Anderson, G.P., J.P. Golden, L.K. Cao, D. Wijesuriya, L.C. Shriver-Lake and F.S. Ligler, 1994a, IEEE Engr. Med. Biol. 13,358. Anderson, G.P., J.P. Golden and F.S. Ligler, 1994b, IEEE Trans. Biom. Engr. 41, 578. Avnir, D., S. Braun, O. Lev and M. Ottolenghi, 1994, Chem. Mater. 6, 1605. Avnir, D., V.R. Kaufman and R. Reisfeld, 1985, J. Non-Cryst. Solids 74, 395. Barreau, J.Y., J.M.D. Costa, I., Desportes, J. Livage,L. Monjour and M. Gentilini, 1994, C.R. Acad. Sci. 317,653.

453

Riekus, Duma, and Zink Ben-David, O., E. Shafir, I. Gilath, Y. Prior and D. Avnir, 1997, Chem. Mater. 9, 2255. Bottcher, H., 2000, J. Prakt. Chem. 342, 427. Braun, S., S. Rappoport, R. Zusman, D. Avnir and M. Ottolenghi, 1990, Mater. Lett. 10,1. Braun, S., S. Shtelzer, S. Rappoport, D. Avnir and M. Ottolenghi, 1992, J. NonCryst. Solids 147/148, 739. Brennan, J.D., 1999, Appl. Spect. 53, 106A. Brinker, C.J., A.J. Hurd, G.C. Frye, P.R. Schunk and C.S. Ashley, 1991, J. Ceram. Soc. Jpn. 99. Brinker, C.J. and G.W. Scherer, 1990, The Physics and Chemistry of Sol-Gel Processing, San Diego, Academic Press. Bronshtein, A., N. Aharonson, D. Avnir, A. Turniansky and M. Altstein, 1997, Chem. Mater. 9, 2632. Carnpostrini, R., G. Caruran, R. Caniato, A. Piovan, R. Filippini and G. Innocenti, 1996, J. Sol-Gel Sci.Technol. 7, 87. Chen, Q., G.L. Kenausis and A. Heller, 1998, J. Am. Chem. Soc. 120, 4582. Chia, S., J. Urano, F. Tamanoi, B. Dunn and J.I. Zink, 2000, J. Am. Chem. Soc. 122, 6488. Chung, K.E., E.H. Lan, M.S. Davidson, B.S. Dunn, J.S. Valentine and J.I. Zink, 1995, Anal. Chem. 67, 1505. Collino R, J. Therasse, P. Binder, F. Chaput, J-P. Boilot and Y. Levy, 1994, J. Sol-Gel Sci. Technol. 2, 823. Dave, B., B. Dunn and J.S. Valentine, J.I. Zink, 1996, In Nanotechnology: Molecularly Designed Materials, Chow G-M and Gonsalves KE, eds. Washington DC, American Chemical Society. Dave, B.C., B. Dunn, J.S. Valentine and J.I. Zink, 1994, Anal. Chem. 66, 1120A. Dave, B.C., J.M. Miller, B. Dunn, J.S. Valentine and J.I., Zink, 1997, J. Sol-Gel Sci. Technol. 8, 629. Diaz, A.N. and M.C.R. Peinado, 1997, Sens. Actuators B 38/39, 426. Diaz, A.N., M.C.R. Peinado and M.C.T. Minguez, 1998, Anal. Chem. 363, 221. Dickey, F.H., 1949, Proc. Nat. Acad. Sci. 35,227. Dislich, H., 1988, In Sol-Gel Technology for Thin Films, Fibers, Preforms, Electronics, and Specialty Shapes, Klein LC, ed., Park Ridge, New Jersey: Noyes Publications, 50. Dunn, B., J.M. Miller, B.C. Dave, J.S. Valentine and J.I., Zink, 1998, Acta Mater. 46, 737. Dunn, B. and J.I. Zink, 1997, Chem. Mater. 9, 2280. Eggers, D.K. and J.S. Valentine, 2001, Prot. Sci. 10, 250. Ellerby, L.M., C.R. Nishida, F. Nishida, S.A. Yamanaka, B. Dunn and J.S. Valentine, 1992, Science 255, 1113. Fennouh, S., S. Guyon, C. Jourdat, J. Livage and C. Roux, 1999, C.R. Acad. Sci. Paris 2, 625. Ferretti, S., S.K. Lee, B.D. MacCraith, A.G. Olivia, D.J. Richardson and D.A. Russell, 2000, Analyst 125, 1993. 454

Optically Based Sol-Gels Gerritsen, M., J.A. Jansen, A. Kros, R.J.M. Nolte and J.A. Lutterman, 1998, J. Invest. Surg. 11,163. Gottfried, D.S., A. Kagan, B.M. Hoffman and J.M. Friedman, 1999, J. Phys. Chem. B 103, 2803. Grant, S.A. and R.S. Glass, 1997, Sens. Actuators 45, 35. Grant, S.A. and R.S. Glass, 1999, IEEE Trans. Biom. Engr. 46, 1207. Heller, J. and A. Heller, 1998, J. Am. Chem. Soc. 120, 4586. Hench, L.L., R.J. Splinter,W.C. Allen and T.K. Greenlee, 1971, J. Biom. Mater. Res. Symp. 2, 117. Hench, L.L. and J.K. West, 1990, Chem. Rev. 90, 33. Husing, N., E. Reisler and J.I. Zink, 1999, J. Sol-Gel Sci. Technol. Jordan, J.D., R.A. Dunbar and F.V. Bright; 1996, Anal. Chem. 332, 89. Kambhampati, D.K., T.A.M. Jakob, J.W. Robertson, M. Cai, J.E. Pemberton and W. Knoll, 2001, Langmuir 17, 1169. Khan, I., C.F. Shannon, D. Dantsker, A.J. Friedman, J. Perez-Gonzalez-deApodaca and J.M. Friedman, 2000, Biochem. 39, 16099. Klibanov, A.M., 1986, Chemtech 354-359. Krihak, M.K. and M.R. Shahriari, 1996, Electron. Lett. 32, 240. Lan, E.H., B.C. Dave, J.M. Fukuto, B. Dunn, J.I. Zink and J.S. Valentine,1999, J. Mater. Chem. 9, 45. Lan, E.H., M.S. Davidson, L.M. Ellerby, B. Dunn, J.S. Valentine and J.I. Zink, 1995, Mater. Res. Soc. Proc. 330, 289. Lan, E.H., B. Dunn and J.I. Zink, 2000, Chem. Mater. 12, 1874. Lehninger, A.L., D.L. Nelson and M.M. Cox, 1993, Principles of Biochemistry. New York: Worth Publishers. Lin, C.T., C.M. Catuara, J.E. Erman, K.C. Chen, S.F. Huang and S.J. Wang, 1996, J. Sol-Gel Sci. Technol. 7, 19. Livage, J., C. Roux, J.M. DaCosta, I. Desportes and J.F. Quinson, 1996, J. SolGel Sci. Technol. 7, 45. MacCraith, B.D., C.M. McDonagh, G. O'Keefe, A.K. McEvoy, T. Butler and F.R. Sheridan, 1995, Sens. Actuators B 29, 51. McCulloch, S. and D. Uttamchandani, 1997, IEEE Proc. Optoelectronics, 144, 162. McCulloch, S. and D. Uttamchandani, 1999, IEEE Proc. Sci. Meas. Technol. 146, 123. Narang, U., J.D. Jordan, F.V. Bright and P.N. Prassad, 1994a, J. Phys. Chem. 98, 8101. Narang, U., R. Wang, P.N. Prassad and F.V. Bright, 1994b, J. Phys. Chem. 98, 17. Nishida, F., J.M. McKieman, B. Dunn, J.I. Zink, C.J. Brinker and A.J. Hurd, 1995, J. Am. Ceram. Soc. 78, 1640. Piszkiewicz, D., 1997, Enzyme Catalysis and Kinetics. Kinetics of Chemical Enzyme-Catalyzed Reactions, New York, Oxford University Press. Pope, E.J.A., K. Braun and C.M. Peterson, 1997, J. Sol-Gel Sci. Technol. 8, 635. Pouxviel, J.C., B. Dunn and J.I. Zink, 1989, J. Phys. Chem. 93, 2134. 455

Rickus, Duma, and Zink Reetz, M.T., A. Zonta, J. Simpelkamp, A. Rufinska and B. Tesche, 1996, J. SolGel Sci. Technol. 7, 35. Reisfeld, R., R. Zusman, Y. Cohen and M. Eyal, 1988, Chem. Phys. Lett. 147, 142. Rickus, J.L., E. Lan, A.J. Tobin, J.I. Zink and B. Dunn, 2001, Mat. Res. Soc. Symp. Proc. 662, in press, available online at www.mrs.org. Rottman, C., G. Grader,Y.D. Hazan, S. Melchior and D. Avnir, 1999, J. Am. Chem. Soc. 121, 8533. Rowe-Taitt, C.A., J.P.Golden, M.J. Feldstein, J.J. Cras, K.E. Hoffman and F.S. Ligler, 2000, Biosens. Bioelectron. 14, 785. Sampath, S. and O. Lev, 1997, J. Electroanal. Chem. 426, 131. Samuni, U., M.S. Navati, L.J. Juszczak, D. Dantsker, M. Yang and J.M. Friedman, 2000, J. Phys. Chem. 104, 10802. Sharma, A. and S.G. Schulman, 1999, Introduction to Fluorescence Spectroscopy, New York, Wiley, 145. Shibayama, N. and S. Saigo, 1995, J. Mol. Biol. 251,203. Shuler, M.L. and F. Kargi, 1992, Bioprocess Engineering, Englewood Cliffs, New Jersey, PTR Prentice Hall. Stryer, L., 1988, Biochemistry, New York, W.H. Freeman and Company. Tan, W., Z-Y. Shi, S. Smith, D. Bimbaum and R. Kopelman, 1992, Science 258, 778. Turniansky, A., D. Avnir, A. Bronshtein, N. Aharonson and M. Altstein, 1996, J. Sol-Gel Sci. Technol. 7, 135. Wang, B., B. Li, Q. Deng and S. Dong, 1998a, Anal. Chem. 70, 3170. Wang, J., P.V.A. Pamidi and K.R. Rogers, 1998b, Anal. Chem. 70, 1171. Wang, R., U. Narang, P.N. Prasad and F.V. Bright, 1993, Anal. Chem. 65, 2671. Wang, W., 1999, Int. J. Pharmac. 185, 129. Weetall, H.H., 1996, Biosens. Actuators 11,327. Wu, S., L.M. Ellerby, J.S. Cohan, B. Dunn, M.A. E1-Sayed and J.S. Valentine, 1993, Chem. Mater. 5, 115. Wu, S., J. Lin and S.I. Chan, 1994, Appl. B iochem. B iotechnol. 47, 11. Yamanaka, S., N.P. Nguyen, B. Dunn, J.S. Valentine and J.I. Zink, 1996, J. SolGel Sci. Technol. 7, 117. Yamanaka, S.A., F. Nishida, L.M. Ellerby, C.R. Nishida, B. Dunn and J.S. Valentine, 1992, Chem. Mater. 4, 495. Zhong, J. and D.C. Greenspan, 2000, J. Biomed. Mater. Res. (Appl. Biomater.) 53, 694.

456

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 15

MEMBRANE-BASED BIOSENSORS

BRUCE A. CORNELLPH.D.

Ambri Ltd, 126 Greville Street, Chatswood, New South Wales, 2067 Australia

Membrane-based biosensors operate by reporting on changes in the a.c. ionic conduction of a molecular membrane. These membranes comprise one or more lipid layers and are usually chemically tethered to an electrical conductor such as gold. An alteration in the ion concentration on the inner side of the membrane induces a proportionate alteration in the image charge at the gold surface. Conventional a.c. impedance spectroscopy between the gold and a counter electrode in the bathing solution may thus be used to measure changes in the ionic conductivity of the membrane. The magnitude of these changes may be increased by separating the membrane from the gold surface using a hydrophilic spacer. Many schemes have been reported for modulating the membrane ion conduction in response to detecting a specific target molecule. In particular, we have reported a generic immunosensing device in which the ion channel, gramicidin A, is assembled into a tethered lipid membrane and coupled to an antibody targeting a compound of diagnostic interest. The binding of the target molecule causes the conformation of the gramicidin channels to switch from predominantly conducting dimers to predominantly non-conducting monomers. The approach permits the quantitative detection of an extensive class of target species, including proteins, bacteria, drugs, toxins and DNA sequences. Incorporating ion specific carriers within the membrane permits the measurement of the concentration of a wide range of specific electrolytes.

457

Comell 1. Overview of Membrane-based Biosensors

Biological sensory systems function by registering changes in the electrical conductivity of specialized cell membranes (Avrone et al., 1995). A well documented example is the acetylcholine-triggered cation channel involved in cross synaptic nerve conduction (Reiken et al., 1996). An important feature of this mechanism is its inherent amplification with a single molecule potentially triggering the passage of up to a million ions per second across an otherwise impermeable membrane. The first proposed use of a synthetic molecular membrane as the basis for a chemical sensor was by del Castello et al. 1966. These authors reported the dependence of the admittance of a classical solvent formed bilayer lipid membrane (BLM) on the presence of a range of proteinprotein interactions between the membrane and species in the bathing solution. The approach failed however, to address the major practical issue of stabilizing the membrane against mechanical damage. Yager (1988) and Zaitsev et al. (1988) incorporated the light-sensitive bacterial proton pump, bacteriorhodopsin, into liposomes, stabilized with polymerisable phosphatidylcholines. These were observed to pump protons on being photocycled. The first reported attempt at developing a practical membrane-based biosensor device was by Ligler et al. (1988). Aimed at the detection of biological toxins, it comprised alamethicin and a calcium channel complex within a polymerized bilayer lipid membrane supported on a porous silicon substrate. The porous silicon provided both a mechanical support for the molecular membrane whilst also providing a reservoir for the trans-membrane flow of ions. The poor stability of the receptormembrane complex limited the range of applications of the device. The poor stability arose from both the intrinsically labile calcium channel complex and the mechanical instabilities of the supporting membrane. The stabilization of the bilayer lipid membrane has been a central theme in the development of membrane-based biosensors. Many strategies have been developed, but most focus on physisorbing or chemically attaching a layer of hydrocarbon to a silicon (Heysel, et al., 1995), hydrogel (Lu, et al., 1996), polymer (Naumann et al. 2001a, 2001b) or metal surface, (Steinem et al., 1996) and then fusing a second layer of mobile lipids onto the tethered monolayer to form a tethered bilayer lipid membrane. Reviews by Sackman (1996) and Plant (1999) describe the literature on BLM stabilization. These approaches have become progressively more sophisticated and there are now many reports of patterning techniques to produce membrane arrays possessing different localized functionalities.

2. The Ion Channel Switch ICS TM Biosensor

The low molecular weight bacterial ion channel gramicidin (Wallace, 1998) has been used (Cornell et. al., 1997; Woodhouse et al., 1998; Cornell et al., 1999; 458

Membrane-based Biosensors Woodhouse et al., 1999), as the basis of a biosensor platform with a range of applications for the detection of low molecular weight drugs, large proteins and microorganisms. The Ion Cha)nnel Switch (ICS TM) biosensor employs a disulphide lipid monolayer tetJ~red to a gold surface. The membrane is separated from the gold surface by a polar spacer which provides a reservoir for ions permeating through the membrane. The transduction mechanism depends on the properties of Gramicidin A within BLM' s. Gramicidin monomers diffuse within the individual monolayers of the BLM. The flow of ions through gramicidin only occurs when two non-conducting monomers align to form a conducting dimer. Within the biosensor, the gramicidins within the tethered inner leaflet of the lipid bilayer are also tethered. The arrival of analyte crosslinks antibodies attached to the mobile outer layer channels to those attached to the membrane spanning lipid tethers. Because of the low density of tethered channels within the inner membrane leaflet, this anchors them distant, on average, from their immobilized inner layer channel partners. Gramicidin dimer conduction is thus prevented and the admittance of the membrane decreases. Applying a small alternating potential between the gold substrate and a reference electrode in the test solution generates a charge at the gold surface and causes electrons to flow in an external circuit. The outer leaflet of the biosensor membrane is effectively a two-dimensional liquid crystal. This permits the antibodies attached to the mobile ion channels to scan a significant membrane area in a few minutes. Each mobile channel thus has access to multiple antibody sites and responds more rapidly than if the transduction mechanism were triggered simply by binding to the antibodies attached to the mobile channels. The speed of the sensor response is improved in direct proportion to the number of binding sites accessible to the mobile channels. This property is effectively a molecular analogue of an electronic multiplexer. The membrane stability is primarily enhanced by tethering the inner membrane leaflet to the gold surface. However, additional stability is achieved by substituting a major fraction of the tethered lipids with "archaebacterial lipids". These are lipids modelled on constituents found in bacteria capable of surviving extremes of temperature and hostile chemical environments. Characteristics of these lipids are that the hydrocarbon chains span the entire membrane (Kushawaha et al., 1981) and that all ester linkages are replaced with ethers (De Rosa et al., 1983). BLM films have previously been formed from archaebacterial lipids (Gliozzi et al., 1983) and resulted in membranes that are stable to temperatures in excess of 90~ Most studies of the ICS TM biosensor have used antibody Fab' fragments as the receptor. However, the approach is generic and has been demonstrated to operate using oligonucleotide probes, heavy metal chelates and cell surface receptors.

459

Comell A commercial device is currently under development employing this approach and is principally aimed at rapid, quantitative measures of the time-critical diagnostic measures in whole blood. However, the operation of the device is not constrained to just blood and may be used with any electrolyte containing biological fluid including, serum, urine and saliva. The specificity of the response is dependent on the specificity of the receptor.

3. Membrane Biosensor Fabrication A stable membrane incorporating ion channels can be self-assembled (Philp and Stoddart (1996) on a clean, smooth gold surface using a combination of sulphurgold chemistry and physisorption.. Each of these stages is considered in the following sections.

3.1. Preparation An important first step in the fabrication of a reliable biosensing device is the production of a contaminant and oxide free, molecularly smooth gold surface. There is an extensive literature on producing good quality gold surfaces. The following articles provide an overview of the main points. Guo et al. (1994) report on the quality of vapor-deposited and sputter-coated thin gold films on mica, oven heated in ultrahigh vacuum at up to 500~ The Au surface roughness became atomically flat at 450-500~ as shown by AFM. Contrary to previous reports, it was found that surface contamination of the gold by organics and gold oxide was more significant in obtaining good high impedance films than simply the molecular smoothness of the gold. It is suggested, however, that grain boundaries in the "pebbled" gold surfaces they observed in the non-annealed gold may capture impurities which cause leakiness in alkanethiol films. Immersion for 10-15 minutes in hot piranha solution (HzSO4/H202), followed by an electrochemical stripping of any resulting oxide resulted in gold that would support a high-impedance self-assembled membrane

(SAM). Ron and Rubenstein (1994) also address the risk of forming alkane thiol monolayers over pre-oxidized gold. Exposure to oxidants such as oxygen plasma, UV/ozone or piranha solution removes organic contaminants but causes the formation of gold oxide. Gold oxide may be removed by prolonged immersion in ethanol or by using electrochemical stripping. Failure to remove all traces of oxide prior to depositing an alkanethiol monolayer causes the oxide to be trapped beneath the alkanethiol film. Furthermore the monolayer isolates the oxide, preventing its reduction by electrochemistry. The properties of trapped. Au203 are different for thiols and disulphides. Whilst alkanethiols and dialkyldisulphides both form compact monolayers on reduced gold, only 460

Membrane-based Biosensors alkanethiols form compact monolayers on oxidized gold. As seen below, the ICS TM sensor technology employs disulphides which requires rigor in eliminating gold oxide. A remarkable quality check of the uniformity of gold films was reported by Gupta and Abbott (1996). They used a nematic liquid crystal, 4'-cyano-4'pentylbiphenyl (5CB) to report on disorder within columna polycrystalline gold. SAMs of C15 and C9 alkanethiols were used as a support for 5CB. By observing the 5CB coated Au between a polarizer and an analyser, they were able to use the birefringence of the nematic liquid crystal to visualize variation in the azimuthal angle of the director of grains within the gold film. Gold obliquely deposited at 50 ~ was compared with uniformly coated gold, deposited onto samples mounted on a two-axis rotating stage. The azimuthal grain order was seen to exist over centimetre scales for the obliquely deposited gold. By contrast SAMs prepared on films of gold deposited with no preferred direction caused organised nematic phases to form over at best, the 10-100 ~ n scale. Other techniques such as contact angle measurements and AFM showed no detectable difference between the two types of gold. It is interesting to note that Aizenberg et al. (1998) actually employ variations in the topology of various metallic films to control the local disorder in SAMs to permit patterning of their surfaces. Zhang et al. (1998) have shown that 03 is at least an order of magnitude more effective than other oxygen-containing species at generating labile products typified as oxidized sulphur head groups which are only weakt~ adsorbed and may be easily rinsed from the surface. Ron and Rubenstein (1998a) observed that gold surfaces exposed to a laboratory atmosphere for only minutes were found by XPS to develop a layer of contamination approximately 0.6 nm thick and composed of oxygen and carbon. In a later article, they (Ron and Rubenstein, 1998b) addressed this problem using an electrochemical technique which oxidises surface contaminants, cleans the surface of oxide, and then accelerates the deposition of the alkane monolayer. Three reactions are important: 2Au + 31-120- 6e --~I--]Au2 O3 + 6H + gold oxidation CH3CH2OH- 2e- > CH3CHO + 2H+ ethanol oxidation Au203 + eCH3CH2OH > 2Au(0)+ 3CH,3CHO+ 3H20 ethanol oxidation by gold Initially, the gold is electrochemically oxidised in water. At-0.3 V, the gold oxide is reduced in ethanol. By switching the potential at the gold to an oxidizing +1.45 V and by using a fast liquid exchange flow cell to synchronously introduce an alkanethiol into the solution, a high quality electrically sealing monolayer is formed in 1-2 seconds.

461

Comell Lee et al. (1998) studied the effect of variable topography of the gold surface on monolayers of alkanethiols. Alkanethiolates on gold, oxidize in air, in the dark, to form sulfinates and sulfonates. The kinetics of oxidation depend on the morphology of the underlying gold, increasing dramatically with a decrease in the size of the grains and the amount of Au(111) on the surface. The difference in kinetics of oxidation is sufficiently great that the authors suggest it may be an explanation for the variation in the quality of alkanethiol SAMs reported by different groups. Grain boundaries of evaporated polycrystalline gold films were identified as an important catalytic site for oxidation. To prepare a smooth surface, they annealed the gold with a small butane flame. This both cleaned the gold and facilitated its recrystallization into large regions of A u ( l l l ) . They produced "epitaxially" grown gold surfaces by slow (1 ,~/s) thermal evaporation on "scratch free" mica at 2.5 x 10.7 mbar at 300 ~ until 1000 A of gold had been deposited. The surface was transformed into defect free Au(111) over 1500 x 1500/~ areas. These regions were substantially more robust against oxidation and electrical leakage of the alkanethiol SAMs. The importance of flow on the electrochemical pre-treatment of gold is reported by Hoogvliet et al. (2000). They reaffirm that the state of the gold surface, and in particular the surface roughness, is important for the reproducible formation of high quality self-assembled monolayers on gold. They report a pulsed potential pretreatment which results in a two-fold reduction in roughness of mechanically polished surfaces. They use a flow cell and a 100 ms triple pulse sequence of +1.6, 0.0 and -0.8 V relative to a counter electrode. Pulsing for 2000-5000 s under flow conditions is required to achieve good smoothing.

3.2. Tethered bilayer lipidmembrane An important element in the stability of membrane-based biosensors is the use of a tethered monolayer, which in the present example is an alkane disulphide that stabilizes the bilayer lipid membrane at the electrode surface. This section will review some of the key articles in this very extensive literature. A further strategy for stabilizing tethered bilayer lipid membranes is to use a significant fraction of ether-linked, membrane-spanning tethered lipids. The use of membrane-spanning lipids results in the stabilization of the lamellar phase for the non-tethered lipids and a resistance to insertion of additional material into the pre-formed membrane assembly. Tethered membranes may be formed that possess excellent stability over many months. A further strategy for long term storage is drying followed by rehydration prior to use.

3.2.1. Tethered membrane components. Ulman (1996) has reviewed the early literature on the formation of self-assembled monolayers of gold-tethered alkanethiols. There are many alternative approaches to tethering to gold which are not discussed here, such as the use of polymer and gel supported bilayers (Naumann et al. 2001a, 2001b; Erdelen et al., 1994). However, the convenience 462

Membrane-based Biosensors and extensive experience available on the sulphur-gold tethering mechanism provides a good foundation for the present discussion (Lingler et al., 1997). Poirier and Pylan (1996) studied the growth of methyl-terminated homologs n = 5 and 9 and six methylene unit homologs X = OH and COOH. The growth of the films follows a two-step process that begins with condensation of low-density crystalline islands with the molecular axis lying parallel to the gold surface. At saturation coverage of this phase the monolayer undergoes a phase transition to a denser phase with a realignment of the molecular axis to orient normal to the gold. Vacancies in the gold lattice translate to holes of substantial width (3-5 nm) in the final condensed alkane thiol monolayer. Tsai and Lin (2001) studied the adsorption induced by Au203interactions with polar species such as-OH, -COOH, and -PO3H2 on the terminating groups of the SAMs and found they caused serious packing disorder in the alkanethiols films. This they ascribed to the interactions preventing the complete reordering of the alkanethiols. They suggest this effect may be minimized by plasma oxidizing the entire surface and then reducing it to metallic gold by ultrasonic agitation in oxygen-free ethanol. This result indicates the need for good control over the assembly process to achieve a reproducible SAM. Fenter et al. (1994) report grazing incidence X-ray diffraction data of CH3 (CH2)9SH self-assembled monolayers on Au(111). It rather surprisingly revealed a disulphide head group structure. The alkane chain was found to possess a gauche bond, thought to result from the need to reconcile the hexagonal packing of the alkane chains with the S-S head group packing. This was not expected since other studies have suggested that S-S groups are reduced to thiols on binding to the gold. Nuzzo et al. (1987) had previously demonstrated the existence of cleavage on adsorption to gold surfaces. Steric effects may have a significant contribution to the reduction of disulphides to thiols at gold surfaces. The major difference between disulphide and thiol SAMs is in the kinetics of the film formation favoring the thiols by 75:1 (Bain et al., 1989). A significant challenge in the fabrication of devices based on SAMs is to have control over the relative composition of the adsorbed film. Folker et al. (1994) have examined the relationship between the composition of a two-component self-assembled monolayer on gold and the composition of the solutions from which they were formed is examined. A two-component monolayer of HS(CH2)21CH3 and HS(CH2)llOH was formed from solutions in ethanol. The study demonstrated the theoretically anticipated outcome that for a twocomponent system of alkanethiols on gold, well equilibrated with alkane thiols in solution, a single phase is preferred at equilibrium; phase separated two component monolayers are not observed. Thus, under the ideal conditions of a strongly cooperative film formation at the gold surface and equilibrium film growth, little or no control would exist over the relative numbers of the different component species in a mixed film on the gold. 463

Comell Two approaches are available to mitigate against this effect: the first to assemble the film under kinetic conditions and the second to reduce the cooperativity of the assembly process. As will be seen later, the ICS TM biosensor comprises phytanyl chain lipids and also a substantial ethylene glycol spacer group, and comprises four tethered species which will reduce the cooperativity of the formation of the mixed film. The electrochemical sealing ability of the phytanyl groups used in the ICT biosensor further reduces the cooperativity of the film formation, and the bulk and dynamic disorder of the phytanyl side chains provide a better and seal against electrochemical stripping. (Braach-Maksvytis and Raguse, 2000). Fuyuan and Lennox (2000) have further explored this effect and report data for SAMs prepared from 5 mM solutions of HSC~502H in C16SH in ratios of 10%90% deposited over either >24 hours in the absence of applied potential, and the same components prepared over <10 minutes, at an applied potential of +0.6V relative to a Ag/AgC1 reference in Ru (NH3)6C1/1M KBr solution. The driven assembly of SAMs favors the kinetic, metastable regime. The potentially driven SAMs formed an excellent seal in <10 minutes compared to no seal existing under the same conditions but in the absence of a potential. This was also evident in the composition of the binary film being uncontrolled in the absence of applied potential; however, with the 0.6 V applied to the gold during assembly, the composition was controlled proportionately to the relative concentration of the two components. Potential-driven film formation has been used to create patterned SAMs. Hsueh et al. (2000) describe an electrochemically directed adsorption process to selectively form SAMs on a particular gold electrode in the presence of another nearby electrode. The monolayers were very similar in thickness, wettability, blocking of heterogeneous electron transfer and elemental composition to analogous SAMs formed by chemisorption. The authors claim that advantages include: spatial control over coverage, short time required for coating (less than 1 minute) and even an ability to form SAMs on gold that is not freshly evaporated. Other mechanisms that create a spread of properties across a gold surface include the topographical and morphological properties of the gold. Walczak et al. (1995) used voltammetric techniques for the reductive desorption of alkanethiolate monolayers from gold electrodes and found that different adsorbate binding sites exist. These difference sites were associated with lattice defects in the gold surface and were also influenced by oxide sites at which thiol desorption was 2-3 times easier. Dishner et al. (1996) showed that during the self-assembly process, sulphur atoms adopt a (3 !/5 x 3V~)R30 ~ structure, i.e. an hexagonal lattice commensurate

464

Membrane-based Biosensors with the Au(111) structure but rotated 30 ~ relative to the gold lattice. The chains of the thiolates extend into space with the same all-trans conformation observed in crystalline paraffins and polyethylene. They report that scanning tunneling microscopy (STM) observations have revealed that the chemisorption (presumably desorption) of the thiol is accompanied by the formation of 3-10 nm pits. They describe the elimination of these pits by UV photolysis. The photolysis oxidizes the thiol to its corresponding sulfonate which can be rinsed from the surface with ethanol or water, leaving the surface smooth and 'pit' free. They suggest this occurs due to accelerating the lateral diffusion of the gold causing the pits to fill in. The pits have dimensions on the nanometer scale, typically 3.3 nm dia., involve nearly 100 gold atoms and typically represent approximately 10-15% of the surface area. These authors reference Bucher et al. (1994) as having produced defect-free films by heating SAMs of n-octadecylthiol to near the boiling point of ethanol at 77~ No impedance measures were reported to indicate whether damage had occurred to the electrical seal of the film at these temperatures. We have observed the introduction of gold in previously gold-free ethanol solutions resulting from soaking gold-coated slides in alkanethiol and disulphide containing solutions.

3.2.2. Assembly of bilayer lipid membrane. Sackmann (1996) and Meuse (1998) have provided reviews on the development over the last decade of the literature on supported membranes. Meuse et al. (1998) prepared stabilized BLMs comprising an alkanethiol inner monolayer and a dimyristoylphosphatidylcholine, mobile outer layer. These authors termed these structures hybrid bilayer membranes (HBMs) and demonstrated that they may be formed from lipid vesicles or by transfer from air-water interfaces. These bilayers are non-interdigitated and by vibrational spectroscopy indicate a more ordered structure equivalent to the effect of lowering temperature. Steinem et al. (1996) studied a family of both thiolated and nonthiolated phospholipids to generate tethered and supported bilayer membranes. Three strategies were demonstrated: 1) The gold surface was initially covered with a chemisorbed monolayer of an alkanethiol such as ODT or a thiolated PC such as 1, 2-dimyristoly-sn-glycero-3phosphothiolethanolamine (DMPTE). Following this a second lipid layer was deposited by: Langmuir-Schaefer technique, from lipid solution in ndecane/isobutanol, by lipid--detergent dilution or by fusion of vesicles. 2) Charged molecules carrying thiol-anchors were attachment to the gold surface by chemisorption, or 3) Lipid bilayer vesicles containing a thiolated phospholipid were directly deposited. The ion channel gramicidin A was co-deposited using the 2 ~d technique described above. Membrane capacitances of 0.55 lxF/cm2 +__ 10% were obtained independent of the technique employed. Membranes seals were obtained to frequencies of approximately 30 Hz for a typical bilayer; when doped with gramicidin, the seal could be raised to approximately 100 Hz. Meuse et al. (1998) formed hybrid bilayers by the interaction of phospholipid with the hydrophobic surface of a self-assembled alkanethiol monolayer on gold. These membranes were characterized in air using atomic force microscopy, 465

Cornell spectroscopic ellipsometry and reflection-absorption infrared spectroscopy. From this analysis, the added phospholipid was one monolayer thick, was continuous, and exhibited molecular order similar to that in phospholipid bilayers. When reintroduced into water and characterized using neutron reflectivity and impedance spectroscopy, it was found that the hybrid bilayers also reflected essentially the same properties of phospholipid bilayers. The bilayer leakage level was not reported below 10 Hz. Plant (1993) measured the capacitance of phospholipid/alkanethiol bilayers and found them to closely mimic that obtained from solvent free phospholipid bilayers. The introduction of the 106M mellitin was observed by cyclic voltammetry to change the bilayer from being impermeable to being highly conductive to K3Fe(CN)6 in 1 M KC1. Stelzle et al. (1993) published the first detailed report of the proposed use of a supported bilayer membrane as a biosensor. Using a monolayer of a carboxy mercaptan deposited onto a 100 nm thick evaporated gold film on a silanized glass substrate, positively charged vesicles of dioctadecyldimethylammonium bromide were fused to form a supported bilayer membrane. Impedance spectroscopy from 2000 Hz to 1 Hz revealed sealed membranes down to frequencies of approximately 100 Hz. At this point defects in the membranes required a more complex network to explain the data.

3.3. Chemistries of membrane components Before describing the assembly of the ICS biosensor, the chemical structures of the tethered and mobile components and the gramicidin are reviewed. Representative examples of the materials from which the device is assembled are shown in Figure 1. The synthesis of most of the compounds has been described elsewhere (Raguse et al., 1998, 2000; Bums et al., 1999). They are divided into two families, one that is tethered to the gold surface and one that is physically absorbed to the surface but free to diffuse in the two dimensional plane of the membrane. The structures in Figure 1 provide examples of both. TM

The concentration of 1 is 350 pM and of 5 is 2 mM. Typical mole ratios of l:2(R1):2(R2):3 are 40,000:400:1:1 and of 5:6:4 are 28,000:12,000:1. In compounds 1-3, attachment of the membrane to the gold substrate is via a disulphide moiety. The use of this disulphide compound brings both a tether for the membrane lipids and introduces a benzyl spacer group that lowers the twodimensional packing density of the assembled membrane. The lower packing density facilitates the entry of ions into the space between the membrane and the gold surface. This space is termed the reservoir. Other factors which influence the properties of the reservoir are discussed below in Section 5.7.

466

Membrane-based Biosensors A further series of tethering compounds are shown in Figure 2. Compounds such as 7, termed the C11 series, possess both the sulphur attachment chemistry and an additional binding energy generated by the Van der Waals attraction of the C l l sequence for each other and for the gold surface. In addition these compounds protect the surface of the gold from the electrolyte solution, lowering the surface charge and minimising deleterious electrochemical effects. The reduction in surface charge also significantly alters the reservoir performance. In compounds 7 and 8, all-ether reservoir linkers are substituted for the succinate groups of 1-4. This eliminates instabilities that can arise from the hydrolysis of ester groups and significantly extends the storage lifetime of the membrane. The all-ether reservoir also beneficially alters the reservoir properties. The area per lipid molecule is determined by a balance of contributions from at least three regions within the membrane, the sulphur-gold interface, the spacer molecules and from the interaction of the hydrocarbon chains within the body of the membrane. The area per lipid can range from 20 ~2 to 150 ~2, depending upon the bulk and position of the dominant packing constraint. Membrane-spanning lipids such as 2 and half membrane-spanning lipids such as 1, 5 and 6 are mixed in different ratios to adjust the membrane packing. Packing at various levels within the membrane may also be adjusted by the series shown in 9 and by the spacers 10a and 10b shown in Figure 3. As discussed above, the phytanyl chain lipids used here are commonly found in thermophilic archibacteria (Dante et al., 1995). They are chemically stable at high temperatures (Stetter, 1996), and the bulkiness of the methyl substituents reduce temperature dependencies in the membrane disorder around the normal operating temperatures for the biosensor of 20~ to 30~ They also provide an excellent electrical seal for impedance measurements. Biphenyl linkers at the mid-plane of the bilayer makes the structure more rigid and prevents the membrane-spanning lipids from entering and emerging on the same side of the membrane (Jordan, 1999). An extensive literature exists on the physical chemistry and ion transport properties of the linear gramicidins (Koeppe and Andersen, 1996), Wallace (1999). Gramicidin A is a low molecular weight pentadecapeptide (MW1882), which occurs in the soil bacteria Bacillus brevis (Katz and Demain, 1977).

467

Cornell

Tethered species

le

s-~O~o~o~o~~O~o~o~o~o , , ~ o ~ S...~

2.S~

0

0

'~

0

0,1---..,0,1---,..0 - ?0 0 ~ 0

(~

~""0~ ~ O ' V ~

s,-~

aa

o

Rlor2

o - v ~. v v 1" v v 0 , . % 7 _ L ~ , o

.

O

S ~ 0""0"~" ~ O ' ~ f i l

0~-- 0-~.0,~-0--~ 0 i ~ ~ 0 . ,

s--O

o

~

o

ramic id in

o

Mobile species

0

4. ~amicidin

/

R2

-

O

+

0_~, 0 ~

N(CH)a3

~ Where for both mobile and tethered species"

R~=H

H 0 0 H 0 R2 =_ OIg,,..,,,v,-~N/1.,'-v,.x.,~NuV,,,v,,~. NI1.,.-,,,.,,,'-,,.~N~ 0 H 0 H

H qO NH b-

Figure 1. Chemical structures of representative compounds from which the tethered membrane is formed. Compounds 1-3 depict tethered species and 4-6 the mobile species. A minor component of 2 and all of 4 are linked to R2, an aminocaproyl-linked biotin. 468

Membrane-based Biosensors

Cll 7.

8. ~/x~''6~~''ff'/~''ff''~''ff'r ~

o

-

J

o

o

Figure 2. Additional tethered compounds. In 7 a Cll alkane provides more secure tethering, shields the substrate from electrolysis and improves the reservoir properties. In compounds 7 and 8, the succinate groups are replaced with all ether linkers extending the storage lifetime and improving the reservoir properties. Compound 9 permits adjustment of the packing constraints within the hydrocarbon region of the membrane.

lOa

s.~oH

I SyOH 0

lOb

Figure 3. Spacer groups used to vary the two-dimensional packing within the membrane.

Depending on its solvent environment, the conformation of gramicidin ranges from random coil to a family of intertwined helices. However, when in a bilayer lipid membrane it forms a single coil, right-handed a-helix. This structure is triggered by four tryptophans in positions 9, 11, 13 and 15 anchoring the C terminus of the helix at the lipid-water interface. The structure is further stabilised by nine internal hydrogen bonds between the adjacent turns of its 2.5 turn helix. Unlike mammalian proteins this structure is made possible by the alternating L-D amino-acid sequence found in bacteria. The conducting form of gramicidin is an n - n dimer o.f ~t-helices, with three hydrogen bonds per monomer contributing to a six hydrogen bond stabilisation of a five-turn dimer that traverses the hydrocarbon core of the BLM. As a conducting dimer, gramicidin selectively facilitates the transport of monovalent cations (Myers et al., 1972).

469

Comell

Figure 4. BLM measures of the single-channel conduction of native and modified gramicidins (left) and examples of structures containing aminocaproyl (C-F) or tetraethylene glycol (B) linker groups (right). The single-channel current is shown in parenthesis adjacent to each structure.

Gramicidin is an ideal choice for engineering a device based on ion channel switching. Unlike large mammalian channels, gramicidin is chemically very stable in both its primary and secondary structure. It can be modified using standard synthetic organic chemistry techniques and then introduced into a bilayer lipid membrane which causes it to spontaneously to fold into the ion conducting form. Ethanol maintains gramicidin as a random coil at below mM concentrations. It is important to ensure that when forming the a-helix in a BLM that the gramicidin has not been transferred from a solvent such as dioxane which may require many hours to convert from the conformation adopted in dioxane. In order to utilise the properties of gramicidin as an ion channel switch, it is necessary to attach linkers for both the tethered and the mobile species. Many analogues of gramicidin A have been synthesised. The common attachment point is the ethanolamine hydroxyl group. The linker structures have mainly been variants on aminocaproyl and tetraethyleneglycol groups respectively for the mobile and tethered gramicidins. Each is tested for single-channel conduction to ascertain the conductivity per channel. Of the many variants made of gramicidin, provided the linker attachment was to the C terminus ethanolamine hydroxyl, the conductance fell within a range of approximately +10% of the native gramicidin. Modifications too near the channel entrance can eliminate conduction. Suarez et al. (1998) have reported a similar 'C' terminus biotin-labelled gramicidin.

3.4.

Antibody fragments

Fab' fragments are used in preference to whole antibodies. The use of whole antibodies can cause species cross-reactivity. A test sample of blood in which anti-mouse antibodies have been raised as a consequence of, for example, a food 470

Membrane-based Biosensors

Figure 5. A fresh gold surface is exposed to an ethanol solution of the tethering species 1-3, given in Section 3.3 above, for 10 minutes. This produces the inner and part of the outer leaflet of the bilayer membrane. Following an alcohol rinse, a second ethanol solution brings the mobile elements, 4 and 5 of the membrane. Rinsing with water spontaneously forms a lipid bilayer structure. Some of the lipids span the membrane, whilst the remainder are mobile within the two-dimensional plane of the membrane. Antibody fragments are then added in the aqueous solution and tethered using a streptavidin-biotin attachment (not shown). For large analyte detection involving a "sandwich assay" configuration, assembly is by adding an equimolar mixture of the two antibody fragments. contamination can result in the a cross-linking of the species-dependent Fc portion of the antibody, causing a false response. Whole antibodies are enzymatically cleaved at the hinge region and biotinylated. Attachment of these fragments to the membrane is usually performed using streptavidin-biotin complexes. This provides convenient and readily available linker chemistry, although other linker chemistries have proven satisfactory for certain applications. These include covalent coupling and a novel organometallic coupling chemistry. 3.5. Biosensor assembly As described above in Section 3.2, the assembly of the lipid bilayer membrane can be achieved using a variety of approaches. All depend upon using a tethered hydrocarbon layer on the substrate to achieve a tight bonding of the mobile leaflet of the membrane. In the present example of the ICS biosensor assembly of the bilayer membrane is occurs during an ethanol/water rinse. This is shown schematically in Figure 5. TM

471

Cornell

Figure 6. The binding of analyte to the antibody fragments causes the conformation of gramicidin A to shift from conductive dimers to non-conductive monomers. This causes a loss of conduction of ions across the membrane. A competitive assay has also been devised in which the analyte causes the population of channel dimers to increase.

