Proteins, Cells and Materials
VSP Utrecht, The Netherlands, 2003
Proteins, Cells and Materials
Dr. John L. Brash
VSP BV P.O. Box 346 3700 AH Zeist The Netherlands
©VSP BV 2003
First published in 2003
ISBN 90-6764-381-5
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior permission of the copyright owner.
Printed in The Netherlands, on acid-free paper, by Ridderprint BV, Ridderkerk.
CONTENTS Foreword
ix
Introduction
xi
Letter
xiii
I. Proteins PEO-like plasma polymerized tetraglyme surface interactions with leukocytes and proteins: in vitro and in vivo studies M. Shen, L. Martinson, M.S. Wagner, D.G. Castner, B.D. Ratner and T. A. Horbett
3
Limits of detection for time of flight secondary ion mass spectrometry (ToF-SIMS) and X-ray photoelectron spectroscopy (XPS): detection of low amounts of adsorbed protein M.S. Wagner, S.L. McArthur, M. Shen, T.A. Horbett and D.G. Castner
27
Photoimmobilization of biomolecules within a 3-dimensional hydrogel matrix X. Cao and M.S. Shoichet
49
Fibrinogen adsorption by PS latex particles coated with various amounts of a PEO/PPO/ PEO triblock copolymer M. Bohner, T.A. Ring, N. Rapoport and K.D. Caldwell
63
Effects of base material, plasma proteins and FGF2 on endothelial cell adhesion and growth P.A. Underwood, J.M. Whitelock, P.A. Bean and J.G. Steele
77
Acoustics of blood plasma on solid surfaces M. Andersson, A. Sellborn, C. Fant, C. Gretzer and H. Elwing
95
Development of small alginate microcapsules for recombinant gene product delivery to the rodent brain C.J.D. Ross and P.L. Chang
107
II. Cells Integrin α2β1 on rat myeloma cells modulates interaction of α4β1 integrin with vascular cell adhesion molecule-1 but not fibronectin B.M.C. Chan, V.L. Morris, D. Hangan-Steinman, B. Jarvie, M. Cialacu, J. Laansoo, G. Hunter, W. Wan and S. Uniyal
119
vi
Contents
Interactions of corneal epithelial cells and surfaces modified with cell adhesion peptide combinations L. Aucoin, C.M. Griffith, G. Pleizier, Y. Deslandes and H. Sheardown
137
Polyelectrolyte multilayer films modulate cytoskeletal organization in chondrosarcoma cells D. Vautier, V. Karsten, C. Egles, J. Chluba, P. Schaaf , J.-C. Voegel and J. Ogier
153
Improved blood compatibility and decreased VSMC proliferation of surface-modified metal grafted with sulfonated PEG or heparin H.J. Lee, J.-K. Hong, H.C. Goo, W.K. Lee, K.D. Park, S.H. Kim, Y.M. Yoo and Y.H. Kim 173 Characterization of poly(ethylene oxide) brushes on glass surfaces and adhesion of Staphylococcus epidermidis H.J. Kaper, H.J. Busscher and W. Norde
187
Tissue-culture surfaces with mixtures of aminated and fluorinated functional groups. Part 2. Growth and function of transgenic rat insulinoma cells (βG I /17) J.R. Bain and A.S. Hoffman
199
III. Materials Elastomeric biodegradable polyurethane blends for soft tissue applications J.D. Fromstein and K.A. Woodhouse
229
Influence of surface morphology and chemistry on the enzyme catalyzed biodegradation of polycarbonate-urethanes Y.W. Tang, R.S. Labow, I. Revenko and J.P. Santerre
245
Cardiopulmonary bypass technology transfer: musings of a cardiac surgeon F.D. Rubens
267
Surface analysis methods for characterizing polymeric biomaterials K. Merrett, R.M. Cornelius, W.G. McClung, L.D. Unsworth and H. Sheardown
283
Glucose binding to molecularly imprinted polymers H. Seong , H.-B. Lee and K. Park
313
The effect of oxidation on the enzyme-catalyzed hydrolytic biodegradation of poly(urethane)s R.S. Labow, Y. Tang, C.B. McCloskey and J.P. Santerre
327
Novel dendrimer based polyurethanes for PEO incorporation X. Duan, C.M. Griffith, M.A. Dubé and H. Sheardown
343
Identification of biodegradation products formed by L-phenylalanine based segmented polyurethaneureas S.L. Elliott, J.D. Fromstein, J.P. Santerre and K.A. Woodhouse
367
Contents
vii
Bioresorbable polymeric stents: current status and future promise R.C. Eberhart, S.-H. Su, K.T. Nguyen, M. Zilberman, L. Tang, K.D. Nelson and P. Frenkel
389
Tissue-culture surfaces with mixtures of aminated and fluorinated functional groups. Part 1. Synthesis and characterization J.R. Bain and A.S. Hoffman
403
Deterioration of polyamino acid-coated alginate microcapsules in vivo J.M. van Raamsdonk, R.M. Cornelius, J.L. Brash and P.L. Chang
419
Water structure around enkephalin near a GeO2 surface: a molecular dynamics study A.M. Bujnowski and W.G. Pitt
441
Towards practical soft X-ray spectromicroscopy of biomaterials A.P. Hitchcock, C. Morin, Y.M. Heng, R.M. Cornelius and J.L. Brash
463
A new vascular polyester prosthesis impregnated with cross-linked dextran D. Machy, P. Carteron and J. Jozefonvicz
483
This page intentionally left blank
FOREWORD
This Festschrift celebrates the 65th birthday of John L. Brash. The field of biomaterials owes much to the contributions that he has made over a 40 year career that includes research done primarily at McMaster University, but also at E. I. DuPont de Nemours and Company and Stanford Research Institute. John has had a broad influence internationally as his work on cardiovascular biomaterials and plasma protein adsorption has received worldwide recognition. The editors, publisher and the many colleagues who are participating in this multi-issue Festschrift wish him well on this special occasion. We anticipate John Brash maintaining a leadership role in biomaterials research and we look forward to his continuing contributions to the field. These articles were first published in 2002– 2003 in several issues of Journal of Biomaterials Science, Polymer Edition. We are most pleased now to make the entire Festschrift available in book form. M. VERT T. TSURUTA S. L. COOPER
This page intentionally left blank
Introduction It is with a great deal of pleasure that I write an introduction in this special issue dedicated to Dr. John Brash on the occasion of his 65th birthday. To say that John’s work over the years has advanced the field of polymeric biomaterials would be an understatement. His sphere of influence and the respect that people have for him and for his work is global. He has always been willing to extend his expertise and to open his facilities to others and has had numerous collaborations over his career. Former students and post doctoral fellows have established strong research programs in academia and made significant industrial contributions. He is on editorial boards for a number of scientific journals, is the journal editor for Colloids and Surface B: Biointerfaces, and through membership on grants and other committees, has helped to bring biomaterials research to the forefront in Canada. To date, John has supervised 13 doctoral students and 18 master’s students, many of whom are represented in this journal. He has published more than 100 papers in refereed journals and is co-editor of two books on the behaviour of proteins at interfaces. His influence in the biomaterials community has been recognized by the Clemson Award for Basic Research from the US Biomaterials Society in 1994, an honorary degree, Docteur Honoris Causa, from the Univeristé Paris Nord in 1996 and by an appointment as University Professor, McMaster University’s highest research honour, in 2001. John received his BSc (1958) and PhD (1961) in the area of kinetics of free radical polymerization from the University of Glasgow. He spent two years as a postdoctoral fellow at the National Research Council of Canada, where he worked on photochemistry and chemical kinetics, and one year working with E. I. Du Pont de Nemours and Company. In 1964, he joined the Stanford Research Institute where he began working on a program on artificial organ research, which included the development of new membranes for hemodialysis and investigations of the mechanism of surface induced thrombogenesis. The former was supported by the Artificial Kidney Program of the NIH and the latter by the Artificial Heart Program, National Heart and Lung Institute. In subsequent work for the Medical Devices Applications Program, John directed the development of a system of biocompatible polyether urethane elastomers for use in circulatory assist devices. Fabrication technology for balloons and cups was also worked out. Another phase of the project was concerned with sulfonated polymers of controlled content and distribution of functional groups.
xii
Introduction
In 1972, John joined the faculty of McMaster University with a joint appointment in Chemical Engineering and Pathology, with teaching responsibilities in Chemical Engineering and a research program in biomedical engineering that bridges the two disciplines. At McMaster, John has conducted research in the field of polymeric biomaterials with an emphasis on cardiovascular implant materials. Much of his efforts have focused on detailed studies of blood-surface interactions, particularly those of plasma proteins. His multifaceted approach recognizes the importance of hemodynamics, transport and surface phenomena in the gross effects that result from blood surface contact. He has spent sabbaticals at the Centre de Recherches sur les Macromolécules, Centre Nationale de la Recherche Scientifique in Strasbourg France (1978– 1979), the University of Linkoping, Linkoping Sweden (1994– 1995?) and at CSIRO in Melbourne and Sydney Australia in 2001. In a communication to me, Allan Hoffman’s comment was that John is too young to retire. This echoes the sentiments of many of his colleagues and collaborators. I hope he does not really retire for a very long time to come. HEATHER SHEARDOWN
Letter Dear John: It is now ten years ago since you contributed a thirteen-page article to my Festschrift, so it is about time for me to contribute at least one letter to yours. In that article you introduced flow, and since then I spent the rest of my natural research life trying to do the same, with Ed Leonard and others. Now, more or less but not entirely away from it all, I sometimes wonder how much in this area we have contributed to the welfare of humanity, and how much humanity would care. I think I earned about one million dollars in my entire lifetime studying interface reactions, so I like to imagine myself going from door to door over the entire United States, telling people that for two cents I can tell them what happens to their fibrinogen when their blood touches glass. I would never have collected the needed funds, so that I would never have been able to tell even myself what happens. Obviously therefore, I am grateful to our government for the support that allowed me to enjoy myself all those years, but especially for the occasion to meet and work with you. Nobody else I know works with that precious combination of intensity and relaxation we should all work with. To me, one of the most enjoyable episodes was our study of ring formation by IgG between a lens and a slide. I had already published this, and asked you to try it out. You never succeeded. The problem appeared not to be the source or preparation of IgG or of the substrate, or you. When you visited us at Columbia I was — thank heavens! — able to demonstrate the phenomenon for you, you then took some of our IgG and I think an anodized tantalum coated slide as well with you to Toronto, and presto! it still did not work when you tried it in your lab. The whole thing would have been frustrating if experienced with almost anybody else, but for me, with you, it was just fun. Thank you, thank you. I was happy to see you have broadened your field of interest. In the name of that same humanity, I thank you. And I hope we will meet more often! Warmest regards, also to all those around you, LEO VROMAN Fort Worth, Texas, 7/4/01
This page intentionally left blank
Part I
Proteins
This page intentionally left blank
PEO-like plasma polymerized tetraglyme surface interactions with leukocytes and proteins: in vitro and in vivo studies MINGCHAO SHEN 1 , LAURA MARTINSON 1 , MATTHEW S. WAGNER 2 , DAVID G. CASTNER 1,2 , BUDDY D. RATNER 1,2 and THOMAS A. HORBETT 1,2,∗ 1 Department 2 Department
of Bioengineering, University of Washington, Seattle, WA 98195, USA of Chemical Engineering, University of Washington, Seattle, WA 98195, USA
Received 5 December 2001; accepted 18 March 2002 Abstract—Polyethylene oxide (PEO) surfaces reduce non-specific protein and cell interactions with implanted biomaterials and may improve their biocompatibility. PEO-like polymerized tetraglyme surfaces were made by glow discharge plasma deposition onto fluorinated ethylene propylene copolymer (FEP) substrates and were shown to adsorb less than 10 ng/ cm2 of fibrinogen in vitro. The ability of the polymerized tetraglyme surfaces to resist leukocyte adhesion was studied in vitro and in vivo. Polymerized tetraglyme and FEP were implanted subcutaneously in mice and removed after 1 day or 4 weeks. Histological analysis showed a similar degree of fibrous encapsulation around all of the 4-week implants. Darkly stained wells were present in the fibrous tissues at the tissue-material interface of both FEP and tetraglyme. Scanning electron micrographs showed that in vivo macrophage adhesion to polymerized tetraglyme was much higher than to FEP. After 2-hour contact with heparinized whole blood, polymorphonuclear leukocyte (PMN) adhesion to polymerized tetraglyme was much higher than to FEP, while platelet adhesion to polymerized tetraglyme was lower than to FEP. When PMNs isolated from blood were suspended in 10% autologous plasma, cell adhesion to polymerized tetraglyme was higher than to FEP; however when the cells were suspended in heat inactivated serum, cell adhesion to FEP was higher than to polymerized tetraglyme. The surface chemistry of polymerized tetraglyme did not change after 2-hour blood contact, but displayed nitrogen functional groups after 1-day implantation and became slightly degraded after 4-week implantation. The surface chemistry of FEP did not change significantly after blood contact or implantation. Loosely bound proteins such as fibrinogen on polymerized tetraglyme may contribute to the adhesion of PMNs and macrophages and ultimately to fibrous encapsulation (the foreign body response) around the implants. Key words: PEO; RF plasma deposition; surface modification; non-fouling; foreign body response; macrophage; protein adsorption. ∗ To
whom correspondence should be addressed. E-mail:
[email protected]
4
M. Shen et al.
INTRODUCTION
Macrophages are considered the central mediators of the foreign body response to implanted biomaterials [1, 2]. Macrophages are found adherent to the surface of many implants [3, 4], where they can undergo cytokine release [5] and fusion to form foreign body giant cells [6]. The characteristic foreign body capsule that forms around implants is thought to be triggered by macrophage adhesion and activation [2]. Polymorphonuclear leukocytes (PMNs) may also be involved in the foreign body response since they mediate acute inflammatory response to injuries and can also adhere to implanted biomaterial surfaces [7, 8]. After contacting synthetic materials such as polyurethane and Dacron, PMNs can secrete reactive oxygen species such as superoxide ions that can induce surface degradation [9, 10]. Due to the highly adhesive nature of macrophages [11] and PMNs [12], it has been difficult to engineer biomaterial surfaces that can inhibit cellular attachment, fibrous encapsulation, and the foreign body response to implanted materials. Cell adhesion proteins that adsorb non-specifically to implanted biomaterials are believed to mediate leukocyte adhesion to surfaces [2, 11, 13]. Polyethylene oxide (PEO)-like surface coatings have been considered promising to prevent nonspecific protein adsorption and cell adhesion to biomaterial surfaces [14, 15]. Although some degree of reduced cell adhesion has been shown on PEO-containing surfaces [16, 17], others did not achieve decreases in leukocyte adhesion [18, 19], probably because protein adsorption to those surfaces was not low enough. Several in vivo studies on PEO-containing surfaces fail to demonstrate reduced cell-surface interactions. For example, in a cage implant system, macrophage cell density on triblock copolymers containing increasing amounts of PEO was much higher than the control surface [20]. Polyurethane polymers containing PEO units did not reduce cell adhesion when implanted intramuscularly in rats [5]. On poly(propylene fumarate-co-ethylene glycol) hydrogels, macrophage density was higher on surfaces with greater PEO content, and there was much higher neutrophil adhesion to PEO-containing materials than the poly(propylene fumarate) control [21]. Truly non-fouling surfaces that completely inhibit protein adsorption are needed to control biological interactions with biomaterials [22]. For example, in previous studies in our lab, we found that platelet adhesion still occurred on surfaces with only 10 ng/ cm2 of adsorbed fibrinogen [23]. We have also applied this criteria to monocytes and shown that greatly reduced short term monocyte adhesion in vitro on a PEO-like glow discharge plasma polymerized tetraglyme surface was only obtained when the surfaces reduced protein adsorption to less than 10 ng/ cm2 [24]. In this study we investigated whether such highly non-fouling plasma polymerized tetraglyme surfaces would resist protein adsorption and leukocyte adhesion in vivo. Cell adhesion to and fibrous encapsulation of subcutaneously implanted materials were measured. Cell adhesion to polymerized tetraglyme surfaces from whole blood and from washed PMNs in plasma or serum was also evaluated. In addition, surface analysis of the explanted materials was performed to determine if surface degradation contributed to the fouling of biomaterial surfaces.
PEO-like plasma polymerized tetraglyme surface interactions
5
MATERIALS AND METHODS
Buffers and reagents Phosphate buffered saline (PBS), May-Grünwald stain, Giemsa stain, Triton X-100, Histopaque 1077, heat-inactivated fetal bovine serum, trypan blue, RPMI 1640, and dextran were purchased from Sigma (St. Louis, MO). Glutaraldehyde (3%) solution for scanning electron microscopy (SEM) contained 0.1 M cacodylate, pH = 7.4. Tetraglyme (CH3 O(CH2 CH2 O)4 CH3 ) was purchased from Aldrich (Milwaukee, WI). Surfaces The FEP film, tetrafluoroethylene-hexafluoropropylene-copolymer (CF(CF3 )CF2 (CF2 -CF2 )n )m , was a gift from E. I. Du Pont de Nemours & Co., Inc. (Circleville, OH). The film was cut into 6.4 mm diameter disks and cleaned by successive 10-min ultra-sonications in methylene chloride (× 2), acetone (×2) and methanol (×2). Radio-frequency glow discharge (RFGD) plasma deposition of tetraglyme on FEP was done as previously described [24]. Tetraglyme was plasma polymerized on FEP for 1 min at 80 W then 10 min at 10 W. The polymerized tetraglyme surfaces were found to inhibit protein adsorption and monocyte adhesion [24]. Fibrinogen adsorption to polymerized tetraglyme from a 0.03 mg/ml fibrinogen solution was less than 10 ng/ cm2 . Prior to implantation, FEP and plasma deposited tetraglyme samples were soaked in 70% ethanol overnight, rinsed in sterile PBS, and screened for endotoxin contamination using a Pyrotell® Limulus Amebocyte Lysate (LAL) assay kit (Associates of Cape Cod Inc., Falmouth, MA) sensitive to 0.06 endotoxin Unit (EU)/ml. The samples were found to contain less than 0.06 EU /ml of endotoxin. Implantation All surgical instruments were cleaned and autoclaved prior to surgery and soaked in 70% ethanol between animal surgeries. Strict aseptic techniques were used for material implantation. Healthy, 6-week old C57Bl6 male mice (B&K Universal, Kent, WA) were used. The animals were anesthetized with isoflurane gas (Abbott Laboratories, North Chicago, IL) for the 4-week implantation study, or with a cocktail of ketamine (Fort Dodge Animal Health, Fort Dodge, IA) and xylazine (Phoenix Pharmaceutical, Inc., St Joseph, MO) for the 1-day implantation study. The 6.4 mm samples were surgically implanted subcutaneously on the backs of mice, each mouse implanted with two different materials, polymerized tetraglyme and FEP (one disk of each). The incision site was prepared by shaving and swabbing with Betadine. A single 1– 1.5 cm incision was made midline on the back of a mouse and two subcutaneous pockets were created by blunt dissection lateral to each side of the incision. One implant was placed in each pocket, and the incision was closed with sterile wound
6
M. Shen et al.
clips. Animals were allowed to recover prior to returning to housing cages. Animals were given food and water ad lib for the remainder of the study. Explantation After 1 day or 4 weeks, animals used for subcutaneous implant studies were sacrificed by CO2 asphyxiation. The wound clips were removed, and animal skin was cut open to access the implants. Fibrous capsules surrounded all of the 4-week implants. For histological analysis, the 4-week samples were retrieved en-block so as not to disturb the biomaterial/ host tissue fibrous capsule interface. The intact tissues were fixed in methyl Carnoy’s for 2 days at 4◦ C. The fixed tissues were prepared using standard paraffin-embedding and sectioning techniques and stained with hematoxylin and eosin (H&E) or Masson’s Trichrome. The cross section of the tissue-implant interface was viewed with an upright Nikon light microscope. For each cross section, the thickness of the fibrous capsule on both the skin and muscle sides of the implants was measured at 4 locations (1 mm apart) near the midportion of the material and the average capsule thickness was calculated for each sample. To determine the surface chemistry of implanted samples and adherent cell type and number, additional FEP and polymerized tetraglyme samples were implanted and removed after 1 day or 4 weeks. During explantation, the fibrous capsule tissues around the implants were nicked open with the tip of forceps to expose the edges of the samples. The implants were not integrated with the surrounding fibrous tissues and were easily pulled out of the capsule by the forceps. The implants were kept in sterile PBS until further analysis of the morphology of the adherent cells and material surface chemistry as described below. Cell adhesion from whole blood Whole blood from healthy human donors was drawn by venipucture into Vacutainers containing sodium heparin (Becton Dickinson, Franklin Lakes, NJ). Polymerized tetraglyme samples made at 10, 20, or 60 W of plasma power and untreated FEP samples were incubated in whole blood in a 48-well polystyrene plate (Corning Inc., Corning, NY) for 2 h at 37◦ C. The samples were dip-rinsed successively in three petri dishes containing PBS to wash away bulk blood cells. No red blood cells were visible on the samples after rinsing. The adherent cell number was characterized by measuring total lactate dehydrogenase (LDH) activity associated with the samples. The LDH method was previously used to measure platelet and leukocyte adhesion to surfaces [11, 23, 24]. Adherent cells were lysed with 1% Triton X-100 for 20 min and the total LDH activity was measured with a cytotoxicity kit (Roche, Indianapolis, IN) at optical density 490 nm with reference optical density at 650 nm [11]. The morphology of the adherent cells and material surface chemistry were also analyzed as described below.
PEO-like plasma polymerized tetraglyme surface interactions
7
PMN adhesion Whole blood from healthy human donors was drawn by venipuncture into Vacutainers containing EDTA (Becton Dickinson). Human PMNs were isolated according to a published method [25]. The cells were isolated with the following steps: density gradient centrifugation over Histopaque 1077, sedimentation in dextran, and hypotonic cell lysis of residual red cells. The cells were >95% viable as determined by Trypan blue exclusion. The isolated cells were suspended at 2 × 106 /ml in RPMI 1640 media containing 10% FBS or autologous plasma. Autologous plasma was prepared from heparinized blood from the same blood donor. PMN suspensions of 400 µl were added to the 48-well plate holding FEP or polymerized tetraglyme samples and incubated at 37◦ C in humidified 5% CO2 atmosphere for 1 h. The surfaces were washed by filling and aspirating the wells 3 times with RPMI 1640. Adherent cell number was determined with the LDH assay using a calibration curve based on PMNs suspended in PBS. Cellular image analysis The samples from the implant and whole blood studies were analyzed with microscopy to determine the type and amount of cells adherent on the surfaces. For SEM, the samples were dip-rinsed in PBS three times and fixed with glutaraldehyde for 24 h at 4◦ C. The samples were rinsed in deionized water three times, dehydrated in a graded series of ethanol-water solutions, critical point dried, sputter coated with gold-palladium, and visualized with a JEOL JSM-6300F scanning electron microscope. The acceleration voltage was 15 keV. For light microscopy, the samples were placed in a 24-well plate, fixed for 6 min with methanol and air-dried. Fixed cells were stained with May-Grünwald for 1 min, rinsed with PBS twice, stained with Giemsa for 5 min, and rinsed with deionized water twice. The samples were dried overnight and imaged with an inverted Nikon light microscope. Surface chemistry analysis Surface analysis was done at the National ESCA and Surface Analysis Center for Biomedical Problems (NESAC /BIO) at the University of Washington. The implants and the samples exposed to whole blood were dip-rinsed in PBS and sonicated for 20 min in PBS containing 1% Triton X-100, except that the 1-day implanted samples were exposed to Triton X-100 without sonication. The samples were further washed 3 times in deionized water and air-dried before being analyzed with ESCA and ToF-SIMS. Control FEP and polymerized tetraglyme samples were incubated with 0.03 mg/ ml of fibrinogen for 2 h at 37◦ C, rinsed with PBS and deionized water, and air-dried before surface analysis. ESCA analysis was done with an SSX-100 surface analysis instrument (Surface Science Instruments or SSI, Mountain View, CA) using a monochromatized AlKα
8
M. Shen et al.
X-ray source and an electron flood gun for charge neutralization. The photoelectron take-off angle for the analysis was 55◦ . The hydrocarbon peak for the ESCA spectra acquired in this study was referenced to a binding energy 285.0 eV. Surface elemental compositions and high resolution carbon functional group compositions were calculated using SSI software. For ToF-SIMS analysis, positive ion ToF-SIMS spectra were acquired using a PHI Model 7200 reflectron time of flight secondary ion mass Spectrometer (Physical Electronics, Eden Prairie, MN) with an 8 keV Cs+ primary ion source. The primary ion dose was maintained below 1012 ions/ cm2 to insure that all spectra were acquired under static SIMS conditions [26]. Spectra were acquired at three different spots for each surface. Spectra were acquired over a horizontal raster size + + of 200 × 200 µm. The positive spectra were calibrated to the CH+ 3 , C2 H3 , C3 H5 , + C7 H7 peaks. A pulsed electron flood gun was used for charge neutralization for all samples. The major peaks associated with polymerized tetraglyme and FEP were identified by their m/z ratios. Statistical analysis To determine the significance of the data, unpaired F -tests and 2-tail t-tests were performed with Microsoft® Excel 98 (Microsoft Corp., Redmond, WA). The level of significance was set at p < 0.05.
RESULTS
Histological tissue analysis Figure 1 shows the cross section view of the implanted materials and the surrounding fibrous tissues. The thickness of the implants is indicated with 2-headed arrows. Sample preparation for histology caused an artificial separation (the open space) between the tissues and the implanted material surface. A thin, dense layer of fibrous/ granulous tissues was observed around both FEP (Fig. 1a) and polymerized tetraglyme (Fig. 1b). The fibrous tissues ran parallel to the smooth implant surfaces. The implants were not integrated into the surrounding connective tissue. No new blood vessel growth from the fibrous tissues towards the implants was observed. For FEP, darkly stained cells were present at the material-tissue interface but not directly on the implant surface (Fig. 1a). For polymerized tetraglyme, darkly stained adherent cells were observed on one side of the tetraglyme sample (Fig. 1b). There were also stained cells within the fibrous tissues. Next to the thin fibrous tissue and away from the implants was more loosely arranged connective tissue. On the skin side, hair follicles were scattered among the collagen fibers. On the muscle side, adipose tissues were occasionally present. The collagenous fibrous capsules around implanted FEP and polymerized tetraglyme were about 30– 40 µm in thickness and were not significantly different from
PEO-like plasma polymerized tetraglyme surface interactions
Figure 1. Collagen tissues and cells around implanted (a) FEP (b) plasma polymerized tetraglyme shown by Masson’s Trichrome staining. The tissues on the top side of FEP and the bottom side of polymerized tetraglyme are the skin side tissues. The hair follicles are not shown. The thickness of the implants is indicated by the arrows. The spaces between the implants and tissues are artifacts from histology preparations. Scale bars represent 100 µm.
9
10
M. Shen et al.
Figure 2. The thickness of fibrous tissue capsules around 4-week implanted FEP (white bars) and polymerized tetraglyme (black bars) was measured with light microscopy. The capsule thickness from the skin and the muscle side was measured. The data represent mean ± S.D.; n = 4 FEP or polymerized tetraglyme samples that were implanted in 4 mice. No differences in capsule thickness were found.
Figure 3. Representative scanning electron micrographs of adherent cells on FEP and polymerized tetraglyme surfaces: Adherent cells on FEP (a) or polymerized tetraglyme (b) after 1-day subcutaneous implantation in mice. Adherent cells on FEP (c) or polymerized tetraglyme (d) after 4-week subcutaneous implantation in mice. Enlarged micrograph of adherent cells on polymerized tetraglyme (e) after 4-week subcutaneous implantation. Adherent cells on FEP (f) or polymerized tetraglyme (g) after 2-hour blood contact. Scale bars represent 10 µm in a, b, e, f, g, and 100 µm in c, d.
PEO-like plasma polymerized tetraglyme surface interactions
11
each other (Fig. 2). The fibrous capsules displayed no significant difference in thickness on either the skin or the muscle side of the implants. Leukocyte adhesion in vivo Cell adhesion to FEP and polymerized tetraglyme samples was observed with SEM (Fig. 3a– g) and light microscopy (Fig. 4a– f). Cell adhesion to 1-day subcutaneously implanted samples was lower on FEP than on polymerized tetraglyme, as shown by SEM (Fig. 3a and b) and light microscopy (Fig. 4a and b). The adherent cells on polymerized tetraglyme appeared to be mainly PMNs, indicated by stained cells that displayed multi-lobular polymorphic cell nuclei. No platelets or red blood cells were observed on either surface. Cell adhesion to 4-week implanted samples was also much higher on tetraglyme than on FEP (Fig. 3c and d). In fact, there were hardly any adherent cells on the FEP surface. The adherent cells on polymerized tetraglyme appeared to be mainly macrophages, as the cells were about 15– 20 µm in diameter, well spread, and displayed ruffled cell membrane characteristic of macrophages (Fig. 3e). No foreign body giant cells were observed on the surfaces. Leukocyte adhesion in vitro After contact with whole blood, the adherent cells on FEP were mainly platelets, while the adherent cells on polymerized tetraglyme were mainly PMNs. The
Figure 3. (Continued).
12
M. Shen et al.
Figure 4. Representative light microscopy images of Giemsa-May-Grunwald stained FEP and polymerized tetraglyme samples: Adherent cells on FEP (a) or tetraglyme (b) after 1-day subcutaneous implantation in mice. Adherent cells on FEP (c) or polymerized tetraglyme (d) after 2-hour blood contact. Adherent PMNs on FEP (e) or polymerized tetraglyme (f) after 1-hour adhesion from 10% autologous plasma. Scale bars represent 100 µm.
PEO-like plasma polymerized tetraglyme surface interactions
13
Figure 5. Cell adhesion to FEP or polymerized tetraglyme surfaces after 2-hour heparinized blood contact. Adherent cells were measured by the LDH method. The data represent mean ± S.D.; n = 4 samples. The LDH activity on 10 W or 60 W polymerized tetraglyme samples was significantly higher than that on FEP control (∗ p < 0.01). The LDH activity on 20 W polymerized tetraglyme was significantly higher than that on 10 W tetraglyme (∗∗ p < 0.01).
adherent cells on FEP were flat (Fig. 3f) compared to the cells on polymerized tetraglyme, which were fully spread and showed ruffled cell membrane (Fig. 3g). Leukocyte adhesion from whole blood to FEP was much lower than to polymerized tetraglyme (Fig. 4c and d). PMNs appeared to be the main type of cells adherent on polymerized tetraglyme, judging by the multi-lobular polymorphic nuclei staining and cell size of about 10 µm diameter (Fig. 4d). The adherent cells on FEP were much smaller and were mainly platelets. No platelets were found on polymerized tetraglyme. No red blood cells were observed on either surface. The LDH method was used to measure total adherent cell LDH content on FEP and three types of polymerized tetraglyme surfaces deposited at different powers. The amount of fibrinogen adsorbed from 0.03 mg/ml of fibrinogen to polymerized tetraglyme samples deposited at plasma powers of 10, 20 or 60 W was 3.4, 17, and 209 ng/ cm2 respectively. Total LDH activity was the lowest on FEP, and increased in order of 60 W, 10 W, and 20 W polymerized tetraglyme samples (Fig. 5). Due to the presence of various cell types, the LDH activity only approximated the total cell number on the surfaces. Since the LDH content in platelets is much lower than in leukocytes, the LDH data agreed with observations above that more leukocytes adhered to polymerized tetraglyme than to FEP.
14
M. Shen et al.
Figure 6. Adhesion of human PMNs in 10% heat-inactivated FBS to FEP (white bars) or plasma polymerized tetraglyme (black bars) after 1 hour. Cell adhesion was measured by the LDH method. The data represent mean ± S.D.; n = 5 samples.
PMN adhesion from washed cell preparations depended on the type of proteins in the media. When 10% autologous plasma was used in cell media, adhesion after 1-hour was higher to polymerized tetraglyme than to FEP (Fig. 4e and f). The corresponding LDH values that correlated to cell number on FEP and polymerized tetraglyme were 0.71 and 1.29 respectively. However, when the cells were in 10% heat inactivated FBS media, adhesion after 1-hour was slightly higher to FEP than to polymerized tetraglyme with (p = 0.067, 2-tail t-test) or without (p = 0.276, 2-tail t-test) preadsorbed fibrinogen (Fig. 6). However, these differences were not statistically significant. ESCA analysis FEP and polymerized tetraglyme surfaces were characterized by ESCA (Table 1a– b and Fig. 7a– d). FEP contained 34.5% carbon and 65.5% fluorine (Table 1a) and displayed fluorinated carbon groups [24]. Polymerized tetraglyme contained 70.5% carbon and 29.4% oxygen (Tables 1b) and displayed mainly ether carbon [24]. After 2-hour adsorption with 0.03 mg/ ml of fibrinogen, FEP displayed 10.9% nitrogen and 13.7% oxygen (Table 1a), showing that a large amount of proteins adsorbed to FEP. Polymerized tetraglyme resisted fibrinogen adsorption, as the surface chemistry remained the same after 2-hour exposure to the fibrinogen solution (Table 1b). After 2-hour contact with whole blood, sonication in Triton X-100 and rinsing, the surface chemistry of both FEP and polymerized tetraglyme was similar to untreated samples (Tables 1a and b). Very small amounts of fluorine
PEO-like plasma polymerized tetraglyme surface interactions
15
Table 1. ESCA elemental compositions of FEP (a) and polymerized tetraglyme (b). The control samples were not treated. The samples adsorbed with fibrinogen were rinsed with PBS and deionized water. The samples implanted or exposed to blood were sonicated in Triton X-100 and rinsed in deionized water, except that the 1-day implant samples were not sonicated when lysed with Triton X-100. The data represent mean ± SD; n = 4 samples except polymerized tetraglyme adsorbed with fibrinogen Materials
%C
%O
%F
%N
(a) FEP control 2-hr in fibrinogen 2-hr in blood 1-day implant 4-week implant
34.5 ± 0.5 57.4 ± 1.3 33.1 ± 0.1 40.3 ± 0.9 32.8 ± 0.3
nd 13.7 ± 0.5 nd 0.5 ± 0.9 nd
65.5 ± 0.6 18.0 ± 1.9 66.9 ± 0.1 58.7 ± 1.6 67.2 ± 0.3
nd 10.9 ± 0.2 nd 0.5 ± 0.1 nd
(b) Polymerized tetraglyme control 2-hr in fibrinogen 2-hr in blood 1-day implant 4-week implant
70.5 ± 0.9 70.5 70.1 ± 1.4 70.4 ± 2.1 64.1 ± 2.8
29.4 ± 0.9 29.5 28.5 ± 0.5 24.8 ± 3.9 24.7 ± 3.6
nd nd 0.8 ± 1.2 0.6 ± 1.2 5.0 ± 5.2
nd nd 0.6 ± 0.7 4.1 ± 3.4 6.3 ± 2.8
Note: nd means that this element was not detected during analysis.
and nitrogen were detected on polymerized tetraglyme (Table 1b). The ESCA C1s spectrum of blood-exposed polymerized tetraglyme surfaces was similar to untreated polymerized tetraglyme surfaces (Fig. 7a). After 1-day subcutaneous implantation of FEP and Triton X-100 treatment to remove any adherent cells, the carbon content increased and fluorine content decreased (Table 1a). There was only a slight increase of oxygen nitrogen on FEP. Compared to untreated FEP, the hydrocarbon peak in the ESCA C1s spectrum of 1-day implanted FEP increased from 0 to 25% (Fig. 7b). Implanted polymerized tetraglyme displayed a significant increase of nitrogen content on the surface (Table 1b), indicating adsorption of proteins to the polymerized tetraglyme surface. There was a small fluorine signal from the polymerized tetraglyme surface, similar to that of blood-contacting polymerized tetraglyme. The oxygen content of polymerized tetraglyme decreased from 29.4% to 24.8%. Compared to untreated polymerized tetraglyme that normally contained 65– 75% of ethercarbon content at 286.5 eV binding energy [24], the polymerized tetraglyme implanted for one day contained about 40% ethercarbon (Fig. 7c). After 4-week subcutaneous implantation of FEP followed by sonication and Triton X-100 lysis, the carbon and fluorine content remained the same as untreated FEP (Table 1a). There was no detectable nitrogen or oxygen on the implanted FEP. A small hydrocarbon peak was observed on implanted FEP (Fig. 7d). After 4 weeks of implantation, there was a significant increase of nitrogen and fluorine content on polymerized tetraglyme (Table 1b). The presence of 6.3% nitrogen on
16
M. Shen et al.
(a)
(b) Figure 7. The resolved ESCA C1s spectra of polymerized tetraglyme after 2-hour contact with whole blood (a), FEP after 1-day implantation (b), polymerized tetraglyme after 1-day or 4-week implantation (c), and FEP after 4-week implantation (d). The samples implanted or exposed to blood were sonicated in Triton X-100 and rinsed in deionized water, except that the 1-day implant samples were not sonicated when lysed with Triton X-100.
the polymerized tetraglyme surface indicated adsorption of proteins. The fluorine signal suggested slight degradation of the polymerized tetraglyme coating on FEP substrate, although the amount of fluorine was low compared to control FEP.
PEO-like plasma polymerized tetraglyme surface interactions
17
(c)
(d) Figure 7. (Continued).
The polymerized tetraglyme oxygen content decreased from 29.4% to 24.7% after implantation. The ethercarbon content was about 40%, similar to 1-day implanted samples (Fig. 7c).
18
M. Shen et al.
ToF-SIMS analysis The ToF-SIMS spectra of polymerized tetraglyme and FEP (Fig. 8a– d) contained a large number of peaks. The major peaks associated with polymerized tetraglyme and FEP were identified by their m/z ratios and are listed in Table 2a– b. Overall, the peaks from whole blood exposed or implanted polymerized tetraglyme include
(a)
(b) Figure 8. ToF-SIMS positive ion spectra: FEP (a) and polymerized tetraglyme (b) after whole blood contact for 2 h, 1-day implanted polymerized tetraglyme (c), and 4-week implanted polymerized tetraglyme (d). The samples implanted or exposed to blood were sonicated in Triton X-100 and rinsed in deionized water, except that the 1-day implant samples were not sonicated when lysed with Triton X-100.
PEO-like plasma polymerized tetraglyme surface interactions
19
chemical species from both polymerized tetraglyme and nitrogen-containing groups derived from adsorbed proteins (Table 2a). The nitrogen-containing peaks include m/z at 56 (C3 H6 N+ ), 70 (C4 H10 N+ ), and 86 (C5 H12 N+ ). The peaks from whole blood exposed or implanted FEP contained mainly fluorinated carbons and no nitrogen-containing groups (Table 2b). The ToF-SIMS spectrum of FEP after 2-hour blood contact displayed mainly fluorinated carbon peaks (Fig. 8a) that are typical of untreated FEP. The spectrum of polymerized tetraglyme after 2-hour blood contact (Fig. 8b) was similar to untreated
(c)
(d) Figure 8. (Continued).
20
M. Shen et al.
polymerized tetraglyme, but a nitrogen-containing peak at 70 m/z (C4 H10 N+ ) was detected. For 1-day implanted samples, the ToF-SIMS spectrum of FEP was similar to the control FEP and displayed mainly fluorinated carbon peaks and no nitrogencontaining peaks. The 1-day implanted polymerized tetraglyme samples displayed nitrogen-containing peaks at m/z 70 and 86 (Fig. 8c). The relative peak intensity Table 2a. Identities of ToF-SIMS positive ion spectra major peaks from polymerized tetraglyme samples before or after implantation Peaks (m/z)
Chemical structure
15 27 29 31 41 43 45 55 56 59 70 71 86 89 101 103
CH+ 3 C 2 H+ 3 C 2 H+ 5 CH3 O+ C 3 H+ 5 C 2 H3 O+ CH3 -O-CH+ 2 C 4 H+ 7 C 3 H6 N+ CH3 -O-CH2 -CH+ 2 C4 H10 N+ C 4 H7 O+ C5 H12 N+ CH3 -O-CH2 -CH2 -O-CH+ 2 C 5 H9 O+ 2 CH3 -(O-CH2 -CH2 )+ 2
Table 2b. Identities of major ToF-SIMS positive ion spectra peaks from FEP samples before or after implantation Peaks (m/z)
Chemical structure
12 31 50 69 93 100 119 131 169 181
C+ CF+ CF+ 2 CF+ 3 C3 F + 3 C2 F + 4 C2 F + 5 C3 F + 5 C3 F + 7 C4 F + 7
PEO-like plasma polymerized tetraglyme surface interactions
21
at m/z 59 (CH3 OCH2 CH+ 2 ) associated with the tetraglyme monomer structure was still the highest. For 4-week implanted samples, the ToF-SIMS spectrum of FEP was similar to the control FEP and displayed mainly fluorinated carbon peaks and no nitrogencontaining peaks. Implanted polymerized tetraglyme displayed nitrogen-containing peaks at m/z 56, 70, and 86 (Fig. 8d). The relative peak intensity at m/z 59 associated with the tetraglyme monomer structure was lower than the nitrogen containing peak at m/z 70.
DISCUSSION
Although PEO-containing polymers that reduced protein adsorption have been shown to reduce platelet adhesion [16, 17, 27], reducing leukocyte adhesion to surfaces has been more difficult to achieve [19– 21]. In a recent study, we showed that plasma polymerized tetraglyme effectively reduced monocyte adhesion in vitro [24], while older studies from this lab showed that platelet adhesion to this type of surfaces was also greatly reduced [28]. However, despite ultra low uptake of fibrinogen (<10 ng/ cm2 ) to polymerized tetraglyme in vitro [24], this study showed that polymerized tetraglyme failed to resist leukocyte adhesion during subcutaneous implantation, whole blood contact, or exposure to blood plasma containing PMNs. In addition, unpublished studies of these materials in collaboration with L. Tang showed they failed to resist leukocyte adhesion after one day intraperitoneal implantation. These results are in agreement with others who showed increased macrophage density [20] and PMN presence [21] to implanted PEO-containing polymers. Several factors may explain increased leukocyte adhesion to polymerized tetraglyme in vivo, despite its resistance to protein adsorption in vitro. First, most in vitro macrophage adhesion studies do not adequately simulate in vivo conditions in regards to the role of adsorbed adhesion proteins. Thus, although fibrinogen is known to be an important mediator of macrophage and PMN adhesion to biomaterial surfaces [8, 11] and is present in blood plasma and in the body fluids, it is typically absent from macrophage cell culture studies in vitro. Most macrophage adhesion studies use serum, from which fibrinogen is absent [29– 31]. Since macrophages and PMNs can bind fibrinogen via β2 integrin receptors [32– 34], either fibrinogen or fibrin could play important roles in leukocyte adhesion to implanted materials, as depletion of fibrinogen significantly reduced leukocyte adhesion to 1-day implants [8]. One study also showed that compared to PMNs in PBS buffer, PMNs suspended in 10% plasma exhibited significantly increased attachment to a PEO-containing polyurethane surface [35]. When we used non-traditional in vitro culture conditions, we got better agreement between our in vitro and in vivo results with tetraglymes. Thus, our study showed that leukocyte adhesion from PMNs suspended in 10% autologous plasma, from whole blood, or 1-day subcutaneous implantation to polymerized tetraglyme and FEP are in good agreement. Cell adhesion to polymerized tetraglyme was higher
22
M. Shen et al.
than to FEP in all these cases in which fibrinogen is present. However, when serum was used during cell adhesion, the opposite was observed, as PMN adhesion to polymerized tetraglyme was slightly lower than to FEP (Fig. 6). Together with previous studies that showed enhanced leukocyte adhesion to PEO-like surfaces in plasma or in vivo [5, 20, 21, 35], our study suggests that the use of plasma rather than serum during in vitro leukocyte studies can provide data that are more relevant to in vivo conditions. A second reason for the divergence between in vivo and in vitro results for tetraglymes may be related to how the role of fibrinogen is simulated in the in vitro studies, and in particular whether fibrinogen was present in the buffer media during macrophage incubation. When we tested for the role of adsorbed fibrinogen in macrophage adhesion to tetraglyme in our previous in vitro studies, we did so by preadsorbing the surfaces with fibrinogen or plasma, but the samples were rinsed in buffer before exposure to monocytes [24]. This artificial step may have rinsed away most of the bulk and loosely bound proteins on polymerized tetraglyme before monocytes encounter the surface. Such rinsing is frequently performed in protein adsorption studies in vitro, but did not occur in the experiments using plasma in media, whole blood, or in vivo animals for cell adhesion studies. Fibrinogen was always present in the cell suspension in these studies. Since fibrinogen was never artificially rinsed away from the substrate surface, it could mediate leukocyte adhesion. We now turn to a comparison of our results to others in the literature. The observation of increased leukocyte adhesion to PEO-containing surfaces from complex biological media is not limited to polymerized tetraglyme but was reported for several other PEO surfaces [20, 21]. Furthermore, a recent study in our lab on self-assembled monolayers (SAMs) showed that PMN adhesion from 10% plasma to PEO-terminated SAM (HS-(CH2 )11 -(OCH2 CH2 )4 -OH) was higher than to methyl-terminated SAM (HS-(CH2 )11 -CH3 ): 2748 ± 100 vs. 2019 ± 129, p < 0.01 (2-tail t-test). The reasons for such leukocyte behavior are not clear. It has been proposed that PEO surfaces that are not completely, 100% non-fouling may enhance the biological activity of the small amounts of the adhesive proteins that still do manage to adsorb, so that even though the amount of protein is in fact lower than on the control, the total biological activity is greater on the PEO surface [36, 37]. This is partly because the adsorbed adhesion proteins on PEO surfaces are prevented from undergoing as much spreading onto the substrate by the PEO chains that partly cover the surface, whereas a surface with no PEO chains allows the adhesion protein to spread fully over the substrate and consequently become less active. The cell adhesion data to several different types of polymerized tetraglyme surfaces from whole blood support this hypothesis. Plasma polymerized tetraglyme at 20 W adsorbed only 17 ng/ cm2 of fibrinogen but led to the highest degree of cell adhesion (Fig. 5). Most importantly, we still do not know the requirement for biomaterial surfaces to be non-fouling in vivo. The small amount of adsorbed fibrinogen on polymerized tetraglyme (less than 10 ng/ cm2 ) can still
PEO-like plasma polymerized tetraglyme surface interactions
23
be sufficient to support leukocyte adhesion from whole blood or in vivo. This was at least true for platelet adhesion, which occurred on surfaces with only 10 ng/ cm2 of adsorbed fibrinogen [23]. How do our data for PMNs compare to previous studies? Adherent PMNs were found on polymerized tetraglyme after both whole blood exposure and 1-day implantation in mice, indicating that acute inflammatory responses play important roles in host response to implanted PEO surfaces. This agrees with a previous study in which PMNs were found to be the main type of adherent cells on titanium after exposure to whole blood [38]. Since PMNs are known to secrete reactive oxygen radicals [10], they may cause degradation of polymerized tetraglyme coating on FEP. Surface analysis showed that coating degradation did not occur after 2-hour blood contact, but signs of degradation appeared after 1-day and 4-week implantation as indicated by the fluorine signals from the FEP substrate. Very low amounts of proteins adsorbed to polymerized tetraglyme after 2-hour contact with blood, as the polymerized tetraglyme surface chemistry remained essentially unchanged. However the polymerized tetraglyme did not resist protein adsorption after 1-day or 4-week implantation, shown by the presence of nitrogen signals on the surface. Surface degradation could be one of the factors that lead to adsorption of proteins to polymerized tetraglyme in vivo. Our studies also indicate differences in how proteins affect monocyte adhesion to FEP in vitro and in vivo. In vitro fibrinogen adsorption to FEP from 0.03 mg/ ml fibrinogen measured with I-125 fibrinogen was 20 fold higher than to polymerized tetraglyme, and FEP surfaces adsorbed with fibrinogen displayed nitrogencontaining peaks when analyzed with ToF-SIMS [24]. However, in the in vivo implant study, FEP had much lower leukocyte adhesion than polymerized tetraglyme after 1-day and 4-week implantation. This agrees with our previous finding that monocyte adhesion from media containing serum was much lower to FEP than to tissue culture polystyrene [24]. Surprisingly, the surface chemistry of FEP after blood contact or implantation (followed by Triton X-100 treatment) was the same as before any treatment. Triton X-100 probably did not affect the adsorbed protein significantly, since polymerized tetraglyme samples exposed to the same rinsing procedure displayed nitrogen signals. It is unclear why no nitrogen signals were detected on FEP after whole blood or implantation studies, since fibrinogen adsorbed FEP displayed 10.9% of nitrogen. The detection limits of ESCA or ToF-SIMS for nitrogen signals on fluorinated substrates may contribute to such phenomena, as unpublished data by Wagner et al. showed that ToF-SIMS could only detect about 100 ng/ cm2 of adsorbed proteins on Teflon. Nonetheless, FEP was not a non-fouling material. Platelet adhesion was observed on FEP in contact with whole blood, but not on the polymerized tetraglyme surface (Figs 3, 4). In addition, histological analysis showed that leukocytes were present in the fibrous tissues surrounding the FEP implant, although they did not appear to have direct contact with the substrate (Fig. 1a).
24
M. Shen et al.
We began these studies with the idea that prevention of monocyte adhesion would reduce the foreign body reaction to tetraglymes, but our results and others to date, emphasize that the mechanisms of host response to implanted materials remain unclear [4]. We cannot yet correlate leukocyte adhesion with the degree of fibrous encapsulation around implants. More studies are needed to understand the mechanism of fibrous encapsulation around soft tissue implants. It is possible that activated macrophages in the vicinity of implants can stimulate fibroblasts to secrete collagenous tissues to wall off the implants [4, 39]. Whether macrophages adhered directly on the implant surface (polymerized tetraglyme) or stayed in the fibrous tissues without directly contacting the substrate (FEP), the degree of fibrous encapsulation was similar for both materials. It seems that the adhesion of macrophages to implanted materials was not necessary for the macrophages to orchestrate the foreign body response. Understanding the role of macrophages in the foreign body response and engineering surfaces that can modulate macrophage responses still remain difficult tasks but are critical for biomaterial scientists to develop truly biocompatible materials [3, 4]. Recent studies using porous materials to modulate macrophage response to surfaces [40] may provide new insight to develop biocompatible materials that can control macrophage behavior, resist fibrous encapsulation, and promote wound healing. The excellent non-fouling behavior of polymerized tetraglyme surfaces was reaffirmed in our in vitro studies. Although such surfaces may not be appropriate in vivo unless further understanding leads to their improvement, their in vitro stability, ease of application and extremely non-adhesive nature make them viable candidates for use in non-implant applications such as diagnostics, biosensors, MEMS and array technologies. In conclusion, this is the one of the first studies to explant long-term soft-tissue implants for direct adherent cell identification and material surface analysis. PEOlike plasma polymerized tetraglyme that reduced fibrinogen adsorption to less than 10 ng/ cm2 in vitro failed to reduce PMN or macrophage adhesion in the presence of plasma, whole blood, or when implanted in mice. Leukocyte adhesion to implants was higher to polymerized tetraglyme than to control FEP. Fibrous encapsulation to 4-week subcutaneously implanted polymerized tetraglyme and FEP were similar to one another. Implanted polymerized tetraglyme became fouled and the coating was slightly degraded. The use of plasma rather than serum during leukocyte adhesion in vitro produced data that were in agreement with leukocyte adhesion from whole blood and in vivo. To make truly biocompatible materials, it is important to elucidate the role of loosely bound adhesive proteins on polymerized tetraglyme that may lead to increased PMN and macrophage adhesion and also to understand the correlation of leukocyte adhesion to the foreign body response. Acknowledgement We gratefully acknowledge financial support from the NSF through UWEB EEC9529161. We thank the assistance from Dr. Stephen Golledge and NESAC /BIO
PEO-like plasma polymerized tetraglyme surface interactions
25
(NIH grant RR01296 from the National Center for Research Resources) for support of the surface analysis experiments, Stephanie Lara and the Department of Pathology for SEM, the Department of Histopathology for histology, and Dr. Kip Hauch for light microscopy. We also thank Iris Garcia, Dr. Sally McArthur, Winston Ciridon, and Yuguang Wu for their valuable technical assistance and discussions, and thank the donors for blood donations.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33.
J. M. Anderson and K. M. Miller, Biomaterials 5, 5 (1984). J. M. Anderson, Problems Gen. Surg. 11, 147 (1994). N. P. Ziats, K. M. Miller and J. M. Anderson, Biomaterials 9, 5 (1988). B. D. Ratner, J. Biomed. Mater. Res. 27, 837 (1993). J. A. Hunt, G. Meijs and D. F. Williams, J. Biomed. Mater. Res. 36, 542 (1997). Q. Zhao, N. S. Topham, J. M. Anderson, A. Hiltner, G. Lodoen and C. R. Payet, J. Biomed. Mater. Res. 25, 177 (1991). L. Tang, A. H. Lucas and J. W. Eaton, J. Lab. Clin. Med. 122, 292 (1993). L. Tang and J. W. Eaton, J. Exp. Med. 178, 2147 (1993). K. H. Lee and C. C. Chu, J. Biomed. Mater. Res. 49, 25 (2000). S. S. Kaplan, R. E. Basford, E. Mora, M. H. Jeong and R. L. Simmons, J. Biomed. Mater. Red. 26, 1039 (1992). M. Shen and T. A. Horbett, J. Biomed. Mater. Res. 57, 336 (2001). C. Eriksson and H. Nygren, J. Lab. Clin. Med. 137, 56 (2001). T. A. Horbett, Cardiovasc. Pathol. 2, 137S (1993). G. R. Llanos and M. V. Sefton, J. Biomater. Sci. Polymer Edn 4, 381 (1993). D. Leckband, S. Sheth and A. Halperin, J. Biomater. Sci. Polymer Edn 10, 1125 (1999). A. Kidane, G. C. Lantz, S. Jo and K. Park, J. Biomater. Sci. Polymer Edn 10, 1089 (1999). K. Park, H. S. Shim, M. K. Dewanjee and N. L. Eigler, Biomater. Sci. Polymer Edn 11, 1121 (2000). S. M. O’Connor, S. J. Patuto, S. H. Gehrke and G. S. Retzinger, Ann. NY Acad. Sci. 831 (1987). C. R. Jenney and J. M. Anderson, J. Biomed. Mater. Res. 49, 435 (2000). B. Ronneberger, W. J. Kao, J. M. Anderson and T. Kissel, J. Biomed. Mater. Res. 30, 31 (1996). L. J. Suggs, M. S. Shive, A. Garcia, J. M. Anderson and A. G. Mikos, J. Biomed. Mater. Res. 46, 22 (1999). L. G. Cima, J. Cell Biochem. 56, 155 (1994). W. B. Tsai, J. M. Grunkemeier and T. A. Horbett, J. Biomed. Mater. Res. 44, 130 (1999). M. Shen, Y. V. Pan, M. S. Wagner, K. D. Hauch, D. G. Castner, B. D. Ratner and T. A. Horbett, J. Biomater. Sci. Polymer Edn 12, 961 (2001). T. P. Stossel, T. D. Pollard, R. J. Mason and M. Vaughan, J. Clin. Invest. 50, 1745 (1971). G. Marletta, S. M. Catalano and S. Pignataro, Surf. Interface Anal. 16, 407 (1990). K. D. Park, W. G. Kim, H. Jacobs, T. Okano and S. W. Kim, J. Biomed. Mater. Res. 26, 739 (1992). G. P. Lopez, B. D. Ratner, C. D. Tidwell, C. L. Haycox, R. J. Rapoza and T. A. Horbett, J. Biomed. Mater. Res. 26, 415 (1992). A. K. McNally and J. M. Anderson, Proc. Natl. Acad. Sci. 91, 10119 (1994). C. R. Jenney, K. M. DeFife, E. Colton and J. M. Anderson, Biomed. Mater. Res. 41, 171 (1998). C. R. Jenney and J. M. Anderson, J. Biomed. Mater. Res. 44, 206 (1999). D. C. Altieri, P. M. Mannucci and A. M. Capitanio, J. Clin. Invest. 78, 968 (1986). D. C. Altieri, J. Immunol. 147, 1891 (1991).