4. Principle of Operation Many attempts to engineer receptor-based gated ion channels have been reported. Mechanisms range from anti-channel antibodies that disrupt ion transport (Buffer et al., 1996), to molecular plugs that block the channel entrance (Lopatin et al., 1995). All mechanisms so far proposed have had a very limited range of application and require re-engineering for each new analyte. The ICS TM, whilst using ion channel transduction provides a mechanism that may be adapted to many different classes of target.

4.1 Detection of large analytes Large analytes include proteins, hormones, polypeptides, microorganisms, oligonucleotides, DNA segments and polymers. In the same manner that an ELISA "sandwich" assay may be developed based on a complementary antibody pair, the ICS biosensor may be adapted to the detection of any antigenic target for which a suitable antibody pair is available. The bacterial ion gramicidin A is assembled into a tethered lipid membrane and coupled to an antibody targeting a compound of diagnostic interest. The binding of the target molecule causes the conformation of the gramicidin channels to switch from predominantly conducting dimers to predominantly non-conducting monomers. TM

4.1.1. Examples of performance. With MW 28kDa, thyroid stimulating hormone comprises an cc and 13 subunit to which a matched pair of antibodies is used, each targeting different non overlapping epitopes. The response to TSH is shown in Figure 7.

472

Membrane-based Biosensors

. . . . . . .

0.1 pM

2 pM

312,p M~,.~,~

Time

Figure 7. Thyroid stimulating hormone (TSH) response curves. The response to TSH over the initial 300s following the addition (arrow) of analyte.

Ferritin is the principal iron transporting protein in human serum.

With a molecular weight of -- 450 kDa, ferritin is one of the largest soluble proteins regularly measured clinically. It contains 24 equivalent subunits each accessible to cross linking such that only a single Fab' type is required to elicit a response. The titration curve seen in Figure 8 shows the response of the sensor to ferritin in patient serum. A comparison of the response of the ICS biosensor with a Diagnostics Products Inc. analyzer is given in Figure 9 taken from 100 patients selected to provide a spread of ferritin values over the clinical range. TM

Performance constraints. Assuming an adequate mass transport of analyte to the membrane and a surface density of tethered antibodies [T] significantly in excess of antibodies linked to the mobile channels [G~], the idealized behavior of the biosensor may be modeled by a family of coupled equations. The most significant of these is given in Figure 10.

4.1.2.

The two-dimensional (2D) reactions on the membrane surface are generally faster than the three-dimensional (3D) reaction rates between the analyte in solution and the surface (Hardt 1979). It is therefore anticipated that for low and medium analyte concentrations that the 3D processes will be rate limiting. At high concentrations, 2D processes will become important. These limiting conditions may be estimated from expressions (1) to (3) below.

473

Comell 0.01

-r

9 r~

r~

0.001

O

J.

T

0.0001 _ .

10

100

1000

10000

Concentration Ferritin / Human Sera (pM) Figure 8. Response rate (/sec) for a range of concentrations of ferritin in human serum. The rate is obtained by normalizing the initial slope of the response curve to the initial admittance. The initial slope of the response is obtained within the first 180-300s depending on the analyte concentration.

o

1000 + o r

9=- 100 9 ,,,.4

=

~

10

r..) ~ 1

Correlation coefficient = 0.!

1

1

1

10

100

1000

Immulite measured Ferritin concentration pM Figure 9. A comparison of estimated Ferritin concentrations from 100 patients measured by a Diagnostics Products "Immulite" and the ICS biosensor. The patients were chosen to provide a wide spread of clinical values. The measurements were made in serum at 30~ TM

474

Membrane-based Biosensors At low concentrations, expression (1) is dominant and the flux of analyte captured at the membrane surface is given by [T] k3D [A] molecules/cmE/s. At high concentrations expression (3) is dominant and the response is independent of [A]. When expressions (1) and (3) are similar, analyte capture by the tethered antibody and its cross linking to mobile gramicidin occur at comparable rates. This produces a response proportional to [A] rE. A further limiting condition is the lifetime of the dimeric channel given in expression (2) by (k-~2d)1. Under these conditions, the response time is also independent of analyte concentration [A]. It is thus straightforward to numerically simulate the device behavior for large analyte detection. These simulations are shown in Figures 11 (a) - (c). From Figure 11, it can be seen that for the conditions chosen here, at low analyte concentrations, two-to-three decades of linear response to analyte are available. At high concentrations, approaching nanomolar, the gramicidin dimer lifetime, k2a-1 or the cross-linking rate of the mobile ion channels to the tethered antibodies, k2D"1 will be limiting. The variables available in establishing the operating conditions are: the surface density of tether sites IT], the surface density of mobile channels [GM ] and the "on rate" of the chosen antibody kaD. This example used ferritin, and thus only a single epitope k2D is required. The "on rate" of the antibodies attached to the mobile channels in general need not be as large as those on the primary tethered capture sites [T] since 2D processes are more effective. The ratio [T]/[G~] 'amplifies' the apparent capture rate of analyte from what would otherwise be the simple first order rate constant, k3D[A] to k3D[A][T]/[GM]. The maximum amplification in this configuration approaches 102-103 and depends on the length of the linkers and whether streptavidin has been used as a coupling protein. At higher analyte concentrations, in the range nanomolar to micromolar, the amplification may be adjusted downwards by lowering, even below unity, the [T]/[GM] ratio. The curves shown in figure 11 were obtained for the relatively low Fab' capture density of [T] = 1.5 x 101~molecules/cm2. This is to circumvent a dependence on viscosity. With the introduction of flow, 1 ~l/min. to the analyte stream the mass transfer limitation can be overcome and the capture density [T] increased by an order of magnitude with a proportionate increase in the response.

475

Cornell

K3D [T ] + [A]

[T*A]

where K3D = k3D/k "13D(1)

k -13D

k2d [GM] + [Gx]~--~ [GD] k -12d

where Kd = kz~/k lza

(2)

K2D [T'A] + [GM]

~ [T*A*GM] k -12D

where KED = k2D/k "IA (3)

where the following symbols denote: [T]

the surface density in molecules/cm 2 of the tethered antibodies attached to the membrane spanning lipid; [h] the analyte Modifications to the channel lifetime are possible through altering the membrane thickness or the channel length, concentration in mole/L; [T'A] the surface density in molecules/cm e of complexes of analyte with a tethered antibody [GM] the initial surface density in molecules/cm 2 of mobile gramicidins linked to antibody and diffusing over the outer surface of the membrane; [GT] the surface density in molecules/cm 2 of gramicidins tethered to the gold film on the electrode. [GD] the surface density in molecules/cm 2 of conducting dimers of gramicidin. k3D, k-13D the three dimensional forward and reverse reaction rates for analyte capture by tethered antibodies k2d and k "~Zd the forward and reverse reaction rates for tethered and mobile gramicidins forming dimers. kEDandk "~:o the forward and reverse reaction rates for tethered and mobile antibodies by forming a "sandwich" complex with gramicidin A. K3D, Kd and Km the equilibrium constants for interactions (1), (2) and (3).

Figure 10. Expressions describing the detection of analyte (1), the formation of gramicidin dimers from monomers (2) and the formation of cross-linked monomers of gramicidin to tethered antibodies following the capture on analyte at the sites of the tethered antibody. The quantitative measure of analyte concentration may be taken as the initial rate of current increase. 476

Membrane-based Biosensors

. . . .

0.01

9

,

tool "isec -I 6e5 2e5

0.001

6e4

o 9

......................... i

2e4 6e3

~ 0.0001 Z

O.O000I

~ - 1

S-1 0.01

,.....--.- 0.08s-I 0.04s-1 0.02s-I

0.001 0.01 s"l

"

Patient sample

9

PBS

9

Calibratoz solutione

O.OOgs -1 ~ 0.0001

0.00001 T~te ~ o n s t a n t

0.01

cm 2 m o l ' l s e e ' l

~ . le-7

. le-8 a) a~

0.001

le-9

o

~ le-lO

r

le-lI

e~0.0001

0.00001

0.

I

10

100 1000 10000

Ferritin C o n c e n t r a t i o n (10Ir

Figure 11. A c o m m o n set of measurements simulated using [T] = 1.5 x 101~ molecules cm "2, k3D = 5 x 105 M l s "1, k3D'l=103s "1, k2D = 10"Tcm2molecule'ls "l, k2d = 1.25 x 10 11 cm2molecule'lsl, k "12d = 1.25 x 10"2S, [GM] = 1 X 108 molecules crn "2, [GT] = 3 x 101~ molecules c m 2. In (a) k3D, in (b) k2d, and in (c) k2D are varied to demonstrate limiting conditions.

477

Cornell

Figure 12. In the absence of analyte, the mobile channels cross-link to anti-digoxin Fab'S anchored at the tether sites. Dimer formation is prevented and the conductance o the membrane falls. The introduction of analyte competes of the hapten and increases the membrane conductance.

, a~F+~ + ' t ~ ' 4 q ' 4 4 - ~ + .+--r, 9+.1:-~*

.o .-r w

\

0

r,.)

b- Fab

digoxin

! I

i i

+,+.

l 4. ~ ++§

4"++ +-~ ++.4..t.~++.~1 "

9

0

200 400 600 Time (seconds)

800

0

,

9

|

'

!

9

!

9

!

9

! ...... 9

200 400 600 Time (seconds)

|

9

i

800

Figure 13. ICS TM system configured for the cardiac stimulant digoxin, (MW 781) linked at the 3" position to gramicidin through a flexible tetra-aminocaproyl group (left). During fabrication, when the biotinylated Fab' fragment is added, the conduction is reduced due to cross-linking of mobile gA-digoxin with tethered antibody fragments. The addition of digoxin displaces the gA-digoxin, and the conduction returns (right). Addition artefacts resulting from matrix effects are differentially cancelled.

4.2. Detection of small analytes For target analytes with low molecular weights such as therapeutic drugs where the target is too small to use a two-site sandwich assay, a competitive adaptation of the ICS T M is available. This is shown in Figure 12. 478

Membrane-based Biosensors

4.2.1. Performance constraints. Equation 4 adds a further reaction to the list in Figure 10.

[T*GMh]

[T] + [GMhI

(4)

k-~2 where the following symbols denote: [GMh]

k "lAh

the surface concentration of mobile gramicidins with attached hapten analogues of digoxin. is the "on rate" of the digoxin hapten for the tethered anti digoxin FaD. H is the survival lifetime of the hapten/Fab' complex.

Competition for the tethered antibodies between analyte and hapten on the mobile gramicidins establishes an equilibrium which determines the surface density of the complex [T*G~]. As with the large analyte example the current output from the sensor is determined by the surface density of gramicidin dimers [GD]. The quantitative measure of analyte concentration may be taken as either the initial rate of current increase. With the competitive, small analyte system (Woodhouse et al., 1999), the 'amplification' effect described above is not available. However, it is still possible to adjust the sensitivity of the device over a considerable range by manipulation of component surface densities.

4.2.2. Platform Response. One of the major advantages of the ICS biosensor is TM

the breadth of applications based on a common transduction mechanism. The principal requirement is an attachment chemistry that will not inactivate the receptor. Well over 50 antibodies have now been adapted to the ICS~ biosensor, as have metal chelates for heavy metal detection, lectins such as concanavalin A for glucose detection, oligonucleotide probes for DNA strand detection, and extracellular cell surface receptors for growth factor detection.

479

Cornell

~ ...................................................................

.'"

,"

:d

(.)

s

~

.......

12.5 ng mL-1

.......

6.25 ng mL-I

---

3.12ngmL-1

" "~'

--

---

J

S

~

~

~

"-

1.56ngmL-1

- 0.78 ng mL-1

u

" - - - 9 0.39 ng mL-1

:/

0

.~

s

.~':t. .

~"

..,"

9 0.19 ng mL-1 ~'-

~-~'..d.. . . - " "

500

1000

15'00

2600

0 . 0 n g m L "1

2500

Time (s)

Figure 14. Reponses to a range of concentrations of digoxin spanning the clinical range of 1 ngmL 1 .

6O 50

Input [gA-dig] = 5E8 - , - Model k2D h = 5E8 --4- M o d e l kEDh= 5E9 M o d e l k2Dh = 5E 10

molecules c m -2 molecules "2cm s l molecules 2 cm s 1 molecules 2 cm s"1

Input [gA-dig] = --,-- Model kEDh = --'-- Model kEDh = ---b-Model kEDh=

molecules c m "2 molecules "2c m s "1 molecules "2cm s 1 molecules "2cm s "1

O

40

3o r3 20

Experimental nominal gA-dig

lO

A

o

. . . . . . . .

1E-9

1E9 5E8 5E9 5E 10

!

. . . . . . . .

1E-8

!

1E-7

. . . . . . . .

!

1E9 m o l e c u l e s cm-2

I'l 5 E 8 m o l e c u l e s c m 2

1E-6

[Digoxin b-Fab] (nM) Figure 15. Superposition of the results of the response to the addition of the b-Fab' with numerical simulation. [gA-dig] refers to the surface density of digoxin labelled mobile gramicidins, k2Dh refers to the two dimensional "on rate" of the anti-digoxin Fab' for the digoxin-labelled gramicidin. Two nominal experimental conditions are shown for the mobile gramicidin surface density.

480

Membrane-based Biosensors 50 40 O o,..~

o~

30

t:m

'

g

$

. . .

'

20 10 -. Model I

I'

Jl

i

,i

I

~

v

o

~

u

~ ''I

i

I

,,

I

!

I0

Concentration (nM) Figure 16. Endpoint gating ratios. The % change in conduction upon the addition of digoxin at the concentration is shown, superimposed on the numerical model described in Woodhouse et al. (1999). The response time to analyte challenge is essentially fixed by the off rate k "lAh of the hapten from the Fab'. In the present example of digoxin, k l # --~ 50 s.

(c) TSH

(a) oligonucleotides

%

\

\

L..

0 ~ ~oo" l?oo" l;o0" ,_'ooo Time (seconds)

\

\

rj

0

400

800

1200

Time (seconds)

0" ",.'d ' "~0;" ~ ' " Time (seconds)

Figure 17. Examples of responses to challenges of different classes of target analyte. (a) A 19-base oligonucleotide probe specific for the 16S rRNA sequence of Listeria monocytogenenes biotinylated at the 5" end via a 23-atom phosphoamidase linker. The response seen is to a 5 nM challenge of a 52-mer target sequence. The double-stranded target was heated to 95~ and cooled to approx 60~ The biotinylated probe on the gramicidin hybridised more rapidly than the its complementary sequence, resulting in the gating shown. (b) Biotinylated Fab' directed to E coli. The response shown was elicited by a challenge of 105 cells/ml. (c) B iotinylated complementary pair of F,b's directed to two epitopes on the 13subunit of TSH. The response shown was elicited to 100 pM TSH. 481

Cornell

901

92.2 ......................................Etohwash ............i........................... 92.4[ Lip[diayer .~......... 92.6 .........

............ ..... Thickness change = 2nm

93.0 ..................................... .............................. 93.2! . . . . . . 93.4 ' " . ..... , ~ . . . 93.0 93"8 ! 94.0 0

"" .

. 1

. 2 Time (min)

ODT

.

3

4

Figure 18. Ellipsometry measurement, A, related to thickness of the outer layer. An ethanol rinse of a bilayer membrane assembled on octadecane thiol (ODT) resulted in a 2 nm change in thickness. This is close to that expected for a fluid monolayer. In all instances, the upper curve is a control in which an anti-theophylline antibody was used and the lower was the active measure.

5. Characterisation 5.1. Ellipsometry Ellipsometry permits a measure of the membrane thickness. Direct calibration using alkane thiols from C8 to C,8 yield a thickness for the biosensor membrane of 4 nm. The mobile outer layer contributes 2 nm and is measured by the change in thickness on rinsing off the layer with ethanol. The overall thickness of the tethered bilayer is measured at approximately 12 n m - 15 nm, although there are many assumptions of the dielectric constant for a tethered membrane separated from the electrode by a polar linker in estimating this value.

5.2. Impedance spectroscopy Membrane thickness can also be determined using the membrane capacitance derived from modeling the impedance spectrum over a swept frequency range of typically 0.1 H z - 1000 Hz. Both phase and modulus are measured and fitted. Simple RC networks provide a fit with <2% residual. A measured capacitance of 0.5 [xF/cm2 is obtained, assuming a relative dielectric constant of 2.2 for the membrane chains again estimates an overall membrane thickness approximately 4nm. The electrical equivalent circuit of a tethered bilayer membrane is shown in Figure 19. 482

Membrane-based Biosensors

Figure 19. Equivalent electrical circuit used here to characterize the tethered lipid bilayer membrane. The various elements are as described in Krishna et al. (2001). The key elements are the conduction element GM which describes the ion flow across the membrane, dominated by the gramicidin channels, the membrane capacitance CM, the Helmholtz capacitance, Ch, of ions crowding against the gold electrode and the diffuse capacitance Cd of ions within a concentration cloud decaying back into the reservoir space. For an excitation of typically 50 mV, the equivalent circuit may be approximated as an effective Helmholtz capacitance, Ch of 3.5 ~tF/cm2, in series with the membrane capacitance, of 0.6 ~tF/cm2 .

Figure 20. The impedance of a capacitor varies according to Z=89 A lOgl0Z/logfl0 axes yields a linear relationship with a slope of-1 and an intercept at 89 of 8 9 logl0C. At high frequencies C = Cm and at low frequencies C = Ch where Cm and Ch refer to the capacitance of the membrane and Helmholtz layers, respectively (left). By observing the frequency at which the impedance profile crosses from Cm-dominated to Ch-dominated behavior, it is possible to follow the change in Gm arising from the ion channels. The crossing point is conveniently identified by the frequency f.~ at which the phase angle ~, between the excitation and the resultant current, is minimized (right).

483

Cornell

- n - 2.8e10 innergA _A- 7.0e10 innergA ,

~ 40

30

00

~------------

- A

A

2010 ~

E r

O C O o,,.

(K2D= 1.0X109 )from titrating outer layer

~~

1 ~ ~

o

c r c O

rn~

E!

I

9

I

9

I'

9

0.0/ ~l.0xlb 2.0xlb 3.0xlb 4.0xlb 5.0xltl 6.0xlb 7.0xlb 8.0xlb

L)

K2D

N o m i n a l n u m b e r of o u t e r l a y e ~ l A m o l e c u l e s z p e r c m

Figure 21(a). The gramicidin dimerisation rate constant, K2D, derived from titrating the membrane conduction against the nominal inner layer gramicidin concentration in molecules crn 2 from 1 x 10 ~~ crn -2 to 8 x 101~ cm 2 for 7 x 108 cm 2 and 1.4 x 109 cm -2 nominal outer layer gramicidin densities.

i

--O-- 7e8outergA --/k-- 1.4e9outergA

i

A =.

20-

O

(K

_E

2D

=

9 )fromtitratingi ....

7.0X10

layer

O C

~ O

o

om

C O U C

o '

i

'

o.oJl.oxlo,o 2.oxlo~ a.oxlo,o K

J

I

'

'

I

4.oxlo,o 5.oxlo'o 6.oxlo'~ 7.oxlo'~ 8.o•1620

Nominal number of inner layer

gA molecules per cm

z

2D

Figure 2 l(b). The gramicidin dimerisation rate constant K 2 D derived from titrating the membrane conduction against the nominal outer layer gramicidin concentration in molecules cm -2 from 10 9 c r n "2 to 8 x 10 9 c m "2 for 2.8 x 101~ 2 and 7 x 101~crn 2 nominal inner layer gramicidin density. 484

Membrane-based Biosensors The validity of this equivalent circuit as a good description of the tethered, conductive bilayer lipid membrane is shown in Figure 20, in which the impedance profile and the phase relationship are plotted.

5.3. Component titrations An important issue when trying to fabricate self-assembling systems is to determine and control the ratio of components in the film. Many factors will influence this: the composition of the coating solutions, the condition of the surface, and the conditions under which the deposition occurs. A very useful measure available in the gramicidin system is the ability to measure directly the ratio of gramicidin to non-gramicidin species within the film, based on changes in the electrical conductivity. As shown in Figure 21, it is possible to titrate Gm against the gramicidin concentration in the depositing solution. The conduction is found to be dependent on the insertion of gramicidin into the inner, outer or both layers. The derived K2o for the gramicidin monomer-dimer interaction obtained from Figure 21 is 7 x 1 0 9 c m "2 if the inner layer gramicidins are taken as reference and 109 cm -2 if the outer layer gramicidins are taken as reference. This would suggest that the tethered gramicidin surface density is lower than nominal; however, absolute estimates of the membrane composition require other techniques such radiolabelling.

5.4. Reservoir characteristics The conductance of the tethered membrane assembled from the compounds shown in Figure 1 is insensitive to ion type, indicating that it is not limited by the conductivity of the channel. The reservoir properties dominate the ion channel conduction. The composition of the reservoir linker significantly influence the magnitude and properties of the diffuse layer capacitance (Krishna et al., 2001) and the magnitude and voltage dependency of the apparent ion channel conductance. The ion channel conductance appears greater by a factor of eight to ten for the all-ether reservoirs, as seen in compound 8 in Figure 2. Spacers such as compounds 10a and 10b in Figure 3 improve the apparent channel conduction, but they also add complexity to the coating solutions and introduce variation in the composition of the tethered membrane.

485

Cornell Table 1. The effect of surface type on non-specific binding of 42 nm streptavidin. The tethered bilayer is predominantly a 70:30 mix of compounds 5 and 6 in Figure 1. This mix of phosphorylcholine and hydroxyl head groups results in binding below the level of detection. Other surfaces examined are: MSLPC, a membrane-spanning lipid, such as compound 2 in Figure 1, also possessing a phosphorylcholine headgroup; MSLOH, a further membrane-spanning lipid possessing a hydroxyl headgroup; ODT, octadecyl thiol; and OH-C11-SH. Surface

Signal (degree)

Tethered bilayer .... 0.02

Apparent Thickness <0.5

surface Coverage <1%

.....

MSL-PC

0.02

<0.5

<1%

MSL-OH '

0.15

3.5

8%

ODT

0.90

20.0

45 %

OH-C 11-SH

0.10

2.3

Both the all-ether and succinate-linked reservoirs improve their ability to store ions with the application of a 100 mV-300 mV potential, negative on the gold electrode. The effect of this potential is to increase the apparent conductivity of the ion channels. The reservoir species containing the C I 1 linkers have a greatly reduced dependence on this potential, in proportion to the lowered surface charge at the gold surface arising from the C l l coating. A consequence of the advantage of a reduced dependence is a lowered dynamic range between the Helmholtz and membrane capacitance.

5.5. Non-specific binding and matrix effects Ellipsometry has been used to determine the level of non-specific binding. Streptavidin provides a convenient model as the presence of biotin on the surface provides a control for the effect of binding. Table 1 shows the surface coverage derived from the ellipsometry data, on various tethered membranes. The tethered bilayer comprises the composition given in Figure 1. From Table 1, it is evident that a substantial fraction of phosphorylcholine groups within the tethered bilayer surface significantly reduces nonspecific binding of streptavidin. Figure 22 shows the effect of including biotinylated lipids in the membrane. The specific binding increases linearly with the inclusion of biotinylated lipids.

486

Membrane-based Biosensors -1.2

f-..

-1.0

fr~.~ [SA] = 42 '

-0.8

28

1:0

24

~'9

2

I'I ~, ',~' ~,j'~1:20

'"

20 16

.<

~o

-0.6

,| ..

~ -0.4

|

,-.I

m

8

N -0.2

s

~t .

l

0

'

I

2

'

.... i

4

:1000

-0

|

'

r.o

4

t

0.0

<

I

6

',

'l

'

~

'1

;

1O

I

12

;

I

14

'

i

16

Time, minutes Figure 22. The effect of including biotinylated membrane-spanning lipid, MSL4XB within the tethered monolayer of MSLPC. The apparent thickness of the film progressively increases reflecting the binding of the streptavidin to the membrane surface. 0 pM Ferritin / Blood .~.,~

o

50 pM Ferritin / Blood

o

O

Time (seconds)

Figure 23. Comparison of a challenge of whole blood contalning a 50 pM spiked ferritin sample to a whole blood control without ferritin.

487

Cornell

Figure 24. Schematic of a tethered BLM in which each of the components have a Cll sequence at the gold surface. Assemblies of thiols employing the C11 sequence are able to withstand extensive storage in ethanol at 50~ with no loss of material to solution. Disulphide films stored under comparable conditions lost a substantial amount.

The low nonspecific binding characteristics of the phosphoryl head group tethered membranes results in an ability to detect target analytes in whole blood without pre-preparation of the sample and minimal matrix effects. An example is shown in Figure 23.

5.6. Stability The stability of the tethered membrane depends on many factors: the lifetime of the sulphur-gold bond, the chemical stability of the various linkers tethering membrane, and the ability of the membrane to resist mechanical damage and loss of the mobile outer layer. Schlenoff et al. (1995) have reported on the rate of desorption of alkanethiols using 35S. In high purity TSF, desorption rates are 10.5 s1. The X-ray data of Fenter et al. (1994) on the structure of disulphides on gold is discussed, and it is concluded that both thiols and disulphides desorb as disulphides. For clean gold surfaces, they argue that defect sites bind sulfurs more tightly and that a defect-

488

Membrane-based Biosensors 16 14 d:l

~'

w

_

(storage at room temperature)

l

12

~

N

10

u

8

~

6

|

! -

0

T

~

10

~

20

1

E~her linked _ ~

30 Days

40

50

l

.,.

60

Figure 25. Frequency at minimum phase, used as a measure of electrical leakage, of a tethered bilayer with benzyldisulphides and sulfur attachment chemistries as seen in compounds 1, 2 and 3 in Figure 1. However, no gramicidin species were used in the outer layer, only compounds 5 and 6 in Figure 1. This provides a measure of the lifetime for which the membrane will maintain its electrical seal.

rich surface would therefore result in a more stable monolayer. The contribution of defect sites to the stability of a tethered BLM is not discussed. Practical recommendations for increasing stability include monolayers with multipoint attachments to the surface and also the choice of the solvent in which the monolayer is stored. Small amounts of water in an organic storage solvent has a strong impact in slowing the desorption rate. No loss of material was observed after storage in air. The use of a C 11 segment to a thiol or disulfide species, as shown in compound 7 in Figure 2, provides a substantial additional avidity for the retention of the monolayer components. Figure 24 shows schematically a tethered BLM in which all components have a C11 spacer incorporated at the gold surface. Also shown in Figure 24 and in compounds 1, 2 and 3 in Figure 1 are the succinate linkers. Replacement of these by all-ether linkers such as shown in compound 8 in Figure 2, have a further substantial effect on the stability of the membrane. The sealed membrane fabricated from the ether linkers is stable beyond three months when fully hydrated. Storage was at room temperature in the presence of 0.01% sodium azide. As indicated the upper curve comprises succinate linkers. In the data obtained in the

489

Comell

Figure 26. Response rate of sensor, fully assembled as shown in Figure 5, dried and rehydrated. The performance is compared to a control sensor which underwent no drying. Prior to drying, the bathing solution was exchanged with a solution of 10 mg/ml BSA 100 mM trehalose solution in 15 mM sodium phosphate pH 7.4 (Ligler et al., 1998).

lower curve, these have been replaced with ether linkers in the attachment of the membrane to the gold. As a further strategy for stabilizing the sensor, it is also possible to return sensing function on rehydrating a fully assembled freeze-dried sensor. This requires the full assembly of a hydrated sensor, up to the stage where analyte is to be added. The sensor is rinsed with a storage solution and then freeze dried. As seen in Figure 26, following rehydration, sensor function returns. Freezing a fully hydrated sensor also appears to cause no loss of function upon being thawed and returned to operating temperature. Under these circumstances, no cryoprotectants are necessary and excellent recovery has been obtained with phosphate buffered saline. These approaches provide strategies for the long-term storage of devices.

6. Future Opportunities Miniaturisation and patterning are two major opportunities for tethered membrane technologies. Davies and Khamsehpour (1991) have demonstrated ion beam milling at a resolution as small as 10 nM. Padeste et al. (1996) have created nanometre-sized gold electrodes by the simple technique of using 490

Membrane-based Biosensors

C}

O

O

~L

O

O2

NO2, O ~

OR ~o ~

O

0

.u~ht.,,.

Figure 27. UV irradiation may be used to selectively deprotect biotin sites and permit their fictionalisation using a streptavidin-biotin attachment.

nanometre-sized latex spheres as a lift-off mask. Kane et al. (1998) have patterned proteins and cells using soft stamps and channels made from elastomeric material. Aizenberg et al. (1998) have reported patterning SAMs by patterning the topography of their metallic supports. Toby et al. (1999) have used microcontact printing of lipophilic self-assembled monolayers for the attachment of biomimetic lipid layers to surfaces. Ufheil (2000) has generated microstructures of solid-supported lipid layers using a SAM pattern generated by what he terms scanning electrochemical microscopy. Due to the slow kinetics of nickel etching in acidic media, an electrochemical reaction is scribed over a surface, achieving a spatial resolution of 4 microns. These features were visualised by fusing fluorescent liposomes onto the surface. He et al. (2000) used SAMs as a passivation layer to selectively etch micro-patterns using cyclic voltammetry. Hongyou et al. (2000) describe techniques for rapid prototyping and production of nanometre'sized arrays. A useful complement to the ability to create micrarrays is the ability to pattern these arrays with selected functionalities. Figure 27 shows a technique employed in the ICS TM biosensor in which the streptavidin-biotin linkage is blocked unless irradiated by UV. This can be applied through a patterned mask to permit attachment to selected sites. Effenberger et al. (1998) has used optically activated preparation and patterning of SAMs. Groves (1997, 2000) have controlled cell adhesion and growth with micropatterned supported BLM's. Phosphatidylserine is used to specify regions for growth of cells. The same group has also developed a technique for defining barriers to diffusion in tethered BLM' s. The functionalities which may be brought to tethered bilayers is becoming extensive. Motesharei et al. (1997) have fabricated peptide nanotubes within supported SAMs. Gr~itzel (1998) determination of the surface concentration of crown ethers in supported lipid membranes by capacitance measurements. 491

Comell Scheibler et al. (1999) have fabricated topographical templates for chemiselective ligation of antigenic peptides to self-assembled monolayers. Stora et al. (1999) incorporated the Ompf porin channel from E. coli, into a tethered BLM and were able to alter its conduction using the channel blocker colicin. The detection of single ion channel currents have not yet been reported. However, with the reduction of electrode dimensions to the micrometre scale, single-channel noise can be detected from gramicidin in tethered BLMs. Schmidt et al. (2000) have fabricated a chip-based sensor for the functional analysis of single channels. These authors use etched silicon to measure single state currents of alamethicin. Glazier et al. (2000) have reconstituted the pore-forming toxin a-hemolysin in supported bilayers. Naumann (1999) et al. have incorporated active cytochrome C into a tethered membrane as one of the first examples of a functionally active biomimetic surface. A further class of device are becoming apparent that could be seen as spin-offs of self assembling tethered membranes Jirage et al. (1997) have fabricated nanotubes based on microporous polymers coated with gold and lined with SAMs. Lee and Martin (2001), have controlled ion transport in nanoporous membranes coated with SAMs to control solute flux.

7. References

Aizenberg, J., J. Black and G.M. Whitesides, 1998, Nature 394, 868. Avrone, E. and J.P. Rospars, 1995, Biosystems 36, 101. Bain C.D., Evall J. and Whitesides G.M.J., 1989, Amer. Chem. Soc. 111, 7155. Braach-Maksvytis V. and B.J. Raguse, 2000, Amer. Chem. Soc. 122, 9544. Buffer, J., S. Kahlert, S. Tzartos, A. Maelicke, and C. Franke, 1996, J. Physiol.London, 492, 107. Bums, C.J., L.D. Field, K. Hashimoto, B.J. Petteys, D.D. Ridley and M. Rose, 1999, Aust. J. Chem. 52, 387. Bums, C.J., L.D. Field, J. Morgan, D.D. Ridley and V. Vignevich, 1999, Tetrahedron Lett. 40, 6489. Bums, C.J, L.D. Field, K. Hashimoto, B.J. Petteys, D.D. Ridley and S. Sandanayake, 1999, Synthetic Commun. 29, 2337. Bums, C.J., L.D Field, K. Hashimoto, B.J. Petteys, D.D. Ridley and M. Rose, 1999, 52, 387. Bums, C.J., L.D. Field, J. Morgan, D.D. Ridley. And V. Vignevich, 1999, Tetrahedron Lett. 40, 6489. Chun, K. Y. and P. Stroeve, 2001, Langmuir 17, 5271. Clark, T.D., J. Tien, D.C. Duffy, K.E. Paul and G.M. Whitesides, 2001, J. Amer. Chem. Soc. 123, 7677. Cornell, B.A., V.L.B. Braach-Maksvytis, L.G. King, P.D.O. Osman, B. Raguse, L. Wieczorek and R.J. Pace, 1997, Nature 387,580. Cui, Y., Q. Wei, H. Park, and C. M. Lieber, 2001, Science 294, 1289. 492

Membrane-based Biosensors Dante S., M. Derosa, E. Maccioni, A. Morana, C. Nicolini, F. Rustichelli, V.I. Troitsky and B. Yang, 1995, Mol. Crystals Liquid Crystals, 262, 191. Davies, S.T., Khamsehpour, B., 1991, Science 254, 1300. Dishner, M.H., F.J. Feher and J.C. Hemminger, 1996, Chem. Commun., 1971. Effenberger, F., G. Gt~tz, B. B idlingmaier and M. Wezstein, 1998, Chem. Int. Ed., 37, 2462. Erdelen, C., L. Haussling, R. Naumann, H. Ringsdorf, H. Wolf, J. Yang, M. Liley,, J Spinke and W. Knoll, 1994, Langmuir, 10, 1246. Fente, P., A. Eberhardt and P. Eisenberger, 1994, Science 266, 1216. Folker, J.P., P.E. Laibinis, G.M. Whitesides and J. Deutch, 1994, J.Phys. Chem. 98, 563. Fuyuan F. and R.B. Lennox, 2000, Langmuir 16, 6188. Glazier, S.A., D.J. Verah, A.L. Plant, H. Bayley, G. Valincius and J.J. Kasianowicz, 2000, Langmuir 16,10428. Gr~itzel, M., Langmuir 1998, 14, 2573. Grove, J.T., L.K. Mahal and C.R. Bertozzi, 2001, Langmuir 17, 5129. Groves, J.T., N. Ulman and S.G. Boxer, 1997, Science 275, 651. Guo,L.H., J.S. Facci, G. McLendon and R. Mosher, 1994, Langmuir 10, 4588. Gupta, V.K. and N.L. Abbott, 1996, Langmuir 12, 2587. Hardt, S.L., 1979, Biophys. Chem. 16, 239. He, H.X., Q.G. Li, Z.Y. Zhou, H. Zhang, S.F.Y. Li, and Z.F. Liu, 2000, Langmuir 16, 9683. Heysel, S., H. Vogel, M. Sanger and H. Sigrist, 1995, Protein Sci. 4, 2532. Hongyou, F., L. Yunfeng, A. Stump, S.T. Reed, T. Baer, R. Schunk, V. PerezLuna, G.P. L6pez and C.J. Brinker, 2000, Nature 405, 56. Hoogvliet, J.C., M. Dijksma, B. Kamp and W.P. van Bennekom, 2000, Anal. Chem. 72, 20161. Hsueh, C.C., M.T. Lee, M.S. Freund and G.S. Ferguson, 2000, Agew. Chem. Int. Ed. 39, 1228. Jordan, R., 1999, Langmuir 15, 2095. Jirage, K.B., J.C. Hulteen and C.R. Martin, 1997, Science 278, 655. Kane, R.S., S. Takayama, E. Ostuni, D. E. Ingber G.M. Whitesides, 1999, Biomaterials, 20, 2363. Katz, B., and R. Demain, 1977, Bacteriol. Rev. 41,449. Koeppe, R.E. and O.S. Andersen, 1996, Ann. Rev. B iophys. B iomol. Structure 25, 231. Krishna, G., J. Schulte, B.A. Cornell, R. Pace, L. Wieczorek and P.D. Osman, 2001, Langmuir, 17, 4458. Lee, M.T., C.C Hsueh and M.S. Freund, Ferguson, G.S., 1998, Langmuir 14, 6419. Lee, S.B. and C.R. Martin, 2001, Chem. Materials 13, 3222. Ligler, F.S., T.L. Fare, K.D. Seib, J.W. Smuda, A. Singh, P. Ahl, M.E. Ayers, A. Dalziel and E. Sackman, 1996, Science 271, 5245.

493

Comell Ligler, F.S., G.P. Anderson, P.T. Davidson, R.J. Foch, J.T. Ives, K.D. King, G. Page, D.A. Stenger and J.P. Whelan, 1998, Environ. Sci. Technol. 32, 2461. Lingler, S., I. Rubenstein, W. Knoll and A. Offenhausser, 1997, Langmuir 13, 7085. Lopafin, A.N., E.N. Makhina and C.G. Nichols, 1995, J. Gen.Physiol. 106, 923. Lu, X.D., A.L. Ottova and H.T. Tien, 1996, Bioelectrochemistry Bioenergetics 39, 285. Meuse, C.W., S. Krueger, C.F. Majkrzak, J.A. Dura, J. Fu, J.T. Connor and A. Plant, 1998, Biophys. J. 74, 1388. Meuse, C.W., G. Niaura, M.L. Lewis and A.L. Plant, 1998, Langmuir, 14, 1604. Motesharei, K. and M.R. Ghadiri, 1997, J. Amer. Chem. Soc. 119, 11306. Myers, V.B. and D.A. Haydon, 1972, Biochem. Biophys. Acta, 27, 313. Naumann, C.A., W. Knoll and C.W. Frank, 2001, Biomacromolecules 2, 1097. Naumann, C.A., O. Prucker, T.J. Lehmann Rtihe, W. Knoll and C.W. Frank, 2001, Biomacromolecules. Naumann, R., E . K . A . Schmidt Jonczyk, K. Fendler, B. Kadenbach, T. Liebermann, A. Offenhausser and W. Knoll, 1999, B iosens. B ioelectron. 14, 651. Padeste, C., S. Kossek, H.W. Lehmann, C.R. Musil, J. Gobrecht and L.J. Tiefenauer, 1996, Electrochem. Soc. 143, 3890. Philp, D. and J.F. Stoddart, 1996, Angew. Chem. Int. Ed. 35, 1155. Plant, A., 1993, Langmuir 9, 2764. Poirier, G.E. and E.D. Pylan, 1996, Science 272, 1145. Raguse, B, V.L.B. Braach-Maksvytis, B.A. Cornell, L.G. King, P.D.O. Osman, R.J. Pace and L. Wieczorek, 1998, Langmuir 14, 648. Raguse, B., P.N. Culshaw, J. Prashar and K. Raval, 2000, Tetrahedron Lett. 41, 2971. Raguse, B., V.L.B. Braach-Maksvytis, B.A. Cornell, L.G. King, P.D.O. Osman, R.J. Pace and L. Wieczorek, 1998, Langmuir 14, 648. Raguse, B., P.N. Culshaw, J. Prashar and K. Raval, 2000, Tetrahedron Lett. 41, 2971. Reiken, S.R., B.J. Vanwie, H. Sutissna, D.F. Moffett, A.R. Koch, M. Silbert and W.C. Davies, 1996, Biosens. Bioelectron. 11, 91. Ron, H. and I.J. Rubenstein, 1998, Chem. Soc., 120, 13444. Ron, H. and I.J. Rubenstein, 1994a, Langmuir 10, 4566. Ron, H. and I.J. Rubenstein, 1994b, Langmuir 10, 4586. Ron, H., S. Matlis and I. Rubenstein, 1998, Langmuir 14, 1116. Sackmann, E., 1996, Science, 271, 43. Sackmann, E. and M. Tanaka, 2000, Biotechnology, 18, 58. Scheibler, L., P. Dumy, M. Boncheva, K. Leufgen, H.J. Mathieu, M. Mutter, and H. Vogel, 1999, Angew. Chem. Int. Ed., 38, 696. Schmidt, C., Mayer, M., Vogel, H., 2000, Nature, Sept. 3137. Schlenoff, J.B., M. Li and H.J. Ly, 1995, Amer. Chem. Soc. 117, 12528.

494

Membrane-based Biosensors Spinke, J., J. Yang, H. Wolf, M. Liley, H. Ringsdorf and W. Knoll, 1992, Biophys. J. 63, 1667. Steinem, C., A. Janshoff, W.P. Ulrich, M. Sieber and H.J. Galla, 1996, Biochim. Biophys. Acta. 1279, 169. Stelzle, M., G. Weissmuller and E. Sackmann, 1993, J. Phys. Chem. 97, 2974. Stetter, K.O., 1996, FEMS Microbiol. Rev. 18, 149. Stora, T., J.H. Lakey and H. Vogel, 1999, Angew. Chem. Int. Ed. 38,389. Suarez, E., E.D. Emmanuelle, G. Molle, R. Lazaro and P. Viallefont, 1998, J. Peptide Sci. 4, 371. Toby, A., A. Jenkins, N. Boden, R.J. Bushby, S.D. Evans, P.F. Knowles, R.E. Miles, S.D. Ogier, H. Sch/3nherr and G.J. V ancso 1999, J. Amer. Chem. Soc. 121, 5274. Tsai, M.Y. and J.C. Lin, 2001, J. Colloid Interface Sci. 238, 259. Ufheil, J., F.M. Boldt, M. BOrsch, K. Borgwarth and J. Heinze, 2000, Bioelectrochemistry, 52, 103. Ulman, A., Chem. Rev., 1996, 96, 1533 Walczak, M.M., C.A. Alves, B.D. Lamp and M.D. Porter, 1995, J. Electroanal. Chem. 396, 103. Wallace, B.A., 1998, J. Structural Biol. 121,123. Yager, P., 1988a, Adv. Exp. Med. Biol. 238, 257. Yager, P., 1988b, Med. Instrumen. 22, 247. Zaitsev, S.I., S.V. Dzekhitser and V.P. Zubov, 1998, Bioorg. Khim. 1496, 850. Zhang, Y., R.H. Terrill, T.A. Tanzer and P.W. Bohn, 1998, J. Amer. Chem. Soc., 120, 2654.

495

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 16

PEBBLE NANOSENSORS FOR REAL TIME INTRACELLULAR CHEMICAL IMAGING +MURPHYBRASUEL,*+RAOULKOPELMAN,PH.D., *MARTINA. PHILBERT,PH.D., +JONATHANW. AYLOTT,PH.D., +HEATHERCLARK, PH.D., +ILANAKASMAN,M.S., +MATTHEWKING,+ERICMONSON, PH.D., *JAMESSUMNER,+HAOXu, *MARIONHOYER,PH.D., *TERRYJ. MILLER,M.S. AND*RONTJALKENS,PH.D. *Department of Chemistry *Department of Environmental Health Sciences The University of Michigan, Ann Arbor, Michigan 48109-1055 USA

PEBBLE nanosensors (Probes Encapsulated By Biologically Localized Embedding) are submicron-sized optical sensors designed specifically for minimally invasive analyte monitoring in viable, single cells with applications for real time analysis of drug, toxin, and environmental effects on cell function. PEBBLE nanosensors is a general term that describes a family of matrices and nano-fabrication techniques used to miniaturize many existing optical sensing technologies. The main classes of PEBBLE nano-sensors are based on matrices of cross-linked polyacrylamide, cross-linked decyl methacrylate, and sol-gel silica. These matrices have been used to fabricate sensors for H § Ca e§ IC, Na+, Mg z§ Zn e+, Cl-, NOe, Oe, NO, and glucose that range from 20 nm to 600 nm in size. A host of delivery techniques have been used successfully to deliver PEBBLE nanosensors into mouse oocytes, rat alveolar macrophages, rat C6-glioma, and human neuroblastoma cells.