26
M. Shen et al.
34. E. F. Plow and L. Zhang, J. Clin. Invest. 99, 1145 (1997). 35. F. Lim and S. L. Cooper, ASAIO Trans. 37, M146 (1991). 36. G. Altankov, V. Thom, T. Groth, K. Jankova, G. Jonsson and M. Ulbricht, J. Biomed. Mater. Res. 52, 219 (2000). 37. V. Thom, G. Altankov, T. Groth, K. Jankov, G. Johnsson and M. Ulbricht, Langmuir 16, 2756 (2000). 38. H. Nygren, C. Eriksson and J. Lausmaa, J. Lab. Clin. Med. 129, 35 (1997). 39. D. W. H. Riches, in: The Molecular and Cellular Bioogy of Wound Repair, R. Clark (Ed.), p. 95, 2nd edn. Plenum Press, New York (1996). 40. J. H. Brauker, V. E. Carr-Brendel, L. A. Martinson, J. Crudele, W. D. Johnston and R. C. Johnson, J. Biomed. Mater. Res. 29, 1517 (1995).
Limits of detection for time of flight secondary ion mass spectrometry (ToF-SIMS) and X-ray photoelectron spectroscopy (XPS): detection of low amounts of adsorbed protein MATTHEW S. WAGNER 1 , SALLY L. MCARTHUR 2 , MINGCHAO SHEN 2 , THOMAS A. HORBETT 1,2 and DAVID G. CASTNER 1,2,∗ 1 Department
of Chemical Engineering, National ESCA and Surface Analysis Center for Biomedical Problems, University of Washington, Box 351750, Seatle, WA 98195-1750, USA 2 Department of Bioengineering, National ESCA and Surface Analysis Center for Biomedical Problems, University of Washington, Box 351750, Seatle, WA 98195-1750, USA Received 14 August 2001; accepted 10 December 2001 Abstract—Characterization of biomaterial surfaces requires analytical techniques that are capable of detecting a wide concentration range of adsorbed protein. This range includes detection of low amounts of adsorbed protein (<10 ng/ cm2 ) that may be present on non-fouling biomaterials. X-ray Photoelectron Spectroscopy (XPS) and Time of Flight Secondary Ion Mass Spectrometry (ToF-SIMS) are surface sensitive techniques capable of detecting adsorbed proteins. We have investigated the lower limits of detection of both XPS and ToF-SIMS on four model substrates each presenting unique challenges for analysis by XPS and ToF-SIMS: mica, poly(tetrafluoroethylene), allyl amine plasma polymer and heptyl amine plasma polymer. The detection limit for XPS ranged from 10 ng/ cm2 of fibrinogen (on mica) to 200 ng/ cm2 (on allyl amine plasma polymers). The detection limit for ToFSIMS ranged from 0.1 ng/ cm2 of fibrinogen to 100 ng/ cm2 , depending on the substrate and data analysis. Optimal conditions provided detection limits between 0.1 ng/ cm2 and 15 ng/ cm2 on all of the substrates used in this study. While both techniques were shown to be effective in detecting protein, the sensitivity of both XPS and ToF-SIMS was shown to be dependent on substrate surface chemistry and the organization of the adsorbed protein film. This study specifically highlights the applicability of ToF-SIMS in the characterization of low level protein adsorption. Key words: ToF-SIMS; XPS; protein adsorption; quantitation; limit of detection.
∗ To
whom correspondence should be addressed. E-mail:
[email protected]
28
M. S. Wagner et al.
INTRODUCTION
Protein adsorption measurements have been an integral component of biomaterials evaluation. Several techniques have been used to study protein adsorption onto biomaterial surfaces including Fourier transform infrared spectroscopy (FTIR) [1, 2], enzyme-linked immunosorbent assay (ELISA) [3, 4], SDS-PAGE with immunoblotting [5], radiolabeled protein [6], atomic force microscopy (AFM) [7], surface matrix-assisted laser desorption ionization mass spectrometry (MALDIMS) [8], X-ray photoelectron spectroscopy (XPS) [9– 12], ellipsometry [13], surface plasmon resonance (SPR) [14, 15] and secondary ion mass spectrometry (SIMS) [16– 19]. Each technique has its own advantages and disadvantages and their mechanisms of detecting adsorbed proteins can greatly influence their sensitivity. X-ray photoelectron spectroscopy (XPS) is a useful technique for the characterization of adsorbed protein films due to its surface sensitivity (80– 100 Å) and quantitative analysis. However, XPS does not have the chemical specificity to readily identify different adsorbed proteins [20]. Time of Flight Secondary Ion Mass Spectrometry (ToF-SIMS) is a surface sensitive (outer 10– 15 Å) technique that is readily capable of identifying different adsorbed proteins due to its high mass resolution and chemical specificity [17, 21]. Furthermore, ToF-SIMS is highly sensitive, with detection limits on the order of 100– 1000 ppm for biological compounds in tissue [22] and 106 –108 atoms/ cm2 in inorganic samples [23]. Several excellent general reviews of XPS [24, 25] and ToF-SIMS [26– 29] are available. This article explores the utility of XPS and ToF-SIMS for the measurement of protein adsorption onto biomaterials. Particular attention will be focused on the detection limits for these two techniques on several different surface chemistries. These surfaces were chosen because their surface chemistry specifically affects both the organization of the adsorbed protein film and the sensitivity of XPS and ToFSIMS.
MATERIALS AND METHODS
Substrate preparation Mica (SPI supplies, West Chester, PA) was cut into 1 cm2 squares and poly(tetrafluoroethylene), PTFE, (Berghof/ America, Concord, CA) was punched into 6.4 mm diameter disks for both 125I-radiolabeled and unlabeled protein adsorption. The mica samples were freshly cleaved prior to protein adsorption. The PTFE samples and silicon wafer substrates for plasma polymer deposition were ultrasonically cleaned sequentially in methylene chloride, acetone and methanol, then dried and stored under nitrogen until use. Allyl amine and heptyl amine (Aldrich, Milwaukee, WI) plasma polymers (AApp and HApp, respectively) were deposited onto the polished side of freshly cleaned 1cm2 silicon wafers (Silicon Sense, Inc., Nashua, NH) and onto both sides of 6.4 mm diameter perflourinated ethylene-propylene
Protein detection by ToF-SIMS and XPS
29
copolymer, FEP, (Du Pont, Circleville, OH) disks for unlabeled and radiolabeled protein adsorption, respectively. Plasma polymer films were deposited using a radio frequency glow discharge (RFGD) reactor operated at a monomer pressure of 250 mT and a reactor power of 80 W for one minute followed by deposition at 10 W for five minutes using a monomer flow rate of 20 standard cubic centimeters per minute. Details of the chamber construction and general parameters have been detailed elsewhere [30]. Preparation of 125 I-labeled fibrinogen Human fibrinogen (Enzyme Research Laboratories, South Bend, IN) was radiolabeled using Na125I (Amersham Pharmacia Biotech Inc., Piscataway, NJ) and the modified iodine monochloride (ICl) technique of Horbett [31]. A 2 : 1 molar ratio of ICl to fibrinogen was used. Unincorporated 125I was separated from the labeled fibrinogen by two passes through Econo Pac® 10 DG desalting columns (Bio-Rad Laboratories, Hercules, CA). The iodinated fibrinogen was stored at −70◦ C and used within two weeks of preparation. Protein adsorption Radiolabeled and unlabeled fibrinogen were adsorbed onto the substrates from citrate phosphate buffered saline with sodium azide and sodium iodide (CPBSzI, 0.11 M NaCl, 0.01 M NaI, 0.01 M sodium citrate, 0.01 M sodium phosphate (monobasic), 0.02% NaN3 , pH = 7.4) [6] at 37◦ C for two hours. For the 125 I-labeled fibrinogen adsorption experiments, enough 125 I-labeled fibrinogen was added to a stock 100 µg/ ml fibrinogen solution to result in a specific activity of no less than 25 cpm / ng. The stock solution for both the unlabeled and radiolabeled protein adsorption was subsequently diluted using CPBSzI buffer to 10 µg/ ml, 1 µg/ ml, 100 ng/ ml, 10 ng/ ml and 1 ng/ ml. The samples from the 125 I-labeled fibrinogen adsorption experiments were rinsed with CPBSzI buffer using a device described previously [6]. At least four replicates were measured for each protein concentration used in this study. Bound 125 I-labeled fibrinogen was measured using a Model 1185 Gamma Counter (TM Analytic, Elk Grove, IL). Amounts of adsorbed fibrinogen were calculated from the retained radioactivity (corrected for background and decay), the specific activity of the protein solution and the surface area of the sample. Samples from the unlabeled fibrinogen adsorption experiments were rinsed twice in stirred buffer solution to remove loosely bound protein and three times in stirred deionized water to remove buffer salts. Control samples were run in parallel and exposed to the same buffer, temperature and rinsing conditions as the samples with adsorbed protein. X-ray photoelectron spectroscopy (XPS) Survey (0– 1000 eV, analyzer pass energy = 150 eV) and high resolution C1s (275– 295 eV, analyzer pass energy = 25 eV) XPS spectra were obtained using
30
M. S. Wagner et al.
a Surface Science X-Probe instrument (SSI, Mountain View, CA) equipped with a monochromatized AlKα1,2 X-ray source and an electron flood gun for charge neutralization. The X-ray spot size was elliptical with a major axis of 1700 µm and minor axis of 1000 µm. Unless otherwise noted, the electron take-off angle (measured between the surface normal and the axis of the electron energy analyzer lens) for all of the XPS spectra in this study was 55◦ . When noted, angleresolved XPS spectra were acquired at an electron take-off angles of 0◦ , 55◦ , 68◦ and 80◦ . Linear background subtraction was used for the survey and high resolution C1s spectra. The raw peak intensities obtained from XPS measurements were corrected for differences in photoionization cross-sections using manufacturer supplied sensitivity factors to obtain surface atomic compositions for all XPS data reported in this study. The major component (presumed to be hydrocarbon) of each C1s spectrum was binding energy referenced to 285.0 eV. Two spots were analyzed on each of at least four samples for every concentration used in this study. Time of flight secondary ion mass spectrometry (ToF-SIMS) ToF-SIMS spectra were acquired using a PHI Model 7200 Time of Flight Secondary Ion Mass Spectrometer (Physical Electronics, Eden Prarie, MN) equipped with an 8 keV Cs+ primary ion source and pulsed flood gun for charge neutralization. Positive and negative ion ToF-SIMS spectra were acquired over an area of 200 µm × 200 µm while maintaining a primary ion dose less than 1012 ions/ cm2 to insure static SIMS conditions [32]. The mass resolution (m/m) at the C4 H8 N+ (m/z = 70) and C2 H− (m/z = 25) peaks were typically above 4000 in the positive and negative ion spectra, respectively. Positive and negative ion ToF-SIMS spectra + + + − − were calibrated to the CH+ 3 , C3 H3 , C3 H5 and C7 H7 peaks and CH , CN and − CNO peaks, respectively, before further analysis. Three spots were analyzed on each of at least four samples for every concentration used in this study. Data analysis Statistical significance of the data was determined using an F -test to determine equality of variance followed by the appropriate two-tailed Student’s t-test to determine equality of means. The detection limit for both XPS and ToF-SIMS is defined as the lowest fibrinogen surface concentrations at which the surface spectrum is statistically significantly different from the control sample. The features of the surface spectra analyzed were dependent on the analysis type and surface chemistry. For XPS data analysis, the N1s intensity was normalized by a unique substrate peak, when possible (Al2s for mica and F1s for PTFE), to account for variations in total signal intensity from spectrum to spectrum. Since the amine plasma polymer substrates contained nitrogen, the N1s to C1s ratio was used to indicate fibrinogen adsorption. For negative ion ToF-SIMS on the PTFE samples, the peaks related to the protein fragmentation (CN− and CNO− ) were summed and normalized by the F− or F− 2 peaks to correct for changes in total signal intensity
Protein detection by ToF-SIMS and XPS
31
from spectrum to spectrum. The CNO− peak was normalized by the CN− peak for the amine plasma polymer substrates as described below. Positive ion ToF-SIMS spectra were analyzed using Principal Component Analysis as described below. Principal Component Analysis (PCA) was used to analyze the positive ion ToFSIMS spectra and was performed using the PLS Toolbox v. 2.0 (Eigenvector Research, Manson, WA) for MATLAB (the MathWorks, Inc., Natick, MA). The positive ion ToF-SIMS spectra were normalized to the sum of the selected peaks and mean-centered before PCA. PCA calculates the linear combination of ToFSIMS peaks that describe the major directions of variation in the data set. The spectra are then assigned values (called scores) on these new axes (called principal components, PCs). The scores describe the relationship between the samples while the relationship between the ToF-SIMS peaks and the PCs are given by the loadings. The scores on the first PC were used to track the fibrinogen adsorption because the first PC captures the largest degree of variation in the ToF-SIMS spectra, the source of which was assumed to be due to the fibrinogen adsorption. Further descriptions of PCA can be found in [33, 34] and descriptions of PCA applied to ToF-SIMS spectra of adsorbed protein films can be found in [17, 21].
RESULTS
Surface chemistry of the substrates The atomic compositions of the control samples for each data set as determined by XPS are given in Table 1. These samples were all exposed to the buffer at the same conditions as the protein adsorption (2 h at 37◦ C) and rinsed as described above (in stirred buffer and deionized water). The mica control samples are composed of silicon, aluminum and potassium with a significant amount of adventitious carbon. The small amount of sodium in the mica spectra is due to adsorption of buffer salts which remain on the surface even after extensive rinsing. The PTFE control samples were composed entirely of fluorine and carbon with a fluorine to carbon atomic ratio near the expected 2 : 1. The atomic composition of the HApp control samples was Table 1. Atomic compositions for the substrates used in this study. Compositions are for buffer soaked control substrates and are given as averages ± standard deviations Substrate
Mica PTFE Silicon Wafer Allyl Amine Heptyl Amine
Atomic concentration (%) C1s
N1s
O1s
F1s
Si2s
Al2s
K2s
Na (Auger)
13 ± 2 35 ± 2 15 ± 1 72 ± 2 84 ± 1
— — — 13 ± 1 10 ± 1
51 ± 2 — 33 ± 0 10 ± 1 6±0
— 65 ± 2 — — —
15 ± 1 — 51 ± 2 5±2 —
16 ± 1 — — — —
3±0 — — — —
3±2 — — — —
32
M. S. Wagner et al.
significantly different from that of the silicon wafer substrate, showing that a thick film was deposited. The AApp control samples had higher levels of nitrogen and oxygen than the HApp control samples in addition to some silicon, showing that this plasma polymer film was thinner than the XPS sampling depth. Protein adsorption:
125
I results
Protein adsorption onto the four surfaces was measured using 125 I-labeled fibrinogen. Figure 1 shows the fibrinogen adsorption for the four substrates used in this study using fibrinogen solutions of various concentrations. At a solution concentration of 1 ng/ ml, the surface concentrations on all surfaces were 1 ng/ cm2 or lower. At 100 µg/ ml, the surface concentrations varied significantly (note the vertical axis in Fig. 1 is logarithmic), ranging from 350 ng/ cm2 (for PTFE) to 1650 ng/ cm2 (for mica). The high fibrinogen adsorption values are greater than monolayer surface coverage. This may be due to nonspecific adsorption of free iodine, which would mean that the adsorbed fibrinogen amounts may be overestimated. An overestimation of the adsorbed fibrinogen amounts would result in a detection limit lower than those calculated in this study. The adsorption on all of the substrates follow a Freundlich adsorption isotherm typical for protein adsorption.
Figure 1. Fibrinogen adsorption onto mica, PTFE, HApp, and AApp after incubation for two hours at the plotted solution concentrations. Note that this is a log– log plot and the fibrinogen adsorptions are significantly different at high fibrinogen solution concentrations.
Protein detection by ToF-SIMS and XPS
33
Protein adsorption: XPS results For all of the surface analysis results reported in this study, the detection limit is defined as the lowest fibrinogen surface concentration at which the surface spectrum is statistically different from the control samples. Figure 2a shows the atomic percent of nitrogen and the intensity ratios of the N1s peak to the Al2s peak as measured from XPS for fibrinogen adsorbed onto mica. The detection limit of XPS for fibrinogen was approximately 10 ng/ cm2 on mica based on both the atomic nitrogen composition and the N /Al ratio (p 0.01). The nitrogen atomic percent of these samples was approximately 1% at the limit of detection. Figure 2b shows the atomic percent of nitrogen and the intensity ratios of the N1s peak to the F1s peak from XPS for fibrinogen adsorbed onto PTFE. While the samples are statistically different from the buffer-soaked control and the unsoaked, clean samples down to 1 ng/ cm2 (p 0.01), the nitrogen atomic percentage of these samples drops below 0.5% at approximately 100 ng/ cm2 . Due to errors in elemental quantitation that can occur in XPS measurements near the instrumental detection limit (0.1– 0.5 atomic %) [24] and the variability in this data at these
(a) Figure 2. Surface nitrogen detected by XPS analysis of fibrinogen adsorbed onto (a) mica (b) PTFE and (c) allyl amine and heptyl amine plasma polymer surfaces. N/Al and N/ F atomic ratios are also given for mica and PTFE substrates, respectively. Asterisks (*) indicate samples that are highly significantly different from the control samples (p 0.01) while crosses (+) indicate samples that are different from clean samples (p 0.01).
34
M. S. Wagner et al.
(b)
(c) Figure 2. (Continued).
Protein detection by ToF-SIMS and XPS
35
surface concentrations, the detection limit must be regarded as approximately 10– 25 ng/ cm2 . Analysis of the XPS protein adsorption data on the nitrogen-containing AApp and HApp samples is more challenging due to the lack of a unique signal to differentiate between the substrate and the protein. While both contain nitrogen, the ratios of the elements vary significantly, especially the nitrogen to carbon ratios. Figure 2c shows the N1s to the C1s intensity ratio for fibrinogen on the amine plasma polymer surfaces. The detection limit for fibrinogen on the HApp surface is between 10 and 100 ng/ cm2 (p 0.01). The N1s/ C1s ratio changed very little throughout the entire range of fibrinogen adsorbed onto the AApp surface. However, by examining the high resolution C1s spectra, fibrinogen adsorption can be tracked by the appearance of the amide peak at a binding energy of 288.2 eV from the polypeptide backbone in the fibrinogen molecule (Fig. 3a). Similar presence of the amide peak was observed on the HApp substrate (Fig. 3b). On the AApp surfaces, the appearance of a definable amide peak corresponds to a surface concentration of 200 ng/ cm2 . Protein adsorption: positive ion ToF-SIMS results PCA was used to analyze the positive ion ToF-SIMS spectra of the fibrinogen adsorbed onto the test surfaces. PCA analyzes the intensities of the peaks in the positive ion ToF-SIMS spectra and finds the major axes of variation in the data set (i.e. the PCs). For the mica and PTFE data sets, a selection of peaks from the protein [17] and several substrate peaks were used in the analysis. The protein peaks and substrate peaks used for the mica and PTFE analysis are given in Table 2. For AApp and HApp, the entire spectrum from 0– 200 m/z was used due to a possible overlap between fragments generated from the protein and fragments generated from the substrate. Figure 4a shows the scores on the first PC for the positive ion ToF-SIMS spectra of fibrinogen adsorbed onto mica as a function of the fibrinogen surface concentration. The scores on the first PC of the positive ion spectra (capturing 87% of the variance in the data set) indicate a detection limit for fibrinogen on mica of 0.1 ng/ cm2 (p 0.01). Later PCs contain no discernable trend with the surface concentration of fibrinogen. Figure 4b shows the loadings for the first PC, which give the relationship between the ToF-SIMS peaks and that PC. The negatively loaded substrate peaks are correlated with the samples with negative scores (i.e. the substrate and the samples with low fibrinogen surface concentrations) and the positively loaded protein peaks are correlated with the samples with positive scores (i.e. the samples with high fibrinogen surface concentrations). Since the peaks from the substrate are distinct from the peaks from the protein, ToF-SIMS is readily able to detect very low amounts of adsorbed fibrinogen on the mica surface. PCA was also used to analyze the positive ion spectra of fibrinogen adsorbed onto PTFE. The first PC captures 53% of the variation in the data set and suggests a detection limit for fibrinogen of approximately 100 ng/ cm2 (p 0.01,
36
M. S. Wagner et al.
(a)
(b) Figure 3. High resolution C1s XPS spectra for fibrinogen on (a) allyl amine and (b) heptyl amine plasma polymer surfaces. Note the introduction of the amide peak at 288.2 eV with the adsorption of fibrinogen onto the plasma polymer substrates.
Protein detection by ToF-SIMS and XPS
37
Table 2. Positive ion peaks selected for Principal Component Analysis of adsorbed fibrinogen films on mica and PTFE. Underlined peaks (69 and 131) were omitted from the protein on PTFE data due to overlap + with fluorocarbon peaks (CF+ 3 and C3 F5 , respectively) Source
Fragment
44: C2 H6 N+ + + + 43: CH3 N+ 2 , 73: C2 H7 N3 , 100: C4 H10 N3 , 101: C4 H11 N3 , + + 112: C5 H8 N3 , 127: C5 H11 N4 + Asparagine (Asn, N) 70: C3 H4 NO+ , 87: C3 H7 N2 O+ , 88: C3 H6 NO+ 2 , 98: C4 H4 NO2 + Aspartic Acid (Asp, D) 88: C3 H6 NO2 Cysteine (Cys, C) 76: C2 H6 NS+ Glutamine (Gln, Q) 84: C4 H6 NO+ Glutamic Acid (Glu, E) 84: C4 H6 NO+ , 102: C4 H8 NO+ 2 Glycine (Gly, G) 30: CH4 N+ + + Histidine (His, H) 81: C4 H5 N+ 2 , 82: C4 H6 N2 , 110: C5 H8 N3 Isoleucine (Ile, I) 86: C5 H12 N+ Leucine (Leu, L) 86: C5 H12 N+ Lysine (Lys, K) 84: C5 H10 N+ Methionine (Met, M) 61: C2 H5 S+ Phenylalanine (Phe, F) 120: C8 H10 N+ , 131: C9 H8 O+ Proline (Pro, P) 68: C4 H6 N+ , 70: C4 H8 N+ Serine (Ser, S) 60: C2 H6 NO+ , 71: C3 H3 O+ 2 Threonine (Thr, T) 69: C4 H5 O+ , 74: C3 H8 NO+ + Tryptophan (Trp, W) 130: C9 H8 N , 159: C10 H11 N+ , 170:C11 H8 NO+ Tyrosine (Tyr, Y) 107: C7 H7 O+ , 136: C8 H10 NO+ Valine (Val, V) 72: C4 H10 N+ , 83: C5 H7 O+ Alanine (Ala, A) Arginine (Arg, R)
Mica PTFE
27: Al+ , 28:Si+ , 39: K+ + + + 12: C+ , 31: CF+ , 50: CF+ 2 , 62: C2 F2 , 69: CF3 , 74: C3 F2 , + + + + + 81: C2 F3 , 93: C3 F3 , 100: C2 F4 , 112: C3 F4 , 119: C2 F5 , + + + + + 124: C4 F+ 4 , 131: C3 F5 , 143: C4 F5 , 155: C5 F5 , 162: C4 F6 , 169: C3 F7 , + + 181: C4 F7 , 193: C5 F7
data not shown). Later PCs contain no discernable trend with the fibrinogen surface concentration, with the variation being due to fluctuations in the relative intensities of the fluorocarbon ions from the PTFE substrate. The loadings show that the peaks correlated with the PTFE control and samples with low fibrinogen surface concentrations result from the PTFE substrate while peaks correlated with the samples with high amounts of fibrinogen adsorbed result from the adsorbed fibrinogen (data not shown). The degradation of the sensitivity of the ToF-SIMS for fibrinogen adsorbed onto PTFE when compared to mica may have been due to two reasons: patchy coverage of the protein on the PTFE [10] and the high secondary ion yield of fluorocarbon ions in ToF-SIMS [35]. Figure 5 shows the ToF-SIMS spectra of fibrinogen adsorbed onto (a) mica and (b) PTFE at fibrinogen surface concentrations of approximately 100 ng/ cm2 . The fluorocarbon peaks in Fig. 5b are present in significantly higher amounts than the mica substrate (Al+ , Si+ and
38
M. S. Wagner et al.
(a)
(b) Figure 4. (a) Scores and (b) loadings from the first PC from the positive ion ToF-SIMS spectra of fibrinogen adsorbed onto mica. Asterisks indicate samples that are highly significantly different from the control samples (p 0.01).
Protein detection by ToF-SIMS and XPS
39
(a)
(b) Figure 5. Positive ion ToF-SIMS spectra for fibrinogen adsorbed onto (a) mica and (b) PTFE with surface concentrations of approximately 100 ng/ cm2 . Note that the fluorocarbon ions in (b) are significantly more intense than the substrate ions in (a) at the same protein surface concentration.
40
M. S. Wagner et al.
K+ ) peaks in Fig. 5a, even though the same amount of protein is adsorbed. For example, the ratio of the intensity of the Al+ peak to the C4 H8 N+ peak (typically the strongest peak in a fibrinogen ToF-SIMS spectrum) is approximately 2.3 on mica while the ratio of the CF+ peak to the C4 H8 N+ peak is approximately 7.0 on PTFE. PCA was also used to analyze the positive ion ToF-SIMS spectra of fibrinogen adsorbed onto the amine plasma polymer surfaces. The scores in Figure 6a on the first PC for the fibrinogen adsorbed onto the HApp (capturing 94% of the variance in the data set) show that the detection limit for fibrinogen is approximately 2 ng/ cm2 (p 0.01). As the scores increase from negative to positive numbers, the amount of adsorbed fibrinogen increases. The loadings in Fig. 6b show that positive ions from the amino acids (e.g. 28: CH2 N+ , 44: C2 H6 N+ , 55: C3 H3 O+ , 70: C4 H8 N+ , 84: C5 H10 N+ and 130: C9 H8 N+ ) correspond to samples with positive scores (i.e. samples with high fibrinogen surface concentrations) while positive ions attributable + + + to the plasma polymer (e.g. 29: C2 H+ 5 , 43: C3 H7 , 55: C4 H7 , 69: C5 H9 and 81: + C6 H9 ) correspond to samples with negative scores (i.e. the control and samples with low fibrinogen surface concentrations). PCA of the positive ion ToF-SIMS spectra of fibrinogen adsorbed onto the AApp substrates show similar results with a detection limit of approximately 20 ng/ cm2 (p 0.01, data not shown). The loadings also describe the peaks with the most significant differences between the samples with high fibrinogen surface concentrations and the control samples. Protein adsorption: negative ion ToF-SIMS results Negative ion ToF-SIMS was also used to analyze the PTFE, AApp and HApp samples. The mica samples were not analyzed using negative ion ToF-SIMS due to the excellent sensitivity of the positive ion mode. Furthermore, a unique signal from the substrate is not present in the negative ion spectra of fibrinogen adsorbed onto mica. Figure 7a shows the ratio of the sum of the intensities of the CN− (m/z = 26) and CNO− (m/z = 42) peaks to the F− (m/z = 19) peak or F− 2 (m/z = 38) peak in the negative ion ToF-SIMS spectra of fibrinogen adsorbed onto PTFE. This data suggests a detection limit of approximately 100 ng/ cm2 when using the F− peak and 2 ng/ cm2 when using the F− 2 peak (p 0.01). The secondary ion yield of atomic fluorine in the negative ion mode may also influence the detection limit of negative ion ToF-SIMS for adsorbed fibrinogen. Furthermore, the scatter in this data was also significant, especially with higher amounts of adsorbed fibrinogen, suggesting patchy coverage of the substrate. The negative ion ToF-SIMS spectra of fibrinogen adsorbed onto the amine plasma polymer substrates also contain information on the amount of adsorbed protein. As evidenced in the XPS data, protein adsorption resulted in the introduction of amide groups on the surface. In analyzing the negative ion data from these nitrogen containing surfaces, the ratio of the intensity of the CNO− (m/z = 42, from the amide linkage in the protein backbone) to CN− (m/z = 26, from the substrate and the protein) ions was utilized. Figure 7b shows the variations in the ratio of the
Protein detection by ToF-SIMS and XPS
41
(a)
(b) Figure 6. (a) Scores and (b) loadings for the first PC of positive ion ToF-SIMS spectra of fibrinogen adsorbed onto the heptyl amine plasma polymer. Asterisks indicate samples that are highly significantly different from the control samples (p 0.01).
42
M. S. Wagner et al.
(a)
(b) Figure 7. (a) Ratio of the intensities of the CN− and CNO− peaks to the F− or F− 2 peak from the negative ion ToF-SIMS spectra of fibrinogen adsorbed onto PTFE. (b) Ratio of the intensity of the CNO− peak to CN− peak from the negative ion ToF-SIMS spectra of fibrinogen adsorbed onto the allyl and heptyl amine plasma polymer substrates. Asterisks indicate samples that are highly significantly different from the control samples (p 0.01).
Protein detection by ToF-SIMS and XPS
43
CNO− to CN− peak intensities with increasing fibrinogen adsorption. The detection limit of the negative ion ToF-SIMS for fibrinogen is approximately 1 ng/ cm2 on AApp and 15 ng/ cm2 for HApp (p 0.01). DISCUSSION
The detection limits of surface analytical techniques for adsorbed protein are especially critical in the biomaterials field. With the advent of non-fouling biomaterial surfaces, there is a continued focus on accurately determining the efficacies of specific surface treatments. The results of this study show that the detection limits of proteins in XPS and ToF-SIMS are highly substrate dependent. The surface chemistry of the substrate is important in the organization of the adsorbed protein layer and performance of the surface analytical technique. The sensitivity of the XPS for adsorbed fibrinogen varied widely across the four substrates used in this study. The low detection limit on mica was not unexpected due to the smoothness of the surface. However, the detection limit on PTFE was significantly poorer than on mica. Since the surface chemistry of the substrate does not affect the photoelectron signal from the adsorbed fibrinogen, the surface coverage and/ or surface roughness presumably affect the sensitivity of the XPS. Previous studies have shown that adsorbed proteins on fluorocarbon substrates form patchy films [10], suggesting that patchy coverage plays a role in the detection limit on PTFE. Furthermore, the PTFE substrates used in this study were observed to be macroscopically rough. Since all of the XPS spectra acquired for this study were obtained at an electron take-off angle (the angle between the surface normal and the electron energy analyzer) of 55◦ , the sampling depth was approximately 45 Å (assuming a sampling depth of 80 Å at 0◦ and a sampling depth that varies as the cosine of the electron take-off angle [24]). At this sampling depth, it is assumed that both the substrate and the adsorbate will be sampled. Use of a higher electron take-off angle (e.g. 80◦ , where the sampling depth is approximately 15 Å), may enable more of the protein layer and less of the substrate layer to be sampled and may result in higher sensitivity for the adsorbed fibrinogen. However, surface roughness would be expected to influence the data at these lower sampling depths due to shadowing and an inexact knowledge of the electron take-off angle [36], making data interpretation more difficult. Angle-resolved XPS was attempted on the PTFE samples with fibrinogen adsorbed from a 100 ng/ ml solution, but no difference was noted between the high take-off angle (low sampling depth) spectra of the sample with protein and the control sample. Film thickness and fractional coverage calculations [9– 12, 20] may also improve the detection and quantitation of adsorbed proteins, but these calculations require an independent measurement of the surface concentration of the adsorbed protein (typically acquired by 125 I radiolabeling of the protein). The high relative standard deviations in the XPS measurements at fibrinogen surface concentrations less than 100 ng/ cm2 make the determination of an absolute limit of detection difficult. Since the average nitrogen
44
M. S. Wagner et al.
atomic percent was greater than those of the control for these surface concentrations (0.3– 0.4 atomic percent vs. 0.1 atomic percent), the detection limit for fibrinogen adsorbed onto PTFE is in the range on 10– 25 ng/ cm2 . Amounts of adsorbed protein can be quantified using XPS when the adsorbed protein amounts are between the detection limit and a monolayer. Previous authors have demonstrated that a nonlinear, Beer’s law approach is useful for probing the organization of adsorbed proteins on both PTFE and mica surfaces [9– 11]. A similar approach is applicable for the quantitation of the amounts of adsorbed proteins. Adsorbed fibrinogen amounts can be quantified on mica down to 10 ng/ cm2 while this limit is 100 ng/ cm2 on PTFE. The difficulty in measuring protein adsorption on amine plasma polymer surfaces was not unexpected. Since the AApp had a higher nitrogen content than the HApp, the detection limit was significantly higher than on all of the other substrates. In fact, the presence of the adsorbed fibrinogen was only able to be detected in the high resolution C1s spectra by the presence of the amide peak or at very high surface concentrations via the N1s/ C1s atomic ratio. Since the HApp had a lower nitrogen content, increases in the nitrogen to carbon ratio were indicative of the presence of adsorbed fibrinogen, although significant changes in the ratio were only seen at a much higher surface concentration than on mica. The results highlighted that while XPS can be used to detect protein on a variety of surfaces, the presence of nitrogen in the substrate or a patchy surface coverage limits its detection capabilities. In all cases, ToF-SIMS was equal to or more sensitive than XPS for the detection of adsorbed fibrinogen, with the improvement of sensitivity for ToF-SIMS over XPS of typically an order of magnitude or more. The lower limit of sensitivity for ToF-SIMS (achieved on the mica substrate) of 0.1 ng/ cm2 is equivalent to 0.1 attomoles (10−18 moles) of fibrinogen (calculated by taking into account the size of the analysis area). Mass spectrometry techniques with samples in the liquid phase, such as Matrix Assisted Laser Desorption/ Ionization (MALDI) and Electrospray Ionization (ESI), have sensitivities in the femptomolar to attomolar range [37]. The sensitivity of ToF-SIMS is similar to these techniques, but the analyte is in the adsorbed phase. ToF-SIMS provides a powerful technique for analyzing adsorbed protein in very low surface concentrations, even when signal from the substrate is similar to the signal from the protein (as is the case with the amine plasma polymers). However, in cases such as PTFE, the substrate may negatively affect the sensitivity of the ToF-SIMS. The ToF-SIMS data can also be used for quantitation of adsorbed fibrinogen amounts between the detection limit and monolayer surface concentrations. Since the scores continue to change throughout this entire concentration range, it is possible to use them for quantitation of the amount of adsorbed protein, though any calibration curve generated using this method will be specific to a particular protein due to differences in fragmentation patterns between proteins. Ferrari and Ratner have previously demonstrated that multivariate calibration techniques can be used to quantify adsorbed protein by ToF-SIMS [38]. However, quantitation is
Protein detection by ToF-SIMS and XPS
45
limited to monolayer and submonolayer surface concentrations due to the sampling depth of ToF-SIMS (10– 15 Å). The high secondary ion yield of the fluorocarbon ions in ToF-SIMS, referred to as the matrix effect [35], results in the suppression of the secondary ions from the protein, drastically reducing the sensitivity of the ToF-SIMS. It is expected that other fluoropolymers would have similar matrix effects, resulting in a reduced sensitivity for ToF-SIMS on these substrates. Furthermore, Briggs et al. have shown that Cl− ions in the negative ion ToF-SIMS spectra of a poly(methyl methacrylate)poly(vinyl chloride) blend have a greater sampling depth than that typically assumed for static ToF-SIMS analysis (10– 15 Å) [39]. Likewise, the F− ions may be assumed to have an increased sampling depth in the negative ion ToF-SIMS spectra, resulting in more of the PTFE substrate being sampled and a decreased sensitivity for the adsorbed fibrinogen. The ratio of the intensities of the CN− and CNO− peaks to the F− 2 peak does not suffer from the same problem, so this ratio shows a sensitivity of approximately 2 ng/ cm2 (p 0.01). Charge neutralization (for insulating materials) also can dramatically effect the mass resolution and sensitivity of the ToF-SIMS, with highly insulating samples (such as PTFE) exhibiting much poorer sensitivity. Furthermore, as with the XPS data, surface roughness can affect sample charge neutralization and signal. The PTFE substrate is the most challenging for ToF-SIMS data interpretation because the chemistry of the substrate and the organization of the adsorbed protein film both influence the sensitivity of the ToFSIMS instrument. The mica substrate is also challenging due to the tight binding of sodium ions which may affect the secondary ion yields from the protein and substrate in the ToF-SIMS (i.e. matrix effects). For the PCA of the positive ion ToF-SIMS spectra of fibrinogen on mica, the data acquired for each concentration was analyzed by PCA. For every concentration, the first PC separated the spectra with high amounts of sodium from those with low amounts of sodium. The spectra with high amounts of sodium in each group were then discarded for subsequent analysis and these data were reported above. However, if the samples with high salt were not discarded, the sensitivity of the ToF-SIMS remains the same for the mica surface, but the standard deviations around each group are larger, suggesting a contribution to the variability in the spectra due to the presence of sodium. Univariate analysis was also used to analyze the positive ion spectra of the mica and PTFE samples using the ratio of the sum of the protein peaks to the sum of the substrate peaks. The univariate analysis showed the same sensitivity as the PCA for these data sets, but with increased standard deviations at each surface concentration (data not shown). PCA has a noise-filtering effect which reduces the standard deviations at each surface concentration and may be important in determining very low surface concentrations of adsorbed proteins (<10 ng/ cm2 ). Univariate analysis was not attempted for the amine plasma polymer surfaces due to their complexity and the overlap of fragments from both the substrate and the adsorbed fibrinogen.
46
M. S. Wagner et al.
In this case, PCA was essential in analyzing the positive ion ToF-SIMS spectra from these surfaces. Another aspect of the ‘limit of detection’ problem for XPS and ToF-SIMS is the ‘limit of identification’, that is, the surface concentration limit at which the surface spectra can be used to correctly identify the type of adsorbed protein on the surface. XPS cannot readily differentiate one protein from another, but we have previously shown that ToF-SIMS can readily differentiate between proteins [17, 21, 40]. While the packing of the adsorbed fibrinogen probably did not influence its detection by ToF-SIMS, the packing (and therefore the protein conformation) will probably have a significant effect on the identification of proteins at low surface concentrations.
CONCLUSION
Detection of low amounts of adsorbed protein is a challenging analytical problem that must be addressed when analyzing non-fouling biomaterial surfaces. We have shown the following: (1) XPS is sensitive to the adsorption of protein on surfaces that are chemically different from proteins. The organization of the adsorbed protein layer and the chemical composition of the substrate must be considered when using XPS to detect adsorbed proteins. (2) ToF-SIMS is a highly sensitive method for the detection of proteins on many different types of surfaces. Again, knowledge of the organization of the adsorbed protein layer and the influence of the surface chemistry on the ToFSIMS process is essential for the proper interpretation of the ToF-SIMS data. (3) Principal Component Analysis simplified the analysis of the positive ion ToFSIMS spectra by determining new axes that described the greatest variation in the data set. Table 3. Summary of detection limits for XPS and ToF-SIMS (all values in ng of fibrinogen per square centimeter) Substrate
XPS
Positive ion ToF-SIMS
Negative ion ToF-SIMS
Mica PTFE Heptyl Amine Plasma Polymer Allyl Amine Plasma Polymer
10 10– 25b 10– 100 200d
0.1 100 2 10
—a 100, 2c 15 1
a Negative
ion ToF-SIMS was not utilized to analyze fibrinogen on mica. is detectable at 10– 25 ng/ cm2 , but variance in the data makes achieving statistical significance difficult. c Results are dependent on whether the F− or F− peak is used in the data analysis. 2 d XPS results were based on the amide peak in the high resolution C1s spectra. No significant difference was found in the atomic composition of the samples. b Nitrogen
Protein detection by ToF-SIMS and XPS
47
(4) The sensitivity of ToF-SIMS is typically enhanced over XPS by 1-2 orders of magnitude. A summary of the sensitivities for XPS and ToF-SIMS is found in Table 3. Both XPS and ToF-SIMS are essential techniques for understanding the organization and composition of the adsorbed proteins. ToF-SIMS provides superior detection limits with the possibility of identifying the adsorbed protein, even at very low surface concentrations. Acknowledgements Funding for this research was provided by the National ESCA and Surface Analysis Center for Biomedical Problems (NESAC /BIO-NIH Grant RR-01296) through the National Center for Research Resources and NHLBI grant HL19419. Dr. Stephen Golledge is gratefully acknowledged for discussions. Winston Ciridon is gratefully acknowledged for performing the RFGD plasma polymer depositions.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
R. Barbucci, M. Casolaro and A. Magnani, Clin. Mater. 11, 37 (1992). K. K. Chittur, Biomaterials 19, 357 (1998). M. Balcells, D. Klee, M. Fabry and H. Hocker, J. Colloid Interface Sci. 220, 198 (1999). A. Welle, M. Grunze and D. Tur, J. Colloid Interface Sci. 197, 262 (1998). S. R. Mulzer and J. L. Brash, J. Biomater. Sci. Polymer Edn 1, 173 (1990). T. A. Horbett, in: Techniques of Biocompatibility Testing, D. F. Williams (Ed.), p. 183. CRC Press, Inc., Boca Raton (1986). C. A. Siedlecki and R. E. Marchant, Biomaterials 19, 441 (1998). P. Kingshott, H. A. W. StJohn, R. C. Chatelier and H. J. Griesser, J. Biomed. Mater. Res. 49, 36 (2000). H. Fitzpatrick, P. F. Luckham, S. Eriksen and K. Hammond, J. Colloid Interface Sci. 149, 1 (1992). R. W. Paynter, B. D. Ratner, T. A. Horbett and H. R. Thomas, J. Colloid Interface Sci. 191, 233 (1984). B. D. Ratner, T. A. Horbett, D. Shuttleworth and H. R. Thomas, J. Colloid Interface Sci. 83, 630 (1981). J.-E. Sundgren, P. Bodo, B. Ivarsson and I. Lundstrom, J. Colloid Interface Sci. 113, 530 (1985). H. Elwing, Biomaterials 19, 397 (1998). R. J. Green, M. C. Davies, C. J. Roberts and S. J. B. Tendler, Biomaterials 20, 385 (1999). R. J. Green, R. A. Frazier, K. M. Shakesheff, M. C. Davies, C. J. Roberts and S. J. B. Tendler, Biomaterials 21, 1823 (2000). J.-B. Lhoest, E. Detrait, P. v. d. B. d. Aguilar and P. Bertrand, J. Biomed. Mater. Res. 41, 95 (1998). J.-B. Lhoest, M. S. Wagner, C. D. Tidwell and D. G. Castner, J. Biomed. Mater. Res. 57, 432 (2001). D. S. Mantus, B. D. Ratner, B. A. Carlson and J. F. Moulder, Anal. Chem. 65, 1431 (1993). C. D. Tidwell, D. G. Castner, S. L. Golledge, B. D. Ratner, K. Meyer, B. Hagenhoff and A. Benninghoven, Surf. Interface Anal. 31, 724 (2001). R. W. Paynter and B. D. Ratner, in: Surface and Interfacial Aspects of Biomedical Polymers, J. D. Andrade (Ed.), p. 189. Plenum Press, New York (1985).
48 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40.
M. S. Wagner et al. M. S. Wagner and D. G. Castner, Langmuir 17, 4649 (2001). C. M. John and R. W. Odom, Inter. J. Mass Spec. Ion. Proc. 161, 47 (1997). M. A. Douglas and P. J. Chen, Surf. Interface Anal. 26, 984 (1998). B. D. Ratner and D. G. Castner, in: Surface Analysis — The Principal Techniques, J. C. Vickerman (Ed.), p. 43. John Wiley & Sons, Chichester (1997). N. H. Turner and J. A. Schriefels, Anal. Chem. 70, 229R (1998). A. Adriaens, L. vanVaeck and F. Adams, Mass Spec. Rev. 18, 48 (1999). B. Hagenhoff, Biosens. & Bioelect. 10, 885 (1995). D. Leonard and H. J. Mathieu, Fresenius J. Anal. Chem. 365, 3 (1999). L. vanVaeck, A. Adraens and R. Gijbels, Mass Spec. Rev. 18, 1 (1999). G. P. Lopez and B. D. Ratner, Langmuir 7, 766 (1991). T. A. Horbett, J. Biomed. Mater. Res. 15, 673 (1981). G. Marletta, S. M. Catalano and S. Pignataro, Surf. Interface Anal. 16, 407 (1990). J. E. Jackson, A User’s Guide to Principal Components. John Wiley & Sons, New York (1991). S. Wold, K. Esbensen and P. Geladi, Chemom. Intell. Lab. Syst. 2, 37 (1987). D. Leonard, P. Bertrand, M. K. Shi, E. Sacher and L. Martinu, Plasmas and Polymers 4, 97 (1999). C. S. Fadley, Prog. Surf. Sci. 16, 275 (1984). A. L. Burlingame, R. K. Boyd and S. J. Gaskell, Anal. Chem. 70, 647R (1998). S. Ferrari and B. D. Ratner, Surf. Interface Anal. 29, 837 (2000). D. Briggs, I. W. Fletcher, S. Reichlmaier, J. L. Agulo-Sanchez and R. D. Short, Surf. Interface Anal. 24, 419 (1996). J.-B. Lhoest, M. S. Wagner and D. G. Castner, in: 12th International Conference on Secondary Ion Mass Spectrometry, Brussels, Belgium, p. 935 (1999).
Photoimmobilization of biomolecules within a 3-dimensional hydrogel matrix X. CAO 1,3,∗ and M. S. SHOICHET 1,2,3,† 1 Department
of Chemical Engineering and Applied Chemistry, University of Toronto, 4 Taddle Creek Road, Toronto, Ontario, M5S 3G9, Canada 2 Department of Chemistry, University of Toronto, 4 Taddle Creek Road, Toronto, Ontario, M5S 3G9, Canada 3 Institute of Biomaterials and Biomedical Engineering, University of Toronto, 4 Taddle Creek Road, Toronto, Ontario, M5S 3G9, Canada Received 12 November 2001; accepted 18 March 2002 Abstract—It has been recognized that a three-dimensional cell invasive scaffold that provides both topographical and chemical cues is desirable in regenerative tissue engineering to encourage cell attachment, migration, regrowth and ultimately tissue repair. Carbohydrate hydrogels are attractive for such applications because they are generally biocompatible and able to match the mechanical properties of most soft tissues. Although carbohydrate hydrogels have been previously modified with cell adhesive peptides and proteins, complicated hydrogel matrix activation was required prior to biomolecule coupling and, perhaps more importantly, the overall immobilization yield was low at ∼1%. In this study, we report the photo-immobilization of a model biomolecule, ovalbumin (OVA), to agarose gel. We describe two methods of modification where the photoactive moiety is coupled to either the protein (i.e. OVA) or the matrix (i.e. agarose) prior to immobilization. We found that the photo-immobilization yield depends on the location of the photoactive moiety. Using photoactive OVA, 1.8% of the OVA initially incorporated into the agarose gel is immobilized; using photoactive agarose, 9.3% of the OVA initially mixed with the agarose is immobilized. The latter is a significant improvement over previous yields and may be useful in attaining our goal of immobilizing a biomolecule gradient for guided tissue regeneration. Key words: Hydrogel; photochemistry; benzophenone; immobilization; nerve regeneration; tissue engineering; 3D scaffold.
∗ Current
address: Center for Engineering in Medicine, Harvard Medical School, Massachusetts General Hospital, Shriners Burns Hospital, 51 Blossom Street, Boston, MA 02114, USA. † To whom correspondence should be addressed. Phone: 416-978-1460. Fax: 416-978-4317. E-mail:
[email protected]
50
X. Cao and M. S. Shoichet
INTRODUCTION
Cells respond to their local environment, including both topographical features, such as three-dimensional (3D) organization of neighboring matrix textures [1], and chemical features, such as cues from fibrillar protein networks known as extracellular matrix (ECM) [2]. For regenerative tissue engineering, a three-dimensional cell invasive biomaterial scaffold that provides both chemical and topographical cues is critical for cell migration, regeneration and tissue organization. For example, we showed that a 3D biodegradable scaffold enhanced bone regeneration in vitro [3] and others have shown that a cell permissive hydrogel matrix [4], within a hollow fiber membrane, promoted nerve regeneration between two severed nerve ends. Hydrogels, such as agarose, are particularly attractive for use in soft tissue engineering applications because they are predominantly biocompatible [4]. Being inert to protein adsorption and physically stronger than most of the non-chemicallycrosslinked hydrogels at the same concentration, agarose has been widely used as a packing material in chromatography (e.g. Sepharose) [5– 7] and as a model hydrogel matrix to study cell permissive scaffolds for potential applications in regenerative tissue engineering [4]. Furthermore, because agarose is thermally reversible, it can be molded into different geometries for specific tissue engineering applications. Several strategies have been pursued to modify agarose via the hydroxyl groups. For example, biomolecules were covalently bound to cyanogen bromide-activated hydroxyl groups [8– 10]; however, because cyanogen bromide is highly toxic, 1,1 carbonyldiimidazole (CDI) activation is now more common. Although CDI activation resulted in a high degree of both activation and biomolecule immobilization yield [6, 11], it also resulted in crosslinking, which affected the thermal reversibility of agarose [12]. This has limited the utility of CDI activation for tissue engineering applications. Alternatively, UV irradiation has been applied to photoimmobilize cell adhesive peptides to agarose matrices. For example, Borkenhagen et al. [13] derivatized CDPGYIGSR with the photoactive benzophenone (BP) and immobilized the photoactive polypeptide conjugate to agarose matrices by UV irradiation in situ. This approach took advantage of relatively simple photochemistry to immobilize biomolecules to a 3D agarose gel, thereby providing an alternative way to derivatize hydrogel matrices for applications in tissue engineering. Unfortunately, this approach resulted in a low immobilization yield of ∼1% [12]. We recently reported the preparation of a well-defined diffusible neurotrophic factor concentration gradient in a three-dimensional agarose gel for guided neurite outgrowth [14]. We believe that an immobilized neurotrophic factor concentration gradient will also direct neurite outgrowth in a manner similar to the diffusible NGF concentration gradient, based on recent published data [15, 16]. While others have immobilized biomolecule concentration gradients on two-dimensional surfaces [17, 18], no one has immobilized a three-dimensional gradient, to the best of our knowledge. Creation of a 3D gradient is one of our ultimate goals, but before we could attempt this, we had to first determine a method to increase the
3D photoimmobilization of biomolecules
51
immobilization yield beyond the previously reported 1% [12]. We chose to use ovalbumin (OVA), as our model biomolecule, to study immobilization. Specifically, we compared the efficiency of two UV irradiation approaches to immobilize OVA within a 3D agarose gel matrix. By modifying OVA with the photoactive BP prior to immobilization in agarose, we were able to achieve an immobilization yield of 1.8%. By modifying agarose with BP prior to immobilizing OVA, we were able to increase the immobilization yield to 9.3%. The amount of protein immobilized to a substrate, due to BP photolysis, is almost directly proportional to that initially incorporated [19]. We have chosen to report the percentage of molecules immobilized as the measure of photoimmobilization yield because our ultimate objective is to immobilize a concentration gradient of neurotrophic factors in a three-dimensional hydrogel matrix to promote and guide nerve regeneration.
EXPERIMENTAL
General considerations Two different approaches were taken to photoimmobilize OVA to a 3D agarose matrix: (1) photoactive OVA and (2) photoactive agarose, as shown is Fig. 1. Materials and methods All chemicals were purchased from Sigma (St. Louis, MO) and used as received, unless otherwise indicated. 1-ethyl-3-(3-dimethylaminopropyl) carbodiimidehydrochloride (EDC) was purchased from Pierce (Rockford, IL) and used as received. Deionized distilled water was obtained from a Millipore Milli-RO 10 Plus and MilliQ UF Plus (Bedford, MA) and used at 18 M resistance. Photoactive OVA approach In this approach, OVA was first modified with the photoactive BP, then OVA-BP was dispersed into an agarose hydrogel and photoimmobilized in situ by UV-irradiation. Preparation of photoactive OVA OVA was modified with benzophenone (BP), by reacting the primary amine groups of OVA with 4-benzoylbenzoic acid succinimide ester (NHS-BP). In a typical reaction, a predetermined amount of NHS-BP was dissolved into 2 ml of N,Ndimethylformamide (DMF, anhydrous), to which 10 ml of 2.15 mg/ ml (w / v) OVA in phosphate-buffered saline (PBS, pH 7.4) solution was added dropwise with stirring. The reaction was carried out at 4◦ C under continuous agitation for a predetermined duration and was quenched by adding 50 µl ethanolamine. The resulting product (i.e. photoactive OVA-BP) was purified using 3000 molecular weight cutoff dialysis tubing (Spectrum, CA) against PBS. The bath was periodically changed
52
X. Cao and M. S. Shoichet
(a)
(b) Figure 1. Two approaches were used to photoimmobilize biomolecules to agarose: (a) in the photoactive biomolecule (OVA) approach, the photoactive moiety (*) is conjugated with the biomolecule (") prior to immobilization; and (b) in the photoactive agarose approach, the photoactive moiety (*) is covalently bound within the agarose gel prior to biomolecule (") photo-immobilization.
and monitored by UV-vis spectrophotometry until there was no (appreciable) absorbance at 260 nm associated with unreacted NHS-BP. All procedures were performed under subdued lights to avoid any unintended photoreaction. Two parameters, the molar ratio of NHS-BP to OVA and the reaction time, were varied to control the degree of modification of OVA-BP. Reactions were conducted at NHSBP to OVA molar ratios of 1 : 1, 2 : 1, 10 : 1, 20 : 1, 40 : 1, and 60 : 1 for 12 h. In a separate set of experiments, reactions were performed for 1 h, 2 h, 4 h, 8 h, 12 h, 24 h, and 36 h for a NHS-BP to OVA molar ratio of 20 : 1. Estimation of the degree of substitution The degree of substitution of OVA by NHS-BP was estimated according to Stocks et al. [20], by determining the unreacted primary amine groups of OVA using a fluorimetric assay. Briefly, photoactive OVA-BP was dissolved in PBS solution and was serially diluted to 0.3 mg/ ml, 0.15 mg/ ml, 0.075 mg/ ml and 0.0375 mg/ml. One half ml of fluorescamine in acetone (0.3 mg/ ml, w /v) was added to 1.5 ml of OVA-BP solution at each concentration. The mixture was then mixed by vortex. After 8 min, the fluorescence in the mixture was measured on a fluorimeter (Turner,
3D photoimmobilization of biomolecules
53
Model 450) with an excitation wavelength of 360 nm and emission at 420 nm. Identical procedures were repeated for unmodified OVA to determine the total titratable amine content. Radioactive labeling of photoactive OVA Tyrosine residues of photoactive OVA (OVA-BP) were labeled as previously described with radioactive iodine (125 I) [21]. Briefly, 6 ml of OVA-BP in PBS solution (2 mg/ ml) was reacted with 1 mCi of carrier-free Na 125I (Amershan Pharmacia Biotech, QC, Canada) in the presence of Iodobeads (Pierece, Rockford, IL) for 15 min. Free iodide was removed by successive passes through columns packed with anion-exchange resin (Dowex 1-X8, Aldrich). The radiolabeled OVA-BP was used to determine the photo-immobilization yield of OVA-BP in the agarose matrix. Determination of the duration of the photo-irradiation In order to determine the appropriate duration of photo-irradiation, photoactive OVA-BP was dissolved in PBS and was exposed to UV irradiation at 355 nm at different exposure times. The UV-vis spectra of the irradiated OVA-BP were monitored, and the gradually reduced absorbance at 260 nm associated with BP photolysis upon UV irradiation was used to determine the appropriate UVirradiation time. Determination of photoimmobilization yield of OVA-BP One ml of 2% agarose solution was mixed with 1 ml of I-125 radiolabeled OVABP in PBS solution (2 mg/ml) to form 2 ml of a 1% agarose solution. The resulting solution was cast into a 35 mm Petri dish and allowed to set at room temperature overnight. This procedure formed a 2 mm thick gel disk (hydrated state) that was then cut into pieces and separately weighed. The agarose/ benzophenonemodified OVA disks (OVA-BP-1) were irradiated at 355 nm for 45 min using a UV reactor equipped with 8 mercury lamps (Rayonet, Branford, CT). These samples were washed with washing buffer (0.1 M sodium bicarbonate and 0.1 M sodium iodide) until the radioactivity of the buffer was less than twice that of the background noise. Radioactivity was measured by a LKB 1282 Compugamma universal gamma counter (Fisher Scientific). Two additional agarose/ OVA-benzophenone disks served as controls: (1) where an identical disk was not irradiated but was washed, as described above (OVA-BP-2) and (2) where an identical disk was neither irradiated nor washed (OVA-BP-3). The radioactivities in all three agarose/ OVA-benzophenone disks (i.e. OVA-BP-1, OVA-BP-2 and OVA-BP-3) were then measured. Comparisons between the weight normalized radioactivities among the three disks were used to determine both the photoimmobilization yield (i.e. OVA-BP-1 vs. OVA-BP-3) and any nonspecific binding (i.e. OVA-BP-2 vs. OVA-BP-3).