1. TechnicalConcept PEBBLEs (Probes Encapsulated By Biologically Localized Embedding) are spherical nanosensors that entrap within a matrix optical sensing components for the real time monitoring of ions and small molecules in biological systems. 497

Brasuel, Kopelman, Philbert, et al.

Figure 1. Two potassium-selective fiber optic sensors inserted in a 100 lain mouse oocyte. PEBBLEs were developed specifically for biological applications and fill a role that lies between pulled optical fiber sensors and free molecular probes (naked indicator dye molecules). The PEBBLE concept hinges on two related and distinct roles. First and foremost, the PEBBLE protects the cell from the toxicity inherent in some of the "molecular dyes" and at the same time protects the indicator dye from cellular interferents, such as protein binding. The second role, which is possible because the PEBBLE creates a separate sensing phase (the PEBBLE matrix), distinct from the cellular environment, is that multiple dyes, ionophores, and other components can be combined to create complex sensing schemes. These complex sensing schemes can include reference dyes to allow ratiometric imaging, or ionophore/chromoionophore combinations that allow for the use of highly selective, non-fluorescent ionophores. Both the protection (of both dye and cell) and the powerful sensing flexibility come in a nano-package, which, in terms of minimal mechanical and physical perturbation, is closer to "free molecular dyes" than most other sensing platforms! However, the nanosensor preserves the excellent chemical sensing and biocompatibility of macro-sensors and surpasses their performance in terms of response time and absolute detection limit.

1.1. Mechanical, physical, and chemical perturbation Common methods for single cell analysis include pulled optical fibers (micro and nano optodes) and electrodes. While the tips of these sensors may have nanometer dimensions, the mechanical and physical perturbation of the cell is perpetrated by the necessity of punching holes in cellular membrane and by the entire penetration volume of the sensor, taking up a significant percentage of intracellular space. The damage due to sensor perturbation has the potential of 498

PEBBLE Nanosensors having as much effect on the cell as the stimulus under study. Take for example the mouse oocyte (Figure 1), where the tips of the inserted fiber optic sensors can be as small as 200 nm, but because of the penetration depth of 50 jxm (this coneshaped volume grows as the third power of the penetration depth) the penetration volume is about 20,000 ~tm 3. The cell volume of a 100 ~xm diameter oocyte is about 524,000 ktm 3, thus the penetration volume is more than 1% (nearly 4%) of the cell volume and most mammalian cells are much smaller! It is not hard to see that monitoring more than one analyte in a single cell, using pulled nano-optodes or pulled capillary microelectrodes, and still maintaining cell viability would be a significant challenge. Though we have found pulled fiber optic nano-optodes useful in the monitoring of single analytes in live cells, PEBBLEs were developed to use the same sensing technologies while significantly reducing the mechanical perturbation of the cells. Take, for example, decyl methacrylate-based PEBBLEs, which at the high end are 600 nm in diameter; their total volume of 0.2 gm 3 is only 1 ppm of the oocyte cell volume. In direct comparison, we could use 22,000 PEBBLEs for the same penetration volume of one pulled nano-optode. In reality we use much less than this. Furthermore the 20-40 nm acrylamide PEBBLEs have a physical volume perturbation of only l ppb or less! In terms of chemical perturbation, the first concern is that the "probe" itself may be toxic to the cell or to specific cellular organelles. Injection of fluorescent indicator "naked" dyes into a cell, combined with ratiometric imaging, confocal microscopy, two-photon fluorescence, or fluorescence lifetime imaging, has provided insights into concentrations and spatial locations of ions throughout a single cell. One challenge in using these optical methods is that each dye for use in intracellular measurements must be chosen carefully and assessed for its ability to enter a cell and then provide accurate and reliable information from within the cell. Problems such as toxicity to the cell, intracellular sequestration into specific organelles, non-ratiometric properties, protein binding and intracellular buffering must be evaluated (Graber et al., 1986; CohenKashi et al., 1997; Overly et al., 1995; Morelle et al., 1994). While some indicator dyes work well within a cell, many fluorescent probes suffer from the above problems, thus limiting the indicator dyes available for reliable intracellular measurements. For example, common calcium probes can buffer a cell when used in high concentrations (Ross, 1993), and pH indicators such as carboxyfluorescein and BCECF (a polar fluorescein derivative) can be affected by self-quenching or by binding to proteins. By entrapping the indicator dyes within a polymer matrix, PEBBLEs provide a method for minimizing many of the undesired interactions that occur between fluorescent probes and the cell. The matrix allows ions or neutral analyte species to diffuse through and bind with the indicator, but prevents mobility of the indicator dyes into and throughout the cell, thus avoiding sequestration and self499

Brasuel, Kopelman, Philbert, et al. Table 1. Simple ionophore-based approximation for the number of 600 nm K§ and Na§ PEBBLEs and 60nm Ca2§ PEBBLEs that can be put in a single mouse oocyte while perturbing the ion concentration by 1% or less. Analyte K* '

-" #'ionophores PEBBLE 1:36• 106[20mM/l(g]

LNa + Ca 2+

1.36• 3. [0.042mM/kg] '

ionsin ' ~ %perturbation cell PEBBLE 2.26x1013 6.03x1"0-~ 2.42• ~2 "5.62X10"5

8.07x10 ~

.

3.71x10 -7 .

.

.

.

.

.

.

i

'#PEBBLES cell 1.66x105 i

i

i-,

i.78xiO4 .

2.70x106

quenching. In short, the matrix protects the cell from the dyes and the dyes from the cell components. A less obvious consideration is that the sensing of analyte molecules requires binding of the analyte of interest to a selective "sensing" molecule. This is true for both "naked" dyes and any equilibrium-based sensor, including PEBBLEs. In effect, these probes (PEBBLEs and "naked" dyes) act as chelators of cellular components. Most of our PEBBLE sensors rely on bulk equilibrium between sample volume and the sensor. In confined volumes, the concern is that the sensor will possibly change the analyte-activity by removing analyte from the sensing environment (because sensor response is dependent on bulk equilibrium). For the purpose of this explanation, a perturbation of the intracellular environment by 1% will be considered as the maximum allowable perturbation. As a simple limiting case, one can assume that the K value of the extraction of analyte into the polymer membrane of the PEBBLE is very large. In this limiting case, every ionophore has extracted an analyte ion. Table 1 shows the results of calculations regarding the extent of perturbation that a single 600 nm liquid polymer (decyl methacrylate) PEBBLE would have on the analyte of interest, based on the ionophore concentration in each type of sensor (K § and Na+). The table also includes the number of PEBBLE sensors that would be required to perturb the system by 1% (assuming a cell with 40 ~m radius). The table also shows the data for a 60 nm diameter acrylamide-based Ca 2§ PEBBLE (same cell size, note that Ca 2§ concentration is very low). From Table 1, it is immediately evident that the small size of the PEBBLE allows us to a high cellular loading of PEBBLEs, without excessive chemical perturbation of the cellular environment. It is possible to go to cells much smaller than the mouse oocyte and still have enough PEBBLEs present to get a sizable signal without mechanical or chemical perturbation of the cells. The ability to measure changes in concentration of analytes in the confined volumes

500

PEBBLE Nanosensors

Figure 2. Schematic representation of PEBBLE delivery methods. A) Gene gun delivery, B) Picoinjection, C) Liposomal delivery and D)Phagocytosis.

of intracellular space, with little perturbation of the cellular space, results in an absolute detection limit of zeptomoles, or less, of analyte. We describe below applications of three families of PEBBLEs based on three kinds of matrices: hydrophilic, hydrophobic and amphiphilic. They are made of polyacrylamide, cross-linked decyl methacrylate, and sol-gel, respectively, and each has its application niche.

1.2. PEBBLE delivery One of the most important considerations when applying PEBBLE nano-sensors to single cell studies is the delivery of the PEBBLEs to the cell. The many methods that have been explored include gene gun, picoinjection, liposomal delivery, and sequestration into macrophages. The method of PEBBLE delivery by gene gun can best be thought of as a shotgun method (Figure 2A). PEBBLEs are dried on a plastic (delivery) disk. This disk is set in front of a rupture disk. Helium pressure is then built up behind the rupture disk, which ruptures at a specific helium pressure and propels the PEBBLEs from the plastic disk into a cell culture in shotgun fashion. The gene gun can be used to deliver one to thousands of PEBBLEs per cell into a large number of cells in a very short amount of time (dependent on the concentration of PEBBLEs on the delivery disk) (Clark et al., 1998, 1999a; Brasuel et al., 2001; 501

Brasuel, Kopelman, Philbert, et al. Xu et al., 2001). Cell viability is excellent, 98% viability compared to control cells (Clark et al., 1999a), for small numbers of PEBBLEs, and hinges directly on the number of PEBBLEs delivered, the delivery pressure, and the chamber vacuum. The PEBBLE momentum determines whether the PEBBLEs are mainly internalized in the cytoplasm or in the nucleus. Picoinjection is used to inject pL volumes of PEBBLE-containing solution into single cells (Figure 2B). This method of delivery is dependent on the fabrication of pulled capillary "needles", through the use of a pipette-puller and a microforge. The smallest volume deliverable is 10 pL, and the most concentrated PEBBLE solution to work in the pulled capillary syringe is 5 mg/ml PEBBLEs. Assuming a PEBBLE mass of 4.2x10~Sg, the injection of 10 pL of 4.19• -4 mg/rnl solution should give one PEBBLE in the cell. If the maximum of 5 mg/ml is used, 1,200 PEBBLEs/pL are injected, that is, 12,000 PEBBLEs in the 10 pL minimum injection. The maximum number of PEBBLEs one can put in is dependent on the volume of solution that can be injected without damaging the cell. Picoinjection can give a wide range of PEBBLE concentrations in the cell, but because each cell must be individually injected the method is time consuming and tedious. Cell viability is good (Clark et al., 1999a). Commercially available liposomes can be used to deliver PEBBLEs to cells (Figure 2C). In order to encapsulate the PEBBLEs, the liposomes are prepared in a solution of PEBBLEs. The liposomes are then placed in the cell culture and the liposomes fuse with the cell membranes and empty their contents (the PEBBLEcontaining solution) into the cell. Three factors play a key role in determining the number of PEBBLEs delivered to each cell with this method: the original concentration of the PEBBLEs, the concentration of liposomes placed in the cell culture, and the length of time the liposomes are left with the cells (Clark et al., 1999a, 1999b, 1999c). The parameters must be tailored for each cell line used in order to obtain the desired concentration of PEBBLEs in the cells. While it would be difficult to deliver a single PEBBLE to each cell with this method, it does seem that a low end of between 10-50 PEBBLEs per cell would be possible, with the high end being the maximum number of PEBBLEs the cell could take without losing viability. Liposomal delivery is useful for delivering low to high concentrations of PEBBLEs to many cells simultaneously. The challenge is in tailoring the delivery for the concentrations desirable for the cell line being used. Cell viability is excellent. Macrophages, a specialized immune system cell, take up PEBBLEs automatically (Figure 2D). The number of PEBBLEs that each macrophage takes up is dependent on the concentration of the PEBBLE solution and the amount of time the macrophages are incubated with the PEBBLEs. The advantage of this delivery method is that one can easily deliver varying concentrations of PEBBLEs to rnacrophages. The disadvantage is that it is only

502

PEBBLE Nanosensors

Figure 3. A rat alveolar macrophage with calcium PEBBLE sensors (60x). A) Nomarski illumination and B) Fluorescence illumination. Graph. The increasing intracellular calcium, indicated by increasing-intensity of calcium PEBBLEs in alveolar macrophage following stimulation with 30 #g/ml concanavalin A.

useful for macrophages (which are hard to culture). This method also provides excellent cell viability (Clark et al., 1999a). 1.3. Chemical imaging in live cells The ability of PEBBLE sensors to measure intracellular analytes has been demonstrated in mouse oocytes and rat alveolar macrophages, as well as in neuroblastoma, myometrial and glioma cells (Clark et al., 1998, 1999a, 1999b, 1999c, Brasuel et al., 2001; Xu et al., 2001). All currently applied PEBBLE technologies have relied on fluorescence emission ratios for signal transduction. Cell imaging was accomplished with either an Olympus FluoView 300 scanning confocal microscope system equipped with an Ar-Kr and He-Ne (or similar system) or an Olympus inverted fluorescence microscope, IMT-II (Lake Success, NY), using Nikon 50mm f/1.8 camera lenses to project the image into an Acton 150mm spectrograph (Acton, MA) with spectra read on a Princeton Instruments, liquid nitrogen-cooled, 1024 X 256 CCD array (Trenton, NJ). The Olympus fluorescent microscope (or similar system with an upright microscope) was used to obtain PEBBLE spectra. Following are examples of real time cell imaging, using standard fluorescent microscopy techniques. The first PEBBLEs produced were acrylamide-based, and one of the first examples of their successful application to cells was with macrophages. Alveolar 503

Brasuel, Kopelman, Philbert, et al. macrophages were obtained from rats lung lavage with Krebs-Henseleit buffer. Macrophages were maintained in a 5% CO2, 37~ incubator in Dulbecco's Modified Eagle Medium (DMEM) containing 10% fetal bovine serum and 0.3% penicillin, streptomycin and neomycin. PEBBLE suspensions ranging from 0.31.0 mg/ml were prepared in DMEM and incubated with alveolar macrophage overnight. Macrophage images were then taken on a confocal microscope and spectra of the same cells were obtained on the fluorescent microscope. Acrylamide PEBBLEs selective for calcium (containing Calcium Crimson) were used in order to monitor calcium in phagosomes within rat alveolar macrophage, because of the ease in which macrophage phagocytose particles. This method for delivering the PEBBLEs into cells provided a simple, yet important, test of the PEBBLE sensors in a challenging (acidic) intracellular environment. Macrophage that had phagocytosed 20 nm calcium-selective PEBBLE sensors (Figure 3) were challenged with a mitogen, Concanavalin A (Con A), inducing a slow increase in intracellular calcium, which was monitored over a period of 20 minutes (Figure 3). PEBBLE clusters confined to the phagosome enabled correlation of ionic fluxes with stimulation of this organelle. The calcium PEBBLE in the macrophage experiment clearly demonstrates a time resolved observation of a biological phenomenon within a single, viable cell. One can clearly obtain relevant time domain data with a fluorescence microscope, spectrograph and CCD. With a confocal microscope system and the appropriate dye/filter sets, one can attain both temporal and spatial resolution, as demonstrated below. Calcium PEBBLEs have been used, containing Calcium Green-1 (Molecular Probes) dye, in combination with sulforhodamine dye, as sensing components. Calcium Green fluorescence increases in intensity with increasing calcium concentrations while the sulforhodamine fluorescence intensity remains unchanged, regardless of biologically relevant concentration of ions, pH, or other cellular component. Thus, the ratio of the Calcium Green/sulforhodamine intensity gives a good indication of cellular calcium levels regardless of dye, PEBBLE concentration, or fluctuations of light source intensity. Color Plate 1 (end of chapter) shows a confocal microscope image of human C6 glioma cells containing calcium green/sulforhodamine PEBBLEs. The PEBBLEs were delivered by liposomes to the cytoplasm of the cells. In the image, the sulforhodamine fluorescence shows up red (reference peak) and calcium green (green fluorescence) increases with increasing calcium intensity (both dyes confined in PEBBLEs). The toxin, m-dinitrobenzene (DNB), was introduced to the left side of the image and allowed to diffuse to the right. The effect of DNB is the disruption of mitochondrial function, followed by the uncontrolled release of calcium associated with onset of the mitochondrial permeability transition (MPT) (Clark et al., 1999c). Calcium PEBBLEs were used to determine that the half-maximal rate of calcium release (ECs0) occurred at a 10-fold lower

504

PEBBLE Nanosensors concentration of m-DNB in human SY5Y neuroblastoma cells than in human C6 glioma cells (Clark et al., 1999c). The above acrylamide PEBBLE matrix has proven to work with any hydrophilic sensing components. However, it is not able to take advantage of the rich history of electrochemical sensors which has identified a host of highly selective, hydrophobic, ionophores. In many cases, the selectivity of these ionophores has yet to be matched by hydrophilic dyes (hydrophilic chromoionophores). Highly selective intracellular (and extracellular) hydrophilic indicator dyes are limited to a small set of analytes, such as pH and calcium. While the use of PEBBLEs instead of traditional free "naked" indicators results in protection beneficial to both the cell and the dye, it does not solve the selectivity problems. For instance, hydrophilic potassium indicators will not work in the presence of significantly higher sodium concentrations, and, conversely, sodium indicators will not work in the presence of high potassium concentrations (Haugland, 1993). Obviously, this has serious implications for both intracellular (e.g., high potassium/sodium ion ratios) and extracellular (e.g., high sodium/potassium) applications. Moreover, for many important analyte ions, such as nitrite, no satisfactory color indicators are available. The above problem has beensolved in optodes by using in tandem an optically silent ionophore (which is highly selective) and an adjacent, optically visible agent that plays the role of a spectator or reporter dye. While the principles of such tandem sensing schemes were worked out by Bakker and coworkers (Bakker and Simon, 1992; Morf et al., 1989; Buhlmann et al., 1998), Suzuki (Kurihara et al., 1999; Suzuki et al., 1990), and Wolfbeis (Mohr et al., 1997a, 1997b), the first demonstration of such a sensing scheme on the nanoscale occurred with the pulled optodes developed by Shortreed et al. (1996, 1997). The extension of these principles to PEBBLEs required the optimization of a new liquid polymer matrix, decyl methacrylate (Brasuel et al., 2001). The result was a 600 nm PEBBLE that is selective for potassium, based on the potassium ionophore 2-dodecyl-2-methyl-l,3-propanediylbis[N-[5'nitro(benzo15-crown-5)-4'-yl]carbamate] (BME-44) and chromoionophore IT/ or 9(diethylamino)-5-[(2-octoyldecyl)imino]benzo[a]phenoxazine (ETH 5350). The first application of the liquid polymer class of PEBBLEs was the observation of potassium uptake in rat C6-glioma cells (Brasuel et al., 2001). Micron-sized liquid polymer beads based on either polyvinyl chloride (PVC) or dodecyl acrylate have also been developed (Tsagkatakis et al., 2001; Peper et al., 2001).

505

Brasuel, Kopelman, Philbert, et al.

Figure 4. Ratio data of decyl methacrylate K§ PEBBLEs in C6-glioma cells during the addition of kainic acid (50 I.tl of 0.4 mg/ml) at 20 s and at 60 s. Ratios were converted to log (aK§ § using solution calibration of the PEBBLEs. Log (aK§ +) is seen to increase after kainic acid addition [and subsequent K§ channel openings]. Inset, Confocal image of decyl methacrylate K§ PEBBLE fluorescence, overlaid with Nomarski image of rat C6-glioma cells. 488 nm excitation, 580 nm long pass filter.

Decyl methacrylate PEBBLEs were delivered into rat C6-glioma cells, using a BioRad (Hercules, CA) Biolistic PDS-1000/He gene gun system, with a firing pressure of 650 psi, and a vacuum of 15 Torr applied to the system. Lrnmediately following PEBBLE delivery, cells were placed on an inverted fluorescent microscope. The gating software for the CCD was set to take continuous spectra at 1.3 s intervals. After 20 s and 60 s, 50 ~tl of 0.4 mg/ml kainic acid was injected into the microscope cell. Kainic acid is known to stimulate cells by causing the opening of ion channels. Confocal microscopy was used to determine the localization of the PEBBLE sensors after gene gun delivery (Brasuel et al., 2001). Figure 4 (inset) shows the confocal fluorescent image of the PEBBLEs, overlaid with a Nomarski differential interference contrast image of the cells. The image indicates that the PEBBLE sensors are localized in the cytoplasm of the glioma cells. Figure 4 shows the PEBBLE sensors inside the cells responding to kainic acid addition to the cell medium, after 20 and 60 seconds. One can see § increases, indicating either an increase in K § concentration or that log (aK§ 506

PEBBLE Nanosensors a decrease in H § concentration (increase in pH). The amount of kainic acid added is not known to affect the pH of cells in culture and kainic acid by itself has no effect on the sensors. Thus the change is likely due to increasing intracellular concentration of K § the expected trend. The membrane of C6glioma cells can initiate an inward rectifying K § current, induced by specific K § channels, a documented role in the control of extracellular potassium (Emmi et al., 2000). Thus, when stimulated with a channel opening agonist, the K § concentration within the glioma cells is indeed expected to increase. Sol-gel, the newest PEBBLE matrix, gives the flexibility of being able to tailor the properties of the matrix to accept either hydrophilic or hydrophobic dyes (For oxygen, their dynamic range is much wider than that of similar acrylamide PEBBLEs) (Xu et al., 2001). It is also proven as a matrix compatible with the use of protein-based sensors (Uhlmann et al., 1997). U s i n g the gene gun, sol-gel PEBBLEs were inserted, under conditions similar to the decyl methacrylate PEBBLEs delivery protocol, into rat C6-glioma cells, in order to monitor oxygen. A ratiometric sol-gel PEBBLE sensor ([Ru(dpp)3] 2§ oxygen-sensitive dye and Oregon Green 488-dextran reference dye) was used. Color Plate 2 (end of chapter) shows the confocal images of C6 glioma cells containing sol-gel PEBBLEs under Nomarski illumination overlaid with: (a) The green fluorescence of Oregon Green 488-dextran and (b) the red fluorescence of [Ru(dpp)3] 2§ It can be seen that the cells still maintained their morphology after the gene gun injection of PEBBLEs and showed no sign for cell death. The dyes were excited, respectively, by reflecting the 488 nm Ar-Kr and the 543 nm HeNe laser lines onto the specimen, using a double dichroic mirror. The Oregon Green fluorescence from the PEBBLEs inside the cells (Panel A) was detected by passage through a 510 nm long-pass and a 530 nm short-pass filter, and the fluorescence of [Ru(dpp)3] 2§ (Panel B) through a 605 nm (45 nm band-pass) barrier filter. A 40X, 1.4 NA oil immersion objective was used to image the Oregon Green and [Ru(dpp)3] 2§ fluorescence. The distribution of PEBBLEs shown in the overlaid images demonstrates that the green and red fluorescence in Color Plate 2, both panels, were truly from PEBBLEs inside cells. It should be noted that most of the PEBBLEs were loaded into the cytoplasm, but there were also some in the nucleus. The graph in Color Plate 2 shows the response of oxygen sensitive sol-gel PEBBLEs, inserted inside rat C6 glioma cells, to changing intracellular oxygen concentrations. After gene gun injection, the cells were immersed in DPBS (Dulbecco's Phosphate Buffered Saline) and a spectrum was taken of these cells, using 480 + 10 nm excitation light. The air-saturated DPBS was then replaced by nitrogen-saturated DPBS, to cause a decrease in the intracellular oxygen concentration, and the response of the oxygen PEBBLE sensors inside the cells was monitored during a time period of 2 minutes. As can be seen, the fluorescence intensity of [Ru(dpp)3] 2§ went up successively as the oxygen level inside the cells decreased. Average intracellular oxygen concentrations were 507

Brasuel, Kopelman, Philbert, et al. Table 2: Real time measurements of intracellular oxygen. Average intracellular oxygen concentrations (ppm) Cells in air-saturated buffer 7.9_+2.1 Ceils in N2-saturated buffer (after 25 6.5+_ 1.7

sec) Cells in N2-saturated buffer (after 120

< 1.5

sec) i

Air-saturated buffer solution

8.8+0.8

determined on the basis of a Stern-Volmer calibration curve, obtained using the fluorescence microscope-Acton spectrometer system, and are summarized in Table 2. The comparatively large errors are due to the low resolution of the spectrometer. We note that our measured intracellular oxygen value (when ceils were in air-saturated DPBS) is comparable with the value of-7.1 ppm measured inside the much larger islets of Langerhans (Jung, 1999). These results show that the PEBBLE sensors are responsive when loaded into ceils and that they retain their spectral characteristics, enabling a ratiometric measurement to be made (Xu et al., 2001).

2. History PEBBLEs are a new technology and do not yet to have much of a history. As a brief introduction to the development of the PEBBLE concept, PEBBLEs are a direct outgrowth of the pulled fiber optical technology developed for biosensing by Tan et al. (1992, 1999) and continued through the work of Rosenzweig (Rosenzweig and Kopelman, 1995, 1996), Shortreed (Shortreed et al., 1996, 1997), and Barker (Barker et al., 1997, 1998). The distinct advantages of having nano-scale dimension sensors were formalized in paper by Dourado and Kopelman (1996). Summarized, below are the advantages of sensing with nanoscale sensors. In most instances, there is an explicit functional dependence of specific optode characteristics on the optode radius (r). For instance, the absolute detection limit decreases with r 3 ( g o o d ! ) , the response time reduces with r 2 (good!), while the signal-to-noise ratio decreases with r (bad!) but n o t r 3 (luckily!) under standard working conditions. Other features that improve with downsizing include sample volume, sensitivity, invasiveness, spatial resolution, enzyme activity, dissipation of heat in sensor and/or sample, non-toxicity and materials cost. Features that may worsen include fluorophore leaching and photo-damage to sensor and/or sample. Methods for overcoming these disadvantages make use of the shorter response times of small optodes and forward optical signal collection with small samples (utilizing standard lab microscopes). Eventually it was determined that cell viability would continue to limit experimentation due to the necessity of having optical fiber mechanically

508

PEBBLE Nanosensors displacing cell membrane during the duration of intracellular studies. This meant that the smallest vibrations in the optical set up would lead to leakage around the optical fiber into or out of cell. It was because of this that Clark, Kopelman, and coworkers developed the "fiberless" nano-optodes, which became know as PEBBLEs (Probes Encapsulated By Biologically Localized Embedding). The initial PEBBLEs were based on the poly-acrylarnide, and the rest of this manuscript describes the impetus of designing additional PEBBLE matrices (Clark et al., 1998). While PEBBLEs were a direct extension of pulled fiber optodes, the use of nano and micro spheres in sensing platforms is not a uniquely PEBBLE phenomenon. The use of microspheres for analytical sensing had already proven valuable in the development of other sensing technologies. Walt and co-workers have developed miniaturized sensor arrays of silica microspheres on the end of an imaging optical fiber (Albert and Walt, 2000). Each silica microsphere (3-5 ktm diameter) is an individually addressable sensing unit at the end of an imaging optical fiber (750 ~tm diameter). The entire sensing package is used for vapor detection with the possibility of using pattern recognition analysis with arrays of many different kinds of sensors and has the added advantage of being able to use the simultaneously obtained data from several similar sensors to improve S/N ratios. Seitz has developed polymer microspheres that swell or shrink depending on analyte concentrations (Seitz et al., 1999). Entrapped in a hydrogel, the properties of these microspheres allow the use of numerous unique sensing schemes. Both the works of Walt and Seitz use microsensors as a part of an ensemble sensing scheme. In the case of PEBBLEs, the nano-sensors stand alone as a complete sensing packages. Thus, PEBBLEs are "free floating" polymer microspheres. The first demonstration of "free-floating" single sensing spheres for analyte detection was a pH optochemical sensor demonstrated by Sasaki and Kopelman in 1996, based on fluorescein entrapped in a polyacrylamide nanoparticle (Sasaki et al., 1996). The first published reports on PEBBLEs appeared in 1998 (Clark et al., 1998). Other recent developments in sensing that are similar to PEBBLEs include the sensing particles developed by Rosenzweig and the liquid polymer sensing particles developed by Bakker. McNamara and Rosenzweig (1998) demonstrated the use of liposomes (unilamellar phospholipid vesicles) for sensing agent encapsulation. A more robust version of the liposome sensor was developed by coating polystyrene nanoparticles with fluorescent optochernical phospholipids (1.6 ~tm diameter) for pH detection in murine macrophages (McNamara et al., 2001). Bakker et al. (2001) have published papers on two types of individual liquid polymer microsensors. One version is fabricated through the solvent casting of PVC in aqueous solution to create 1 ktm diameter PVC sensing beads for potassium detection (Tsagkatakis et al., 2001). The second liquid polymer microsensors were prepared by photoinitiated, dispersion polymerization of dodecyl acrylate (DDA) and the cross-linker 509

Brasuel, Kopelman, Philbert, et al. hexanedioldiacrylate (HDDA) to create "plasticizer free" ion-selective sensors (1-10~tm in diameter) (Peper et al., 2001).

3. State of the Art

PEBBLEs are a recent advance in nano-optode technology. Production relies on emulsion and dispersion fabrication techniques. The nano-emulsion/dispersion process for preparing PEBBLEs is subtle, and there is no universal method for making hydrophilic, hydrophobic, and amphiphilic nanospheres that contain the correct matrix and chemical components, in their proper proportions. Thus, switching from single dye containing hydrophilic polyacrylamide nanospheres to multi-component, hydrophobic, liquid polymer sensors or to inert glass sol-gel sensors is not yet a routine procedure. Specific methods for producing sensors from all these matrices are described below as well as the related response mechanisms for each type of sensor. 3.1. Acrylamide PEBBLEs

3.1.1. Acrylamide PEBBLE production. The production of acrylamide PEBBLEs is based on the nano-emulsion techniques studied by Daubresse (Daubresse et al., 1994), which can reliably produce bimodal distributions of nanosized particles. Some control over particle size and shape can be gained by adjusting the surfactant levels. The polymerization solution consists of 0.4 mM fluorescent ionophore (any hydrophilic dye selective for the analyte of interest), 27% acrylamide (polymer), 3% N,N-methylenebis(acrylamide) (cross-linker), in 0.1 M phosphate buffer, pH 6.5. One mL of this solution is then added to a solution containing 20 mL hexane, 1.8 mmol dioctyl sulfosuccinate (surfactant; sodium salt) and 4.24 mmol Brij 30 (surfactant). The solution is stirred under nitrogen for 20 min, while cooling in an ice bath. The polymerization is initiated with 24/zL of a 10% ammonium persulfate solution and 12 ~tL TEMED, then the solution is allowed to stir at room temperature for 2 hours. Hexane is removed by rotary evaporation, then the probes are rinsed of surfactant with ethanol, to give a product consisting of 20 and 200 nm probes (Clark et al., 1998, 1999b). Figure 5 shows both SEM images and results of dynamic light scattering measurements of particles produced by the nano-emulsion process. Figure 5 demonstrates the ability to deliver by gene gun, both 200 nm and 20 nm sized PEBBLEs. If the application requires PEBBLEs of a particular size, the PEBBLEs can be separated by filtration through various sizes of porous polymer.

510

PEBBLE Nanosensors

Figure 5. (A-B) Transmission electron micrographs of acrylamide PEBBLE sensors embedded biolistically (via gene gun), at 900 psi, into the cytoplasm of neuroblastoma cells: (A) Two 200 nm PEBBLEs, near or inside the cell nucleus (B) One 20 nm PEBBLE next to a 1~ lysosome in the cell cytoplasm. Original magnification is indicated on the figure and the inset. (C) Plot of differential mass fraction of a different batch of PEBBLEs (dark line) vs. hydrodynamic radius. Size distribution was determined by Asymmetric Field Flow Fractionation followed by multi-angle static light scattering measurements on the separation channel output. The distribution shows that over 85% of the particles (by mass) are below 60- nm in diameter. The light line shows a test sample of 50 nm and 150 nm diameter latex particles (Duke Scientific) in a 2/1 mass ratio for comparison.

3.1.2. Acrylamide PEBBLE calibration. The protection of sensing elements from interference due to protein binding or cellular sequestration allows the calibration of the PEBBLE sensors in solution, with confidence that this same calibration will be valid during intracellular applications. It is important to note that the interference due to protein binding or large molecule/heavy metal interactions are eliminated [but this does not improve the selectivity]. Most acrylamide PEBBLE calibrations are performed on a Fluoro-Max 2 spectrofluorometer (ISA Jobin Yvon-Spex, Edison, NJ, USA) with the excitation and emission slits set at 2 nm bandwidth. The polyacrylamide polymer PEBBLEs are hydrophilic and water can diffuse freely through them. In these PEBBLEs a dye that has a chromometric response to the analyte is entrapped in the pores of the hydro-gel. Extraction of analyte ions into the hydro-gel is not a consideration because water and small ions are allowed to diffuse freely through the hydro-gel. What does occur is the 511

Brasuel, Kopelman, Philbert, et al. formation of a chromoionophore-analyte complex, similar to the response of the "naked" dye in solution. The dynamic range and selectivity of the PEBBLE is dependent on the KD of the dye with respect to the analyte and any interfering ions. The system shown below represents a calcium sensor, a common ion to sense with these acrylamide PEBBLEs as a result of the large number of water soluble calcium dyes available. Below is the response mechanism and a description of KD.

[Ca2+](aq) + C(acrylamide) ~ K o

" [C(Ca)2+](acrylamide)

_ [C](acyrlamide) [ Ca2+ ](aq)

_

(1)

(2)

[ C ( Ca ) 2+ ]
There is a wide range of available fluorescent dyes with good selectivity for pH and calcium. Thus calcium and pH sensing presented a good opportunity to compare PEBBLE response to that of "naked" dye in similar calibration environments. Complete details on these comparisons can be found (Clark et al., 1999b, 1999c). Tables 3 and 4 summarize the results for pH-sensitive dyes and calcium dyes, respectively. Both dye solutions and PEBBLEs utilized the inert sulforhodamine 101 dye as an internal standard. It can be seen that, with few exceptions, in PEBBLEs the slope of the calibration decreases, resulting in reduced sensitivity of the measurements, but the linear ranges are not significantly affected by incorporating the dye in the PEBBLE. The acrylamide matrix does not adversely affect the reversibility of the indicator dye and can be tailored to entrap the dye for periods long enough to do sensing in live cells. Figure 6A illustrates the reversibility of pH PEBBLEs containing the fluorescent pH probe CDMF (5-(and -6-)carboxy-4',5'-dimethylfluorescein and the reference dye sulforhodamine 101 (SR). The pH of the solution was changed between pH 6.4 and pH 7.0 for several cycles. The CDMF emission maximum was ratioed against the SR emission maximum and plotted, clearly demonstrating full reversibility of the sensors. Assays of the leaching of CDMF and calcium green PEBBLEs (each containing the sulforhodamine 101 reference dye) show that less than 50% of the dye leaches from the PEBBLEs in a 48 hr time period. On the time scale of the current single cell experiments, mostly a couple of hours, the dye loss is acceptable. With comparable leaching rates for indicator and reference dye, the problem is minimized due to the ratiometric sensing scheme used with these PEBBLEs (Clark et al., 1999b, 1999c). Figure 6B shows zincsensitive polyacrylamide sensors, containing immobilized Newport Green, and Texas Red-dextran as a reference dye. The maximal percent increase in the "Intensity Ratio" is 50%, which is less than observed with the Newport Green dye in solution (which has an increase of-250%). The data indicate a linear

512

P E B B L E Nanosensors Table 3. Summary of pH sensors. Results of calibrations of five pH-sensitive dyes and the corresponding acrylamide P E B B L E sensors. An internal standard, sulforhodamine 101 was added to each dye solution and was contained within the polymeric matrix of each P E B B L E sensor. '

-

pH Indcator -cNF

,,

. . . . . . . . . . . . . . . . . . .

Linear Range p H units 7.0 - 7.7

....... (dye)

L .

.

.

.

.

.

.

,,,,

slope" • SD

,,

-

,,

.

intercept

7.8 • 0_4

'

.

.

.

r~

n

-53

0'.98

6

CDMF + SR (dye)

6.2 - 7.4

3.3•

-19

0.96

12

B C P C F + SR ( d y e )

6.5 - 7.6

1.3•

-7.3

0.99

6

F S A + SR ( d y e )

5.8 - 7 . 2

3.1•

-17

0.99

7

SNAFL (dye)

7.2 - 7.7

-0.75 + 0.01

6.8

0.99

4

CNF (PEBBLEs)

7.0 - 7.7

0.24•

-0.66

0.96

5

0.67•

-3.0

0.99

6

0.43•

-1.6

0.90

6

3.4•

-17

0.99

5

C D M F + SR 6.2 - 7.4 (PEBBLEs) B C P C F + SR 6.2 - 7.2 (PEBBLEs) F S A + SR 5.8 - 7 . 0 (PEBBLEs) SNAFL 7.2 - 8.0 _(PEBBLEs) "Ratio o f n o r m a l i z e d f l u o r e s c e n c e i n t e n s i t y vs. .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

-0.49 + 0.02

.

4.6

. . . . . . . . . . . . . . . . . . . .

-

. , . . . . . .. .. . .

.

7

0.99 .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

pH.

CNF: 5-(and 6-) carboxynaphthofluorescein, CDMF: 5-(and -6-) carboxy-4",5"dimethylfluorescein, BCPCF: 2",7"-bis-(2-carboxypropyl)-5-(and-6)-carboxyfluorescein, FSA: fluorescein-5-(and -6)-sulfonic acid, SNAFL: 5-(and-6)-carboxy SNAFL| SR: sulforhodamine 101

Table 4. Results of calibrations of three calcium-selective dyes and the corresponding acrylamide P E B B L E sensors. An internal standard, sulforhodamine 101 was added to each dye solution and was contained within the polymeric matrix of each P E B B L E sensor.

....

i

.

;|;,

i

i

,

,,

,

i ;

Calcium Indicator

|....

Linear Range slope + SD' intercept # M calcium -Calcium Green . . . . . . . 0 - 0 . 1 5 ..................30 +017 . . . . . . . . i"7..... + SR (dye) Calcium Orange 0 - 0.15 1.5 + 0.03 1.0 + SR (dye) Calcium Green 5N 3 - 30 0.010 + 0.05 0.99 + SR (dye) Calcium Green 0 - 0.15 7.3 + 0.05 0.97 + SR (PEBBLEs) Calcium Orange 0 - 0.1 1.3 + 0.05 1.0 + SR (PEBBLEs) Calcium Green 5N 0- 5 0.022 + 0.007 1.0 + SR ~PEBBLEs ) . . . . . . . . . . . . . . . . . . . . .

513

r2

n J-

0.94

4

0.95

6

0.95

7

0.99

6

0.79

5

0.99

4

Brasuel, Kopelman, Philbert, et al.

1.4 1.2 1.00.8-

~ 2

0.6.

0.40.20.0

,

0

,

_

1

,

,

2

3

,,,

4

,

,

5

6

Number

Acqul=ition

Newport Green/Texas Red dextran PAA PEBBLEs Calibrated i n 5 . 1 m M MOPS pH 7.13

~

1.2

[

,

I

'

1

1

T

J

~

-: . . . . . .

~;----1~-

;

-1--=-

I I

' -' . . . . . .

~ ~ . . . . . .

' "

-'-

I-

-' . . . . . . . . . . . . . .

~

-,---

1"11 1.0

"~

0.9

~0.8

-I _'L

:l

"~ 0.7 ~ --

0.6

0

' . . . . . . . . . . . . . . ~- . . . . . . . . . . . . . . . . . . . '

I ,

-

-, .......

_

i

.

.

.

.

i _i

I L

, ......

.

.

.

/

.

.

.

.

.

.

, Z

t

'

! !

- -

i

,

, -,---

L

_1_

I

q,t _

_

_j

,

,

,

,

i

20

40

60

80

100

Zn =* (Ida)

Figure 6. Demonstration of acrylamide PEBBLE reversibility and calibration. (A) The pH of a solution of CDMF/SR PEBBLEs was cycled between pH 6.4 to pH 7.0 for several measurements. The ratios of the two peaks in the spectrum attained from each of the measurements were then plotted, illustrating that the PEBBLE sensors are reversible. (B) 3-mg/mL Newport Green/Texas Red PEBBLE suspension calibrated with zinc nitrate. The samples were excited at 508 nm on a fluorometer, and the emission spectra were collected. From the emission spectra, the ratios of intensities were collected.

range between 10 and 40 lxM Zn 2+, with a detection limit of 4 t~I. Although the sensor has a low affinity for zinc, these PEBBLEs may be useful for neural analysis where zinc levels can reach micromolar concentrations (Sumner et al., 2001). The batch-to-batch reproducibility of poly-acrylamide matrix PEBBLE sensors is very good. The standard deviation between batches matched the deviation found in a single batch. Despite the practice of calibrating each new batch, one can have confidence that once a procedure is optimized for a particular application, the protocol will be reliable to produce the required sensors (Clark et al., 1999b). To date consistent acrylamide PEBBLE sensors for pH, calcium, oxygen, and zinc have been produced. 514

PEBBLE Nanosensors

3.2. Decyl methacrylate tandem PEBBLEs The use of fluorescent indicator molecules in encapsulated form (acrylamide PEBBLEs) has proven valuable in the study of a number of intracellular analytes (Clark et al., 1998, 1999a, 1999b; Sumner et al., 2001) (H § Ca 2§ Mg 2§ Zn 2§ 02); however, there are many ions for which no fluorescent indicator dye is sufficiently selective or even available. An alternate class of tandem optical nano-sensors was thus required. Decyl methacrylate-based sensors, with much higher selectivity for most ions, rely on chemical equilibrium (or steady state) among different sensor components. These sensors act as "bulk optodes" or ion selective optodes (ISO) where the matrix (hydrophobic liquid polymer) contains a selective lipophilic ionophore ("optically silent"), a fluoroionophore and an ionic additive (Brasuel et al., 2001). The operation of the entire system is based on having a thermodynamic equilibrium that controls ion exchange (for sensing cations) or ion co-extraction (for sensing anions), i.e., an equilibrium-based correlation between different ion species. To achieve sensor miniaturization, fluorescence (rather than absorbance) has been utilized. Fabrication of PEBBLEs based on the described scheme, in the sub-micron size range, has been demonstrated for K § and Na § and extension to existing ionophores for other cationic, as well as anionic, analytes simply involves an optimization of the matrix for the required sensing elements (Brasuel et al., 2001). Larger liquid polymer sensing particles have also been recently developed by Eric Bakker and co-workers, utilizing a dispersion method to cast PVC particles and plasticizer-free dodecyl acrylate microspheres for ion sensing (Tsagkatakis et al., 2001; Peper et al., 2001).