54
X. Cao and M. S. Shoichet
Photoactive agarose approach In this approach, the photoactive moiety (i.e. BP) was immobilized to the agarose matrix first, and then native (or unmodified) OVA was photoimmobilized to the matrix. Preparation of photoactive poly(allylamine) [PA-BP] A 0.75 ml solution of 0.035 M NHS-BP in N,N-dimethylformamide (DMF) was added drop-wise to a 10 ml, 0.075% (w /w) poly(allylamine) (PA, Mw ∼ 17 000 g/ mol) in PBS solution, with continuous stirring at room temperature. Care was taken to avoid precipitation by controlling the rate of NHS-BP addition to the poly(allylamine) (PA) solution. The PA /NHS-BP solution was left to react overnight under subdued lights and then quenched by adding 50 µl ethanolamine. The resulting product (PA-BP) was purified by dialysis using 3000 molecular weight cut-off dialysis tubing (Spectrum, CA) against PBS. The bath was periodically changed and monitored by UV-vis until there was no appreciable absorbance at 260 nm due to unreacted NHS-BP. Oxidation of agarose OH to agarose COOH The primary hydroxyl groups of agarose (agarose OH) were oxidized to carboxylic acid groups (agarose COOH) according to a modified published procedure [22, 23]. Briefly, 1 g of agarose was suspended in 50 ml of water and stirred. Twenty mg of 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO) was added to the agarose suspension with 0.1 g of sodium bromide and 2.5 ml (13% w /w) sodium hypochlorite solution. The oxidization reaction was then initiated by adjusting the pH of the suspension to pH 10.8 with 1.0 M NaOH and the pH was maintained by continuously titrating 0.5 M NaOH to the suspension under well-stirred conditions. Since the oxidation quantitatively produced carboxylic acids that needed to be neutralized by NaOH, the extent of the oxidation was followed by the amount of the 0.5 M NaOH required to maintain the solution at pH 10.8 [24]. For example, if 6.4 ml of 0.5 M NaOH were added, it would indicate that the oxidation was complete, i.e. 100% of the primary hydroxyl groups were converted to carboxylic acid groups. Unless otherwise indicated, all agarose COOH samples had 0.8 ml of 0.5 M NaOH added which translates to 12.5% of the hydroxyl groups oxidized to carboxylic acid groups. The oxidation reaction was conducted at 0◦ C and the reaction was quenched by the addition of sodium borohydride (NaBH4 ) at the desired degree of oxidation. The oxidized agarose suspension was washed 5– 6 times by centrifugation-washing cycles in deionized distilled water. Subsequently, the oxidized gel was freeze-dried and stored at −20◦ C in the dark prior to use. FTIR absorbance spectra (Galaxy Series 5000 spectrometer) were taken of thin agarose films, to confirm the oxidation of agarose OH to agarose COOH. Thin agarose COOH films were prepared from 1% agarose solution, cast onto ZnSe disks and air-dried.
3D photoimmobilization of biomolecules
55
Coupling of poly(allyl amine)-benzophenone (PA-BP) to agarose COOH PA-BP was coupled to agarose COOH by EDC activation. Briefly, 0.1 g of agarose COOH was dissolved in 10 ml of water by microwaving for 45 s and then 0.3904 g (0.002 mol) of 2-(N-morpholino)-ethanesulfonic acid (MES) were added followed by the addition of 15 mg (7.8 × 10−5 mol) of EDC for 5 min. Then 10 ml of 0.075% (w /w) PA-BP in PBS solution were added. The reaction was carried out overnight at 37◦ C and the products were purified by dialysis (MWCO 50k, Spectrum, CA) against PBS. The resulting photoactive agarose gel is referred to as agarose-BP. Determination of photoimmobilization yield of native OVA to agarose-BP One ml of a 2% photoactive agarose solution in PBS (agarose-BP) was mixed with 1 ml of I-125 radiolabeled OVA in PBS solution (2 mg/ ml) to form 2 ml of a 1% agarose solution. This resulting solution was cast into a 35 mm Petri dish and set at room temperature overnight in a hydrated environment. The resulting 2 mm thick gel disk was then cut into pieces and separately weighed. The agarose-BP / OVA disks (agarose-BP-1) were irradiated at 355 nm for 45 min using a UV reactor equipped with 8 mercury lamps (Rayonet, Branford, CT). These samples were washed with washing buffer (0.1 M sodium bicarbonate and 0.1 M sodium iodide) until the radioactivity of the buffer was less than twice that of the background noise. Two additional agarose-BP / OVA disks served as controls: (1) where an identical disk was not irradiated but was washed, as described above (agarose-BP-2) and (2) where an identical disk was neither irradiated nor washed (agarose-BP-3). Radioactivity in all three disks (i.e. agarose-BP-1, agarose-BP2 and agarose-PB-3) was measured as before. Comparisons between the weightnormalized radioactivities of the three samples were used to determine both the photoimmobilization yield (agarose-BP-1 vs. agarose-BP-3) and the non-specific binding of OVA to agarose (agarose-BP-2 vs. agarose-BP-3). RESULTS
The goal of this study is to demonstrate the feasibility of photo-immobilizing biomolecules within a three-dimensional agarose gel matrix. In order to achieve this goal, two approaches were pursued using the OVA protein as the model biomolecule: photoactive OVA and photoactive agarose. Since we envision creating a protein gradient that is immobilized by photochemistry, we have focused on photoimmobilization yield and not on amount because the latter is proportional to the concentration of protein present [19]. Increasing the % yield is the challenge. Photoactive OVA approach In the photoactive protein approach, OVA was modified with BP, to form photoactive OVA-BP. OVA-BP was then dispersed in an agarose matrix and photoimmobi-
56
X. Cao and M. S. Shoichet
lized to the matrix in situ. The coupling of the BP to OVA was confirmed by UV-vis spectrophotometry with a significant absorbance at 260 nm. Figure 2 shows the UV-vis spectra of OVA-BP. Spectrum A shows a single broad absorbance peak due to the adjacent peaks at 260 nm of BP and 280 nm of OVA in OVA-BP molecule. No residual NHS-BP was detected in the washing buffer after the dialysis purification process, as evidenced by a lack of absorbance at 260 nm. It is unlikely that the significant absorbance observed at 260 nm of OVA-BP (inside the dialysis tubing) was due to unreacted NHS-BP because the dialysis tubing that was used had a MWCO of 3000 g/ mol, unreacted NHS-BP has a molecular weight of 323.2 g/ mol and OVA has an average molecular weight of 45 000 g/mol. As shown in Fig. 2, as the BP moiety was photolysed with increased UV exposure time, the absorbance at 260 nm decreased and reached a plateau value between 45 and 60 min. Thus, 45 min was chosen as the appropriate UV irradiation time in this study and was used in both photoactive OVA and photoactive agarose approaches. The degree of substitution of OVA by NHS-BP was determined by a fluorescent assay. when fluorescamine reacts with primary amines, it forms stable fluorophores, the fluorescence of which is proportional to the primary amine content [20]. Taking advantage of this reaction, we determined the number of remaining primary amine groups of OVA-BP by reacting it with fluorescamine and compared the fluorescent signal measured to a calibration curve, which was generated in a parallel experiment with native OVA without BP modification, following the identical procedure. Based
Figure 2. UV-vis spectra of OVA-BP were followed to determine the appropriate UV exposure time for photoimmobilization. The exposure times in the direction of the arrow are: 0, 3, 15, 30, 45, 60 min. (A) indicates the OVA-BP without UV irradiation; note the broadened peak.
3D photoimmobilization of biomolecules
57
on the number of primary amine groups of native OVA available for reaction with fluorescamine and the number of amine groups still present on OVA-BP, we were able to calculate the degree of substitution of OVA with BP by subtraction. As shown in Fig. 3, the amount of BP coupled to OVA can be controlled by either the molar ratio between NHS of NHS-BP and OVA (Fig. 3a) or reaction time (Fig. 3b). It is interesting to note that the substitution reactions between NHS-BP and OVA seem to reach a plateau at 14 amine groups per OVA molecule (Fig. 3b). Neither
(a)
(b) Figure 3. The degree of OVA substitution by NHS-BP can be controlled by either (a) the molar ratio of NHS to OVA or (b) reaction time.
58
X. Cao and M. S. Shoichet
an increase in NHS-BP:OVA molar ratio nor an increase in reaction time resulted in more than 14 OVA amines being substituted with NHS-BP. The increasing steric hindrance associated with the increasing bulky hydrophobic BP groups may limit further modification. The photoactive OVA-BP was immobilized to the agarose matrix by UV irradiation at 355 nm for 45 min. Using I-125 labeled OVA-BP, we calculated that only 1.8% of the photoactive OVA originally incorporated into the agarose matrix was immobilized and 0.5% of this could be attributed to the non-specific interaction of OVA with the agarose gel. This translates to that an immobilization amount of 0.4 × 10−6 mol of OVA in 1 l of agarose gel. Photoactive agarose gel approach In the photoactive agarose approach, agarose OH was oxidized to agarose COOH and poly(allylamine) was modified with benzophenone (PA-BP). The two, i.e. agarose COOH and PA-BP, were coupled using EDC and then native OVA was dispersed into the photoactive agarose gel and immobilized by UV irradiation. The hydroxyl groups of agarose were oxidized to carboxylic acids using the TEMPO radical as a mild oxidizer. The degree of oxidation was determined by the amount of NaOH added to maintain a constant pH 10.8 during the reaction. The oxidation reaction has been shown to proceed at an optimal rate at pH 10.8 [25]. Figure 4 compares the FTIR spectra of different agarose gels at different degrees of oxidation, thereby confirming the success of the TEMPO catalyzed reaction, converting agarose OH to agarose COOH. As shown in Fig. 4, the ratio of the carbonyl (C O) stretch peak at 1650 cm−1 to the in-plane bend of the hydroxyl (O H) peak at 1360 cm−1 increased with oxidation. For example, the ratio of the (C O) peak to the (O H) peak was (A) 0.67 for un-oxidized agarose; (B) 0.87 for 12.5% oxidized agarose COOH; and (C) 1.89 for 100% oxidized agarose COOH. It should be noted that the carbonyl peak at 1650 cm−1 overlaps with the scissoring motion of H O H, which may result in some artifact in the spectra due to the presence of water in the sample (cf. Fig. 4A). We found that the physical properties of the agarose gel changed with oxidation. Table 1 correlates the relationship between the degree of modification with the physical appearance of the modified agarose gel and its ability to re-gel. The modified agarose gel showed a significant increase in water solubility, especially at higher degrees of modification and thus a decreased ability to gel by the usual hydrogen bonding mechanism. Inclusion of PA-BP, at the concentrations studied, did not significantly alter the physical properties of the gel further. Our observations agree with those of Bellamkonda et al. [4] who showed that gel strength and porosity were not significantly affected by the presence of biomolecules. Interestingly, the 100% oxidized agarose had crosslinking properties that are similar to those of alginate. For example, the 100% oxidized agarose could be crosslinked with divalent calcium (using 1 nM CaCl2 ) to form a weak gel. The amine groups of poly(allylamine) were modified with benzophenone and then dialyzed to remove any unreacted NHS-BP. Little to no absorbance was observed at
3D photoimmobilization of biomolecules
59
Figure 4. Comparison of FTIR spectra of agarose at different degrees of oxidation: (A) native agarose; (B) 12.5% oxidation; (C) 100% oxidation. The peaks at 1650 cm−1 and 1360 cm−1 are the C O stretch and O H in-plane bend, respectively. Table 1. Correlation of the degree of oxidation, physical appearance and ability to re-gel of oxidized agarose Volume of NaOH added (ml)a
Theoretical degree of oxidation
Physical appearance after modification
Ability to re-gel after re-solubilization
6.4 3.2 1.6 0.8
100% 50% 25% 12.5%
transparent liquid viscous slag/ paste gel gel
N/ A Nob Yes — weak gel Yes — strong gel
a NaOH b Only
concentration of 0.5 M. gelled at pH 4.7 and 4◦ C (overnight).
260 nm (due to the unreacted NHS-BP) in the dialysate washing buffer, indicating quantitative conversion. Assuming 100% conversion, we estimate that 60 BP groups were present per poly(allyamine) which is similar to what has been reported in the literature [26] — i.e. 80 BP molecules per poly(allylamine) molecule. (This result is significantly greater than the maximum of 14 BP molecules incorporated per OVA molecule achieved in the photoactive OVA approach.)
60
X. Cao and M. S. Shoichet
Native OVA was immobilized to the photoactive agarose matrix (agarose-BP) by UV irradiation at 355 nm for 45 min. Using I-125 labeled OVA, we determined that 9.3% of the photoactive OVA originally incorporated into the agarose matrix was immobilized and ∼0.5% of this could be attributed to the non-specific interaction of OVA with the agarose gel. This corresponds to 2.1 × 10−6 mol of OVA immobilized within 1 l of agarose gel. Thus the photoimmobilization yield increased relative to that which had been previously reported in the literature of 1% [12] to 1.8% for the photoactive OVA approach and to 9.3% for the photoactive agarose approach.
DISCUSSION
We studied two general photoimmobilization methods, using OVA as the model molecule for its immobilization in a 3D hydrogel matrix. In comparison with other immobilization methods, photoimmobilizatin has the advantage of being simple and versatile to a variety of functional groups (or no functional groups at all). Furthermore, concerns about the bioactivity of UV-irradiated biomolecules have largely subsided after it was shown that the bioactivity is mostly unaffected when molecules are irradiated at wavelengths of 350 nm or longer [27]. Recently, patterned biomolecule surfaces were created by UV irradiation of photoactive-biomolecule conjugates through a photomask [28]. Moreover, agarose has been modified with a tripeptide, arginine-glycine-aspartic acid (RGD), via photo-irradiation [13]. This resulting agarose-RGD matrix elicited significant neurite outgrowth from dorsal root ganglion cells (DRGs) relative to native agarose gel, yet the immobilization yield was only about 1% [12]. We believe that this low yield could be attributed to the limited interactions between the hydrophobic photoactive BP and the hydrophilic environment (1% agarose and 99% water) in which the photoactive moiety was dispersed. To determine whether we could improve the photoimmobilization yield, we immobilized OVA using either a photoactive OVA or a photoactive agarose. Since effective photoimmobilization can only take place when the activated BP triplet radical is within 3.1 Å of an agarose molecule [29], the interactions between the BP and agarose likely dictate photoimmobilization yield. Thus we explored two ways to improve the interaction between hydrophilic agarose and hydrophobic BP, and thereby increase the photoimmobilization yield. In the photoactive OVA approach, OVA-BP was mixed with agarose prior to UVirradiation and resulted in a 1.8% yield. While this yield is slightly greater than that previously achieved [12], the yield is low. We postulate that this low yield may result from the BP moiety being ‘buried’ within the OVA core and thus unavailable for reaction with agarose. In the photoactive agarose approach, poly(allylamine) was modified with BP which was subsequently bound to agarose COOH prior to binding OVA by photo-irradiation. This method alleviates, at least in part, the poor hydrophobichydrophilic interaction between BP and agarose. Furthermore, we speculate that
3D photoimmobilization of biomolecules
61
the hydrophobic BP groups can better interact with the amphiphilic protein vs. the hydrophilic agarose by having greater proximity between the photoactive moiety and the protein molecule. This may explain why the greater immobilization yield of 9.3% was achieved using the photoactive agarose approach. Additionally, the use of poly(allylamine) effectively increased the amount of BP incorporated into agarose and this in turn improved the photoimmobilization coupling yield of OVA and agarose. Although it has been demonstrated that the inclusion of peptides/ proteins will not significantly change the porosity and mechanical strength of agarose gel, we found that the oxidation of agarose OH to agarose COOH decreased agarose’s ‘gel– ability’ (cf. Table 1). This problem could be circumvented by controlling the oxidation to lower degrees. We found that at a 12.5% degree of oxidation, agarose gelled with satisfactory mechanical integrity and thermal reversibility while having sufficient carboxylic acid groups for further modifications. Photochemistry has been widely used to pattern biomolecules with certain spatial features. For example, Borkenhagan and colleagues [13] photoimmobilized cell adhesive peptides in agarose gel using laser patterning. By varying UV exposure time, a concentration gradient of oligopeptide was immobilized on a 2-dimensional substrate over a 0.3 mm distance [17, 18]. We are also interested in creating a photoimmobilization protein gradient and focused on controlling the interactions between hydrogel matrices, biomolecules and photoactive moieties to improve the photimmobilization yield. Since the amount of protein photo-immobilized is proportional to the amount of protein present [19], it is possible o immobilize a biomolecule concentration gradient in 3-dimensions in situ. In ongoing studies, we are investigating methods to optimize the creation of a stable concentration gradient in a 3D hydrogel matrix.
CONCLUSIONS
We successfully modified agarose with OVA using photoimmobilization technology. We improved the coupling yield to 9.3%, which is a significant improvement over previously published yields of 1%. By using poly(allyamine), we were able to increase the number of BP groups introduced to agarose and available for reaction with OVA. This methodology is versatile to any protein or neurotrophic factor because it is the matrix that is modified, and not the bioactive molecule, prior to photoimmobilization. We are currently exploring methods of immobilizing a concentration gradient of neurotrophic factors for guided axonal regeneration. Acknowledgements The authors gratefully acknowledge partial financial support from the Ontario Neurotrauma Foundation and the Natural Sciences and Engineering Research Council of Canada.
62
X. Cao and M. S. Shoichet
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
D. M. Brunette and B. Chehroudi, J. Biomech. Eng. 121, 49 (1999). E. Ruoslahti, Science 276, 1345 (1997). C. E. Holy, M. S. Shoichet and J. E. Davies, J. Biomed. Mater. Res. 51, 376 (2000). R. V. Bellamkonda, J. P. Ranier and P. Aebischer, J. Neurosci. Res. 41, 501 (1995). M. T. W. Hearn, G. S. Bethell, J. S. Ayers and W. S. Hancock, J. Chromatogr. 185, 463 (1979). G. S. Bethell, J. S. Ayers, M. T. W. Hearn and W. S. Hancock, J. Chromatogr. 219, 353 (1981). N. Ubrich, P. Hubert, V. Regnault, E. Dellacherie and C. Rivat, J. Chromatogr. 584, 17 (1992). J. Porath, R. Axen and S. Ernback, Nature 215, 1491 (1967). R. Axen, J. Porath and S. Ernback, Nature 214, 1302 (1967). C. A. Schall and J. M. Wiencek, Biotechnol. & Bioengin. 53, 41 (1997). G. S. Bethell, J. S. Ayers and W. S. Hancock, J. Biol. Chem. 254, 2572 (1979). R. V. Bellamkonda and R. F. Valentini, in: Tissue Engineering Methods and Protocols, J. R. Morgan and M. L.Yarmush (Eds), p. 101. Humana Press, New Jersey (1999). M. Borkenhagen, J. F. Clemence, H. Sigrist and P. Aebischer, J. Biomed. Mater. Res. 40, 392 (1998). X. Cao and M. S. Shoichet, Neurosci. 103, 831 (2001). G. Gallo, F. B. Lefcort and P. C. Letourneau, J. Neurosci. 5445 (1997). Y. Ito, Biomaterials 20, 2333 (1999). C. B. Herbert, T. L. McLernon, C. L. Hypolite, D. N. Adams, L. Pikus, C. C. Huang, G. B. Fileds, P. C. Letourneau, M. D. Distefanno and W. S. Hu, Chem. & Biol. 4, 731 (1997). C. L. Hypolite, T. L. McLernon, D. N. Adams, K. E. Chapman, C. B. Herbert, C. C. Huang, M. D. Distefanno and W. S. Hu, Bioconj. Chem. 8, 658 (1997). J. M. R. Parker and R. Hodges, J. Protein Chem. 3, 465 (1984). S. J. Stocks, A. J. M. Jones, C. W. Ramey and D. E. Brooks, Anat. Biochem. 154, 232 (1986). Y. W. Tong and M. S. Shoichet, J. Biomed. Mater. Res. 42, 87 (1998). P. L. Bragd, A. Besemer and H. van Bekkum, Carbohydr. Res. 328, 355 (2000). A. Isoga and Y. Kato, Cellulose 5, 153 (1998). P. S. Chang, J. F. Robyt, J. Carbohydr. Chem. 15, 819 (1996). A. E. J. Nooy, A. Besemer and H. Bekkum, Carbohydr. Res. 269, 89 (1995). Y. Ito, in: Trans. Sixth World Biomater. Congress, p. 455 (2000). H. Sigrist, A. Collioud, J.-F. Clemence, H. Gao, M. Sanger, G. Sundarababu, Optical Eng. 34, 2339 (1995). Y. Ito, G. Chen and Y. Imanish, Bioconj. Chem. 9, 277 (1998). G. Dorman and G. D. Prestwich, Trends in Biotech. 18, 64 (2000).
Fibrinogen adsorption by PS latex particles coated with various amounts of a PEO/PPO/PEO triblock copolymer M. BOHNER 1,∗ , T. A. RING 2 , NATALYA RAPOPORT 1 and K. D. CALDWELL 1,†,‡ 1 Center for Biopolymers at Interfaces, University of Utah, Salt Lake City, UT 84112, USA 2 Department of Chemical and Fuels Engineering, University of Utah, Salt Lake City,
UT 84112, USA Received 10 January 2002; accepted 3 April 2002 Abstract—Polystyrene (PS) latex particles of different sizes were adsorption coated with the polymeric surfactant Pluronic F108 (PEO129 -PPO56 -PEO129 ). The commercial surfactant was found to have a bimodal molecular weight distribution. However, the maximum surface concentrations resulting from adsorption of either the purified high molecular weight component or the composite were identical. An increase in the copolymer surface concentration on 252-nm particles was found to decrease their fibrinogen uptake exponentially. At maximum copolymer surface concentration, the fibrinogen uptake was two orders of magnitude lower than that of bare particles (down from 3.3 mg/ m2 to 0.03 mg/ m2 ). This surface protection was equally effective whether the adsorption involved the bimodal polymer surfactant or the purified high molecular weight fraction. The PEO tail mobility was investigated with electron paramagnetic resonance (EPR), and found to increase with an increase in polymer surface concentration. The comparatively slow motion of the PEO chains at low surface concentration indicated that not only the PPO block, but also the PEO blocks interacted hydrophobically with the PS surface. When the copolymer surface concentration was increased, the PEO tails were gradually being released, acquiring higher mobility as the surface became covered by the more strongly binding PPO blocks. Results obtained with F108 coated particles of different sizes showed that particle size had a significant effect on the fibrinogen uptake, with larger particles showing larger fibrinogen uptakes. Key words: Adsorption; isotherm; particle; polymer; chain mobility; fibrinogen.
∗ Present
address: Dr. h.c. Robert Mathys Foundation, Bischmattstr. 12, CH-2544 Bettlach, Switzerland. † Present address: Center for Surface Biotechnology, Uppsala University, PO Box 577, 75123 Uppsala, Sweden. ‡ To whom correspondence should be addressed.
64
M. Bohner et al.
INTRODUCTION
In the Biomaterials field of research, few persons have generated as much insight into the interactions between plasma proteins and polymeric surfaces as has John Brash. It is an honor to be able to participate in the celebration of this eminent scientist and generous colleague. The ability of poly(ethylene glycol) (PEG) to provide a protein repelling shield to surfaces of various kinds has been of significant interest in the biomaterials field, and has stimulated a large number of research groups during the past decades to study the nature of this repulsion with the aim of optimising surface protection. Long before PEG, or poly(ethylene oxide)(PEO), was immobilized to surfaces, the highly water soluble polyether was known to exclude proteins and other polymers from its immediate vicinity in aqueous environments [1– 4]. For a given monomer concentration, this exclusion increases with polymer molecular weight up to values of around 6000 Da, beyond which it becomes virtually molecular weight insensitive. The polymer is known to interact with the surrounding water via hydrogen-bond formation, whereby its ether oxygens participate as acceptors. From calorimetric observations it has been suggested that each EO monomer associates with three water molecules in the presence of excess water [5], making the polymer highly water-logged and characterized by a large excluded volume. The covalent surface attachment of PEG is by now a well-accepted way to reduce protein adsorption on surfaces in contact with biological fluids. This reduction has shown to be more effective the longer the PEG chain, up to molecular weights of around 3500– 5000 Da [6, 7]. In this respect, the surface exclusion resembles that of the polymer in solution. Optimizing the repulsion of end-linked polymer chains in terms of both chain length and grafting density is however a difficult proposition, since the two properties tend to influence one another, so that longer chains result in lower grafting density [8]. A related characteristic with postulated influence over the repelling power of a PEG surface layer is chain mobility [9– 12]. According to a repulsion model proposed by Andrade et al. [13, 14], there is an entropy penalty associated with the adsorption of a bulky protein molecule on a surface covered with highly mobile polymer chains. In the absence of a net bond formation between the protein and polymer chains, the increased crowding caused by an approaching protein, and the resulting loss of chain dynamics, tends to counteract the adsorption. The desire to understand surface protection by PEG has therefore in many quarters become a search for the optimal grafting density, chain length, and chain mobility. The ability to engineer self-assembled PEG-terminated alkane monolayers (SAMs) on ideally flat metal surfaces has offered some new and interesting insight into the protective qualities of the EO moiety [15, 16]. In work with gold films covered with oligo(ethylene glycol)-terminated SAMs, the group of Whitesides [17, 18] has used surface plasmon resonance (SPR) detection to examine the protein adsorption from aqueous solution, and observed little or no uptake. From observations on such highly ideal surfaces the authors conclude that neither chain length nor mobility have an influence on the repulsion. Rather, the effect is pro-
Fibrinogen adsorption by PS latex particles
65
posed to result from the structure of water proximal to the surface layer [19– 21]. Particular support for this notion is given by the group’s comparative studies of SAMs on silver and gold substrates, where the silver substrate is seen to force the oligo(ethylene glycol) into a planar configuration, while the gold substrate induces the formation of a helix [20]. The two configurations have different influence on their vicinal water, and in protein uptake studies the gold surface shows repelling characteristics while the silver surface appears highly adsorbing. It is clear that both the physical and chemical nature of a surface plays a role, not only in its ability to be protected by PEG, but in how this protection is executed as well. Polymeric biomaterials, and surfaces for diagnostic and analytical use, are typically not as ideally flat as the gold surfaces used in SPR analysis. Instead, they are generally molded into, at times intricate, forms suitable for device fabrication. Such surfaces are difficult to derivatize to a uniform and predetermined degree, and their protection has to be provided in a different way. Due to the hydrophobic nature of many polymeric materials, the protection has been afforded by adsorption, through hydrophobic interaction, of some PEO-containing block copolymers whose anchoring block is a hydrophobic polymer, e.g. poly(propylene oxide) (PPO) [22– 26]. The simple adsorption has been shown to yield complexes of such stability that they can withstand continuous circulation in the blood stream of test animals for numerous hours [23, 26, 27]. This intriguing quality prompted a series of studies in our laboratory with the aim of establishing the optimal PEO chain length and surface density for protein repulsion [28– 30]. In conformity with more recent observations of the adsorption of PEO-PPO-PEO triblock surfactants at the air-water interface [31] we found the close packing on hydrophobic polystyrene (PS) surfaces to be independent of the PEO chain length, and instead primarily governed by the length of the PPO block. By coating PS nanoparticles with triblocks whose centers were of comparable size, but whose flanking blocks were of different lengths, it was thus possible to examine the effect of PEO chain length on protein repulsion in a manner independent of surface density. Similar to the exclusion efficiency of soluble PEO, which appears to taper off at molecular weights above 5000 Da, the efficiency of adsorbed, PPO-linked PEO became less sensitive to molecular weight for chain lengths above 100 EO units (4400 Da) in terms of its ability to suppress adsorption of the plasma protein fibrinogen. Of the triblocks tested (all of the Pluronic® type) the most effective surface protectant was Pluronic F108 with the general composition (EO)129-(PO)56 -(EO)129 . The question of the role of polymer dynamics on this repulsion came in renewed focus by our preliminary ESR studies of spin-probe labelled F108 adsorbed to PS particles of different size [29]. Here, adsorption complexes involving a small (69 nm) and a large (272 nm) particle fraction gave rise to different surface densities of F108, in turn resulting in different PEO chain mobilities. In addition, the higher chain mobility on the smaller particle appeared coupled to a lower fibrinogen uptake by this particle. If, as in these observations, a lower surface concentration of a specific PEO chain length were to give rise to a higher chain mobility, and thereby to a more
66
M. Bohner et al.
effective protein repulsion, there might exist an optimal surface density for any of these protective coatings. Such an optimum would be similar to the one observed by Blume and Cvec [32] for the longevity of in vivo circulating PEG-ylated liposomes, which reached a maximum at a PEG surface concentration below the one maximally achievable. The present work was undertaken to examine this hypothesis. Specifically, we wished to systematically vary the surface concentration of Pluronic F108 on one substrate type, and examine the effects of this concentration on chain mobility and protein repulsion. Similar to our previous work, fibrinogen was selected as the probe protein due to its important function in the blood’s response to e.g. implanted cardiovascular devices [23].
EXPERIMENT
Materials Pluronic F108 was kindly provided by BASF Corp (N◦ 583062, Lot N◦ WP10549B). The molecular weights of its PEO and PPO blocks are nominally about 5.68 and 3.25 kDa respectively, for a total molecular weight of approximately 14.6 kDa. The F108 molecular weight distribution was measured by gel permeation chromatography (GPC), using a liquid chromatography controller LCC-500 linked to two P-500 pumps and one peristaltic pump P-1 (all devices from Pharmacia, Sweden). The column Superose 12 was purchased from Pharmacia (N◦ 17-053801) and calibrated with PEO standards of known molecular weight. Sodium azide (NaN3 ; Sigma N◦ S-2002, lot N◦ 43H0291), TRIZMA® base (Sigma N◦ 1503, lot N◦ 64H5767), and sodium chloride (Mallinckrodt AR® N◦ 7581, lot N◦ 7581KMJC) were used to prepare the elution buffer (0.05 M TRIS + 0.5 M NaCl; pH adjusted to 8.0 with concentrated HCl). The refractive index of the eluate was measured with a differential-refractometer (Knauer). F108 was found to consist of two fractions of which the larger was collected and dialysed using a polymer membrane of type Spectra/ Por® 7, purchased from Spectrum (N◦ 132 104, Mw cut-off = 1 kDa), and used for adsorption experiments under the name of ‘large F108 component’. The latex particles were purchased from Seradyn and Bangs Laboratories (see Table 1), while the fibrinogen was obtained from KABI Vitrum (Sweden). Sample preparation The adsorption experiments were made in Eppendorf tubes (1.5 ml). All reactants added to the tubes were weighed. Deionized water and the desired amounts of F108 solution (taken from a previously-prepared stock solution) were rapidly mixed in an Eppendorf tube. The latex particles were then added and the sol (total volume 1 ml) was mixed end-over-end for 24 h. The particle concentration was always 0.40% w /w. However, the initial F108 concentration varied depending on the size of the
Fibrinogen adsorption by PS latex particles
67
Table 1. Listing of the particles used in this study. The particle size was measured by PCS and SedFFF Nominal size (nm)
PCS size (nm)
SedFFF size (nm)
Manufacturer
127 165 214 252 314 360
136.2 ± 0.8 167.6 ± 1.2 212.7 ± 1.3a 251.6 ± 1.7a 348.1 ± 2.1a 357.8 ± 2.4a
— — 214.4 ± 0.2 251.5 ± 0.4 351.7 ± 1.5 355.2 ± 1.3
Bangs Bangs Bangs Seradyn Bangs Bangs
a These
samples were first filtered by SedFFF.
particles: 0.075% w /w for nominal sizes 127, 165 and 214 nm; 0.050% w /w for nominal sizes 314 and 360 nm; and in the range of 0.004 to 0.050% w /w for the nominal size of 252 nm. Altogether, four adsorption isotherms were measured with 252 nm particles: two with F108, one with spin-labeled F108, and one with the large F108 component. All samples, except those coated with the spin-labeled F108, were then coated with fibrinogen. For the other particle sizes, two repeats of three samples were prepared. Before exposure to protein, the F108 coated samples were washed with a phosphate buffer solution (PBS, I = 0.15 M, pH 7.4) using multiple centrifugation/ resuspension steps (16000g in an Eppendorf Model 5414 table top centrifuge). After the first centrifugation, 0.6 ml of the supernatant was removed and its F108 content measured by means of the Baleux method [34] (see below). Knowing the initial F108 concentration, and assuming that the PS latex particles were totally deflocculated, the amount of F108 on the particle surfaces could be easily calculated. During the centrifugation/ resuspension steps, the particles were concentrated by a factor of 10 (4 mg particles in 0.1 ml PBS). This was necessary in order to accurately measure the protein uptake by the F108-coated latex particles. Uptake measurements were based on both direct and indirect analysis and were made after incubation of the particle sols with 0.9 ml aliquots of a fibrinogen solution (0.3 or 1 mg/ ml) under end-over-end mixing for 2 or 4 h. The supernatants were then separated from the particles by centrifugation, as above (16000g for 3 h). In the indirect analysis the protein concentrations of the supernatants were determined by UV spectroscopy (Perkin-Elmer Lambda 6 spectrophotometer) at a wavelength of 280 nm, using a solution of known concentration (determined by amino acid analysis, AAA) as a standard. The amount of fibrinogen adsorbed on the particles was then calculated from the difference in concentration before and after exposure to the particles. The surface area of the particles was assumed to be that corresponding to a monodisperse particle dispersion. In the direct analysis, the pelleted particles were carefully washed by multiple centrifugation/ resuspension steps in which the supernatants were removed between spins and replaced by fresh
68
M. Bohner et al.
buffer. After 4 washing cycles, the particles were freeze-dried and submitted to AAA in order to measure their protein content. Characterizations The size of particles was measured by means of photon correlation spectroscopy (PCS) using a Brookhaven Model BI-100M goniometer with a BI-2030AT autocorrelator at 25 ◦ C. If the samples were not previously run through a sedimentation field-flow fractionation (SedFFF) [28] unit, the suspensions used for the measurements (in 0.1% FL-70 solution; FL-70 purchased from Fischer Scientifics) were passed through a 3 µm filter paper. At least 10 measurements were made on each sample. At least 3 repeats were done (usually on different days), so that the particle size was obtained from an average of at least 30 measurements. The sizes of bare particles were also measured by SedFFF according to the procedure described by Li and Caldwell [28]. The SedFFF samples contained 2.5% w /w particles, and 3% w /w F108. The PCS and SedFFF sizes are given in Table 1. The depletion method used to measure the F108 surface concentration on the PS latex particles has been described by Baleux [34] and will therefore be referred to as ‘the Baleux method’ (BM). When an iodine solution is added to a PEO compound, a reddish brown complex forms which can be quantified spectrophotometrically. A sample of the F108 solution (supernatant obtained in ‘Materials’) was diluted with deionized water (final volume: 5 ml) so that the final F108 concentration was in the range 0 to 20 mg/ l (0 to 0.002% w /w). A 1.25 ml aliquot of a KI3 solution, prepared by dissolving 1 g I2 (Mallinckrodt AR® N◦ 1008, lot N◦ 1008KECX) and 2 g KI (Mallinckrodt AR® N◦ 1127, lot N◦ 1127KLHE) in 100 ml of deionized water, was then added to the diluted F108 solution. After a minimum of 5 min, the absorbance was measured at a wavelength of 500 nm. The F108 concentration of the diluted solution, and therefore of the initial solution, could be calculated using a calibration curve made fresh for each batch of reagent. The AAA procedure has been summarized by Caldwell et al. [35]. It can be briefly described as consisting of three steps: (1) 20 h hydrolysis of the freeze-dried sample in 6N HCl at 105 ◦ C; (2) derivatization with phenyl isothiocyanate; (3) analysis by a reverse phase liquid chromatography system (Hewlett– Packard, Model HP1050). The total protein content was determined as the sum of recorded amino acid residues. This value will not include the content of tryptophan, which is destroyed in the hydrolysis. To measure the dynamics of the coating layer, the PEO tails were spin-labelled by introducing 3-carboxy-proxyl free radicals at the chain termini [29], and the probe mobilities were measured by electron paramagnetic resonance (EPR). EPR spectra were recorded at room temperature with an X-band Bruker (Billerica, MA) ER200 SRC EPR spectrophotometer with the following instrument parameters: central field at 3475 G; sweep width, 100 G at 200 s; 100 kHz modulation frequency with modulation amplitude typically set to a quarter of a line width; incident microwave power was set to 0.5– 2 mW to avoid saturation. Rotational correlation time (τR ) in
Fibrinogen adsorption by PS latex particles
the fast motion regime was calculated using the following equation [36]: τc = 6.73H0 I0 /I−1 − 1 10−10 , s,
69
(1)
where H0 is peak-to-peak width of the midfield line (G), and I0 and I−1 represent the intensities of the mid- and high field peaks, respectively.
RESULTS AND DISCUSSION
Adsorption of Pluronic F108 The size exclusion analysis of our Pluronic F108 sample indicated the product to be bimodal in its molecular weight distribution, as seen from Fig. 1. As a consequence of this complex composition, the adsorption isotherm that was determined from contacts between solutions of this composite product in different concentrations and fixed amounts of 252-nm PS particles departs from the shape of a Langmuir isotherm [37– 39], as seen from the lower two traces in Fig. 2. At low F108 concentrations, both components adsorb on the PS latex surface [37]. When the surface is fully covered with F108 (in this case at a surface density of about 0.5 mg/ m2 ), the molecules with longer PPO anchoring blocks begin to displace those with shorter PPO blocks in a process that continues until the surface is totally covered with longer polymer chains (at a surface density of about 1.8 mg/ m2 in these measurements). This displacement effect is clearly seen by comparing the isotherms of the composite sample with that determined for the large F108 component alone (Fig. 2). The linear domain so prominently featured in the middle of the composite isotherm has disappeared in the isotherm for the purified large fraction, even though its initial and final surface densities are identical to those of the composite F108. The shifts in the F108 adsorption isotherm provoked by the presence of a small F108 component has been thoroughly discussed elsewhere [39].
Figure 1. F108 GPC chromatogram. The largest and the smallest fraction (left peaks) have molecular weights of 13700 ± 100 and 6100 ± 100 Da, respectively. This has been determined using the calibration curve: log(Mw ) = 6.67 − 0.0839 t (t = elution time; r 2 = 0.993). The right peak corresponds to the solvent.
70
M. Bohner et al.
Figure 2. Adsorption isotherms of F108 (!,P — two independent determinations) and of the large F108 component (1) on PS latex particles. Each mark corresponds to an average of two concentration measurements done on the same sample. Table 2. Particle size effect on protein uptake PCS size (nm)
F108 surface concentration (mg/ m2 )
Fibrinogen surface concentration (mg/ m2 )
136.2 ± 0.8 167.6 ± 1.2 212.7 ± 1.3 251.6 ± 1.7 348.1 ± 2.1 357.8 ± 2.4
1.90 (0.06)a 2.18 (0.06) 2.72 (0.07) 1.96 (0.10) 1.78 (0.13) 2.26 (0.41)
0.008 (0.004)a 0.011 (0.003) 0.019 (0.015) 0.038 (0.008) 0.046 (0.009) 0.041 (0.029)
a Standard
deviation (n = 6).
It should be noted that the plateau concentration of F108 reported here for the Seradyn 252 nm particles is lower than that typically measured in our laboratories (e.g. 2.8 mg/ m2 for the 272 nm particles in Ref. [29], or 2.72 mg/ m2 for the 214 nm particles in Table 2). Protein adsorption The fibrinogen uptake by the F108-coated PS latex particles continuously decreases with increasing polymer surface concentration, whether of the composite F108 or of the large F108 component alone, and eventually reaches a plateau (Fig. 3). At low F108 surface coverage the surfactant tends to adsorb to more than one particle at a time (unpublished observations), causing a significant degree of flocculation. This is the reason why no results are shown for low supernatant concentrations of F108. The fibrinogen uptakes measured by AAA were always slightly lower than those measured by UV spectroscopy (Fig. 4). Contrary to the UV spectroscopic method,
Fibrinogen adsorption by PS latex particles
71
Figure 3. Effect of the F108 concentration on the fibrinogen uptake by F108-coated PS latex particles. The results were obtained by AAA for: (!,P — two independent determinations) F108, and for: (1) the large F108 component.
Figure 4. Comparison of the results obtained by AAA and UV spectroscopy.
which accounts for all protein removed from solution, whether by adsorption to particles or to container walls, the method using AAA is a direct measurement of the amount of protein left on the surface of the particles after careful washing. The difference between the two methods could, therefore, in part be due to some fibrinogen desorbing during the washing steps that preceded the AAA. A comparison of Figs 2 and 3 shows that the minimum in protein uptake appears to coincide with the maximum in F108 surface coverage. Rather than the hypothesized existence of an optimal surface concentration of F108, as the PEO chains transition from a ‘mushroom’ to a ‘brush’ configuration according to the notion of deGennes [40], the data in Fig. 5 seem to indicate a complete covariance between the decrease in fibrinogen uptake and the increase in polymer close-packing on these PS particles.
72
M. Bohner et al.
In addition, the figure shows the surface protection to be equally efficient, regardless of whether the polymer adsorption had taken place from solutions of the composite surfactant or from solutions containing only the purified high molecular weight F108 component. At maximum F108 surface coverage, the fibrinogen uptake is decreased by two orders of magnitude, and is found to be close to 0.03 mg/ m2 . This result is similar to findings by Li et al. [25, 41], who observed a maximum surface coverage of 3.3 mg/m2 on uncoated PS latex particles and a minimum protein uptake of 0.021 and 0.038 mg/ m2 on F108-coated latex particles (respective diameters: 69 and 272 nm) after exposure to fibrinogen solutions containing 0.5 mg/ ml. It should be noted that these fibrinogen surface concentrations are comparable to those found on protected surfaces of different kinds, e.g. PEOgrafted silica (<0.05 mg/ m2 ) [42] and octa(ethylene oxide) terminated SAMs (<0.1 mg/m2 ) [43]. The influence of particle size on fibrinogen uptake that was noted in our previous work [25, 41] was verified in the present study, as seen in Table 2. Unfortunately, due to the rather large scatter in the measured F108 uptakes for the differently sized particles, it is impossible from these data to postulate an explanation for the observed protein uptakes based on polymer surface arrangement. However, whether due to the presence of more mobile polymer chains on the smaller particles than on the large ones, or simply due to the geometric consequences of fitting large protein molecules (the fibrinogen molecule is ca. 40 nm along its major axis) onto small particles, the trend of increasing protein uptake with increasing particle size seems well established by the data in Table 2. Similar observations have been reported by Blunk et al. [44], who studied the uptake of plasma proteins (unfractionated) by similarly coated PS particles. The adsorbing protein in the latter case consisted not only of fibrinogen, but also of the numerous smaller plasma components, and the geometric argument must therefore be used with some caution.
Figure 5. Effect of F108 surface density on the fibrinogen uptake of F108-coated PS latex particles. The results were obtained by AAA for: (!,P — two independent determinations) F108, and for: (1) the large F108 component.
Fibrinogen adsorption by PS latex particles
73
Surface dynamics The adsorption isotherm for the spin-labeled F108 is nearly identical to that of the non-labeled F108 (Fig. 6). Although the plateau appears to be slightly lower with the spin-labeled F108, the experimental error makes it impossible to know whether this effect is due to the presence of the spin probe or to the experimental procedure. The similarities between the two isotherms suggest, however, that the terminally attached spin labels at the PEO blocks of the F108 copolymer do not significantly modify its packing on the surface. The rotational frequency (i.e. the inverse of the rotational correlation) is a measure of the mobility of the spin labels. Since these labels are attached to the ends of the PEO chains, the data shown in Fig. 7 are assumed to reflect the mobility of the PEO chain ends in the F108-PS adsorption complexes. From this figure, the maximum value for (τR )−1 is found to be about 0.8 × 10+10 s−1 , which corresponds to the
Figure 6. Adsorption isotherm of spin-labelled F108 on PS latex particles (F). Comparison with the adsorption isotherm of unlabeled F108 (!).
Figure 7. Variation of the PEO tail mobility (expressed as (τR )−1 ) as a function of the amount of spin-labelled F108 adsorbed on the surface of the PS latex particles.
74
M. Bohner et al.
value measured by Li et al. for 272-nm PS particles coated with spin-labeled F108 (0.9 × 10+10 s−1 ) [29]. However, contrary to expectations, the mobility is seen to increase with an increasing density of PEO chains on the surface. According to our original hypothesis, a surface density that was low enough to allow the PEO chains to establish a ‘mushroom’ configuration, in the deGennes terminology [40], should have displayed a higher chain mobility than a more close-packed arrangement in which the chains were forced into ‘brush’-like structures. To be sure, this type of behavior would only occur in cases where the end-group tethered polymer chains had no affinity for the substrate. In the present case, therefore, the possibility exists that a low surface density of PEO chains would allow these slightly hydrophobic (critical surface tension 43 dyn/ cm compared to the surface tension of 73 dyn/ cm for water) polymers to establish a significant degree of hydrophobic interaction with the bare PS substrate [45– 47]. If so, the mobility of end-attached PEO on PS surfaces would be analogous to that already observed for PEO chains grafted onto silica surfaces. In the latter case the EPR technique was used to show that when the polymer chains are present at low surface density they are relatively immobile, but that their mobility increases with an increase in surface concentration [48]. In the case of silica substrates, the fettering of the PEO chains at low concentration was postulated to be due to hydrogen bonding between ether oxygens in the chains and silanols on the substrate surface.
CONCLUSION
In accordance with previous findings, this study has shown that the surface density of the Pluronic F108 triblock copolymer (PEO-PPO-PEO), adsorbed on PS latex particles, has a strong effect on their fibrinogen uptake, with a monotonic decrease in protein uptake resulting from an increase in F108 surface concentration. Specifically, a Pluronic F108 surface density of 2 mg/ m2 is found to reduce the fibrinogen uptake of F108-coated PS latex particles by as much as two orders of magnitude, compared to that seen on unshielded particles. This reduction is identical, whether the shielding is accomplished through adsorption of unfractionated F108 or its purified high molecular weight component, which is the major constituent in the commercial Pluronic F108 product. The increased shielding efficiency of the polymeric surfactant is directly correlated with an increase in its PEO tail mobility. This is contrary to our hypothesized existence of an intermediate surface concentration with a maximum in chain mobility and repulsion efficiency. Since the mobility instead is found to increase monotonically with the close-packing of PEO chains on the PS surface it is impossible to interpret the surface shielding as being due to either polymer dynamics or to steric exclusion effects. Doubtless, both effects are operative in these systems. The relatively slow motion of the PEO chains at low surface concentration is an indication of the weak hydrophobic interaction known to exist between PEO and PS.
Fibrinogen adsorption by PS latex particles
75
As the surface concentration of F108 increases, the stronger hydrophobicity of its PPO center blocks give rise to a gradual displacement of the adhering PEO flanking blocks, which thereby gain in mobility. Fibrinogen uptake by differently sized F108 coated PS particles appears to increase with increasing size. The reason for this effect, which has been seen by others as well, is unclear at this point and requires further investigation. Acknowledgements The authors would like to thank Dr. Jenq-Thun Li for his collaboration, Dr. Pavla Kopeckova for the GPC measurements and Dr. Lei Shi for the AAA measurements. Marc Bohner gratefully acknowledges the ‘Fond National Suisse pour la Recherche Scientifique’ for giving him a one-year grant.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.
R. Bhat and S. N. Timasheff, Protein Science 1, 1133 (1992). K. C. Ingham, Arch. Biochem. Biophys. 186 (1), 106 (1978). D. H. Atha and K. C. Ingham, J. Biol. Chem. 256 (23), 12108 (1981). C. R. Middaugh, W. A. Tisel, R. N. Haire and A. Rosenberg, J. Biol. Chem. 254, 367 (1979). N. B. Graham, M. Zulfiger, N. E. Nwachuku and A. Rashid, Polymer 30, 528 (1989). W. R. Gombotz, W. Guanghui, T. A. Horbett and A. S. Hoffman, J. Biomed. Mat. Res. 25, 1547 (1991). C.-G. Gölander and E. Kiss, J. Colloid Interface Sc. 121 (1), 240 (1988). J. D. Andrade, D. L. Coleman, P. Didisheim, S. R. Hanson, R. Mason and E. W. Merrill, Trans. Am. Soc. Artif. Internal Organs 27, 659 (1981). E. W. Merrill, Ann. N. Y. Acad. Sci. 283, 6 (1977). S. Nagaoka, Y. Mori, H. Takiuchi, K. Yokota, H. Tanzawa and S. Nishiumi, in: Polymer as Biomaterials, S. W. Shalaby, A. S. Hoffman, B. D. Ratner and T. A. Horbett (Eds), p. 361. Plenum Press, New York (1984). V. P. Torchilin and M. I. Papisov, J. Liposome, Res. 4, 725 (1994). M. A. Carignano and I. Szleifer, Colloids and Surfaces B. Biointerfaces 18, 169 (2000). S. I. Jeon, L. H. Lee, J. D. Andrade and P. G. de Gennes, J. Coll. Interface Sci. 142, 149 (1991). S. I. Jeon and J. D. Andrade, J. Coll. Interface Sci. 142, 159 (1991). K. L. Prime and G. M. Whitesides, Science 252, 1164 (1994). R. Valiokas, S. Svedhem, S. C. T. Svensson and B. Liedberg, Langmuir 15, 3390 (1999). E. Otsuni, L. Yan and G. M. Whitesides, Colloids and Surfaces B. Biointerfaces 15, 3 (2000). C. Roberts, C. S. Chen, M. Mrksich, V. Martichonsk, D. E. Ingber and G. M. Whitesides, J. Am. Chem. Soc. 120, 6548 (1998). A. J. Pertsin, M. Grunze and I. A. Garbuzova, J. Phys. Chem. B 102, 4918 (1998). P. Harder, M. Grunze, R. Dahint, G. M. Whitesides and P. E. Laibinis, J. Phys. Chem. B 102, 426 (1998). K. Feldman, G. Hähner, N. D. Spencer, P. Harder and M. Grunze, J. Am. Chem. Soc. 121, 10134 (1999). M. Amiji and K. Park, Biomaterials 13 (10), 682 (1992). J. Lee, P. A. Martic and J. S. Tan, J. Colloid Interface Sci. 131 (1), 252 (1989). J. H. Lee, J. Kopecek and J. D. Andrade, J. Biomed. Mater. Res. 23, 351 (1989).
76
M. Bohner et al.
25. J. S. Tan, D. E. Butterfield, C. L. Voycheck, K. D. Caldwell and J.-T. Li, Biomaterials 14, 823 (1993). 26. L. Illum, L. O. Jacobsen, R. H. Müller, E. Mak and S. S. Davis, Biomaterials 8, 113 (1987). 27. N. Watrous-Peltier, J. Uhl, V. Steel, L. Brophy and E. Merisko-Liversidge, Pharm. Res. 9 (9), 1177 (1992). 28. J.-T. Li and K. D. Calwell, Langmuir 7, 2034 (1991). 29. J.-T. Li, K. D. Calwell and N. Rapoport, Langmuir 10, 4475 (1994). 30. J.-T. Li, K. D. Calwell and J. S. Tan, in: Particle Size Distribution II, T. Provder (Ed.), Ch. 16. American Chemical Society, Washington, DC (1991). 31. S. H. O’Connor, S. H. Gehrke and G. S. Retzinger, Langmuir 15, 2580 (1998). 32. G. Blume and G. Cevc, Biochim. Biophys. Acta 157, 1146 (1993). 33. P. K. Weathersby, T. A. Horbett and A. S. Hoffman, J. Bioeng. 1, 395 (1977). 34. M. B. Baleux, C. R. Acad. Sci. Paris 274, 1617 (1972). 35. K. D. Caldwell, J. M. Li, J.-T. Li and D. G. Dalgleish, J. Chromatogr. 63, 604 (1992). 36. J. H. Freed, in: Spin Labeling. Theory and Applications, L. J. Berliner (Ed.), p. 53. Academic Press, NY (1976). 37. M. A. Cohen-Stuart, J. M. H. M. Scheutjens and G. J. Fleer, J. Polym. Sci.: Polym. Phys. Edn 18, 559 (1980). 38. V. Hlady, J. Lyklema and G. J. Fleer, J. Colloid Interface Sci. 87 (2), 395 (1982). 39. M. Bohner, T. A. Ring and K. D. Caldwell, Macromolecules (submitted). 40. P. G. de Gennes, Adv. Coll. Interf. Sci. 27, 189 (1987). 41. J.-T. Li, Plasma protein interactions with copolymer stabilized colloids, PhD dissertation, University of Utah (1993). 42. K. Holmberg, F. Tiberg, M. Malmsten and C. Brink, Colloids and Surfaces A 297, 123– 124 (1997). 43. S. Svedhem, Biomolecular interactions at model interfaces: a self-assembled monolayer approach, PhD dissertation, Linköping University (2001). 44. T. Blunk, M. Lück, J. E. Diederichs, B. W. Müller and R. H. Müller, Proceed. Intern. Symp. Control. Rel. Bioact. Mater. 47, 22 (1995). 45. S. Wu, in: Polymer Interface and Adhesion, Table 5.5. Marcel Dekker, New York (1982). 46. J. Israelachvili, in: Intermolecular and Surface Forces, Table XVII. Academic Press, London (1985). 47. R. L. C. Wang, H. J. Kreuzer and M. Grunze, J. Phys. Chem. B. 101, 9767 (1997). 48. A. Afif, H. Hommel, A. P. Legrand, M. Bacquet, E. Gailliez-Degremont and M. Morcellet, J. Colloid Interface Sci. 211 (2), 304 (1999).
Effects of base material, plasma proteins and FGF2 on endothelial cell adhesion and growth P. ANNE UNDERWOOD ∗ , JOHN M. WHITELOCK, PENNY A. BEAN and JOHN G. STEELE CSIRO Molecular Science, Sydney, NSW, Australia Received 8 January 2002; accepted 22 May 2002 Abstract—Blood-contacting materials rapidly acquire a coating of plasma proteins which can lead to local platelet activation and thrombus formation. This phenomenon seriously limits the usefulness of small diameter synthetic vascular grafts. One solution to this problem is to pre-seed or encourage in situ colonisation of the material with endothelial cells to maintain a non-thrombogenic surface. We have investigated the effect of contact with plasma and serum on the subsequent ability of human endothelial cells to adhere to model hydrophobic and hydrophylic plastic surfaces, and the effect of surface bound fibroblast growth factor 2 (FGF2) on endothelial cell proliferation. Cell adhesion was mainly dependent on adsorbed fibrinogen or vitronectin, depending on the polymer surface, and correlated with antibody binding to these molecules rather than quantitative surface concentrations. Cell proliferation was directly correlated with surface bound FGF2. Surface binding of the latter was controlled both by the chemical nature of the polymer surface and by the presence of FGFbinding molecules adsorbed on the surface. FGF2 bound specifically to surface-adsorbed fibrinogen, fibronectin and vitronectin as well as to pre-coated heparan sulphate proteoglycan, perlecan. Binding was significantly inhibited by plasma and serum which contained high levels of FGF2 binding proteins. To be effective in supporting endothelialisation of vascular grafts in vivo, surface-bound FGF2 would need to be protected from surface dissociation into the circulating blood. Key words: Protein adsorption; fibronectin; fibrinogen; vitronectin; growth factor.
INTRODUCTION
Blood-contacting materials rapidly acquire a coating of plasma proteins. The surface properties of such materials effect selective adsorption of protein species which may be unrelated to their representation in the bulk phase of plasma [1, 2]. These adsorbed proteins control the subsequent biological activity of the material surface. In the case of synthetic vascular grafts commonly manufactured from ∗ To
whom correspondence should be addressed. E-mail:
[email protected]
78
P. A. Underwood et al.
poly(ethylene terephthalate) (PET, Dacron) or expanded polytetrafluoroethylene (ePTFE), the result is often platelet adhesion and activation and initiation of blood clotting cascades. This can lead to occlusion of the conduit or embolisation as local thrombi are dislodged from the surface. In the longer term, vascular hyperplasia can occur, which tends to progress from the anastomoses and can also result in occlusion. These responses limit the minimum functional size of the internal diameter of synthetic vascular grafts and hence the anatomical locations where they can be used. One potential solution to this problem is to generate a non-thrombogenic surface by seeding synthetic grafts with endothelial cells [3]. Seeding cells from the patient at the time of sugery has not been successful to date, due to poor retention of cells on the surface and incomplete endothelial coverage [4]. Clinical success has been achieved by harvesting endothelial cells from the patient one month before surgery and growing them in vitro in the graft to obtain a stable confluent endothelium before implantation [5]. Clearly this approach cannot be used for emergency procedures. A number of groups are investigating ways to improve the outcome of endothelial cell seeding at the time of surgery and/ or to encourage transmural and anastomotic migration and colonisation of endothelial cells in situ. These have involved pre-coating synthetic surfaces with fibronectin to increase cell adhesion, and linking heparin to the surface and binding vascular growth factors such as fibroblast growth factor 2 (FGF2) or vascular endothelial growth factor (VEGF) [6, 7]. Such methods encourage endothelial proliferation in vitro but have yet to be proven successful in animal models [8]. Some success in animal models has been achieved by incorporating heparin with FGF1, FGF2 or VEGF in fibrin or gelatin gel coatings of grafts [4, 9, 10]. In these studies it is unclear which surface molecular interactions drive endothelial cell adhesion, and which factors govern maintenance or loss of growth factors from the surface — both essential elements of endothelial colonisation. In previous studies we have investigated the role of proteins surface-adsorbed from serum upon subsequent endothelial cell adhesion [11]. In the present study we have compared the effects of exposure of hydrophobic polystyrene (PS) and hydrophilic tissue culture polystyrene (TCPS) surfaces to serum and plasma, upon human arterial endothelial cell adhesion. These materials have similar water contact angles to those described for ePTFE and PET respectively [12– 14]. We have also determined the effect of pre-coating the surface with a sub-endothelial heparan sulphate proteoglycan, perlecan, upon FGF2 binding and endothelial proliferation.
EXPERIMENTAL
Materials Bovine fibronectin (used to coat surfaces for routine cell culture), was purified from fresh bovine plasma by affinity chromatography on gelatin Sepharose
Endothelial response to surface proteins
79
(AMRAD Pharmacia Biotech, Melbourne, Victoria, Australia) as described [15]. Human vitronectin was purified from fresh-frozen plasma by affinity chromatography using monoclonal antibody HV2 as described [16, 17]. Human perlecan was purified from culture medium conditioned by human endothelial cells as previously described [18]. Purified human fibronectin, fibrinogen, serum albumin and immunoglobulin G were obtained from Sigma Aldrich (Castle Hill, NSW, Australia). Reconstituted proteins were stored in sterile aliquots at −70◦ C. Recombinant Fibroblast Growth Factor 2 (FGF2) was obtained from R&D Systems, Minneapolis, USA and 125I labelled FGF2 was from Amersham, Castle Hill, NSW, Australia. Foetal bovine serum (FBS), low endotoxin, was from Commonwealth Serum Laboratories, Melbourne, Australia. FBS was sequentially depleted of vitronectin and fibronectin using an anti vitronectin monoclonal antibody affinity column and gelatin Sepharose as previously described [19]. Human Serum was prepared from samples of whole blood obtained from polycythaemic patients at Royal North Shore Hospital, Sydney, eight samples being pooled together and filter sterilised. A pool of human plasma was prepared from three samples of sterile freshfrozen plasma obtained from the Red Cross Blood Bank, Sydney. Aliquots of serum and plasma were stored at −20◦ C. Protein A Sepharose CL-4B was obtained from AMRAD Pharmacia Biotech, Melbourne, Victoria, Australia. NHS biotin was from Pierce, Rockford, IL, USA and Na 125 I (100 mCi/ml) was from Australian Isotopes, Sydney. The tetrazolium compound MTS [3-(4,5-dimethylthiazol-2yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] was obtained from Promega (Madison, Wisconsin, USA). The electron transfer reagent PMS (phenazine methosulphate) was obtained from Sigma-Aldrich. Crystal violet (CV, C.I. 42555) was obtained from British Drug Houses (Poole, England). Monoclonal antibodies A22, specific for bovine fibronectin and HV2, specific for human vitronectin have been previously described [16, 19]. Monoclonal antibody PE 24/ 19, specific for human fibrinogen was prepared from mice immunised with platelet extract as previously described [20]. Monoclonal antibody 3E1, specific for human fibronectin was obtained from Gibco BRL Life Technologies (Rockville, MD, USA). None of these antibodies interfered with cells adhesion to their respective ligands. Rabbit anti mouse immunoglobulins (RAM) conjugated to horseradish peroxidase (HRP) was obtained from Dako, (Carpinteria CA, USA) and Streptavidin conjugated to horseradish peroxidase from Amersham. Peroxidase substrates ABTS (2,2 -azino-bis(3-ethylbenz-thiazoline-6-sulfonic acid) and C1N (4-chloro-1-napthol) were from Sigma Aldrich. 96 well tissue culture polystyrene (TCPS) plates (167008) and flasks were from Nunc, Roskilde, Denmark. 96 well polystyrene (PS) microtitration plates (Linbro/ Titertek 76-232-05) were from ICN Biomedicals (Australasia) and 96 well polyvinyl chloride (PVC) U-shaped microtiter plates (2401) were from Dynex Technologies (Chantilly, VA, USA). All standard laboratory chemicals were analytical grade.
80
P. A. Underwood et al.
Labeling of proteins Human fibrinogen, fibronectin and vitronectin were labelled with 125I using the chloramine T method as previously described [21]. Final concentrations of labelled proteins were 10 µg/ ml with activities of 20 000 cpm / µl (fibronectin), 25 000 cpm / µl (fibrinogen) and 45 000 cpm / µl (vitronectin). Proportions of the label as surface-bindable protein were calculated on polyvinyl plates each time the labelled proteins were used, as previously described [21], and ranged from 0.53 to 0.32 (fibronectin), 0.50 to 0.39 (fibrinogen) and 0.45 to 0.28 (vitronectin) over the experimental period. These values were used to correct estimates of surfaceadsorbed protein. Iodine labelled proteins were used within 3 weeks. The same three proteins were labelled with biotin as previously described [21], using a 50 molar excess of NHS biotin for 10 minutes. Aliquots of labelled proteins were stored at −70◦ C. Maximal binding of all labelled proteins from dilute solutions exceeded 95%. Integrity of labelled proteins and binding curves from dilute solutions Doubling dilutions of unlabelled fibronectin, fibrinogen and vitronectin were prepared in phosphate buffered saline (PBS, Dulbecco’s A, pH 7.1), spiked with 125 I or biotin labelled proteins. Two series of iodinated proteins were prepared with 10 000 cpm or 50 000 cpm for each per 50 µl of diluted protein. Biotin labelled proteins were used at 1 µg/ ml (fibronectin), 0.75 µg/ ml (fibrinogen) or 0.4 µg/ ml (vitronectin). Adsorption of the labelled proteins to PVC plates at 50 µl per well, was determined after a 3 day incubation at 4◦ C as previously described [21]. Adsorption curves were plotted for both iodine and biotin spiked protein dilutions. These were used to test the integrity of binding from the different concentrations of iodine label and to compare the binding curves using iodine vs biotin labels. Cell culture Primary cultures of human umbilical arterial endothelial cells (HUAEC) were prepared from fresh umbilical cords delivered by Caesarian section at Royal North Shore Hospital, Sydney, as described [22] using 0.1% collagenase (Sigma C-6885). Cells were routinely grown in 80 cm2 TCPS flasks, precoated for two hours at 37◦ C with 5 ml bovine fibronectin at 10 µg/ ml in PBS. The culture medium was Medium 199 with Earle’s Salts (Gibco BRL, Life Technologies) containing 20% FBS, 100 units/ ml penicillin, 100 µg/ ml streptomycin sulphate (ICN Biomedicals), 100 µg/ ml heparin (H-3149, Sigma) and 2% bovine brain extract (prepared according to [23]). Cells were passaged at a 1 : 5 split ratio after disaggregation with 0.125% trypsin, 0.02% EDTA (T /EDTA), and used between passages 6 and 10.
Endothelial response to surface proteins
81
Calculation of critical surface concentrations of adhesive proteins for HUAEC adhesion Three dilution series were set up for each adhesive protein. The first was of biotinylated proteins alone in PVC wells starting at 1 µg/ ml (fibronectin), 0.75 µg/ ml (fibrinogen), 0.4 µg/ ml (vitronectin), incubated for 3 days at 4◦ C. Standard adsorption curves based on >95% protein binding were constructed from these. The second series was of unlabelled proteins spiked with a single concentration of biotinylated proteins, as above. These were incubated in PS or TCPS wells for 2 h at 37◦ C followed by 24 h at 4◦ C, and used together with the standard curves to estimate concentrations of adsorbed proteins. The third series was the same as the second, replacing the biotin label with unlabeled protein. Following coating and blocking with BSA, these plates were used to estimate cell adhesion as follows. Disaggregated HUEAC were resuspended in medium 199 containing 10% serum depleted of fibronectin and vitronectin, and added to wells at 3 × 104 cells per well. Plates were incubated for 24 h in the culture incubator. Adhered cells were stained with crystal violet, and absorbance of solubilized dye measured at 595 nm, as described [24]. We have previously determined that no HUAEC proliferation occurs in this period [25]. Cell adhesion was estimated as percent of the maximum reading for each adsorbed protein. Cell adhesion was plotted against surface concentration of adsorbed protein and the surface concentration of each protein required to support 70% of maximal HUAEC adhesion was calculated.
Estimation of concentration of cell adhesion proteins in human plasma and serum The concentration range of fibrinogen in human plasma has been reported as 2.5– 5 mg/ ml (Ciba Geigy Scientific Tables 1971), while that of vitronectin and fibronectin have been estimated at 200– 400 µg/ ml [26, 27]. The concentration of fibronectin in serum has been reported to be lower than that in plasma due to cold precipitation and incorporation into the fibrin clot [28]. The concentration of these proteins in our pooled plasma and serum was estimated as follows. Twofold dilution series (starting at 10%) of plasma and serum were made in PBS and duplicate 2 µl volumes spotted onto Immobilon P transfer membranes (Millipore, Bedford, MA, USA). An artificial 10% ‘plasma’ solution was made by dissolving 300 µg/ ml fibrinogen, 4 mg/ml albumin, 1 mg/ ml immunoglobulin G, 30 µg/ ml vitronectin and 30 µg/ ml fibronectin in PBS. This was serially diluted and spotted in parallel. Adsorbed fibrinogen, fibronectin and vitronectin were detected using monoclonal antibodies PE 24/ 19 ascites 1/1000, 3E1 ascites 1/ 1000 and HV2 at 2 µg/ ml, RAM conjugated HRP and C1N as substrate. Concentrations of these proteins in the plasma and serum were estimated by comparison of intensity of spots.
82
P. A. Underwood et al.
Estimation of adsorption of fibrinogen, fibronectin and vitronectin from plasma and serum onto PS and TCPS using 125 I labelled proteins Wells of PS or TCPS plates were coated with 50 µl human fibronectin, perlecan (each at 10 µg/ ml in PBS) or PBS alone for 2 h at 37◦ C followed by 24 h at 4◦ C. After rinsing with PBS, 50 µl per well of undiluted plasma or serum, 10% plasma or serum in medium 199, or 1% BSA in PBS, each containing 100 ng/ ml FGF2, were added and further incubated for 2 h at 37◦ C. The plasma and serum samples contained added iodinated fibrinogen, fibronectin or vitronectin at the rate of 20 000 to 100 000 cpm per well. Following incubation and 4× PBS washes, individual wells were cut out of the plates with a hot wire and surface concentrations of these proteins were calculated from the cpm bound per well, corrected for bindability of label, as described above. Relative estimation of adsorption of fibrinogen, fibronectin and vitronectin from plasma and serum onto PS and TCPS using monoclonal antibodies Wells of PS and TCPS plates were coated, and treated with plasma, serum or BSA containing FGF2 as in the previous section, omitting the 125 I labels. Following the plasma/ serum treatments, wells were rinsed with PBS and further blocked with 200 µl 1% BSA in PBS. Adsorbed proteins were then detected by ELISA using antibodies PE 24/ 19 at 4 µg/ ml, 1E3 at 1/ 2000, or HV2 at 4 µg/ ml followed by RAM conjugated HRP 1/1000 and ABTS. Controls included (1) normal mouse serum at 1/ 2000 and (2) omitted primary antibody. After correction for controls, absorbances were expressed as a proportion of the maximum absorbance observed for each antibody. We determined that, for each of the antibodies used, there was no significant difference in their abilities to bind to protein adsorbed at similar surface concentrations on PS compared to TCPS (i.e. they bound to epitopes equally accessible on either surface). Therefore relative quantitative comparisons between the two surfaces are valid. Effect of added plasma on monoclonal antibody detection of coated protein PS wells were coated with 50 µl fibrinogen, fibronectin or vitronectin at 10 µg/ ml in PBS for 2 h at 37◦ C followed by 24 h at 4◦ C. They were then treated with 50 µl 10% plasma or 1% BSA for 2 h at 37◦ C, followed by an additional block of 200 µl of 1% BSA. The ability of the specific antibodies to detect the coated protein was then estimated by ELISA as above. Adhesion and proliferation of HUAEC on treated surfaces Wells of PS and TCPS plates were coated and treated with plasma, serum or BSA containing FGF2 as in the previous section. Following the plasma and serum treatments, wells were washed ×2 with medium 199. HUAEC were added at 3 × 103 per well in medium containing 10% double-depleted serum, but no heparin
Endothelial response to surface proteins
83
or FGF additive. Thus the only available FGF was that adsorbed to the surface of the wells during the plasma/ serum treatments. Plates were incubated in the culture incubator for 6 days with replacement of medium at 3 days. At 24 h, 3 days and 6 days replicate wells were used to estimate adhered cell number, by both conversion of MTS, followed by crystal violet staining as previously described [25]. Cell adhesion at 24 h was expressed as a percentage of the maximal adhesion observed on fibronectin coated wells. Proliferation after 3 and 6 days was expressed as percent increase over the measurement at 24 h for each treatment. Estimation of FGF2 binding to treated surfaces Wells of PS and TCPS plates were coated, and treated with plasma, serum and BSA containing FGF2, as for cell adhesion and proliferation. 125 I labelled FGF2 was added with the unlabeled FGF2 at the rate of 50 000– 75 000 cpm per well. Following incubation at 37◦ C for 2 h and 4 PBS washes, wells were cut out of the plate and bound 125I-FGF2 measured. Estimation of FGF2 binding to coated cell adhesion molecules Wells of PS and TCPS plates were coated with 50 µl of human fibrinogen, fibronectin, vitronectin, perlecan, serum albumin or immunoglobulin G at 10 µg/ ml in PBS for 2 days at 4◦ C. After washing with PBS and blocking with 200 µl per well 1% BSA for 1 h, 45 000– 60 000 cpm of iodinated FGF2 were added per well in 50 µl 1% BSA and incubated at room temperature for 2 h. Additional wells coated with fibrinogen, fibronectin, vitronectin or perlecan were treated in parallel with iodinated FGF2 in 1% BSA containing 10 µg/ ml fibrinogen, fibronectin, vitronectin or perlecan respectively as competitors. Surface bound iodinated FGF2 was estimated as above.
RESULTS
Characteristics of surface adsorption of cell adhesion proteins Saturation binding curves of fibronectin, vitronectin and fibrinogen on PVC were obtained on PVC wells with proteins labelled with either 125 I or biotin. Furthermore, similar saturation binding profiles were obtained with each protein at either concentration of 125 I label, indicating that the iodine labelling process did not change the protein adsorption characteristics. Binding saturated at approximately 20 µg/ ml (1587 ng/ cm2 ) for both fibronectin and fibrinogen, and at 5 µg/ ml (397 ng/ cm2 ) for vitronectin. The biological activity of these proteins adsorbed from dilute solutions to PS and TCPS, as measured by the adhesion of HUAEC, is shown in Table 1. The ratios in Table 1 give a relative measure of the efficiency of each protein adsorbed to the two surfaces to support cell adhesion. In each case the TCPS surface was more efficient
84
P. A. Underwood et al.
than the PS surface. Fibrinogen was less effective on TCPS compared to the two other proteins. The surface adsorption to PS and TCPS of the three cell adhesion proteins from plasma and serum are shown in Table 2, derived using 125I labelled molecules. Adsorption of these proteins from undiluted plasma (and to a lesser extent from serum) often resulted in surface concentrations exceeding the saturation concentration obTable 1. Biological activity of adsorbed adhesive proteins expressed as Surface Concentration required for 70% cell adhesion Surface
Fibrinogen
Fibronectin
Vitronectin
PS TCPS Ratio PS/TCPS
39.7 (6.86) 23.8 (2.86) 1.67
31.7 (2.71) 9.52 (1.47) 3.33
41.7 (4.48) 11.1 (1.22) 3.76
HUAEC were adhered for 24 h at 37◦ C. Entries are mean ng/ cm2 from 2 experiments. Values in brackets are SEM calculated by Analysis of Variance of each protein in the 2 experiments, 3 replicates per experiment. Table 2. Adsorption of cell adhesive proteins from plasma or serum onto PS and TCPS, in ng/ cm2 Protein source
Fibrinogen PS
TCPS
Fibronectin PS TCPS
Vitronectin PS
TCPS
10079 (816) na 612 (21.4) na
550 (92.4) 232 (68.0) 132 (35.7) 11.9 (0.7)
1905 (215) 73 (6.0) 145 (9.4) 24.6 (2.9)
450 (106.4) 651 (131.9) 53.2 (4.9) 38.1 (5.9)
815 (26.9) 713 (54.9) 104 (2.3) 151 (11.9)
(B) Precoated with Fibronectina plasma 7936 (2746) 8175 (499) serum na na 10% plasma 485 (18.4) 659 (50.1) 10% serum na na 1% BSA na na
1397 (330) 773 (218) 379 (37.1) 257 (33.1) 222b (3.3)
2302 (82.9) 336 (82.4) 374 (23.9) 264 (37.2) 206b (10.3)
733 (183.3) 582 (156.1) 54.8 (4.5) 56.3 (15.9) na
873 (27.9) 502 (19.1) 100 (7.3) 118 (3.5) na
(C) Precoated with Perlecan 10% plasma 646 (103) 757 (84.8)
131 (3.9)
132 (9.2)
70.6 (6.2)
111 (10.5)
(A) Precoated with PBS plasma 6357 (2132) serum na 10% plasma 677 (37.9) 10% serum na
Precoating was done for 2 h at 37◦ C followed by overnigt at 4◦ C. Protein solutions in column 1 were added after precoating, and incubated for 2 h at 37◦ C. a Amount of fibronectin on the surface includes that precoated as well as adsorbed from protein solutions. b Amount of FN on the surface due to precoat alone. Entries are the means of 2 experiments, with 3 replicates per experiment. Entries in brackets denote SEM calculated by Analysis of Variance of each protein/ treatment in the 2 experiments. Entries in bold denote surface concentration of particular protein is greater than the maximum estimated at surface saturation for PVC plates (see text).
Endothelial response to surface proteins
85
served with individual proteins on PVC wells, suggesting multi-layering and/ or protein interactions. Pre-coating the surfaces with fibronectin or perlecan (PN) at ∼12.5% saturation made no significant difference to the subsequent adsorption of proteins from plasma or serum (paired T tests, p > 0.05). Cell adhesion to plasma and serum treated surfaces Adhesion of HUAEC to surfaces treated with plasma and serum, using the same conditions as for Table 2, is shown in Fig. 1. The numbers above each bar indicate the potential adhesiveness of the treated surface. These were derived by taking the surface concentration of a protein from Table 2 and dividing it by the appropriate entry in Table 1 to obtain an index of adhesiveness. This was summated for all three proteins for each surface/ treatment combination. Thus for 10% plasma on PS, 677/39.7 + 132/31.7 + 53.2/41.7 = 22.5, entered as 23 over the 10% plasma bar in Fig. 1A. This equation assumes that the cell adhesiveness of a protein mixture equals the sum of its individual parts. This assumption is invalidated by the data in Fig. 1A. It is clear that the degree of cell adhesion does not correlate with the amount of potentially cell adhesive protein on the surfaces treated with whole plasma or serum, compared with dilute solutions. Thus the potential adhesion index values of PS treated with 100% serum, PS treated with 10% plasma and TCPS treated with 10% serum were similar (23, 23 and 16 respectively), yet cell adhesion was not above background in the first condition but close to maximal in the second and third. Furthermore the potential adhesiveness score of PS treated with 100% plasma was 188, yet cell adhesion only reached 50% of the maximum. In contrast to the differential effects of the plasma and serum treatments of uncoated surfaces on subsequent cell adhesion, pre-coating with fibronectin resulted in similar levels
Figure 1. Cell adhesion to PS and TCPS surfaces treated with plasma or serum. Bars are percent adhesion at 24 h after seeding, taking the highest absorbance value (TCPS coated with FN followed by 10% plasma) as 100%. Means ± SEM of 3 experiments. Numbers above the bars represent the potential adhesion index of the surface (see text). Abbreviations: P, plasma; S, serum; FN, fibronectin; PN, perlecan.
86
P. A. Underwood et al.
of cell adhesion to all subsequently treated surfaces (Fig. 1B, p > 0.05, Analysis of Variance). The basis of this lack of correlation observed with complex mixtures, between cell adhesive protein density and HUAEC attachment was further explored. If multilayering of proteins was occurring in response to exposure of surfaces to undiluted proteins, it was possible that the profile of cell-adhesive proteins contactable by cell surface receptors was not representative of the content of adhesive proteins in the multi-layer. Conversely, during the incubation period of the cell adhesion assay, a change in the surface profile of adsorbed proteins could have occurred (in contrast to the rapid saline washes which preceded estimation of adsorbed proteins by bound 125 I label). To test for these possibilities, we used monoclonal antibodies to give relative quantitative measures of the amount of surface proteins which would be accessible to cell surface receptors (IgG and integrins are of similar molecular size). These results are shown in Fig. 2 and each quantity measured (cell adhesion or antibody binding) has been expressed relative to the maximum observed for that quantity across the treatments. Figures 2A (for PS) and 2B (for TCPS) show that for each combination of surface and protein solution, there is a close correlation between cell adhesion and the specific antibody-binding of at least one of the adhesive proteins. The protein for which the correlation was observed, however, differed between different treatments. On TCPS it was anti-vitronectin antibody binding that correlated with cell adhesion for the 10%, 100% serum and 100% plasma treatments, whereas on PS it was anti-fibrinogen antibody binding that correlated with cell adhesion for 100% plasma treatment. Of particular note in these assays was the strong binding of fibronectin and fibrinogen antibodies on TCPS treated with 10% plasma, but not with undiluted plasma, despite the initially much
Figure 2. Cell adhesion to PS and TCPS surfaces treated with plasma or serum, compared with antibody detection of cell adhesion molecules on the surface. Bars are means ± SEM of 3 experiments. Cell adhesion expressed as proportion of maximum observed (as for Fig. 1). Optical densities for individual antibodies binding to molecules on TCPS surfaces were multiplied by PS /TCPS ratios in Table 1 and then scaled as proportions of the maximum value for each antibody. Abbreviations: P, plasma; S, serum; FN, fibronectin.
Endothelial response to surface proteins
87
higher surface concentrations of these proteins with the latter treatment. This was also observed with detection of fibronectin on PS, but not with fibrinogen. These results were consistently reproducible. Clearly fibrinogen played an important role in promoting cell adhesion to plasma treated PS, whereas serum treated PS lacked this molecule and cell adhesion was poor. These experiments demonstrated that antibody binding was a more reliable indicator of potential cell adhesiveness of a protein-treated surface, than is estimation of the total surface concentration of adhesive proteins, even correcting for the differences in adhesive potency of proteins on the different surfaces. On the surfaces pre-coated with fibronectin, similar levels of cell adhesion were detected independently of subsequent serum or plasma treatment (Fig. 1B). When probed with the anti-fibronectin antibody, maximal absorbance (comparable to that seen with coated fibronectin followed by 1% BSA (Fig. 2A)) was observed in all treatments (not shown). Proliferation of cells adhered to plasma or serum treated surfaces Subsequent to cell adhesion, the proliferation of HUAEC is highly dependent on the presence of vascular growth factors such as FGF2 or VEGF. We tested the ability of HUAEC to proliferate on surfaces treated with plasma or serum, to which FGF2 was added. FGF2 was not added to the HUAEC growth medium so that cell proliferation would be solely dependent on FGF2 bound to the surface from the plasma/ serum mixtures. Thus we expected to observe proliferation only on the surfaces pre-coated with perlecan — an FGF-binding molecule with similar affinity for FGF2 as heparin. The results of estimates of proliferation at three and six days post cell seeding are shown in Fig. 3. On perlecan pre-coated surfaces, proliferation to day 3 and maintenance to day 6 was observed with FGF2 added in 10% plasma on both polymer surfaces, with TCPS displaying a significant advantage.
Figure 3. Cell proliferation after 3 or 6 days on uncoated and protein-precoated PS and TCPS surfaces treated with plasma or serum. Data are increase in cell number expressed as percentage of that at 24 h after cell seeding. Means ± SEM of 3 experiments. Abbreviations: P, plasma; S, serum; FN, fibronectin; PN, perlecan. Note only those treatments which resulted in significant cell adhesion (see Fig. 2), could be tested for proliferation.
88
P. A. Underwood et al.
Figure 4. Binding of FGF2 to PS and TCPS surfaces treated with plasma or serum, estimated using 125 I labelled FGF2. Data are means ± SEM of 4 experiments. Abbreviations: P, plasma; S, serum.
Unexpectedly, some surfaces which had not been pre-coated with perlecan did support cell proliferation but this was dependent upon both the polymer and the composition of the protein treatment. On surfaces which had received no pre-coat (Fig. 3A), significant proliferation to day 3 with maintenance to day 6, was observed on TCPS surfaces treated with FGF2 in undiluted or 10% plasma (Fig. 3A), but was not seen on TCPS treated with FGF2 in serum (10% or 100%), or on PS treated with FGF2 in plasma. On pre-coated surfaces (Fig. 3B), TCPS precoated with either fibronectin or perlecan supported proliferation to day 3 with maintenance to day 6 whatever protein mixture had been used to deliver the FGF2. Pre-coating with perlecan as opposed to fibronectin conferred no particular growth advantage on this surface. On pre-coated PS surfaces, however, pre-coating with perlecan conferred a distinct growth advantage over pre-coating with fibronectin, unless the latter was followed by FGF2 added in 1% BSA, as opposed to plasma or serum (Fig. 3B). To determine the factors controlling these unpredicted proliferation results, we measured the surface binding of FGF2 to the variously treated surfaces using 125 I labelled FGF2. These results are shown in Fig. 4. For all treatments FGF2 binding was significantly higher to TCPS than to PS, the differences ranging from twofold (perlecan pre-coat, FGF2 added in 1% BSA) to thirty eight fold higher on TCPS (no pre-coat, FGF2 added in 10% serum). On PS the highest FGF2 binding was observed with pre-coated perlecan and FGF2 added in 1% BSA (Fig. 4A). FGF2 added in 10% plasma showed reduced binding. FGF2 added in 1% BSA showed increased binding to pre-coated fibronectin compared with background binding to untreated PS. This was also reduced in the presence of plasma and further reduced in the presence of serum. Undiluted plasma and serum exhibited greater inhibition of FGF2 binding than dilute solutions. These differential effects were statistically significant (p < 0.05, Analysis of Variance and Student Newman Keul’s test). On TCPS, FGF2 bound equally well from 1% BSA or 10% plasma to pre-coated perlecan, pre-coated fibronectin or uncoated surfaces (p > 0.05, Analysis of Variance and Student Newman Keul’s test). Binding to pre-coated fibronectin or
Endothelial response to surface proteins
89
Figure 5. Correlation of cell proliferation to 3 days with concentration of FGF2 bound on the surface, combining data from Figs 3 and 4. Circles: treatments where cell numbers decreased from day 3 to day 6. Triangles: treatments where cell number did not significantly change between days 3 and 6. Squares: treatments where cell numbers increased from day 3 to day 6. *Treatments where cell adhesion was due solely to adsorbed vitronectin (TCPS + serum or 10% serum, cf. Fig. 2).
uncoated TCPS was reduced by undiluted plasma, serum or 10% serum (p < 0.05, Analysis of Variance and Student Newman Keul’s test). The greatest reduction in FGF2 binding was seen in the presence of undiluted serum. When HUAEC proliferation to day 3 was plotted against surface-bound FGF2 (Fig. 5) a clear correlation was observed. Not only was proliferation dependent on the amount of surface-bound FGF2, but increased proliferation to day 6 only occurred at the highest FGF2 surface concentrations. Two exceptions to this relationship were clearly seen (marked by* in Fig. 5). These were two conditions where HUAEC adhesion was solely dependent on surface-bound vitonectin (serum and 10% serum treatments of TCPS). We have shown in previous studies that HUAEC grow poorly on vitronectin for reasons unrelated to availability of growth factors [29]. These results demonstrated that the underlying polymer chemistry was exerting a profound effect on surface binding of FGF2 and also suggested that components of plasma and serum were capable of binding FGF2. When surface-adsorbed, these components would be likely to increase FGF2 binding; when in solution they would exert competitive inhibition. To explore this more fully we investigated the binding of FGF2 to individual molecules which are present at relatively high concentrations in plasma and/ or serum. These results are shown in Fig. 6. On PS, FGF2 bound to the uncoated surface, pre-coated human serum albumin (HSA) or immunoglobulin G (IgG) at a relatively low rate (Fig. 6A). Pre-coated fibrinogen, fibronectin or vitronectin bound twofold higher amounts. In the case of fibrinogen and fibronectin, FGF2 binding was reduced by about 50% by the presence of these molecules in solution at 10 µg/ ml (p < 0.05, paired T tests). In contrast soluble vitronectin was
90
P. A. Underwood et al.
Figure 6. Binding of FGF2 to individual proteins coated on PS and TCPS. Data are means ± SEM of triplicate samples from a representative experiment. Abbreviations as for previous figures or given in the text.
without inhibitory activity. FGF2 binding to perlecan was significantly higher than to any of the coated plasma molecules and was totally prevented by the presence of soluble perlecan (p < 0.05, Analysis of Variance and Student Newman Keul’s test). On TCPS, binding of FGF2 to uncoated or coated surfaces was uniformly high (Fig. 6B), with perlecan coating conferring only a small (but significant, p < 0.05, Analysis of Variance and Student Newman Keul’s test) advantage over the uncoated surface. As on the PS surface, soluble fibrinogen or fibronectin reduced FGF2 binding to these coatings and soluble vitronectin was without effect. Perlecan competition was not tested on this surface.
DISCUSSION
Surface adsorption of cell adhesive proteins Proteins adsorbing to surfaces from pure solutions often exhibit saturation binding governed by Langmuir or Freundlich isotherms [2]. We observed saturation binding of fibrinogen, fibronectin and vitronectin from solutions of purified proteins to PVC. Binding saturated at surface concentrations compatible with monolayer coverage (side on or end on) values calculated by others [30– 33]. As PVC shows higher surface adsorption of plasma proteins compared to polystyrene, polyethylene or TCPS surfaces under similar conditions [31, 33, 34], we assumed that saturation binding levels on PVC represented a generous upper limit for monolayer coverage of plasma proteins on PS or TCPS. When iodine labels were used to measure surface adsorption of cell adhesive proteins from undiluted plasma or serum onto PS or TCPS, surface concentrations in gross excess of either monolayer coverage or previously reported values [2] were observed. Horbett [2] has suggested that surface concentrations of adsorbed plasma proteins in excess of 500 ng/ cm2 are likely to be due to the presence of small fibrin clots. While this may explain the high values
Endothelial response to surface proteins
91
observed with plasma, protein: protein interaction and multilayering cannot be ruled out — particularly with the high surface concentrations of vitronectin adsorbed from undiluted serum (no fibrinogen could be detected in our serum samples). Endothelial cell adhesion We have previously described the superior efficiency of venous endothelial cell adhesion on vitronectin or fibronectin adsorbed to TCPS compared to the same surface concentrations of these molecules on PS [11]. With dilute solutions of serum there was good correlation between the amount of adsorbed protein and endothelial cell adhesion (correcting for conformational effects, expressed by the molecular potency of the particular molecule on the particular surface). The current study demonstrates the same phenomenon with arterial endothelial cell adhesion and extends the observations to adsorbed fibrinogen and dilute plasma. In contrast, however, we found that the correlation broke down when undiluted plasma or serum was used. With either plasma or serum, surface adsorbed concentrations of adhesive proteins were in vast excess of threshold levels needed for cell adhesion, yet cells adhered to some surfaces and not to others. Alternatively, in place of 125 I labels to measure total protein adsorbed (adjusted for molecular adhesion potency), we used binding of specific antibodies to detect adsorbed proteins (adjusted for molecular adhesion potency). Under these conditions we found good correlation between antibody binding and cell adhesion. On this basis, the contributions of different adsorbed proteins to cell adhesion could be detected in differing treatments and some unexpected conclusions emerge. Thus HUAEC adhesion to plasma treated PS was due to adsorbed fibrinogen (with vitronectin being inaccessible), whereas on TCPS it was due to adsorbed vitronectin (with fibrinogen being inaccessible). Antibody-detectable vitronectin was also correlated with the successful HUAEC adhesion to TCPS treated with either diluted (10% v/ v) or undiluted serum, and the failure of cell adhesion to similarly treated PS. Thus it appeared that accessibility of adsorbed adhesion proteins to cellular adhesion receptors and specific antibodies was similar. Treatment of either polymer surface with 10% plasma produced maximal cell adhesion and maximal binding of antibodies to all three adhesion molecules. Whether the lack of correlation between the biological events and the estimated total potentially adhesive protein on the surface was due to protein layering limiting accessibility to all molecules on the surface, or a dramatic loss of protein content due to exchange/ dissociation of weakly bound protein from the surface during the incubation steps subsequent to initial protein adsorption, is not known. The higher availability of fibronectin and fibrinogen on surfaces treated with 10% plasma compared with 100% plasma (see Fig. 2) may be due to differences in protein : protein interactions at different concentrations, or different dissociation rates from the surface. The correlation between antibody binding activity and HUAEC adhesion stands in stark contrast to the lack of correlation between measured bound protein and cell adhesion. Correlation of platelet adhesion
92
P. A. Underwood et al.
with surface-adsorbed fibrinogen measured by antibody binding rather than total fibrinogen on the surface, has been described for plasma treated surfaces using both antibodies specific for the platelet adhesion domain [35] and antibodies specific for several different fibrinogen domains [36]. These effects were thought to be largely conformational rather than reflecting epitope availability. Of additional interest was our finding that although pre-coating of PS or TCPS surfaces with fibronectin or perlecan did not significantly affect subsequent adsorption of proteins from plasma or serum, the coated proteins were still detectable to the same level by antibody binding. HUAEC adhesion to fibronectin (pre-coated overnight) was unaffected by any subsequent plasma or serum treatments. This suggests that the pre-coated proteins were firmly adsorbed and not appreciably dislodged by subsequent exposure to plasma or serum. This is in agreement with previous reports of increased binding strength with increased protein residence time [2]. Surface binding of FGF2 and endothelial proliferation Our results showed that endothelial cell proliferation was closely correlated with the amount of FGF2 bound to the surface. Similar in vitro correlation has been reported by others who used surface-bound heparin to localise FGF2 [6, 7, 37]. Our results additionally show that FGF2 binding to the surface is complicated by both the nature of the polymer surface itself and the local fluid phase protein composition. Thus we show that FGF2 bound avidly to the oxidised hydrophilic TCPS surface such that little advantage was gained by the presence of the specific FGF-binding molecule, perlecan. In contrast, perlecan conferred a significant advantage in FGF2 binding when coated on the hydrophobic PS or PVC surfaces. The presence of serum or plasma during FGF2 binding significantly reduced the amount of FGF2 bound to the surface, particularly on PS. This was due to the presence of FGF2-binding proteins in plasma and serum. The binding of FGF2 and VEGF by fibrinogen has been recently described [38, 39], but to our knowledge binding of FGF2 by fibronectin or vitronectin has not been previously noted. Fibrinogen and fibronectin in the solution phase each specifically inhibited FGF2 binding to the solid phase. Although binding of FGF2 to these plasma proteins was of significantly lower affinity than to perlecan, the high concentrations of these proteins in circulating plasma is likely to exert a significant dissociative effect on surface-bound FGF2. Alpha 2 macroglobulin, which has a plasma concentration approaching that of fibrinogen, has also been reported to bind (and inactivate) FGF2 [40]. Consequently surfaces which bind FGF2 in vitro and confer a growth advantage for endothelial cells, may not perform as expected when exposed to blood flow in vivo. Under these conditions vascular grafts which are coated with FGF2 presented on a binding molecule within a gel rather than on an exposed surface, may retain the growth factor for longer and display more effective endothelial colonisation in vivo [4, 37, 41]. Alteration of the surface chemistry of the polymer to increase FGF2 binding may also improve retention on the surface. Our findings suggest that it is important to test polymers which are to be endothelialised for vascular use, for their cell adhesive and growth
Endothelial response to surface proteins
93
factor retaining properties under in vitro conditions closely resembling those in vivo, before testing in animal models. Acknowledgements We are indebted to the staff of the Maternity unit of the Royal North Shore Hospital, Sydney for supply of umbilical cords for endothelial cell isolation and to the department of Haematology for the supply of blood for serum and plasma preparation. We also thank the Australian Red Cross Sydney for blood and plasma samples. We thank Meg Evans and Graham Johnson for critically reading the manuscript.
REFERENCES 1. J. L. Brash, in: Proteins and Interfaces: Physicochemical arid Biochemical Studies, J. L. Brash and T. A. Horbett (Eds), p. 490. American Chemical Society (1987). 2. T. A. Horbett, Cardiovasc. Pathol. 2, 137s (1993). 3. C. Gillis-Haegerstrand, S. Frebelius, A. Haegerstrand and J. Swedenborg, J. Vasc. Surg. 24, 226 (1996). 4. H. P. Greisler, J. Control. Release 39, 267 (1996). 5. J. Meinhart, M. Deutsch and P. Zilla, ASAIO J. 43, M515 (1997). 6. G. W. Bos, N. M. Scharenborg, A. A. Poot, G. H. M. Engbers, T. Beugeling, W. G. van Aken and T. Feijen, J. Biomed. Mater. Res. 44, 330 (1999). 7. M. Ishihara, Y. Saito, H. Yura, K. Ono, K. Ishikawa, H. Hattori, T. Akaike and A. Kurita, J. Biomed. Mater. Res. 50, 144 (2000). 8. A. Poot, M. Wissink, G. Engbers, W. van Aken, M. Visser, J. van Bockel, H. Koerten and J. Feijen, in: Trans. 6th World Biomat. Congress, p. 620 (2000). 9. D. A.Weatherford, J. E. Sackman, T. T. Reddick, M. B. Freeman, S. L. Stevens and M. H. Goldman, Surgery 120, 433 (1996). 10. S. Masuda, K. Doi, S. Satoh, T. Oka and T. Matsuda, ASAIO J. 43, m530 (1997). 11. J. G. Steele, B. A. Dalton, G. Johnson and P. A. Underwood, J. Biomed. Mater. Res. 27, 927 (1993). 12. D. K. Pettit, T. A. Horbett and A. S. Huffman, J. Biomed. Mater. Res. 26, 1259 (1992). 13. S. I. Ertel, B. D. Ratner, A. Kaul, M. B. Schway and T. A. Horbett, J. Biomed. Mater. Res. 28, 667 (1994). 14. J. G. Steele, G. Johnson, C. McFarland, B. A. Dalton, T. R. Gengenbach, R. C. Chatelier, P. A. Underwood and H. J. Griesser, J. Biomater. Sci. Polymer Edn 6, 511 (1994). 15. E. Ruoslhati, E. G. Hayman, M. Pierschbacher and E. Engvall, J. Methods in Enzymol. 82, 803 (1982). 16. C. A. Morris, P. A. Underwood, P. A. Bean, M. Sheehan and J. A. Charlesworth, J. Biol. Chem. 269, 23845 (1994). 17. P. A. Underwood, P. A. Bean, S. M. Mitchell and J. M. Whitelock, J. Immunol. Meth. 247, 217 (2001). 18. J. Whitelock, L. D. Graham, J. Melrose, A. D. Murdoch, R. V. Iozzo and P. A. Underwood, Matrix Biol. 18, 163 (1999). 19. P. A. Underwood, J. G. Steele, B. A. Dalton and F. A. Bennett, J. Immunol. Meth. 127, 91 (1990). 20. C. D. McFarland, M. Jenkins, H. J. Griesser, R. C. Chatelier, J. G. Steele and P. A. Underwood, J. Biomater. Sci. Polymer Edn 9, 1207 (1998).
94
P. A. Underwood et al.
21. P. A. Underwood and J. G. Steele, J. Immunol. Meth. 142, 83 (1991). 22. J. R. Weis, B. Sun and G. M. Rodgers, Thrombosis Res. 61, 171 (1991). 23. T. Maciag, J. Cerundolo, S. Ilsley, P. R. Kelly and R. Forand, Proc. Natl. Acad. Sci. 76, 5674 (1979). 24. W. Kueng, E. Silber and U. Eppenberger, Analyt. Biochem. 182, 16 (1989). 25. P. A. Underwood and P. A. Bean, In vitro Cell. Dev. Biol.-Animal. 34, 200 (1998). 26. M. C. Shaffer, T. P. Foley and D. W. Barnes, J. Lab. Clin. Med. 103, 783 (1984). 27. N. DiGirolamo, P. A. Underwood, P. J. McCluskey and D. Wakefield, Diabetes 42, 1606 (1993). 28. E. G. Hayman, M. D. Pierschbacher, S. Suzuki and E. Ruoslahti, Exp. Cell Res. 160, 245 (1985). 29. P. A. Underwood, P. A. Bean and L. Cubeddu, J. Cell. Biochem. 82, 98 (2001). 30. A. Schmitt, R. Varoqui, S. Uniyal, J. L. Brash and C. Pusineri, J. Colloid Interface Sci. 92, 25 (1983). 31. K. Park, D. F. Mosher and S. L. Cooper, J. Biomed. Mater. Res. 20, 589 (1986). 32. W. G. Pitt, S. H. Spiegelberg and S. L. Cooper, in: Proteins at Interfaces Physicochemical and Biochemical Studies, J. L. Brash and T. A. Horbett (Eds), p. 324. American Chemical Society (1987). 33. W. G. Pitt, D. J. Fabrizius-Homan, D. F. Mosher and S. L. Cooper, J. Colloid Interface Sci. 129, 231 (1989). 34. W. Breemhaar, E. Brinkman, D. J. Ellens, T. Beugeling and A. Bantjes, Biomaterials 5, 269 (1984). 35. W. B. Tsai, J. M. Grunkemeier and T. A. Horbett, J. Biomed. Mater. Res. 44, 130 (1999). 36. J. N. Lindon, G. McManama, L. Kushner, E. W. Merrill and E. W. Salzman, Blood 68, 355 (1986). 37. K. Doi and T. Matsuda, J. Biomed. Mater. Res. 34, 361 (1997). 38. A. Sahni, L. A. Sporn and C. W. Francis, J. Biol. Chem. 274, 14936 (1999). 39. A. Sahni and C. W. Francis, Blood 96, 3772 (2000). 40. P. A. Dennis, O. Saksela, P. Harpel and D. B. Rifkin, J. Biol. Chem. 264, 7210 (1989). 41. M. J. B. Wissink, R. Beernink, J. S. Pieper, A. A. Poot, G. H. M. Engbers, T. Beugeling, W. G. van Aken and J. Feijen, Biomaterials 22, 2291 (2001).
Acoustics of blood plasma on solid surfaces MARCUS ANDERSSON 1,∗ , ANDERS SELLBORN 1 , CAMILLA FANT 1 , CHRISTINA GRETZER 2 and HANS ELWING 1 1 Department
of Cell and Molecular Biology/ Interface biophysics, Lundberg Laboratory, Göteborg University, Box 462, SE-405 30 Göteborg, Sweden 2 Department of Biomaterials/ Institute of Surgical Sciences, The Sahlgrenska Academy, Göteborg University, Box 412, SE-405 30, Göteborg, Sweden Received 6 February 2002; accepted 7 May 2002 Abstract—We have quantified surface associated coagulation of human blood plasma with a recently developed methodological system consisting of a Quartz Crystal Microbalance with Dissipation monitoring (QCM-D), a method that measures the weight of adsorbed molecules on surfaces as a function of frequency shifts of a quartz crystal. Further, it measures the damping energy (i.e. viscoelasticity) of the adsorbed layer. Four different surfaces where studied: Heparin (Hep) surface as an active inhibitor of clot formation, titanium (Ti) surfaces that are known to activate the intrinsic pathway, polystyrene (PS) surfaces and poly(urethane urea) (PUUR) surfaces. The experiments were initiated by applying citrated human plasma at the sensor surfaces; calcium was then added to initiate coagulation. The Hep surfaces showed no apparent indication of clot formation during one hour of incubation at room temperature. However, on Ti surfaces we observed an early and rapid change in both frequency shift and viscoelastic properties of the coagulating plasma. We inhibited the intrinsic pathway activation by using corn trypsin inhibitor (CTI), which is specific for factor FXIIa in the bulk phase, which prolonged the coagulation times for all non-heparinized surfaces. We have also found a peculiar initial plasma protein interaction phenomenon on Ti surfaces. The described methodology would be very efficient for basic studies of surface associated coagulation and as a screening method for new biomaterials. Key words: Coagulation; surface activation; QCM-D; FXII inhibition; HMWK deposition.
INTRODUCTION
The composition of adsorbed proteins and activation of body defence systems plays an important role in the bioactivity of an implanted biomaterial. Among these systems, there have been much interest in complement activation and surface induced coagulation. We have recently developed a technique to quantify the ∗ To
whom correspondence should be addressed. Phone: +46-31-7732566. Fax: +46-31-7732599.
96
M. Andersson et al.
complement activation of different types of surfaces [1], using Quartz Crystal Microbalance with Dissipation monitoring (QCM-D) [2– 4]. In this study we modified the QCM-D sensor surfaces with materials of different physico-chemical characteristics, suitable as model surfaces for biomaterial research. To continue our investigation of blood-surface interaction with the QCM-D technique, we have now focused on the kinetics of coagulation on surfaces. Others [5, 6] have used the quartz crystal microbalance to measure the coagulation of blood or blood plasma on gold surfaces. Coagulation resulted in a dramatic decrease of resonance frequency and an increase of dissipation. However, in these studies the coagulation was initiated by adding thromboplastin, which makes it impossible to determine the effect of surface induced coagulation. Two different pathways can initiate the coagulation system. There is the extrinsic pathway, which involves endothelial damage, and there is also the intrinsic pathway [7], which is activated upon contact with surfaces. The intrinsic pathway includes high molecular weight kininogen (HMWK), factor XII (FXII), factor XI (FXI) and prekallikrein (PK). Contact associated coagulation on charged surfaces, such as titanium and glass, usually takes place rapidly, while for hydrophobic polymers, contact associated activation is prolonged [8]. Detection of the contact associated coagulation at a surface can be made in different ways, for example detection of kallikrein formation [9] and deposition of high molecular weight kininogen [7, 10, 11]. Also corn trypsin inhibitor (CTI), a specific FXII-inhibitor active in the bulk phase [8, 12– 15], has been used to study the pharmaco-dynamic aspects of intrinsic pathway activation. Platelets also play a central role in blood coagulation as binding sites for factor X and as an initiator of the coagulation. However, even a supposedly ‘platelet free’ blood plasma seems to contain platelet fragments enough to ensure a near normal coagulation response [16]. The aim of the present study was to study the contact associated blood plasma coagulation on different modified QCM-D sensor surfaces; and to investigate the intrinsic pathway activation in the presence of CTI and to determinate surface associated HMWK.
MATERIALS AND METHODS
Preparation of plasma and solutions We used citrated blood plasma, rather than whole blood, in order to isolate the intrinsic pathway, without the interference of blood cells. Fresh, citrate, anticoagulated, human blood from one healthy donor, which appeared to have a normal coagulation function, (first 5 ml was discarded to avoid any disturbance from tissue factor) was centrifuged (Universal 16R, 4880 rpm, 37 ◦ C, 11 min), the plasma supernatant was collected and immediately frozen (−80 ◦ C) until further use. Thawed plasma was used immediately. For the experiments of HMWK adsorption, the plasma was diluted to 10% in Veronal buffered saline (VBS) (0.15 M NaCl,
Acoustics of blood plasma on solid surfaces
97
1.8 mM Na-5,5-diethylbarbiturate, 3.1 mM 5,5-diethylbarbituric acid, 0.5 mM NaN3 , pH 7.4). CTI (Dia-Service AB, Göteborg, Sweden) was diluted to 5 mg/ml in 20 mM TrisHCl, 0.03 M NaCl, pH 8.2. Antibody preparation directed against HMWK was diluted 1 : 50 in VBS (a-HMWK IgG fraction, The Binding Site, Birmingham, UK). Preparation of surfaces All the surfaces were prepared on gold plated QCM-D crystals (Q-Sense AB, Göteborg, Sweden). Four different kinds of surfaces were used, titanium (Ti), heparin coated (Hep), spin-coated polystyrene (PS) and poly(urethane urea) (PUUR). The Ti [17, 18] were sputtered (Q-Sense AB, Göteborg, Sweden) to a thickness of about 300 nm. The heparin-coated surfaces (a kind gift from Rolf Larsson, Corline Systems AB, Uppsala, Sweden) was used as a negative control surface [19– 21]. The Corline heparin surface is prepared by irreversible adsorption of a macromolecular conjugate composed of seventy heparin molecules covalently linked to a 50 kD polymer carrier. The surface concentration of heparin is approximately 0.5 µg/ cm2 , which is equivalent to a film thickness of about 10 nm. The heparin was coated on titanium. The underlying gold surface used for the spin-coated polymers, was cleaned in an UV /ozone chamber for 10 min, followed by immersion in a 5 : 1 : 1 mixture of milli-Q water, H2 O2 (30%) and NH3 (25%) for 10 min at 70 ◦ C. General purpose FDA-approved PS [22] microtitration plates (Sero-Wel, Bibby Sterlin Ltd, Uk) was dissolved in toluene (0.5% w /w PS in toluene). PUUR consisting of repeated soft and hard segments was synthesized in solution using a two-step polymerization method. Polycaprolactone diol (PCL) as the soft segment and 4,4 -diphenylmethane diisocyanate (MDI), 1,3-diaminopropane (1,3-DAP) as the hard segment. The solid polymer was then dissolved in dimetylformamide (5% w /w PUUR in dimetylformamide). The polymers (50 µl) were spin-coated (Chemat Technology spin-coater KW-4A, 2000 rpm and 5000 rpm respectively for 2 min) on the cleaned gold surfaces. The thickness of the polymer layers was about 20 nm. The hydrophobicity of the surfaces was examined by measuring the diameter of a 10 µl sessile water drop [23] (Table 1). Table 1. Advancing contact angle estimations by measuring the diameter of 10 µl sessile water drop (n = 3) Surface
Contact angle () degrees
SD
Ti PUUR PS Hep
10.2◦ 67.1◦ >85◦ <5◦
0.52 2.4 — —
98
M. Andersson et al.
The QCM-D technique and measurements The QCM-D technique is a method, which is basically a piezoelectric quartz crystal set in a resonant lateral motion with a predefined fundamental frequency of 5 MHz. Simultaneous frequency (f ) and dissipation (D) measurements are made by periodically switching on and off the voltage over the crystal. The decay signal is recorded and numerically fitted to an exponentially damped sinusoidal curve. When molecules adhere to the surface, the resonant frequency (f ) decreases. The frequency (f ) is used to calculate the adhered mass (m) per area (A), using the Saurbrey equation (1) [4]; Cf m = , A nr
(1)
where nr is the overtone number (= 1, 3, . . .) and C is the mass-sensitivity constant (17.7 ng cm−2 Hz −1 ). The measured mass includes the mass of associated water [24]. In addition, the decay-time is measured (τ ), and is used to calculate the dissipation (D) (2). D=
1 . πf τ
(2)
The dissipation gives information about the viscoelastic properties of the adlayer in relation to the adsorbed mass. This has turned out to be a very valuable as well as reliable source of information. The penetration depth of the QCM-D can be calculated by using equation (3); η , (3) δ= πfρ where δ is the penetration depth, η is the shear viscosity and ρ is the density of the layer. This results in a penetration depth in plasma of about 170 nm for the third overtone. The Ti resonates at about the same frequency as the underlying gold, and has therefore about the same penetration depth. A Quartz Crystal Microbalance with Dissipation monitoring apparatus, QCM-D (model D300, Q-Sense AB, Göteborg, Sweden) was used to study the kinetics of plasma clot formation. The fibrin clot measurements were performed in an open QCM-D measurement cell, while the initial protein adsorption study was carried out in a closed, temperature controlled QCM-D measurement cell (Q-Sense, Göteborg, Sweden). All the experiments were conducted in room temperature (+22.0 ◦ C ± 0.2 ◦ C), and after achieving a stable baseline of frequency and dissipation registration. All the measurements were conducted at least three times.