3.2.1. DecylMethacrylate PEBBLE production. A batch of decyl methacrylate PEBBLE sensors typically consists of 210 mg of decyl methacrylate, 180 mg hexanedioldimethacrylate, 300 mg of dioctyl sebacate, with 10-30 mmol/kg each of ionophore, chromoionophore, and ionic additives added after spherical particle synthesis. The spherical particles are prepared by dissolving decyl methacrylate, hexanedioldimethacrylate, and dioctyl sebacate (DOS) in 2 ml of hexane. To a 100 ml round bottom flask, in a water bath on a Coming pc-351 hot plate stirrer, 75 ml of pH 2 HC1 is added along with 1,793 mg of PEG 5'000 monomethyl ether and stirred and degassed. The monomer cocktail dissolved in hexane is then added to the reaction flask (under nitrogen), stirred at full speed, and water bath temperature is raised to 80~ over 30-40 minutes. Six mg of potassium peroxodisulfate is then added to the reaction and stirring is reduced to medium speed. The temperature is kept at 80~ for two more hours, and then the reaction is allowed to return to room temperature and stir for 8-12 hours. The resulting polymer is suction-filtered through a Fisherbrand glass microanalysis vacuum filter holder with a Whatman anodisc filter (0.2 ~m pore diameter).

515

Brasuel, Kopelman, Philbert, et al.

Figure 7. SEM of gold-coated decyl methacrylate PEBBLEs produced in emulsion polymerization. Left, with PEG 5'000 monoethyl ether as surfactant. Size distribution of 500 nm + 40 nm = 48% and 600 nm + 40 nm = 37% for the PEBBLEs produced. Right,same procedure without PEG as a steric stabilizer. The polymer is rinsed three times with water and three times with ethanol to remove excess PEG and unreacted polymer. Tetrahydrofuran (THF) is then used to leach out the DOS and then the PEBBLEs are again filtered and rinsed. They are allowed to dry in a 70~ oven overnight. Dry polymer is then weighed out, and DOS, ionophore, chromoionophore and ionic additive are added to this dry polymer, so that the resulting polymer has 40% DOS, 20 mmole/kg ionophore, 10 mmole/kg chromoionophore, and 10 mmole/kg ionic additive. Enough THF is added to this mixture so as to just wet the PEBBLEs. The PEBBLEs are allowed to swell for eight hours and then the THF is removed by rotoevaporation. The resulting PEBBLE sensors are rinsed with double distilled water and allowed to air dry. For sizing, the PEBBLEs are suspended in a 50/50 water/ethanol solution. A few drops of this suspension are evaporated on a glass cover slip and sputtered with gold. Then the SEM images were taken on a Hitachi S-3200N Scanning Electron Microscope. Figure 7 clearly illustrates the utility of PEG 5'000 as a steric stabilizer for PEBBLE fabrication.

3.2.2. Decyl methacrylate (DMA) PEBBLE calibration. The fluorescence response scheme used to follow analyte binding of non-fluorescent ionophores in DMA-based PEBBLE sensors closely follows previous work on PVC-based fiber optic sensors selective for potassium and sodium ions (Buhlmann et al., 1998; Shortreed et al., 1996, 1997). The hydrogen-ion selective chromoionophore competes with the optically silent ionophore as cations enter the liquid polymer matrix. A lipophilic additive maintains ionic strength in the matrix and aids in preventing the co-extraction of anions. It allows for charge neutrality in the membrane without negative counter-ions being brought from the solution into the membrane (Bakker and Simon, 1992). The work described here takes advantage of an indicator with two fluorescence emission maxima (~.1,3.2), giving a relative intensity that changes with the degree of protonation (1-I). This degree of protonation, I7, can be evaluated in terms of 516

PEBBLE Nanosensors

Figure 8. Left: Normalized emission spectra from suspended K § PEBBLE sensors using the pH chromoionophore ETH5350 for ion-correlation spectroscopy in tandem with B ME-44 for the monitoring of K§ activity. The spectra show response in going from 10 mM KC1, 2.0 M KC1 (well beyond saturation of the sensor), all buffered at pH 7.2 with 10 mM Tris buffer. Right: (a) Response of BME-44-based decyl methacrylate PEBBLEs to potassium (o), and sodium (4), along with theoretical curves. Theoretical curves are constructed solving equation 5 for ax,+.(I§ being K § or Na § in this case), for a given value of 1-I. The solid lines delimit values for log (aK+/aH§ typically found in intracellular media and the dashed lines delimit the typical extracellular ratios (Ammann, 1986). (b) Response of K§ PEBBLEs to standard additions of KC1 in Tris buffer (o) compared to a similar experiment run in a constant background of 0.5 M Na § ( 9 ). The theoretical lines are drawn using Equation 5 with log K~ i = -3.3.

the fluorescence intensity ratio, Fx2~x~, given by the protonated chromoionophore intensity Fx2 and the deprotonated chromoionophore intensity Fx~ (See Figure 8 for spectra) (Shortreed et al., 1996). The experimentally obtained spectra are normalized to the iso-emmisive point of the dye and Equation 3 is used to determine the degree o f protonation of the dye at any given analyte concentration at a known pH (Brasuel et al., 2001). Superscripts P and D denote the completely protonated state and completely deprotonated state of the chromoionophore, respectively, lack of superscript denotes intermediate points.

517

Brasuel, Kopelman, Philbert, et al.

F~ F~ n=

F~~

F~

(3) -1

fa~o

fxl

The degree of protonation (FI) of the indicator spectra obtained from the PEBBLE calibration is related to the analyte concentration by using the theoretical treatment of ion-exchange sensors developed by Simon, Bakker and colleagues (Bakker and Simon, 1992; Buhlmann et al., 1998; Morf et al., 1989; Shortreed et al., 1996). For the incorporation of a selective neutral ionophore into a matrix, along with a selective chromoionophore for indirect ion monitoring (ion exchange sensors), the metal ion activity a~~+in solution (see Equation 4) is a function of the hydrogen ion activity an+ in solution, the interfering cations ajz+ (where K~ is the selectivity coefficient for the interfering ion) and the constants [Ltot], [Ctot], [Rtot-], which are total ionophore (ligand) concentration, total chromoionophore concentration, and total lipophilic charge site concentration in the membrane. Note that [CH] is the protonated chromoionophore concentration and [C] is the free base concentration. It is assumed that all components added during the matrix PEBBLE swelling procedure go into the matrix. The parameter FI has been def'med (Shortreed et al., 1996) as the relative portion of the protonated chromoionophore, YI =[CI-I]/[Ctot]. It follows that

I( ~

[e~o,]_(rl)[C,o,]

1 I(l_H)ax+l ~

a r + ""J aJ~" = ~K~_,

(4)

v(tL, o,J- 1"11t [R-,o,]-(H)[C,o,]})

which, for primary and interfering ions with charges of 1, simplifies to:

ai v~ +-,~o

o tl(tLtotlll

a/-=---(rI-'-l)a,,§ Kexch

-1

9

(5)

[R-to, l-l-I[Cto,]

Calibration of a K § sensor based on these principles is shown in Figure 8 along with normalized spectra (to demonstrate the ratiometric nature of the sensor). For potassium sensing, the chromoionophore is ETH5350, the ionophore is BME-44, and the lipophilic additive is potassium tetrakis-[3,5bis(trifluoromethyl)phenyl] borate (KTFPB) (Shortreed et al., 1997; Brasuel et al., 2001). The data points for potassium and sodium responses are plotted along with corresponding theoretical curves based on Equation 5. The constant K~x~his 518

PEBBLE Nanosensors determined from a line fit to the experimental data. Then the theoretical curve is plotted using the experimentally determined Kexchand the constants Rtot, Ctot and Ltot so as to find the expected a~§ for a given value of 1-I. Dashed lines delimit typical extracellular activity ratios and the solid lines delimit the intracellular levels (log (aK+/aH+)) (Ammann, 1986). It was found that the response matches well with the theory, which is gratifying considering the small size of the systems. The dynamic range at pH 7.2 extends from 0.63mM to 0.63M aK+. The log of the selectivity for potassium vs. sodium, determined by measuring the horizontal separation of the response curves at H = 0.5, is -3.3. This value, when used to plot the expected response in a sample with a constant 0.5M Na § interference, matches the experimental data obtained (see Figure 8). This value indicates a selectivity similar to or better than that obtained for other and larger matrices incorporating BME-44, e.g. -3.1 in PVCbased fiber optic work, and-3.0 in PVC-based microelectrodes (Brasuel et al., 2001; Shortreed et al., 1997). It also exactly matches the value given in the review by Buhlmann, Pretsch, and Bakker (Buhlmann et al., 1998) for a thin PVC film sensor. This selectivity should be more than sufficient for measurements in intracellular media where potassium concentration (Ammann, 1986) is about 100 rnM and sodium is about 10 mM. It can be easily demonstrated that most ionophores available for use in ionselective electrodes can be used in the development of PEBBLEs. This is illustrated by using the same PEBBLE formulations described except for replacing the BME-44 ionophore with sodium ionophore IV, (DD-16-C5 ;2,3:11,12-didecalino- 16-crown-5 ;2,6,13,16,19-Penta-oxapentacyclo[ 18.4.4.47'12.01'2~ dotria-contane), and the chromoionophore ETH5350 with a Chromoionophore II or 9-dimethylamino-5-[4-16-butyl-2,14-dioxoaeicosyl)phenylimino]-5H-benzo[a]phenoazine (ETH 2439)/1,1 '-dioctadecyl-3,3,3' ,3 'tetramethyl-indocarbocyanine perchlorate (DiI) combination (a ratiometric combination based on the inner filter effect (Shortreed et al., 1996)) as seen in Figure 9. The Na § decyl methacrylate PEBBLEs are formulated to have 30 mM/kg ionophore, 15 mM/Kg chromoionophore, and 15 mM/Kg negative lipophilic additive. Calibrations to NaC1 and KC1 were carried out in Tris buffer, pH 7.3, and theoretical response curves were constructed using Equation 5, after using Equation 3 to plot the experimental data (as with the K § PEBBLEs). It is shown that these PEBBLEs have a log selectivity of approximately-2.0, or are 100 times more selective for sodium than potassium. This selectivity works well for extracellular applications but intracellular applications will require a more selective ionophore.

519

Brasuel, Kopelman, Philbert, et al.

r-C .8-

-1

1

3

5

7

9

Log [Na+]/aH+ Figure 9. Response of sodium ionophore IV-based decyl methacrylate PEBBLEs to sodium (l-q), and potassium ( A ), along with theoretical curves. Theoretical curves are constructed solving Equation 5 for alv.. (I§ being Na§ or K§ in this case), for a given value of FI.

3.3. Sol-gel PEBBLEs Sol-gel glass has also been used as the matrix for the fabrication of PEBBLE nanosensors, because of the superior properties it has over organic polymers. Sol-gel glass is a porous, high purity, optically transparent and homogeneous material (Uhlmann et al., 1997), thus making it an ideal choice as a sensor matrix for quantitative spectrophotometric measurements. Also, it is chemically inert, photostable and thermally stable, compared with polymer matrices. The preparation of sol-gel "glasses" is technically simple and tailoring the physicochemical properties (i.e., pore size or inner-surface hydrophobicity) of sensor materials can be achieved easily by varying the processing conditions and the amount or type of reactants used. This enables the pore sizes to be optimized such that the analyte is able to diffuse easily and interact with the sensing molecules whilst the latter are prevented from leaking out of the matrix (also true for polyacrylamide based sensors). Furthermore, this "glass" is produced under so-called soft chemical conditions, allowing the inclusion of biomolecules. A range of sol-gel sensor configurations has been described in the literature, including monoliths, thin films, miniaturized probe-tips and powders (Uhlmann et al., 1997). Immobilization of the sensing reagent in a supportive matrix is a critical step in the fabrication of optical sensors. It can be achieved by either chemical or physical entrapment of the fluorescent dye molecules in the pore structures of the sol-gel network. An important advantage of physical entrapment is the minimal alteration in the spectral and binding properties of the sensing molecules, due to the weak interactions with the supporting matrix. A key 520

PEBBLE Nanosensors

Figure 10. A typical SEM image of PEGylated St6ber silica nanoparticles. The scale bar is 1 l.tm. The particles are prepared by using PEG MW 5000 monomethyl ether (2 g), methanol (99.9%, 24 ml), ammonium hydroxide (30% wt of ammonia, 6 ml), and TMOS (99.9%, 0.2 ml). advantage of sol-gel sensors is that the "soft" chemical techniques used are ideal for the entrapment of enzymes and proteins. For instance, one can incorporate oxidase enzymes into proven ratiometric oxygen sensing sol-gel PEBBLEs (Xu et al., 2001) for the sensing of analytes like glucose (using the glucose oxidase enzyme).

3.3.1. Sol-gel PEBBLE production. For the production of oxygen sensitive solgel PEBBLEs, the reaction solution consists of polyethylene glycol (PEG) MW 5000 monomethyl ether (3 g), absolute ethanol (6 ml), Oregon Green-dextran MW 10,000 (0.1 mM), [Ru(dpp)3] 2+ (0.4 mM), and 30% wt. ammonia water (3.9 ml) with ammonia serving as catalyst and water being one of the reactants. Upon mixing, the solution becomes transparent and tetraethyl orthosilicate (TEOS) (0.5 ml) is added dropwise to initiate the hydrolysis of TEOS. The solution is then stirred at room temperature for 1 hour to allow the sol-gel reaction to reach completion. A liberal amount of ethanol is then added to the reaction solution and the mixture is transferred to an Amicon ultrafiltration cell (Millipore Corp., Bedford, MA). A 100 kDa membrane is used to separate the reacted sol-gel particles (PEBBLEs) from the unreacted monomers, PEG, ammonia and dye molecules, under a pressure of 10 psi. The PEBBLEs are further rinsed with 500 ml ethanol to ensure that all unreacted chemicals have been removed from the PEBBLEs. The PEBBLE solution is then passed through a suction filtration system (Fisher, Pittsburgh, PA) with a 2 lxm filter membrane to separate the larger size particles from the smaller ones. The filtrate (containing the smaller 521

Brasuel, Kopelman, Philbert, et al. particles) is filtered again, this time with a 0.02 lxm filter membrane, to collect the particles which are then dried to yield a final product consisting of sol-gel PEBBLEs in the size range of 100-600 nm in diameter. Figure 10 shows a scanning electron micrograph of gold-coated sol-gel particles produced using optimized component ratios for 100 gm PEBBLEs (Xu et al., 2001). 3.3.2. Sol-gel PEBBLE calibration. The oxygen sensing sol-gel pebbles are based on the quenching of luminescence by oxygen. The oxygen quenching process is ideally described by the linear Stem-Volmer equation:

I0fl = 1 + Ksv p[Oz]

(6)

where I0 and I are the luminescence intensities in the absence and presence of oxygen at a partial pressure of p[O2], respectively, and Ksv is the Stern-Volmer constant, which depends directly upon the diffusion constant of oxygen, the solubility of oxygen, and the quenching efficiency and lifetime of the excitedstate of the fluorophore (Demas, 1991). Sol-gel PEBBLE sensor response is determined from the ratio (R) of the fluorescence intensities of [Ru(dpp)3] 2+ to Oregon Green 488-dextran. The overall gas phase quenching response, QG, is given by: QG = (Ir~2- Io2) / In2

(7)

Where IN2and lo2 denote intensities in 100% N2 and 100% 02, respectively. The measured value of QG for the sol-gel PEBBLEs is -92%. While sol-gel based oxygen PEBBLEs have retained the sensing properties of thin film sol-gel sensor for gas phase sensing, for biological applications the main interest is in solution-based oxygen sensing (Xu et al., 2001). Figure 11 shows the response of the TEOS-based sol-gel PEBBLEs to dissolved oxygen (DO). The QDO for the sol-gel oxygen PEBBLEs is -80%. (Quenching response to dissolved oxygen, Qoo, is defined in a similar way to Q~ (gas phase sensing), where IN2 (intensity at 100% N2) and lo2 (Intensity at 100% 02) are replaced by I in fully deoxygenated water and I in fully oxygenated water, respectively). This value represents a slight reduction in performance vis-ft-vis gaseous oxygen; however it is a great improvement with regards to TEOS-based sol-gel films (McEvoy et al., 1996). These sol-gel films had an excellent response to oxygen in the gas phase (QG = 90%), but a poor quenching response to dissolved oxygen (QDo = 20%). MacCraith and co-workers have subsequently reported significant improvements to the QDO ratio by preparing organically modified sol-gel

522

PEBBLE Nanosensors

3-

450000 ~"

400000

"~

350000

"~

3OOOOO

9-

2.000

3.0 - I

2.5

2.0

~

I~

i ,,~

J["

1.5

"~

100ooo

~>"

50000 1,1..

O

t"

/ , 500

-

, ......... 9 550

", 600

. . . . . 850

9 700

o.o

'

~e' ........... -" 0

,0,

2'o

3'0

go

Oxygen Concentration (ppm)

Wavelength ( n m )

Figure 11. Left. Aqueous phase emission spectra of sol-gel oxygen PEBBLEs excited at 488 nm. Top line: PEBBLE solution purged with N2; middle line: PEBBLE solution purged with air; bottom line: PEBBLE solution purged with 02. Right. Stern-Volmer plot of relative fluorescence intensity ratios for ratiometric sol-gel oxygen PEBBLEs in aqueous phase.

(Ormosil) films using methyltriethoxysilane (MTEOS) and ethyltriethoxysilane (ETEOS) as the precursors. The success of the Ormosil films in raising the QDO ratio to 70-80% is largely attributed to the increased hydrophobicity of the film which reduced the water solubility in the film and enhanced the partitioning of oxygen out of solution and into the film (McDonagh et al., 1998). It is thought that the high QDO response of the TEOS sol-gel PEBBLEs might be caused by the PEG content of the sensing matrix, PEG playing a role analogous to the Ormosil precursors and thus partitioning the oxygen preferentially into the solgel PEBBLEs. It is well known that oxygen has a higher solubility in organic liquids than in water (Merck, 1996), so it should dissolve much better in an organic phase compared to an aqueous phase. In summary, doping the sol-gel PEBBLEs with PEG adds organic components to the sensing matrix, thus encouraging the partitioning of oxygen into the matrix and increases the accessibility of oxygen to the entrapped indicator dye molecules. This is in addition to the role PEG plays in preventing particle aggregation during PEBBLE sensor fabrication. Figure 11 (right) shows the Stern-Volmer plot of fluorescence intensity ratios to oxygen concentrations. Although the performance of the sol-gel PEBBLEs is slightly reduced in the aqueous phase, as opposed to the gas phase, the sensors still demonstrate good reversibility and reproducibility (Xu et al., 2001). The dashed line in Figure 11 (right) shows the extent of the biologically relevant regime of oxygen concentrations. We note that in this regime (from 0 to ~30 ppm oxygen), the Stem-Volmer plot is quasi-linear (r2 = 0.988). The sensors showed at least 95% recovery each time that the sensing environments were changed among air-, O2-, or N2-saturated sensor solutions. 523

Brasuel, Kopelman, Philbert, et al. 4. Advantages and Limi,tations of PEBBLE Sensors

As mentioned, the main advantages of the PEBBLE sensor are first, protection of the cell from the sensing elements and protection of the sensing elements from cellular interferents, and second, the ability to combine components to accomplish complex sensing schemes. The greatest advantage, from a sensing standpoint, is that all of this is accomplished in a sensor with nanometer dimensions! The advantage gained by protecting the cell from the sensing elements is selfevident, especially when the selective sequestration of some dyes into cellular organelles is considered. What may be less intuitive is the interference of cellular components, especially protein, with the function of sensing dyes (Graber et al., 1986). The comparison of the function of free dyes to both acrylamide and sol-gel PEBBLEs utilizing the dyes as sensing elements clearly demonstrates the advantage of preventing macro-molecule interaction with sensing components. One example of the effect of protein binding is demonstrated with 5- (and 6-) carboxynaphthofluorescein (CNF) pH dye. CNF is a highly photostable, ratiometric dye for pH, which is not used for intracellular applications because of the error induced by macromolecule binding. However, protected in an acrylamide matrix, CNF becomes a viable tool for intracellular pH study. Figure 12 illustrates the benefit of entrapping CNF in an acrylamide matrix (PEBBLE). Incubation of the free dye with as little as 0.01% albumin induces alterations in emission ratios of almost 90% (pH was maintained constant as measured with a standard pH electrode), which is an error equivalent to 1 pH unit. The same dye, protected in the PEBBLE, shows minimal perturbation by albumin, with the resulting error equivalent to about 0.01 pH units (Clark et al., 1999b). Another example of the benefit of sensing element protection by the acrylamide matrix is found when comparing the Zn PEBBLE (based on Newport Green) to naked Newport Green in bovine serum albumin (BSA) solution. Aliquots of the BSA solution were added to either a 3 mg/mL PEBBLE suspension or a 125 nM Newport Green dye solution and the resulting fluorescence was monitored. Although Newport Green has good selectivity over intracellular ions, the dye itself is prone to artifacts resulting from non-specific binding of proteins, such as (BSA), as shown in Figure 12. Monitoring the peak of Newport Green at 530 nrn, there is a substantial increase in the peak intensity with each successive addition of BSA. In addition to the increase of intensity, there is also a 4 nm shift in peak wavelength. The PEBBLEs containing the Newport Green dye, however, are unaffected by the additions of BSA, which is confirmed by the peak wavelength remaining the same. As little as 0.02% BSA causes a major change in the Newport Green dye intensity (i.e. >200% increase) but the intensity of the

524

PEBBLE Nanosensors

3.5

160 CNF

140 ~-

dye~

3.0

120

2.5

o :=

100 ,..o

n"

._o

,c

o

20

N

o

CNF P E S e L E , ~ .'-'r"-e,'-":

ooo

o.ol

........;

0.02

::-

/

o;3

o.~

~

9

o os

% Albumin

/

-.-o

O.S ~ 9

4r . . . . . .

/

1.o

,,o

~ .........

,,&/-"

2.0

1.5

...,

~

....

~ - - - ~ - - - ~ - - - ~ - - - o

-~..o,,,..,,~.,'~.,,~s

o.o ~ l , T , , . 0

....

.... 0.02

-.-N.~., o ,

-. . j. . . . . . . . . . . .

,,

0.04

0.1

0.06 0.08 BSA ( w / v %)

,-~,., 0,12

Figure 12. Effect of protein on sensors: Left: Adding as little as 0.01% albumin to a solution of CNF dye molecules causes almost a 90% change in the fluorescence intensity ratio of this pH-sensitive dye, even though the pH of the solution remains constant. Under the same conditions the PEBBLEs containing the CNF dye are not affected by the addition of albumin. Right: Peak emission intensity of Newport Green, 530-nm, monitored on a fluorometer. Spectra are acquired after each successive aliquot of a 10% (w/v) bovine serum albumin solution. The BSA concentrations are plotted versus the peak intensities. As little as 0.02% BSA causes over a 200% increase in Newport Green dye intensity, but the intensity of the Newport Green embedded in acrylamide PEBBLE remains unchanged, even at BSA concentrations above 0.10%. Newport Green embedded in the sensor remains unchanged, even at BSA concentrations above 0.10% (Sumner et al., 2001). As expected, it has been found that the sol-gel matrix of PEBBLE sensors also prevents macromolecules such as proteins, from diffusing through the matrix. The matrix thus protects the entrapped dyes from the intracellular environment, preventing interference with the fluorescent properties of the dyes. Without this shielding of a dye, its fluorescence would behave unpredictably inside a given cell, making calibration of even ratiometric dyes difficult or impossible. As with the acrylamide PEBBLEs the effects of non-specific protein binding have been investigated by the addition of bovine serum albumin. The results show that adding as little as 0.14% BSA to a solution containing [Ru(dpp)3] 2§ and Oregon Green 488-dextran dye (at the same molar ratio in solution as inside the PEBBLEs) changes the fluorescence intensity ratio of the two dyes by a factor of more than 2.3 (i.e., an increase of over 130%). This change is mostly due to the change in the fluorescence intensity of [Ru(dpp)3] 2§ after the addition of BSA, while the intensity of Oregon Green remains basically unchanged. However, under the same conditions, the PEBBLE sensors containing these two dyes are not affected by the addition of BSA and a change in fluorescence intensity ratio of at most 4% is observed when even an increased concentration of BSA (0.23%) is added, demonstrating the same protection as given to dyes in the acrylamide matrix. The susceptibility of the sol-gel PEBBLEs to heavy metal ions (Hg 2§ and Ag § and to one of the notorious collisional quenchers (I) were also examined. 525

Brasuel, Kopelman, Philbert, et al. Hg(NO3)2, Ag(NO)3 and KI were added to a PEBBLE solution and to a free dye solution of [Ru(dpp)3] 2§ (same concentrations as in PEBBLEs) up to a concentration of about 200 p2Pm. There was a 5-10% decrease in the fluorescence intensity of the [Ru(dpp)3] § free dye each time, while for the PEBBLEs no measurable effect could be observed (Xu et al., 2001). There is not a similar "naked" dye/PEBBLE-protected dye comparison for liquid polymer DMA-based PEBBLEs. The hydrophobic components can not be used "naked" in aqueous solution. The separate phase (the liquid polymer) is essential for the decyl methacrylate sensing mechanism and allows for the complex sensing schemes that effect the ion-exchange and co-extraction mechanisms (Buhlmann et al., 1998). The nanometer dimensions of the PEBBLEs of all three matrices give a useful advantage over traditional, monolithic optodes in terms of response time. In order to follow biological perturbations in real time, a fast response is required from the PEBBLE sensors. Most PEBBLE sensors depend on bulk-equilibrium, between sensor and solution phase (the oxygen sensor depends on steady-state). The diffusion of analyte inside acrylamide and sol-gel should be similar to that in aqueous solution phase, while the diffusion of analyte in the hydrophobic decyl methacrylate matrix is much slower. In all cases, the small size of the PEBBLE sensors gives a rapid response time, despite the need for bulk equilibrium. Figure 13 shows the response times of the various matrices. A reliable method for determining the response time of acrylamide sensors to Ca 2§ used an Olympus IX50 inverted microscope equipped with a mercury arc lamp and a PMT. Calcium selective probes were premixed with a caged calcium ion, and this solution was inserted into a quartz capillary. The calcium was uncaged with a pulse of UV light from a Quanta-Ray 10 ns Nd/Yag laser (Quanta-Ray, Mountain View, CA) equipped with a frequency tripler and coupled into an optical fiber positioned over the capillary. In Figure 13A the response time of the PEBBLE sensors was compared to that of the free dye (no polymer matrix), so as to separate the diffusion time through solution from the diffusion time through the matrix. As can be seen from Figure 13A, the 90 % response time of the PEBBLE sensor to the increase in free calcium is on the order of o n e m i l l i s e c o n d o r less. Theoretically, with an approximate diffusion constant of 10 -6 cm2/sec, the average diffusion time should be about ten microseconds for a 100 nm radius sensor, and 100 nanoseconds for a 10 nm radius sensor (Clark et al., 1999b). The measured transition times for sol-gel (Figure 13C) are on the order of 20 to 30 s, but these times are much longer than the intrinsic response time of the PEBBLEs, due to the significant contribution of the time used to saturate the solution with 02 or N2. It is difficult to measure the exact response time, because changing between oxygenated and deoxygenated PEBBLE solutions takes time. 526

PEBBLE Nanosensors

41 B

I/I

9

t~ /n

~. 30 ,~20 0 Ig

15

._o

0

17

,

,

,

5

10

15

19

. . . . . . . . ,

20

25

Time in S e c o n d s

0.7 O

C

0.6. G:

o

o.5 0.4

II111-11

i.--Im

tl-| ....

Time

,

(minute)

Figure 13. (A) Response time of acrylamide PEBBLE sensors. Calcium was released using a single 10 ns UV pulse from a Nd/Yag laser which photolysed the cage, releasing free calcium into a solution of PEBBLEs. The observed response time, less than 1 msec, was indiscernible from that of the corresponding dye not entrapped in a polymer matrix. (B) Response time of K§ PEBBLEs to added KC1 and added buffer solutions. It can be seen that response in the forward direction is about 0.5 s (0-40 mM KC1, A) and the reverse (see inset) slightly longer (40-20 mM KC1, B), 0.8s. (C) Reversibility of sol-gel PEBBLE sensor response to dissolved oxygen. 527

Brasuel, Kopelman, Philbert, et al. The measured transition times (including the time of saturating the solution) are only an upper limit of the response time. The PEBBLE sensors should intrinsically have shorter response times than previously reported thin film and fiber optic sol-gel sensors (on the order of seconds or minutes) simply because of the smaller sizes of these sensors. According to the Einstein diffusion equation, where X2=2D'c, a shorter diffusion length X (which is directly related to the size of the sensor) results in a much shorter time for oxygen molecules to diffuse through the sensing matrix (which is basically the response time). A lower limit can thus be estimated, using D=2xl0 9 m2/sec (diffusion constant of oxygen in water) and X = 3 x l 0 7 m , giving x=20xl0 6 sec, i.e., a response time in the microsecond range. An upper limit can be estimated considering that the PEBBLE sensor dimensions are 10-100 times smaller than thin film sensors, and have a spherical shape. This should give the PEBBLE sensors a response time in the millisecond range (Xu et al., 2001). For decyl methacrylate based K + sensors, using BME-44 as ionophore, ETH5350 as the chromoionophore and KTFPB as the ionic additive, the ratio of the protonated chromoionophore to free base was analyzed vs. time in response time measurements. It was found (Figure 13B) that in going from log aK+/aH+ = 3.6 to 5.7, the response time (10%-90% signal change) was about 0.5 s (for a concentration change of over 2 decades). In the reverse direction, the response was about 0.8 s (Figure 13B). This fast, sub-second response time of the PEBBLEs is a direct result of their small size. Diffusion in decyl methacrylate is in the range of 108 cm2/s, with small variations that depend on cross-linker content (Ambrose and Meyerhoff, 1996, 1997). Thus, for a PEBBLE radius of about 300 nm, one expects a diffusion time of about 10 -3 S. This is consistent with the experimental values, which are again upper limit values due to the solution mixing times (Brasuel et al., 2001). To summarize, using acrylamide or sol-gel PEBBLEs, one can monitor changes with response times on the microsecond timescale or, using the decyl methacrylate PEBBLEs, one can monitor changes on the 0.5s time scale. Leaching of dye molecules out of the PEBBLE matrix is the greatest limitation of PEBBLE sensors; it is a major concern and is very dependent on the dye and matrix combination. Factors such as the molecular size of the dye (small dyes can more readily diffuse through the pores and leak out of the matrix) and the solubility of the dye in the matrix and in water play a significant role. Assays of the leaching of CDMF and calcium green from polyacrylamide PEBBLEs (each containing the sulforhodamine 101 reference dye) show that less than 50% of the dye leaches from the PEBBLEs in a 48 hr time period. On the time scale of the current single cell experiments, mostly a couple of hours, the dye loss is acceptable. With comparable leaching rates for indicator and 528

PEBBLE Nanosensors reference dye, the problem is minimized due to the ratiometric sensing scheme used with these PEBBLEs (Clark et al., 1999b). Decyl methacrylate-based potassium sensors have a lifetime of 30 minutes, due to component leaching from the liquid polymer membrane. This is consistent with the lifetime of PVC-based optodes of the same composition (Shortreed et al., 1997). After 30 minutes, the sensor response can deviate up to 7% from the initial calibration data at lower K § concentration. After 90 minutes, the deviation is up to 13% at lower K § concentrations. The deviations are smaller at larger K § concentrations (Brasuel et al., 2001). Sol-gel supports provide excellent stability with respect to dye leaching (Ingersoll and Bright, 1997). In particular, ruthenium complexes often have excellent stability inside the sol-gel matrix and in agreement with previous reports (Klimant et al., 1999; McDonagh et al., 1998; Ingersoll and Bright, 1997; Bossi et. al., 1999; Murtagh et al., 1998) the indicator dye [Ru(dpp)3] 2§ shows no signs of leaching. For the reference dye, the large size of the dextran molecular backbone to which the Oregon Green dye molecules are bound should greatly reduce leaching. According to the dilution factor, a rough estimate provides an upper limit of 1% for the amount of dye molecules leached out of the sensing matrix over a three-day period (Xu et al., 2001). In summary, what has been determined for three matrices is that, at the short end of the lifetime scale, the decyl methacrylate PEBBLEs have a lifetime of about half an hour, at midrange the acrylamide PEBBLEs have a lifetime of about 24 hours, and the sol-gel PEBBLEs have a lifetime of more than three days. However, as the PEBBLEs are single-use sensors made for quick measurements inside in vitro cells that survive only a short period of time, this is acceptable. It should also be noted that such well documented steps as covalently attaching the dye to the matrix polymer backbone and adding cages or lipophilic tails (i.e., using dextran with hydrophilic polymers or adding lipophilic tails to lipophilic dyes) can be accomplished to increase PEBBLE sensor lifetime when needed. As PEBBLEs get smaller and smaller, the concentration of dye should remain the same, but fewer and fewer dye molecules are available for generating signal or for participating in equilibrium-based thermodynamic sensing schemes. It is known that with currently available intensified CCD (Charge Coupled Device) technology, at least 5 dye molecules per PEBBLE are required to image a single PEBBLE. This limit has already been reached with our acrylamide-based PEBBLEs. Calculations based on the dye concentration present in the acrylamide emulsion predict that for an acrylamide PEBBLE of 20 nm, on average, there is less than one dye molecule per PEBBLE. We do not attempt to image these particles singly, but rather use an ensemble of acrylamide PEBBLEs for imaging. For single acrylamide PEBBLE work, we have used 200 nm PEBBLEs, which contain approximately 100 dye molecules. In the case of liquid 529

Brasuel, Kopelman, Philbert, et al. polymer (DMA) PEBBLEs, there is a minimum number of required molecules for the thermodynamic equilibrium sensing scheme to be useful. It is not yet known what the minimum size of decyl methacrylate PEBBLEs is, but it will be obtained as the size of the liquid polymer PEBBLEs is decreased. The PEBBLE sensing technology is fairly new and the flexibility in design provided by the available matrices and components has a long way to go before all possibilities for this technology are discovered.

5.

The Future of PEBBLE Technology

The obvious next step in PEBBLE technology is to apply the PEBBLE techniques currently used to other biologically interesting ions and work continues on all three matrices to develop sensors for all ions of biological relevance. Further developments in progress are given below. As an example of an enzyme-based sensor, oxidase enzymes have been incorporated into the acrylamide PEBBLEs along with the oxygen sensing components. In this manner, enzyme selectivity is used to detect molecules of interest while the steady state oxygen sensor gives the means to indirectly monitor the analyte concentration based on the rate of enzyme activity. Glucose-sensitive PEBBLEs have been designed to be ratiometric by entrapping a reference fluorescent dye (Oregon Green 488-dextran), in addition to the oxygen sensitive dye Ru[(dpp)SO3Na]3 and the glucose oxidase enzyme, within a polyacrylamide (PAA) matrix. The common glucose sensing scheme involves the employment of glucose oxidase, which catalyzes the oxidation of glucose according to the following equations, and oxygen is consumed during the process:

d - glucose + 02 cJ.~___> d - gluconolactone + H202

(8)

d- gluconolactone + HzO --->d-gluconic acid

(9)

The measurement of the reduced oxygen level by the oxygen-sensitive dye Ru[(dpp)SO3Na]3, when glucose is oxidized by the enzyme, serves as an indirect indication for the glucose concentration (Figure 14). The calibrations of the glucose PAA PEBBLEs were taken on a FluoroMax-2 spectrofluorometer (ISA Jobin Yvon-Spex, Edison, NJ), slits set to 2 nm for both the emission and excitation. During the calibration, aliquots of a glucose solution 530

PEBBLE Nanosensors

u

1600 g

":" 1400

W r

11

1200

I m

_= looo

8 ~

O

".

0.0

'1

,

~,

L

0.5

|,

1.0

~

,.

!

1.S

...... .

,

!

--

2.0

Glucose added (raM)

Figure 14. Plot showing the response of PAA glucose PEBBLEs to the increasing glucose concentrations. were added to a cuvette that contained a PAA glucose PEBBLE solution. The changes in the ratios between the emission peak intensifies of Ru[(dpp)SO3Na]3 and Oregon Green 488-dextran, due to the changes in the glucose concentration in the PEBBLE solution, were recorded for data analysis. Modifications of the PEBBLE outer shell can serve as both a platform for targeting the PEBBLEs to specific biological locations (using antibodies) or for the development of unique sensing schemes of species too reactive to enter the polymer membrane. The attachment of antibodies to the surface of acrylamide PEBBLEs is facilitated through the use of biotin-avidin binding. This strategy involves biotinylating the PEBBLEs and the antibodies and then using avidin to bridge the two. The use of avidin as a bridging molecule is possible because of its multiple sites that bind biotin with great affinity. In order to biotinylate the acrylamide PEBBLEs, primary amine groups are introduced to the PEBBLE surface through an N-(3aminopropyl)methacrylamide copolymer, in order to produce aminefunctionalized polyacrylic acid (AFPAA) PEBBLEs. Under mildly basic conditions, the AFPAA PEBBLEs can simply be introduced to a succinimidyl ester conjugate of biotin, and the primary amines react with the ester to form a covalent bond. A similar strategy is used to produce acrylamide PEBBLE sensors for hydroxyl radical. The hydroxyl radical is one of the most reactive and potentially damaging species found in biological systems. It is primarily formed by reactions involving superoxide and/or hydrogen peroxide and copper or iron ions. A challenge in creating a sensor for the hydroxyl radical is its high reactivity. Our current detection scheme for hydroxyl radical is a PEBBLE-based assay 531

Brasuel, Kopelman, Philbert, et al. CCA 7-OH-CCA o

~~0

0-F o~H 385 nm HO"

0 ~

"0"

~0

Figure 15. Conversion of non-fluorescent CCA to the fluorescent species 7-OH-CCA by reaction with the hydroxyl radical (Maneyich et al., 1997).

using the molecule coumarin-3-carboxylic(CCA) to detect OH. CCA is a nonfluorescent aromatic compound that reacts with OH to produce 7-OH-CCA, which is highly fluorescent and emits around 450 nm (Figure 15). In most of the PEBBLEs, the dye is entrapped inside the matrix, but for the hydroxyl radical probe, CCA is attached to the outside of the PEBBLE. This is done because of the extreme reactivity of the hydroxyl radical. As with the biotinylated PEBBLEs, CCA-coated PEBBLEs are produced by using AFPAA PEBBLEs and a succinimidyl ester of CCA (coumarin-3-carboxylic acid succinimidyl ester (SECCA)). The succinimidyl ester in SECCA reacts with the primary amine groups on the PEBBLE to produce an amide bond, effectively coating the PEBBLE with CCA. Initial experimentation has shown that the functionalized CCA PEBBLEs increase fluorescence in the presence of hydroxyl radical producing reactions. The conjugation of multiple dyes (20-40+ per PEBBLE) allows for the monitoring of OH production over time in a localized area and may allow for the imaging of single PEBBLEs. This has a great advantage in the measurement of OH radical, because of the localized production and the short diffusion radius of the OH radical in biological systems. Work continues on increasing the scope of analytes that PEBBLEs can detect. PEBBLEs are a new technology. Their possible applications and scope of impact are just beginning to be realized.

532

PEBBLE Nanosensors

Color plate 2, Figure 6. Confocal images of rat C6 glioma cells loaded with sol-gel PEBBLEs by gene-gun injection. (A) Nomarski illumination overlaid with Oregon Green fluorescence of PEBBLEs inside cell. (B) Nomarski illumination overlaid with [Ru(dpp)3] 2§ fluorescence of PEBBLEs inside cells. Right Fluorescence spectra of a typical ratiometric sensor measurement of molecular oxygen inside rat C6-glioma cells; bottom line: cells (loaded with sol-gel PEBBLEs) in air-saturated DPBS; middle line: cells in N2-saturated DPBS, 25 seconds after replacing the air-saturated DPBS; top line: cells in N2-saturated DPBS, after 2 minutes.

533

Brasuel, Kopelman, Philbert, et al.

6. Acknowledgments The authors thank Steve Parus for instrumentation expertise, Susan Barker for technical assistance, Professor Mark E. Meyerhoff and Theresa M. Ambrose for help in the initial stages of utilizing DMA and Maria J. Moreno for assistance with the data normalization formalism for DMA PEBBLEs, Rhonda Lightle and Chris Edwards for transmission electron microscopy, the University of Michigan Transgenic Core for technical assistance, and University of Michigan Electron Microbeam Analysis Laboratory (funded in part by NSF grant EAR-9628196) for use of the SEM.. We also gratefully acknowledge NIH Grants 2R01-GM50300 (Kopelman) and R01-ES08846 (Philbert) for funding.

7. References Albert, K. J. and D.R. Walt, 2000, Anal. Chem. 72, 1947. Ambrose, T. M. and M.E. Meyerhoff, 1996, Electroanal. 8, 1095. Ambrose, T. M. and M.E. Meyerhoff, 1997, Anal. Chem. 69, 4092. Ammann, D., 1986, Ion-Selective Microelectrodes, Springer, Berlin, 346. Bakker, E. and W. Simon, 1992, Anal. Chem. 64, 1805. Barker, S. L. R., M.R. Shortreed and R. Kopelman, 1997, Anal. Chem. 69, 990. Barker, S. L. R., B.A.Thorsrud and R. Kopelman, 1998, Anal. Chem. 70, 100. Bossi, M. L., D., E. Marta and P.F. Aramendia, 1999, J. Photochem. Photobiol. A 120, 15. Brasuel, M., R. Kopelman, T.J. Miller, R. Tjalkens and M.A. Philbert, 2001, Anal. Chem. 73, 2221. Buhlmann, P., E. Pretsch and E. Bakker, 1998, Chem. Rev 98, 1593. Clark, H. A., S.L.R. Barker, R. Kopelman, M. Hoyer and M.A. Philbert, 1998, Sens. Actuators B Chem. 51, 12. Clark, H. A., M. Hoyer, S. Parus, M. Philbert and R. Kopelman, 1999a, Mikrochim. Acta 131, 121. Clark, H. A., M. Hoyer, M. Philbert and R. Kopelman, 1999b, Anal. Chem. 71, 4831. Clark, H. A., R. Kopelman, R. Tjalkens and M.A. Philbert, 1999c, Anal. Chem. 71, 4837. CohenKashi, M., M. Deutsch, R. Tirosh, H. Rachmani and A. Weinreb, 1997, Spectrochim. Acta A Mol. Spectrosc. 53, 1655. Daubresse, C., C. Granfils, R. Jerome and P. Teyssie, 1994, J. Colloid Interface Sci. 168, 222. Demas, J. N. and B. A. DeCrraff, 1991, Anal. Chem. 63, 829A. Dourado, S. and R. Kopelman, 1996, Proc. SPIE 2836, 2. Emmi, A., H.J. Wenzel and P.A. Schwartzkroin, 2000, J. Neurosci. 20, 3915. Graber, M. L., D.C. DiLillo, B.L. Friedman and E. Pastoriza-Munoz, 1986, Anal. Biochem. 156, 202.