Acoustics of blood plasma on solid surfaces
99
RESULTS
Blood plasma coagulation at modified QCM-D sensor surfaces To study the plasma clot formation, citrated plasma (100 µl) was incubated on the coated sensor surface. After 2 minutes (a stable baseline of frequency and dissipation registration was achieved) a fresh solution of plasma (90 µl) and CaCl2 (10 µl, 0.25 M in MilliQ) was added. The results were presented in Fig. 1. On the Ti surfaces we observed a rapid decrease in frequency followed by a relatively rapid stabilisation of frequency at about −500 Hz. Prolonged incubation for 25 minutes did not result in further decrease in frequency. There was also a visible coagulation of blood plasma after 10 minutes on the Ti surfaces. At the Hep surfaces we observed, as expected, no change in frequency shift with a prolonged incubation of blood plasma and there were no visible coagulation of plasma after 80 minutes of incubation. The decrease in frequency at the PS and PUUR surfaces typically occurred after a 10– 20 minutes lag phase. After that, the decrease in frequency levelled off at about f = −1300 Hz, which was a very large frequency shift compared to the adsorption of a monolayer of proteins [3]. It might be suggested that the curve showed multilayers of protein (fibrin). We also observed a gentler slope in the case of the polymer surfaces, compared to that of the Ti surface, which coincided with the prolonged coagulation times. What also coincided with a prolonged coagulation time was the final value of the frequency shift. This
Figure 1. Blood plasma coagulation measured as frequency shifts (third overtone 15 MHz) registered by the QCM-D for four surfaces; Ti, Hep (control), PS and PUUR.
100
M. Andersson et al.
Figure 2. Kinetic effects of CTI (40 µg/ ml) added to plasma on coagulation time on different surface modifications (n = 3) as indicated in the figure. Time reference is set when 50% of total frequency shift has occurred (coagulation50 ).
might suggest a difference in the assembly of the fibrin structure on the Ti surface compared to that on the polymer surfaces. There was also visible coagulation of plasma on the Ti, PUUR and PS surfaces as detected when the surfaces were rinsed at the end of the experiments. All the experiments in Fig. 1 were repeated at least three times with very little difference in the reaction patterns. In further experiments we expressed the coagulation activity on the QCM-D sensor surfaces as the time (Coagulation50 time) needed for obtaining 50% decrease in frequency after adding Ca2+ . The values thus obtained were 10 min, 25 min and 40 min for Ti, PUUR and PS, respectively. Effect of CTI on coagulation time at different coated QCM-D surfaces The CTI is known to be a specific inhibitor to FXIIa and is used here to investigate the degree of intrinsic pathway activation on our experimental surfaces. We repeated the experiments presented in Fig. 1 with the addition of CTI (40 µg/ ml) in plasma 4 minutes before applying the plasma mixture to the coated QCM-D surfaces. The Hep surfaces were not included in these experiments. The results of the experiments were presented in Fig. 2. We found a significant increase in coagulation time on the Ti surfaces when CTI was added. The coagulation time was also significantly increased on the PUUR and PS surfaces. The relative change of coagulation time was independent of the kind of material used (CTI affects the bulk concentration of factor XIIa and does not affect the factor XII already bound to the surface). Detection of adsorbed HMWK on coated QCM-D surfaces The traditional picture of the intrinsic pathway activation includes the association HMWK to the activating surfaces [7, 10, 11]. But since CTI prolonged the coagulation time also for the PS and PUUR surfaces (Fig. 2) we investigated the quantity of HMWK association on the coated QCM-D surfaces. After 30 minutes of incubation in plasma (without adding Ca2+ ), the surfaces were rinsed and incubated
Acoustics of blood plasma on solid surfaces
101
Figure 3. Detection of HMWK on various surfaces viewed at the third overtone (15 MHz). 10% plasma (in VBS-buffer) was added, incubated for 30 min incubation, rinsed shortly (‘r’), and antiHMWK was added. Note that the Ti surface and the PUUR surface shows an early (3 min after addition of plasma) transient desorption of mass while PS surface doesn’t.
with rabbit anti-human HMWK. The results were shown in Fig. 3. It was obvious that PUUR and PS surfaces adsorbed very little anti-HMWK whereas the Ti surface adsorbed a significant amount of HMWK antibodies. Thus it might be indicated from these experiments that the traditional picture of the intrinsic pathway activation is not entirely adequate (see discussion). Analysis of frequency and dissipation of adsorbed plasma proteins A unique feature of the QCM-D method is the possibility to analyze continuously both frequency (adsorbed protein and associated water) and dissipation (viscoelastic properties of adsorbed proteins). If an adsorbed protein layer gives a proportional contribution of adsorbed mass and viscoelasticity, a D/f plot of an adsorption process will have a linear relationship. However, if the relation between D and f is changed during the adsorption it will appear as a non-linear relationship in a D/f plot. When such extended D/f plots were made for the results from the first two experiments (coagulation without and with CTI) we found linear relationships between D and f . Thus, the fibrin clot formation on the modified QCM-D surfaces did not result in any non-linear curve of the D/f plots. This was however not the case with 10% citrated plasma (without Ca2+ ), where a non-linear D/f plot was found (Fig. 4A and B). For Ti an initially linear curve
102
M. Andersson et al.
Figure 4. (A and B) Frequency shift (third overtone 15 MHz) vs. dissipation plots for the first 15 min of plasma (10%) protein adhesion on Ti (squares) and PS (diamonds) surface (every 10th data point is marked).
transcended to a non-linear mass-displacement curve (after about 3 min). Figure 4B is an enlargement of Fig. 4A. Notice that during the increase of frequency (mass displacement) (3– 15 min) of the Ti surface, the damping energy is fairly constant. This displacement is not seen on PS surface.
Acoustics of blood plasma on solid surfaces
103
DISCUSSION
We have used the QCM-D as a method for real time clot formation measurements on different surfaces. As expected, the titanium surface resulted in a rapid clot formation while the polymer-coated surfaces exhibited a considerably longer time to form the fibrin clot. Our control (covalently bound heparin) showed no apparent signs of coagulation. It was indicated that CTI sensitive intrinsic pathway (FXIIa) might be important for all the artificial surfaces. This observation correlated to the study made by Hong et al. [8]. They also observed a CTI inhibitor sensitive trombin-antitrombin complex formation in association with both titanium and the more hydrophobic poly-vinylchloride surfaces. In our second set of experiments, where we tried to detect HMWK attached to the surface using polyclonal antibodies, we found, as expected, high amounts of antibodies (toward HMWK) associated on the titanium surface but low association of antibodies on the polymer surfaces. As described above, the propagation of the surface sensitive clot formation is dependent on several things, whereof the two most crucial features are the auto-activation of FXII on the surface and the binding of HMWK to the surface. It was observed in this study that the polymer surfaces were affected by CTI sensitive FXII, but the coagulation seemed to proceed without any HMWK present at the surface. One hypothesis is that residues from platelets, still active in the plasma, in combination with material properties can allow the coagulation process to continue. In a recent study, more platelets were detected on Ti than on polyvinyl chloride surfaces [8] and increased amounts of platelet derived growth factor (PDGF) and β-thromboglobulin were detected in association with Ti [8]. Further, it is also reported that presence of CTI reduces the release of β-thromboglobulin from platelet α-granule [15]. We found a non-linear relation of the D/f plot of adsorption of diluted plasma on the Ti surfaces (Fig. 4). The D/f plots initially appeared linear when plasma proteins were adsorbed. Three minutes after addition of plasma it seems that there was an increase in frequency without a corresponding decrease in dissipation. This might be interpreted as some kind of mass desorption, which does not affect the damping energy. At the present we do not have a molecular explanation to this phenomena but we cannot exclude that it might be associated with the HMWK and the intrinsic pathway. It has previously been reported that the composition of the adsorbed protein layer deposited from citrated is time-dependent [25], and that there is a displacement of fibrinogen by HMWK on negatively charged glass [26]. Taken together, this suggests a more complex nature of the clotting mechanism than the traditional model of intrinsic pathway. The QCM-D technology is a very attractive alternative or complement to optical methods, such as ellipsometry [27, 28] or surface plasmon resonance, [5, 29, 30] for the investigation of protein adsorption, immune complement activation and surface associated blood coagulation on experimental surfaces. The sensitivity of QCM-D for detection of protein adsorption, is, compared to the optical methods, in the same
104
M. Andersson et al.
range [1]. An additional advantage of the QCM-D method is that it is relatively insensitive to poor optical properties of the test solution. Thus, the QCM-D method allows experiments using whole blood [5] and therefore it is easy to study proteinsurface interactions on a wide range of sensor surface modifications, viewed in real-time [1]. We consider the QCM-D method a very valuable tool in biomaterial research and in investigations of bio-interfaces in general.
CONCLUSION
A great advantage of the QCM-D method is that the end effect of contact associated coagulation, fibrin formation, can be determined at the same surface that initiated the coagulation cascade. This opened up several new experimental possibilities in research about the molecular background of the intrinsic pathway. We have also showed that the method is well suited as a screening method for new materials. Acknowledgements We wish to thank Entific Medical System AB (Göteborg, Sweden) and Vinnova for research funding and Artimplant AB for providing the spin-coater and help with the polymers. We also wish to express our gratitude to Rolf Larsson at Uppsala University (Uppsala, Sweden). Parts of the methodology used in this study have been submitted for patent pending. Parts of this study have been presented at ESB biomaterial conference (London, September 2001).
REFERENCES 1. A. Sellborn, M. Andersson, C. Fant, C. Gretzer and H. Elwing, Colloid Surface B. (submitted). 2. M. Rodahl, F. Hook, C. Fredriksson, C. A. Keller, A. Krozer, P. Brzezinski, M. Voinova and B. Kasemo, Faraday Discuss. 107, 229 (1997). 3. F. Hook, M. Rodahl, P. Brzezinski and B. Kasemo, J. Colloid Interface Sci. 208, 63 (1998). 4. G. Saurbrey, Z. Phys. 155, 206 (1959). 5. T. P. Vikinge, K. M. Hansson, P. Sandstrom, B. Liedberg, T. L. Lindahl, I. Lundstrom, P. Tengvall and F. Hook, Biosens. Bioelectron. 15, 605 (2000). 6. T. J. Cheng, H. C. Chang and T. M. Lin, Biosens. Bioelectron. 13, 147 (1998). 7. R. W. Colman and A. H. Schmaier, Blood 90, 3819 (1997). 8. J. Hong, J. Andersson, K. N. Ekdahl, G. Elgue, N. Axen, R. Larsson and B. Nilsson, Thromb. Haemost. 82, 58 (1999). 9. M. Lestelius, B. Liedberg, I. Lundstrom and P. Tengvall, J. Biomed. Mater. Res. 28, 871 (1994). 10. B. Walivaara, I. Lundstrom and P. Tengvall, Clin. Mater. 12, 141 (1993). 11. P. Tengvall, A. Askendal, I. Lundstrom and H. Elwing, Biomaterials 13, 367 (1992). 12. D. Basmadjian, M. V. Sefton and S. A. Baldwin, Biomaterials 18, 23 (1997). 13. D. J. Schneider, P. B. Tracy, K. G. Mann and B. E. Sobel, Circulation 96, 2877 (1997). 14. Y. Hojima, J. V. Pierce and J. J. Pisano, Thromb. Res. 20, 149 (1980). 15. J. Hong, A. Larsson, K. N. Ekdahl, G. Elgue, R. Larsson and B. Nilsson, J. Lab. Clin. Med. 138, 139 (2001).
Acoustics of blood plasma on solid surfaces
105
16. K. W. van der Kamp, K. D. Hauch, J. Feijen and T. A. Horbett, J. Biomed. Mater. Res. 29, 1303 (1995). 17. B. Walivaara, A. Askendal, I. Lundstrom, and P. Tengvall, J. Biomater. Sci. Polymer Edn 8, 41 (1996). 18. P. Tengvall, in: Titanium in Medicine: Material Science, Surface Science, Engineering, Biological Responses and Medical Applications, D. M. Brunette (Ed.), p. 457. Springer, Berlin (2001). 19. K. Christensen, R. Larsson, H. Emanuelsson, G. Elgue and A. Larsson, Biomaterials 22, 349 (2001). 20. W. J. van Der Giessen, H. M. van Beusekom, R. Larsson and P. Serruys, Curr. Interv. Cardiol. Rep. 1, 234 (1999). 21. M. Johnell, G. Elgue, R. Larsson, A. Larsson, S. Thelin and A. Siegbahn, J. Thoracic. and Cardiovasc. Surg. (2002) (in press). 22. J. M. Grunkemeier, W. B. Tsai and T. A. Horbett, J. Biomed. Mater. Res. 41, 657 (1998). 23. C. Dahlgren and T. Sunqvist, J. Immunol. Meth. 40, 171 (1981). 24. F. Hook, B. Kasemo, T. Nylander, C. Fant, K. Sott and H. Elwing, Anal. Chem. 73, 5796 (2001). 25. L. Vroman and A. L. Adams, J. Biomed. Mater. Res. 3, 43 (1969). 26. L. Vroman, A. L. Adams, G. C. Fischer and P. C. Munoz, Blood 55, 156 (1980). 27. H. Elwing, Biomaterials 19, 397 (1998). 28. P. Tengvall, I. Lundstrom, C. Freijlarsson, M. Kober and B. Wesslen, J. Mater. Sci.-Mater. M 4, 305 (1993). 29. K. M. Hansson, T. P. Vikinge, M. Ranby, P. Tengvall, I. Lundstrom, K. Johansen and T. L. Lindahl, Biosens. Bioelectron. 14, 671 (1999). 30. T. P. Vikinge, K. M. Hansson, J. Benesch, K. Johansen, M. Ranby, T. L. Lindahl, B. Liedberg, I. Lundstom and P. Tengvall, J. Biomed. Opt. 5, 51 (2000).
This page intentionally left blank
Development of small alginate microcapsules for recombinant gene product delivery to the rodent brain C. J. D. ROSS 1,∗ and P. L. CHANG 1,2,† 1 Department 2 Department
of Biology, McMaster University, Hamilton, Ontario, Canada of Pediatrics, McMaster University, Hamilton, Ontario, Canada
Received 12 November 2001; accepted 1 May 2002 Abstract—A novel form of gene therapy using encapsulated recombinant cells in alginate microcapsules has proven effective in treating several animal models of human diseases. For treating neurological deficits in rodents with this technology, the size of the microcapsules has to be reduced for implantation in the central nervous system (CNS) to bypass the blood– brain barrier. This article reports the development of small alginate microcapsules suitable for implantation into the mouse CNS. By varying the encapsulation protocol, recombinant cells could be encapsulated in microcapsules ranging in diameter from 5 to 2000 µm. The optimal size for implantation was determined to be 100– 200 µm, based on the smallest, homogeneously sized, cell-filled microcapsules that could pass the 500 µm inner diameter of a CNS-implantation needle. Compared with medium-sized (500– 700 µm) microcapsules, these small microcapsules packed more tightly together with less inter-capsule space, resulting in an increased number of cells and a higher rate of recombinant gene product secretion per volume of microcapsules. The small microcapsules also displayed increased mechanical strength, compared with large microcapsules. These excellent in vitro properties of small 100– 200 µm microcapsules warrant further in vivo investigation into the feasibility of using immuno-isolation gene therapy to deliver recombinant gene products to the rodent CNS. Key words: Cell therapy; gene therapy; microencapsulation; alginate; immuno-isolation.
INTRODUCTION
Delivery of therapeutic gene products to the central nervous system (CNS) is important for treatments of neurodegenerative diseases, such as the lysosomal storage diseases [1]. However, the blood– brain barrier (BBB) impedes the passage ∗ Present address: Department of Medical Genetics, University of British Columbia, Vancouver, British Columbia, Canada. † To whom correspondence should be addressed. Department of Pediatrics, Health Sciences Centre Room 3N18, McMaster University, 1200 Main St. West, Hamilton, Ontario, Canada L8N 3Z5. Phone: (905)525-9140, ext. 22278. Fax: (905)521-1703. E-mail:
[email protected]
108
C. J. D. Ross and P. L. Chang
of many substances into the brain, thereby limiting the application of most current gene-transfer protocols primarily aimed at delivering genes to peripheral organs. Attempts have been made to introduce foreign gene products directly into the CNS, bypassing the BBB. First proposed in 1987, grafting genetically-modified autologous cells in the CNS can produce recombinant marker gene expression for up to 8 weeks [2, 3]. However, the CNS is not totally immunologically isolated and the need for expensive, patient-specific, genetically modified cells may limit the cost-effectiveness of such an approach. In contrast, immuno-isolation gene therapy could treat many patients who require the same recombinant gene product with a single ‘universal’ recombinant cell line [4]. Polymer-encapsulated cells have been implanted in the brain to produce recombinant nerve growth factor [5] and recombinant ciliary neurotrophic factor [6, 7]. Most of the over 50 lysosomal storage diseases affect the CNS with severe neurodegenerative consequences [1, 8]. Deficiency of β-glucuronidase causes the neurodegenerative lysosomal storage disease mucopolysaccharidosis VII (MPS VII) [9]. Immuno-isolation gene therapy using alginate– poly-L-lysine– alginate (APA) microcapsules reduces the peripheral disease in the mouse model of MPS VII [10]. However, β-glucuronidase is unable to bypass the BBB, and neither β-glucuronidase activity nor clinical improvements are observed in the CNS. Potentially, APA microcapsules could be implanted into the brain to provide a source of β-glucuronidase for the CNS. However, APA microcapsules manufactured through current techniques are generally too large (500– 1500 µm in diameter) for safe implantation into the rodent CNS. The smallest alginate microcapsules that have previously been characterized and used in vivo were 315 µm in diameter [11]. Since alginate microcapsule size depends on several parameters, including: the diameter of the alginate-extrusion needle, the airflow rate, and the outer diameter of the airflow nozzle [12], we now report the development of smaller APA microcapsules (100– 200 µm in diameter) through modifications of the production protocol. These microcapsules turned out to have properties that are superior to the regular-sized capsules and are ideal for implantation into the rodent CNS.
MATERIALS AND METHODS
Cells Mouse 2A50 fibroblasts (gift from W. S. Sly, St. Louis, MO, USA), transfected with pMSXND-MβG to express mouse β-glucuronidase [13], were maintained in DMEM supplemented with 0.1 g/ l sodium pyruvate, 2.2 g/ l sodium bicarbonate, 1.2 mM L-glutamine (Gibco, BRL), and 3.2 µM methotrexate. Alginate microcapsules Cells were encapsulated in standard APA microcapsules [10], except that the starting cell number and the airflow over the extruded alginate were increased for
Microcapsules
109
smaller microcapsules. Normal (500– 700 µm) microcapsules were generated with a standard 2×106 cells per ml of alginate and a regular airflow (3 l/ min) over the tip of the needle from which the alginate solution was extruded. Small (100– 200 µm) microcapsules were generated with a 2.5-fold higher starting cell number and a higher airflow (6 l/ min). Very small (<100 µm) microcapsules were also generated with a 2.5-fold higher starting cell number and the highest airflow (7– 8 l/ min). Unlike standard microcapsules, the smaller microcapsules did not settle during washes, so all washes were shortened to leave time for centrifugation (30 s, 20 × g, Sorval RT6000B, 300 rpm). The short centrifugation allowed the supernatant of each wash solution to be removed before the next wash. Analysis of microencapsulated cells The number of microcapsules, cell number, and cell viability using the Trypan blue exclusion test were determined as previously described [10]. Recombinant gene product secretion The rate of β-glucuronidase secretion from microencapsulated cells was determined as previously described [10, 14]. Microcapsule osmotic pressure test The resistance of microcapsules to osmotic stress from exposure to hypotonic solutions was as previously described [15]. All statistical analyses were carried out using the Student’s t-test.
RESULTS
Development of small alginate microcapsules Smaller alginate microcapsules were developed for implantation into the intraventricular space of mice. To reduce the size of the alginate beads, the airflow was increased over the tip of the needle from which the alginate solution was extruded. The increased airflow broke off the alginate beads more rapidly, producing smaller microcapsules. By varying the airflow, initial experiments produced microcapsules ranging from 5 to 3000 µm in diameter. The high airflow (>7 l/ min) necessary to produce very small microcapsules (<100 µm) produced a heterogeneous population of microcapsules, ranging widely from 5 to 1000 µm in diameter (Fig. 1a). Although microcapsules of a desired size could be separated from this heterogeneous mixture by varying the time that the microcapsules were allowed to settle, this procedure was time-consuming and laborintensive. However, by controlling the airflow at a slightly lower rate, uniformly sized 100– 200 µm microcapsules could be generated that did not require further
110
C. J. D. Ross and P. L. Chang
Figure 1. Production of different microcapsule sizes. Varying the airflow over the tip of the alginate extrusion needle generated various sizes of alginate microcapsules. A high airflow (>7 l/ min) produces very small microcapsules; however, the microcapsules were not uniformly sized, ranging from 5 to 1000 µm diameter (a). A slightly lower airflow (6 l/min) produced homogeneously sized small microcapsules (100– 200 µm) (b). Standard alginate microcapsules (500– 1200 µm) are generated with an airflow of 3 l/ min (c). Scale bar = 500 µm.
size selection (Fig. 1b). Thus, 100– 200 µm microcapsules were the smallest homogenously sized microcapsules that could be quickly and reliably generated. However, in microcapsules less than 200 µm in diameter, up to 50% were empty microcapsules devoid of cells. These small empty microcapsules could not be separated from cell-carrying microcapsules. Since these empty microcapsules would reduce the amount of cells delivered in a given volume of microcapsules, the efficiency of the implant would be decreased. Hence, the initial density of cells in the alginate solution was increased 2.5-fold to circumvent this problem. The number of empty microcapsules was subsequently reduced to less than 10%. Finally, it was determined that microcapsules less than 200 µm in diameter were suitable for injection through a 500 µm inner-diameter injection needle suitable for murine intraventricular implantation in the CNS. Microcapsules of a larger size could not be efficiently loaded into the injection needle. Since 100 µm was the smallest microcapsule diameter that could be reliably generated, and microcapsules of less than 200 µm were suitable for in vivo injection, small 100– 200 µm diameter microcapsules were optimal for rodent CNS implantation. Microcapsule packing efficiency The empty space between microcapsules was reduced in smaller microcapsules (Table 1). As a result, the number of microcapsules per volume of packed microcapsules was significantly increased. Very small (<100 µm) microcapsules could be packed to more than 300 microcapsules per microliter, as opposed to 20– 50 for small microcapsules and 4– 6 for medium microcapsules (Fig. 1c). The efficacy of the ‘payload’ is thus dramatically improved by reducing the diameter from a standard 500 µm capsule to a small 200 µm capsule.
Microcapsules
111
Table 1. Theoretical properties of various microcapsule sizes Capsules
Diameter (µm)
Very small 50 Small 150 Medium 600 Large 1200
Volume (µl) 4/3 πr 3
Cubic space (µl) (2r)3
0.07 1.8 113.1 904.8
0.13 3.4 216.0 1728.0
Unoccupied space (µl) (2r)3 − 4/3πr 3
Relative unoccupied space per capsule
Surface area (mm2 ) 4πr 2
Relative surface area to volume ratio
0.06 1.6 102.9 823.2
13 824 512 8 1
8 71 1131 4524
24 8 2 1
Small microcapsules (∼150 µm) have a theoretical 64-fold smaller volume, cubic space, and unoccupied volume than medium microcapsules (∼600 µm diameter) based on a cuboidal packing model. Small microcapsules also have a 16-fold smaller surface area; however, they have a four-fold greater surface area to volume ratio. Very small microcapsules (50 µm in diameter) have an even greater surface area to volume ratio.
Encapsulated cells Although cell viability did not differ between small (100– 200 µm) and medium (500– 700 µm)-size microcapsules (Fig. 2a), the increased packing efficiency of small versus medium microcapsules (51.6 vs. 4.3 microcapsules per microliter) significantly increased the number of cells per volume of packed microcapsules, by up to 9.4-fold at week 2 (Fig. 2b). In both types of microcapsules, cells grew quickly until reaching a peak capacity (Fig. 2c). The capacity of small microcapsules peaked by 10 days (770 ± 50 cells per microcapsule), while the medium microcapsules peaked on day 21, with three-fold more cells per microcapsule (2500± 110 cells per microcapsule). Thus, even with a reduced peak number of cells per microcapsule, the small microcapsules sustained a significantly higher number of cells per volume of packed microcapsules. Production of recombinant gene product The production of recombinant gene products from the encapsulated cells was monitored by comparing the rates of β-glucuronidase secretion between small (100– 200 µm) and medium (500– 700 µm) microcapsules. Although there was no significant difference in the rate of enzyme production per cell in either microcapsule type, the higher cell number per volume of small microcapsules significantly increased the resulting rate of gene product secretion per volume of microcapsules (Fig. 2d). β-Glucuronidase secretion per volume of microcapsules was up to 8.7-fold higher in small microcapsules at week 2, compared with medium microcapsules. Although this level of enzyme production from small microcapsules declined over time, it remained higher than that of medium microcapsules up to 4 weeks post-encapsulation (end of experiment).
112
C. J. D. Ross and P. L. Chang
Figure 2. In vitro properties of small microcapsules. The viability of cells encapsulated within small (100– 200 µm) and medium (500– 700 µm) alginate microcapsules was similar (a). However, up to ten times as many small microcapsules could be packed into the same volume as medium microcapsules. As a result, the small microcapsules had a significantly higher cell density (up to 9.4fold higher) (b), even though individually each small microcapsule held only 30% as many cells (c). As a result, the small microcapsules exhibited a significantly higher rate of recombinant gene product secretion per milliliter of microcapsules (up to 8.7-fold higher) (d).
Microcapsule strength The small and medium microcapsules had a similar microcapsule wall thickness (11 ± 5 vs. 13 ± 2 µm; small vs. medium microcapsules, p = 0.18). The microcapsule strengths of very small (<100 µm in diameter), small (100– 200 µm), and medium microcapsules (500– 700 µm) were compared using the osmotic pressure test [15]. Although the strength of very small microcapsules was at, or above, the range of the assay, it was clear that very small microcapsules were significantly stronger than both small and medium microcapsules (Fig. 3a). Small microcapsules were significantly stronger than medium microcapsules (p < 0.001), with an average six-fold increase in strength (Figs 3b and 3c) when their resistance to osmotic stress in water was compared. There is an inverse relationship between microcapsule size and microcapsule strength (Fig. 4).
Microcapsules
113
Figure 3. Strength of varying microcapsule sizes. The strengths of very small (<100 µm), small (100– 200 µm), and medium (500– 700 µm in diameter) alginate microcapsules were compared using the osmotic pressure test. The very small microcapsules (a) were clearly the strongest, with almost no breakage, and were at least 16% stronger than small microcapsules (b) (p < 0.0001). The small microcapsules were an average six-fold stronger than the medium microcapsules (c) (p < 0.01). The difference between very small and small microcapsules may have been underestimated, because the strength of the very small microcapsules was at, and likely above, the maximum range of the assay.
Figure 4. Microcapsule strength compared with microcapsule size. The relative strength of the three different alginate microcapsules was inversely proportional to the microcapsule size (r 2 = 0.9998). The relative microcapsule strength increased roughly 16% per 100 µm decrease in diameter.
DISCUSSION
Although the small microcapsules in this study were developed for the primary purpose of fulfilling the requirement for implantation in small rodent brains with stereotaxic guidance [16], they were discovered to confer several additional advantages over larger microcapsules. Overall, the encapsulated cell viability and recombinant gene product secretion per cell were similar between small (100– 200 µm)
114
C. J. D. Ross and P. L. Chang
and medium (500– 700 µm) microcapsules. This indicated that the size of the capsules (within the range under study in this report) does not affect the nutritional status of the encapsulated cells or product diffusion from the capsules. The four-fold reduction in diameter from the medium (500– 700 µm diameter) to the small microcapsules (100– 200 µm) increased the observed small microcapsule packing density by 12-fold, compared with the medium microcapsules. Although the theoretical reduction in unoccupied space between small microcapsules estimates a 64-fold greater microcapsule packing density, this estimate does not take into account the fluid nature of small microcapsules in suspension (Table 1). Nonetheless, Kepler’s conjecture states that the ratio of unoccupied space between spheres is constant for all sphere diameters. In other words, the increased packing density would be inconsequential if the number of cells per microcapsule were equivalently reduced. However, the small microcapsules exhibited only an average three-fold reduction in cells per microcapsule, compared with medium-size microcapsules. Thus, this three-fold reduction in cell capacity in combination with the 12-fold increase in packing density would be expected to result in a net four-fold higher ‘payload’ of cells per ml of microcapsules. Consequently, even though the small microcapsules held only 30% as many cells in each individual microcapsule, the small microcapsules packed an actual 5.8-fold more cells per ml of microcapsules (up to 9.4-fold more) (Fig. 2b) and secreted an average 4.1-fold more gene product per ml of microcapsules (up to 8.7-fold more) (Fig. 2d). The ability to sustain the higher number of cells per volume of small microcapsules was likely the result of the increased surface area to volume ratio. The 4.1-fold and 5.8-fold increases in average gene product secretion per ml and in cells per microcapsule correlate with the theoretical four-fold increase in the surface area to volume ratio, compared with the medium microcapsules (Table 1). In this study, the rate of recombinant gene product secretion from small microcapsules noticeably declined after 10 days. This trend began after the small microcapsules reached a peak number of cells per microcapsule, after which the cell viability and cell number gradually began to decline an average of 12% and 17% a week, respectively. Since the small microcapsules were sustained in the same volume of medium as medium-size microcapsules, the up to 9.4-fold higher cell number in the small microcapsules may have outgrown the nutrient supply more rapidly than that of the medium-size microcapsules. Even though the medium was changed regularly, the medium microcapsules, with fewer cells per volume, could have had an advantage over the small microcapsules. In vivo, the nutrient supply provided through the lymphatic system and the cerebral spinal fluid would be continuously exchanged and not subject to the in vitro limitations of episodic medium changes. Hence, it is possible that in vivo, the small microcapsules would in fact exhibit an improved rate of nutrient and waste exchange, compared with larger microcapsules, because of the increased surface area to volume ratio of small microcapsules. While the small and medium microcapsules had a similar microcapsule wall thickness, the the smaller diameter and volume of the small microcapsules re-
Microcapsules
115
sulted in a three-fold higher ratio of microcapsule wall thickness to microcapsule diameter (6.2 × 10−2 vs. 1.9 × 10−2 ; small vs. medium microcapsules wall thickness/ diameter) and a 51-fold higher ratio of wall thickness to volume (174 × 10−7 vs. 4 × 10−7 ; small vs. medium microcapsule wall thickness/ volume). Hence, the increased strength of smaller microcapsules (Figs 3b and 3c) is likely due to the relative increase in microcapsule wall thickness in relation to microcapsule size. Although very small microcapsules (<100 µm) appear the strongest (Fig. 3a), our current airflow-dependent protocol is time-consuming and limited to small-scale production. Using filtration rather than centrifugation for the washing steps and alternative methods of capsule production [17– 20] should speed up the protocol. Smaller alginate microcapsules could reduce the size of the microcapsule injection device, thereby reducing the surgical trauma on the brain. As a result, smaller microcapsules could increase the chance of observing intended behavioral corrections resulting from therapeutic gene product delivery, versus observing unintended deterioration as a consequence of trauma from surgery. Smaller APA alginate microcapsules also demonstrate an improved biocompatibility, as demonstrated by a reduced pericapsular fibrosis reaction when smaller (350 µm in diameter) microcapsules were compared with larger (1200 µm) microcapsules for the encapsulation of islets [20]. In conclusion, the use of smaller alginate microcapsules provides additional potency of gene product delivery and microcapsule strength. A four-fold reduction in the microcapsule size (from 600 to 150 µm) resulted in a four-fold higher capacity to sustain cells, and each 100 µm reduction in diameter from 600 to 50 µm produced a 16% relative increase in microcapsule strength. The potentially improved nutrient and waste exchange in vivo, increased number of cells per implant volume, and greater mechanical stability may lead to more efficient recombinant product delivery, longer survival of the microcapsules in vivo, smaller total implant volume with less ensuing surgical trauma, and improved accessibility to sizerestricted implantation sites, such as the brain in small animal models. This strategy has been successful for correcting the behavioral abnormality in a mouse model of a human neurodegenerative disease, mucopolysaccharidosis type VII [16]. Acknowledgements This work was supported by the Medical Research Council (CIHR) of Canada and the Ontario Mental Health Foundation (fellowship awarded to CJDR).
REFERENCES 1. E. F. Neufeld and J. Muenzer, in: The Metabolic Basis of Inherited Disease, C. R. Scriver, A. L. Beaudet, W. S. Sly and D. Valle (Eds), pp. 1565– 1587. McGraw-Hill, New York (1989). 2. F. H. Gage, J. A. Wolff, M. B. Rosenberg, L. Xu, J. K. Yee, C. Shults and T. Friedmann, Neuroscience 23, 795 (1987).
116
C. J. D. Ross and P. L. Chang
3. L. C. Doering and P. L. Chang, J. Neurosci. Res. 29, 292 (1991). 4. P. L. Chang, in: Somatic Gene Therapy, P. L. Chang (Ed.), pp. 203– 223. CRC Press, Boca Raton, FL (1995). 5. I. Date, T. Shingo, T. Ohmoto and D. F. Emerich, Exp. Neurol. 147, 10 (1997). 6. D. F. Emerich, S. R. Winn, P. M. Hantraye, M. Peschanski, E. Y. Chen, Y. Chu, P. McDermott, E. E. Baetge and J. H. Kordower, Nature 386, 395 (1997). 7. P. Aebischer, M. Schluep, N. Deglon, J. M. Joseph, L. Hirt, B. Heyd, M. Goddard, J. P. Hammang, A. D. Zurn, A. C. Kato, F. Regli and E. E. Baetge, Nature Med. 2, 696 (1996). 8. V. Gieselmann, Biochim. Biophys. Acta 1270, 103 (1995). 9. W. S. Sly, B. A. Quinton, W. H. McAlister and D. L. Rimoin, J. Pediatr. 82, 249 (1973). 10. C. J. Ross, L. Bastedo, S. A. Maier, M. S. Sands and P. L. Chang, Hum. Gene Ther. 11, 2117 (2000). 11. F. A. Leblond, G. Simard, N. Henley, B. Rocheleau, P. M. Huet and J. P. Halle, Cell Transplant. 8, 327 (1999). 12. G. H. Wolters, W. M. Fritschy, D. Gerrits and R. van Schilfgaarde, J. Appl. Biomater. 3, 281 (1991). 13. J. H. Grubb, J. M. Kyle, L. Cody and W. S. Sly, FASEB J. 7, A1255 (1993). 14. J. H. Glaser and W. S. Sly, J. Lab. Clin. Med. 82, 969 (1973). 15. J. M. Van Raamsdonk and P. L. Chang, J. Biomed. Mater. Res. 54, 264 (2001). 16. C. J. Ross, M. Ralph and P. L. Chang, Exp. Neurol. 166, 276 (2000). 17. B. R. Hsu, H. C. Chen, S. H. Fu, Y. Y. Huang and H. S. Huang, J. Formos. Med. Assoc. 93, 240 (1994). 18. A. King, S. Sandler, A. Andersson, C. Hellerstrom, B. Kulseng and G. Skjak-Braek, Diabetes Care 22, B121 (1999). 19. J. P. Halle, F. A. Leblond, J. F. Pariseau, P. Jutras, M. J. Brabant and Y. Lepage, Cell Transplant. 3, 365 (1994). 20. R. Robitaille, J. F. Pariseau, F. A. Leblond, M. Lamoureux, Y. Lepage and J. P. Halle, J. Biomed. Mater. Res. 44, 116 (1999).
Part II
Cells
This page intentionally left blank
Integrin α2β1 on rat myeloma cells modulates interaction of α4β1 integrin with vascular cell adhesion molecule-1 but not fibronectin BOSCO M. C. CHAN 1,2 , VINCENT L. MORRIS 2,3 , DOLORES HANGAN-STEINMAN 1,2 , BRENNA JARVIE 1,2 , MIHAELA CIALACU 1,2 , JAAN LAANSOO 1,2 , GREGORY HUNTER 1,2 , WANKEI WAN 4 and SHASHI UNIYAL 1,2,∗ 1 Biotherapeutic Research Group, The John P. Robarts Research Institute, London, Ontario, Canada 2 Department of Microbiology and Immunology, University of Western Ontario, London, Ontario,
Canada 3 Departments
of Oncology and Medical Biophysics, University of Western Ontario, London, Ontario, Canada 4 Department of Chemical and Biochemical Engineering, University of Western Ontario, London, Ontario, Canada Received 26 September 2001; accepted 19 November 2001 Abstract—It is well established that α2β1 integrin functions as a receptor for collagen and laminin; whereas α4β1 integrin binds fibronectin and vascular cell adhesion molecule-1 (VCAM-1). In the present study, we showed that rat myeloma YB2/ 0 cells constitutively expressed α4β1 but not α2β1 integrin. Transfection of cDNA of mouse α2 integrin subunit resulted in the expression of heterologous α2β1 integrin on YB2/0 cells (YBmα2). The expression of α2β1 conferred YBmα2 cells the ability to interact with collagen and laminin. In comparison with mock transfected YB2/ 0 cells (YBpF), YBmα2 cells exhibited increases in the binding and migration on VCAM-1; in contrast, both YBpF and YBmα2 were similar in their interactions with fibronectin or fibronectin fragment FN-40 that contains the binding site for α4β1 integrin. The interaction of α4β1 with VCAM-1 was further stimulated upon ligation with α2β1-specific mAb. The use of specific inhibitory mAb demonstrated the role of α4β1 in mediating the observed interactions with fibronectin and VCAM-1. Therefore, results show that expression of α2β1 differentially regulated α4β1 integrin function by stimulating its interactions with VCAM-1 but not fibronectin. The in vivo significance of α2β1 integrin expression was demonstrated by intravital videomicroscopy showing that ligation of α2β1 enhanced α4β1-mediated extravasation of YBmα2 cells in the liver. Key words: Integrins; extracellular matrix substrates; cell adhesion; cell migration; extravasation.
∗ To
whom correspondence should be addressed. E-mail:
[email protected]
120
B. M. C. Chan et al.
INTRODUCTION
An understanding of how cells interact with each other and with biomaterial surfaces is important in many aspects of biotechnology. It impacts on the design of materials for biomedical devices ranging from vascular prostheses, wound dressing to scaffolds for tissue engineering [1– 5]. The biocompatibility of biomaterials is closely related to cell functions when the cell is in contact. Surface characteristics of materials including topography, chemistry or surface energy, play an essential part in cell adhesion on biomaterials. Thus, cell adhesion, spreading and migration represent the early phase of cell-biomaterial interactions. It is therefore important to understand the relationship between cell adhesion and physico-chemical properties of biomaterial surfaces [3, 4]. In addition, it is now clear that extracellular matrix (ECM) proteins such as collagen, fibronectin and laminin impact cell survival, growth and differentiation [6, 7]. Cell interactions are improved when the biomaterials are coated/ absorbed with ECM proteins [8– 10]. Members of the β1 integrin family have been grouped on the basis of association of unique α subunit with the common β1 integrin subunit. β1 Integrins function as the major receptors for ECM substrates and for certain members such as α4β1/ α4β7 and α5β1 integrins, they also mediate interactions with other cells [6, 11, 12]. Integrinmediated adhesion confers the ability of cells to remain stationary or undergo cell movement; optimal cell movement requires intermediate level of integrin adhesive function [13– 15]. An objective of this study is a better understanding of how β1 integrins coordinate in mediating adhesion and migration on ECM substrates. Fibronectin and vascular cell adhesion molecule-1 (VCAM-1) represent the major ligands for α4β1 integrin [16– 19]. On leukocytes, α4β1 exhibits low affinity for fibronectin and VCAM-1 [19, 20]. Integrin α4β1 plays a major role in the recirculation of leukocytes [21– 25]. The initiation of extravasation involves binding of L-selectin to its ligands, GlyCAM-1, CD34 and MAdCAM-1, which results in the rolling and tethering of leukocytes on the endothelial surface [26– 29]. As well, engagement of L-selectin promotes α4β1 integrin binding to fibronectin and β2 integrin function in binding ICAM-1 [23, 30– 33]. When expressed on activated T lymphocytes, α4β1 integrin mediates the initial rolling/ tethering stage of interaction with the endothelium [34]. Mapping by monoclonal antibodies (mAbs) reveals that the binding sites of α4β1 for fibronectin and VCAM-1 are not identical [35]. In addition, results from using activating β1-specific mAb suggest that the threshold level of stimulation for α4β1 binding VCAM-1 is lower than the threshold level of stimulation for α4β1 binding fibronectin [36]. This suggests that the interaction of α4β1 integrin with fibronectin and VCAM-1 can be differentially regulated. However, at present, it is not clear whether α4β1 function in the binding of VCAM-1 and fibronectin can be differentially regulated in a physiological setting. Functional linkage among cell surface molecules represents a key mode of regulation of integrin functions. On leukocytes, both α Lβ2(LFA-1) and αvβ3 integrins have been shown to downregulate α4β1 adhesive function for fibronectin
Integrin α2β1 stimulates α4β1 binding of VCAM-1
121
and VCAM-1, respectively [37, 38]. Thus, the modulation of α4β1 integrin function requires either a sustained activation of LFA-1 or crosslinking of αvβ3 integrins. The crosstalks among integrins are consistent with a relatively ordered involvement of different cell surface receptors in mediating the different stages of extravasation [23, 39]. Integrin α2β1 functions as a receptor for collagen and laminin [40– 42]. Expression of α2β1 re-organizes collagen matrix in gel contraction assay, mediates postextravasation cell movement within tissue parenchyma and modulates tumor activity in vivo [43– 46]. In addition, the expression of α2β1 induces the expression of α6β4 integrin and production of matrix metalloproteinase-1 [47, 48]. It has been shown that α2β1 integrin function may be regulated by other integrin members at the cell surface [49, 50]. Thus, ligation of αIIbβ3 results in a transdominant inhibition of α2β1 integrin function in Chinese hamster ovary (CHO) cells [49]. In comparison, binding of α5β1 to fibronectin stimulates α2β1 integrin function on monocytes [50]. In the present study, we demonstrate that α2β1 integrin expression can modulate α4β1 integrin function. Thus, expression of α2β1 differentially stimulates the adhesive/ migratory function of α4β1 on VCAM-1 but not fibronectin. As well, ligation of α2β1 further stimulates α4β1-dependent migration on VCAM1 and extravasation of cells in the liver.
MATERIALS AND METHODS
Reagents RPMI 1640, fetal bovine serum, L-glutamine, antibiotics, human fibronectin, rat collagen type I, mouse laminin 1 and Geneticin G418 were purchased from Life Technologies (Gaithersburg, MD). The mAb BMA2.1 specific for mouse α2β1 integrin was obtained from Chemicon International (Temecula, CA). The hamster mAbs HMa5-1 specific for rat α5β1 integrin, Ha1/ 29 specific for mouse and rat α2β1 integrin, and Ha2/ 5 specific for rat β1 integrin subunit were purchased from PharMingen (San Diego, CA). The blocking mouse mAb TA-2 specific for rat α4 integrins was kindly provided by Dr. T. Issekutz (Dalhousie University, Canada). Recombinant VCAM-1 was a kind gift from Dr. F. Takei (University of British Columbia, Canada). (2 ,7 )-bis (carboxyethyl) 5,6 carboxy fluorescein (BCECFAM) was obtained from Sigma Chemical Co. (St. Louis, MO). Cell culture and transfection The rat myeloma cell line YB2/0 was obtained from the American Type Culture Collection (Rockville, MD). The preparation of stable YB2/ 0 transfectant cells expressing the mouse α2/rat β1 integrin (YBmα2) has been described in detail in a previous study [51]. Briefly, YBmα2 cells were prepared by transfection of YB2/0 cells with cDNA of mouse α2 integrin subunit using the expression vector
122
B. M. C. Chan et al.
pFneo and Lipofectin reagent (Life Technologies). Both YBmα2 and the YB2/ 0 cells transfected with the expression vector pFneo alone (YBpF) were maintained in RPMI medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ ml penicillin, 100 µg/ ml streptomycin and 2 mg/ml geneticin G418. Flow cytometry and immunoprecipitation Flow cytometry for detection of integrin expression was carried out using a Becton Dickenson FACScan as previously described [46]. Briefly, cells were incubated with mAbs to specific integrins for 30 min at 4◦ C. Cells were then washed and incubated with the appropriate F(ab)2 fragments of fluorescein (FITC)-conjugated antibodies; all antibodies were used at pre-determined saturating concentrations. The levels of fluorescence from immunostaining with integrin-specific mAbs were compared with that from isotype-matched control mAbs. The levels of integrin expression were analysed using the CellQuest software. Results shown in Fig. 1 are representative of 5 experiments. Immunoprecipitation was carried out according to previously published procedures [15, 46]. Briefly, cells were surface labeled with 125I using the lactoperoxidase method and lysed in 0.5% non-ionic detergent NP-40 containing proteinase inhibitors: aprotinin (1 unit/ ml), leupeptin (0.1 M), and phenylmethylsulfonyl fluoride (2 mM). β1 Integrins were immunoprecipitated using specific mAbs. Immune complexes were isolated by using immunobeads conjugated with the appropriate secondary antibodies. Bound materials were washed, eluted with SDS-PAGE sample buffer and fractionated by SDS-PAGE (6% gel) under non-reducing conditions using 1 × 106 cell equivalent per sample. Results were then visualized by autoradiography. Adhesion and migration assays Cell adhesion assays were carried out as previously described [46, 52, 53]. Briefly, cells were labeled with fluorescent dye BCECF. A total of 5 × 104 labeled cells per well were added and allowed to adhere to wells coated with varying concentrations of ECM proteins for 45 min at 37◦ C. Non-adherent cells were removed by gentle washing with plain RPMI medium. Bound fluorescence was then measured using a Fluorescence Concentrator Analyser (IDEXX Lab., Westbrook, ME). Net fluorescence was obtained by subtracting the background adhesion to BSA-coated wells. Cell adhesion was then expressed as the number of bound cells per unit area based on the fluorescence from 5 × 104 labeled cells after a similar subtraction of background fluorescence. Each experimental condition was carried out in triplicates and each experiment was repeated a minimum of three times. Cell migration assays were carried out according to established procedures using the 48-well migration chamber (NeuroProbe, Cabin John, MD) [15, 54]. Briefly, polyvinylpyrrolidone-free polycarbonate filters of 8 µm pore size were coated on both sides with varying concentrations of ECM proteins in 0.1 M NaHCO3
Integrin α2β1 stimulates α4β1 binding of VCAM-1
123
Figure 1. Expression of α2β1 integrin on YB2/0 transfectant cells. The expression of α2β1, α4β1, α5β1 and β1 integrins on YB2/ 0 transfectant cells (YBpF and YBmα2) was measured by flow cytometry using anti-α2β1 mAbs BMA2.1 and Ha1/ 29, anti-α4β1 mAb TA-2, anti-α5β1 mAb HMα5-1 and anti-β1 integrin subunit (Ha2/ 5) (dark line). For control, cells were immunostained by the corresponding isotype-matched control mAbs (dotted line). YBmα2 cells were stained by both mAbs to α2β1; whereas no staining was detectable on YBpF cells. Both YBpF and YBmα2 were stained by TA-2 and Ha2/ 5 but not HMα5-1.
124
B. M. C. Chan et al.
overnight. The filters were washed with PBS, air-dried, and re-assembled with the bottom wells containing 30 µl of RPMI/1% BSA. Cells were then added to the top chamber at 1 × 105 cells in 50 µl per well and allowed to migrate into the ECM-coated filter for 6 h at 37◦ C. At the end of the migration assay, cells that did not migrate into the filter were mechanically scraped off. Cells that migrated into the ECM-coated filter were stained with Harris’ hematoxylin. Cell migration was determined as average of the number of cells counted from five random fields (400×, high-power field) from each of the quadruplicate wells by light microscopy using a gridded objective. Each experiment was repeated a minimum of three times. Intravital videomicroscopy Intravital videomicroscopy (IVM) was carried out as previously described [15, 46, 54]. Briefly, YBpF and YBmα2 cells were labeled with Fluoresbrite carboxylate microspheres (0.059– 0.067 µm; YG; Polysciences, Inc., Warrington, PA) and preincubated with the indicated concentration of mAb BMA2.1 or the isotypematched control mAb IgG1 for 1 h at 4◦ C. The cells were then washed with RPMI, and 3 × 105 cells were injected into the mesenteric vein of anesthetized nu/ nu mice (The Jackson Laboratory, Bar Harbor, ME). Cells in the liver were visualized using an inverted microscope (Zeiss Axiovert 135, Empix Imaging , Inc.; Mississauga, Canada) equipped with epifluorescence and a fiber optic light guide for transillumination. The microcirculation and location of labeled cells were viewed using a videocamera (Hamamatsu C2400, Empix Imaging) and recorded using Super VHS videotapes. The percentage of cells that had extravasated into the liver was determined 3h after cell injection by analysis of a minimum of 50 cells from each mouse using a minimum of 3 mice per experimental group.
RESULTS
β1 integrin expression on YBpF and YBm α2 cells The expression of α2β1 integrin on YB2/0 transfectant cells (YBmα2) had been characterized primarily by immunoprecipitation using antiserum against the cytoplasmic domain of α2 integrin subunit [51]. To quantitate the expression of β1 integrins on YB2/0 transfectant cells, flow cytometry was performed using integrin-specific mAbs (Fig. 1). MAb BMA2.1 specific for mouse α2β1 integrin immunostained YBmα2 cells; in comparison, YB2/0 cells transfected with only the expression vector pFneo (YBpF) were not immunostained by BMA2.1 (Figs 1A and B). Similar results were obtained using mAb Ha1/ 29, which binds both mouse and rat α2β1 (Figs 1C and D). Therefore, results show that YB2/0 cells do not constitutively express α2β1 integrin and that expression of α2β1 on YBmα2 cells was due to transfection of cDNA of mouse α2 integrin subunit. In addition, both YBmα2 and YBpF were immunostained at comparable levels of fluorescence
Integrin α2β1 stimulates α4β1 binding of VCAM-1
125
by α4β1-specific mAb TA-2 (Figs 1E and F), but were not immunostained by α5β1-specific mAb HMa5-1 (Figs 1G and H). Therefore, both YBmα2 and YBpF expressed comparable levels of α4β1 integrin and neither transfectant cell expressed α5β1 integrin. As expected, both YBmα2 and YBpF were immunostained by the pan-β1 integrin mAb Ha2/5 (Figs 1I and J). Results were confirmed by immunoprecipitation experiments using the corresponding mAbs (Fig. 2). Thus, using mAb BMA2.1, labeled bands corresponding to α2 and β1 integrin subunits were immunoprecipitated from YBmα2 but not YBpF cell lysate; no 125 I-labeled materials was immunoprecipitated using normal rat Ig. Results from these studies showed that transfection of cDNA of mouse α2 integrin subunit resulted in the expression of mouse α2/rat β1 heterodimers with no detectable effects on the expression of endogenous α4β1 integrins.
Figure 2. Expression of α2β1 integrins on YBpF and YBmα2 cells. YB2/ 0 cells transfected with the expression vector pFneo (YBpF) or pFneo containing the cDNA of mouse α2 integrin subunit (YBmα2) were 125 I-labeled at the cell surface. Cell lysates were immunoprecipitated by normal rat IgG (NRIgG) or rat mAb to mouse α2β1 (anti-α2β1) and analysed by SDS-PAGE under nonreducing conditions. 125 I-labeled mouse α2 and rat β1 integrin subunits were immunoprecipitated from YBmα2 but not the control YBpF transfectant cells.
126
B. M. C. Chan et al.
Expression of α2β1 integrin promotes YB2/0 transfectant cell adhesion to collagen, laminin, VCAM-1 but not fibronectin To determine the effect of α2β1 integrin expression, the levels of cell adhesion to collagen, laminin, as well as the substrate ligands for α4β1, fibronectin and VCAM-1, were compared between YBmα2 and YBpF cells in static adhesion assays (Fig. 3). YBmα2 cells bound collagen and laminin in a concentration dependent manner; the level of binding to collagen was higher than that to laminin (Figs 3A and B). In comparison, YBpF cells, which lacked α2β1 integrin expression, exhibited little, if any, adhesive function on collagen and laminin. Therefore, results are consistent with the previous study demonstrating α2β1 integrin function upon expression on YB2/0 transfectant cells [51]. Our results show that α4β1 integrin expression was not affected by the expression of α2β1 integrin; thus, α4β1 was expressed at comparable levels between YBmα2 and YBpF cells (Fig. 1). To assess α4β1 integrin function, the binding of YBmα2 and YBpF to fibronectin and VCAM-1 were compared. As shown in Fig. 3C, YBmα2 and YBpF were comparable in binding fibronectin. In contrast, YBmα2 exhibited significantly greater binding activity to VCAM-1 than YBpF (p < 0.005)
Figure 3. Adhesion properties of YBmα2 cells on collagen, laminin, fibronectin and VCAM-1 substrates. YB2/0 transfectant cells (YBpF and YBmα2) were characterized for their adhesion to wells coated at varying concentrations of collagen (A), laminin (B), fibronectin (C) and VCAM-1 (D). YBmα2 cells expressing α2β1 integrin adhered to collagen and laminin substrates at levels higher than YBpF cells (A and B). YBpF and YBmα2 cells exhibited comparable levels of adhesion to fibronectin substrate (C). YBmα2 cells expressing α2β1 integrin bound VCAM-1 at higher levels than YBpF cells lacking the expression of α2β1 (D). Mean values and SDs (bars) are shown.