534

PEBBLE Nanosensors Haugland, R. P., 1993, Molecular Probes Handbook of Fluorescent Probes and Research Chemicals, Molecular Probes, Inc., Eugene, OR, 679. Ingersoll, C. M. and F.V. Bright, 1997, Chemtech, 27, 26. Jung, S.-K., W. Gorski, C.A. Aspinwall, L.M. Kauri and R.T. Kennedy, 1999, Anal. Chem. 71, 3642. Klimant, I., F. Ruchruh, G. Liebsch, A. Stangelmayer, and O. S. Wolfbeis, 1999, Mikrochim. Acta, 131, 35. Kurihara, K., M. Ohtsu, T. Yoshida, T. Abe, H. Hisamoto and K. Suzuki, 1999, Anal. Chem. 71, 3558. Maneyich, Y., K.D. Held, J.E. Biaglow, 1997, Radiation Res. 148, 580-591. McDonagh, C., B. D. MacCraith and A.K.McEvoy, 1998, Anal. Chem. 70, 45. McEvoy, A. K., C. McDonagh and B.D. MacCraith, 1996, Analyst, 121,785. McNamara, K. P., T. Nguyen, G. Dumitrascu, J. Ji, N. Rosenzweig and Z. Rosenzweig, 2001, Anal. Chem. 73, 3240. McNamara, K. P. and Z. Rosenzweig, 1998, Anal. Chem. 70, 4853. The Merck Index, 12th edition, 1996, Merck & Co, NJ. Mohr, G. J., F. Lehmann, R. Ostereich, I. Murkovic and O.S. Wolfbeis, 1997a, Fresenius J. Anal. Chem. 357,284. Mohr, G. J., I. Murkovic, F. Lehmann, C. Haider and O.S. Wolfbeis, 1997b, Sens. Actuators B Chem. 39, 239. Morelle, B., J.M. Salmon, J. Vigo and P. Viallet, 1994, Cell Biol. Toxicol. 10, 339. Morf, W. E., K. Seiler, B. Lehmann, C. Behringer, K. Hartman and W. Simon, 1989, Pure Appl. Chem. 61, 1613. Murtagh, M. T., M. R. Shahriari and M. Krihak, 1998, Chem. Mater. 10, 3862. Overly, C. C., K.D. Lee, E. Berthiaume and P.J. Hollenbeck, 1995, Proc. Nat. Acad. Sci.~USA 92, 3156. Peper, S., I. Tsagkatakis and E. Bakker, 2001, Anal. Claim. Acta 442, 25-33. Rosenzweig, Z. and R. Kopelman, 1995, Anal. Chem. 67, 2650. Rosenzweig, Z. and R. Kopelman, 1996, Anal. Chem. 68, 1408. Ross, W. N., 1993, Biophys. J. 64, 1655. Sasaki, K., Z.-Y. Shi and R. Kopelman, 1996, Chem. Lett. 2, 141. Seitz, W. R., M.T.V. Rooney, E.W. Miele, H. Wang, N. Kaval, L. Zhang, S. Doherty, S. Milde and J. Lenda, 1999, Anal. Chim. Acta 400, 55. Shortreed, M., E. Bakker and R. Kopelman, 1996, Anal. Chem. 68, 2656. Shortreed, M. R., S. Dourado and R. Kopelman, 1997, Sens. Actuators B Chem. 38-39, 8. Sumner, J. P., J.W. Aylott, E. Monson and R. Kopelman, 2002, Analyst, 127, Advance Article. Suzuki, K., H. Ohzora, K. Tohda, K. Miyazaki, K. Watanabe, H. Inoue and T. Shirai, 1990, Anal. Chim. Acta 237, 155. Tan, W., R. Kopelman, S.L.R. Barker and M.T. Miller, 1999, Anal. Chem. 71, 606A. Tan, W., Z.-Y. Shi, S. Smith, D. Birnbaum and R. Kopelman, 1992, Science 258, 778. 535

Brasuel, Kopelman, Philbert, et al. Tsagkatakis, I., S. Peper and E. Bakker, 2001, Anal. Chem. 73, 315. Uhlmann, D. R., G. Teowee and J. Boulton, 1997, J. Sol-Gel Sci. Technol. 8, 1083. Xu, H., J.W. Aylott, R. Kopelman, T.J. Miller and M.A. Philbert, 2001, Anal. Chem. 73, 4124.

536

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All fights reserved

CHAPTER 17

COLLOIDAL SEMICONDUCTOR QUANTUM DOT CONJUGATES IN BIOSENSING

1 HEDI MATTOUSSI, PH.D., ~M. KENNETH KUNO,

PH.D., 2ELLENR. GOLDMAN,PH.D., 2GEORGEP. ANDERSON, PH.D. AND 2j. MATTHEW MAURO, PH.D.

~Division of Optical Sciences 2Center for Bio/Molecular Science and Engineering U.S. Naval Research Laboratory Washington, DC 20375 USA

This chapter reviews progress in bio-related applications of luminescent colloidal quantum dots (QDs). The material reviewed undoubtedly represents only the prologue of an unfolding story, as quantum dots are a relatively recent discovery and their biological applications are newer still. Nonetheless, a significant body of research literature exists pointing the way toward future advances. We begin with a basic introduction to quantum dots, including their synthesis and some characteristic physical properties, then follow with a review of bio-related work involving semiconductor nanocrystals published todate. Work involving preparation and use of QD-protein conjugates in cellular imaging, quantitative immunoassays, and in early-stage energy transfer applications is reviewed, as well as uses of QD-DNA conjugates as nanoscale building blocks. A listing of early patents in this area is also included for those who contemplate utilizing these materials in the commercial arena. Advantages and limitations in bio-related applications are presented based on the current state-of-the-art in QD technology.

537

Mattoussi, Kuno, Goldman, et al.

1. Principles of Operation 1.1. Introduction Once purely a novelty in the realm of low dimensional semiconductor physics, quantum dots (QDs) have now come of age. The last decade has seen tremendous scientific interest and progress in understanding these semiconductor nanoparticle materials as well as initial attempts to develop and commercialize various applications (Yoffe, 1993, 2001; Alivisatos, 1996; Efros et al., 2000; Murray et al., 2000). Prompting this trend has been a growing realization of the technological importance of understanding the chemistry, physics and biology of materials at the nanometer scale, an area collectively known as nanoscience. The drive for expanding our understanding of semiconductor nanocrystals has also been spearheaded by potential applications for these materials in devices such as absorption filters (Borrelli et al., 1987, Hall et al., 1988), light emitting diodes (Colvin et al., 1994; Dabboussi et al., 1995; Schlamp et al., 1997; Mattoussi et al., 1998a), and photovoltaic cells (Greenham et al., 1996). QD bioconjugates, materials comprised of luminescent colloidal QDs conjugated with biomolecules, can be used in applications such as detection and quantitation of soluble substances, in bioimaging, and potentially, in a range of diagnostics applications. Successful integration of these promising materials into these and other emerging biotechnological areas will necessitate a thorough understanding of the properties of these hybrid bioinorganic systems, requiring multidisciplinary and coordinated efforts in chemistry, physics and materials sciences. Biological tagging using fluorophores is useful in many biotechnological applications, including immunoassays, disease diagnosis, drug development, and cell and tissue imaging in both single and multiplex approaches (Schrock et al., 1996;, Hermanson, 1996). For instance, recent flow cytometry work (Roederer et al., 1997) using a multi-laser excitation system and a multi-color labeling scheme, allowed concurrent observation of ten parameters involving cellular antigens, demonstrating the high level of sophistication possible using dye labels. Furthermore, microarray-based gene analysis using multiple fluorescent probes has become a critical technology in the burgeoning genomics field (Lobenhofer et al., 2001). Virtually all available organic light-emitting dyes, however, have inherent functional limitations such as narrow excitation bands and broad redtailing photoluminescence spectra, low resistance to photodegradation, and the necessity for individually tailoring synthesis and conjugation procedures for each fluorophore. Thus, there remains a need for new and improved types of fluorescent labeling materials. Semiconductor nanocrystals (e.g., CdSe-ZnS core-shell QDs) represent a promising alternative in certain bioanalytical and imaging applications (Bruchez et al., 1998; Gerion et al., 2001). These very bright photoluminescent materials have readily tunable spectral properties, high 538

Quantum Dot Bioconjugates for Biosensing photobleaching thresholds, and biocompatibility. Colloidal QDs made of ZnSe, CdS, CdSe, CdTe, and HgSe emit light over a wide range of wavelengths in the visible and near IR (Hines et al. 1998; Henglein et al., 1982; Weller et al., 1986; Rosetti et al., 1983, 1984; Murray et al., 1993; Rogach et al., 1997; Mikulec, 1999; Eychmialler et al., 2000). In addition, their essentially continuous absorption envelope allows simultaneous excitation of several different colors of QDs with a single wavelength, making them naturally suitable for multiplexing applications. In this chapter, we first describe some basic features and unique properties of colloidal semiconductor QDs, followed by a short history outlining some of the most important early bio-related studies. We then present the current status of known research efforts that involve using luminescent colloidal QD bioconjugates in biosensing and bioimaging.

1.2. Chemistry and physics of semiconductor quantum dots 1.2.1. Description. Colloidal semiconductor quantum dots are small, spherical, crystalline particles of a given material consisting of hundreds to thousands of atoms. They are neither atomic nor bulk semiconductors, but may best be described as artificial atoms. Their properties originate from their physical size, which ranges from 10 to ~ 100 A in radius and is often comparable to or smaller than the bulk Bohr exciton radius (Woggon, 1997; Gaponenko 1998; Yoffe 2001; Efros et al., 2000). As a consequence, QDs no longer exhibit their bulk parent optical or electronic properties. Instead, they exhibit novel electronic properties due to what are commonly referred to as quantum confinement effects. These effects originate from the spatial confinement of intrinsic carriers (electrons and holes) to the physical dimensions of the material rather than to bulk length scales. One of the better-known confinement effects is the increase in semiconductor band gap energy with decreasing particle size; this manifests itself as a sizedependent blue shift of the band edge absorption and luminescence emission with decreasing particle size (Figure 1). This size-dependent absorption can be understood by using a simple analogy to a quantum mechanical particle in a one-dimensional box of length L. In this model, a carrier is localized within a potential minimum between two infinite barriers. Due to this spatial confinement, the energies of the carriers are quantized to discrete values, proportional to the inverse of the square of the length of the box (En 0r n2/L2, with n = 1,2,3 .... ). An extension to QDs is achieved by considering a three dimensional box (or sphere) where the potential minimum represents the QD and the barrier to escape originates from the abrupt termination of the QD at its surface.

539

Mattoussi, Kuno, Goldman, et al.

Figure 1. Photoluminescence spectra for five different core sizes of CdSe-ZnS quantum dots in water solutions. All samples were excited at 350 nm. Core radii were extracted from small angle x-ray scattering (SAXS) data (Mattoussi et al., 1998b).

For a spherical QD with an infinite potential barrier one obtains the following expression for the electron and hole energy levels in the particle: ~12~'~"

(1)

2me,ha Here ~,, is the nth root of the spherical Bessel function of order l, me.h is the effective mass of the electron (e) or hole (h) and a is the radius of the QD. One therefore predicts discrete quantized electron-hole transitions, an increase in effective band gap (or HOMO-LUMO transition in molecular terms) with decreasing particle size and conversely a decrease in spacing between states with increasing size. It should be noted that the Coulomb interaction between the confined electron and hole alters these energies, but since this term scales as 1/a, it is essentially a small perturbation to Equation 1, which varies more strongly with size (i.e., 1/a2). A number of excellent review articles exist on the subject, particularly those by Yoffe (1993, 2001) Gaponenko (1998) and Efros (Efros et al., 2001); the interested reader is referred to them for more information.

1.2.2. Nanocrystal synthesis. QDs were first discovered in doped silicate glasses by Ekimov and Onuschenko (Ekimov et al., 1980, 1981, 1982, 1983, 1984, 1985a,b, 1993, 1996). In their,seminal work, a supersaturated solution of copper and chlorine compounds in glass was heated at high temperatures to cause the controlled precipitation of CuC1. Additional heating of the melt allowed them to 540

Quantum Dot Bioconjugates for Biosensing systematically create collections of small crystalline CuC1 particles ranging in size from tens to hundreds of angstroms, initially denoted as quantum droplets. The particles have since become known as quantum dots, although alternative names exist in the literature, including nanoparticles, nanocrystals, nanocrystallites and Q-dots. Today, a wide variety of methods such as e-beam lithography, x-ray lithography, molecular beam epitaxy (MBE), ion implantation, sonochemistry, and growth in size-restricted environments are available for making small nanocrystallites not only of semiconductors but also of metals. Some of the more common techniques are outlined below (and in Table 1) with particular emphasis on preparations yielding colloidal QDs that have surface capping/passivating molecules (ligands). Using these ligands allows tailoring of QD solubility in a variety of solvents, permits facile solution processing and can, in some cases, make them amenable to biological manipulations. Growth of QDs in glass melts is achieved by doping the melt with salts of the desired material (Ekimov et al., 1980, 1982, 1986; Borrelli et al., 1987). The temperature of the glass is then rapidly dropped to generate small nuclei of the semiconductor. The glass then undergoes a secondary heat treatment over temperatures ranging from 400 to 1000~ to induce the nuclei to grow, forming small spherical crystalline particles of semiconductor dispersed in amorphous glass matrices. Advantages of this technique include h!ghly crystalline particles and the ability of the glass host to support very large (hundreds of angstroms) QDs. A serious disadvantage is that the QDs cannot be easily manipulated after their synthesis. They remain trapped in the glass and there are few possibilities for treating the material once made, for example, to alter surface chemistry or improve their size distribution. In parallel with the discovery of QD growth in glasses, it was found that semiconductor nanoparticles could be grown within inverse micelles (Figure 2) (Henglein, 1982; Rosetti et al. 1983,1984; Kotov et al., 1993,1994; Pileni et al., 1992). This technique exploits natural geometrical structures created by waterin-oil mixtures upon adding an amphiphilic surfactant such as sodium dioctyl sulfosuccinate (AOT). By varying the water content of the mixture, it was shown that the size of the water droplets suspended in the oil phase could be varied systematically. This led to the idea of using these self-enclosed water pools as micro-reactors for carrying out nanoscale sustained chemical reactions. In the case of QDs, it was found that adding metal salts to the water pools could cause nucleation and growth of colloidal nanocrystalline particles. Advantages of this technique include reactions carried out at room temperature and, more importantly, the ability to isolate the QDs after their synthesis. The inverse micelle preparation was therefore a significant advance in the development of QDs, giving researchers access to the surface chemistry of the particles for additional functionalization and manipulation.

541

Mattoussi, Kuno, Goldman, et al. Table 1. Major known types of Colloidal Group II-VI semiconductor QDs, method of preparation and representative literature citation. Some of these materials have been used in bio-related experiments.

Nanocrystal type CdS

i

Preparation meti~od Silica glass

Ekimov et al., 1985; Potter et al., 1988; Liu et al., 1990, Persans et al., 1989; Zhao et al., 1991 Rosetti et al. 1983, 1984, Weller et al., 1986; Misawa et al., 1991; Woggon et al., 1993 Murray et al., 1993; Artmeyev et al., 1995

Aqueous solutions, inverted micelles Polymer and high temperature coordinating solutions Sol-gel glass

Nogami et al., 1990; Minti et al., 1991; Spanhel et al., 1992; Mathieu et al., 1995 ;Wang et al., 1989; Herron et al., 1989 Gurevich et al., 1992

Semiconductor-glass composite films CdSe

Ekimov et al., 1985; Borrelli et al., 1987; Gaponenko et al., 1993 Murray et al., 1993

Silica glass High temperature coordinating solutions Polycrystalline films

CdSeovercoating

CdTe

ZnSe

i

Literature citation

Hodes et al., 1987

ZnSe using hybrid micelle/organometallic ZnS using high temperature coordinating solutions CdS using high temperature coordinating solutions

Kortan et al., 1990

Silicate glass Semiconductor-glass Composite films High temperature coordinating solutions

Potter et al, 1988; Liu et al., 1991 Ochoa et al., 1996

High temperature coordinating solutions

Chestnoy et al., 1986; Hines et al., 1998

Hines et al., 1996; Dabbousi et al., 1997 Peng et al., 1997

Murray et al., 1993; Mikulec et al., 1999

542

Quantum Dot Bioconjugates for Biosensing

/~ = AOT CdCI2 + H2S Water/Isooctane Room Temperature

Figure 2. Growth of CdS quantum nanoparticles in inverse micelles. Other materials such as CdSe and CdTe have also been prepared using the inverse micelles approach.

In the early 1990's it was shown by the Bawendi group (Murray et al., 1993) and confirmed shortly thereafter (Bowen Katatri et al., 1994) that an organometallic synthesis based on pyrolysis of metal-organic precursors could yield CdSe QDs with a size distribution of 8-10% as made, with distributions that could be improved during post-reaction processing to values as small as 5 % (Murray et al., 1995). This preparation followed on the early micelle advances, yielding highly crystalline particles which were significantly improved in terms of their fluorescence quantum yield (QY). Colloidal QDs could now be made with room temperature quantum yields on the order of 5-10% (and low temperature QYs near unity), making fluorescence-based applications of QDs viable for the first time. This technique is widely used to generate QDs used in bio-related applications, and is described here (Figure 3). In general, a solution of dimethylcadmium (CdMez) and trioctylphosphine selenide (TOPSe), diluted in trioctylphosphine (TOP), is rapidly injected into a hot stirring solution of trioctylphosphine oxide (TOPO). The rapid introduction and concomitant temperature drop resulting from adding these reagents result in discrete temporal nucleation of CdSe seeds. After reagent injection, the temperature of the solution is raised to 280-300~ in order to grow the particles. The high temperature growth promotes highly crystalline QD cores. Growth is monitored through UV/visible spectroscopy and when the desired size is reached (as monitored by the peak wavelength of the first absorption feature), the temperature is dropped below 100~ to arrest the growth. A more detailed description of a typical laboratory scale organornetallic preparation of QDs is provided in the three steps described below: 543

Mattoussi, Kuno, Goldman, et al.

TOPO CdMe 2 + TOPSe

300-350 ~

Figure 3. High temperature organometallic growth of colloidal CdSe quantum nanocrystals, as described first by Murray et al. (1993).

1. In a glovebox, under nitrogen, a 1M stock solution of trioctylphosphine selenide (TOPSe) is prepared by adding 7.9 grams of amorphous Se (99.99%) shot to 100 ml of trioctylphosphine (TOP, 90-95%). An injection solution is formulated by adding 170-200 ~1 CdMe2 and 3.5-4 ml 1M TOPSe to ~15 ml of TOP. The reagents are mixed and loaded into a syringe equipped with a large gauge needle for rapid injection. 2. On a Schlenk line, a 100 ml three-neck flask is loaded with -20-30 grams of TOPO (90%) and heated to 150-180~ for 2-3 hours under vacuum while stirring in order to remove water. When dry, the flask is backf'llled with inert gas (typically N/) and the temperature is raised to 300-350~ in preparation for precursor injection. The loaded syringe is removed from the glovebox and its contents quickly injected into the flask. Upon injection, there is a vigorous evolution of gas followed by a rapid color change of the solution to light yellow. The temperature falls to ~ 250~ and an absorption spectrum shows sharp features with the peak of the first transition usually located between 470 and 490 rim.

3. The temperature is raised to 290-300~ to allow for growth and annealing of the QDs. During growth, samples are periodically removed and their UV/visible absorption spectra taken. The peak position of the first absorption feature is noted, as well as the relative width of the transition, which gives a measure of a sample's size distribution. Occasionally there is a decrease in the growth rate accompanied by an increase in the relative size distribution. To overcome this growth bottleneck, the temperature is raised by several degrees. When the peak of the first absorption feature reaches a wavelength maximum indicative of a desired size, the temperature is dropped below 100~ to arrest crystal growth. 544

Quantum Dot Bioconjugates for Biosensing Recently, Peng and coworkers have developed a modified organometallic synthesis that is less dependent on the purity of the TOPO and avoids the use of pyrophoric CdMe2 precursor (Peng et al., 2001a, Qu et al., 2001). In their synthesis, high purity TOPO and controlled amounts of cadmium coordinating ligands, e.g., hexylphosphonic acid (HPA) or tetradecylphosphonic acid (TDPA), are combined in the preparation flask. Cadmium compounds such as cadmium oxide (CdO) or cadmium acetate [Cd(Ac)2] are added at a relatively low temperature (e.g., at 140~ and the mixture heated to generate Cd 2§ ions before addition of TOPSe results in nanocrystal nucleation and growth. This procedure is promising; however, additional work is still needed before reproducible high quality CdSe and CdSe-ZnS QDs can be routinely prepared. To obtain material with low size dispersity, growth of CdSe QDs is often followed by size selective precipitation (Murray et al., 1993). This involves adding a "bad" solvent for the TOP/TOPO-capped nanocrystals, such as methanol, to a preparation of QDs whereupon larger particles in the mixture precipitate first due to preferential Van der Waals interactions. Smaller particles remain in solution until enough MeOH is added to drive most of the QDs out of solution. Use of size-dependent QD precipitation enriches for populations of desired nanocrystal size. Repeated precipitations can reduce the overall size distribution of QD synthetic mixtures to values of ~5%. This method is widely employed for polymers and colloids to reduce polydispersity after synthesis. Size and size distribution measurements are usually carried out using transmission electron microscopy (TEM) (Murray et al., 1993) and/or small angle x-ray scattering (SAXS). TEM tends to provide slightly smaller values for the inorganic core than SAXS because TEM does not take into account the amorphous outermost atomic layer on the nanocrystal surface (Mattoussi et al., 1998b). It was discovered in the mid-1990's that passivating QDs with an additional thin layer made of a wider band gap semiconductor could improve the surface quality of the particles (by providing a better passivation of surface states), resulting in dramatic enhancements of the fluorescence quantum yield. Although the principle was previously known from semiconductor bandgap engineering (Steigerwald et al., 1988; Kortan et al., 1990), the optimal set of conditions for creating strongly fluorescent overcoated QDs was not realized until the seminal work of Hines and Sionnest (Hines et al., 1996), when they showed that overcoating CdSe QDs with ZnS improved quantum yields to values of 30% or greater. This was shortly followed by other studies that described additional characterization of CdSe QDs overcoated with ZnS (Dabbousi et al., 1997) and CdS (Peng et al., 1997). In brief, the procedure for overcoating colloidal QDs with another semiconductor involves the following steps. A dilute solution of QDs is made up in an appropriate coordinating solvent (TOPO, for instance). The temperature of the 545

Mattoussi, Kuno, Goldman, et al.

CdSe QDs

ZnEt 2, (TMS)2S ...... TOPO, 140-180 ~

Figure 4. Overcoating of CdSe quantum dots with ZnS using high temperature solution route.

solution is raised to --150~ but kept lower than 200~ to prevent further growth of the QDs. A dilute solution of Zn (or Cd) and S precursors is then slowly introduced into the hot stirring QD solution. The high dilution and relatively low temperature of the mixture prevent separate nucleation of ZnS or CdS quantum dots. Once the precursors have been added, the temperature is lowered to ~ 80~ and the reaction vessel left undisturbed for several hours. Noticeable improvements in the PL quantum yield are apparent after several hours of heat annealing (Figure 4). A typical laboratory scale ZnS overcoating process for CdSe QDs includes the following steps (Dabbousi et al., 1997): Size-selected CdSe particles dispersed in hexane are added to 5-10 grams of dried, degassed TOPO at -70~ Inside the glovebox, equimolar amounts of diethylzinc (or CdMe2) and hexamethydisilathiane are mixed with -5 mls of TOP. The amount of Zn and S precursors added varies depending on the size of the CdSe QD and is calculated to yield a 2-3 (or more) atomic monolayer coverage on the particle surface. Once this solution has been prepared, the temperature of the QD/TOPO solution is raised to a value between 140~ and 180~ The Zn (Cd) and S precursor solution is then brought out of the glovebox and introduced at a rate of ~ 0.5 ml/min through a separate addition funnel attached to the flask holding the QD/TOPO mixture. Once the addition is complete, the pot temperature is lowered to ~80~ and the mixture is left undisturbed for several hours. The overcoated QDs are subsequently precipitated with methanol prior to further processing.

546

Quantmn Dot Bioconjugates for Biosensing

1.3. Semiconductor nanocrystal properties Initial optical studies of QDs in the late 1980's aimed at correlating sizedependent spectral shifts in absorption with quantum confinement effects. Today, the absorption properties of CdSe QDs are relatively well understood with up to ten excited states in the absorption assigned and theoretical avoided crossings observed (Norris et al., 1996). However, the origin of the band edge emission in CdSe QDs was not immediately understood due to the inability of the same theory to explain unusual size-dependent features in QD fluorescence spectra. This presented a serious challenge to the above model, and it was not until the mid 1990' s, when Chamarro (Chamarro et al., 1995, 1996) realized the importance of electron hole exchange interaction in QD materials that a more comprehensive understanding of CdSe QDs was achieved. Modifications by Efros to that theory led to understanding of previously unrecognized "dark exciton" effects, which explained many unusual features in the emission spectra such as the size-dependent ~ and "global" Stokes shift observed in fluorescence line narrowing and global excitation experiments (Nirmal et al., 1995; Efros et al., 1996; Kuno et al., 1997). In 1996 Nirmal et al. conducted the first single-particle fluorescence studies of isolated TOP/TOPO-capped CdSe and CdSe-ZnS QDs. They discovered that QDs underwent intermittent on/off emission (so-called "blinking") under continuous excitation (Nirmal et al., 1996). Unlike single fluorescent molecules, this behavior could not be attributed to a commonly known effect referred to as quantum jumps (Cook and Kimble, 1985). Instead, the on/off intermittency in QD emission was attributed to Auger ionization of the QD (Chepic et al., 1990; Nirmal et al., 1996; Efros et al., 1997). The blinking effect is still not fully understood, however.

2. History of Bio-Related Applications using QD Bioconjugates The amount of published research involving bio-related uses of semiconductor nanocrystals has expanded rapidly since the initial reports from the laboratories of Alivasatos (Bruchez et al., 1998) and Nie (Chan and Nie, 1998) first appeared. This section begins with a description of early work on the preparation of protein-derivatized water-compatible quantum dots, preliminary QD-protein conjugate characterization and their use in imaging cellular structures. It is followed by description of a simple and useful electrostatically controlled conjugation method developed in our laboratory (Mattoussi et al., 2000). We conclude by describing briefly the original work involving formation and use of QD-DNA conjugates as performed in the Mirkin laboratory (Mitchell et al., 1999).

547

Mattoussi, Kuno, Goldman, et al.

Figure 5. Surface treatment of overcoated quantum dots. A) Silica shell with reactive amine functions; B) Mercaptoacetic acid coverage; C) Capping with dihydrolipoic acid [DHLA]. D) Direct binding of thiol-terminated DNA.

2.1. Preparation and use of QD bioconjugates in cellular imaging Water-soluble nanocrystals derivatized with the actin-binding protein, phalloidin, were used for the first attempts at intracellular imaging in fixed mouse fibroblasts (Bruchez et al., 1998). For conjugating phalloidin with nanocrystal surfaces, the authors started with CdSe-CdS core-shell QDs that were capped with a thin amine-derivatized silica shell to render them both reactive and water-compatible (Figure 5A). The silica-encapsulated, red-emitting CdSe-CdS core-shell particles (4 nm diameter core size and emission maximum at 630 nm) were subsequently biotinylated at their exposed reactive amine sites. Streptavidin was then used effectively as a bridge between actin-bound biotinylated phalloidin and the biotinylated red-emitting QDs in cell imaging work. Simultaneous nuclear staining was achieved using green-emitting QDs coated with an anionic silane, in a process driven by electrostatic and hydrogen binding interactions. Light from a mercury lamp with a 488 nm excitation filter and a single long-pass emission filter were used to image both red (actin-bound) and green (nucleus-bound) QDs at the same time. Importantly, it was also demonstrated with sequential scans using laser scanning confocal microscopy that the red nanocrystal bioconjugates were dramatically more photostable than fluorescein-labeled phalloidin bound to actin fibers under essentially identical conditions. A detailed description of the process used for preparation of silica encapsulated reactive QDs, together with more extensive characterization of their properties, has since been published (Gerion et al., 2001). 548

Quantum Dot Bioconjugates for Biosensing

Figure 6. Schematic of engineered two-domain protein electrostatically complexed with a DHLA-capped CdSe-ZnS quantum dot (based on Mattoussi et al., 2000).

A contemporaneous imaging study employed bio-compatible nanocrystals conjugated with human transferrin to conduct intracellular staining of fixed HeLa cells (Chan and Nie, 1998). For conjugating transferrin with nanocrystal surfaces, the authors started with CdSe-ZnS core-shell dots that had been capped by ligand-exchange with mercaptoacetic acid in order to render them both watercompatible and reactive (Figure 5B). Mercaptoacid-capped red-emitting CdSeZnS core-shell particles (2.1 nm radius and emission maximum of 560 nm) were subsequently coupled with transferrin using 1-ethyl-3-(3-dimethylaminopropyl carbodiimide) hydrochloride (EDC)-promoted condensation chemistry. The conjugates preserved the absorption and photoluminescence properties of the nanocrystals as well as the activity of the transferrin. For QD-transferrin conjugates, functionality was demonstrated by exposing the conjugates to cultured HeLa cells overnight followed by extensive washing and imaging using an epifluorescence microscope equipped with a CCD camera. Imaging revealed that luminescent QDs had entered the cells, presumably via receptor-mediated endocytosis. In the absence of transferrin, i.e., using capped 2.1 nm core radius QDs without a protein coating, only weak cellular autofluorescence was observed. These authors also prepared human IgG-QD conjugates using the same methods. Reaction with human IgG-specific polyclonal antibody resulted in extensive aggregation as observed using fluorescence imaging, undoubtedly due to inter-dot bridging among rnultiply-derivatized QDs.

2.2. Preparation of QD-protein conjugates using a non-covalent strategy An alternative method for preparing functional QD-protein conjugates has been developed in our laboratory at NRL. This system employs "bidentate" dihydrolipoic acid (DHLA) groups to coat QDs (Figure 5C) in combination with two-domain proteins engineered to interact electrostatically with negatively 549

Mattoussi, Kuno, Goldman, et al. charged QD surfaces. In a model system that employed engineered variants of E. coli maltose binding protein (MBP) and DHLA-capped CdSe-ZnS nanocrystals, stable self-assembled QD-protein complexes formed in an efficient and controlled manner (Mattoussi et al., 2000). In order to promote the self-assembly process, a modular MBP-basic leucine zipper chimeric protein was designed and prepared in a recombinant system (MBP-zb) (Figure 6). The strongly positivelycharged C-terminal "tail" present in the novel MBP-zb variant resulted in rapid formation of QD-MBP conjugates that retained both the optical QD properties and the active folded state of the MBP protein. Based on incremental increases in fluorescence quantum yield that occurred upon titration of QDs with increasing amounts of MBP-zb, and from steric considerations, it was estimated that for a nanoparticle of-- 70 A, (core-plus-shell) diameter an average of 12-15 engineered protein molecules could be packed around each nanocrystal (Mattoussi et al., 2001). That number can presumably increase or decrease for larger or smaller QDs, respectively. In the first known use of QD-bioconjugates in a quantitative fluoroassay, a functional assay for maltose was developed that monitored displacement of QDMPB-zb conjugates bound to a cross-linked amylose affinity matrix as various amounts of dissolved maltose were injected into the flowing buffer upstream of the affinity column. Maltose concentrations of injected samples were determined by integrating PL intensity (QD emission) versus time as QD-MBP-zb conjugates were displaced from the column by maltose; limits of detection were on the order of 10 pmol of maltose (Mattoussi et al., 2001). In order to prepare reagents for use in fluoroimmunoassays, an analogous recombinant construct was developed based on the IgG-binding domain of protein G from Staphylococcus. In this construct, the engineered dimeric form was critical in providing two points of attachment for each IgG. The PG-zb protein serves as a very effective bridge between the DHLA quantum dot surface and any type of IgG antibody, resulting in reagents that can be used in general fluorimmunoassay protocols. QD-antibody complexes made with this strategy were used in fluoroimmunoassays in analysis of the protein toxin staphylococcal enterotoxin B (SEB) (Tran et al., 2001; Goldman et al., 2001, 2002a) and in analysis of low levels of 2,4,6-trinitrotoluene (TNT) and hexahydro-l,3,5trinitro-l,3,5-triazine (RDX) dissolved in water (Goldman et al., 2002a,b). In the sandwich-assay format for SEB, the limit of detection was about 200 pmol of the protein toxin. A competition assay performed in microtiter plates for TNT and RDX allowed facile quantitation of the dissolved explosives, with detection limits of 1 ng and 0.2 ng of the explosives, respectively (Goldman et al., 2002a).

2.3. QD-DNA conjugates Sequence-dependent hybridization of deoxyoligonucleotides bound to CdSe-ZnS QDs was first demonstrated in the Mirkin laboratory (Mitchell et al., 1999). Two 550

Quantum Dot Bioconjugates for Biosensing populations of nanoparticles were prepared using either 3' or 5' alkyl thiolterminated 22-mer oligonucleotides for QD surface attachment (Figure 5D). Upon addition of a 24-mer "bridging" or "capture" oligonucleotide designed to be complementary to the outer 12-met sequence of both types of DNA-modified nanoparticle, a cross-linked network of specifically hybridized particles was formed. Aggregation was demonstrated to be specific by control experiments with non-complementary capture DNA in which no evidence of multi-center cluster formation was observed. Cluster formation by specific hybridization resulted in a decrease in integrated fluorescence intensity of 26 + 6%; excimer formation between the DNA linked dots was cited as a possible explanation. These researchers also broke new ground by forming and characterizing mixed aggregates composed of both QD-DNA and Au-DNA conjugates. Temperatureinduced melting of the hybridized DNA present within both QD-oligonucleotide complexes and mixed QD-Au nanoparticle-DNA complexes was studied by monitoring light absorption versus temperature. A sharp duplex melting temperature transition was observed, suggesting that cooperativity effects operate within the complexes due to multiple DNA links per particle. Melting of mixed QD-Au-DNA particles ("A-B" structures) could be observed at one tenth the concentration of a pure QD-DNA system due to the very large molar extinction coefficient associated with the plasmon band of the DNA-derivitized 13 nm Au particles used (2.8 x 108 M 1 crnl). Construction of more complex multicomponent nanostructured materials might be possible using these types of building blocks.

2.4. Seminal patents Although applications of quantum dots to bio-related issues is a relatively new area, several U.S. patents have already been issued for processes involving synthesis and use of quantum dots within the biochemical realm. Table 2 presents eight seminal patents in the area. All have been issued within the last two years. A wide range of areas is covered in these patents; interested readers should consult the U.S. Patent and Trademark Office Home Page (http:!/www.uspto.gov/) for viewing the complete patents.

3. State of the Art 3.1. Types of QDs and their optical properties Colloidal semiconductor nanocrystal quantum dots are in general considered to be spherical in shape. They can be dispersed in a solid matrix such as those prepared by annealing at high temperature in glasses (referred to as quantum droplets), where the nanocrystal growth is driven by precipitation/nucleation, or 551

Mattoussi, Kuno, Goldman, et al. Table 2. Seminal patents describing applications involving the use of QD-conjugates.

Patent Number

Issue Date

Patent Title

Oct.

6,309,701

2001

6,207,392

2001

6,319,607 6,306,610

2001 Oct. 2001

Fluorescent nanocrystal-labeled microspheres for Fluorescence analysis Semiconductor nanocrystal probes for biological applications and process for making and using such probes Purification of functionalized fluorescent nanocrystals Biological applications of quantum dots

6,274,323

Aug. 2001

Mar.

NOV.

Nov.

6,319,426

2001 Nov.

5,990,479

1999

6,114,038

Sept. 2000

Authors/Assignees

E. Barbera-Guillem/ Bio-Pixels, Ltd. S. Weiss, M. Bruchez, P. Alivisatos/ Univ. of California

E. Barbera-Guillem/ Bio-Pixels, Ltd. M. Bawendi, F. Mikulec, V. Sundar/ MIT Method of detecting an analyte M. Bruchez, R. Daniels, S. in a sample using semiconductor nanocrystals as a Empedocles, V. Phillips, E. Wong,, D. detectable label Zehnder/ Quantum Dot Corp. M. Bawendi, F. Water-soluble semiconductor Mikulek, J. Lee/ nanocrystals MIT S. Weiss, M. Bruchez, Organo Luminescent P. Alivisatos/ semiconductor nanocrystal Univ. of California probes for biological applications and process for making and using such probes Functionalized nanocrystals and S. Castro and E. Barbera-Guillem/ their use in detection systems BioCrystal Ltd.

surface functionalized with organic ligands to make them soluble in a variety of organic solvents to make colloidal dispersions. The latter are usually grown using inverted micelles or high temperature solution chemistry routes. Recently, preparation of anisotropic semiconductor nanocrystals made of CdSe (quantum rods) using the organometallic synthesis route was reported. Control of the type of cadmium ligands used (e.g., HPA or TDPA) and the concentration of the cadmium complexes formed are the key elements to growing anisotropic

552

Quantum Dot Bioconjugates for Biosensing nanocrystals (Peng et al., 2000, 2001b.). These quantum rods were reported to have linearly polarized luminescence emission, a property that may have a potential use in biological tagging applications (Hu et al. 2001). The common optical property that characterizes QDs is the size-dependence of their spectroscopic properties (e.g., absorption and phohtoluminescence), which results from quantum confinement of charge carriers within a volume smaller or comparable to the Bohr exciton radius. Within the family of II-VI compounds the range of absorption and emission peaks depends on the materials; ZnS and ZnSe QDs have absorption and emission spectra limited to the UV and blue regions, whereas nanocrystals made of heavier atoms such CdTe or HgSe or hybrids composed of PbSe (III-VI compounds, Murray et al. 2001b) have useful optical properties that extend into the near IR region of the spectrum. For biological labeling, colloidal CdSe-ZnS QDs have been the most widely used in published studies. Experiments using CdS QDs, CdS-overcoated CdSe nanocrystals or CdTe QDs have also been described in a few instances (Mahtab et al., 1996, Bruchez et al., 1998, Mamedova et al., 2001).

3.2. Nanocrystal surface treatment

3.2.1. Inorganic overcoating. For colloidal nanocrystals, organic ligands tend to provide only partial surface passivation, which translates into rather modest photoluminescence yields. Overcoating the native core with a wider band gap semiconducting material provides additional surface passivation and reduces leakage of excitons outside the core. Optimal passivation of the surface states occurs when the growth of the overcoating layer is near-epitaxial, i.e., the lattice mismatch between the core and the shell material is as small as possible. On the other hand, using an overcoating material that has a closely matching lattice structure in order to promote epitaxial growth produces a leakage of the exciton to the overcoating layer, which in turn results in a pronounced red shift of absorption and emission spectra in comparison with the native nanocrystals. This pronounced red shift has an additional disadvantage as it moves the range of useful wavelengths further into the red. For example, when using a CdSe core, overcoating with CdS produces a higher PL yield than the one measured with ZnS. However, minimal red shift of the band edge absorption and emission are measured for CdSe-ZnS QDs compared with CdS-overcoated nanocrystals. Thus, only ZnS-overcoated QDs allow a broad range of wavelengths in the visible to be produced including emission in the blue region of the optical spectrum. 3.2.2. Capping ligandsfor bio-compatibility and conjugateformation. Protecting the nanocrystals' physical and chemical integrity in aqueous media, while simultaneously providing sufficiently reactive surface sites to allow bioconjugation, has been challenging. Several means of accomplishing this have been devised, some of which have already been alluded to. Capping with 553

Mattoussi, Kuno, Goldman, et al. mercaptoacids imparts water solubility and provides carboxyl groups for the condensation chemistry necessary for further covalent modification (Chan and Nie, 1998) (Figure 5B). It has been suggested that the "bidentate" type of interaction of dihydrolipoic acid (DHLA) (Figure 5C) (Mattoussi et al., 2000) or dithiothreitol (Pathak et al., 2001) with the inorganic QD surface results in more water-stable nanocrystals, but no systematic study has been performed. The charged surface provided by DHLA-capped CdSe-ZnS QDs drives electrostatic self-assembly of QD-protein conjugates that, once formed, are surprisingly stable even in high salt solutions (Mattoussi et al., 2000). Porous silica shells have been used for passivation and placement of hydroxyl, amino, thiol, and phosphonate groups in position for bioconjugate formation (Gerion et al., 2001) (Figure 5A). Attachment of thiolated DNA directly to QDs by replacement of surface mercaptoacids has been successfully accomplished (Figure 5D) (Mitchell et al., 1999), and similarly, CdS nanoparticles have been grown in the presence of thiolated peptides, resulting in derivatized fluorescent particles able to recognize cellular receptors (Winter et al., 2001).

3.3. Bioassay work involving quantum dots Quantum dot bioconjugates can function as fluorescent reporters in recognitionbased quantitative and qualitative bioassays. QDs conjugated to both proteins and DNA have been used in a limited number of applications, as described below.

3.3.1. Fluoroimmunoassays using QD-antibody conjugates. Immunoassays utilizing CdSe-ZnS QD-conjugates formed by electrostatically driven selfassembly have been developed (Goldman et al., 2001, 2002a, b; Tran et al., 2001). QD-antibody complexes for use in bioassays have been formed using adaptor proteins as bridges to link QDs with antibodies. Either naturally occurring protein bridges (e.g., avidin) or engineered recombinant protein bridges can be used in this capacity. In practice, mixed-surface QD conjugates have been made with both the antibody-binding adaptor protein and an engineered maltose binding protein derivative (MBP-zb) bound to their surface (Figure 7A). The mixed recognition elements on the particles allow separation of QD-antibody complexes from unbound antibody using affinity chromatography. After saturation of antibody binding sites with IgG (or biotinylated IgG when using the avidin bridge) and purification on a cross-linked amylose column to remove excess unbound IgG, various QD-antibody conjugates have been demonstrated to bind antigen directly adsorbed to microtiter plates. Sandwich- or competitivetype assays were then performed in model systems for analysis of staphylococcal enterotoxin B (SEB) and for detecting low levels of the explosives TNT and RDX dissolved in water (Figure 7B and Figure 8).

554

Quantum Dot Bioconjugates for Biosensing

Figure 7. QD antibody conjugates prepared using molecular bridges. A. Mixed surface conjugate after purification by cross-linked amylose affinity chromatography. B. Schematic of competitive assay for the explosive RDX dissolved in water (Goldman et al., 2002a,b).