Integrin α2β1 stimulates α4β1 binding of VCAM-1
127
(Fig. 3D). Thus, at 10 µg/ ml of VCAM-1, the level of YBmα2 adhesion was greater than 300 cells/ mm2 ; in comparison, the adhesion of YBpF was less than 100 cells/ mm2 at all the concentrations of VCAM-1 tested. Therefore, results show that α4β1 integrin function is enhanced upon expression of α2β1; the stimulated α4β1 function is specific in that only its binding to VCAM-1 but not fibronectin is enhanced. To confirm that the enhanced adhesive function seen on collagen and laminin was mediated by α2β1 integrin, cell adhesion assays were carried out in the presence of blocking α2β1-specific mAb. Thus, cell adhesion to collagen and laminin (5 µg/ ml) were assessed in the presence of varying concentrations of blocking α2β1-specific mAb BMA2.1 (0.01 to 10 µg/ ml). As shown in Fig. 4, BMA2.1 inhibited adhesion of YBmα2 cells to both collagen and laminin; in comparison, isotype-matched control mAb had no significant effect on the levels of adhesion. The concentration of BMA2.1 required for inhibition of collagen binding was higher than that for laminin; at greater than 3 µg/ ml BMA2.1, adhesion to both ECM substrates were reduced to essentially basal levels. Using a distinct mAb to α2β1(Ha1/ 29) at 10 µg/ ml yielded similar levels of inhibition. Results have therefore confirmed and extended previous studies showing that α2β1 integrin promotes YB2/0 transfectant cell adhesion to collagen and laminin substrates.
Figure 4. Involvement of α2β1 integrin in YBmα2 transfectant cell adhesion to collagen and laminin substrates. YBmα2 cell adhesion to collagen (A) and laminin (B) at 5 µg/ ml were characterized in the presence of varying concentrations of blocking α2β1-specific rat mAb (BMA2.1) or isotype-matched controlled mAb (Control mAb). For comparison, adhesion was also characterized in the presence of blocking α2β1-specific hamster mAb (Ha1/ 29) at 10 µg/ ml. Mean values and SDs (bars) are shown.
128
B. M. C. Chan et al.
Expression of α2β1 integrin stimulates YB2/0 transfectant cell migration on VCAM-1 but not fibronectin We and others have demonstrated that the ability of cells to adhere to ECM substrates can modulate cell movement; thus, optimal cell migration occurs at intermediate levels of ECM binding [13– 15]. However, our results demonstrated that α2β1 expression stimulated α4β1-mediated adhesion to VCAM-1 but not fibronectin (Fig. 4). To determine how stimulated α4β1 integrin function impacted cell movement, the migratory properties of YBmα2 and YBpF were compared in migration assays with filters coated with varying concentrations of fibronectin, fibronectin fragment-40 (FN-40) or VCAM-1. As shown in Figs 5A to D, YBmα2 and YBpF cells exhibited similar extents of migration on fibronectin, as well as
Figure 5. Comparison of migratory function of YBpF and YBmα2 cells on fibronectin, FN-40 and VCAM-1 substrates. Migration of YBpF (A, C and E) and YBmα2 cells (B, D and F) was determined in the presence of control mAb IgG1 (10 µg/ml) on polycarbonate filters coated at varying concentrations of fibronectin (A and B), fibronectin fragments containing the recognition site EILDV for α4β1(FN-40) (C and D) or soluble VCAM-1 (E and F). The role of α4β1 was determined by carrying out migration assays in the presence of mAb TA-2 at 10 µg/ml. Values were enumerated from the mean (±SD) of cell counts from five random fields (400×) from each of the quadruplicate wells.
Integrin α2β1 stimulates α4β1 binding of VCAM-1
129
on fibronectin fragment FN-40, which contains the EILDV binding site for α4β1 [16– 18]. On VCAM-1 substrate, YBmα2 cells exhibited cell movement in a concentration dependent manner. In contrast, YBpF cells that bound VCAM-1 at only basal levels also had low levels of migration on VCAM-1 (Figs 5E and F). The presence of mAb TA-2 (10 µg/ ml) inhibited migration of YB2/0 transfectant cells on fibronectin and VCAM-1, thus confirming the role of α4β1 integrin in the observed cell movement. Ligation of α2β1 integrin further stimulates α4β1-mediated migration on VCAM-1 Results from the present study suggest a functional linkage between α2β1 and α4β1 integrins. To determine whether the stimulatory effect of α2β1 can be modulated, α4β1-mediated cell movement was assessed upon ligation of α2β1 using mAb BMA2.1. Thus, migration assays were carried out in the presence of BMA2.1. The presence of BMA2.1 (2 µg/ ml) had no significant effect on YBmα2 migration on fibronectin (0.2 to 20 µg/ ml) and FN-40 (1.87 to 30 µg/ ml); no major difference was observed in the levels of cell movement when compared to using isotype-matched control mAb IgG1 (Figs 6A and B). In contrast, ligation of α2β1 by BMA2.1 further stimulated YBmα2 cell movement on VCAM-1 (Fig. 6C). At 10 µg/ ml VCAM-1, ligation of α2β1 resulted in an approximately 3-fold increase in cell movement in comparison to using isotype-matched control mAb. To confirm that the stimulated cell movement was α4β1-mediated, migration assay was carried out in the presence of mAb TA-2. As shown in Fig. 6D, α4β1-specific mAb (TA-2) inhibited the constitutive migration of YBmα2 cells (IgG1). Ligation of α2β1 stimulated cell movement (BMA2.1) in comparison to using control mAb (IgG1). Inhibition of α4β1 integrin by TA-2 abolished BMA2.1-stimulated cell movement (BMA2.1 + TA-2) to levels approximating that seen in using TA-2 to inhibited the migration of unstimulated cells (TA-2). Ligation of α2β1 integrin enhances α4β1 integrin-mediated extravasation in vivo To determine the in vivo significance of α2β1 ligation in α4β1 integrin function, intravital videomicroscopy (IVM) was performed for a real time assessment of the extravasation of YB2/0 transfectant cells. Thus, fluorescence-labeled YBpF and YBmα2 cells were preincubated with BMA2.1 or control mAb IgG1, injected via the mesenteric vein, and the extents of their extravasation in the liver were monitored by IVM. The percentages of cells that had undergone extravasation were determined 3h after cell injection. YBmα2 cells exhibited higher constitutive extravasation than YBpF (36±3% vs. 28 ± 4%) (Figs 7A and B). Preincubation with BMA2.1 or control mAb IgG1 at concentrations ranging from 0.15 to 20 µg/ ml had no significant effects on YBpF extravasation (Fig. 7A). In contrast, extravasation of YBmα2 cells was enhanced after treatment with BMA2.1 at concentrations greater than 0.3 µg/ ml; thus, treatment with BMA2.1 increased the extravasation by approximately 50% (Fig. 7B). MAb TA-2 was used to assess the role of α4β1
130
B. M. C. Chan et al.
Figure 6. Effect of ligation of α2β1 integrin on YBmα2 cell movement on fibronectin, FN-40 and VCAM-1 substrates. Migration of YBmα2 cells on polycarbonate filters coated with fibronectin (A), FN-40 (B) or VCAM-1 (C) at varying concentrations were assayed in the presence of isotype-matched control mAb IgG1 (Q) or BMA2.1 (") at 2 µg/ml. The role of α4β1 integrin in YBmα2 cell migration on filters coated with VCAM-1 at 1.5 µg/ml was determined by using mAb TA-2 at 10 µg/ml (D). YBmα2 cells were either treated with control mAb IgG1 or α2β1-specific mAb BMA2.1 at 2 µg/ml. Values were enumerated from the mean (±SD) of cell counts from five random fields (400×) from each of the quadruplicate wells.
integrin in BMA2.1-stimulated extravasation. As shown in Fig. 7C, the levels of extravasation were similar between using YBmα2 cells alone or after treatment with control mAb. Treatment with TA-2 had no effect on the constitutive level of extravasation. In comparison, treatment of BMA2.1 significantly stimulated extravasation (p < 0.001), and the stimulated extravasation was reduced to basal levels when in the presence of mAb TA-2 (BMA2.1 + TA-2). DISCUSSION
Crosstalk among integrins impacts diverse cellular functions including cell spreading, adhesion, migration, phagocytosis, and matrix metalloproteinase synthesis [37, 38, 49, 55– 57]. These provide a basis by which distinct integrins coordinate their involvement during the different stages of biological events such as that seen in leukocyte recirculation. Ligation of LFA-1 or αvβ3 integrins has been shown to inhibit the adhesive function of α4β1 integrin for fibronectin and VCAM-1, respectively [37, 38]. In the present study, we demonstrated that α2β1 integrin stim-
Integrin α2β1 stimulates α4β1 binding of VCAM-1
131
Figure 7. Extravasation of YB2/ 0 transfectant cells in the liver. Intravital videomicroscopy was carried out to measure the extravasation of YB2/ 0 transfectant cells after treatment of YBpF (A) or YBmα2 cells (B) with either control mAb IgG1 (Q) or BMA2.1 (") at varying concentrations. Each value represents mean percentage (±SD) of cells that extravasated at 3 h after injection of 3×105 cells into nu/ nu mice via a mesenteric vein. A minimum of 50 cells were analysed from each of the 3 mice used per experimental group. The role of α4β1 in the extravasation of YBmα2 cells was determined by using mAb TA-2 at 10 µg/ml (C). YBmα2 cells were either injected without prior treatment (Cells alone) or after preincubation with control IgG1 or BMA2.1 at 3 µg/ ml. The number of mice per experimental group are shown (n). Values are mean percentage (±SD) of cells that extravasated at 3 h after cell injection.
132
B. M. C. Chan et al.
ulated α4β1 integrin-dependent adhesion and migration on VCAM-1 but not on fibronectin substrate. The stimulation of α4β1 function was not due to changes in the level of its expression. The increase in the binding of YBmα2 cells to VCAM-1 was associated with enhanced extravasation in the liver. On resting leukocytes, integrins at the cell surface are either inactive or only partly active, and their ligand binding function requires stimulation [20, 58]. Thus, early studies have demonstrated that T lymphocytes exhibit β1 integrindependent adhesion to ECM substrates upon antigen presentation or stimulation by antibody to CD3 [20, 59]. The inside-out signaling of β1 integrin function is consistent with studies showing that Ras activation can modulate the functional state of β1 integrins at the cell surface [60– 62]. As well, α2β1 integrin has been shown to signal Ras activation [63]. In the present study, expression of α2β1 enhanced the adhesive/ migratory activity of α4β1 on VCAM-1 substrate. The α2β1 at the cell surface were at least partly active as shown by their ability to mediate YB2/0 transfectant cell adhesion/ migration on both collagen and laminin substrates. It therefore appears that α4β1 integrin function is enhanced by the constitutive signaling properties of active α2β1; stimulation of α4β1 function is further enhanced upon ligation with mAbs to α2β1. In one study, αvβ3 integrin occupancy on pro-T cell line FTF1 has been shown to inhibit binding of α4β1 to VCAM-1, which in turn, increases α4β1-mediated migration on VCAM-1 substrate [38]. In comparison, our results show that YBmα2 cell migration on VCAM-1 is associated with an increase in the adhesive function of α4β1. This may reflect distinct signaling properties between α2β1 and αvβ3 integrins. Alternatively, this may be related to differences in the adhesive strength of α4β1 integrins between FTF1 and YBmα2 cells. We and others have demonstrated a biphasic relationship between the adhesive strength and cell movement on ECM substrates; optimal migration occurs at intermediate levels of interaction between β1 integrins and ECM substrates [13– 15]. It is likely that the adhesive strength of α4β1 on YBmα2 is suboptimal; thus, increases in the adhesive function of α4β1 by expression and/ or ligation of α2β1 enhances cell movement. In comparison, α4β1 integrin on FTF1 cells may bind VCAM-1 at levels that are beyond optimal interaction and thus, stimulation of α4β1-mediated cell movement requires a reduction in its adhesive function. On YBmα2 cells, α2β1 expression affects α4β1 interaction with VCAM-1 but not fibronectin. α4β1 Integrin binds the Ig domains 1 and 4 of VCAM-1; whereas, it binds the EILDV motif in the CS-1 region of fibronectin [16– 18, 64, 65]. Mapping of α4β1 by a panel of mAbs indicates that the binding sites for VCAM-1 and fibronectin overlap, but they are not identical [35]. The ligand affinity of integrins may be modulated as a result of conformational changes [52, 66, 67]. Therefore, signaling from α2β1 may induce α4β1 to undergo conformational changes that differentially affect the binding sites for VCAM-1 without affecting the binding sites for fibronectin. Our results are also consistent with studies suggesting that the threshold for stimulation of α4β1 to bind VCAM-1 is lower than the
Integrin α2β1 stimulates α4β1 binding of VCAM-1
133
threshold level of its stimulation for fibronectin [36]. Thus, stimulation of α4β1 binding of VCAM-1 precedes its binding of fibronectin. In one study, ligation of LFA-1 integrin has been shown to inhibit α4β1 binding of fibronectin; however, this study has not determined whether the binding of VCAM-1 by α4β1 is also inhibited [37]. In addition, members of tetraspanins have been shown to associate with β1 integrins [68, 69]. Modulation of tetraspanins impact integrin-mediated adhesion and migration [69– 71]. Integrin α4β1 has been shown to associate with tetraspanins, CD53, CD63, CD81 and CD82 [69]. Results from the present study cannot exclude a role for tetraspanins to mediate the stimulation of α4β1 integrin function by α2β1. Using intravital videomicroscopy, we have previously demonstrated that α2β1 integrin has a major role in modulating the postextravasation migration of human rhabdomyosarcoma RD transfectant cells [46]. Thus, after extravasation into the liver, RD transfectant cells expressing α2β1 integrin appear to wrap around the sinusoid and fail to migrate through the liver parenchyma. However, expression of α2β1 had no effect on the extravasation of RD cells. In comparison, the present study reveals a role of α2β1 in the extravasation of YBmα2 cells by stimulating α4β1 to interact with VCAM-1. Therefore, the effect of α2β1 on extravasation in the liver is cell type dependent. It is likely that on YB2/0 cells, α4β1 are only partly active; whereas α4β1 on RD cells are fully active and expression/ ligation of α2β1 does not further stimulate α4β1 function. Upon expression of α2β1, YB2/0 transfectant cells exhibited enhanced extravasation into the liver; ligation of α2β1 further enhanced extravasation of YBmα2 cells. It is well established that LFA-1 collaborates with α4β1 in leukocyte extravasation [21, 34, 37]. In addition, ligation of LFA-1 downregulates α4β1 adhesive function [37]. More recently, it has been found that α4β1 binding of VCAM-1 stimulates LFA-1 function [72]. Results from flow cytometry indicated that LFA-1 integrin is also expressed on YB2/ 0 cells (unpublished data). Therefore, α4β1 integrin is not likely the sole receptor responsible for the extravasation of YBmα2. This may explain our observation that blocking mAb to α4β1(TA-2) had no major effect on YBmα2 extravasation. Without ligation of α2β1 to stimulate α4β1 function, LFA-1 may have a greater role than α4β1 in YBmα2 extravasation in the liver. However, the role of α4β1 in mediating extravasation increases upon ligation of α2β1. Thus, binding of BMA2.1 enhanced α4β1-dependent extravasation of YBmα2 cells. It is well established that α2β1 expression is induced upon activation of T lymphocytes; on resting T lymphocytes, its expression is low, if any [73– 75]. Results from the present study suggest that α2β1 may alter the relative contribution between α4β1 and LFA-1 in leukocyte recirculation. This is consistent with studies demonstrating differences between resting and activated T lymphocytes in utilizing α4β1 integrin for extravasation [34]. Collagen and laminin represent the major interstitial matrix proteins. It is therefore expected that α2β1 integrin impacts the ability of cells to navigate and distribute within the stroma. We have previously demonstrated that depending on
134
B. M. C. Chan et al.
its adhesive strength and/ or the cell type involved, α2β1 can modulate the ability of cells to undergo migration within tissue parenchyma [15, 46]. Thus, expression of α2β1 inhibits RD but enhances human erythroleukemia K562 transfectant cell migration within the liver parenchyma. It has recently been suggested that α2β1 integrin has an important role in the formation of immunological synapse and coordinate the migration and interaction of T lymphocytes with antigen presentation cells [76]. In addition, α2β1 may impact the process of intravasation by potentiating α4β1 integrin to interact with VCAM-1 as the cell exits into the blood system. Results from the present study showing a potential role of α2β1 in cell recirculation may be useful in the design of vascular prosthesis. Acknowledgement This work was supported by the Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council of Canada and the Cancer Research Society.
REFERENCES 1. J. A. Hubbell, Biotechnology 13, 565 (1995). 2. K. James and J. Kohn, MRS Bull. 21, 22 (1996). 3. C. Cochrane, M. G. Rippon, A. Rogers, R. Walmsley, D. Knottenbelt and P. Bowler, Biomaterials 20, 1237 (1999). 4. E. Cenni, D. Granchi, G. Ciapetti, E. Verri, D. Cavedagna, S. Gamberini, M. Cervellati, A. DiLeo and A. Pizzoferrato, Biomaterials 18, 489 (1997). 5. R. B. O’Neal, J. J. Sauk and M. J. Somerman, J. Oral Implantol. 18, 243 (1992). 6. R. O. Hynes, Cell 69, 11 (1992). 7. J. L. Guan and H. C. Chen, Int. Rev. Cytol. 168, 81 (1996). 8. N. Faucheux, C. Correze, B. Haye and M. D. Nagel, Biomaterials 22, 2993 (2001). 9. M. Shen and T. A. Horbett, J. Biomed. Mater. Res. 57, 336 (2001). 10. G. Rainaldi, P. Filippini, A. Ferrante, P. L. Indovina and M. T. Santini, J. Biomed. Mater. Res. 55, 104 (2001). 11. K. M. Yamada and S. Miyamoto, Curr. Opin. Cell Biol. 7, 681 (1995). 12. M. Ruppert, S. Aigner, Hubbe, H. Yagita and P. Altevogt, J. Cell Biol. 131, 1881 (1995). 13. P. A. DiMilla, J. A. Stone, J. A. Quinn, S. M. Albelda and D. A. Lauffenburger, J. Cell Biol. 122, 729 (1993). 14. S. P. Palecek, J. C. Loftus, M. H. Ginsberg, D. A. Lauffenburger and A. F. Horwitz, Nature 385, 537 (1997). 15. W.-C. Ho, C. Heinemann, D. Hangan, S. Uniyal, V. L. Morris and B. M. C. Chan, Mol. Biol. Cell 8, 1863 (1997). 16. J.-L. Guan and R. O. Hynes, Cell 60, 53 (1990). 17. E. A. Wayner, A. Garcia-Pardo, M. J. Humphries, J. A. McDonald and W. G. Carter, J. Cell Biol. 109, 1321 (1989). 18. A. P. Mould, A. Komorya, K. M. Yamada and M. J. Humphries, J. Biol. Chem. 266, 3579 (1991). 19. M. J. Elices, L. Osborn, Y. Takada, C. Crouse, S. Luhowskyj, M. E. Hemler and R. R. Lobb, Cell 60, 577 (1990). 20. Y. Shimizu, G. A. Van Secenter, K. J. Horgan and S. Shaw, Nature 345, 250 (1990).
Integrin α2β1 stimulates α4β1 binding of VCAM-1
135
21. R. Alon, P. D. Kassner, M. W. Carr, E. B. Finger, M. E. Hemler and T. A. Springer, J. Cell Biol. 128, 1243 (1995). 22. P. D. Kassner, R. Alon, T. A. Springer and M. E. Hemler, Mol. Biol. Cell 6, 661 (1995). 23. T. A. Springer, Cell 76, 301 (1994). 24. C. Weber, R. Alon, B. Moser and T.A. Springer, J. Cell Biol. 134, 1063 (1996). 25. F. W. Luscinskas, G. S. Kansas, P. Ding, P. Pizcueta, B. E. Schleiffenbaum, T. F. Tedder and M. A. Gimbrone, J. Cell Biol. 125, 1417 (1994). 26. M. B. Lawrence and T. A. Springer, Cell 65, 859 (1991). 27. L. A. Lasky, M. S. Singer, D. Dpwbenko, Y. Imai, E. J. Henzel, C. Fennie, N. Gillett, S. R. Watson and S. D. Rosen, Cell 69, 927 (1992). 28. S. Baumhueter, M. S. Singer, W. Henzel, S. Hemmerich, M. Renz, S. D. Rosen and L. A. Lasky, Science 262, 436 (1993). 29. E. L. Berg, L. M. McEvoy, C. Berlin, R. F. Bargatze and E. C. Butcher, Nature 366, 695 (1993). 30. I. B. Mazo, J. C. Gutierrez-Ramos, P. S. Frenette, R. O. Hynes, D. D. Wagner and U. H. von Andrian, J. Exp. Med. 188, 465 (1998). 31. P. A. Giblin, S. T. Hwang, T. R. Katsumoto and S. D. Rosen, J. Immunol. 159, 3498 (1997). 32. R. Alon, R. C. Fuhlbrigge, E. B. Finger and T. A. Springer, J. Cell Biol. 135, 849 (1996). 33. A. Hafezi-Moghadam, K. L. Thomas, A. J. Prorock, Y. Huo and K. Ley, J. Exp. Med. 193, 863 (2001). 34. C. Berlin, R. F. Bargatze, J. J. Campbell, U. H. von Andrian, M. C. Szabo, S. R. Hassien, R. D. Nelson, E. L. Berg, S. L. Erlandsen and E. C. Butcher, Cell 80, 413 (1995). 35. R. Pulido, M. J. Elices, M. R. Campanero, L. Osborn, S. Schiffer, A. Garcia-Pardo, R. Lobb, M. E. Hemler and F. Sanchez-Madrid, J. Biol. Chem. 266, 10241 (1991). 36. A. Masumoto and M. E. Hemler, J. Biol. Chem. 268, 228 (1993). 37. J. C. Porter and N. Hogg, J. Cell Biol. 138, 1437 (1997). 38. B. A. Imhof, D. Weerasinghe, E. J. Brown, F. P. Lindberg, P. Hammel, L. Piali, M. Dessing and R. Gisler, Eur. J. Immunol. 27, 3242 (1997). 39. Y. Shimizu, W. Newman, T. V. Gopal, K. J. Horgan, N. Graber, L. D. Beall, G. A. van Seventer and S. Shaw, J. Cell Biol. 113, 1203 (1991). 40. Y. Takada and M. E. Hemler, J. Cell Biol. 109, 397 (1989). 41. M. J. Elices and M. E. Hemler, Proc. Natl. Acad. Sci. USA 86, 9906 (1989). 42. L. R. Languino, K. R. Gehlsen, E. Wayner, W. G. Carter, E. Engvall and E. Ruoslahti, J. Cell Biol. 109, 2455 (1989). 43. B. M. C. Chan, P. D. Kassner, J. A. Schiro, H. R. Byers, T. S. Kupper and M. E. Hemler, Cell 68, 1051 (1992). 44. J. Schiro, B. M. C. Chan, W. Roswit, P. D. Kassner, A. Pentland, M. E. Hemler, A. Eisen and T. Kupper, Cell 67, 403 (1991). 45. B. M. C. Chan, N. Matsuura, Y. Takada, B. R. Zetter and M. E. Hemler, Science 251, 1600 (1991). 46. D. Hangan, S. Uniyal, V. L. Morris, I. C. MacDonald, C. von Ballestrem, T. Chau, E. E. Schmidt, A. F. Chambers, A. C. Groom and B. M. C. Chan, Cancer Res. 56, 3142 (1996). 47. H. Sun, S. Santoro and M. M. Zutter, Cancer Res. 58, 2224 (1998). 48. O. Langholz, D. Rockel, C. Mauch, E. Kozlowska, H. Bank, T. Krieg and B. Eckes, J. Cell Biol. 131, 1903 (1995). 49. F. Diaz-Gonzalez, J. Forsyth, B. Steiner and M. H. Ginsberg, Mol. Biol. Cell 7, 1939 (1996). 50. R. Pacifici, J. Roman, R. Kimble, R. Civitelli, C. M. Brownfield and B. Cinzia, J. Immunol. 153, 2222 (1994). 51. J. Edelman, B. M. C. Chan, S. Uniyal, H. Onodera, D. Z. Wang, N. F. St. John, L. Dmjanovich, D. B. Latzer, R. W. Finberg and J. M. Bergelson, Cell Adh. & Commun. 2, 131 (1994). 52. B. M. C. Chan and M. E. Hemler, J. Cell Biol. 120, 537 (1993). 53. S. Kawaguchi and M. E. Hemler, J. Biol. Chem. 268, 16279 (1993).
136
B. M. C. Chan et al.
54. D. Hangan, V. L. Morris, L. Boeters, C. von Ballestrem, S. Uniyal and B. M. C. Chan, Cancer Res. 57, 3812 (1997). 55. S. D. Blystone, F. P. Lindberg, S. E. LaFlamme and E. J. Brown, J. Cell. Biol. 130, 745 (1995). 56. P. Huhtala, M. J. Humphries, J. B. McCarthy, P. M. Tremble, Z. Werb and C. H. Damsky, J. Cell Biol. 129, 867 (1995). 57. S. Z. Domanico, A. J. Pelletier, W. L. Havran and V. Quaranta, Mol. Biol. Cell 8, 2253 (1997). 58. M. L. Dustin and T. A. Springer, Nature 341, 619 (1989). 59. B. M. C. Chan, J. G. P. Wong, A. Rao and M. E. Hemler, J. Immunol. 147, 398 (1991). 60. P. E. Hughes, M. W. Renshaw, M. Pfaff, J. Forsyth, V. M. Keivens, M. A. Schwartz and M. H. Ginsberg, Cell 88, 521 (1997). 61. P. J. Keely, E. V. Rusyn, A. D. Cox and L. V. Parise, J. Cell Biol. 145, 1077 (1999). 62. Z. Zhang, K. Vuori, H.-G. Wang, J. C. Reed and E. Ruoslahti, Cell 85, 61 (1996). 63. C. Kapron-Bras, L. Fitz-Gibbon, P. Jeevaratnam, J. Wilkins and S. Dedhar, J. Biol. Chem. 268, 20701 (1993). 64. L. Osborn, C. Vassallo and C. D. Benjamin, J. Exp. Med. 176, 99 (176). 65. R. H. Vonderheide and T. A. Springer, J. Exp. Med. 175, 1433 (1992). 66. N. L. Kovach, T. M. Carlos, E. Yee and J. M. Harlan, J. Cell Biol. 116, 499 (1992). 67. A. G. Arroyo, A. Garcia-Pardo and F. Sanchez-Madrid, J. Biol. Chem. 268, 9863 (1993). 68. F. Berditchevski, M. M. Zutter and M. E. Hemler, Mol. Biol. Cell 7, 193 (1996). 69. B. A. Mannion, F. Berditchevski, S.-K. Kraeft, L. B. Chen and M. E. Hemler, J. Immunol. 157, 2039 (1996). 70. H. Hasegawa, T. Nomura, K. Kishimoto, K. Yanagisawa and S. Fujiya, J. Immunol. 161, 3087 (1998). 71. R. L. Yauch, F. Berditchevski, M. B. Harler, J. Reichner and M. E. Hemler, Mol. Biol. Cell 9, 2751 (1998). 72. D. M. Rose, V. Grabovsky, R. Alon and M. H. Ginsberg, J. Immunol. 167, 2824 (2001). 73. M. E. Hemler, D. Glass, J. S. Coblyn and J. G. Jacobson, J. Clin. Invest. 78, 696 (1986). 74. M. E. Hemler and J. G. Jacobson, J. Immunol. 138, 2941 (1987). 75. M. E. Hemler, C. Huang and L. Schwarz, J. Biol. Chem. 262, 3300 (1987). 76. M. Dustin, P. M. Allen and A. S. Shaw, Trend in Immunol. 22, 192 (2001).
Interactions of corneal epithelial cells and surfaces modified with cell adhesion peptide combinations L. AUCOIN 1 , C. M. GRIFFITH 2 , G. PLEIZIER 3 , Y. DESLANDES 3 and H. SHEARDOWN 1,4,∗ 1 Department
of Chemical Engineering, McMaster University, 1280 Main St. W., Hamilton ON, L8S 4L7, Canada 2 University of Ottawa Eye Institute, Ottawa ON, Canada 3 ICPET, National Research Council of Canada, Ottawa ON, Canada 4 Department of Pathology and Molecular Medicine, McMaster University, 1280 Main St. W., Hamilton ON, L8S 4L7, Canada Received 17 September 2001; accepted 27 February 2002 Abstract—In order to facilitate the adhesion of corneal epithelial cells to a poly dimethyl siloxane (PDMS) substrate ultimately for the development of a synthetic keratoprosthesis, PDMS surfaces were modified by covalent attachment of combinations of cell adhesion and synergistic peptides derived from laminin and fibronectin. Peptides studied included YIGSR and its synergistic peptide PDSGR from laminin and the fibronectin derived RGDS and PHSRN. Surfaces were modified with combinations of peptides determined by an experimental design. Peptide surface densities, measured using 125-I labeled tyrosine containing analogs, were on the order of pmol/ cm2 . Surface density varied as a linear function of peptide concentration in the reaction solution, and was different for the different peptides examined. The lowest surface density at all solution fractions was obtained with GYRGDS, while the highest density was consistently obtained with GYPDSGR. These results provide evidence that the surfaces were modified with multiple peptides. Water contact angles and XPS results provided additional evidence for differences in the chemical composition of the various surfaces. Significant differences in the adhesion of human corneal epithelial cells to the modified surfaces were noted. Statistical analysis of the experimental adhesion results suggested that solution concentration YIGSR, RGDS, and PHSRN as well as the interaction effect of YIGSR and PDSGR had a significant effect on cell interactions. Modification with multiple peptides resulted in greater adhesion than modification with single peptides only. Surface modification with a control peptide PPSRN in place of PHSRN resulted in a decrease in cell adhesion in virtually all cases. These results suggest that surface modification with appropriate combinations of cell adhesion peptides and synergistic peptides may result in improved cell surface interactions. Key words: Corneal epithelial cell; Cell adhesion; YIGSR; RGDS; PDSGR; PHSRN; surface modification; poly (dimethyl siloxane).
∗ To
whom correspondence should be addressed.
138
L. Aucoin et al.
INTRODUCTION
An implantable keratoprosthesis could be used to restore sight in cornea blind patients for whom other treatment options are likely to fail or for whom donor tissue is not available. Due to the unique biology and physiology of the cornea, such a device should interact with corneal cells of different types and perform multiple functions including transmission and refraction of light and protection of the inner structures of the eye. Extrusion of the device as a result of internal globe pressure and failure to promote adequate wound healing at the implant margin is a common cause of device failure [1]. Since the anterior surface of the cornea is covered by a stratified, squamous, nonkeratinized epithelium, the anterior surface of a prosthetic device should support fundamental corneal epithelial cell functions, including adhesion, proliferation, migration and differentiation [2, 3]. This layer would play an important role in the inhibition of bacterial infection and appropriate coverage would be a significant factor in the inhibition of epithelial downgrowth [4– 8]. Promising in vitro corneal epithelial adhesion, and outgrowth results have been obtained following surface modification with cell adhesion peptides including RGDS and YIGSR [9– 12] and with cell adhesion molecules [9, 13– 16]. It has been shown that the well known RGD [17] sequence and the sequence PHSRN [18] are present on adjacent loops of two FIII modules in fibronectin and bind synergistically to numerous integrins containing various α and β subunit combinations. While PHSRN has no apparent adhesive activity on its own, its presence can lead to, for example, an approximately 100-fold increase in cell adhesion to fibronectin via α5 β1 [19]. Similarly, synergistic peptides including PDSGR [20] have been found for the laminin binding peptide YIGSR. Unlike PHSRN however, PDSGR does have reported cell adhesive activity. Kao et al. [21] studied the response of macrophages to surfaces modified with oligopeptides based on the primary and tertiary structure of human and plasma fibronectin including peptides incorporating RGD and PHSRN separated by a glycine spacer. The results demonstrated the importance of the synergy between the RGD and PHSRN and the relative orientation of the two peptides in supporting macrophage fusion to form foreign body giant cells. Heath et al. [22] showed that bone cells interacting with a PHSRN-RGD sequence were more active than those on RGD-PHSRN, RGD or PHSRN alone. Tong and Schoichet [23] studied the immobilization of the laminin derived cell adhesive peptide YIGSR and outgrowth stimulating peptide IKVAV alone and in combination on neural cell response. They found that surfaces modified with a 1 : 1 mixture of YIGSR and IKVAV peptides promoted a greater number of neurites per cell and longer neurites than surface modified with one peptide type. The results of these studies demonstrate that modification with combinations of cell adhesion peptides as well as cell adhesion and synergistic peptides may promote cell — biomaterials interactions to a greater extent than modification with single peptides. In the cornea, the epithelial basement membrane is composed primarily of laminin and Type IV collagen, with fibronectin playing a role during wound healing [24].
Corneal epithelial cells and cell adhesion peptide combinations surfaces
139
Therefore, we hypothesize that modification of potential keratoprosthetic surfaces with combinations of laminin derived and fibronectin derived cell adhesive and synergistic peptides will enhance corneal epithelial cell adhesion and promote cellular interactions. Polydimethyl siloxane (PDMS) was selected as a substrate for these studies based on its ophthalmic compatibility, transparency, oxygen permeability and previous application in keratoprosthetic applications [25, 26]. Surfaces were modified with combinations of YIGSR, RGDS, PDSGR and PHSRN and the adhesion of corneal epithelial cells to these modified surfaces were studied.
MATERIALS AND METHODS
Surface preparation PDMS (Dow Corning, Toronto Canada) membranes were prepared according to directions provided by the manufacturer. Briefly, the monomer and curing agent were mixed in a 10 : 1 (m : m) ratio and placed under vacuum for approximately 1 h to remove entrapped air. The mixture was then placed in a plastic petri dish and allowed to cure overnight. Surfaces were hydroxylated for subsequent modification by microwave frequency plasma polymerization of allyl alcohol as described in Wickson and Brash [27]. The reactor was evacuated to a pressure of 40 µm Hg and the argon flow started at a rate of 235 sccm (standard cubic centimeters per minute). At a pressure of approximately 60 µm Hg, the glow discharge was initiated at a power of 20 W for 5 min. The allyl alcohol flow was subsequently initiated at a flow rate of 0.9 sccm for a 10 min period. Since oxygen-containing plasmas can give rise to carbonyl groups during plasma polymerization regardless of the initial monomer structure [28], surfaces were treated overnight in aqueous (0.26 M) sodium borohydride (NaBH4 ) to convert carbonyl to hydroxyl groups [29]. Following this treatment, surfaces were rinsed thoroughly with water, dried under vacuum overnight at room temperature, and stored in covered containers until use. Combinations of the cell adhesion and synergistic peptides YIGSR, RGDS, PHSRN and PDSGR were attached using tresyl chloride chemistry as in our previous study [12]. The combinations were selected using the statistical design software CARD. This design was selected as permits the examination of a wide range of variables with relatively few experimental runs, providing important information about major trends and promising directions for further experimentation. Using the design with appropriate repeats therefore allowed us to accurately examine the effect of all four peptides alone and in various concentration combinations. In the current study, additional repeats were performed to improve the statistical accuracy of the study. To increase the accuracy of the statistical evaluation, surfaces for these repeats were prepared and cultured separately from those in the original design. For attachment, the surfaces were activated by reaction with trifluoroethane sulfonyl chloride in dry acetone (200 µl in 5 ml) with pyridine (400 µl) for 1 h. Surfaces were then rinsed with increasing amounts of 1 mM HCl (10– 100%) in
140
L. Aucoin et al.
Table 1. Summary of surfaces prepared in experimental design Sample 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
% YIGSR 25 100 0 0 0 0 50 50 50 0 0 0 62 13 13 13 33 33 33 0 50 50 0
% RGDS 25 0 100 0 0 0 50 0 0 50 50 0 13 62 13 13 33 33 0 33 0 0 50
% PDSGR 25 0 0 100 0 0 0 50 0 50 0 50 13 13 62 13 33 0 33 33 50 0 50
% PHSRN 25 0 0 0 100 0 0 0 50 0 50 50 13 13 13 62 0 33 33 33 0 50 0
Note that 21 is a repeat of 8, 22 is a repeat of 9 and 23 is a repeat of 10 as prescribed by the experimental design. Contact angle measurements were also made on Samples 1– 7, 11, 12, 13, 19, PDSGRYIGSR and PHSRNRGD as well as on hydroxylated PDMS. XPS analysis was performed on Samples 17– 20.
acetone for five minutes each time followed by a rinse in 100% sodium bicarbonate buffer (0.2 M, pH 9). Activated surfaces were subsequently modified by reaction with the appropriate peptide solution for 24 h at room temperature. Modified surfaces were rinsed well with buffer and with β-mercaptoethanol. Peptide solutions were prepared from 100 ng/ ml stock solutions and the total peptide concentration in each reaction solution was 100 ng/ ml. A summary of the surfaces prepared is presented in Table 1, with fractions representing mass fractions. Reaction volumes were selected in order that there was at least a 10 fold excess of peptide based on the surface areas and approximate surface hydroxyl concentrations as determined by XPS. All surfaces in the design were prepared for cell analysis, with some additional randomly selected repeats as shown and described above. Further evaluation of the efficacy of systems was performed using additional control surfaces, prepared using PPSRN rather than PHSRN. These surfaces were also tested for cell interactions. Additional randomly selected samples were prepared for surface characterization using XPS and water contact angles. For comparison,
Corneal epithelial cells and cell adhesion peptide combinations surfaces
141
Table 2. Surfaces prepared using radiolabeled peptides for quantification of attachment
1 2 3 4 5 6 7 8 9
YIGSR
RGDS
25 25 25 62 62 62 13 13 13
0 25 25 0 13 13 0 13 13
125 I GYRGDS 25 0 0 13 0 0 62 0 0
PHSRN 25 0 25 13 0 13 13 0 13
125 I GYPHSRN 0 25 0 0 13 0 0 62 0
PDSGR 25 25 0 13 13 0 13 13 0
125 I GYPDSGR 0 0 25 0 0 13 0 0 62
Each sample was prepared in triplicate.
surfaces were also prepared using the linear combination peptides RGDS-PHSRN and YIGSR-PDSGR. Quantification of peptide density To quantify peptide surface densities, the tyrosine containing peptide analogs GYRGDS, GYPHSRN and GYPDSGR were radiolabeled with 125 I using the iodogen method. Free iodide was removed using Sep-Pack columns (Millipore Waters, MA). The surfaces summarized in Table 2 were prepared using combinations of the labeled and unlabeled peptides. Following extensive rinsing, the amount of attached peptide was quantified by measuring the radioactivity using a gamma counter. Water contact angles Advancing and receding sessile drop water contact angles were measured on the unmodified, hydroxylated and several of the peptide modified PDMS surfaces with a Ramé Hart NRL C.A. goniometer (Mountain Lakes NJ) Milli-Q water (18 M) was used with a drop volume of approximately 0.02 ml. X-ray photoelectron spectroscopy (XPS) XPS analysis was performed at the National Research Council of Canada. The surface of the samples was analyzed using a KRATOS AXIS HS X-ray photoelectron spectrophotometer (Kratos, Manchester UK). The size of the analyzed area was approximately 1 mm2 . Monochromatized Al K radiation was used for excitation and a 180◦ hemispherical analyzer with a three-channel detector was employed. The X-ray gun was operated at 15 kV and 20 mA. The spectrophotometer was operated in Fixed Analyser Transmission (FAT) mode throughout the study using electrostatic magnification. Surface and high-resolution spectra were collected using a 160 and 20 eV pass energy respectively. The pressure in the analyzer chamber was
142
L. Aucoin et al.
10−8 to 10−9 torr. An electron flood gun was used to neutralize the charge during the experiment. Binding energies were referenced to the carbon-carbon bond that was assigned a binding energy of 285 eV. Atomic composition was estimated using standard software provided with the instrument using the following sensitivity factors: 0.25 for C1s, 0.66 for O1s and 0.42 for N1s relative to F1s at 1.00. Peak deconvolution was performed using the software provided with the instrument. In vitro cell adhesion The effect of surface modification with peptide combinations on interactions with human corneal epithelial cells was examined. All materials used in cell culture, if not received sterile from the manufacturer, were sterilized by autoclaving. Polymer samples were placed in the wells of a 24 well tissue culture place and human corneal epithelial cells, were seeded on the surfaces at a density of 105 cells per well. The hybrid Adenovirus 12-SV40 immortalized human epithelial cell line of ArakiSasaki et al. [30] dubbed ‘HCE’ was used. This line has been shown to have similar properties to normal corneal epithelial cells. Cell culture medium Keratinocyte Serum-Free Medium, supplemented with bovine pituitary extract and epidermal growth factor (EGF) (Canadian Life Technologies, Burlington ON). Samples were examined on a daily basis for cell attachment, spread and proliferation. For further examination, the cells were fixed in Davidson’s fixative (10% neutral formalin, acetic acid (95%) and water in a 1 : 3 : 2 ratio and stained with haemotoxylin and eosin (H&E). Cell surface density was assessed using digital images captured by light microscopy and analysed morphometrically using the Northern Eclipse software to obtain fractional surface coverage in a blinded fashion. The cultured cells were also immunohistochemically stained for cytokeratins 3 and/ or 12 with the AE5 monoclonal antibody (ICN, Aurora OH), to confirm the cell type. A broad-based antibody (pan cytokeratin, Sigma Chemical Co., St. Louis, MO) was used to recognize cytokeratin in undifferentiated or immature cells as well as in differentiated cells.
RESULTS
The transparency of some of the surfaces was compromised by the modification procedure, with several of the surfaces becoming translucent following the modification. There did not appear to be a pattern associated with these changes. Otherwise, surfaces did not appear to be affected macroscopically by the modification procedure. Peptide surface density Peptides surface densities, evaluated using combinations of unlabeled and 125 I labeled peptides are summarized in Fig. 1. As expected, as the concentration of
Corneal epithelial cells and cell adhesion peptide combinations surfaces
143
Figure 1. Peptide surface density measured using 125-I labeled versions of the peptides. There are significant differences in the peptide surface densities at all peptide solution concentrations, with GYPDSGR having the highest density and GYRGDS having the lowest density at all reaction solution concentrations. The peptide surface density of each of the peptides seems to be a linear function of the peptide solution concentration.
the peptide in the reaction solution increased, the resultant peptide surface density increased for all three of the peptides studied. Furthermore, for all of the peptides, this increase occurred in a relatively linear fashion as a function of reaction solution concentration (R 2 > 0.98 for all of the peptides examined). There were, however, peptide specific differences with respect to surface reaction. It can be seen at the same concentration in the reaction solution all cases, GYRGDS was found on the surface at the lowest density and GYPDSGR was found at the highest density. These differences were significant at all concentrations examined (α < 0.01). Water contact angles Water contact angle measurements, summarized in Fig. 2, show several notable factors. The unmodified PDMS surfaces are quite hydrophobic, with a measured advancing water contact angle of 96.3 ± 3.2◦ (n > 10) and receding water contact angle of 80.1 ± 4.0◦ . Following hydroxylation of the surface by plasma polymerization of allyl alcohol, a significant (p < 0.005) decrease in both the advancing and receding water contact angles occurred to 77.4 ± 1.0 and 57.8 ± 1.5 (n > 6). Modification with the peptides also resulted in significant changes in the water contact angles compared with the hydroxylated surfaces in some but not all cases. However, there were significant differences in both the advancing
144
L. Aucoin et al.
Figure 2. Advancing and receding water contact angles measured on combination peptide modified surfaces. There are significant differences in both the advancing and receding water contact angles as a result of modification with the different peptide combinations. Fractions represent % mass in reaction solution YIGSR : RGDS : PDSGR : PHSRN.
(p 0.001) and receding (p < 0.05) water contact angles on the surfaces modified with different peptide combinations. X-ray photoelectron spectroscopy XPS analysis of several of the modified surfaces, summarized in Table 3, provides additional evidence that the surfaces are modified with the peptide combinations. Relative to the unmodified PDMS surface, the hydroxylated PDMS surface shows a slight increase in the C1s signal and a decrease in the Si2p signal. While the O1s signal decreased relative the control surface, this may be more representative of the hydrophobicity of the XPS environment than of the effectiveness of the plasma polymerization of allyl alcohol in light of the significant decreases in the water contact angles. The high-resolution C1s envelope showed the appearance of a peak at 285.8, indicative of a C O bond, as would be expected on a poly allyl alcohol surface. A significant amount of silicon remained on the hydroxylated surfaces, likely the result of the depth of XPS analysis and the fluidity of the polymer. Attachment of the peptides, alone and in combination resulted in significant changes in the surface as evaluated by XPS. A small but significant N1s signal was noted on all but one of the peptide-modified surfaces as would be expected. The nitrogen levels, on the order of 1%, were however considerably less than would be expected on surface containing a peptide monolayer which is likely a reflection of
Corneal epithelial cells and cell adhesion peptide combinations surfaces
145
Table 3. Summary of XPS results on combination peptide modified surfaces Sample
C1s
Unmodified Hydroxylated 33/ 33/ 33/ 0∗ 33/ 33/ 0/ 33∗ 33/ 0/ 33/ 33∗ 0/ 33/ 33/ 33∗
Total 47.7 47.2 35.5 41.3 42.0 46.3
285.0 42.2 43.9 27.7 33.3 37.0 39.1
285.7 5.5 3.3 4.5 5.5 3.9 5.0
287.3
2.4 2.5 1.1 1.8
O1s
Si2p
26.9 28.6 38.7 34.2 32.2 29.3
25.5 24.2 24.9 24.5 25.0 23.6
N1s
289.6
0.8
1.0 0.6 0.8
* YIGSR/RGDS/PDSGR/PHSRN.
the inevitable presence of a significant amount of underlying polymer. There were significant changes in the C1s high-resolution envelopes on the peptide modified surfaces compared with the hydroxylated surfaces. Specifically, on the peptide modified surfaces there was an increase in the peaks corresponding to carbon bonded to nitrogen or oxygen via a single bond (286.3 eV) and the appearance of the peak corresponding to carboxylic and amide functions (288.8 eV) as would be expected. While there were slight differences between the different peptide combinations, no trends were apparent as would be expected from the size of an analyzed area. In all cases the amount of silicon on these surfaces remains relatively constant at approximately 24.5 atom percent, likely a reflection of the relatively small peptide size, the surface morphology and the analysis depth. In vitro cell culture studies Fractional coverage of human corneal epithelial cells on the various modified surfaces is summarized in Fig. 3. Several of the samples were repeated, and in some cases there were multiple repeats. In all but one case, the fractional adhesion between identical surfaces did not differ by more the 10% despite the fact that many of the repeats were prepared and were cultured separately from the surfaces prepared for the main experimental design. These repeated results were included in the statistical analysis of the data. However due to the low number of repeats based on the experiment design parameters (n 4 in all cases and n = 1 for most samples) error bars are not shown on the figure. It should be noted that while there were small differences in the spreading of the cells on the different surfaces, these were generally unremarkable. There was no adhesion noted on the unmodified surface as would be expected. Surprisingly, even in the presence of serum (not shown), there was no adhesion on these surfaces. There were however differences in cell coverage on the various modified samples. Statistical analysis of the data using a mixture design suggests surface modified with combinations of peptides resulted in greater fractional coverage with cells than surfaces modified with single cell adhesion peptides (p 0.01). The YIGSR, RGDS and PHSRN fractions in the reaction mixture
146
L. Aucoin et al.
Figure 3. Summary of cell adhesion results to combination peptide modified surfaces. Differences in the level of corneal epithelial cell adhesion were noted on the various surfaces as a result of modification with cell adhesion and synergistic peptides in various combinations. Fractions represent % mass in reaction solution YIGSR : RGDS : PDSGR : PHSRN.
were determined to significantly affect the resultant cell coverage. While PDSGR concentration in the reaction mixture was not shown to have a significant effect on fractional cell coverage, there was a YIGSR-PDSGR interaction effect. This is not surprising since these two peptides have been shown to act synergistically [20]. However, the combination of RGDS and PHSRN, which have also been suggested to show synergy [18, 19], did not significantly affect fractional corneal epithelial cell coverage. It is also interesting to note that the peptides combinations resulting the greatest fractional cell coverage included three or four peptides with RGDS and/ or YIGSR. Modification of the surfaces with the peptides PDSGR-YIGSR and PHSRN-RGDS gave similar levels of adhesion to surfaces modified with the same ‘free’ synergistic peptides in a 50 : 50 ratio, although in both cases the fractional epithelial cell coverage with the long chain combination peptides was slightly less than that noted on the surfaces modified with free peptides in combination. These differences are not likely significant. The effect of replacing the fibronectin derived synergistic peptide PHSRN with an inactive control on fractional corneal epithelial cell coverage is summarized in Fig. 4. In most cases, surfaces modified with the active peptide showed greater cell adhesion as expected. It is not surprising that there are also cells on the surfaces modified with combinations including the inactive control however, since
Corneal epithelial cells and cell adhesion peptide combinations surfaces
147
Figure 4. Comparison of corneal epithelial cell adhesion to surfaces modified with peptide combinations including the RGDS synergistic peptide PHSRN and the non-synergistic control peptide PPSRN. In most cases, adhesion to the surfaces modified with the non-synergistic control was less than adhesion to the surfaces modified to include the synergistic peptide. Fractions represent % mass in reaction solution YIGSR : RGDS : PDSGR : P(P-H)SRN.
the presence of the other peptides will also affect the interactions of the surface with the cells. A very low fractional coverage was noted on the surface modified with 100% PPSRN as expected. However, there was significant adhesion on the PHSRN surface, which is somewhat surprising given the reported lack of cell binding activity of this peptide [19].
DISCUSSION
The importance of tissue integration and specifically an epithelial layer in the development of a corneal onlay or keratoprosthesis has been established in a number of studies [31]. Various surface modification techniques including plasma surface modification [3, 25, 26] and covalent attachment of cell adhesion proteins [9, 14– 16] and peptides [11, 12] have been used to promote the interactions of corneal epithelial cells with the surfaces in question. In previous studies, modification with collagen or collagen derivatives, fibronectin or the fibronectin binding peptide RGDS, as well as the laminin-derived peptide YIGSR has given promising in vitro results. Despite these results however, maintaining a confluent monolayer following in vivo implantation of various devices has not been successfully achieved. This is
148
L. Aucoin et al.
likely the result of a complex interplay of factors including the surface chemistry and topography [32– 35] as well as the porosity of the underlying substrate [36– 40], providing adequate nutrient flux to the cells. The goal of surface modification with cell adhesion peptides is to mimic as closely as possible the surface under natural conditions while avoiding the complications associated with the use of proteins derived from animal tissues including inflammation and nonspecificity of attachment. It is hoped that under these conditions, preseeded cells will ultimately generate an appropriate matrix and perform the functions of the natural cornea. Given that the composition of the corneal epithelial basement membrane is primarily laminin and Type IV collagen, while fibronectin dominates under wound healing conditions, it is reasonable to hypothesize that combinations of laminin derived, fibronectin derived and/ or collagen derived peptides may mimic the corneal surface more closely, resulting in epithelial cell interactions and in the generation of a stable and physiologically active epithelial layer. The peptides used in the present study were selected based on their presence in the extracellular matrix of a normal human cornea under various conditions. The synergism between of the fibronectin derived peptides RGDS and PHSRN and that between the laminin derived peptides YIGSR and PDSGR have been shown in various studies [18– 20]. The importance of this synergism in interactions with corneal epithelial cells on synthetic surfaces was examined in the current work. While we and others have demonstrated the potential for single peptide modified surfaces, there have been relatively few studies where combinations of peptides such as those used in the current work have been examined [21– 23]. Furthermore, we are aware of no studies examining the specific peptide combinations used in this work or studies in which the various synergistic peptides are attached to the surfaces alone rather than as ‘super peptide’ combinations (i.e. RGDS and PHSRN alone rather than as RGDSPHSRN or PHSRNRGDS with or without a glycine spacer molecule). The use of an experimental design also permits us to examine a range of combinations and concentrations not previously evaluated. Longer time point experiments rather than shortterm initial adhesion studies were used in this work in order to clearly define the differences as well as to provide information about the long-term interactions between the cells and the surfaces. While the peptide modified surfaces showed a poorer response than tissue culture polystyrene, which showed complete coverage at the end point in all cases, there was a clear difference between the surfaces modified with the single peptide versus the multi-peptide solutions. Statistically, the combination of laminin derived peptides YIGSR and PDSGR showed the greatest effect on surface coverage with corneal epithelial cells, in agreement with our previous results which suggest that the presence of the laminin derived YIGSR had a greater influence on the adhesion of corneal epithelial cells than the fibronectin derived RGDS [12]. YIGSR, RGDS and PHSRN were also found to significantly affect adhesion despite that fact that PHSRN has been shown not to have a cell adhesive effect on its own. Furthermore, in support of the hypothesis that combinations of cell adhesion peptides may better
Corneal epithelial cells and cell adhesion peptide combinations surfaces
149
mimic the natural extracellular matrix, the surfaces that resulted in the greatest levels of corneal epithelial cell adhesion included multiple peptides. Surfaces with all four of the peptides, with the either RGDS or YIGSR in the greatest quantity showed particularly good corneal epithelial cell adhesion. XPS, water contact angles and surface modification with radioactively labeled peptides demonstrate that the surfaces were modified with multiple peptides. Significant differences in water contact angles were noted on the surfaces modified with different peptide combinations. As well, differences between the different peptide combinations were noted by XPS. Examination of peptide surface interactions with radiolabeled peptides showed interesting patterns of peptide surface interactions. While the radioactively labeled peptides studies are not identical to those used in cell studies due to the need for a tyrosine for labeling, it is assumed that the behaviour of these peptides will be reasonably similar to their non-labeled counterparts, and that the results therefore give a reasonable indication of the composition of the modified surfaces. There was a strong linear correlation between the amount of peptide in the reaction solution and the amount of peptide on the surface. The amount of peptide on the surfaces was clearly peptide dependent suggesting differences in peptide reactivity as a function of the composition. It is difficult to predict the surface densities with the unlabeled peptides however due to the presence of the glycine spacer and tyrosine group. Surface peptide concentrations were on the order of pmol/ cm2 , consistent with peptide surface densities measured in other studies [41, 42] and above the minimum levels suggested to be necessary for cell adhesion [43]. It is therefore apparent that the plasma polymerization reaction is generating a sufficient surface hydroxyl concentration to give a surface with a high enough peptide density to support adhesion of corneal epithelial cells. The substrate selected for these studies, poly dimethyl siloxane (PDMS), was chosen based on its previously demonstrated ophthalmic compatibility, transparency and oxygen permeability. However the porosity and therefore nutrient permeability of the materials are such that it is unlikely that this will be suitable material for long-term implantation, and modifications will be necessary to improve these properties. The surface modification procedure, involving plasma polymerization of allyl alcohol and attachment of the peptides via tresyl chloride chemistry should yield relatively substrate independent surfaces and can also be applied to other substrates, although the interactions of the cells with surface modified PDMS are superior to the interactions with surface modified pHEMA from our previous studies [12]. In this study, epithelial cells were plated on the surfaces under serum-free conditions, in order to avoid exposure to adhesive proteins in the serum including fibronectin and vitronectin. While the presence of laminin and/ or fibronectin in the culture medium may have provided information about the nature of the binding between the cells and the substrates through interactions between cell surface receptors and the adhesion proteins blocking interactions with the surface, interactions between these receptors and proteins adsorbed on the surface cannot
150
L. Aucoin et al.
be ruled out. Therefore, the use of serum-free conditions only allowed us to study the interactions of corneal epithelial cells and surface-bound peptides without the interference of extraneous and possibly confounding peptide groups. From the results, it is therefore clear that combinations of appropriate cell adhesion and synergistic peptides can improve corneal epithelial cell adhesion to synthetic surfaces. However, while it seems likely in all cases that the peptides are mediating initial attachment of the cells, allowing them to secrete matrix proteins that further promote attachment. This has been previously shown with corneal epithelial cells [44]. While it is possible that this initial attachment is non-specific in all cases, the differences in coverage between the various combinations and the lack of coverage on the PDMS control make it seem reasonable that there is some specificity in the initial attachment imparted by the presence of the peptides in different combinations. Furthermore, the dramatic decreases in the attachment when PPSRN rather than PHSRN was used provide additional evidence that there is some specificity to the attachment in the presence of the peptides. In summary, PDMS surfaces modified with a combination of cell adhesion and synergistic peptides derived from laminin and fibronectin were found to improve adhesion of corneal epithelial cells compared to modification with single peptides only. The combination of YIGSR and PDSGR was shown to have a particularly significant effect on the fractional coverage of the surfaces with the cells. Statistically, the amount of YIGSR, RGDS and PHSRN alone were also shown to have a significant effect on cell coverage. Surface modification with the different peptides was confirmed by XPS, water contact angles and radiolabeling. The radiolabeled peptide results demonstrate that the surface density of peptide on the surface varies in a linear fashion with the solution concentration of peptide. However, the surface mass density was shown to be a function of the peptide in the solution. In general, surface concentrations were on the order of pmol/ cm2 . The use of combinations of peptides on surfaces therefore has the potential to improve cell surface interactions, potentially by more closely mimicking the composition of the extracellular matrix. Acknowledgements The authors acknowledge the technical assistance of Lulu Burstzyn, Manish Bharati and Rena Cornelius. We thank Dr. John Brash and Dr. John MacGregor for useful discussions. Funding from the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged.
REFERENCES 1. A. W. Lloyd, R. G. A. Faragher and S. P. Denyer, Biomaterials 22, 769 (2001). 2. M. Franco, P. F. Nealey, S. Campbell, A. I. Teixeira and C. J. Murphy, J. Biomed. Mater. Res. 52, 261 (2000). 3. S. D. Lee, G. H. Hsiue, C. Y. Kao and P. C. T. Chang, Biomaterials 17, 587 (1996).
Corneal epithelial cells and cell adhesion peptide combinations surfaces
151
4. H. M. Leibowitz, V. Trinkaus-Randall, A. Tsuk and C. Franzblau, Prog. Ret. Eye Res. 13, 605 (1994). 5. R. Latkany, A. Tsuk, M.-S. Sheu, I.-H. Loh and V. Trinkaus-Randall, J. Biomed. Mater. Res. 36, 29 (1997). 6. C. R. Hicks, J. H. Fitton, T. V. Chirila, G. J. Crawford and I. J. Constable, Surv. Ophthalmol. 42, 175 (1997). 7. C. R. Hicks, X. Lou, S. Platten, A. B. Clayton, S. Vijaayasekaran, H. J. Fitton, T. V. Chirila, G. J. Crawford and I. J. Constable, Australian and New Zealand Journal of Ophthalmology 25, S50 (1997). 8. J. M. Legeais, I. Drubaix, B. Briat, G. Renard and Y. Pouliquen, Journal Francais d’Ophthalmologie 20, 42 (1997). 9. H. Kobayashi and Y. Ikada, Curr. Eye Res. 10, 899 (1991). 10. V. Trinkaus-Randall, M. Vural, J. Capecchi, C. Franzblau and H. M. Leibowitz, Invest. Ophthalmol. Vis. Sci. (Suppl.) 32, 1983 (1991). 11. G. Johnson, M. Jenkins, K. M. McLean, H. J. Griesser, J. Kwak, M. Goodman and J. G. Steele, J. Biomed. Mater. Res. 51, 612 (2000). 12. K. Merrett, C. M. Griffith, Y. Deslandes, G. Pleizier and H. Sheardown, J. Biomater. Sci. Polymer Edn 12, 647 (2001). 13. V. Trinkaus-Randall, J. Capecchi, A. Newton, A. Vadasz, H. Leibowitz and C. Franzblau, Invest. Ophthalmol. Vis. Sci. 29, 393 (1988). 14. H. Kobayashi and Y. Ikada, Biomaterials 12, 747 (1991). 15. R. Z. Xie, D. F. Sweeny, G. J. Beumer, G. Johnson, H. J. Griesser and J. G. Steele, Aust. NZ J. Ophthalmol 25 (Suppl. 1), S46 (1997). 16. H. Thiessen, K. McLean, G. Johnson, J. G. Steele and H. J. Griesser, Society for Biomaterials, Meeting Transactions 22, 450 (1999). 17. E. Ruoslahti, Ann. Rev. Cell. Dev. Biol. 12, 697 (1996). 18. S. Aota, M. Nomizu and K. M. Yamada, J. Biol. Chem. 269, 24756 (1994). 19. S. D. Redick, D. L. Settles, G. Briscoe and H. P. Erikson, J. Cell. Biol. 149, 521 (2000). 20. H. K. Kleinman, J. Graf, Y. Iwamoto, M. Sasaki, C. S. Schasteen, Y. Yamada, G. R. Martin and F. A. Robey, Arch. Biochem. Biophys. 272, 39 (1989). 21. W. J. Kao, D. Lee, J. C. Schense and J. A. Hubbell, J. Biomed. Mater. Res. 55, 79 (2001). 22. D. J. Heath, S. Halstenberg, J. C. Schense, J. W. Kao, S. Downes and J. Hubbell, in: Society for Biomaterials 25th Annual Meeting, Abstracts 22, p. 140 (1999). 23. Y. W. Tong and M. S. Schoichet, Biomaterials 22, 1029 (2001). 24. D. L. Mooradian, J. B. McCarthy, J. D. Cameron, A. P. N. Skubitz and L. T. Furcht, Invest. Ophthalmol. Vis. Sci. 29, 393 (1988). 25. G.-H. Hsiue, S. D. Lee, P. C. T. Chang and C.-Y. Kao, J. Biomed. Mater. Res. 42, 134 (1998). 26. P. C. T. Chang, S. D. Lee and G. H. Hsiue, J. Biomed. Mater. Res. 39, 380 (1998). 27. B. M. Wickson and J. L. Brash, Colloids and Surfaces A: Physiochem and Eng Aspects 156, 201 (1999). 28. D. Yu and R. E. Marchant, Macromolecules 22, 2957 (1989). 29. S. Yuan and R. E. Marchant, J. Appl. Polym. Sci. Appl. Polym. Symp. 54, 77 (1994). 30. K. Araki-Sasaki, Y. Ohashi, T. Sasbe, K. Hayashi, H. Watanabe, Y. Tano and H. Handa, Invest. Ophthalmol. Vis. Sci. 36, 614 (1995). 31. T. V. Chirila, C. R. Hicks, P. D. Dalton, S. Vijaysekaran, X. Lou, Y. Hong, A. B. Clayton, B. W. Ziegelaar, J. H. Fitton, S. Platten, G. J. Crawford and I. J. Constable, Prog. Poly. Sci. 23, 447 (1998). 32. R. Z. Xie, D. F. Sweeney, H. J. Griesser and J. G. Steele, Aust. NZ J. Ophthalmol. 26, S47 (1998). 33. J. H. Fitton, B. A. Dalton, G. Beumer, G. Johnson, H. J. Griesser and J. G. Steele, J. Biomed. Mater. Res. 42, 245 (1998).
152
L. Aucoin et al.
34. B. A. Dalton, M. D. M. Evans, G. A. McFarland and J. G. Steele, J. Biomed. Mater. Res. 45, 384 (1999). 35. M. D. M. Evans, B. A. Dalton and J. G. Steele, J. Biomed. Mater. Res. 46, 485 (1999). 36. R. Turss, J. Friend and C. H. Dohlman, Exp. Eye Res. 9, 254 (1970). 37. S. L. Lane and R. L. Lindstrom, Int. Ophthalmol. Clin. 31, 37 (1991). 38. B. E. McCarey, Refract. Surg. 31, 87 (1991). 39. M. B. McDonald, B. E. McCarey, B. Storie, R. W. Beuerman, B. Salmeron, G. Van Rihj and P. M. Knight, J. Cataract. Refract. Surg. 19, 213 (1999). 40. D. F. Sweeney, R. Z. Xie, D. J. O’Leary, A. Vannas, R. O’Dell, K. Schindhelm, H. Cheng, J. G. Steele and B. A. Holden, Invest. Ophthalmol. Vis. Sci. 39, 284 (1997). 41. R. McMillan, B. Meeks, F. Bensebaa, Y. Deslandes and H. Sheardown, J. Biomed. Mater. Res. 54, 272 (2001). 42. S. Saneinejad and M. Shoichet, J. Biomed. Mater. Res. 42, 13 (1998). 43. S. P. Massia and J. A. Hubbell, J. Cell Biol. 114, 1089 (1991). 44. J. G. Steele, G. Johnson, H. J. Griesser and P. A. Underwood, Biomaterials 18, 1541 (1997).
Polyelectrolyte multilayer films modulate cytoskeletal organization in chondrosarcoma cells DOMINIQUE VAUTIER 1,∗ , VERONIQUE KARSTEN 1 , CHRISTOPHE EGLES 1 , JOHANNA CHLUBA 1 , PIERRE SCHAAF 2 , JEAN-CLAUDE VOEGEL 1 and JOËLLE OGIER 1 1 INSERM U 424, UFR Odontologie, 11 Rue Humann, 67085 Strasbourg Cedex, France 2 Institut Charles Sadron (CNRS/ULP), 6 Rue Boussingault, 67083 Strasbourg Cedex, France
Received 6 December 2001; accepted 18 March 2002 Abstract—The aim of this study was to evaluate polyelectrolyte multilayer films as interfaces for implants. Polyelectrolyte multilayers were built up with different terminating layers by alternate deposition of oppositely charged polyelectrolytes on which chondrosarcoma (HCS-2/8) cells were grown in the presence of serum. Films formed by an increasing number of layers were investigated. The terminating layer was made of one of the following polyelectrolytes: poly-sodium-4-styrenesulfonate (PSS), poly-L -glutamic acid (PGA), poly-allylamine hydrochloride (PAH), or poly(L -lysine) (PLL). Cell viability, inflammatory response, adherence, and cytoskeletal organization were studied. Induction of interleukin-8 (IL-8) secretion was detected on PAH and PLL ending polyelectrolyte films. Early cellular adherence was enhanced with PGA, PAH, PLL, and, to a lower extent, PSS terminating layers. Adherence was independent of the number of layers constituting the films. The presence of actin filaments and vinculin focal adhesion spots was observed on PSS or PAH ending films. They were respectively partially and totally absent on PGA and PLL terminating multilayer architectures. For PLL ending films, vinculin and actin organization was clearly dependent on the number of deposited layers. The results of this study suggest that PSS ending multilayered films constitute a good interfacial micro-environment at the material surface for HCS-2/8 cells. Key words: Polyelectrolyte multilayer films; cell adhesion; actin; vinculin.
INTRODUCTION
The chemical modification of biomaterial surfaces became a major challenge in the last decade, allowing broad medical applications for implant and tissue engineering [1– 3]. Bioactive molecules such as enzymatic peptides, for example, ∗ To whom correspondence
should be addressed. E-mail:
[email protected]
154
D. Vautier et al.
have been incorporated into the bulk of biodegradable materials [4]. In other approaches, only surface properties are modified. Self-assembled monolayers (SAMs) or Langmuir– Blodgett techniques have been commonly employed to produce new interfaces [5, 6]. More recently, a new versatile method of self-assembled architectures based on the alternate deposition of polyanions and polycations has been developed for the build up of multilayered polyelectrolyte films [7]. Besides varying film thickness, roughness, and porosity, it is also possible to incorporate in the film architecture functionalized macromolecules, enzymes, and proteins embedded at different depths [8, 9]. This concept gives rise to complex nanoarchitectures with specific biomimetic properties. It has, for example, been shown that α-melanocortin peptide hormone covalently coupled to poly(L-lysine) conserves full biological activity when deeply embedded into poly(L-lysine)/ poly(Lglutamic acid) (PLL /PGA) multilayer architectures [10]. Aimed to be used in contact with biofluids and/ or tissues, interfaces need to be evaluated in terms of biocompatibility properties in in vivo or in vitro conditions and in terms of cellular interaction with the substratum [11, 12]. The cell adhesion process contributes strongly to the clinical success of the implanted biomaterials [13]. During the adhesion of cells to material, specific cellular processes, such as signal transduction responses [13], can be deeply influenced by the interface microenvironment. The responses are mediated through cell-surface integrin receptors, which bind the cell to surface-adsorbed extracellular matrix proteins (ECMs) and connect the actin filament cytoskeleton to signalling molecules, including among them the cytoskeletal proteins vinculin and paxillin [14– 16]. Successful formation of the cytoskeletal organization, in addition to other early adhesion-dependent events [17, 18], is a major indicator of efficient ‘outside– in’ signal occurring between the extracellular environment (e.g. ECMs adsorbed on the material surface) and the cell [19, 20]. Physiological fluids (e.g. serum) contain numerous soluble proteins that react spontaneously with solid surfaces in very dynamic ways and contribute to the attachment and spreading of the cells [21– 23]. Most of the investigations concerning cellular interactions with a polyelectrolyte have been performed on adsorbed polyelectrolyte monolayers. For example, the coating of PLL monolayer is one of the most commonly used methods to promote cellular adhesion. Up to now, only a few studies have been devoted to the behavior of cells on polyelectrolyte multilayers. In particular, the influence of polyelectrolyte multilayer films on cellular cytoskeletal organization has not yet been studied. Therefore, in the present work we propose to investigate cell cytoskeletal actin and vinculin organization in early adhesion events on polyelectrolyte multilayer films. Different polycations such as polyethyleneimine (PEI), poly-allylamine hydrochloride (PAH), and poly(L-lysine) (PLL) and polyanions such as poly-sodium 4-styrenesulfonate (PSS) and poly-L-glutamic acid (PGA) were used to build up these architectures. The chosen cellular model was a cell line of human chondrosarcoma (HCS-2/8) having a cartilage phenotype. HCS-2/8 cells possess the ability
Polyelectrolyte multilayer films modulate cytoskeletal organization
155
to synthesize cartilage-specific proteoglycans and type II collagen [24]. Here, we evaluated viability parameters such as cell nucleus morphology, cell inflammatory response, and adherence.
MATERIALS AND METHODS
Cell culture HCS-2/8 cells, a human chondrosarcoma-derived chondrocyte-like cell line [24], were routinely grown in Gibco BRL’s minimum essential medium with Eagle’s Salts (MEM, Life Technologies), 10% fetal calf serum (FCS, Life Technologies), 50 U /ml penicillin, and 50 µg/ ml streptomycin (Biowhittaker) in an atmosphere of 5% CO2 and 95% air at 37 ◦ C. Polyelectrolyte film deposition The following polyelectrolytes were used for the build up of the multilayer films: polyethyleneimine (PEI; Mr 750 × 103 , Aldrich), polysodium 4-styrenesulfonate (PSS; Mr 70 × 103 , Aldrich), poly-allylamine hydrochloride (PAH; Mr 70 × 103 , Aldrich), poly-L-glutamic acid (PGA; Mr 54.8 × 103 , Sigma), and poly(Llysine) (PLL; Mr 23.4 × 103 , Sigma). PEI, PSS, and PAH solutions were prepared at 5 mg/ ml, whereas PGA and PLL solutions were used at 1 mg/ml. All polyelectrolytes were dissolved in 1 M NaCl. Glass coverslips (CML, France) were pretreated for 15 min at 100 ◦ C with 10−2 M SDS and 0.12 N HCl, followed by extensive doubly distilled H2 O rinsing. Glass coverslips were deposited in 24well plastic plates (NUNC) and immersed for 20 min in 300 µl of the appropriate polyelectrolyte solution. After each polyelectrolyte layer deposition, the wells were rinsed three times for 5 min with distilled water. A first set of film build-ups started by the deposition of a precursor film made up of PEI-(PSS /PAH)2 . The architecture was completed by further addition of either PSS, PGA /PLL /PGA, PGA /PAH, PGA /PLL, or PGA /PEI layers. In the following, these films will be described by the last deposited layer and denoted with the corresponding polyelectrolyte abbreviation (PSS, PGA, PAH, PLL or PEI) (type 1 architectures). As the control surface, a PAH or PLL monolayer and PLL /PSS or PLL /PGA bilayers, directly deposited on the solid surface, were made. A second set of polyelectrolyte multilayer architectures was prepared by changing the number of polyelectrolyte layers. These films were made up of (PAH /PSS)n or (PLL /PGA)n -PLL, with n = 1, 5, 10, and directly assembled on the glass surface (type 2 architectures). All of these different build-ups of polyelectrolyte multilayer architectures were controlled by optical wave-guide light-mode spectroscopy [10, 25]. Films were sterilized for 15 min by UV light (254 nm), stored at 4 ◦ C, and usually used within 1 week.
156
D. Vautier et al.
Test of cell viability HCS-2/8 cells (105 cells per well) were seeded on different polyelectrolyte multilayer architecture-coated coverslips, placed into 24-well plates, and maintained for 24 h. As a positive control, apoptosis was induced on PEI-terminating layers, which are known to be cytotoxic [26]. Cells were fixed with 3.7% paraformaldehyde (Boehringer), permeabilized with 0.1% Triton X-100 (Sigma), washed with phosphate-buffered saline (PBS), and finally stained with 2 µg/ ml Hoechst 33258 (Sigma) to detect condensed nuclei of apoptotic cells [27]. At the end of the procedure, coverslips were mounted in VectaShield (Vector) on glass slides and visualized using a Zeiss Axioplan 2 microscope with an MC 80 DX photographic attachment. Cytokine production The amount of IL-8 released after 24 h by HCS-2/8 cells (105 cells per well) in 24well plates coated with the different polyelectrolyte films was measured in culture supernatants by enzyme-linked immunoassay (ELISA, Biosource) according to the manufacturer’s recommended procedures. We used TNF-α preconditioned HCS-2/8 (10 ng of TNF-α per ml of cell culture, 24 h activation) as a positive control. These pretreated cells were washed out of the drug before seeding them into ELISA plates. In graphs, mean values with standard errors for one experiment as normalized data for an identical total cell number per well are given. Cell adhesion assay Quantification of attached cells labeled with horseradish peroxidase (HRP) was performed according to the method described by Löster et al. [28]. Briefly, HCS2/ 8 cells were washed with serum-free MEM, suspended subsequently in serumfree MEM containing 2 mg/ ml HRP, and incubated at 37 ◦ C for 30 min. After a further wash with serum-free MEM, HCS-2/8 cells, in 10% serum MEM, were plated at a density of 105 cells/ ml on the films. After 30 min of culture, the absorbance was measured. Unattached cells were removed after two washes with PBS at 37 ◦ C. Wells containing HRP-labeled attached cells on the different polyelectrolyte multilayer film-coated glass coverslips were incubated with 500 µl/ well of HRP-substrate buffer [10 mg OPD Sigma, 25 µl of 6% (v/ v) H2 O2 solution in 20 ml of 0.1 M sodium citrate buffer (pH 6.0), 0.5% (v/ v) Triton X-100]. After 15 min of reaction in the absence of light, the reaction was stopped by adding 125 µl of 1 M H2 SO4 per well and the OD of the reaction product was measured at 490 nm. The absorbance of the dye varied linearly with the number of adhering cells per well. Immunofluorescence The cells attached on the films were rinsed briefly in PBS and fixed in 3.7% formaldehyde diluted in PBS for 15 min at room temperature. The cells were then
Polyelectrolyte multilayer films modulate cytoskeletal organization
157
permeabilized for 10 min with 0.1% Triton X-100 in 3.7% formaldehyde diluted in PBS at room temperature. Rinsing was performed with PBS, followed by two 15-min incubations with 10% fetal calf serum and 15 min staining by TRITCphalloidin (5 µg/ ml, P1951, Sigma) and a final wash in PBS. For detection of vinculin, the cells were incubated with the primary antibody anti-human vinculin (V9131, Sigma), 1/ 100 diluted in PBS, washed in PBSand incubated for 30 min with the secondary antibody conjugated to fluorescein (Nordic, Tilburg), diluted 1/ 40 in PBS. Finally, cells were rinsed in PBS and mounted in VectaShield (Vector). Samples were observed with a Nikon Microphot-FXA fluorescence microscope.
RESULTS
Cell viability on polyelectrolyte multilayer films When HCS-2/8 cells were observed after 24 h of culture on PSS (Fig. 1A), PGA (not shown), PAH (not shown) or PLL (Fig. 1B) terminating layers (type 1 architectures), the nucleoplasm of the cells was uniformly labeled with the specific staining (Hoechst 33258). The aspect of the nucleus was very similar to that observed on untreated glass coverslips (Fig. 1C). In contrast, apoptotic condensed HCS-2/8 cell nuclei were observed after 24 h of culture on PEI terminating layers (Fig. 1D). Cytokine production on polyelectrolyte multilayer films In order to check the behavior of the polyelectrolyte multilayer architectures towards pro-inflammatory processes, we tested their effect on IL-8 secretion by HCS-2/8 cells after 24 h contact. In the culture supernatant of HCS-2/8 cells grown on PLL or on PAH terminating films (type 1 architectures), the IL-8 level was increased by 2.5 and 3 times, respectively, in comparison with the IL-8 level measured in the supernatant of HCS-2/8 cells grown on control surfaces without a film (‘IL-8 control surface’) (Fig. 2). The IL-8 level measured with PGA or PSS terminating layers was not significantly different from the ‘IL-8 control surface’. Adhesion of HCS-2/ 8 cells on polyelectrolyte multilayer films The cellular adhesion of HCS-2/8 cells was quantified by a HRP adhesion assay (see the Materials and Methods section). After 30 min of contact, comparatively to an uncoated glass surface, the cellular adhesion on PAH, PGA, and PLL ending films was increased by about 4 times (Fig. 3) and by 2 times on type 1 architectures
158
D. Vautier et al.
(A)
(B) Figure 1. Assessment of apoptosis by staining with H33258. HCS-2/8 cells were cultured for 24 h in MEM 10% FCS on PSS (A), PLL (B), and PEI (D) ending polyelectrolyte films, or on glass coverslips (C) fixed and stained with H33258. (D) Apoptotic condensed nuclei (arrow). Bar = 4 µm.
Polyelectrolyte multilayer films modulate cytoskeletal organization
(C)
(D) Figure 1. (Continued).
159
160
D. Vautier et al.
Figure 2. Effect of polyelectrolyte multilayer films (type 1 architectures) on IL-8 production by HCS-2 cells. Cells were seeded at a density of 105 cells per well in 96-well plates, pretreated with polyelectrolyte films as described in the Materials and Methods section. Cell culture supernatants were collected after 24 h of culture and analyzed by ELISA for the presence of IL-8. TNF-α-preconditioned HCS-2 cells (10 ng/ ml, 24 h) were used as a positive control for cytokine production. All films started by the precursor film PEI-(PSS/PAH)2 (wave symbol). The ending polyelectrolyte layers are indicated under the different columns. The cytokine level of each sample was normalized to the total cell number per well. Results are means of triplicate well (three measurements per well) determinations of one independent experiment. The error bars represent the standard errors of the means derived from independent experiments. *Statistically significant difference with respect to the other groups (n = 3, p < 0.001, ANOVA, Tukey test).
(Fig. 3). When the assay was monitored after 2 h, the measurements of the absorbance reached 2.3 OD on untreated surface, 2.1 OD on PSS terminating films, 2.29 OD on PGA terminating films, 2.3 OD on PAH terminating films, and 2.4 OD on PLL terminating films (type 1 architectures). These data suggested that the total number of attached cells was similar for all substrates. These results also indicate that the ending polyelectrolyte layer of the architecture modulates only early adhesion events. The influence on cellular adhesion of the number of deposited polyelectrolyte layers was next studied using PSS and PLL terminating layers. In this evaluation, (PLL /PGA)n -PLL or (PAH /PSS)n multilayer architectures were built up with n = 1, 5 or 10 (type 2 architectures). For both kinds of architecture, no significant differences were found, whatever the number of deposited layers (Fig. 4A, PSS ending films; Fig. 4B, PLL ending films).
Polyelectrolyte multilayer films modulate cytoskeletal organization
161
Figure 3. Cell adhesion on various polyelectrolyte ending multilayer films (type 1 architectures). All build-ups started by a precursor PEI-(PSS/PAH)2 film (wave symbol). The ending polyelectrolyte layers are indicated under the different columns. Cell attachment was monitored after 30 min postseeding. Results are means of triplicate well (four measurements per well) determinations of two independent experiments. The error bars represent the standard error of the two means derived from the two independent experiments. *Statistically significant difference with respect to the other groups (n = 24, p < 0.001, ANOVA, Student Newman– Keuls method).
Cytoskeletal organization The interactions of HCS-2/8 cells with their substratum were also evaluated by means of the cytoskeletal organization of vinculin and actin proteins. HCS-2/8 cells were cultured for 4 h on surfaces coated with monolayers (PAH or PLL), bilayers (PLL /PSS or PLL /PGA), or multilayer architectures (see the Materials and Methods section) in which the ending polyelectrolyte was varied (PSS, PGA, PAH or PLL) (type 1 architecture). On uncoated surfaces, cells presented a large variety of morphologies with predominant rounded aspects. A punctate distribution of vinculin is known to be associated with focal adhesion complexes [14]. As can be seen in Fig. 5A, HSC-2/8 cells newly attached on glass coverslips contain focal adhesion spots of vinculin localized at the base of the cell body. These structures were also observed when cells attached on PLL /PSS (not shown) or PLL /PGA bilayered films (Fig. 5B) and on multilayered PSS ending architectures (Fig. 5C). In contrast, the spots of vinculin were considerably reduced in size in HCS-2/8 cells attached on multilayered architectures ending with PGA (not shown). With PAH monolayer or multilayered architectures ending with PAH (not shown), the
162
D. Vautier et al.
Figure 4. Cell adhesion on polyelectrolyte multilayer films [varying the number of layers and ending with PSS (A) or PLL (B)]. The polyelectrolyte multilayer architectures are indicated under the different columns. Cell attachment was monitored after 30 min post-seeding. Results are means of triplicate well (four measurements per well) determinations for two independent experiments. The error bars represent the standard error of the two means derived from the two independent experiments. *No significant difference with respect to the other groups (n = 24, p < 0.001 ANOVA, Student Newman– Keuls method).
Polyelectrolyte multilayer films modulate cytoskeletal organization
163
(A) Figure 5. Vinculin immunostaining in HCS-2/ 8 cells adhered to polyelectrolyte monolayer, bilayer or multilayer films, varying the ending polyelectrolyte (type 1 architectures). HCS-2/8 cells were plated on (A) glass coverslips, or on (B) PLL/PGA and (D) PLL, or on polyelectrolyte multilayer films starting with the PEI-(PSS/PAH)2 precursor film and ending with (C) PSS and (E) PGA/PLL. HCS2/ 8 cells were cultured for 4 h at 37 ◦ C, fixed, and stained for vinculin. (A– D) Focal adhesion spots of vinculin (arrowhead). Comparable cellular observations were made in two independent experiments. Bar = 5 µm.
vinculin complexes were also similar to those described for cells adhering on glass coverslips. On PLL monolayers, several cells still presented characteristic vinculin focal adhesion (Fig. 5D), whereas for PLL ending multilayered architectures this cytoskeletal organization was not detected (Fig. 5E). Cells in contact with uncoated surfaces exhibited a fine parallel actin filament network (Fig. 6A). Cellular morphology and actin filament organization remained quite similar when cells attached on PLL /PSS (not shown) or on PLL /PGA bilayers (Fig. 6B). On PSS ending multilayer architectures, cells had actin cables localized at the periphery of the cell bodies together with an actin punctate distribution inside the cell body (Fig. 6C). However, on PGA ending multilayer architectures, little polymerized actin was seen (not shown). An actin punctate distribution was also clearly seen on PAH (not shown) or PLL monolayers (Fig. 6D), or with PAH (not shown) and PLL (Fig. 6E) ending multilayered architectures.
164
D. Vautier et al.
(B)
(C) Figure 5. (Continued).
Polyelectrolyte multilayer films modulate cytoskeletal organization
(D)
(E) Figure 5. (Continued).
165
166
D. Vautier et al.
(A) Figure 6. Actin staining in HCS-2/ 8 cells adhered to polyelectrolyte monolayer, bilayer or multilayer films, varying the ending polyelectrolyte (type 1 architectures). HCS-2/8 cells were plated on (A) glass coverslips, or on (B) PLL/PGA and (D) PLL, or on multilayered polyelectrolyte starting with the PEI-(PSS /PAH)2 precursor film and ending with (C) PSS and (E) PGA/ PLL. HCS-2/8 cells were cultured for 4 h at 37 ◦ C and then fixed. Actin was stained with TRITC-phalloidin. (A– C) Actin filaments (arrow); (D, E) actin dots (arrowhead). Comparable cellular observations were made in two independent experiments. Bar = 5 µm.
DISCUSSION
Since biomaterial/ cell interactions are determinant for cellular behavior such as cell adhesion [29], proliferation [30], and differentiation [31], we evaluated the viability, adhesion, and cytoskeleton organization of HCS-2/8 human chondrocyte-like cells with interfaces made of different polyelectrolyte multilayer films in the presence of serum. The structural and physico-chemical properties of multilayered polyelectrolyte architectures have been extensively investigated in the literature in the last decade. As a general rule for such films, each new polyanion (respectively polycation) addition renders the surface alternatively negatively (respectively positively) charged. This also holds for the PSS /PAH system, with zeta potential values alternating between −20 mV and +20 mV [32], and between −50 mV and +50 mV for the PGA /PLL system [33], after each new polyanion or polycation addition. The film thickness or amount of polyelectrolyte deposited per layer is largely dependent on various parameters such as ionic strength, the pH for weak polyacids or polybases, and film
Polyelectrolyte multilayer films modulate cytoskeletal organization
(B)
(C) Figure 6. (Continued).
167
168
D. Vautier et al.
(D)
(E) Figure 6. (Continued).
Polyelectrolyte multilayer films modulate cytoskeletal organization
169
growth mechanisms (linear or super increase regimes). Thus, for the PSS /PAH system, under comparable experimental conditions, the thicknesses determined by means of optical wave-guide light-mode spectroscopy (OWLS) were about 40 nm for PEI(PSS /PAH)5 and 70 nm for PEI(PSS /PAH)10. For the PGA /PLL system growing in a linear manner, the thicknesses obtained for PEI(PGA /PLL)5 and PEI(PGA /PLL)8 were about 100 and 280 nm, respectively [33]. Film roughnesses were also estimated by means of atomic force microscopy (AFM) and the maximal height amplitudes were of the order of 10– 12 nm for films having thicknesses of about 40 and 70 nm as estimated by OWLS. For the PGA /PLL system, the height amplitudes of the film profile increase with the number of deposited layers and then reach a constant value of about 25 nm for PEI(PGA /PLL)5 and PEI(PGA /PLL)10 [33]. Cell morphology and inflammatory responses were investigated to assess cell viability in the presence of such films. As previously observed for other cell models [26], on PSS, PAH, PGA, and PLL ending layers, HCS-2/8 cells did not exhibit an apoptotic morphology. However, cells secreted a pro-inflammatory cytokine, IL-8, when grown on PAH or PLL ending films. We observed that compared with adhesion to glass, early cell adhesion was enhanced by interfaces made of polyelectrolyte films ending with PAH, PGA, and PLL, and to a lower extent with PSS ending films. These results confirm that early adhesion is promoted on positively charged surfaces [26]. The cell attachment assay (HRP assay) used here was based on the total number of attached cells on a given surface. In these conditions, we found that the number of layers (n = 1, 5 or 10) did not influence cell adhesion, either with (PAH /PSS)n or with (PLL /PGA)n /PLL polyelectrolyte multilayer architectures. These results suggest that cells detect principally the ending layer and do not sense the underlying oppositely charged layers; they also suggest that cells were relatively insensitive to the surface roughness or porosity changes related to the number of deposited polyelectrolyte layers (Lavalle, personal communication). However, the HRP adhesion assay did not necessarily reflect attachment forces, which have been shown, in our laboratory, to be dependent on the number of layers as well as on the charge of the ending layer. The micro-peeling approach allows quantitative estimation of the cellular adhesion onto polyelectrolyte multilayer architectures. In accordance with the present work, stronger adhesion was found on positively charged ending multilayer films than on negatively charged ones, but differences in the cellular adhesion responses by changing the number of layers in the film were also measured. Decreased rupture forces with increased layer numbers for the positively charged PLL and no variation for the negatively charged PGA ending multilayers were measured (Richert, personal communication). The mechanism of cell attachment, spreading, and migration on surfaces with various physico-chemical properties, surface characteristics, and in the presence of adsorbed protein has been extensively studied [34]. Cell adhesion likely occurs via extracellular matrix proteins [fibronectin (FN), vitronectin, laminin, and albumin]
170
D. Vautier et al.
adsorbed onto the surface [35]. Previous studies have shown that most of the cell adhesion activity on synthetic surfaces was generated by surface-bound proteins containing serum [36]. A recent study by McClary et al. showed that the surface of carboxyl-terminated alkylthiol SAMs yielded higher levels of FN adsorption that correlated with high degrees of cell attachment spreading than methyl-terminated SAMs [20]. For osteoblasts, phosphorylation of paxillin, a cytoskeletal adaptor protein, could be induced by serum proteins, FN, and vitronectin adsorbed on a charged polystyrene surface [37]. In this study, analysis of the cytoskeletal actin and vinculin focal adhesion organization demonstrated differences in response with the charge of the ending layers and also with the number of deposited layers. Negatively charged ending layers preserved both actin filaments and vinculin spots for PLL /PSS or PLL /PGA bilayer and PSS ending multilayer build-ups, but not for PGA ending multilayer films. In this latter condition, the spots of vinculin were considerably reduced. PAH and PLL ending layers were less favorable for the polymerized form of actin. Interestingly, the most striking difference in the cytoskeletal organization, related to the number of layers constituting the films, was observed for PLL ending films. Cells contained vinculin adhesion spots and also some actin filaments after binding on a PLL monolayer. In contrast, they showed neither vinculin adhesion spots nor actin polymeric forms on PLL ending multilayer architectures. It has been reported that the focal adhesion kinase (FAK) [38, 39], the tensin protein [40], and the mitogen-activated protein kinase (MAPK) [41] were not phosphorylated in cells plated onto PLL. Tensin and FAK are important components of the focal adhesion sites, which become tyrosine-phosphorylated upon integrin-mediated attachment to ECM proteins. MAPK is activated and translocated to the nucleus in integrinmediated signal transduction pathways. Therefore, Krause et al. [41] proposed that cell adhesion to PLL does not involve integrins, but ionic charges. In accordance with these data, our results suggest that the adhesion mechanism is rather chargedependent on PLL multilayered ending films. PLL could exert an inhibitory effect on vinculin adhesion spots and actin filament organization. It will be interesting to determine whether PLL and PSS multilayered ending films involve an integrindependent adhesion mechanism in the presence of serum. In this case, contributions of serum-derived adhesion proteins such as fibronectin, vitronectin, laminin, and albumin would need to be evaluated. During cell adhesion, ECM produced by HCS2/ 8 cells could also participate in the adhesion mechanism, as well as ECM proteins present in serum. Considering this possibility, one cannot exclude a possible influence of the polyelectrolyte architecture on ECM production. On PLL monolayers, some cells presented spots of vinculin. This difference could be explained by the adsorption process of monolayers compared with multilayer films. Indeed, by AFM investigations, it was found that after deposition of PEI(PGA /PLL)1 on a silica surface, the surface was not completely coated with polyelectrolytes [33]. However, the same authors found that the polyelectrolyte multilayer architecture completely covered the silica surface after deposition of the
Polyelectrolyte multilayer films modulate cytoskeletal organization
171
fifth bilayer. Consistent with these data, our results suggest that cells adsorbed on a PLL monolayer could sense that silica was not completely covered with PLL and form focal adhesion sites. In contrast, on multilayer PLL ending films, cells were unable to sense the silica surface and did not develop focal adhesion sites.
CONCLUSION
We have examined chondrosarcoma cell interactions with polyelectrolyte multilayered surfaces, evaluating the biocompatibility of the films and the cytoskeletal organization of the HCS-2/8 cells in an in vitro cell culture system with serum. The results of this study demonstrate that the different polyelectrolyte ending multilayers tested did not induce a nucleus apoptotic morphology. However, PAH and PLL ending films activated IL-8 secretion. Early cell adhesion was optimal on positively charged (PAH and PLL) compared with negatively charged ending polyelectrolytes (PSS), and was not influenced by the number of polyelectrolyte layers deposited. In contrast, the number of layers could modulate cytoskeletal actin and vinculin. Finally, PLL ending multilayer films appear to be less biocompatible in terms of cytoskeletal organization than PSS. PSS ending architectures could be new and interesting interfaces for implants. Acknowledgements We thank Dr. C. Picart (University Louis Pasteur, Strasbourg, France) for stimulating discussions. We acknowledge Mrs N. Fournier for developing photographs and Mrs C. Affolter for technical support (University Louis Pasteur, Strasbourg, France). This study was supported by grants from Ligue Nationale Contre le Cancer (Comité du Bas-Rhin), from the program INSERM –CNRS ‘Ingéniérie Tissulaire’, and from the IFRO (Institut Français pour la Recherche Odontologique).
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.
R. Langer and J. Vacanti, Science 260, 920 (1993). N. Peppas and R. Langer, Science 263, 1715 (1994). N. Angelova and D. Hunkeler, Trends Biotechnol. 17, 409 (1999). J. C. Schense, J. Bloch, P. Aebischer and A. Hubbell, Nature Biotech. 18, 415 (2000). M. Mrksich, Curr. Opin. Colloid Interface Sci. 2, 83 (1997). M. Lösche, Curr. Opin. Solid State Mater. Sci. 2, 546 (1997). G. Decher, Science 277, 1232 (1997). F. Caruso, K. Niikura, D. N. Furlong and Y. Okahata, Langmuir 13, 3427 (1997). T. Cassier, K. Lowack and G. Decher, Supramol. Sci. 5, 309 (1998). J. Chluba, J.-C. Voegel, G. Decher, P. Erbacher, P. Schaaf and J. Ogier, Biomacromolecules 2, 800 (2001). 11. B. Groessner-Schreiber and R. S. Tuan, J. Cell Sci. 101, 209 (1992). 12. L. Chou, J. D. Firth, V. J. Uitto and D. M. Brunette, J. Cell Sci. 108, 1563 (1995).
172 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41.
D. Vautier et al. J. A. Hubbell, Biotechnology 13, 565 (1995). C. H. Damsky and Z. Werb, Curr. Opin Cell Biol. 4, 772 (1992). S. Miyamoto, S. K. Akiyama and K. M. Yamada, Science 267, 883 (1995). D. A. Calderwood, S. J. Shattil and M. H. Ginsberg, J. Biol. Chem. 275, 22610 (2000). L. Kornberg, H. S. Earp, J. T. Parsons, M. Schaller and R. L. Juliano, J. Biol. Chem. 267, 23439 (1992). M. D. Schaller and J. T. Parsons, Curr. Opin. Cell Biol. 6, 705 (1994). E. A. Clark and J. S. Brugge, Science 268, 233 (1995). K. B. McClary, T. Ugarova and D. W. Grainger, J. Biomed. Mater. Res. 50, 439 (2000). J. Andrade (Ed.), Surface and Interfacial Aspects of Biomedical Polymers. Plenum Press, New York (1985). M. Mrksich, C. S. Chen, Y. Xia, L. E. Dike, D. E. Ingber and G. M. Whitesides, Proc. Natl. Acad. Sci. USA 93, 10775 (1996). A. S. Curtis and J. V. Forrester, J. Cell Sci. 71, 17 (1984); C. H. Damsky and Z. Werb, Curr. Opin. Cell Biol. 4, 772 (1992). M. Takigawa, K. Tajima, H. O. Pan, M. Enomoto, A. Kinoshita, F. Suzuki, Y. Takano and Y. Mori, Cancer Res. 49, 3996 (1989). C. Picart, G. Ladam, B. Senger, J.-C. Voegel, P. Schaaf, F. J. G. Cuisinier and C. Gergely, J. Chem. Phys. 115, 1086 (2001). P. Tryoen-Toth, D. Vautier, Y. Haikel, J.-C. Voegel, P. Schaaf, J. Chluba and J. Ogier, J. Biomed. Mater. Res. (in press). Z. Darzynkiewicz, G. Juan, X. Li, W. Gorczyca, T. Murakami and F. Traganos, Cytometry 27, 1 (1997). K. Löster, C. Schüler, C. Heidrich, R. Horstkorte and W. Reutter, Anal. Biochem. 244, 96 (1997). G. Rainaldi, P. Filippini, A. Ferrante, P. L. Indovina and M. T. Santini, J. Biomed. Mater. Res. 55, 104 (2001). Y. Mitsui, K. Yamamoto, M. Yamamoto and K. Matuoka, Adv. Exp. Med. Biol. 190, 567 (1985). R. Herrero, M. R. Moncelli, R. Guidelli, M. Carla, A. Arcangeli and M. Olivotto, Biochim. Biophys. Acta 1466, 278 (2000). G. Ladam, P. Schaad, J. C. Voegel, P. Schaaf, G. Decher and F. Cuisinier, Langmuir 16, 1249 (2000). Ph. Lavalle, C. Gergely, F. J. G. Cuisinier, G. Decher, P. Schaaf, J. C. Voegel and C. Picart, Macromolecules (in press). K. Anselms, Biomaterials 21, 667 (2000). T. A. Horbett, Cardiovasc. Pathol. 2, 137 (1993). P. A. Underwood and F. A. Bennett, J. Cell Sci. 4, 641 (1989). D. W. Sommerfeldt, K. J. McLeod, C. T. Rubin and M. Hadjiargyrou, Biochem. Biophys. Res. Commun. 285, 255 (2001). K. Burridge, C. E. Turner and L. H. Romer, J. Cell Biol. 119, 893 (1992). M. Ouwens, H. M. M. Mikkers, G. C. M. Van der Zon, M. Stein-Gerlach, A. Ullrich and A. Maassen, Biochem. J. 318, 609 (1996). S. M. Bockholt and K. Burridge, J. Biol. Chem. 268, 14565 (1993). A. Krause, E. A. Cowles and G. Gronowicz, J. Biomed. Mater. Res. 52, 738 (2000).
Improved blood compatibility and decreased VSMC proliferation of surface-modified metal grafted with sulfonated PEG or heparin HEE JUNG LEE 1 , JONG-KYU HONG 1 , HYUN CHUL GOO 1 , WON KYU LEE 1 , KI DONG PARK 2 , SOO HYUN KIM 1 , YOUNG MI YOO 1 and YOUNG HA KIM 1,∗ 1 Biomaterials
Research Center, Korea Institute of Science and Technology, P.O. Box 131, Cheongnyang, Seoul 130-650, Korea 2 Department of Molecular Science and Technology, University of Ajou, Suwon 442-749, Korea Received 29 October 2001; accepted 28 May 2002 Abstract—Although the technique of coronary stenting has remarkably improved long-term results in recent years, (sub)acute thrombosis and late restenosis still remain problems to be solved. Metallic surfaces were regarded as thrombogenic, due to their positive surface charges, and stenosis resulted from the activation and proliferation of vascular smooth muscle cells (VSMCs). In this study, a unique surface modification method for metallic surfaces was studied using a self-assembled monolayer (SAM) technique. The method included the deposition of thin gold layers, the chemisorption of disulfides containing functional groups, and the subsequent coupling of PEG derivatives or heparin utilizing the functional groups of the disulfides. All the reactions were confirmed by ATR-FTIR and XPS. The surface modified with sulfonated PEG (Au-S-PEG-SO3) or heparinized PEG (Au-S-PEGHep) exhibited decreased static contact angles and therefore increased hydrophilicity to a great extent, which resulted from the coupling of PEG and the ionic groups attached. In vitro fibrinogen adsorption and platelet adhesion onto the Au-S-PEG-SO3 or Au-S-PEG-Hep surfaces decreased to a great extent, indicating enhanced blood compatibility. This decreased interaction of the modified surfaces should be attributed to the non-adhesive property of PEG and the synergistic effect of sulfonated PEG. The effect of the surface modification on the adhesion and proliferation of VSMCs was also investigated. The modified Au-S-PEG-SO3 or Au-S-PEG-Hep surfaces also exhibited decreased adhesion of VSMCs, while the deposited gold layer itself was effective. The enhanced blood compatibility and the decreased adhesion of VSMCs on the modified metallic surfaces may help to decrease thrombus formation and suppress restenosis. It would therefore be very useful to apply these modified surfaces to stents for improved functions. A long-term in vivo study using animal models is currently under way. Key words: Metallic surface modification; self-assembled monolayer technology; sulfonated PEG; heparin; protein/ platelet adhesion; smooth muscle cell adhesion. ∗ To
whom correspondence should be addressed. E-mail:
[email protected]
174
H. J. Lee et al.
INTRODUCTION
The implantation of metallic stents is widely accepted as post-operative therapy to percutaneous transluminal coronary angioplasty (PTCA) [1– 4]. Although the technique of coronary stenting has remarkably improved long-term results in recent years, (sub)acute thrombosis and late restenosis still remain problems to be solved. Generally, metallic materials are regarded as thrombogenic due to their positive surface charge and/ or the high surface free energy [3, 4]. In addition, late restenosis might result mainly from the activation and proliferation of vascular smooth muscle cells (VSMCs) [1– 4]. Surface-induced thrombosis is initiated by plasma protein adsorption and platelet activation. First the adsorbed proteins may serve as sites for platelet adhesion and activation by interactions with platelet membrane receptors. At the same time, coagulation factors in the plasma are activated to form a fibrin network. Heparin is a well-established anticoagulant used clinically to inhibit the formation of the fibrin network. There have been a number of approaches to utilize heparin by either slow-releasing systems or immobilization methods [5– 10]. Heparin has actually been applied to coronary stents, which have been successful, to some extent, in decreasing the associated (sub)acute thrombosis [8– 10]. Heparin has been also reported to decrease the proliferation of VSMCs [11, 12]. Surface modification using hydrophilic polymers, especially poly(ethylene glycol) (PEG), has been shown to decrease protein adsorption and platelet/ cell adhesion on biomaterials that make contact with blood [13– 30]. The role of PEG was explained by its unique properties, such as the excluded volume on the surface and the flexible hydrophilic chain motion to expel proteins/ cells, in addition to its nontoxicity and non-immunogenecity [14, 15, 23]. Methods for the surface modification by PEG have included simple physical adsorption, a self-assembled monolayer (SAM) [16– 18, 21, 28– 30], and chemical bond formation, such as chemical coupling [20, 22, 26] or graft polymerization [14, 24]. SAM techniques using silanes or gold/ sulfur compounds have been applied to modify glass, metal, and polymers for investigation as models in several areas [31– 33]. Whitesides et al. demonstrated that a SAM composed of oligo-PEG decreased protein adsorption [16– 18]. The surface grafting of PEG by SAM techniques has been investigated not only on polymers [14, 21, 27, 39], but also on metals and glass [20, 28– 30]. All of these studies demonstrated more or less decreased adsorption of proteins and adhesion of platelets, and therefore improved blood compatibilty. Furthermore, in our previous studies, sulfonated PEG (PEG-SO3) grafted polyurethane demonstrated enhanced blood compatibility and biostability to a larger extent than PEG alone; this may be attributed to a synergistic effect of the non-adhesive and mobile PEG chain motions combined with anticoagulant active sulfonate groups [25, 26, 34– 38]. In this study, PEG or sulfonated PEG was coupled on gold surfaces by SAM technology using functional sulfur compounds. Heparin was also immobilized
Improved blood compatibility and decreased VSMC proliferation
175
onto PEG chains containing functional groups. The surfaces coupled with PEG or heparin were characterized by static contact angles, attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy, X-ray photoelectron spectroscopy (XPS), and atomic force microscopy (AFM) [28]. In vitro protein adsorption, platelet adhesion, and VSMC proliferation were investigated to evaluate blood compatibility and the effect on restenosis.
EXPERIMENTS
Materials Stainless steel plates (SS, 1 × 3 cm or 0.5 × 0.5 cm; Kistech Co., Korea) were soaked in chromic acid and washed in distilled water three times before use. Gold-coated SS (Au) was prepared as follows: gold was deposited to a thickness of about 100 µm onto SS with an ion sputter (Hitachi E-1030, Japan), using an argon plasma at about 6 Pa under 15 mA with a deposition rate of about 11 nm / min. Au was soaked in a chromic acid solution for 3 days and then washed three times in distilled water in an ultrasonic bath. Dithiobutyric acid [TBA, (SCH2 CH2 COOH)2], cystamine dihydrochloride [cystamine, (SCH2 CH2 NH2 )2 · 2 HCl], taurine (2-aminoethanesulfonic acid), EDC [1-(3-dimethylaminopropyl)-3ethylcarbodiimide hydrochloride], and triethylamine (TEA) were purchased from Aldrich Chemical Co., USA. Heparin (sodium salt, grade I-A, from porcine intestinal mucosa), bovine serum fibrinogen (Fib, type I-S), fetal bovine serum (FBS), penicillin, and streptomycin were bought from Sigma, USA. Platelet-rich plasma (PRP) was donated by the Central Red Blood Center, Korea and vascular smooth muscle cells (VSMCs) from rat artery were provided by Dr. K. H. Chung, Cardiovascular Research Institute, Yonsei University College of Medicine, Korea. Bicinochroninic acid (BCA; Pierce Chemicals, USA) protein assay kit, Dulbeco’s modified Eagle’s medium F12 (DMEM /F12; Gibco-BRL, USA), bovine serum albumin (Gibco-BRL, USA), and Denacol EX-861 (PEG 1000 with epoxy groups at both ends; Nagase Chemical Ltd., Japan) were used. Diamino-terminated PEG (H2 N-PEG-NH2, molecular weight of PEG 1000; NOF Co., Japan) was dissolved in chloroform for purification, precipitated into diethyl ether, and dried in a vacuum. Then only one amino end-group of H2 N-PEG-NH2 was sulfonated by 1,3-propane sultone as described previously [25]. Briefly, 10% w /v propane sultone in tetrahydrofuran (THF) was added dropwise to 10% w /v H2 N-PEG-NH2 solution in THF and reacted at 50 ◦ C for 6 h. The resulting product precipitated as the reaction proceeded. The filtered product was washed with cold THF and dried overnight at room temperature. The structure of sulfonated PEG (H2 N-PEG-SO3) was confirmed by a nuclear magnetic resonance spectrometer (1 H-NMR, 13 C-NMR, Varian Gemini 200 MHz).
176
H. J. Lee et al.
Immobilizing sulfonated PEG or heparin on the gold surface Figure 1 shows a schematic illustration of the surface modification using dithiobutyric acid (TBA) or cystamine. In method (a) using TBA, Au was immersed in a solution of 100 mM TBA in methanol for 12 h to obtain Au-S-COOH. The mixture was immediately placed into a 20 wt% aq. solution of H2 N-PEG-SO3 containing EDC in a glass tube and the glass tube was sealed under N2 and kept at 60 ◦ C for given periods of time. The H2 N-PEG-SO3 immobilized gold (Au-S-PEGA-SO3) was washed with distilled water three times to remove unreacted PEG derivatives and dried in N2 . In method (b) using cystamine, Au was soaked in a solution of 100 m M cystamine in methanol at room temperature for 12 h. The cystamine-adsorbed gold (Au-SNH2 ) was then rinsed thoroughly with methanol and 0.1 mM KOH to remove the physically adsorbed cystamine and HCl, respectively. After rinsing and drying at 50 ◦ C, Au-S-NH2 was treated with a 100 mM solution of Denacol EX-861 in THF with a few ml of TEA at 50 ◦ C for 48 h. The Denacol-grafted gold plate (Au-S-PEG) was rinsed with THF and soaked in a solution of 50 mM taurine in formamide at 50 ◦ C for 48 h so that the residual epoxy group on Au-S-PEG reacted with taurine to introduce a sulfonate group (Au-S-PEG-SO3). In a similar way, Au-S-PEG having residual epoxy groups was immersed in a solution of 2 m M heparin in formamide to immobilize heparin (Au-S-PEG-Hep). The resulting gold plates, Au-S-PEGSO3 and Au-S-PEG-Hep were rinsed several times with formamide and ethanol and dried in N2 . Static contact angle (θS ) measurements The static contact angles were measured by an apparatus (Digidrop, GBX Scientific Instrument, France) fitted with a video camera (Nikon, Japan). Water drops of 10 µl prepared with a micro-syringe (Perfektum® , Popper & Sons Inc., Japan) were dropped onto three different points of each modified gold surface with a contact time t = 30 s. Attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy and atomic force microscopy (AFM) ATR-FTIR (Bruker FT-IR, IFS 66, Germany) was used to analyze the modified gold surfaces, using a KRS-5 crystal. The modified gold surfaces were observed by AFM (Park Scientific Instruments, CP, USA) in the tapping mode. X-ray photoelectron spectroscopy (XPS) XPS spectra of the modified gold surfaces were obtained on an ESCA 2803-S spectrometer (SSI, USA) with Al Kα X-rays. In order to determine O /C, N /C, S /C, and Na/C stoichiometries, collecting factors of 2.50, 1.68, 1.80, and 8.5 were
Improved blood compatibility and decreased VSMC proliferation
177
used for C1s, O1s, N1s, and S2p3, respectively. Binding energies were referenced to the C H group of the C1s peak components at 285 eV. In vitro fibrinogen adsorption test The modified gold plates were placed into 5 ml plastic disposable syringes and equilibrated with 2 ml of phosphate buffer solution (PBS) overnight. The buffer was replaced by 0.15 mg/ ml bovine serum fibrinogen (Fib), and the syringes were tapped to remove air bubbles, sealed and rotated in a shaking incubator (Daeil Engineering Co., Korea) at 37 ◦ C. By this method, the modified gold surfaces were constantly exposed to the protein solution. After 15 and 30 min, Fib-adsorbed gold plates were rinsed with PBS three times and sonicated in 1% SDS solution and centrifuged. The adsorbed protein concentration was determined using a BCA protein assay kit as described by the manufacturer. In brief, BCA working reagent was prepared by combining 50 parts of buffer solution A, 48 parts of BCA stock solution B (4% w /v BCA in water), and two parts of 4% w /v CuSO4 · H2 O. A set of protein standards (bovine serum albumin) was prepared using PBS as a diluent. A 100 µl volume of each standard sample was pipetted into a 96-well plate. The well plate was incubated at 60 ◦ C for 1 h. The absorbance of the water-soluble purple product was measured at 562 nm by an ELISA (enzyme-linked immunosorbent assay) apparatus (Spectra Max 340, Molecular Device Inc., USA). In vitro platelet adhesion test The surface-modified gold plates were incubated first with PBS and then with PRP by the same method as the above fibrinogen adsorption test. The number of platelets in PRP was 5 × 104 cells/ µl. The depleted platelets in the PRP were counted with a hemacytometer (Bright-Line® , Reichert, USA). The morphologies of the surfaces to which the platelets adhered were observed with a scanning electron microscope (SEM, HI-TACHI S-510). In vitro VSMC proliferation study Vascular smooth muscle cells (VSMCs) between the third and fifth passages were incubated at 37 ◦ C for 3 days in DMEM /F12 supplemented with 10% FBS, 100 U /ml penicillin, and 100 µg/ ml streptomycin. Meanwhile, the modified gold surfaces were sterilized with 70% ethyl alcohol for 24 h and rinsed with PBS three times before seeding. SMCs (number of cells 1×105 ) were gently seeded onto each plate located in a 24-well plate. After 1, 2, and 4 days, the cells on the surfaces were detached by trypsin– EDTA treatment and counted by a hemacytometer. As a control, a tissue culture dish (24-well plate, Falcon, USA) was used. The morphologies of the cells adhered on the surfaces were observed by SEM.