140 120 .,..,

el

"" 100

~

~

80

~

60

E

o L_

10 20 )

9 '

O. 1

i

i,i

i i l l l

i

1

i

i , , | , 1 1

i

10

i , I

, , I I I L ,

1 O0

i

,

L

i

i

!

i/,

1000

Concentration of RDX (ng/ml)

Figure 8. Results of a quantitative competitive assay for RDX using QD-anti-R.DX antibody conjugates, where inhibition of QD-conjugates to BSA-RDX by free RDX was investigated. 100% indicates signal in the absence of RDX. 555

Mattoussi, Kuno, Goldman, et al. Laser scanning confocal microscopy has also been used as the means of detection in sandwich immunoassays utilizing QD-antibody conjugates (Sun et al., 2001). In these experiments, rabbit anti-human antibody was covalently conjugated to CdSe-ZnS core-shell QDs using EDC chemistry; unconjugated antibody was removed using repeated cycles of washing and centrifugation. Rabbit antihuman IgG was immobilized on a glass slide to form the capture layer, followed by exposure to a mixture of human IgG and goat IgG. After incubation and rinsing unbound antibodies, the surface was exposed to a solution of QDs conjugated with rabbit-anti-human antibodies. Reading the fluorescence signal of the glass slide showed a linear increase of the signal with human IgG concentration, indicating that the QD-conjugates specifically recognized immobilized human IgG. IgG samples containing about 2 ~tg/ml antibody could be detected. Comparison against antibody labeled with a conventional fluorophore (fluorescein isothiocyanate) in the same system gave a limit of detection of about 25 ng/ml. Inappropriate selection of the excitation wavelength with respect to the absorption and emission of the nanocrystal employed in the assay may have contributed to reduced assay sensitivity with the QD conjugates. CdS QDs prepared with amino-derivatized polysaccharides (aminodextrans, Amdex) have been conjugated with mouse monoclonal anti-T4 antibody for use in flow cytometry (Sondi et al., 2000). The QD-Amdex-antibody complexes were purified from unconjugated antibody by size exclusion chromatography. Samples of whole blood were lysed and the anti-T4 antibody-Amdex-QD reagent tested for the detection of T4 positive lymphocytes. For unknown reasons, direct cytometric detection using QD fluorescence was not achieved. Signal was seen, however, when sheep-anti-mouse antibody conjugated to phycoerythrin was added to the sample containing anti-T4 antibody-Amdex-QDs, suggesting that the purified QD conjugates were effectively binding to receptor sites on the lymphocytes. 3.3.2. DNA-based systems using QDs as reporters. QDs have been used as probes of DNA structure (Mahtab et al., 1996; Gearheart et al., 2001). Luminescence from CdS QDs was monitored as DNA with different sequences (implying different local structures) was adsorbed to QD surfaces. Adsorption of increasing amounts of DNA to QD surfaces resulted in systematic decreases in emission intensity. DNA sequences with intrinsically kinked structures were found to bind preferentially to nanoparticles rich in surface Cd z§ sites as well as to neutral, mercaptoethanol-capped CdS QDs. DNA molecules lacking [-finks, however, bound to the CdZ§ surface of CdS QDs and not to mercaptoethanolcapped QDs. DNA methylation affected the interaction of Cd2+-rich surfaces of CdS QDs with DNA. Addition of DNA containing hypermethylated triplet repeats d(mCmCG)7 to QDs produced less quenching of the particle luminescence than equal amounts of d(mCGG)7 DNA, even though the binding constants for the two types of DNA were essentially identical.

556

Quantum Dot Bioconjugates for Biosensing In another demonstration of the use of QD-DNA conjugates, DNA conjugated to CdSe-ZnS QDs have been used for the detection of specific chromosome sequences in human sperm cells using fluorescence in situ hybridization (FISH) (Pathak et al., 2000). Prior to DNA conjugation, the nanoparticles were treated with dithiothreitol, presumably resulting in stable "bidentate" interactions of each capping molecule with the ZnS overcoating. Surface hydroxyl groups made available by this process were then activated with carbonyl diimidazole (CDI) and reacted with 5'aminated oligonucleotides to give stable carbamate linkages. QD-oligonucleotide conjugates with sequences specific for the Y-chromosome were used in FISH assays on human sperm cells. Half the cells are expected to contain a Y chromosome and hybridize to the probe, while the other half should contain the X chromosome and should not hybridize with the probe. As predicted, about half the cells showed the fluorescent signal. Only background emission (less than 5% of positive signal) was seen when identical experiments were performed using unconjugated QDs or QDs conjugated to an oligonucleotide having a non-relevant sequence, i.e., a sequence not present in the human genome.

3.4. Bio-imaging applications using quantum dots The two pioneering efforts described in the History section of this review have demonstrated in a "proof-of-principle" mode the utility of using quantum dot bioconjugates as histochemical imaging reagents (Bruchez et al., 1998; Chan and Nie, 1998). In a logical extension of initial cell imaging work, specific sites on the surface of living cultured neuronal cells have been labeled with CdS quantum dots derivatized with two types of recognition molecules (Winter, et al., 2001). In the first case, primary antibodies directed toward the ~ subunit of the tx~l]~integrin that studs the surface of SK-N-SH neuroblastoma cells were treated with secondary antibody-QD complexes. Specific attachment of QDs was verified by bright field and fluorescence optical microscopy. In an effort to reduce the distance between QDs and the cell body, CdS dots were prepared in the presence of synthetic peptide CGGGRGDS, which includes the RGD (Arg-Gly-Asp) sequence known to bind to ~13~ and ~[33 integrins as well as a terminal cysteine residue for interaction with exposed surface atoms of the nanocrystal. Preparation of these peptide-derivatized nanoparticles was performed by singlestep arrested precipitation in the presence of 1" 10 peptide:mercaptoacetic acid; fluorescence anisotropy studies of the prepared particles strongly suggested that peptide attachment had in fact occurred. Microscopy of the cells following exposure to the RGD dots showed a yellow/orange layer of CdS dots coating the blue autofluorescent cells. The control with a non-binding peptide sequence was negative for the staining. Although the thrust of this work involved demonstration of specific binding to living cells, long-range goals involve 557

Mattoussi, Kuno, Goldman, et al. preparation of nerve cell-semiconductor interfaces for use in future nanodevices and sensors. Time-gated imaging of mouse fibroblast cells has been accomplished using silanized 1.8 nm radius (575 nm peak emission) CdSe-ZnS quantum dots (Dahan et al., 2001). In preliminary experiments, the normalized fluorescence decay for the particles used was fit with a triple exponential with components of 3.4, 16.1, and 35.6 ns, corresponding to 1, 50, and 48 percent, respectively, of the emitted photons. Mouse 3T3 fibroblasts incubated overnight with 1 0 - 100 nM nanocrystals were washed and fixed prior to image collection using a custom built stage-scanning time-correlated single-photon counting confocal microscope. Images constructed using all detected photons (i.e., no time gating) showed dense non-specific autofluorescence throughout the cells, while images constructed with photons from the 35-65 ns window exhibited very low backgrounds with isolated bright clusters of QDs. These aggregated QDs were possibly taken up by the growing cells via endocytosis and stored in lysozomes. Time gating resulted in a 15-fold enhancement in signal-to-noise over the ungated data. Enhanced image constrast will likely be crucial to observing single-nanocrystals in cellular environments.

3.5. Energy transfer and quenching studies Solution phase protein-ligand binding models have been studied using CdSe-ZnS QDs conjugated to proteins by monitoring fluorescence resonance energy transfer (FRET) between a QD energy donor conjugated to one binding partner and an organic acceptor dye attached to the other binding partner (Willard et al., 2001). Biotinylated bovine serum albumin (bBSA) was conjugated to the QDs by introduction of free sulfhydryl groups onto the protein; the new thiol groups served as attachment points for the protein to the QD surface. Unconjugated bBSA was removed by filtration through a 100 kD cutoff spin column. Introduction of increasing concentrations of streptavidin labeled with tetramethylrhodamine (TMR) to purified QD-bBSA conjugates resulted in decreases in the QD fluorescence emission. A concurrent increase in the TMR emission was observed, suggesting that energy transfer was occurring. Energy transfer between a fluorescent protein donor (BSA) emitting in the near UV and CdTe QDs was studied in aqueous solutions of QD-BSA conjugates (Mamedova et al., 2001). BSA emission is centered at 340 nm and originates from naturally occuring tryptophan residues. CdTe nanocrystals, prepared using arrested precipitation, were capped with L-cysteine and conjugated with BSA using a one-step linkage procedure employing glutaric dialdehyde, which forms a bridge between the amino groups on the L-cysteine and the lysine moieties on the BSA. The coupling procedure preferentially yielded a 1:1 stoichiometery BSAQD conjugates with a small fraction of 2:1 (BSA:QD) complexes. When the BSA-CdTe conjugate solutions were excited at 290 nm, where both protein and 558

Quantum Dot Bioconjugates for Biosensing QD absorb, emission from the BSA was completely quenched, while that of the QDs showed a two-fold increase with respect to unconjugated nanocrystals. In controls using unconjugated QDs (i.e., unlinked with protein) BSA emission persisted, but only ~ 1/3 of its value measured in the absence of QDs, while QD emission was unaffected. The decrease in the BSA emission was attributed to changes in the pH and presence of Cd 2§ ions, which alters the luminescence efficiency of tryptophan residues. Excitation at longer wavelengths, where BSA absorption is negligible, showed only emission from the nanocrystals. The authors attributed the above observation to resonance energy transfer from BSA to CdTe QD in the QD-BSA complexes. Solution-phase fluorescence quenching assays have been carried out using a dyelabeled variant of the two-domain maltose binding protein (MBP), MBP-tb A75C, bound to DHLA-capped CdSe-ZnS quantum dots. The protein variant, which contained a single cysteine, was specifically labeled at that residue with the nonemitting quencher dye QSY-7 (Tran et al., 2002). Conjugation of MBP-tb and DHLA-capped QDs was driven by electrostatic self-assembly. QSY-7 was chosen as a quenching chromophore due to the overlap between its absorption spectrum and the emission spectrum of the QDs employed (core radius of 21.5,~ and emission maximum centered at 590 nm). Increased quenching of the nanocrystal emission was observed with increasing amounts of quencher-labeled protein bound; the nanocrystals lost about 90% of their signal when 60% of the MBP-tb QD-bound molecules were QSY-7 labeled (Figure 9). When all the conjugated proteins (ca. 5 per QD) were labeled with QSY-7, the nanocrystal emission was nearly fully quenched. These results were attributed to radiationless energy transfer occurring between the QDs and bound MBP-tb/QSY -7. In a second system, the emission from surface-tethered QDs was monitored as QSY-7-1abeled antibodies were bound to immobilized nanocrystals using an antibody-specific molecular adaptor PG-zb (Tran et al., 2002). DHLA-capped QDs were first immobilized on poly-L-lysine-treated glass slides, followed by incubation with PG-zb to form QD-PG-zb conjugates. The surface-bound QDPG-zb complex has a high selectivity for the Fr region of antibodies, and introduction of the QSY-7-1abeled antibodies onto these surfaces resulted in formation of labeled QD-PG-zb-IgG conjugates. Fluorescence quenching occurred systematically as the proportion of QSY-7 labeled IgGs conjugated to surface-bound QDs increased, until all available QD surfaces had been saturated with labeled antibody (Tran et al., 2002). 3.6. QDs in polymerized microspheres for use as micro-barcodes

B io-related applications of microsphere-encapsulated quantum dots have recently been investigated (Han et al., 2001). CdSe-ZnS nanocrystals were embedded in cross-linked polymeric beads (1-2 ~tm diameter size) formed by emulsion 559

Mattoussi, Kuno, Goldman, et al.

6000 5OOO

/5

4000

3000 E

E Q Q t...n

2000 1000 0 500

__

525

550

_

575

600

625

_ - U ~ - - " ~ ". . . . . .

650

Wavolen.qth (nm) Figure 9. PL spectra of solutions in a 2 mm optical path cell containing 30 pmol QDs and increasing molar ratio of QSY-7-1abeled MBP-tb to QD. The total numbers MPB-tb proteins per QD remained fixed (protein to QD ratio of 5). The highest PL intensity was measured in solution containing unlabeled QD-MPB-tb conjugates, and the signal decreased systematically with increasing fraction of labeled proteins, with QSY-7-MBPtb varying from 0 (unlabeled proteins) to 5 (100% QSY-7-1abeled proteins) in the above data.

polymerization of styrene, divinylbenzene and acrylic acid. The process of embedding the nanocrystals involved swelling the beads in a solution of chloroform and butanol in the presence of one or more populations of nanocrystals. Nanoparticles that had migrated into the swelled spheres were trapped upon removal of solvent. The relatively large size of the polymer spheres allowed embedding of a large number of QDs in each bead. Using a single dot size (thus one color), a range of intensity-coded beads could be prepared. Embedding two, three, or more populations of QDs per bead allowed control of the emission intensity and detailed spectral characteristics of the QDbeads, resulting in sets of "color and intensity barcode" polymer beads. A model DNA hybridization system using oligonucleotide probes conjugated to the QD-encoded beads was designed and tested in biological assays to detect target DNA sequences. Target DNA molecules to be quantified were labeled with a fluorescent dye (Cascade Blue). The dye was chosen to have an absorption band that allows its excitation simultaneously with the codes within the QD-beads, while having emission distinguishable from the coding signal (i.e., no overlap between the dye and bead emissions). Assays were performed at the single bead level, yielding both the DNA identity (based on the QD barcode) as 560

Quantum Dot Bioconjugates for Biosensing well as its abundance (based on Cascade Blue emission). The DNA sequence was identified by the coding signal (i.e., optical code of the bead defined by the spectral definition of the emission and the relative intensity of each embedded QD color). Signal from the dye attached to the DNA target accounted for the amount of the target material present in the assayed sample.

3.7. Biomaterials applications of QDs DNA-mediated assembly of CdS quantum dots into carefully designed layered arrays attached to electrode surfaces has been demonstrated, followed by generation of photoinduced current by the layered arrays in the presence of a sacrificial electron donor (Willner et al., 2001). Microemulsion-grown CdS nanoparticles capped with cystamine/thioethanesulfonate were derivatized with an estimated 20-24 thiolated oligonucleotides per nanoparticle. These DNAmodified QDs were tethered by hybridization to a gold electrode surface previously derivatized with thiolated 13-mer capture DNA, forming a first layer of DNA-QDs. Subsequent layers (up to four total) were built up by hybridization using additional oligo-DNA modified nanoparticles; the layering process could be observed using quartz crystal microgravimetry, absorbance and fluorescence methods. Current flow was observed upon illumination, and photocurrent amplitude correlated closely with absorbance spectra of the arrays. The photocurrent could be switched on and off by cycled illumination. Photocurrent generation likely involved the photoejection of conduction-band electrons of CdS particles in contact with or within tunneling distance of the electrode. Improved photocurrent generation (approximately 2-fold) could be obtained by treating arrays with 5 x 10.6 M [Ru(NI--I3)6]3+, which presumably electrostatically interacts with the DNA network. Multiple bound ruthenium complexes may form a conduit for delivery of electrons to the electrode. Finally, a unique assay for DNA was shown, in which specific hybridization of an oligonucleotide (109 M lower limit) was detected by changes in photocurrent in a two-layer CdS-DNA cross-linked array.

4. Advantages and Limitations Colloidal semiconductor quantum dots have a number of important advantages over conventional organic fluorophores. The QD absorption spectrum extends into the UV, regardless of size, making it possible to excite multiple sized (colored) particles with one excitation wavelength; organic fluorophores, on the other hand, often have discrete, widely spaced, singlet transitions. In addition, QDs have large extinction coefficients, which translate to absorption crosssections on the order of 1015 cm2. By contrast, many organic dyes have absorption cross-sections nearly an order of magnitude smaller than 1016 cm2. These properties, as well as the color variation made possible by simply varying the physical size of the particle (as opposed to synthesizing new analogs or 561

Mattoussi, Kuno, Goldman, et al. derivatives of conventional organic dyes) represent major advantages of QDs, especially in light of potential multiplexing applications. With respect to their emission, the quantum yields (QYs) of QDs can be comparable to those of organic dyes with values close to 30%. Although they are not as bright as some of the best organic laser dyes such as Rhodamine 6G (QY ~ 95%), they outperform organic fluorophores in two ways. First, the colloidal QDs have narrow (~30-40 nm full width at half maximum) and symmetric emission spectra. Organic dyes, on the other hand, often have broad, asymmetric spectra with a distinct phonon progression to the red. This is a limiting factor in the case of multiplexing applications due to undesirable spectral cross talk among different detector channels. Second, organic dyes suffer from rapid irreversible photodegradation effects, a process often referred to as "photobleaching". While the causes of this effect are not completely understood, photooxidation and other types of degradative photochemistry effectively destroy dye molecules or quench their emission (Eggeling et al., 1999). Semiconductor QDs are much less susceptible to photobleaching since they are made of inorganic materials. This dramatic difference in photostability has been observed at the ensemble level and at the single particle level, where a single CdSe QD has been observed to emit 108 photons with no evidence of photobleaching. That value decreases to ~ 105106 photons for organic dyes (Kuno et al., 2001). Fluorescence resonance energy transfer (FRET), a phenomenon that involves nonradiative transfer of excited state energy from a donor to an acceptor, has an efficiency that depends on the degree of overlap between the donor emission spectrum and the absorption spectrum of the acceptor. Early experiments reported efficient F6rster energy transfer between neighboring closely packed colloidal QDs of different sizes (Kagan et al., 1996). The tunable and narrow PL spectra of QDs make them potentially very suitable for biosensing applications based on energy transfer, where dye-labeled receptors conjugated to colloidal QDs can report binding events. Experiments demonstrating that QD luminescence can be quenched by surface-bound acceptor dyes in QD-protein conjugates in solution and in solid-phase formats have been described (Willard et al., 2001; Tran et al., 2002). Substantial difficulties are associated with making QDs water-soluble and derivatizable. No consistent protocol for achieving QD water-compatibility has been devised that can be applied to a wide range of QD materials. As discussed above, a number of methods exist for coating CdSe, CdSe-ZnS, CdS, or CdTe nanocrystals. Most published works to date have focused on the use of CdSe-ZnS QDs, with a few exceptions where CdS, CdSe-CdS, or CdTe QDs have been used (Mahtab et al. 1996; Bruchez et al., 1998; Mamedova et al., 2001). However, at this time, materials with cores made of materials other than CdSe tend to have low quantum yields and poor resistance to degradation in aqueous environments. 562

Quantum Dot Bioconjugates for Biosensing In our experience, although CdSe-ZnS QDs capped with DHLA are relatively stable and easy to handle, they have a quite limited functional pH range. For instance, DHLA-capped QDs are stable in basic solutions at pH > 7, but aggregation, often accompanied by loss of luminescence, takes place even in "mildly" acid solutions. Surface functionalization using a porous silica shell is reported to provide stable water-compatible QDs, but the coating process is tedious and tends to result in small amounts of material having a low quantum yield. QD conjugation to proteins has often been carried out using EDC condensation. However, even though this type of chemistry is well established for labeling biomacromolecules with organic dyes, conjugation to QDs can produce irreversible aggregation that may be either immediate or can develop with time (Mattoussi et al., 2000). Non-covalent, electrostatically driven QDbioconjugate self-assembly can ameliorate aggregation problems in some cases, but it requires the use of surface-charged nanoparticles and oppositely charged proteins with the appropriate biological activity. Other potential problems that may complicate analysis and understanding of experiments using QD-conjugates derive from the size dependence of their optical properties (e.g., emission) and from the stoichiometry of the QDbiomolecule complexes. Thus, particles emitting further in the red are larger and thus have different diffusion characteristics than smaller ones. This may, for example, complicate experiments where diffusion and dynamics of particle movement within a cell are important. In addition, QDs conjugated with multiple protein receptors or DNA oligomers per dot will have different mobility and likely experience different avidity effects than 1:1 QD:biomolecule conjugates. This may limit their usefulness in some applications.

5. Potential for Use of Quantum Dots in Bio-related Applications Semiconductor nanocrystals continue to be viewed as potentially extremely useful materials in the realm of biotechnology (Niemeyer, 2001), and the work performed so far in a number of laboratories reaffirms this expectation. Nonetheless, although great strides have been made in the short period of time since the introduction of QDs in biocompatible forms, the state of our knowledge, both in terms of basic science and of nanoengineering technology, is far from mature. Substantial opportunities will exist for new contributions in this dynamic area for the foreseeable future, particularly in the areas of DNA and protein microarray technology, fluorescence-based imaging, and high-throughput drug candidate screening. The major near-term focus will likely be on uses of QD bioconjugates as photostable substitutes for organic fluorophores as well as in multi-color barcode applications. In the former case, expanded usefulness will come when functional problems related to nanocrystal surface chemistry are solved. Remaining 563

Mattoussi, Kuno, Goldman, et al. obstacles in this area include surface oxidation, long-term stability in physiological environments, passivation methods needed to reduce non-specific interactions, and potential toxicity issues related to in vivo use. Further expansion of usefulness would also accrue from development of robust biocompatible near IR and IR-emitting materials. Linked to the these obstacles is the fact that there currently exists no acceptable way to mass-produce biocompatible QDs and QD conjugates inexpensively, a situation which will surely hinder wider usage of these materials in new applications until it is overcome. In the case of potential use of QDs and mixtures of QDs as molecular barcode elements, most likely encapsulated within polymer spheres, an obvious advantage is the inherent multiplex nature of the materials due to their nearly infinite flexibility with respect to excitation wavelengths. This flexibility should translate into simpler and less expensive optical platforms used in barcode applications. In fact, QDs embedded into microbeads have already been tested in this scenario (Han et al., 2001). It should be noted, however, that organic dyebased microsphere technology is quite advanced and may be expected to present considerable competition for analogous QD-based applications. A case in point is the high throughput flow cytometer and associated dye-labeled microsphere reagents commercialized by Luminex (Austin, TX). The system is capable of identifying 100 or more different bead types by the dye ratio contained in each. The amount of fluorescent analyte or antibody bound to each bead is simultaneously measured. In this manner very rapid multiple homogeneous assays can be performed (Kettman et al., 1998; Vignali, 2000; Iannone et al., 2001, Ye et al., 2001). Although semiconductor nanocrystals as replacements for organic fluorophores will likely account for the bulk of their near-term uses, exploitation of some of their other unique properties in bio-related scenarios is possible. As an example, experiments involving QD bioconjugates binding to neural receptors point to the possibility of utilizing the optoelectronic properties of nanocrystals in less obvious, more sophisticated ways (Winter et al., 2001). Finally, understanding the behavior of the nanocrystals and their bioconjugates at the single particle level will contribute to as yet undiscovered applications. We anticipate that fuller understanding of single-dot phenomena such as intermittent blinking, spectral shifts, effects of crystal lattice defects and surface traps, etc., will lead to development of new types of nanosensors and provide materials for additional bio-related applications.

564

Quantum Dot Bioconjugates for Biosensing

6. Acknowledgements We thank Professor M.G. Bawendi from MIT and Drs. B.L. Justus and F.S. Ligler from NRL for the fruitful discussions and useful suggestions. Financial support from the Office of the Naval Research (ONR, Dr. K. Ward), grants # N0001499WX30470, # N0001400WX20094 and # N0001401WX20854, is highly appreciated. The views, opinions, and/or findings described in this report are those of the authors and should not be construed as official Department of the Navy positions, policies or decisions.

7. References Alivisatos, A.P., 1996, Science 271,933. Artmeyev, M.V., S.V. Gaponenko, I.N. Germanenko and A.M. Kapitonov, 1995, Chem. Phys. Lett. 243, 450. BorreUi, N.F., D.W. Hall, H.J. Holland and D.W. Smith, 1987, J. Appl. Phys. 61, 5399. Bowen Katari, J.E., V.L. Colvin and A.P. Alivisatos, 1994, J. Phys. Chem. 98, 4109. Bruchez, M., Jr., M. Moronne, P. Gin, S. Weiss and A.P. Alivisatos, 1998, Science 281, 2013. Chamarro, M., C. Gourdon, P. Lavallard, O. Lublinskaya and A.I. Ekimov, 1995, Nuovo. Cim. Soc. Ital. Fis. D 17, 1407. Chamarro, M., C. Gourdon, P. Lavallard, O. Lublinskaya and A.I. Ekimov, 1996, Phys. Rev. B 53, 1336. Chan, W.C.W. and S. Nie, 1998, Science 281, 2016. Chepic, D.I., A1.L.Efros, A.I. Ekimov, M.G. Vanov, V.A. Kharchenko, I.A. Kudriavtsev and T.V. Yazeva, 1990, J. Luminesc. 47, 113. Chestnoy, N., R. Hull and L.E. Brus, 1986, J. Chem. Phys. 85, 2237. Colvin, V.L., M.C. Schlamp and A.P. Alivisatos, 1994, Nature 370, 354. Cook, R.J. and H.J. Kimble, 1985, Phys. Rev. Lett. 54, 1023. Dabbousi, B.O., M.G. Bawendi, O. Onitsuka and M.F. Rubner, 1995, Appl. Phys. Lett. 66, 1316. Dabbousi, B.O., J. RodriguezViejo, F.V. Mikulec, J.R. Heine, H. Mattoussi, R. Ober, K.F. Jensen and M.G. Bawendi, 1997, J. Phys. Chem. B 101, 9463. Dahan, M., T. Laurence, F. Pinaud, D.S. Chemla, A.P. Alivisatos, M. Sauer and S. Weiss, 2001, Optics Lett. 26, 825. Efros, A.L., M. Rosen, M. Kuno, M. Nirmal, D.J. Norris and M.G. Bawendi, 1996, Phys. Rev. B54, 4843. Efros, A.L. and M. Rosen, 1997, Phys. Rev. Lett. 78, 1110. Efros, A.L. and M. Rosen M., 2000, Ann. Rev. Mater. Sci. 30, 475.

565

Mattoussi, Kuno, Goldman, et al. Eggeling, C., J. Widengren, R. Rigler and C.A.M. Seidel, 1999, In Applied Fluorescence in Chemistry, Biology and Medicine, W. Rettig, B. Strehmel, S. Schrader, H. Seifert, Eds., Springer, Berlin, pp. 193. Ekimov, A.I., A.A. Onuschenko and V.A. Tsekhomskii, 1980, Fiz. Khim. Stekla 6,511. Ekimov, A.I. and A.A. Onuschenko, 1981, JETP Lett. 34, 345. Ekimov, A.I. and A.A. Onuschenko, 1982, Sov. Phys. Semicond. 16, 775. Ekimov, A.I. and A.A. Onuschenko, 1984, JETP Lett. 40, 337. Ekimov, A.I., A.A. Onuschenko, A.G. Pluhkin and A.L. Efros, 1985, JETP Lett. 88, 1490. Ekimov, A.I., A.L Efros and A.A. Onuschenko, 1985, Sol. Stat. Comm. 56, 921. Ekimov, A.I., F.Hache, M.C. Schanneklein, D. Ricard, C. Flytzanis, I.A. Kudryavtsev, T.V. Yazeva, A.V. Rodina and A.L. Efros, 1993, J. Opt. Soc. Am. B-Opt. Phys. 10, 100. Ekimov, A.I., 1996, J. Luminesc. 70, 1. Gaponenko, S.V., U. Woggon, M. Saleh, W. Langbein, A. Uhrig, M. Muller and C. Klingshirn, 1993, J. Opt. Soc. Am. B-Opt. Phys. 10, 1947. Gaponenko, S.V., 1998, Optical Properties of Semiconductor Nanocrystals, Cambridge University Press, Cambridge, 260 pp. Gaponik, N.P., D.V. Talapin, A.L. Rogach, A. Eychmuller, 2000, J. Mat. Chem. 10,2163. Gearheart, L., K.K. Caswell and C.J. Murphy, 2001, J. B iomed. Optics, 6, 111. Goldman, E.R., E.D. Balighian, M.K. Kuno, S. Labrenz, P.T. Tran, G.P. Anderson, J.M. Mauro and H. Mattoussi, 2002, Phys. Stat. Sol., in press. Goldman, E.R., G.P. Anderson, P.T. Tran, H. Mattoussi, P.T. Charles and J.M. Mauro, 2002, Anal. Chem., in press. Goldman, E.R., H. Mattoussi, P.T. Tran, G.P. Anderson and J.M. Mauro, 2001, In Semiconductor Quantum Dots, Ed, S. Fafard, D. Huffaker, R. Leon and R. Noetzel, Mat. Res. Soc. Pro., Pittsburgh, 642, J2.8.1. Greenham, N.C., X.G. Peng. and A.P. Alivisatos, 1996, Phys. Rev. B54, 17628. Gurevich, S.A., A.I. Ekimov, I.A. Kudryavtsev, A.V. Osinskii, V.I. Skopina and D.I. Chepik, 1992, Sov. Phys. Semicond. 26, 57. Hall, D.W. and Borrelli N.F., 1988, J. Opt. Soc. Am. B-Opt. Phys. 5,1650. Han, M., X. Gao, J.Z. Su and S. Nie, 2001, Nature Biotech.19, 631. Henglein, A., 1982, Ber. Bunsen Phys. Chem. 86, 301. Hermanson, G.T., 1996, B ioconjugate Techniques, Academic Press: London. Herron, N, Y. Wang, M.M. Edd, G.D. Stucky, D.E. Cox, K. Moiler and T. Bein, 1989, J. Am. Chem. Soc. 111,530. Hines, M.A. and P. Gnyot-Sionnest, 1996, J. Phys. Chem. B 100, 468. Hines, M.A. and P. Guyot-Sionnest, 1998, J. Phys. Chem. B 102, 3655. Hodes, G., A. Albuyaron, F. Decker and P. Motisuke, 1987, Phys. Rev. B 36, 4215. Hu, J.T., L.S. Li, W.D. Yang, L. Manna, L.W. Wang and A.P. Alivisatos, 2001, Science 292, 2060. 566

Quantum Dot Bioconjugates for Biosensing Iannone, M.A., T.G. Consler, K.H. Pearce, J.B. Stimmel, D.J. Parks and J.G. Gray, 2001, Cytometry 44, 326. Kagan C.R., C.B. Murray and M.G. Bawendi, 1996, Phys. Rev. B 54, 8633. Kettman, J.R., T.Davies, D. Chandler, K.G. Oliver and R.J. Fulton, 1998, Cytometry 33, 234. Kortan, A.R., R. Hull, R.L. Opila, M.G. Bawendi, M.L. Steigerwald, P.J. Carrolla and L.E. Brus, 1990, J. Am. Chem. Soc. 112, 1327. Kotov, N.A., K. Putyera, J.H. Fendler, E. Tombacz and I. Dekany, 1993, Coll. Surf. A-Phys. Eng. Aspects 71, 317. Kotov, N.A., F.C. Meldrum, C. Wu and J.H. Fendler, 1994, J. Phys. Chem. 98, 2735. Kuno, M., J.K. Lee, B.O. Dabbousi, F.V. Mikulec and M.G. Bawendi, 1997, J. Chem. Phys. 106, 9869. Kuno, M., D.P. Fromm, H.F. Hamann, A. Gallagher and D.J. Nesbitt, 2001, J. Chem. Phys. 115, 1028 Liu, L.C. and S.H. Risbud, 1990, J. Appl. Phys. 68, 28. Liu, L.C., M.J. Kim, S.H. Risbud and R.W. Carpenter, 1991, Philosoph. Mag. B 63,769. Lobenhofer E.K., P.R. Bushel, C.A. Afshari and H.K. Hamadeh, 2001, Env. Health Pers. 109, 881. Mahtab, R., J.P. Rogers, C.P. Singleton and C.J. Murphy, 1996, J. Am. Chem. Soc., 118, 7028. Mamedova, N.N., N.A. Kotov, A.L. Rogach and J. Studer, 2001, Nano Lett. 1, 281. Mathieu, H., T. Richard, J. Allegre, P. Lefebvre and G. Arnaud, 1995, J. Appl. Phys. 77, 287. Mattoussi, H., A.W. Cumming, C.B. Murray, M.G. Bawendi, and R. Ober, 1998-11, Phys. Rev. B 58, 7850. Mattoussi, H., L.H. Radzilowski, B.O. Dabbousi, E.L. Thomas, M.G. Bawendi and M.F. Rubner, 1998, J. Appl. Phys. 83, 7965. Mattoussi, H., J.M. Mauro, E.R. Goldman, G.P. Anderson, V.C. Sundar, F.V. Mikulec and M.G. Bawendi, 2000, J. Am. Chem. Soc. 122, 12142. Mattoussi, H., J.M. Mauro, E.R. Goldman, R.M. Green, G.P. Anderson, V.C. Sundar and M.G. Bawendi, 2001, Phys. Stat. Sol. 224, 277. Meyer, M., C. W allberg, K. Kurihara and J.H. Fendler, 1984, J. Chem. Soc. Chem. Commun. 90, 90. Mikulec, F.V., 1999, Ph.D. Dissertation, Massachusetts Institute of Technology. Minti, H., M. Eyal, R. Reisfeld and G. Berkovic, 1991, Chem. Phys. Lett. 183, 277. Misawa, K., H. Yao, T. Hayashi and T. Kobayashi, 1991, Chem. Phys. Lett. 183, 113. Misawa, K., H. Yao, T. Hayashi and T. Kobayashi, 1991, J. Chem. Phys. 94, 4131. Mitchell, G.P., C.A. Mirkin and R.L. Letsinger, 1999, J. Am. Chem. Soc. 121, 8122. 567

Mattoussi, Kuno, Goldman, et al. Murray, C.B., S. Sun, W. Gaschler, H. Doyle, T.A. Betley and C.R. Kagan, 2001, IBM J. Res. Dev. 45, 47. Murray, C.B., C.R. Kagan and M.G. Bawendi, 2000, Ann. Rev. Mater. Sci. 30, 545. Murray, C.B., D.J. Norris and M.G. Bawendi, 1993, J. Am. Chem. Soc. 115, 8706. Niemeyer, C.M., 2001, Angew. Chem. Int. Ed. 40, 4218. Nirmal, M., D.J. Norris, M. Kuno, M.G. Bawendi, A.L. Efros and M. Rosen, 1995, Phys. Rev. Lett. 75, 3728. Nirmal, M., B.O. Dabbousi, M.G. Bawendi, J.J. Macklin, J.K. Trautman, T.D. Harris and L.E. Brus, 1996, Nature 383, 802. Nogami, M., K. Nagasaka and M. Takata, 1990, J. Non-Cryst. Sol. 122, 101. Norris, D.J., N. Yao, F.T. Chamock and T.A. Kennedy, 2001, Nano Lett. 1, 3. Ochoa, R.O., C. Colajacomo, E.J. Witkowski, J.H. Simmons and B.G. Potter, 1996, Sol. Stat. Comm. 98, 717. Pathak, S., S.K. Choi, N. Amheim and M.E. Thompson, 2001, J. Am. Chem. Soc. 123, 4103. Peng, X., M.C. Schlamp, A.V. Kadavanich, U. Banin and A.P. Alivisatos, 1997, J. Am. Chem. Soc. 119, 7019. Peng, X.G., L. Manna, W.D. Yang, J. Wickham, E. Scher, A. Kadavanich and A.P. Alivisatos, 2000, Nature 404, 59. Peng, Z.A. and X.G. Peng, 2001, J. Am. Chem. Soc. 123, 183. Peng, Z.A. and X.G. Peng, 2001, J. Am. Chem. Soc. 123, 1389. Persans, P.D., A. Tu, Y.I. Wu and M. Lewis, 1989, J. Opt. Soc. Am. B-Opt. Phys. 6, 818. Pileni M.P., L. Motte and C. Petit, 1992, Chem. Mater. 4, 338. Potter, B.G. and J.H. Simmons, 1988, Phys. Rev. B 37, 10838. Potter, B.G. and J.H. Simmons, 1988, J. Appl. Phys. 68, 1218. Qu, L., Z.A. Peng and X.G. Peng, 2001, Nano Lett. 1,333. Rodriguez-Viejo, J., H. Mattoussi, J.R. Heine, M.K. Kuno, J. Michel, M.G. Bawendi, and K.F. Jensen, 2000, J. Appl. Phys. 87, 8526. Roederer, M., S. DeRosa, R. Gerstein, M. Anderson, M. Bigos, R. Stovel, T. Nozaki, D. Parks, L. Herzenberg and L. Herzenberg, 1997, Cytometry 29, 328. Rogach, A.L., L. Katsikas, A. Kornowski, D. Su, A. Eychmuller and H. Weller, 1997, Ber. Bunsen. Phys. Chem. 101, 1668. Rossetti, R., S. Nakahara and L.E. Brus, 1983, J. Chem. Phys. 79, 1086. Rossetti, R., J.E. Ellison, J.M. Gibson and L.E. Brus, 1984, J. Chem. Phys. 80, 4464. Schlamp, M.C., X.G. Peng and A.P. Alivisatos, 1997, J. Appl. Phys. 82, 5837. SchrOck, E., S. du Manoir, T. Veldman, B. Schoell, J. Wienberg, M.A. FergusonSmith, Y. Ning, D.H. Ledbetter, I. Bar-Am, D. Soenksen, Y. Garini and T. Ried, 1996, Science, 273,494. Spanhel L., E. Arpac and H. Schmidt, 1992, J. Non-Crys. Sol. 147, 657. Sondi, I., O. Siiman, S. Koester and E. Matijevic, 2000, Langmuir 16, 3107. 568

Quantum Dot Bioconjugates for Biosensing Steigerwald, M.L., A.P. Alivisatos, J.M. Gibson, T.D. Harris, R. Kortan, A.J. Muller, A.M. Thayer, T.M. Duncan, D.C. Douglass and L.E. Brus, 1988, J. Am. Chem. Soc. 110, 3046. Steigerwald, M.L. and L.E. Brus, 1989, Ann. Rev. Mat. Sci. 19, 471. Sun, B., W. Xie, G. Yi, D. Chen, Y. Zhou, and J. Cheng, 2001, J. Irranunol. Meth., 249, 85. Tedeschi, C.; F. Caruso, H. MOhwald and S. Kirstein, 2000, J. Am. Chem. Soc. 122, 5841. Tran, P.T., E.R. Goldman, H. Mattoussi, G.P. Anderson and J.M. Mauro, 2001, Proc. SPIE 4258, 1. Tran, P.T., E.R. Goldman, G.P. Anderson, J.M. Mauro and H. Mattoussi, 2002, Phys. Stat. Sol., In Press. Trindade T., P. O'Brien, N.L. Pickett, 2001, Chem. Mat. 13, 3843. Vignali, D.A.A., 2000, J. Immunol. Methods 243, 243. Wang, Y, N. Herron, W. Mahler and A. Suna, 1989, J. Opt. Soc. Am B-Opt. Phys. 6, 808. Weller, H, H.M. Schmidt, U.Koch, A. Fojtik, S. Baral, A. Henglein, W. Kunath, and K. Weiss, 1986, Chem. Phys. Lett. 124, 557. Willard, D.M., L.L. Carillo, J. Jung and A.Van Orden, 2001, Nano Lett. 1,469. Willner, I., F. Patolsky and J. Wasserman, 2001, Ang. Chem. Intl. Ed. 40, 1861. Winter, J.O., T.Y. Liu, B.A. Korgel and C.E. Schmidt, 2001, Adv. Mater. 13, 1673. Woggon, U., S.V. Bogdanov, O. Wind, K.-H., Schlaad, H. Pier, C. Klingshim, P. Chatziagorastou and H.P. Fritz, 1993, Phys. Rev. B48, 11979. Woggon; U., 1997, Optical Properties of Semiconductor Quantum Dots, Springer-Verlag, Berlin, 251 pp. Ye, F., M.S. Li, J.D. Taylor, Q. Nguyen, H.M. Colton, W.M. Casey, M. Wagner, M.P. Weiner and J. Chen, 2001, Hum. Mutat. 17, 305. Yoffe, A.D., 1993, Adv. Physics, 42, 173. Yoffe, A.D., 2001, Adv. in Physics 50, 1. Zhao, X.S., J. Schroeder, P.D. Persans and T.G. Bilodeau, 1991, Phys. Rev. B 43, 12580.

569

Optical Biosensors: Present and Future F.S. Ligler and C.A. Rowe Taitt (editors) 9 2002 Elsevier Science B.V. All rights reserved

CHAPTER 18

SOFT LITHOGRAPHY AND MICROFLUIDICS

RAVI S. KANE ~ PH.D ABRAHAMD STROOCK1 NOO LI JEON 1 PH.D. DONALD E. INGBER2 PH.D. AND GEORGE M. WHITESIDES1 PH.D 9

9

1Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 0213 8 2Departments of Pathology and Surgery, Children's Hospital and Harvard Medical School Boston, MA 02115

Optical biosensors necessarily involve an interface between synthetic materials and biological systems. This chapter describes the application of soft lithography to create and control this interface. Soft lithography is a set of techniques that includes 1) methods of fabricating microstructures in polymers, especially elastomers, 2) uses of these methods in combination with organic surface chemistry to generate micron-scale patterns on synthetic surfaces, and 3) uses of microfluidic systems to pattern the composition of the fluid medium adjacent to a surface. These techniques allow the immobilization of biomolecules and cells at surfaces with micron-scale resolution, and for the control of the subsequent interaction of these species with liquid media. These techniques are compatible both with optical and electronic materials and with biological systems. This review focuses on the use of soft lithography to fabricate microfluidic systems and to position and manipulate living cells on surfaces.

1. Technical Concept 1.1. Introduction This chapter describes the use of a set of non-conventional (i.e., not based on photolithography) microfabrication techniques known as soft lithography (Xia 571

Kane, Stroock, Jeon, et al. and Whitesides, 1998; Whitesides et al., 2001) to create and control the interfaces between synthetic materials and biological systems. The term, "soft lithography," describes an integrated set of techniques for fabricating microstructures in an elastomeric material, for modifying the chemical properties of surfaces, and for controlling flows of fluid adjacent to surfaces. In the context of the interface between synthetic materials and biological systems, soft lithography makes use of elastomeric stamps, membranes, and microfluidic channels to deposit small molecules, biological molecules, and living cells on synthetic substrates with micron-scale spatial resolution. Microfluidic channels made with soft lithography provide an environment for cell culture in which reagents and analytes can be delivered non-destructively to cells with sub-cellular precision (Duffy et al., 1998; Takayama et al., 1999). These soft lithographic techniques are inexpensive, procedurally simple, and do not require stringent control of the laboratory environment (i.e., a cleanroom is not required). They provide greater flexibility and convenience than photolithography for patterning organic and biological materials. Soft lithographic techniques offer a means for tailoring the interface between the "optical" and the "bio" components of optical biosensors. Figure 1 illustrates this interface schematically. The characteristics of the interface that must be controlled are 1) the position (on the scale of microns) of biological elements with respect to sensing elements, 2) the chemical interaction (both attractive and repulsive) between the surfaces of the materials used in the biosensors and biomolecules and cells, and 3) the local environment of the biological system, in such a way that the biological components remain active. The materials used in creating this interface must also be compatible with both the optical system (that is, they must be transparent, have the correct index of refraction, and form adequate mechanical seals) and the biological system (that is, they must be nontoxic, be selective in their molecular recognition, and have appropriate surface composition). These criteria are met by the materials used in soft lithography. The examples that we use in this chapter concentrate on systems of the class that is shown in Figure 1 in which molecules, cells, and media can all, in principle, be patterned. These systems--instrumented micro-cell culture systems~are immediately useful in fundamental studies of cell biology. They also have the potential to act as sophisticated biosensors and analytical systems. Both cellbased assays, which require repeated examination of individual cells, and biosensors, which rely on the collective observation of multiple cells, will benefit from accurate control of cell location. B iosensors and combinatorial screens may also require surfaces that display specific ligands in a surface that otherwise minimizes non-specific interactions with proteins or cells. Control of the bio/materials interface is also key to solving an important, long-term problem: the design of hybrid systems that combine (or allow direct communications between) living and non-living systems.

572

Soft Lithograpy and Microfluidics

Figure 1. Controlling the interactions of a cell with its environment using soft lithography. Most of the environmental features sensed by the cell can be patterned using soft lithography or devices fabricated using soft lithography: the surface on which proteins adsorb or that presents ligands (patterning by microcontact printing); the identity of neighboring cells (membrane-based patterning or patterning using three-dimensional microfluidic systems); the composition of the extracellular medium (laminar flow patterning in microchannels).

1.2. Soft lithographic methods of microfabrication Controlling the environment experienced by individual cells or groups of cells requires control over the composition of both the surface and the medium on relevant length scales (micrometers for single cells and millimeters to centimeters for groups of cells). We have developed a set of techniques that we call "soft lithography" that is an alternative to photolithography, and that can be used to create microstructures, and to control the surface chemistry of synthetic materials, with a spatial resolution of microns. Soft lithographic techniques are inexpensive, procedurally simple, and do not require stringent control over the laboratory environment. These techniques can be used to pattern both planar and non-planar substrates, and also to pattern the cell culture medium. In soft lithography, elastomeric stamps, microfluidic channels, and membranes prepared by casting or spin coating the liquid prepolymer of an appropriate elastomer against a master that has a patterned relief structure (Figure 2). Most of the research based on soft lithography has used poly(dimethylsiloxane) 573

Kane, Stroock, Jeon, et al.

Figure 2. Schematic diagrams of the soft lithographic approach of fabricating microfluidic channels (A), membranes, and stamps (B). A) A transparency prepared on a high resolution printer (5000 dpi) is used as a photomask. Epoxy photoresist is spun onto a silicon wafer, exposed, and developed to create a master structure. Many (> 100) negative copies of the structure on the master can be formed by molding the structure into poly(dimethylsiloxane) (PDMS), an elastomeric polymer. To form a closed channel, the PDMS mold is sealed to a flat surface either covalently by oxidizing the surfaces in a low temperature plasma or non-covalently by applying pressure. B) Fabrication of a membrane (left) and a stamp (right) from a master. To form a membrane, PDMS is spun onto the master in a thin layer such that the features on the master create holes that tranverse the entire thickness of the layer. On a stamp, the negative of the features on the master are molded in bas-relief on one surface (McDonald et al., 2000; Xia and Whitesides, 1998; Whitesides and Stroock, 2001). (PDMS) as the elastomer, because PDMS is biocompatible, permeable to gases (and can thus be used for cell culture), and inexpensive. PDMS also has good optical characteristics; the cured polymer is transparent from 235 nm to the near infrared (Wu and Whitesides, 2001) and can make tight, weakly scattering seals around embedded optical elements such as optical fibers (Chabinyc et al., 2001). The interfacial properties of PDMS can be readily modified by plasma oxidation and silanization (Chaudhury and Whitesides, 1991). PDMS structures can often be used many times in transferring patterns (we have used the same PDMS stamp in microcontact printing approximately 100 times over a period of several months without any noticeable degradation in its performance), and each master can be used to make a large number of stamps or membranes. The access to photolithographic equipment required (to fabricate masters) in soft lithography is therefore minimal. 574

Soft Lithograpy and Microfluidics

An advantage of soft lithography as a method for patterning cells is that, at the feature sizes required for this application (typically, 2-500 /zm), it is often possible to make photomasks using procedures that are significantly more rapid and less expensive than those commonly used to make chrome masks for conventional photolithography. For the fabrication of masters having feature sizes greater than or equal to 20 /~m, masks can be generated by the highresolution laser printing of patterns (generated using computer programs such as Freehand or AutoCAD) onto flexible transparencies (Qin et al., 1996). The masks can be made in a few hours at a cost as low as $0.25 per square inch. For feature sizes between 10/zm and 20 #m, the optical reduction of images printed onto transparencies generates patterns in microfiche (Deng et al., 1999). Microfiche is then used as the photomask. For feature sizes between 2/xm and 20 #m, a relatively inexpensively approach is to use commercial laser writing to fabricate masters from which PDMS stamps can be molded (Grzybowski et al., 1998). For feature sizes between a few hundred nanometers and 20 ~tm, there are more specialized techniques (Wu and Whitesides, 2001; Love et al., 2001b). The capability to produce features larger than 20/zm rapidly and inexpensively allows researchers to prototype and produce small numbers of simple microstructures and microsystems. This capability, which we call "rapid prototyping", has minimized the barriers to the use of lithographic techniques by biochemists. Soft lithography also facilitates the fabrication of complex structures such as three-dimensional (3D) networks of channels (Love et al., 2001a; Anderson et al., 2000). Figure 3 shows an example of a multi-level network of microchannels made by stacking a stamp and a membrane (Chiu et al., 2000). The elastomeric character of PDMS enables the simple integration of a variety of thin organic materials such as filters and dialysis membranes (Ismagilov et al., 2001; Chiu et al., 2001). 1.3. Molecular control of interfaces

A necessary ingredient for the control of the bio-material interface is a versatile strategy for adding organic functionality to synthetic substrate. We have made extensive use of self-assembled monolayers (SAMs) for this purpose (Dubois and Nuzzo, 1992; Whitesides and Gorman, 1995; Whitesides et al., 1996; Wilbur and Whitesides, 1999; Folch and Toner, 2000; Mrksich, 2000). SAMs are organized organic monolayer films that provide molecular-level control over the composition and properties of the interface. Most studies of SAMs have involved monolayers of alkanethiolates on gold and silver. A benefit of working with SAMs of thiols on thin metal layers is compatibility with surface plasmon resonance (SPR) techniques (Sigal et al., 1996, 1997; Lahiri et al., 1999b; Chapman et al., 2000). SAMs of alkanethiolates on gold and silver are also

575

Kane, Stroock, Jeon, et al.

Figure 3. Three-dimensional microfluidic network for applying reagents to a surface in a discontinuous pattern, a technique called 3D MIMIC. A) Scheme for the fabrication of 3D-microfluidic stamp. The slab and the membrane are formed with different masters (cf. Figure 2B). The membrane contains segments of channel that will be in contact with the surface that is to be patterned. Vertical vias in the membrane connect the lower channels (in contact with surface) to the upper layer of channels in the slab. The aligned stack of the slab and the membrane are sealed (non-covalenfly) to the surface that is to be patterned. B) Schematic illustration of the fluid paths in the 3D network. C) Silicon oxide surface that has been etched with a network of channels such as in (A) and (B) in which different concentrations of etchant (hydrofluoric acid) were run through the three independent channels. Different shades of the etched regions correspond to different thicknesses of the oxide after etching (Chiu et al., 2000).

compatible with microcontact printing, a powerful technique for creating chemical patterns with micron-scale resolution (cf. Section 1.4) (Wilbur et al., 1994; Xia and Whitesides, 1998). Figure 4 outlines three methods developed in our group for attaching ligands covalently to surfaces with controlled orientation and density (Lahiri et al., 1999a, 1999b; Roberts et al., 1998; Yan et al., 1997). Control over the density of groups presented at a surface can be achieved by forming a SAM from a solution containing a mixture of alkanethiols, although phase segregation in the monolayer might affect the surface properties of certain mixed SAMs. In Figures 4A and 4B, the ligand is presented in a background of SAMs terminated with oligomers of ethylene glycol (OEG); SAMs that present OEG resist the nonspecific adsorption of protein (Prime and Whitesides, 1991); such surfaces are called "inert". The presentation of ligands in an inert background is important for performing quantitative, low-noise binding assays at surfaces (Lahiri et al., 1999b).

576

Soft Lithograpy and Microfluidics A. Direct method

B. C o m m o n intermediate method

C. Anhydride method

(CF3CO}~O, Et3N,

DMF, 20 rain, rl

[~

~ 1)SHCIIHr~EGQ

~,

t,AU,

t,

t I

2) 5 HCIIHtzE ~C HzCONH"v--.~>

,

I

s

L

s,

,s~

AU

s,

s

.s

]

!

,

.,

AU

,

~

!

[

,

,

, ,,

Au

,

,__

I

. . . . .

Au

!

Figure 4. Generation of SAMs presenting specific ligands: A) Immersion of a gold substrate in a solution containing a mixture of an alkanethiol terminating in oligo(ethylene glycol) groups (OEG), and a second alkanethiol terminating in a ligand of interest, generates a surface presenting the desired ligand (Roberts et al., 1998). B) A mixed SAM is formed by the immersion of a gold substrate in a solution containing a mixture of EG3OH-terminated and EG6-OCH2COOH-terminated alkanethiols. Activation of the carboxylic acid groups using N-hydroxysuccinimide (NHS) and 1ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), followed by a reaction with an amine-terminated ligand, generates a surface presenting that ligand (Lahiri et al., 1999a, 1999b). C) A surface presenting interchain carboxylic acid anhydrides is generated by treating SAMs that present terminal carboxylic acid groups with trifluoroacetic anhydride. Reacting these activated SAMs with amine-terminated ligands results in a surface presenting a 1:1 mixture of the desired ligand and carboxylic acid groups (Yan et al., 1997).

1.4. Patterning molecules and cells on synthetic surfaces The types of microstructures (Section 1.2) and surface chemistry (Section 1.3) described i n the proceeding sections can be combined to form powerful techniques for patterning organic chemical functionality, biomolecules, and even living cells on surfaces with micron-scale precision. In this section, we outline the use of three such techniques: microcontact printing (lxCP) (Wilbur et al., 1994; Xia and Whitesides, 1998), membrane patterning (MEMPAT)(Ostuni et al., 2000), and three-dimensional micromolding ha capillaries (3D MIMIC)(Chiu et al., 2000). Microcontact printing is a technique that uses the relief pattern on the surface of an elastomeric PDMS stamp to form patterns of SAMs on the 577

Kane, Stroock, Jeon, et al.

Figure 5. Procedure for patterning SAMs by microcontact printing: A stamp is inked with an alkanethiol and placed on a gold (or silver) surface; the pattern on the stamp is transferred to the gold by the formation of a SAM on the regions that contacted the substrate. The bare areas of the gold are exposed to a different alkanethiol to generate a surface patterned with a SAM that presents different chemical functionalities in different regions (Xia and Whitesides, 1998). surfaces of substrates (Figure 5). Patterned SAMs generated by microcontact printing can be used to control the adsorption of proteins on surfaces. L6pez et al. (1993) first used microcontact printing to pattern gold surfaces into regions presenting oligo(ethylene glycol) groups in a background of methyl groups. Immersion of the patterned SAMs in solutions of proteins resulted in the adsorption of proteins only on the methyl-terminated regions. These systems have subsequently been extended to other experiments in cell biology (see Section 3.2) (Chen et al., 1997; Mrksich et al., 1997; Mrksich, 2000). While microcontact printing is a technique that has sufficient resolution to allow the patterning of single cells, in its simplest configuration, it does not allow any changes in the pattern or shape of adsorbed cells. MEMPAT and 3D MIMIC are complementary techniques that not only allow the patterning of cells~individually or in groups---on arbitrary substrates, but also allow studies of the spreading or migration of cells from their initial pattern. MEMPAT makes use of elastomeric membranes~free-standing PDMS films that have through-membrane pores---to pattern proteins and cells on a variety of substrates including plastics and glass (Figure 6) (Folch and Toner, 2000; Ostuni et al., 2000). Bringing an elastomeric membrane into contact with a substrate restricts access of a solution of protein or a suspension of cells to those regions of the substrate exposed through the pores. The deposition of proteins, or the attachment of cells, is therefore restricted to these exposed regions of the substrate. The patterned cells are constrained by the walls of the pores in the membrane. The ability to remove the constraints imposed by the membrane by peeling it away from the substrate, and to observe the subsequent spreading or 578

Soft Lithograpy and Microfluidics

Figure 6. Schematic diagram that describes the use of MEMPAT for plating cells onto a substrate with well defined position and shape. The surface of the membrane and the walls of its holes are coated with bovine serum albumin (BSA). The membrane is placed on a clean surface (e.g., Petri dish) and exposed to a solution of fibronectin (FN). After rinsing with a solution of a phosphate buffer solution, the membrane and the substrate are covered with a suspension of ceils for 24 h. The membrane can be removed without damaging the cells and the protected areas of the substrate can be modified by the adsorption of an adhesive protein that allows the patterned cells to spread (Ostuni et al.,

2000).

Figure 7. Schematic diagram of a "T-Sensor" based on the diffusional mixing between three laminarly flowing streams in a microchannel (Weigl and Yager, 1999). The analyte of interest (e.g., an enzyme) is introduced in the sample stream (bottom). A solution of indicator (e.g., a substrate that becomes fluorescent upon interaction with the enzyme) flows in the detection stream (center). The third stream contains a reference solution. Diffusive mixing of the solutions occurs at the interface of the streams. The product of this mixing (e.g., a fluorophore) will be localized in the diffusively mixed region between the streams. migration of the cells as a function of the composition of the substrate and the cellular environment, may be exploited in several areas of cell biology. 3D 579

Kane, Stroock, Jeon, et al. MIMIC allows the generation of arbitrary and discontinuous patterns of proteins or cells on planar substrates (Figure 3). We had previously developed a soft lithographic technique called MIMIC (micromolding in capillaries) (Kim et al., 1995) for fabricating three-dimensional structures by allowing solutions to flow into microfluidic channels. The use of MIMIC was limited to relatively simple, continuous patterns. To overcome some of the limitations of MIMIC, we developed an analogous technique that makes use of three-dimensional microfluidic systems (Anderson et al., 2000) for patterning (3D MIMIC). Figure 3C illustrates the use of 3D MIMIC to create a discontinuous pattern of etched regions in a silicon oxide layer on silicon. The same methods can be used with solutions of proteins and suspensions of cells to deliver these elements to surfaces in complex, discontinuous patterns (cf. Section 3.2.2).

1.5. Laminar flow in microfluidic systems Networks of microchannels can be used to control the location of fluids on the micron scale. This spatial control can be extended by taking advantage of the laminar character of flows in microchannels (Takayama et al., 1999; Kenis et al., 1999). Since microfluidic channels are small (100 /.tm in cross-sectional dimension), the flow of liquids in them is usually laminar (Bird, 1960); adjacent streams of different composition mix into one another only by diffusion. This characteristic of laminar flows has been exploited by Weigl and Yager (1999) for continuous chemical analysis of solutions flowing in a microchannel (Figure 7). Laminar flows in microchannels also allow the chemical composition of solutions and surfaces to be patterned on scales smaller than the channel itself; we refer to this method of patterning as "laminar flow patterning" (Kenis et al., 2000). Figure 8 demonstrates the use of laminar flow patterning for the electroless deposition of a silver wire on the center of the floor of a microchannel (Kenis et al., 1999). Laminar flow patterning can be extended to biological systems and :allows for control of both cells and the cellular environment---qhat is, it allows control over the nature of molecules that are deposited on the substrate, the nature and position of neighboring cells, and the composition of the extracellular medium. Laminar flow patterning makes it possible to pattern the fluid culture medium itself (cf. Section 3.2.3). Networks of microchannels can be used to generate complicated patterns in the composition of flowing solutions. A strategy for forming gradients in the concentration of solutes in a flow of buffer in a microchannel is shown in Figure 9: A small number (six in the case shown) of solutions of different composition are injected into a network of channels; in the network, streams of the solutions are allowed to divide and mix into one another to form a larger number (24 in the case shown) of streams of solutions of intermediate composition; at the outlet of the network, the multiple streams are allowed to recombine in single channel across which there is, as a result of this "combination", a gradient in the 580

Soft Lithograpy and Microfluidics

Figure 8. Demonstration of patterned surface chemistry achieved with laminar fowing streams in a microchannel. A silver wire deposited in a zigzag channel at the laminar flow interface between solutions containing the components of an electroless silver plating solution (Kenis et al., 1999).

Figure 9. Gradients of solute made with a microfluidic network, a) Schematic diagram of network that transforms 6 input solutions into 24 output solutions that contain intermediate concentrations of solute in the incoming streams. The 24 streams flow laminarly in a broad outlet channel (2160/ma wide), b-d) Demonstration of three types of gradients that can be formed in the outlet channel using fluorescein solutions. On top are fluorescent micrographs of streams flowing in the outlet channel. Below, plots show the corresponding fluorescence intensity profile across the outlet channel at the beginning of the channel (L1, white dotted line) and 800/an downstream from the junction (L2, white dotted line). The fluorescein concentration of the solutions introduced into the inlets of the microfluidic device is shown above the individual gradients (Dertinger et al., 2001).

concentration of the solutes from the original solutions; under laminar flow conditions, this gradient propagates along the channel with only diffusional 581

Kane, Stroock, Jeon, et al. broadening of the gradient profile (Jeon et al., 2000; Dertinger et al., 2001). The gradient in the combined stream can be used to study cell behaviors such as chemotaxis. In the case in which the solutes adsorb on walls of the channel, we have found that the gradient in the combined stream led to gradients of qualitatively similar form in the concentration of the species that are bound to the wall (unpublished results).

2. History 2.1. Self assembled monolayers The history of ordered molecular monolayers is long, but only in the past few decades have convenient methods emerged that allow for the formation of high quality films on solid substrates. Early work on monolayers was based on the Langmuir-Blodgett method in which the film is formed at the liquid-air interface and subsequently transferred to a solid support (Ulman, 1991). This method can lead to films with a high degree of molecular ordering, but the process is complicated and prone to errors. Silanes offer a flexible way to bring organic functional groups to solid surfaces that present hydroxyl groups (Grushka, 1974). The molecules in silane layers are only partially ordered, and silane films are prone to degradation in aqueous buffer due to the hydrolysis of silicon-oxygensilicon bonds. In 1985, Allara and Nuzzo discovered the self-assembly of disulphides on metal surfaces (Allara and Nuzzo, 1985); this discovery led to work with thiols on metals. Long-chain (greater than 10 carbons) alkane thiols form SAMs with crystalline order that are very stable in aqueous medium. SAMs of thiols have been extensively characterized by our group and others (Dubois and Nuzzo, 1992; Whitesides and Gorman, 1995).

2.2. History of patterned surface chemistry Until recently, there were few methods available to pattern the chemical groups presented on a surface with micron-scale resolution: photo-labile groups presented a surfaces could be patterned with exposure through a photomask (Wollman et al., 1994); photolithography could also be used to define patterns in photoresist that would act as a mask for the deposition of metals, oxides, and organics (Muller and Kamins, 1986). The development of SAMs of thiols on metals opened the possibility of using traditional patterning methods such as writing and stamping to create patterns of well ordered molecular layers with sub-micron-scale resolution (Kumar et al., 1992, 1994; Wilbur et al., 1994; Piner et al., 1999).

582

Soft Lithogravy and Microfluidics

2.3. History of microfluidics The initial development of microfluidics (Manz et al., 1991) used fabrication techniques adapted from the electronics industry. With these techniques, channel structures are formed in hard materials such as glass and silicon using photolithography followed by etching; the channels are typically sealed by anodic bonding (Kovacs, 1998). These steps are slow, expensive, and require a cleanroom environment. Hard plastics are also used (see for example, Micronics, www.micronics.net). Most of the early work in microfluidics focused on using electroosmotic flow (Harrison et al., 1993; Jacobson et al., 1994). Over the past few years, we and others have been developing alternative methods based on soft lithography to fabricate microfluidic devices (Delamarche et al., 1997; Duffy et al., 1998; Beebe et al., 2000; Quake and Scherer, 2000) (Figures 2 and 3). These methods, which use PDMS as the principle material, are simple, are inexpensive, and can be performed in a standard laboratory environment. The mechanical flexibility of PDMS makes it appropriate for the fabrication of the movable components that are often required for the control of pressure-driven flows (Unger et al., 2000).

2.4. History of patterning cells Early work on patterning cells on synthetic substrates was done using silanes and photolithography (Kleinfeld et al., 1988); this process required multiple steps. The development of ~tCP simplified the process of patterning cells. We have employed ~tCP to pattern extracellular matrix proteins and to control the position, shape, and function of single living cells (Singhvi et al., 1994; Mrksich et al., 1997; Chen et al., 1997; Dike et al., 1999; Kane et al., 1999; Takayama et al., 2000). Others working in the area of cell patterning include the groups of Toner (Folch and Toner, 2000) and Shakesheff (Patel et al., 1998). The use of cells in sensors has been developed by the groups of Stenger and Kovacs (Jung et al., 1998; Pancrazio et al., 1998).

3. State of the Art 3.1. Microfluidic systems fabricated using soft lithography 3.1.1. Microfluidic channels and components. Soft lithography provides a method for fabricating almost any system of channels that might be needed for microfluidics (Figure 2, of. Section 1.2) (McDonald et al., 2000). For example, channels made in PDMS using soft lithography support electroosmotic flow; these channels can be used for capillary electrophoresis (Duffy et al., 1998). Soft

583

Kane, Stroock, Jeon, et al.

Figure 10. Rotary pump made with soft lithography in PDMS. A) Schematic diagram of a peristaltic pump based on pneumatic valves that take advantage of the elastomeric character of PDMS. The air pressure in each upper channels is controlled independently; high pressure in one of the upper channels locally deforms the lower channel and restricts flow. When actuated sequentially, the three valves act as a peristaltic pump. B) Flow ring. Six valves are activated sequentially to drive fluid around the circle when the input and output channels are blocked. The recirculating flow can be used to increase the residence time of flow in the channel for mixing or for the completion of a slow chemical reaction (Unger et al., 2000; Quake and Scherer, 2000).

lithography also simplifies the fabrication of three-dimensional channels (Figure 3) (Anderson et al., 2000); the Beebe group used soft lithography to fabricate a active components, such as pumps and valves, that are required in microfluidic systems. Chou and Quake have designed a valve and a peristaltic pump using a multilayer structure made from PDMS (Figure 10) (Unger et al., 2000; Quake and Scherer, 2000). Soft lithographic methods make the registration and integration of multiple layers (at the 50 ktm scale) simple (Love et al., 2001b). The elastomeric character of PDMS allows the valve to be actuated with small changes in the pressure in the gas-filled channels.

3.1.2. Integrated microfluidic devices. Soft lithographic methods also facilitate the integration of multiple materials (e.g., PDMS, glass, organic membranes, polymer tubing, and metal films) and non-fluidic components (e.g., lenses, optical fibers, and electrodes) into a single device. An important integration process that is simplified by soft lithography is the connection of external tubing with on-chip microchannels for the introduction and collection of samples. With soft lithography, these connections are made by simply boring holes in the PDMS and press-fitting tubing into these holes; this press-fit seal can withstand several bars of pressure (McDonald et al., 2000).

584

Soft Lithograpy and Microfluidics

Figure 11. Parallel detection using an array of crossing microfluidic channels. A) Schematic drawing of a lxl array of microfluidic channels in which a poly(carbonate) membrane (0.1 ktm pores) separates the two channels at the point at which they cross. The bottom channel contains a substrate entrapped in a gel matrix. B) Detection of enzymatic activity using fluorometric (top) and colorimetric (bottom) methods in a 5x5 array of crossing channels. Substrates for enzymes were immobilized in a 1.25% agarose gel in the lower (horizontal) set of channels. Solutions of different enzymes were allowed to flow by gravity in the upper (vertical) set of channels. The ELF-97-1inked substrates release a fluorescent molecule upon cleavage by an enzyme; these products are visualized under UV illumination (top). BCIP/NBT and X-gal form precipitates upon cleavage; these products are visualized by optical adsorption (bottom).

Figure 11 shows a simple optical biosensor that allows for five tests to be performed on five solutions in parallel. In this device, a poly(carbonate) membrane is sealed between two layers of PDMS, each of which contains a set of microchannels; the membrane allows for diffusive (not convective) exchange of molecules between the channels in the regions in which they cross (Ismagilov et al., 2001). The bottom channels contain substrates for the enzymes of interest in an agarose gel. The sample solutions are allowed to flow in the upper channels. An enzyme is detected when it diffuses from an upper channel into the lower channel and acts on one of the substrates to form either a precipitate or a fluorescent molecule. This type of hybrid system (PDMS-thin film-PDMS) is easy to fabricate using soft lithography because PDMS conforms and seals around the intervening layer. Figure 12 shows an integrated microfluidic device for the separation and detection of fluorescently-labeled proteins; an optical fiber, optical filter, and micro avalanche photo diode (ktAPD) are integrated with the microchannel (Chabinyc et al., 2001). The fiber is molded into the slab of PDMS that contains 585

Kane, Stroock, Jeon, et al.

Figure 12. Microfluidic chip in PDMS for capillary electrophoresis with integrated optical fiber and micro avalanche photo diode (~APD) detector. A) Schematic diagram of experimental setup. The excitation light was provided by a blue LED off-chip. This light was coupled into an optical fiber that was molded into the PDMS chip. Both the excitation and the detection light were filtered with inexpensive polymeric optical filters. The chip has three levels: a slab of PDMS that contains the channel and the fiber, a polymeric optical filter, and a array of laAPDs embedded in a slab of PDMS. The signal from the I.tAPD was processed off-chip. B) Micrograph of microchannel and optical fiber. The serpentine microchannel was filled with fluorescein. The image shows the size of the detection volume (--25 nL) and that the light is coupled from the fiber into the PDMS with minimal scattering (Chabinyc et al., 2001).

the channel. Figure 11B illustrates the clean optical coupling that is achieved between the fiber and the PDMS. A sheet of polymeric filter is sealed (noncovalently) between the slab of PDMS and the lxAPD that is embedded in PDMS; the filter eliminates stray excitation light. The PDMS makes conformal contact with the filter so the interface between the PDMS and the filter is optically smooth.

3.2. Controlling the cellular environment using soft lithography

3.2.1. Spatially constrained cell culture using IzCP. The ability to pattern SAMs by microcontact printing, and the resulting control over the adsorption of adhesive proteins (for example, the extracellular matrix proteins fibronectin, vitronectin, and laminin), enables the patterning of cells on substrates (Figure 13). Mrksich et al. (1997) used microcontact printing to pattern gold substrates into regions comprising SAMs capped with oligo(ethylene glycol) groups that resist the adsorption of proteins and regions comprising SAMs capped with methyl groups that adsorb proteins. After immersing the substrates in a solution of fibronectin, bovine capillary endothelial cells were found to attach only to the methyl-terminated, fibronectin-coated regions of the patterned SAMs.

586

Soft Lithograpy and Microfluidics

Figure 13. Patterned cells on a surface patterned with micro-contact printing. A) A gold surface was patterned into regions of hexadecanethiolate and undecanethiolate terminated with tri(ethylene glycol). Fibronectin (bright) adsorbed on the hydrophobic squares of hexadecanethiolate but not on the tri(ethyleneglycol)-terminated alkanethiolate (dark). Patterned substrates were soaked in a solution of fibronectin, fixed using paraformaldehyde, and immersed in a solution of anti-human fibronectin IgG and then rinsed. The substrates were then placed in contact with a solution of Texas Red| goat anti-rabbit IgG and mounted in fluoromount-G. B) Bovine capillary endothelial (BCE) cells patterned by culturing on a substrate presenting hydrophobic squares of varying sizes that were coated with fibronectin prior to incubation with cells using the procedure described in (A)) (Chen et al., 1997).

The ability to engineer the properties of the interface between mammalian cells and their substrates using microcontact printing has been useful in understanding the effect of cell shape on cell behavior. Singhvi et al. (1994) used the ability to control cell shape by microcontact printing to investigate the effect of cell shape on cell function. They plated primary hepatocytes on substrates (patterned by microcontact printing) presenting square and rectangular islands of laminin surrounded by non-adhesive regions. Cells attached preferentially to the laminincoated regions, and in most cases, conformed to the shape of the island. The size and shape of cells could therefore be manipulated by changing the size and shape of the adhesive islands, without changing the density of the adhesive protein laminin. The synthesis of DNA was highest on unpatterned surfaces, where the cells could spread without restriction, and a decrease in the size of the cells led to a progressive reduction in DNA synthesis. For the smallest islands (< 1600 /zmZ), less than 3 % of the adherent cells entered the DNA synthesis phase of the cell cycle. The size of the cells also affected the differentiated function of hepatocytes, as reflected by the concentration of secreted albumin in the culture supematant: albumin secretion rates increased as the size of the adhesive island was decreased. This study demonstrated that cell shape could influence cell growth and protein secretion independent of any changes in the density of the adhesive protein laminin. Studies of ceils grown on micropatterned substrates 587

Kane, Stroock, Jeon, et al.

Figure 14. A) Optical micrographs of bovine capillary endothelial cells patterned using MEMPAT on a bacteriological petri dish (see Figure 6 for experimental details). The top left frame shows the state of the cells just after the membrane was removed (7 hours after cells were plated). The other frames show the spreading of the unconstrained cells in the hours following the removal of the membrane. (Ostuni et al., 2000) B) Schematic representation of the channel network used to pattern cells by 3D MIMIC (see Figure 3 for generic structure of channels). C) Fluorescence micrograph of human bladder cancer cells (ECVs; labeled (bright) with 5-chloromethylfluorescein diacetate) and bovine capillary endothelial cells (BCEs; labeled (dim) with 1,1'-dioctadecyl-3,3,3',3'tetramethylindocarbocyanine); the cells were patterned on alternating squares of a checkerboard motif. The cells were cultured for 42 h before the fluorescence micrograph was taken. The channel structure is still on the surface. D) A picture of a confluent layer of cells before the 3D microfluidic stamp was removed. E-F) Pictures taken in fluorescence (E)and in phase contrast (F) show the spreading and growth of the two cell types after the removal of the PDMS stamp. The pictures in (E) and (F) were taken 20 h after removal of the stamp. The three images in D-F are registered; the dotted lines show the relative orientation of the patterns. The BCEs spread more rapidly than the ECVs by a factor of 2-3 (Chiu et al., 2000).

have also indicated that the constrained size and shape of a cell influences whether it lives or dies (undergoes apoptosis) (Chen et al., 1997) as well as its differentiated state (Singhvi et al., 1994; Dike et al., 1999).

3.2.2. Patterned cell culture with variable spatial constraints. M E M P A T and 3D MIMIC offer the possibility of plating cells in regions of well defined size, shape, and location (cf. Section 1.4). Furthermore, with both methods the spatial constraint on the cell(s) imposed by the membrane or channel can be released by peeling the membrane (MEMPAT) or channel network (3D MIMIC) away from the substrate to allow the cells to spread and migrate. Figure 14A shows the movement of bovine capillary endothelial cells across the surface after their release from the regions defined by the holes in a PDMS membrane (see Figure 6 for experimental details) (Ostuni et al., 2000). The rate and degree of the spreading of cells could be used as a simple indicator of the state of cells as they interact with their environment in an cell-based effect sensor. 588

Soft Lithograpy and Microfluidics

Figure 15. Patterned plating of cells from laminar flow. A) Patterning different cell types on the floor of a single microchannel. Chick erythrocytes and E. coli were deposited selectively in their designated lanes by patterned flow of cell suspensions. Adherent ceils were visualized with a fluorescent nucleic acid stain (Syto 9). B) Patterning the delivery of a stain to bovine capillary endothelial cells in a microchannel. A suspension of bovine capillary endothelial cells was introduced into channels that were pre-treated with fibronectin and allowed to attach and spread. After removing non-adherent cells by washing with medium, Syto 9 was allowed to flow through one of the inlets as medium was allowed to flow through the other two (Takayama et al., 1999).

3D MIMIC makes it possible to deposit different cell types in close proximity (~ 200 btm) to one another. This technique may therefore be useful in exploring interactions between different types of cells, and valuable in exploring processes such as morphogenesis, angiogenesis, and differentiation. In the experiment shown in Figures 14B-F, 3D MIMIC is used to create a culture of two cell types (cancer cells and capillary endothelial cells) that are relevant for the study of angiogenesis (Chiu et al., 2000). To achieve an alternating pattern of the two types of cell (Figure 14C), suspensions of cells were allowed to fill the two orthogonal sets of channels in the 3D microfluidic network that is shown schematically in Figure 14B (cf. also Figure 3). The cells were allowed to settle and attach to the substrate for 42 hours with the microfluidic network still sealed against the surface. Figures 14D-F show the evolution of cells after the microfluidic network was removed. This type of experiment could be useful for studying the effect of angiogenic factors released by the cancer cells.

3.2.3. Controlling the total environment of a cell with laminar flow patterning. Laminar flow patterning allows for the patterned deposition of cells and the patterned delivery of reagents to cells that are already present on the walls of a microchannel. Figure 15A illustrates patterning of multiple cell types (erythrocytes and E.coli in this case) on the floor of a microchannel (Takayama et al., 1999). In this experiment, streams of the suspensions of the different types of cell were introduced into the channel through independent inlets. As these streams join in the main channel, they flow laminarly along side one another; the populations of the cells remain segregated in streams and they attach to the floor in the distinct regions covered by the different streams. This technique is a 589

Kane, Stroock, Jeon, et al.

Figure 16. PARTCELL. Patterned flow of trypsin over a bovine capillary endothelial cell on the floor of a microchannel. A) The scheme shows the geometry of the channel. B) Micrograph that shows the cell before treatment with trypsin. C) Micrograph that shows the partially detached cell after treatment (Takayama et al., 1999).

simple alternative to 3D MIMIC to plate different cell types in close proximity to one another for studies of cell interaction. Figure 15B illustrates patterning of the culture medium itself. In this experiment, a single type of cell was plated uniformly on the floor of the microchannel by allowing a suspension to flow through all three inlets. Subsequently, a stream of medium containing a fluorescent stain (Syto-9) was allow to flow through only one of the three inlets of the channel as regular media was allowed to flow through the other inlets. In the main channel, the stain remained localized to a third of the width of the channel as the streams of media flowed laminarly along side one another. As is seen in the micrograph, the only cells that were stained were those covered by the stream that carded the stain. This experiment demonstrates the selective delivery of a reagent to only part of a small population of cells. In this experiment, the untreated cells could act as an internal control or the other inlets could be used to selectively deliver another reagent to a distinct or overlapping sub-population of the cells. For large cells (~ 100 ~tm), laminar flow patterning can be used to deliver reagents selectively to parts of a single cell; we call this technique PARTCELL (Figure 16) (Takayama et al., 1999). In the experiment shown, a suspension of cells was allowed to flow through the entire channel and attach to its floor. Subsequently, distinct streams of medium containing trypsin (a protease that cleaves the attachments of the cells to the surface) and medium without trypsin were allowed to flow through the two inlets of the channel. Figures 16B and 16C show a cell that spanned the interface between the two laminarly flowing streams. In the region covered by the stream that contained trypsin, the attachments that held the cell to the surface were cleaved; the cell retracted from this region. PARTCELL provides the new capability of interacting with a cell on the cellular scale in a non-destructive way (Takayama et al., 2001). 590

Soft Lithograpy and Microfluidics

4. Advantages and Limitations The central advantages of soft lithography as compared to conventional microfabrication methods are its simplicity and its broad compatibility with both organic and inorganic systems. The methods described in this chapter can be used with a minimum of specialized equipment and in a typical laboratory environment (not a cleanroom). The abilities to make many copies of a microstructure by molding, and copies of a chemical pattern by stamping, are useful outcomes of soft lithography. The use of SAMs of thiols allows chemical modification of substrates that are compatible with both electrical (e.g., electrochemical) and optical (e.g., SPR and microscopy) measurements. Soft lithography offers a simple means of fabricating complicated microstructures that integrate organic (filters, membranes, gels), inorganic (glass and silicon surfaces, optical elements, electrodes), and biological elements (medium, proteins, living cells). PDMS, the core material in these systems, has a number of useful properties including low cost, low toxicity, transparency from the visible into the near ultraviolet, chemical inertness, versatile surface chemistry, mechanical flexibility, and durability. PDMS forms conformal seals with most smooth surfaces and can be covalently sealed to itself and to glass after a short (one minute) oxidation step in a low temperature plasma. Soft lithographic methods facilitate the fabrication of microfluidic systems in PDMS. The use of flows in these microfluidic devices to deliver reagents to surfaces with micron-scale resolution is particularly interesting for work with living cells; the flows are non-destructive and the walls of the channels are gas permeable. One limitation of soft lithographic methods comes from the incompatibility of PDMS with many organic solvents; solvents such as dichloromethane and tetrahydrofuran penetrate PDMS and cause it to swell. The use of micro-contact printing has been most successful with thiols on metals. Some work has been done with silanes on silicon oxide, but the procedure is more complicated and the resolution is not as high as with thiols (Jeon et al., 1997). PDMS structures might not be durable enough for certain industrial applications; the low cost of these devices means that they can often serve in "one-time-use" applications.

5. Potential for Improving Biosensor Performance Soft lithographic methods can be used in both the fabrication and the operation of biosensors. Many of the structural, mechanical, and optical elements of biosensors could be fabricated simply and inexpensively in PDMS using soft lithographic techniques. For example, an array of micro-lenses could be molded into one side of a thin slab of PDMS that contains a network of microchannels in relief on the other side. If aligned with the channels, the array of lenses could be used to focus the light from a flood illumination source onto small regions of the 591

Kane, Stroock, Jeon, et al. underlying fluidic network (Wu and Whitesides, 2001). Furthermore, this slab could be aligned and sealed to another slab of PDMS in which an array of optical detectors was embedded. Waveguides and photonic structures could also be fabricated and integrated into sensors using soft lithographic methods (Schueller et al., 1999; Yang et al., 2000). Waveguides can act as highly sensitive detectors of binding at surfaces: the evanescent field of the light in a waveguide can excite fluorescence in molecules bound to the surface of the guide; the emission light is then coupled back into the guide with an efficiency that depends, in part, on the geometry of the guide (Golden et al., 1992). Using soft lithography, this geometry could be controlled precisely. Self-assembled monolayers of thiols on gold-coated glass are an attractive system for use with SPR detection. SAMs of thiols are particularly useful for assays of specific binding of proteins to ligands at the surface. Mixed SAMs of thiols that are terminated with a ligand of interest and thiols that are terminated with oligo(ethylene glycol) offer a effective platform for binding assays; in this system, the degree of specific binding can be tuned by changing the concentration of ligand-terminated thiol and non-specific binding is very low. These are both important characteristics in applications in biosensing. We believe that soft lithographic techniques will be particularly important for the development of cell-based effect sensors in which the detection of an agent is based on the response of a living cell. For example, microcontact printing or MEMPAT could be used to plate cells in an initial condition (e.g. spatially constrained) from which they evolve in a known way as a function of their environment. Laminar flow patterning could be used to treat a single culture of cells or even a single cell in parallel with different solutions of interest. This method has the advantage of having a built-in control.

6. Acknowledgments Work from both the GMW group and the Ingber group was support by NSF DMR 9809363 (MRSEC). The Whitesides group was also supported by DARPA, NIH GM 30367, and NSF ECS 9729405. The Ingber group was also supported by NIH CA 45548 and NIH CA 55833. A number of our co-workers have been responsible for the work reported here. Although they are listed in the references, we are especially grateful to Milan Mrksich, Shu Takayama, Robert Chapman, Emanuele Ostuni, Chris Chen, Lin Yan, and Joydeep Lahiri for major contributions to the program.

592

Soft Lithograpy and Microfluidics 7. References

Allara, D. L. and R. G. Nuzzo, 1985, Langmuir 1, 45. Anderson, J. R., D. T. Chiu, R. J. Jackman, O. Cherniavskaya, J. C. McDonald, H. Wu, S. H. Whitesides and G. M. Whitesides, 2000, Anal. Chem. 72, 3158. Beebe, D. J., J. S. Moore, Q. Yu, R. H. Liu, M. L. Kraft, B. H. Jo and C. Devadoss, 2000, Proc. Natl. Acad. Sci. USA 97, 13488. Bird, R. B., 1960, Transport Phenomena, Wiley, New York. Chabinyc, M. L., D. T. Chiu, J. C. McDonald, A. D. Stroock, J. F. Christian, A. M. Karger and G. M. Whitesides, 2001, Anal. Chem. 73,4491. Chapman, R. G., E. Ostuni, S. Takayama, R. E. Holmlin, L. Yan and G. M. Whitesides, 2000, J. Am. Chem. Soc. 122. Chaudhury, M. K. and G. M. Whitesides, 1991, Langmuir 7, 1013. Chen, C. S., M. Mrksich, S. Huang, G. M. Whitesides and D. E. Ingber, 1997, Science 276, 1425. Chiu, D. T., N. L. Jeon, S. Huang, R. Kane, C. J. Wargo, I. S. Choi, D. E. Ingber, and G. M. Whitesides, 2000, Proc. Natl. Acad. Sci. USA 97, 2408. Chiu, D. T., E. Pezzoli, H. Wu, A. D. Stroock, and G. M. Whitesides, 2001, Proc. Natl. Acad. Sci. USA 98, 2961. Delamarche, E., A. Bernard, H. Schmid, B. Michel and H. Biebuyck, 1997, Science 276, 779. Deng, T., J. Tien, B. Xu and G. M. Whitesides, 1999, Langmuir 15, 6575. Dertinger, S. K. W., D. T. Chiu, N. L. Jeon and G. M. Whitesides, 2001, Anal. Chem. 79, 1240. Dike, L. E., C. S. Chen, M. Mrksich, J. Tien, G. M. Whitesides and D. E. Ingber, 1999, In Vitro Cell. Dev. Biol.-Anita. 35, 441. Dubois, L. H. and R. G. Nuzzo, 1992, Annu. Rev. Phys. Chem. 43,437. Duffy, D. C., J. C. McDonald, O. J. A. Schueller and G. M. Whitesides, 1998, Anal. Chem. 70, 4974. Folch, A. and M. Toner, 2000, Annu. Rev. Biomed. Eng. 2, 227. Golden, J. P., L. C. Shriver-Lake, G. P. Anderson, R. B. Thompson and F. S. Ligler, 1992, Opt. Eng. 31, 1458. Grushka, E., 1974, Bonded Stationary Phases in Chromatography, Ann Arbor Science Publication, Ann Arbor, 937 pp. Grzybowski, B. A., R. Haag, N. Bowden and G. M. Whitesides, 1998, Anal. Chem. 70, 4645. Harrison, D. J., K. Fluri, K. Seiler, Z. Fan, C. S. Effenhauser and A. Manz, 1993, Science 261,895. Ismagilov, R. F., J. M. K. Ng, P. J. A. Kenis, and G. M. Whitesides, 2001, Anal. Chem. 73, 5207. Jacobson, S. C., R. Hergenroder, L. B. Koutny and J. M. Ramsey, 1994, Anal. Chem. 66, 1114. Jeon, N. L., S. K. W. Dertinger, D. T. Chiu, I. S. Choi, A. D. Stroock and G. M. Whitesides, 2000, Langmuir 16, 8311. 593

Kane, Stroock, Jeon, et al. Jeon, N. L., K. Finnie, K. Branshaw and R. G. Nuzzo, 1997, Langmuir 13, 3382. Jo, B. H., L. M. Van Lerberghe, K. M. Motsegood and D. J. Beebe, 2000, J. Microelectromech. Syst. 9, 76. Kane, R. S., S. Takayama, E. Ostuni, D. E. Ingber and G. M. Whitesides, 1999, Biomaterials 20, 2363. Kenis, P. J. A., R. F. Ismagilov, S. Takayama, G. M. Whitesides, S. Li and H. S. White, 2000, Acc. Chem. Res. 33, 841. Kenis, P. J. A., R. F. Ismagilov and G. M. Whitesides, 1999, Science, 285, 83. Kim, E., Y. Xia and G. M. Whitesides, 1995, Nature 376, 581. Kleinfeld, D., K. H. Kahler and P. E. Hockberger, 1988, J. Neurosci. 8, 4098. Kovacs, G. T. A., 1998, Micromachined Transducers Sourcebook, CB/McGrawHill, Boston, 911 pp. Kumar, A., H. A. Biebuyck, N. L. Abbott, and G. M. Whitesides, 1992, J. Am. Chem. Soc. 114, 9188. Kumar, A., H. A. Biebuyck and G. M. Whitesides, 1994, Langmuir 10, 1498. Lahiri, J., L. Isaacs, B. Grzybowski, J. D. Carbeck and G. M. Whitesides, 1999a, Langmuir 15, 7186. Lahiri, J., L. Isaacs, J. Tien and G. M. Whitesides, 1999b, Anal. Chem. 71,777. L6pez, G. P., H. A. B iebuyck, R. H~irter, A. Kumar and G. M. Whitesides, 1993, J. Am. Chem. Soc. 115, 10774. Love, J. C., J. R. Anderson and G. M. Whitesides, 200 l a, MRS Bull. 26, 523. Love, J. C., D. B. Wolfe, H. O. Jacobs and G. M. Whitesides, 200 l b, Langmuir 17, 6005. Manz, A., D. J. Harrison, E. M. J. Verpoorte, J. C. Fettinger, H. Ludi and H. M. Widmer, 1991, Chimia 45, 103. McDonald, J. C., D. C. Duffy, J. R. Anderson, D. T. Chiu, H. Wu and G. M. Whitesides, 2000, Electrophoresis 21, 27. Mrksich, M., 2000, Chem. Soc. Rev. 29, 267. Mrksich, M., L. E. Dike, J. Tien, D. E. Ingber and G. M. Whitesides, 1997, Exp. Cell Res. 235,305. Muller, R. S. and T. I. Kamins, 1986, Device Electronics for Integrated Circuits, 2nd ed., Wiley, New York, 524 pp. Ostuni, E., R. Kane, C. S. Chen, D. E. Ingber and G. M. Whitesides, 2000, Langmuir 16, 7811. Pancrazio, J. J., P. P. Bey, D. S. Cuttino, J. K. Kusel, D. A. Borkholder, K. M. Shaffer, G. T. A. Kovacs and D. A. Stenger, 1998, Sens. Actuators BChem. 53, 179. Patel, N., R. Padera, G. H. W. Sanders, S. M. Cannizzaro, M. C. Davies, R. Langer, C. J. Roberts, S. J. B. Tendler, P. M. Williams and K. M. Shakesheff, 1998, Faseb J. 12, 1447. Piner, R. D., J. Zhu, F. Xu, S. Hong and C. A. Mirkin, 1999, Science 283, 661. Prime, K. L. and G. M. Whitesides, 1991, Science 252, 1164. Qin, D., Y. Xia and G. M. Whitesides, 1996, Adv. Mater. 8,917. Quake, S. R. and A. Scherer, 2000, Science 290, 1536.

594

Soft Lithograpy and Microfluidics Roberts, C., C.S. Chen, M. Mrksich, V. Martichonok, D.E. Ingber and G.M. Whitesides, 1998, J. Am. Chem. Soc. 120, 6548. Schueller, O. J. A., X.-M. Zhao, G.M. Whitesides, S.P. Smith and M. Prentiss, M., 1999, Adv. Mater. 11, 37. Sigal, G. B., C. Bamdad, A. Barberis, J. Strominger and G.M. Whitesides, 1996, Anal. Chem. 68, 490. Sigal, G. B., M. Mrksich and G.M. Whitesides, 1997, Langmuir 13, 2749. Singhvi, R., A. Kumar, G.P. Lopez, G.N. Stephanopolous, D.I.C. Wang, G.M. Whitesides and D.E. Ingber, 1994, Science 264, 696. Takayama, S., R.G. Chapman, R. Kaneand G.M. Whitesides, Eds., 2000, Patterning of Cells and Their Environment, Academic Press, San Diego, 209 pp. Takayama, S., J.C. McDonald, E. Ostuni, M.N. Liang, P.J.A. Kenis, R.F. Ismagilov and G.M. Whitesides, 1999, Proc. Natl. Acad. Sci. USA 5545. Takayama, S., E. Ostuni, P. LeDuc, K. Naruse, D.E. Ingber and G.M. Whitesides, 2001, Nature 411, 1016. Ulman, A., 1991, An Introduction to Ultrathin Organic Films Langmuir-Blogett to Self-Assembly, Academic Press, Boston, 442 pp. Unger, M. A., H. Chou, T. Thorsen, A. Schererand S.R. Quake, 2000, Science 288, 113. Weigl, B. H. and P. Yager, 1999, Science 283, 346. Whitesides, G. M., A.J. Black, P.F. Nealey and J.L. Wilbur, 1996, In The Robert A. Welch Foundation 39th Conference on Chemical Research on Nanophase Chemistry, Houston, TX, 109 pp. Whitesides, G. M. and C.B. Gorman, Eds., 1995, Self-Assembled Monolayers: Models for Organic Surface Chemistry, CRC Press, Boca Raton, 713 pp. Whitesides, G. M., E.S. Ostuni, S. Takayama, X. Jiang and D.E. Ingber, 2001, Ann. Rev. Biomed. Eng. 3, 335. Whitesides, G. M. and A.D. Stroock, 2001, Physics Today 54, 42. Wilbur, J. L., A. Kumar, E. Kim and G.M. Whitesides, 1994, Adv. Mater. 6, 600. Wilbur, J. L. and G.M. Whitesides, 1999, In Nanotechnology, Ed., G. Timp, Springer-Verlag, New York, 331 pp. Wollman, E. W., D. Kang, C.D. Frisbie, I.M. Lorkovic and M.S. Wrighton, 1994, J. Am. Chem. Soc. 116, 4395. Wu, M. H. and G.M. Whitesides, 2001, Appl. Phys. Lett. 78, 2273. Xia, Y. and G.M. Whitesides, 1998, Angew. Chem. Int. Ed. Engl. 37, 550. Yan, L., C. Marzolin, A.Terfort, A. and G.M. Whitesides, 1997, Langmuir 13, 6704. Yang, P., G. Wirnsberger, H. Huang, S.R. Cordero, M.D. McGehee, B. Scott, T. Deng, G.M. Whitesides, B.F. Chmelka, S.K. Buratto and G.D. Stucky, 2000, Science 287,465.

595

Index absorption .... 5, 11, 14, 62, 144, 149, 195, 200, 242, 279, 291, 321, 326, 333-334, 338-339, 343, 345, 348, 354, 356, 361, 363, 435, 438, 444, 467, 539-540, 544-545, 548, 550, 552, 554, 557, 560-563 acceptance cone angle .................. 8, 9 acetate .................... 23, 355-357,546 acetylcholine receptor (AChR) ........ 70, 74-75, 79, 116, 269 acetylcholinesterase ....... 18, 40, 49, 7278, 117 actin ................................. 549, 264 aequorin ........................... 313, 316 affinity...18, 40, 49, 68, 72-72, 79, 88, 101, 114, 127, 161, 190-191, 197, 226, 240, 242, 245, 261-272, 299, 326, 332, 334-335, 344, 346, 351, 356, 360-361, 372-377, 386, 390392, 398, 407-413, 422, 429, 439, 442, 515, 532, 551,555-556 affinity chromatography ....... 270, 299, 373, 555-556 aged gels ................................. 433 aggregation...78, 199, 442, 451, 524, 550, 564 alcohol...193, 271, 345-350, 354, 356, 412, 416, 432-437, 451, 461-466, 472, 483, 489, 511, 517, 522, 546547 AlexaFluor ........................ 167, 168 alkaline phosphatase ..... 13, 21, 34, 436, 438,440 alkaloids atropine .............................. 188 sparteine .............................. 188 alkanethiol ........ 461-463, 466, 467, 578-579

amino acid...44, 101, 187-189, 264, 308-309, 314, 318-320, 332, 375, 391,402, 407, 414 amino acids glutamic acid .... 45, 46, 160, 351, 448-452 glutamine..45, 46, 159-161,315, 390 histidine .................. 103, 355,381 proline .......................... 188, 195 tryptophan ......... 185, 193,314, 559 valine ........................... 185, 188 aminopentanoic acid ................... 350 aminopropanoic acid .................. 350 annihilation ECL ..... 180, 193, 199, 200 ANS ................................ 151,159 anthracene...344, 347, 354-355, 359, 360, 420 antibiotics .......... 29, 45, 185, 188, 341 polymyxin B ..................... 74, 79 valinomycin .......................... 71 antibody ..... 7, 11-13, 16, 19-21, 25-29, 35-36, 39, 42-46, 49-51, 69-78, 83, 86-88, 96, 101-118, 124-134, 137, 139, 141, 150, 158, 186-191, 197, 236-242, 259-263, 270, 271, 288289, 297-300, 332, 337, 374-375, 387-389, 398-409, 428, 429, 433, 436-437, 445-448, 452, 458, 460, 471-480, 483, 532, 550, 551, 555560, 565 antimony ................................. 323 aptamer..23, 49, 116, 141,332, 369-396 DNA ................ 116, 376-377,388 RNA ........................ 88, 116, 377 aptamer beacon ......................... 378 aptazyme .... 116, 375,382-289, 392, 393 aptazyme ligases ......... i ........ 386, 388 arsenic .................................... 323 artificial receptor ...... 86, 331-368, 405, 422 artificial tongue ......................... 362

aluminum ...... 13, 21, 34, 200, 323,444 amine...25, 185, 188, 197, 199, 236, 258, 299, 349, 375, 421, 532, 533, 549, 578 597

Index assay configurations competition assay ...... 20, 36, 74-75, 238, 239,271,272, 551 direct assay ...................... 71, 101 displacement assay ...... 80, 102, 125137,445 sandwich assay ...... 21, 36, 72, 74, 88, 101, 103, 191,237,239, 240, 300, 742, 479,557 association constant(Ka)...333-336, 345361 ATP ..... 14, 38-39, 271, 313, 322, 360, 377-384, 389, 392 atrazine ........... 109, 238,300, 412, 414 attenuated total reflectance ...... 96, 225227 attenuated total reflection method...218, 224-229 avidin...25, 34-35, 69, 71, 81, 102-108, 114, 116, 259, 264, 299, 532, 555 neutravidin ..................... 105,267 streptavidin...88, 1 0 7 , 116, 191, 262-268, 318, 389, 472, 476, 487, 488,492, 559 azide ................................. 355, 361,490 Bacillus anthracis ............. 110-112, 192 Bacillus globigii ............................... 110 bacteria .... 12, 14, 23, 29, 66, 73-76, 96, 240, 243, 307, 312-317, 323-326, 450, 458, 460, 468,470 B a c i l l u s a n t h r a c i s ........ 110-112, 192 B a c i l l u s g l o b i g i i .......................... 110

bicarbonate ....................................... 356 bilayer lipid membrane ............. 457-496 binding constant...86, 1 3 2 , 266-267, 299, 301,350-351,354, 390, 557 binding protein...71, 79, 88, 161, 263269, 308,308, 322-326 protein G ............................ 551 acetylcholine receptor .............. 70, 74, 79, 116, 269, DNA-binding ........................ 79 fluorophore-labeled ......... 307,324 glucose/galactose binding protein .................... 159, 390 glutamine binding protein...315,390 maltose binding protein...38, 315, 371,390, 551,555,560 periplasmic binding protein ....... 151, 160, 390 binding site .... 20, 25, 49, 75, 85, 88, 101-104, 114, 133, 148, 152, 159, 166, 238, 242, 264, 265, 308, 318, 321, 332-337, 342, 351, 366, 375376, 390, 391, 398, 403-408, 412, 417-420, 433, 445, 448, 460, 465, 555 biocompatibility .... 443, 447, 453, 499, 540 bioconjugate .... 278, 279, 298-301,555, 564 biological warfare ............ 274, 374, 422 bioluminescence ... 13, 41, 68, 313, 314 biomolecular interaction analysis kinetics...69, 72, 102, 132, 140, 145, 178, 241,242, 263-269-273, 361, 380, 414, 433, 438, 440, 451, 452463, 464, 492 rate constants ....................... 241,263 biotic receptor .................................. 332 biotin...24, 35, 69-71, 81, 104-108, 114-116, 258-268, 299, 315, 318, 375,469,471,472, 487,492, 532 blood...29, 41-43, 77-79, 96, 107, 111, 129-131, 156, 159-161, 193, 207244, 256, 260, 264-267, 270, 301, 344, 345, 361, 371, 378, 387, 442, 443, 447, 450, 454, 461, 464, 471,

B r u c e l l a a b o r t u s .................. 11 O- 112

40-41, 49, 77, 159-160, 240, 323,388,390, 493, 551,590 F r a n c i s e l l a t u l a r e n s i s .... 83, 110-112 H e l i c o b a c t e r p y l o r i ................ 80, 268 L i s t e r i a ................................ 240, 482 S a l m o n e l l a ..... 88, 152, 240, 300, 301 Y e r s i n i a p e s t i s ................ 77, 110, 111 bacteria-based sensing .............. 312-317 barcode ............................. 561,564-565 benzene ............................... 85,197,324 benzoate .................................... 354, 355 E. c o l i . . . 1 9 ,

598

Index charge coupled device (CCD)...30, 33, 48, 64, 99, 113, 114, 228, 273, 294, 452, 504-507, 530, 550 chemiluminescence... 12, 13, 17, 21, 40, 46, 174, I79, 195-196, 199, 200, 313, chemosensor ............................. 331-368 chip arrays ................................ 385, 388 chloramphenicol transferase ............ 313 chloride ...................... 70, 126, 360, 506 chromophore... 188, 193,313,321,332342, 356, 361,415,416, 560 citrate ............................... 193, 354, 359 CLE-ASV ........................................ 168 CMOS camera ..................... 48, 89, 118 cocaine...74, 83, 128, 134, 138, 377, 378 cofactors...18, 39, 101,266, 314, 332, 380, 351,352, 380 colloid ...... 430, 538-546, 553,554, 563 complementarity .............................. 335 complex reversible ............................. 147 confocal microscopy... 170, 549, 557, 559 conformation...12, 273, 309, 319, 340, 378, 381, 382, 401, 414, 438-442, 458.466, 470, 471,473, continuous sensors ........................... 429 copper...165-168, 317, 323, 380, 438, 532, 541, coreactant ECL ......... 182-183, 188, 193 cotinine ............................................ 353 creatinine ................................. 353, 354 critical angle ................ 8, 9, 58, 97, 245 C r y p t o s p o r i d i u m .............................. 193 cyan fluorescent protein (CFP) ...... 316, 321,322 cytochrome c ...................... 26, 433, 451 d-d absorbance bands ......... 37, 165, 166 D-dimer...43, 110, 111,431,437, 447, 448 dendrimers ....................................... 197 deoxyribozyme ................................ 380

474-475, 488-489, 505, 525-526, 557, 559, 575,580, 592 blue fluorescent protein (BFP) ...... 311, 316, 320-324 Brucella abortus ........................ 110-112 cadmium ................... 317, 323, 546, 553 calcium...24, 169, 264, 269, 318-320, 341, 391, 459, 500, 505, 515, 527, 529 calmodulin .......... 38, 264, 315-322, 391 capacitance ....... 401,467, 483-487, 492 Helmholtz capacitance ................. 484 capillary electrophoresis...200, 387, 584, 587 carbohydrate...ll4, 151,159, 191,265, 271,343-347, 361,390, 450 allose ............................................ 345 arabinose ...................... 317,323,344 cyclodextrins ............................... 342 fructose ................................ 344-347 galactose ........................ 31,265, 345 glucosamine ......................... 347-348 glucose .... 17, 25, 31, 42, 45, 117, 158-160, 185, 188, 195, 315, 323344, 347, 361,362, 371,390, 415,480, 498,522, 531,532 lactulose ....................................... 346 maltotriose ........................... 346-347 mannose ............................... 344, 415 monosaccharide ........................... 344 trehalose .............................. 346, 491 trisaccharide ................................. 347 carbonate .................................. 453, 586 carbonic anhydrase...37, 149-151, 154, 160-166, 241,391 carboxylicacid...351, 353, 420, 533, 578 cells...5, 7, 13-19, 26-29, 40, 41, 45, 49, 50, 88, 96, 108-109, 115, 165, 169, 189, 193-194, 243, 253, 256, 258, 264-267, 300, 310, 313-316, 320-326, 374-376, 388, 405, 428, 429, 431, 451, 454, 482, 492, 498509, 513, 530, 539, 550, 558, 559, 572-581,584, 587-593 599

Index diagnostics...41, 43, 96, 109, 174, 175, 189, 194, 196-200, 270, 271, 371, 376, 387,447,458, 461,473,539 dicarboxylic acid ...................... 350-351 diffusion...24, 43, 47, 86, 88, 115, 165, 181, 199, 236, 243, 301, 303, 323, 415, 5435, 439, 451,466, 492, 523, 527,529, 533,564, 580-582 digoxin ........................ 36, 187,479-482 dimethylaminobenzoate .................... 354 diphenylanthracene (DPA)...175, 178, 180, 185, 198, 199 dissociation rate...132, 133, 241, 261263,266, 267, 271,272, DNA .... 7, 22, 29, 33-36, 71-75, 79-83, 88, 96, 102-105, 109, 115-118, 174, 187, 189-192, 195, 236, 259, 265267, 271, 299-300, 109, 312, 317, 321-326, 372-377, 380, 388, 402, 458,473, 480538, 548-552, 555-558, 561-564, 588 double-stranded DNA...22, 73, 372, 374 single-stranded DNA...36, 266, 267, 372-374, 379 triplex DNA ..................... 75, 80, 266 dopants..,429, 436, 437,440, 431,453, 454, dosimeters ........................ 429, 444, 452 drug screening ................. 115, 135,422, drugs ......................... 130, 135, 138, 188 cocaine74, 83, 128, 134, 138, 377, 378 digoxin ................... 36, 187, 479-482 methamphetamine ............... 238,239 theophylline...36, 109, 382, 384, 389, 483, dynamic range...48, 144, 164, 169, 196, 314, 320, 387, 388, 487, 508, 513, 520 E. coli...19, 40-41, 49, 77159-160, 240, 323, 388,390, 551,590 ELECSYS ........................................ 194 electrochemical detection ................. 450 electrochemiluminescence ....... 173-206

annihilation ECL... 180, 193, 199, 200 electron transfer...148, 149, 174-180, 186, 196, 339, 340, 344, 351, 356357, 465 electrostatic...23-24, 262, 335, 339, 371, 402, 403, 407, 408, 549, 555, 560 ellipsometry ..................... 401,467,487 ELONA ............................................ 387 encapsulation...9, 86, 428-430, 436453, 503,510, 516, 549,560, 565, endoscopes ................................... 41, 49 energy transfer... 11, 26, 36, 43, 68, 148, 150, 158, 164-166, 169, 196, 218, 310, 311, 324, 339, 371, 378, 391,416, 538,559, 560, 563 F/Srster ................................. 149, 170 environmental applications...39, 40, 195, 199 enzyme-linked immunosorbent assay 109, 112-114, 192, 263, 271, 300, 387-389, 473 enzyme-linked oligonucleotide assay (ELONA) ............................. 387 enzymes...7, 13-21, 25, 29, 39, 78, 86, 117, 118, 188, 266, 371, 380, 393, 398-405, 409, 428, 429, 433, 438, 439, 440-443, 448, 451-453, 522, 531,586 acetylcholinesterase... 18, 40, 49, 7278, 117 alkaline phosphatase... 13, 21, 34, 436, 438, 440 carbonic anhydrase...37, 149-151, 154, 160-166, 241,391 chloramphenico! acetyltransferase .................... 313 galactosidase ........... 13, 19, 313,450 glucoseoxidasel7, 25, 31, 42, 195439, 522, 531 glutamate dehydrogenase... 449, 552 horseradish peroxidase ....... 431,437 lactamase ..................................... 188 oxidase... 17, 18, 431,437,439, 453, 522, 531 600

Index polarization ......................... 380, 441 fluorescent resonant energy transfer ...11, 36, 38, 42, 46,, 68, 307, 310311, 314-316, 320-326, 415, 416, 559, 563 fluoride ............................. 354, 356 fluoroimmunoassay ................ 57-142 fluorophores... 11, 35, 36, 57-73, 77, 81, 87, 96-98, 106, 114, 117, 126-130, 144-155, 165-166, 307-326, 332342, 347, 353-354, 360-361, 370371, 374-380, 388, 390, 440, 509, 523, 539, 557,562-565, 580 ABD ..................... 149-151,154, 166 acrylodan ..................... 160, 161, 318 Alexa Fluor ......................... 167, 168 coumarin ............................. 101, 533 CPM ............................................ 319 cyanine...69, 99-102, 111-112, 126, 198, dansylamide ......................... 162-164 fluorescein...17, 29, 35, 43, 49, 69, 78-79, 99-102, 105, 117, 126, 164, 316, 319, 322, 360, 379,412, 415,416, 431,444, 447,500, 510, 549, 557,582, 587 lanthanide ions .................... 146, 419 MDCC ................................. 319, 320 MLCT probe ............................... 159 Oregon green... 166, 508, 522-523, 526, 530-532, POPOP ........................................ 146 pyrene ........... 27, 146, 357,360, 411 rhodamine...69, 101, 316, 322, 360, 437, ruthenium...17, 42, 117, 156, 158, 181, 185, 188, 196, 200, 358,360, 530, 562 Francisella tularensis .......... 83, 110-112 frequency synthesizer ...................... 154 galactosidase ................ 13, 19, 313, 450 green fluorescent protein (GFP) ...... 12, 49, 103 gelation .................................... 432-435, gene gun ................... 502, 507,508, 511

reductase ......................................431 serine proteases ............................ 378 superoxide dismutase ................... 438 estrogen ........................................74, 79 estrogen receptor ..........................74, 79 ethanol...348-350, 354, 412-416, 461466, 472, 483, 489, 511,517, 522 evanescent field . . .57-122, 215, 216, 243, 253, 254, 258, 267, 272, 278282, 287-289, 293, 296, 300-303, 388,593 exciplexes ......................................... 185 explosives...77, 124, 126-136, 193, 551,555 RDX...77, 83, 125, 129, 130, 134137,551,555,556 TNT...77, 83, 128-131, 134-136, 193, 431,437,445,447, 551,555 ferritin ............................... 474-476, 488 flow immunosensor .................. 123-142 fluidics...64-68, 82, 84, 87-89, 112, 116, 124, 132, 140, 286, 301, 571597 fluorescence .... 7-22, 28-43, 50, 59-89, 96-117, 126-128, 132, 139, 141, 144-169, 242, 288, 308-325, 332360, 371, 376-392, 411-418, 435, 437, 440-449, 500, 504-509, 516518, 523-533, 544-552, 557-564, 582, 589, 593 emissive state ............... 175, 179, 339 excited state...12, 14, 144, 147, 174186, 195-197, 338-340,.344, 354, 548,563 exponential decay...98, 1 4 5 , 152, 253,254, 258, ground state...ll, 12, 14, 147, 176, 178, 183, 338, intensity...ll, 32, 68, 80, 81, 106, 117,128, 139, 144-147, 159-160, 168, 308, 325, 356, 357, 371, 376-379, 448, 449, 505, 508, 518, 524, 526, 562, 582 lifetime ............ 11, 69, 144-169, 500 microscopy .......................... 164, 169 multiexponential decay ................ 145 601

Index genetic engineering...16, 19, 38, 71, 114, 140, 307-330 glass...5, 8, 9, 20-24, 70, 100-107, 112, 126-127, 195, 214-215, 219, 222, 228-229, 233, 253, 259, 281, 282, 295, 296, 299, 392, 428, 430, 453, 467, 511, 516, 517, 521, 541-543, 557, 560, 579, 584-585, 592, 593 glucose...17, 25, 31, 42, 45, 117, 158160, 185, 188, 195, 315, 323344, 347, 361, 362, 371, 390, 415, 480, 498,522, 531,532 glucoseoxidase...17, 25, 31, 42, 195439, 522, 531 glucose/galactose binding protein... ...................................... 159,390 glutamate dehydrogenase ......... 449, 552 glutamine binding protein ........ 315, 390 glycoproteins .................................... 261 ricin ................................ 83,265,374 gold film...209, 213-215, 236, 388, 399, 401,458-490, 579-588 graded index (GRIN) lens .................. 99 grarnicidin...458, 459, 466, 467, 470493 grating couplers ................ 217-231,279 green fluorescent protein (GFP)...12, 19, 49, 103, 311,313-316, 319-326, 391 hapten ............... 263, 272, 300, 479-482 Helicobacter pylori ..................... 80, 268 heine proteins ................................... 428 hemoglobin ........................... 43, 44, 442 hexylamine ....................................... 350 homogeneous assay .................... 36, 565 hormones... 109, 149, 192, 236, 391, 474 chorionic gonadotropin ................ 109 diethylstilbestrol ............................ 79 estradiol ......................................... 79 estriol ............................................. 79 estrogen ................................... 74, 79 serotonin ...................................... 103 thyroid stimulating hormone .......................... 473, 374,482

horseradish peroxidase ............. 431,437 host/guest chemistry ........................ 333 HPLC...124, 130, 131, 135, 137, 270, 412-413, 189, 200, 412, 413 human immunodeficiency virus (HIV) ............................. 185, 192, 379 hybrid bilayer membrane .............. 466 hydrogen bonding...104, 335,341,351, 353,354,402, 403,417,441,470, hydrolysis... 19-24, 188, 298, 313, 375, 419, 430-436, 468,522, 583 hydroquinone ................... 199, 351,352 ICP-MS ............................................ 161 immobilization...23-26, 47, 49, 58, 62, 70, 75, 87, 88, 96, 102-110, 114-116, 126, 127, 156, 191, 236, 245, 324, 341, 361, 363, 429, 444, 450, 452, 572 immunoassay...21, 60, 64, 73-76, 86, 101, 111, 113, 124, 134-141, 150, 158, 174, 187, 194, 340, 437, 538539 immunornagnetic separation ............ 191 impedance...458, 561, 466-468, 483486 in vitro selection ............... 372-375,386 indicator...7, 17, 23, 30, 31, 39, 42, 47, 49, 158-164, 169, 413, 435, 444, 453, 499-500, 506, 513, 516-519, 524, 529, 530, 580, 589 ink jet printing .................................. 108 inositol triphosphate ......................... 354 integrated optical waveguide (IOW) ..... 100, 104, 115, 116, 233,234 internal charge transfer (ICT)...33, 340, 465 internal reflection element ............ 100 inverse rnicelle ......................... 542, 544 ion channel...458-461, 466, 471, 473, 476, 484-487,493,507 alamethecin gramicidin...458, 459, 466, 467, 470-493 valinomycin .................................. 71

602

Index mass transport .......... 132, 199,256, 474 matrix effects...96, 113, 114, 118, 136, 138, 140, 325,479, 487, 489 membrane-based biosensor ....... 457-496 membranes...23, 28, 67, 96, 126, 127, 146, 315, 354, 410, 414, 458, 460, 463, 466, 467, 487, 489, 492, 493, 503,572-576, 579, 585,592 membrane-spanning lipid...463, 468, 487, 488 mercury .................... 317, 323,527, 549 metal ions Ca(II)...161264, 313-322, 391,498, 501,513, 516, 527 Cd(II)...149, 161, 166, 546, 557, 560 Co(II)...37, 149, 154, 161,165, 166, 315,322, Mg(II) ......................................... 161 Ni(II)...105, 149, 161, 165, 166, 322 Zn(II)...37, 149, 161,166, 267, 315, 316, 321,498,515, 516 methamphetamine .................... 238, 239 micelle .............................. 198,542-544 microarray ................ 109, 388,539, 564 microenvironment ..... 30, 439,443,453, microfabrication...194, 326, 322, 444, 572, 574, 592 microsphere .......... 33-35, 510, 560, 565 microstructures...435, 492572-578,592 microviscosity miniaturization...26, 46, 48, 76, 82, 96, 236, 245,274, 443,444, 491,516 m-nitrobenzoate ............................... 354 ModE repressor protein ................... 266 modulator acousto-optic device .................... 154 Pockels cell ................................. 154 molecular beacon...12, 23, 35, 36, 81, 86, 88, 110, 111,117, 376, 379 molecular device .............. 333,340, 342 molecular imprinted polymer (MIP) ...................................... 397-426 molecular machine ........................... 333

kinetics...69, 72, 102, 132, 140, 145, 178, 241, 242, 263-266, 269, 273, 361, 380, 414, 433, 438, 440, 451, 452, 463, 464, 492 Michaelis-Menten ................ 438,439 lactamase ........................................... ! 88 laminin .............................. 261,587,588 Langmuir-Blodgett films ............ 71,583 lead...5, 28, 48, 114, 117, 118, 197, 244, 273, 274, 376-380, 421, 442, 453,510, 565,583 lectin ................................................. 191 ligand fishing .................... 259, 270, 271 light detectors ....... .............................. 48 light sources...15, 46, 48, 64, 85, 154, 160, electroluminescent source ........... 154 halogen lamp ........................... 63, 82 lasers diode...48, 64, 76, 82, 89, 99, 152, 154, 166, 167, 420 titanium sapphire .................... 169 LED ....................................... 154, 87 xenon lamp .................................... 64 lipid vesicles ..................................... 466 lipopolysaccharide .................. 74, 77, 79 liposomes...259, 268, 269, 375, 459, 492, 503,505, 510 liquid crystal display ........................ 113 Listeria ...................................... 240, 482 low density lipoprotein (LDL) .......... 263 luciferase...13, 19, 39-41, 313, 323, 324 luminescent ..... 175,436, 538-540, 550 luminol ......................... 12, 40, 184, 195 luminophore... 174, 176, 181, 184, 188, 191,193, 196-198 Lutheran glycoproteins ..................... 261 macrophages ............. 498, 502-505, 510 magnetic beads ......................... 191, 194 maltose binding protein...38, 315, 371, 390, 551,555,560 mass spectrometry (MS)... 124, 130, 134, 239, 265,272 603

Index molecular probe ................ 340, 440, 499 molecular recognition...87, 96, 101, 104, 124, 138, 140, 335-338, 341342, 350, 354, 361, 370-376, 388, 390, 573 monolith ...........................................437 multianalyte...5, 10, 26, 29-36, 46-52, 76, 82, 83, 86, 89, 96, 109-112, 115, 197,234, 235,326, 452 multiphoton microscopy ................... 169 Na+/H§ exchanger regulatory factor .............................................260 NADH .......... 14, 18, 185, 188, 189, 449 nano bio-optrodes ......................... 26-28 nanocrystal ........ 546-553,557-560, 564 nanoparticles...510, 539, 542, 544, 551552, 555-558, 562, 564 nanosensor ........................................499 near field optics ................................444 near-field scanning optical microscopy .............................................. 199 nebulin ...............................................264 Neisseria meningitides ..................... 262 neural networks .................................. 52 nicotinamide ..................................... 188 nicotine .........................79, 83,352, 353 nitrate ................................................354 nitrite ........................................354, 506 nucleic acid...7, 12, 13, 16, 19, 22, 25, 36, 46, 51, 76, 79, 80, 86, 96, 101, 116, 186, 191, 192, 197, 259, 266, 370-381, 385, 393, 401-405, 414, 590 nucleic acid hybridization...22.29, 75, 79-81, 86, 87, 101, 102, 109, 196, 197, 288, 376, 381, 551, 558, 561, 562 nucleotide...29, 30, 80, 81, 102, 117, 266, 370, 372, 381, 384-388, 460, 480, 482, 552, 558,561-562 numerical aperture...9, 61, 65, 157, 508, organometallic ...24, 472, 543-546, 553 ORIGEN ................................... 187, 194 overcoating ............... 543-547, 554, 558

oxalate ...................... 181, 187,195, 199 oxidation...17, 70, 107, 149, 175-183, 198, 200, 298-299, 444, 462, 463, 531,565,575,592 oxidation potential ........................... 149 oxygen...13, 17, 19, 42, 117, 158, 184, 193, 199, 200, 308, 442, 444, 445, 461-464, 508, 509, 515, 522-524, 527-531,583 oxygen sensors .......... 442-444, 527, 531 panning ............................................ 263 PEBBLE ................................... 497-536 acrylamide...500, 506, 508, 511512, 514-516, 526, 530-532 calcium ........................................ 505 decyl methacrylate...508, 516, 520, 529-531 oxygen ................................. 508,523 pH .............................................. 513 sol-gel ........... 508,522-526, 529-530 Peltier ....................................... 257,302 peptides...88, 141, 188, 195, 259, 263, 370, 388-391, 399, 402, 452, 493, 555 pesticides... 18, 19, 39, 40, 109, 130, 137, 141 pH .... 1, 12, 17, 18, 23, 26, 30-32, 42, 46, 51, 72, 75, 78, 104, 117, 141, 164, 184, 185, 266, 269, 298, 300, 325, 333-350, 354, 355, 360, 380, 381, 404-405, 422, 433-437, 440444, 451, 453, 491, 500, 505-520, 525,560, 564 phage display ............................... 49, 88 phenols...40, 199, 349, 354, 356, 400, 406, 416 phenylalaninol .................................. 349 phenylethylamine ............................. 349 phenylglycinol ................................. 349 phloretin ............................................. 71 phosphate...38, 103, 105, 309, 315, 319, 354, 359, 360, 375, 380, 390, 447,491,511,580 phosphorescence ....................... 146, 158 photoablation .................................... 106 604

Index prism coupling... 100, 214, 215, 218, 219, 222, 224, 227-231, 253, 254, 259, 284, 285,293,294, 298 propanal ........................................... 348 protein A .................................... 71, 114 protein G ................ 71, 79, 88, 114, 551 protein stability ................ 435,440, 442 proteinase K ..................................... 261 proteomics ............................... 273,298 protozoa ....................... 74, 76, 374, 431 Cryptosporidium ......................... 193 Giardia .................................... 77, 83 pyridine ............................ 266, 350, 418 pyrolysis ........................................... 544 pyrophosphate .................................. 360 Q-dot ......................................... 537-570 quantum dot ........................ 89, 537-570 quantum efficiency... 12, 69, 101, 146, 150, 179, 196-197, 326, 358, 544, 546, 547,551,563, 564, quaternary structure ................. 378, 381 quenchers...35, 36, 81, 117, 146-149, 339,353,374-380, 560 quenching...ll, 36, 42, 43, 68, 69, 81, 103, 117, 126, 146-154, 158, 165, 166, 193, 199, 200, 308, 340, 347, 354-360, 370, 371, 376-378, 417, 418, 437, 445-447, 500, 501, 523, 526, 557, 559, 560, 563 collisional ..... ............................... 147 dynamic ....................................... 147 quenching constant ..................... 146 static .................................... 147, 151 quenching mechanisms electron exchange... 148, 174-180, 186, 196 electron transfer 1...48, 149339, 340, 344, 351,356, 357,465 paramagnetic ............... 148, 19i, 451 proton transfer ............. 148,338, 356 spin-orbit coupling ...................... 148 unpaired electrons ....................... 148 quinones ........................................... 351 Rad51 protein ................................... 266

photobleaching...47, 61, 68, 69, 87, 435, 540, 563 photodiode .................... 64, 69, 195, 228 photoinduced electron transfer (PET) 339, 340, 344, 347, 351, 351, 354357 photolithography... 106, 108, 280, 292, 572-576, 583-584 photoluminescence 1...76-185, 195-198, 539, 547,551-554, 561-563 photomultiplier tube... 16, 28, 48, 64, 99, 155, 187, 191, 194, 195, 323, 527 photopolymerization ....... ........ 29, 31, 32 physiological markers acetylcholine...18, 40, 49, 50, 70, 74, 79, 116.117,269, 459 cardiac troponin I ......................... 111 D-dimer... 110, 111, 431, 437, 447, 448 myoglobin ........................ 42-44, 111 thrombin ..... 116, 378,388, 392, 444 pKa ................... 164, 337,344, 346, 443 polarization...116, 144, 222, 223, 227, 235,295,296, 338, 356, 440 pollutants .......................... 199, 323, 324 poly(methylmethacrylate) (PMMA)... ................................ 62, 111,195 polydimethyl siloxane...108, 111, 575579, 584-589, 592 polyethylene glycol...70, 107, 299, 375, 516, 517, 522, 524 polymerase chain reaction (PCR) .... 22, 79, 80, 192, 196, 203, 206, 312, 372-376, 386 polymyxin B ................................. 74, 79 polystyrene...62, 67, 70, 88, 104, 108, 191,300, 401,510 pore-matrix interface ........................ 441 porosity...429,432, 433,451,452, 521, potassium...281, 335, 342, 350, 506510, 516-520, 530 preorganization ......................... 335,336

605

Index ratiometric imaging .................. 499, 500 reagentless biosensors...307, 370, 371, 376, 379, 385,388, 392 bio-optrode ...................... 5, 26, 36 real-time measurements...48, 86, 114, 128, 138, 140, 161, 167, 168, 241, 242, 246, 253, 260, 272, 371, 376, 380, 498,504, 527 receptors...17, 19, 69, 72, 74, 78, 88, 103, 116, 245, 246, 260, 268, 303, 333, 334, 343-347, 351-354, 360363, 390-393, 398, 402-405, 422, 460, 480, 555,563-565 acetylcholine receptor...70, 74-75, 79, 116, 269 estrogen receptor ..................... 74, 79 ganglioside ................... 103, 105,270 serotonin receptor ........................ 103 reduction...18, 79, 147, 148, 175-184, 200, 262, 391, 439, 449, 461-464, 468,493,523,576, 588 reference dye .... 499, 508, 513, 529, 530 refractive index...9, 57-62, 96-100, 113, 207-224, 228-237, 242-244, 253-257, 271-273, 278-289, 295, 298, 301,302, 436 regeneration of surfaces... 18, 75, 79, 87, 102, 116, 125, 262, 272, 274, 296 resonant angle ........................... 254, 255 resonant mirror ................. 242, 253-276 response time...24-27, 43, 44, 47, 80, 270, 272, 324, 325, 348, 433, 435, 443,476, 482, 499, 509, 527, 529 reverse transcriptase polymerase chain reaction ................................. 312 reverse transcription ................. 372, 374 ribozymes ................. 370, 375,379-385 allosteric .............................. 381,383 hammerhead ........................ 380-386 RM cuvette...159, 227, 254-259, 262264, 267, 271,532 rotational mobility ............ 438, 441-442 rubrene .............................................. 185 Salmonella . . . . . . . . . . 88, 152, 240, 300, 301

scanning electrochemical microscopy ..................................... 199,492 selectivity... 19, 37, 51, 124, 161, 169, 174, 188, 253, 260, 314, 322-326, 332, 335, 343-351, 356-361, 378, 414, 419, 506, 512, 513, 516, 519, 520, 525,531,560 semiconductor...48, 155,232, 280, 282, 295, 298, 538-543, 546, 548, 552553,559, 562, 565 serine protases .................................. 378 sialic acid ......................... 268, 414, 415 signaling aptamers...370, 376-381,389392 sodium...102, 193, 281,335, 342, 350, 444, 490, 491, 506, 511, 517, 519, 520, 542 sol-gel...24, 40, 43, 69, 200, 282, 286, 302, 427-456, 498, 502, 508, 511, 521-530, sol-gel dopants cx-naphtholphthalein .................... 443 Ru(II) 4, 7-diphenyl 1, 10phenanthroline ..................... 437 seminaphthorhodamine- 1 carboxylate ..................... 437, 443 solubility165, 179, 361, 523-524, 529, 542,555 solvatochromism .............................. 338 stability...16, 24, 46, 48, 72, 87, 88, 101-102, 113, 176, 180, 196, 225, 261, 264, 266, 269, 273, 287, 288, 291,295, 341,345, 354, 375,, 401, 405, 407, 438-443, 451, 453, 459, 460, 463,489, 490, 530, 565 Sterne-Volmer equation strokes ...................................... 443,447 succinimidyl esters .... 186, 258, 532533 sugar tweezer ................................... 345 sulfonamide ............................. 164, 165 dansylamide ......................... 162-164 superoxide dismutase ....................... 438 supramolecular chemistry ........ 335, 341

606

Index bungarotoxin ..................... 74, 79, 83 cholera toxin...83, 110-112, 268-270 Naja naja toxin ....................... 79, 83 neurotoxins ........... 40, 74, 78, 79, 83 ricin ............... 83, 111, 112, 265, 374 SEA ............................................. 271 SEB...83, 88, 110, 111, 238, 239, 551,555 transcription...312-314, 318, 325, 372, 374, 379,402 transition metals ............... 335, 341,342 transition state analogs ..................... 375 triethylphosphite .............................. 350 tri-n-propylamine (TPrA) ....... 182-185, 189, 191,194, 197-199 tripeptide .................................. 353, 362 twisted internal charge transfer...339, 355 urea .... 75, 102, 116, 299, 300, 353-356 urine...42, 74, 77-78, 111, 130, 131, 187, 195, 261,362, 461 vaccines ........................................... 270 valinomycin ................................. 74, 79 vascular endothelial growth factor (VEGF) ................................ 387 vibro-stirring ............................ 256, 257 viruses .............................................. 270 virus-like particles (VLP) ................ 270 V-number .......................... 59-62, 66, 87 whole-cell sensing...313-314, 317, 325, 326 xerogel ............................................. 433 yellow fluorescent protein (YFP)...315, 316, 319, 321,322, 325 Yersinia pestis .................... 77, 110, 111 zinc finger ........................ 316, 322, 391

surface plasmon resonance (SPR) .... 68, 96, 113-116, 207-252, 253,254, 297, 298,401,436, 576, 592, 593 surfactants...198, 199, 269, 453, 511, 542 synthesis...326, 337, 341, 343, 362, 372, 374, 275, 400-403, 421, 430, 432, 435, 450, 467, 526, 538-546, 552,553,588 Tamm-Horsfall protein (THP) .......... 261 T-dependent immunogen .................. 262 tethered lipid ..... 458,460, 463,473, 484 tethered lipid bilayer ......................... 484 theophyUine...36, 109, 382, 384, 389, 483 thyroid stimulating hormone (TSH) ....................... ...... 473,474, 482 time-correlated single photon counting .............................................. 152 time-resolved fluorescence lifetime phase...ll, 37, 101, 131, 138, 140, 148-168, 183, 191, 196, 219-223, 226, 227, 235, 236, 241, 244, 255, 278, 282-98, 302, 350, 353, 361, 372, 401, 417, 463, 483-486, 499, 523, 524, 527, 542, 559-560, 563, 577, 588, 589 phase fluorometers .......... 152, 155 frequency ................ 151,152, 157 TCSPC .................................... 152 titin ................................................... 264 TM modes ................................ 273, 295 TMPD ....................................... 179, 180 toluene .............................. 317, 353, 408 total internal reflection (TIR)...8, 9, 96117,215,254, 280, 388 toxins...79, 83, 127,265,269, 274, 317, 324, 452, 493, 498, 505, 551 botulinum toxin ................... 110, 112

607