178
H. J. Lee et al.
RESULTS AND DISCUSSION
Characterization of the modified surfaces A number of methods have been studied for the incorporation of PEG derivatives onto glass, metallic, or polymeric surfaces. As glass and metals do not have functional groups, several approaches were studied: plasma coating or plasma polymerization onto the surfaces; coupling of various silane compounds; and chemisorption of sulfur compounds onto deposited gold layers. Self-assembled monolayers (SAMs) using silanes or sulfur compounds and their applications have been reviewed in detail by Ulman [33]. Silane coupling agents are well known to modify the surface of glass or metal, and recently several groups have demonstrated the incorporation of PEG, using silanes [20, 29, 46]. Gold and silver are very stable against oxidation and other chemical reactions, but they adsorb sulfur compounds very strongly to form quite stable coupling with thiolates, sulfides, or disulfides. Whitesides [16– 18] applied SAMs using thiolated oligo-PEG on gold surfaces to demonstrate the effect of PEG in decreasing protein adsorption; this technology was extended to several studies [28, 30, 39, 40]. Figure 1 illustrates the surface modification scheme. In method (a), dithiobutyric acid (TBA) was adsorbed onto gold layers to yield Au-S-COOH. The COOH
Figure 1. Schematic illustration of the surface modification process.
Improved blood compatibility and decreased VSMC proliferation
179
group introduced was utilized to react with the NH2 group of H2 N-PEG-SO3 to incorporate the sulfonated PEG (Au-S-PEGA-SO3), where EDC was used as a catalyst. The adsorption and reaction were confirmed by FTIR: a COOH peak at 1740 cm−1 or/ and SO3 group at 1342– 1352 cm−1 (asymmetric SO2 stretch), 1165– 1150 cm−1 (symmetric SO2 stretch) and 910– 895 cm−1 (S – O stretch), respectively, as shown in Figs 2a and 2b. In method (b), cystamine containing NH2 groups was adsorbed to obtain Au-S-NH2 , and PEG containing epoxy groups at both ends (Denacol EX-861) was coupled to yield Au-S-PEG via a reaction between NH2 and epoxy groups, which was catalyzed by an amine (TEA). In addition, the other residual epoxy group of PEG was further reacted with taurine to introduce sulfonate groups (Au-S-PEG-SO3 ), or to immobilize heparin (Au-SPEG-Hep). All the above reactions were confirmed again by ESCA; the surface atomic compositions are presented in Table 1. In Table 1, it can be seen that Au-SCOOH and Au-S-NH2 before coupling of the PEG derivatives exhibited relatively large S contents, but the values decreased substantially after the coupling of PEG. The C, O, and S concentrations of the Au-S-COOH surface were 66.8%, 23.7%, and 9.50%, respectively; those of the Au-S-PEGA-SO3 surface were 63.5%, 31.1%, and 5.40%. The grafted sulfonated PEG decreased the C and S concentrations, but increased the O content. These changes indicated the surface coverage of TBA or sulfonated PEG. In the case of Au-S-NH2 , Au-S-PEG, and Au-S-PEG-Hep, the changes in the surface atomic concentrations were similar, but N was not detected on the Au-S-PEG surface. It can be concluded from the FTIR and ESCA data that all the above reactions proceeded as planned.
Figure 2. ATR-FTIR spectra of the modified surfaces; (a) Au-S-COOH, (b) Au-S-PEG-SO3 .
180
H. J. Lee et al. Table 1. Elemental compositions of the modified gold surfaces with sulfonated PEG and heparinized PEG Atomic %a
Samples
Au-S-COOH Au-S-PEGA-SO3 Au-S-NH2 Au-S-PEG Au-S-PEG-SO3 Au-S-PEG-Hep
C
O
S
N
66.8 63.5 62.6 54.0 58.7 54.1
23.7 31.1 22.3 43.8 32.3 31.6
9.50 5.40 9.4 2.2 3.5 5.2
— — 5.8 — 5.5 9.1
O/Cb
S/Cb
0.36 0.49 0.36 0.81 0.55 0.58
0.14 0.085 0.15 0.04 0.094 0.097
a Analyzed b Surface
from survey scan spectra. atomic ratio. Table 2. Static contact angle (θS ) data of the modified gold surfacesa Samples
θS
SS Au Au-S-COOH Au-S-PEGA-SO3 SS Au Au-S-NH2 Au-S-PEG Au-S-PEG-SO3 Au-S-PEG-Hep
56.3 ± 1.5 26.1 ± 1.3 43.0 ± 1.9 wetting 56.3 ± 1.5 47.6 ± 2.6 51.5 ± 2.2 40.0 ± 2.3 13.9 ± 1.0 wetting
a Unit
degree, Average ± standard deviation, n = 5.
Such changes of the surface atomic concentrations described above were directly reflected by the water contact angles, which resulted from the change in surface hydrophilicity. Table 2 lists the static contact angles of the modified gold surfaces. The contact angle of SS was 56.3◦ and those of Au were 26.1◦ and 47.6◦ , indicating that the Au surface may be very sensitive to dust and impurities. Au-S-COOH and Au-S-NH2 showed a moderate decrease of the contact angles, similar to the case of Au-S-PEG containing epoxy end-groups. On the contrary, the Au-S-PEGASO3 , Au-S-PEG-SO3, and Au-S-PEG-Hep surfaces demonstrated very low contact angles, indicating the high hydrophilicity due to the PEG chains and attached ionic end-groups. Furthermore, Au-S-PEGA-SO3 and Au-S-PEG-Hep were completely wetted. Such an increased hydrophilicity of the modified gold surfaces again confirmed the actual surface modification processes. The topographies of the modified gold surfaces were examined by AFM. The control (Au) surface was relatively smooth and had a specific texture (Fig. 3a).
Improved blood compatibility and decreased VSMC proliferation
181
Figure 3. AFM images of the modified surfaces; (a) Au, (b) Au-S-PEG-SO3 , (c) Au-S-PEG-Hep: RMS roughness; (a) 2.231, (b) 1.070, (c) 1.129 nm.
Figures 3b and 3c show the images of the Au-S-PEG-SO3 and Au-S-PEG-Hep surfaces, respectively. They were even smoother than the control. The root-meansquare (RMS) value of the control Au surface was 2.231 nm, but those of Au-SPEG-SO3 and Au-S-PEG-Hep decreased to 1.070 and 1.129 nm, respectively. The RMS data revealed that the modified gold surfaces became even smoother due to the coverage of the incorporated PEG chains. SAM technology using the chemisorption of sulfur compounds onto gold layers provides various methods of introducing a wide range of functional groups on inert metallic surfaces. In this study, a unique method to couple PEG and/ or heparin onto a gold layer deposited on SS was investigated. The PEG or heparin compound grafted on the surface was quite stable during the various characterizations or measurements. Recently, several groups have investigated the stability of the coupling between gold and sulfur compounds [41– 43]. Wirde reported that the coupling was not so stable against oxidation in particular [41]. However, Lee observed that it was stable for a month at room temperature [43]. A study on the long-term stability should be carried out. In vitro protein adsorption and platelet adhesion test When biomaterials make contact with blood, proteins are first adsorbed instantaneously onto the surfaces and deformed. Then platelets adhere to the surfaces, are activated and aggregate, so that platelets may play a major role in thrombus formation. Therefore, a study on protein adsorption and platelet adhesion is the first step in evaluating the blood compatibility of biomaterials [23]. The effect of the modified gold surfaces on protein adsorption was examined using fibrinogen (Fib) as a model protein. The results are presented in Fig. 4. The concentrations of adsorbed Fib were measured after 15 min and 30 min, although the values were not very different. The amount of Fib adsorbed after 30 min onto the Au surface was 0.50 µg/ cm2 (89% of that on SS), slightly smaller than the value on SS (0.56 µg/ cm2 , 100%).
182
H. J. Lee et al.
Figure 4. Fibrinogen adsorption on the modified surfaces.
Figure 5. Platelet adhesion on the modified surfaces.
The Fib concentrations of Au-S-PEG-SO3 and Au-PEG-Hep surface decreased further to 0.23 (41%) and 0.20 µg/ cm2 (39%), respectively, demonstrating the effect of the introduced PEG-SO3 or PEG-Hep. Figure 5 shows the results of in vitro platelet adhesion onto the modified gold surfaces for 15 and 30 min. The percentage of adhered platelets for the 30 min period is higher than that for 15 min. However, the tendency of platelet adhesion
Improved blood compatibility and decreased VSMC proliferation
183
was very similar to protein adsorption. The Au surface presented a smaller percentage of adhered platelets (31% for 30 min) than SS (42%). The Au-SPEG-SO3 (20%) and Au-S-PEG-Hep (13%) surfaces exhibited much less platelet adhesion than the Au surface for 30 min. The Fib adsorption and platelet adhesion on the modified surfaces decreased in the order SS > Au > Au-S-PEG-SO3 > Au-PEG-Hep. Furthermore, the surface modifications using both sulfonated PEG and heparinized PEG were very effective in preventing protein adsorption and platelet adhesion onto the modified gold surface. This effect of PEG or heparin should be attributed to the specific function of PEG or hydrophilic heparin. Hydrophilic PEG chains grafted on surfaces inhibit the attachment of proteins or platelets by an excluded volume effect and flexible dynamic chain motions, as explained in other studies [14, 15]. However, the extent of the decrease of platelet adhesion, or, especially, protein adsorption, may be regarded as not very much compared with other studies. In our previous study on the protein adsorption of modified PU, the amounts of Fib adsorbed on both PU-PEG and PU-PEG-SO3 were much smaller than on PU; PU-PEG-SO3 exhibited a smaller Fib value but a higher amount of adsorbed albumin than those on PU-PEG [35], although PU-PEG-SO3 exhibited better in vivo blood compatibility than PU-PEG [34]. The advantage of the PEG-SO3 grafted system compared with only PEG may be the heparin-like anticoagulant activity of PEG-SO3 itself (14% of free heparin) [36], further decreased platelet adhesion, and increased hydrophilicity, indicating a synergistic effect of PEG chains and SO3 groups [34– 36]. In a different previous study on the platelet adhesion of modified PU, PU-PEG-SO3 always showed a smaller value than PUPEG for various MWs of PEG [25]. Heparin inhibits the function of thrombin and at the same time decreases platelet adhesion, possibly due to hydrophilicity first of all [7]. In vitro vascular smooth muscle cell (VSMC) proliferation study It was investigated whether the late restenosis associated with coronary stents resulted from the stimulation and proliferation of VSMCs after stenting in vivo [2, 3]. Heparin has been reported not only to act as an anticoagulant, but also to decrease VSMC proliferation in vitro and in vivo [11, 12], as heparin may interfere with the action of growth factors in addition to clotting factors or it may act directly on cells to inhibit their division. In addition, PEG and especially PEG-SO3 exhibited decreased interactions with cells and bacteria [25, 37, 44]. Therefore, in vitro VSMC adhesion and proliferation on the modified surfaces were investigated and the results are summarized in Fig. 6. The numbers of VSMCs that adhered to and grew on the modified gold surfaces were compared. When VSMCs (number of cells 1 × 105 ) were seeded on the 24-well culture plate (control), the number of SMCs was 1 × 105 at day 1, but increased to 1.5 × 105 at day 2 and 2.8 × 105
184
H. J. Lee et al.
Figure 6. Vascular smooth muscle cell adhesion and proliferation on the modified surfaces.
at day 4. However, when the same number of SMCs was seeded onto the surface of Au, Au-S-NH2 , Au-S-PEG-SO3, and Au-S-PEG-Hep, respectively, only about 10% of the total SMCs adhered to each surface and they did not proliferate at all. Actually, it was expected that Au-S-PEG-SO3 or Au-S-PEG-Hep might exhibit decreased adhesion of SMCs. We have previously reported that surfaces grafted with PEG or PEG-SO3 decreased the adhesion of corneal stromal cells [44] or bacteria [25], due to the non-adhesive property of PEG or PEG-SO3. Heparin has also been reported to decrease VSMC proliferation as discussed above [11, 12]. However, in this study VSMCs did not adhere on grow so much either on Au or on Au-S-NH2 , indicating a specific effect of the gold surface itself on the adhesion of SMCs. Actually, Sheardown reported that deposited gold surfaces revealed the adhesion of few cells [40]. In addition, gold-coated stents were claimed to be effective against allergies [45]. Therefore, the decreased adhesion of SMCs to Au or Au-S-NH2 might be due to the specific character of gold. The interaction of VSMCs with deposited gold, PEG-SO3 or introduced heparin should be further investigated in detail, maybe by animal models. The surface areas of the metallic sheets applied here might be too small to investigate differences in cell adhesion and proliferation. This unique surface modification method for metallic materials presented a successful incorporation of PEG and/ or heparin demonstrating decreased protein adsorption and platelet adhesion. In addition, the modified surfaces revealed decreased VSMC adhesion and maybe proliferation. Therefore, this method would be very useful to apply to stent coating. A long-term in vivo study using animal models is currently under way.
Improved blood compatibility and decreased VSMC proliferation
185
CONCLUSION
Although the technique of coronary stenting has remarkably improved long-term results in recent years, (sub)acute thrombosis and late restenosis still remain problems to be solved. Restenosis was suggested, due to the activation and proliferation of vascular smooth muscle cells. In this study, a unique surface modification method for metallic surfaces was studied using a SAM technique. The method included the deposition of thin gold layers, the chemisorption of disulfides containing functional groups, and the subsequent coupling of PEG derivatives or heparin utilizing the functional groups of the disulfides. All the reactions were confirmed by ATR-FTIR and XPS. The modified surface with sulfonated PEG (Au-S-PEG-SO3) or heparinized PEG (Au-S-PEG-Hep) presented very low contact angles or complete wetting, and therefore high hydrophilicity. In vitro fibrinogen adsorption and platelet adhesion onto the Au-S-PEG-SO3 or Au-S-PEGHep surfaces decreased to a great extent, indicating enhanced blood compatibility. Such a decreased interaction of the modified surfaces can be attributed to the nonadhesive property of PEG and the synergistic effect of the sulfonated PEG. The effect of the surface modification on the adhesion and proliferation of SMCs was also investigated. The modified Au-S-PEG-SO3 or Au-S-PEG-Hep surfaces also exhibited decreased adhesion of VSMCs, while the deposited gold layer itself was effective. The enhanced blood compatibility and less adhesion and maybe proliferation of VSMC on the modified metallic surfaces might help to decrease thrombus formation and suppress restenosis; therefore it would be very useful to apply to stents with an improved function. A long-term in vivo study using animal models is currently under way. Acknowledgement This work was supported by the Korean Ministry of Science and Technology National Research Laboratory Program No. N19610 and N21820. The authors appreciate Dr. K. H. Chung, Cadiovascular Research Institute, Yonsei University College of Medicine, Korea for supplying vascular smooth muscle cells. We thank also to Nagase Chemical Ltd., Japan for providing Denacol EX-861. REFERENCES 1. 2. 3. 4. 5. 6. 7.
T. Peng, P. Gibula, K. Yao and M. F. A. Goosen, Biomaterials 176, 685 (1996). D. L. Fischman, M. B. Lean, D. S. Daim et al., New England J. Med. 331, 496 (1994). J. C. Palmaz, Am. J. Roentgenol. 160, 613 (1993). P. A. Ribeiro, R. Gallo, J. Antonius et al., Am. Heart J. 125, 501 (1993). K. D. Park, T. Okano, C. Nojiri and S. W. Kim, J. Biomed. Mater. Sci. 22, 977 (1988). M. Pekna, R. Larsson, B. Formgren, U. R. Nilsson and B. Nilsson, Biomaterials 14, 189 (1993). D. K. Han, S. Y. Jeong and Y. H. Kim, J. Biomed. Mater. Res.; Appl. Biomaterials 23 (A2), 211 (1989). 8. W. J. van der Giessen, H. M. M. van Beusekom et al., Semin. Intervent. Cardiol. 3, 173 (1998). 9. R. Blezer, L. Cahalan, P. T. Cahalan and T. Lindhout, Blood Coagulation & Fibronolysis 9, 435 (1998).
186
H. J. Lee et al.
10. P. W. Serruys, H. Emanuelsson, W. J. van der Giessen et al., Circulation 93, 412 (1996). 11. A. D. Snow, R. P. Bolender, T. N. Wight and A. W. Clowes, Am. J. Pathol. 137, 313 (1990). 12. L. A. Pukac, G. M. Hirsch, J. C. Lormeau, M. Petitou, J. Choay and M. J. Karnovsky, Am. J. Pathol. 139, 1501 (1991). 13. E. W. Merill and E. W. Salzman, Am. Soc. Artif. Intern. Org. J. 6, 60 (1983). 14. Y. Mori, S. Nagaoka, H. Takiuchi, T. Kikuchi, N. Noguchi, H. Tanzawa and Y. Noshiki, Trans. Am. Soc. Artif. Intern. Org. 28, 459 (1982). 15. J. H. Lee, H. B. Lee and J. D. Andrade, Prog. Polym. Sci. 20, 1043 (1995). 16. C. Pale-Grosdemage, E. S. Simon, K. L. Prime and G. M. Whitesides, J. Am. Chem. Soc. 113, 12 (1991). 17. K. L. Prime and G. M. Whitesides, J. Am. Chem. Soc. 115, 10714 (1993). 18. P. Harder, M. Grunze, R. Dahint, G. M. Whitesides and P. E. Laibinis, J. Phys. Chem. B 102, 426 (1998). 19. P. Claesson, Colloids Surf. A 77, 109 (1993). 20. T. B. McPherson, H. S. Shim and K. Park, J. Biomed. Mater. Res. Appl. Biomater. 38, 553 (1997). 21. G. Mao, D. W. Grainger and D. Castner, Polym. Prep. Am. Chem. Soc. Div. Polym. Chem. 38, 289 (1997). 22. K. Holmberg, K. Bergstrom, C. Brink, E. Osterberg, F. Tiberg and J. M. Harris, J. Adhesion Sci. Technol. 7, 503 (1993). 23. W. R. Gombotz, W. Guanghui, T. A. Horbett and A. S. Hoffman, J. Biomed. Mater. Res. 25, 1547 (1991). 24. Y. X. Qui, D. Klee, W. Pluster, B. Severich and H. Hocker, J. Appl. Polym. Sci. 61, 2373 (1996). 25. H. J. Lee, K. D. Park, Y. H. Kim et al., Colloids Surf. B 18, 355 (2000). 26. Y. G. Ko, Y. H. Kim et al., Biomaterials 22, 2115 (2001). 27. F. Lim, X.-H. Yu and S. L. Cooper, Biomaterials 14, 537 (1993). 28. Y. H. Kim, H. J. Lee et al., Trans. Soc. Biomaterials 24, 7 (2001). 29. F. Zhang, E. T. Kang, K. G. Neoh and W. Huang, J. Biomater. Sci. Polymer Edn 12, 515 (2001). 30. F. Zhang, E. T. Kang, K. G. Neoh, P. Wang and K. L. Tan, Biomaterials 22, 1541 (2001). 31. C. D. Bain and G. M. Whitesides, Science 240, 62 (1988). 32. C. D. Bain and G. M. Whitesides, J. Am. Chem. Soc. 111, 7164 (1989). 33. A. Ulman, Chem. Rev. 96, 1533 (1996). 34. D. K. Han, S. Y. Jeong, Y. H. Kim, B. K. Min and H. I. Cho, J. Biomed. Mater. Res. 25, 561 (1991). 35. D. K. Han, G. H. Ryu, K. D. Park, U. Y. Kim, B. G. Min and Y. H. Kim, J. Biomed. Mater. Res. 30, 23 (1996). 36. D. K. Han, N. Y. Lee, K. D. Park, S. Y. Jeong, Y. H. Kim, H. I. Cho and B. G. Min, Biomaterials 16, 467 (1995). 37. D. K. Han, K. D. Park, S. Y. Jeong, Y. H. Kim, U. Y. Kim and B. G. Min, J. Biomed. Mat. Res. 27, 1063 (1993). 38. K. D, Park, Y. H. Kim et al., Biomaterials 19, 851 (1998). 39. Y. J. Du and J. L. Brash, Trans. Soc. Biomaterials 22, 71 (1999). 40. R. McMillan, B. Meeks, F. Bensebaa, Y. Deslandes and H. Sheardown, J. Biomed. Mater. Res. 54, 272 (2001). 41. M. Wirde, U. Gelius and L. Nyholm, Langmuir 15, 6370 (1999). 42. R. Voicu, T. H. Ellis, H. Ju and D. Leech, Langmuir 15, 8170 (1999). 43. Y. S. Shon and T. R. Lee, J. Phys. Chem. B 104, 8192 (2000). 44. M. K. Kim, I. S. Park, H. D. Park, W. R. Wee, J. H. Lee and Y. H. Kim, J. Cataract Refract. Surg. 27, 766– 774 (2001). 45. E. Alt, US Patent No. 5,824,045 (Oct. 20, 1998). 46. S. Jo and K. Park, Biomaterials 21, 605 (2000).
Characterization of poly(ethylene oxide) brushes on glass surfaces and adhesion of Staphylococcus epidermidis HANS J. KAPER 1 , HENK J. BUSSCHER 1,∗ and WILLEM NORDE 1,2 1 Department
of Biomedical Engineering, University of Groningen, Antonius Deusinglaan 1, 9713 AV Groningen, The Netherlands 2 Laboratory of Physical Chemistry and Colloid Science, Wageningen University, Dreijenplein 6, 6703 HB Wageningen, The Netherlands Received 28 January 2002; accepted 15 July 2002 Abstract—Poly(ethylene oxide) brushes have been covalently bound to glass surfaces and their presence was demonstrated by an increase in water contact angles from fully wettable on glass to advancing contact angles of 54◦ , with a hysteresis of 32◦ . In addition, electrophoretic mobilities of glass and brush-coated glass were determined using streaming potential measurements. The dependence of the electrophoretic mobilities on the ionic strength was analyzed in terms of a softlayer model, yielding an electrophoretic softness and fixed charge density of the layer. Brush-coated glass could be distinguished from glass by a 2– 3-fold decrease in fixed charge density, while both surfaces were about equally soft. Adhesion of Staphylococcus epidermidis HBH276 to glass in a parallel plate flow chamber was extremely high and after 4 h, 19.0 × 106 bacteria were adhering per cm2 . In contrast, the organisms did not adhere to brush-coated glass, with numbers below the detection limit, i.e. 0.1 × 106 per cm2 . These results attest to the great potential of polymer brushes in preventing bacterial adhesion to surfaces. Key words: Polymer brushes; poly(ethylene oxide); electrophoretic mobility; Staphylococcus epidermidis; bacterial adhesion.
INTRODUCTION
Biomaterials are widely used in modern medicine or the production of artificial organs and in a variety of intra- and extra-corporeal prostheses. However, their application can give rise to biomaterial-centered infections (BCIs) that defy treatment with antibiotics. Therefore, the only effective remedy appears to be the removal of the infected device [1]. The development of biofilms causing BCIs occurs in several ∗ To
whom correspondence should be addressed. Tel.: (31-50) 363-3140; Fax: (31-50) 363-3159; e-mail:
[email protected]
188
H. J. Kaper et al.
steps: firstly, a substratum surface is covered with a conditioning film of surfaceactive molecules, in particular proteins. Secondly, bacteria are transported towards the substratum and adhere. Initial adhesion is through long-range interactions, acting over distances up to a few tens of nanometers between the bacterium and the substratum, after which short-range interactions acting on a sub-nanometer scale become operative. The long-range interactions in bacterial adhesion may be calculated by applying concepts from colloid and interface science [2]. Thus, the DLVO theory, originally formulated to describe the adhesion of inanimate colloidal particles, has been applied to bacterial adhesion with varying success. According to the DLVO theory, the overall bacterium – substratum interaction is governed by contributions from Lifshitz– Van der Waals forces and forces resulting from overlapping electrical double layers. The description of short-range interactions (e.g. hydrogen bonding, ion pairing, hydrophobic interaction) depends on detailed knowledge of the chemistry for each surface involved and such information is usually not available for biological surfaces. Two strategies may be followed to reduce the risk of BCIs and to prevent or delay the adhesion of infectious bacteria to the substratum surface: (i) modifying the surface with (charged) groups that make the surface less attractive for the bacteria [3]; and (ii) introducing steric hindrance that keeps the bacteria at a distance from the surface where long-range attractive interaction forces are reduced to an ineffective magnitude [4, 5]. Steric hindrance may be achieved by decorating the surface with polymer molecules that are attached through an anchor to the surface, whereas the other part (the buoy) is moving freely in the surrounding medium. When the density of the polymer is high enough, the polymer molecules are forced to stretch out and the resulting layer is called a ‘molecular brush’ (see Fig. 1). The brush is essentially penetrable for solvent and low-molecular-weight ions, but depending on its packing density σ and extension L0 , may prevent the deposition of larger components such as protein molecules and bacteria [6]. So far, most of the research, both experimental and theoretical, on biomedical applications of molecular brushes has focused on the prevention of protein adsorp-
Figure 1. Schematic representation of polymer brush molecules. Upon increasing the grafting density, the polymer molecules are forced to stretch out.
Characterization of poly(ethylene oxide) brushes on glass surfaces
189
tion using poly(ethylene oxide) (PEO) as the polymer material. Work on bacterial adhesion includes the experiments by Park et al. [7], who grafted PEO molecules with different lengths and end-groups to a polyurethane surface, and Vacheethasanee and Marchant [5], who modified polyethylene and pyrolytic graphite substrata with a surfactant consisting of a poly(vinyl amine) backbone with grafted PEO and hexanal. In both experiments, surface modification with PEO molecules of sufficient length and grafting density was effective in preventing bacterial adhesion. Several theoretical models exist to predict the interaction of a tethered layer with incoming particles as a function of its parameters, such as the grafting density and polymer chain length [8– 11]. Jeon et al. [8, 9] assumed that the polymer chain is uniformly stretched with the end points located at the layer– solvent boundary. This corresponds to long and densely grafted polymer molecules. Szleifer [10] has pointed out that these assumptions do not hold in many practical situations. He has developed a model based on a self-consistent mean field approach. In this approach, interactions such as the intramolecular interactions between the monomer elements and the interactions between the brush polymer and the protein are taken into account. The model has been worked out for a PEO –lysozyme system. It was found that protein adsorption depends mainly on the grafting density, whereas the thickness of the grafted layer influences the kinetics of the adsorption process. It is doubtful whether this conclusion also holds for bacterial adhesion. Proteins are small enough to reach the substratum by diffusing through the layer. Bacteria, however, can adhere only by compressing the brush layer. Halperin [11] adopted a simple model for the polymer molecule, which was described as a string of non-interacting monomer elements. According to this model, an incoming particle may penetrate the brush and adsorb in the (absolute) primary minimum at the substratum, or it may be trapped in the secondary minimum at the aqueous edge of the brush. Various techniques, offering different degrees of control over the brush lengths versus densities, have been applied to attach the polymer molecules to a surface [12]. These techniques may be divided into non-covalent attachment, where the anchor has a high physico-chemical affinity for the substratum surface, and chemical grafting, where the anchor forms a chemical bond with surface groups. As the binding energy of a chemical bond exceeds that of a physical bond by approximately an order of magnitude, chemically grafted brushes are likely to be more robust. A technique that is often used for chemical grafting is the ‘grafting from’ method, where polymer molecules are grown from the substratum. This can be accomplished, for instance, by covering the surface with covalently linked initiators from which polymer chains grow in a solution containing the monomers, resulting in a polydisperse brush [13], or by using radio-frequency glow discharge plasma deposition to coat the surface with a thin, covalently bound polymer layer [14]. In the latter case, however, depending on the reaction conditions, the polymer molecules will form many cross-links and the resulting layer is sometimes referred to as a ‘surface gel’.
190
H. J. Kaper et al.
Figure 2. Vinyl-terminated poly(ethylene oxide) covalently bound to a hydrated silica surface.
Polymer molecules can be end-grafted to silicon and silica surfaces using an ether bond. This is an example of the ‘grafting to’ technique. Zhu et al. [15] employed a two-step procedure to link the hydroxyl end-group of poly(ethylene oxide) molecules with an ether bond to a silicon surface. Using an even simpler method, Maas et al. [16] linked vinyl-terminated polystyrene molecules to a silica surface by letting the vinyl groups react with the hydroxyl groups at a silica surface that appear as a result of the chemisorption of water to form ether bonds. Grafting densities, σ −1 , of 0.47 and 0.33 nm−2 were obtained using polystyrene of 2000 and 20 000 Da molar mass, respectively. These densities are high enough (i.e. σ −1 > RF2 , where RF is the Flory radius) to make the polystyrene chains overlap so that they stretch out to form a brush. The aim of this study was to graft methacryl-terminated monodisperse poly(ethylene oxide) to glass by linking its vinyl end-group to the hydroxyl groups on the glass surface (see Fig. 2), using the method described by Maas et al. [16], and to assess the usefulness of the resulting brush in preventing initial bacterial adhesion as the onset of infection [1]. The grafted polymer layers were physico-chemically characterized by water contact angles and streaming potentials. The adhesion experiments were carried out in a parallel plate flow chamber using S. epidermidis HBH 276. This bacterial species was chosen because it is among the most relevant strains found in BCIs.
MATERIALS AND METHODS
Poly(ethylene oxide) (PEO) brushes Methacryl-terminated poly(ethylene oxide) with a molar mass of 9800 Da, corresponding to approximately 250 monomer units, and a polydispersity index less than 1.03 was purchased from Polymer Source (Dorval, Quebec, Canada) and used as received. Microscope glass slides of size 76 × 26 × 1 mm (Menzel-Gläser) were used as a substratum surface. The slides were first sonicated in 2% RBS detergent (Omnilabo International, Breda, The Netherlands), rinsed in warm tap water and
Characterization of poly(ethylene oxide) brushes on glass surfaces
191
demineralized water, sonicated again in methanol, and rinsed in demineralized water, to remove oil contamination and fingerprints. Next, possible metallic oxides on the glass surface were removed by submersing the slides in hot (95 ◦ C) nitric acid (65%; Merck, Darmstadt, Germany) for 45 min. Finally, the slides were extensively rinsed with demineralized water and Millipore-Q water and dried in a heat box at 80 ◦ C for 5 h. To graft the PEO chains on the glass surfaces, the slides were covered with a solution of the methacryl-terminated PEO in chloroform (4 mg/ ml). The solvent was evaporated in a stream of nitrogen, after which the slides were annealed overnight in a vacuum at 145 ◦ C. Prior to experiments, excess material was removed by washing the slide with Millipore-Q water and the slides were dried in a stream of nitrogen. Water contact angle measurements Water contact angles were measured at room temperature with a home-made contour monitor using the sessile drop technique. Advancing and receding water contact angles were obtained by keeping the syringe needle in the water droplet (1– 1.5 µl) after positioning it on the surface and by carefully moving the sample until the advancing angle was maximal. Contact angles with water droplets at rest will be referred to as equilibrium contact angles. On each sample, at least ten droplets were placed at different positions. Streaming potential measurements For a solid surface in contact with a liquid, streaming potentials Estr , arising from a forced flow of the liquid under the influence of a pressure p, depend on the electrokinetic potential ζ at the solid– liquid interface according to Estr εε0 = ζ, p ηκsp
(1)
where εε0 is the dielectric permittivity, η is the viscosity and κsp is the specific conductivity of the liquid. By comparison with the electrokinetic model of colloidal particles, the electrophoretic mobility µ for a flat surface is related to the electrokinetic potential ζ by [17] εε0 ζ. (2) µ= η Combining equations (1) and (2) yields the connection between µ and Estr : µ = κsp
Estr . p
(3)
The pressure dependence of the streaming potentials was measured employing a parallel plate flow chamber [18]. The walls of the flow chamber were brush-coated
192
H. J. Kaper et al.
microscope slides (26 × 76 mm) separated by a 0.2 mm Teflon gasket, while two rectangular platinum electrodes (5.0 × 25.0 mm) were located at both ends of the parallel plate flow chamber. Streaming potentials were measured in KCl solutions with ionic strengths of 5, 10, 15, 25, 50, 75 and 100 m M at ten different pressures ranging from 50 to 200 hPa. Each pressure was applied for 10 s in both directions. Ohshima et al. [19] have proposed a model for the electrophoretic mobility of particles with fixed surface charges distributed across an ion-penetrable, porous layer. It unites the theory of electrophoresis of coiled structures of highly solvated polyelectrolytes with that of rigid spheres where the surface charge is located in an infinitesimal thin layer at the surface. In the Ohshima model, the ion-penetrable layer is characterized by its fixed charge density ρ and a parameter 1/λ, referred to as the electrophoretic ‘softness’ of the ion-penetrable layer, which depends on the frictional force exerted on the water when it flows through the ion-penetrable layer. For planar surfaces, under the conditions that (a) the charge densities are relatively low, (b) 1/λ is less than the thickness over which the liquid flow penetrates the soft surface layer and (c) the Debye length, κ −1 , is less than the thickness of the ion-penetrable layer (all being fair assumptions for PEO-grafted surfaces and to a somewhat lesser extent also for glass having a porous, jelly surface in aqueous media with a wide range of ionic strengths), the electrophoretic mobility as a function of the reciprocal Debye length κ is approximated by 2 λ 1 + λ/2κ ρ . (4) µ= 2 1+ ηλ κ 1 + λ/κ For symmetrical electrolytes, κ is related to the ionic strength as κ2 =
2F 2 ci zi2 , εε0 RT
(5)
where F is the Faraday constant, T is the absolute temperature, R is the gas constant, zi is the valency and ci is the concentration of ion i. The most salient feature of equation (4) is the fact that in contrast to the rigid surface model, the electrophoretic mobility does not approach zero as the electrolyte concentration increases. A least-squares fit of electrophoretic mobilities measured as a function of the ionic strength to equation (4) allows the evaluation of the softness of the polymer layer and the space charge density in the soft part of the layer. This is based on the assumption that the values for 1/λ and ρ are invariant with ionic strength. All electrophoretic mobilities and softness values reported are the mean values of three different measurements with separately prepared brushes. Bacterial strain and adhesion experiments Staphylococcus epidermidis HBH276 was cultured in tryptone soya broth (OXOID, Basingstoke, UK). First, a strain was streaked from a frozen stock and grown overnight at 37 ◦ C on a blood agar plate. A colony was used to inoculate 5 ml of
Characterization of poly(ethylene oxide) brushes on glass surfaces
193
Figure 3. Schematic view of the parallel plate flow chamber used in this study.
growth medium, which was incubated at 37 ◦ C in ambient air for 24 h and used to inoculate a second culture in 150 ml of growth medium that was grown for 17 h. The bacteria from the second culture were harvested by centrifugation (5 min, 5000 g) and washed twice with Millipore-Q water. Subsequently, the bacteria were resuspended in PBS (pH 6.8). The suspension was sonicated on ice (10 s) to disrupt aggregates. The concentration of bacteria was determined using a Bürker-Türk counting chamber and adjusted to 3 × 108 bacteria/ ml. Adhesion experiments were carried out using a flow chamber and an image analysis system. Figure 3 shows an exploded view of the flow chamber used. The Teflon gasket between the upper and the lower part of the flow chamber determines the dimensions of the flow channel (175 × 17 × 0.75 mm). The top and bottom collector plates have the dimensions of a common microsope slide: 76×26×1 mm. The top slide is made out of glass. The bottom slide is covered with the surface under study and microbial adhesion can be directly observed using a phase-contrast microscope equipped with a 40× ultra-long working distance objective. A pulse free flow (0.0325 ml/s) was created by hydrostatic pressure and the suspension was recirculated using a peristaltic pump. By means of a valve system, the flasks containing buffer or bacterial suspension can be connected with the flow chamber. Prior to an experiment, all air bubbles were removed from the tubing and the flow chamber and the system was perfused for 60 min with buffer. Subsequently, flow was switched to the bacterial suspension. During deposition, images of the bottom plate were recorded using a 512 × 512 pixel CCD camera and the bacteria present on the surface were counted by dedicated image analysis software [20]. On the one hand, the lower detection limit of the system is determined by the number of CCD pixels per bacterial cell that are necessary for the automatic counting of the bacteria, and on the other hand, by the statistical error in the counting
194
H. J. Kaper et al.
of the bacteria that one is willing to accept. For the current system, a statistical error of 10% results in a lower detection limit of approximately 0.1 × 106 bacteria per cm2 . The initial adhesion experiments were carried out for 4 h, with bacteria suspended in PBS. The initial increase in the number of adhering bacteria was linearly extrapolated to t = 0 to obtain the initial bacterial deposition rate j0 , which represents the number of bacteria transported by convection and diffusion towards the substratum surface that subsequently have been able to adhere. RESULTS AND DISCUSSION
Physico-chemical characterization of the polymer brush Cleaned glass, i.e. the glass surface prior to application of the polymer, was fully wettable with water, while after application of the polymer, the surface was more hydrophobic (see Table 1). The equilibrium water contact angle increased to 43◦ and the advancing and receding water contact angles were 54◦ and 22◦ , respectively. This contact angle hysteresis either suggests [21] partial coverage of the glass surface by the brush, or attests to the mobility of the polymer chains in the brush. The advancing angle and the hysteresis are somewhat higher than the values measured by Park et al. [7], who reported advancing and receding angles of 44.6◦ and 30.2◦ on a polyurethene surface covered PEO with a molecular mass of 1000 Da. Harder et al. [22] found advancing water contact angles of 30– 35◦ for oligo(ethylene oxide). Our PEO-coated surfaces are probably more hydrophobic than coatings made of smaller molecules, because small molecules possess a relatively higher fraction of hydroxyl to carbon groups, creating a more hydrophilic coating. Electrophoretic mobilities Figure 4 shows the electrophoretic mobilities of clean and brush-covered glass as a function of the ionic strength. Both surfaces have a finite electrophoretic Table 1. Physico-chemical characteristics, including water contact angles and electrophoretic properties according to a soft-layer model of polymer brushes on glass, together with adhesion of Staphylococcus epidermidis in a parallel plate flow chamber Water contact angles
Glass Brush
Electrophoretic properties
Bacterial adhesion
θw,eq (degrees)
θw,adv (degrees)
θw,rec (degrees)
ρ (106 C m−3 )
λ−1 (nm)
j0 (cm−2 s−1 )
n4h (106 cm−2 )
Wetting 43 ± 3
Wetting 54 ± 4
Wetting 22 ± 4
−1.5 ± 0.3 −0.6 ± 0.03
3.3 ± 0.3 3.6 ± 0.1
1860 ± 120 bd
19.0 ± 3 bd
± indicates the SD over three separately prepared substratum surfaces, while in the case of bacterial adhesion, also three separately grown cultures were employed. bd, below detection.
Characterization of poly(ethylene oxide) brushes on glass surfaces
195
softness, as can be inferred from the fact that both curves tend to a non-zero electrophoretic mobility at elevated ionic strength. The electrophoretic softness of glass was calculated to be 3.3 nm and that of the brush 3.6 nm (Table 1). The results indicate that glass has a water- and ion-penetrable surface, which is in line with earlier measurements reporting thicknesses of ion-penetrable layers on glass of about 0.7 nm [23], 1.9 nm [24] and 4 nm [25]. Surprisingly, the softness of the brush hardly differs from that of the glass surface, which may point to strong immobilization of water in the polymer matrix of the brush. The soft outer region of the brush has a much lower charge density (ρbrush = (−0.6 ± 0.03) × 106 C /m3 ) than in the soft glass layer (ρglass = (−1.5 ± 0.3) × 106 C /m3 ). This result is expected because PEO is essentially uncharged. The thickness L0 of the brush layer can be estimated using the equation [11] 1/3 , (6) L0 ≈ aN σ a 2 where a is the length of a monomer unit and N is the number of monomer units. The applicability of this equation to approximate the thickness of PEO brushes has recently been demonstrated by Efremova et al. [26]. As we used the same grafting technique as and comparable materials to Maas et al. [16], we assume a similar grafting density. Adopting the grafting density that Maas et al. found for Mw = 20 000 (i.e. σ −1 = 0.33 nm−2 ) and inserting N = 250 and a = 0.278 nm [27], L0
Figure 4. Electrophoretic mobilities of glass and brush-coated glass as a function of the ionic strength of a KCl solution. The lines indicate the best fit to equation (4), i.e. the Ohshima soft-layer model.
196
H. J. Kaper et al.
is calculated to be 21.6 nm. Hence, every polymer molecule in the brush occupies a volume of L0 /σ equal to 65 nm3 . The molar volume of an EO segment is equal to 38.9 cm3 [28]. Therefore, the volume taken by the monomer segments of the PEO molecule is 16 nm3 , corresponding to approximately 25% of the total brush volume. This fraction is high enough to ensure that virtually every water molecule in the brush layer is located in the vicinity of a polymer molecule. Since ethylene oxide is known for its strong hydration, it is thus plausible that the water is strongly immobilized in the soft brush layer. Bacterial adhesion In Fig. 5, a representative example of the results of a bacterial adhesion experiment is given. From these plots one can determine the numbers of S. epidermidis adhering after 4 h, as well as the initial deposition rates j0 , which are calculated by linear regression of adhesion numbers during the first 30 min of desposition. Table 1 summarizes the results of the bacterial adhesion experiments quantitatively. The numbers of S. epidermidis adhering after 4 h to pristine glass are several orders of magnitude higher than to the brush-coated glass. In fact, the brush effectively decreases bacterial adhesion to below the detection limit (0.1 × 106 per cm2 ) for direct counting in the parallel plate flow chamber. The results appear to be consistent with other experiments on bacterial adhesion to ‘brush-like’ layers, although the experiments differ in surface modification techniques and/ or experimental methods to assess the efficacy of the surface layer, making a fair comparison difficult. Hendricks et al. [14] tested a plasma-deposited PEO coating
Figure 5. Example of the deposition kinetics of S. epidermidis HBH276 to glass with (circles) and without (triangles) a PEO brush.
Characterization of poly(ethylene oxide) brushes on glass surfaces
197
Figure 6. Micrograph of S. epidermidis HBH276 adhering to a glass slide, partly covered with a PEO brush. The bar marker corresponds to 10 µm (micrograph by courtesy of A. Roosjen).
on polyetherurethane under flowing conditions and reported reductions of over 99% in the initial adhesion of Pseudomonas aeruginosa with respect to a control, as well as over 90% reductions after 18 h of growth. Park et al. [7] also reported reductions of 90% in the adhesion of S. epidermidis on a polyurethane surface with PEG3.4 kDa modification after allowing the bacteria to grow for 24 h. Surprisingly, a PEG-1 kDa-modified substrata performed significantly worse for S. epidermidis. Finally, in order to illustrate the effectiveness of the brush layer in reducing bacterial adhesion, Fig. 6 shows a micrograph of a partly brush-coated glass slide after 4 h of exposure to a bacterial suspension in the parallel plate flow chamber. Clearly, far less bacteria have adhered to the brush-covered surface than to the bare hydrophilic glass.
CONCLUSIONS
Based on this study, the following conclusions can be drawn: (1) polymer brushes applied on a glass surface can be distinguished from untreated glass by an increased hydrophobicity and a decreased fixed charge density, as derived from an electrokinetic soft-layer model;
198
H. J. Kaper et al.
(2) poly(ethylene oxide) brushes on a glass substratum strongly discourage the adhesion of an S. epidermidis strain. Hence, it can be anticipated that polymer brushes will constitute effective non-adhesive coatings for the control of BCIs.
REFERENCES 1. A. G. Gristina, Science 237, 1587 (1987). 2. M. Hermansson, Colloid. Surf. B: Biointerfaces 14, 105 (1999). 3. B. Gottenbos, H. C. van der Mei, H. J. Busscher, D. W. Grijpma and J. Feijen J. Mater. Sci.: Mater. Med. 10, 853 (1999). 4. A. Razatos, Y.-L. Ong, F. Boulay, D. L. Elbert, J. A. Hubbell, M. M. Sharma and G. Giorgiou, Langmuir 16, 9155 (2000). 5. K. Vacheethasanee and R. E. Marchant, J. Biomed. Mater. Res. 50, 302 (2000). 6. D. Leckband, S. Sheth and A. Halperin, J. Biomater. Sci. Polymer Edn 10, 1125 (1999). 7. K. D. Park, Y. S. Kim, D. K. Han, Y. H. Kim, E. H. B. Lee and K. S. Choi, Biomaterials 19, 851 (1998). 8. S. I. Jeon, J. H. Lee, J. D. Andrade and P.-G. de Gennes, J. Colloid Interface Sci. 142, 149 (1991). 9. S. I. Jeon and J. D. Andrade, J. Colloid Interface Sci. 142, 159 (1991). 10. I. Szleifer, Biophys. J. 72, 595 (1997). 11. A. Halperin, Langmuir 15, 2525 (1999). 12. B. Zhao and W. J. Brittain, Prog. Polym. Sci. 25, 677 (2000). 13. O. Prucker and J. Rühe, Macromolecules 31, 592 (1998). 14. S. K. Hendricks, C. Kwok, M. Shen, Th. A. Horbett, B. D. Ratner and J. D. Bryers, J. Biomed. Mater. Res. 50, 160 (2000). 15. X.-Y. Zhu, Y. Jun, D. R. Staarup, R. C. Major, S. Danielson, V. Boiadjiev, W. L. Gladfelter, B. C. Bunker and A. Guo, Langmuir 17, 7798 (2001). 16. J. H. Maas, M. A. Cohen Stuart, A. B. Sieval, H. Zuilhof and E. J. R. Südhölter, Thin Solid Films (in press). 17. P. C. Hiemenz and R. Rajagopalan, in: Principles of Colloid and Surface Chemistry, 3rd edn, Ch. 12. Marcel Dekker, New York (1997). 18. R. J. van Wagenen and J. D. Andrade, J. Colloid Interface Sci. 76, 305 (1980). 19. H. Ohshima, Colloids Surfaces A: Physicochem. Eng. Aspects 103, 249 (1995). 20. J. M. Meinders, J. Noordmans and H. J. Busscher, J. Colloid Interface Sci. 152, 265 (1992). 21. C. W. Extrand, J. Colloid Interface Sci. 207, 11 (1998). 22. P. Harder, M. Grunze, R. Dahint, G. M. Whitesides and P. E. Laibinis, J. Phys. Chem. B 102, 426 (1998). 23. J. M. Kleijn, Colloids Surfaces 51, 371 (1990). 24. A. T. Poortinga, R. Bos and H. J. Busscher, Colloids Surfaces B: Biointerfaces 20, 105 (2001). 25. D. J. Shaw, Colloid & Surface Chemistry, 4th edn. Butterworth-Heinemann, Oxford (1992). 26. N. V. Efremova, S. R. Sheth and D. E. Leckband, Langmuir 17, 7628 (2001). 27. M. Morra, J. Biomater. Sci. Polymer Edn 11, 547 (2000). 28. D. W. van Krevelen, Properties of Polymers, 2nd edn. Elsevier Scientific Publishing, Amsterdam (1976).
Tissue-culture surfaces with mixtures of aminated and fluorinated functional groups. Part 2. Growth and function of transgenic rat insulinoma cells (βG I /17) JAMES R. BAIN 1,2,∗ and ALLAN S. HOFFMAN 3 1 Sarah
W. Stedman Center for Nutritional Studies, Duke University Medical Center, Durham, NC 27710, USA 2 Department of Pharmacology and Cancer Biology, Duke University Medical Center, Mail Stop DUMC 3813, LSRC Building, Room C348, Durham, NC 27710, USA 3 Department of Bioengineering, Box 352255, University of Washington, Seattle, WA 98195, USA Received 10 January 2002; accepted 12 December 2002 Abstract—Interactions of transplantable cells with synthetic polymers can influence the function of biohybrid artificial organs. This study explored growth and secretion of human insulin by βG I/17 cells cultured on surfaces bearing diamine groups (N2), trifluoropropyl groups (F3) and mixtures of the two. Cells cultured on high-F3 and high-N2 surfaces spread well, grew rapidly and produced >1.8 mol lactate per mol glucose consumed, closely resembling cells grown on the permissive control, glass. On one mixed surface, with a molar ratio of 33 N2 groups : 67 F3 groups, cells had a lower lactate/ glucose ratio, adopted a rounded form, grew slowly and were quick to form emergent aggregates, similar to cultures on the inhibitory control, untreated polystyrene. Cultures on surfaces with higher F3 content secreted the most insulin and, in the case of the highest-F3 surface, showed improved responsiveness to secretagogues. Hormone secretion was roughly 50% greater when cells were grown on F3 surfaces conditioned by earlier cultures of βG I/ 17. Incubation of conditioned surfaces with high concentrations of a polyclonal anti-laminin serum prior to re-plating partially abolished this improvement in secretory function. Polymers bearing trifluoropropyl groups appear to be attractive candidates for use in the artificial endocrine pancreas. Surface coatings that include laminin might promote function of transgenic insulinoma cells in vitro and in vivo. Key words: Artificial endocrine pancreas; transgenic insulinoma cells; cell culture; surface modification; insulin; glucose.
∗ To
whom correspondence should be addressed. Tel.: (1-919) 613-8652; Fax: (1-919) 668-6044; e-mail:
[email protected]
200
J. R. Bain and A. S. Hoffman
INTRODUCTION
Surgeons might someday be able to restore function of diseased tissues by implanting composites of living cells and artificial materials. Since chemistry and morphology of biomaterials can influence the phenotype of transplantable cells, knowledge of these interactions is fundamental to the creation of successful biohybrid organs, including the artificial endocrine pancreas. Type I or insulin-dependent diabetes mellitus arises from autoimmune destruction of the β cells of the pancreatic islets of Langerhans. Millions of humans suffer from Type-I diabetes and the incidence is increasing in diverse societies worldwide [1– 4]. Despite massive investments in research in recent decades, the causes of this devastating disease remain obscure. Insulin-replacement therapy by injection has been the standard treatment for decades. Though injection therapies improve the quality and length of life, they are not able to provide the close control of blood glucose necessary to prevent diabetic complications in later life, including damage to the kidneys, nerves, blood vessels and eyes [3]. Treatment of human patients by transplantation of islets or whole pancreata from humans (allografts) and pigs (xenografts) has had variable success. Challenges include limited donor supply, immune reactions and the fastidious nature and limited growth potential of human β cells [5– 16]. To overcome these challenges, proliferative cells capable of processing and secreting human insulin are attracting interest as platforms for the creation of transplantable β-cell surrogates by genetic engineering or by manipulation of stem cells. Popular candidates include insulinoma cell lines, which are derived from rare β-cell tumors (Fig. 1), along with other cell types that share the insulinoma’s secretory phenotype [3, 5, 7, 10, 11, 13, 14, 16– 31]. Concomitantly, synthetic materials are being developed to serve as anchorage supports or immune-protection barriers to foster survival and function of β-cell surrogates in the human body. Hybrid artificial organs employing mitotically-expanded islets, insulin-secreting cell lines, or fragments of insulinoma tumors have been grown in vitro, implanted in laboratory animals and used to treat several cases of human diabetes [6, 8, 9, 12, 14, 18, 32– 40], but little is known about the effects of device chemistry and design on the phenotype of transplantable insulinoma cells [29, 38– 43]. In the present study, we examined the growth and phenotype of βG I/ 17, a line of genetically modified rat insulinoma cells (RINs), cultured on borosilicate glass modified with organosilane coupling agents to create surfaces rich in diamine groups, trifluoropropyl groups and mixtures of the two [44]. Portions of this report are reproduced from our preliminary presentation [45]. Note that certain control data on cell metabolism are common to Ref. [43] and the present study, since the two investigations were concomitant. MATERIALS AND METHODS
Culture surfaces As described in the accompanying paper [44], borosilicate glass discs were silanized to create culture surfaces bearing diamine groups (N2), trifluoropropyl groups
Insulinoma culture on aminofluorosiloxanes
201
Figure 1. Presumed interrelationships among rat insulinoma (RIN) cell lines discussed in the text. All derive from a single, radiation-induced tumor in the New England Deaconess Hospital (NEDH) rat. These and related RIN lines have become mainstays of diabetes research, worldwide. RIN line βG I/ 17 was used in the present study.
(F3) and mixtures of the two. Etched and silanized glass discs were mounted for cell culture in 24-well, surfaced-oxidized, tissue-culture plates (Falcon Multiwell 3047, Becton Dickinson Labware, Franklin Lakes, NJ, USA). Discs were held in place by fluoroelastomeric o-rings made of poly(vinylidene fluoride-cohexafluoropropylene), size 014 (Viton A, Kontes Glass Company, Vineland, NJ, USA), as described earlier [43]. Cell line and cell culture The lineage leading to the engineered insulinoma cell line βG I/ 17 began with a rat insulinoma (RIN) that developed in an NEDH rat (New England Deaconess
202
J. R. Bain and A. S. Hoffman
Hospital) after high-dose X-irradiation [46] (Fig. 1). The cell line, RINr, was established from this insulinoma after nine serial tumor transplants in NEDH rats [7, 47]. At low passage, a RINr clone, RINr1046-38, or RIN-38 for short, demonstrates glucose-induced secretion of native rat insulin [48, 49]. Genetic engineers at BetaGene (Dallas, TX, USA), created βG I/ 17 from RIN-38 by transfection with a single human proinsulin cDNA (Fig. 1), as confirmed by Southern-blot analysis [50]. Transgene expression is driven by the cytomegalovirus promoter/ enhancer element. βG I/ 17 contains approx. 410 ng insulin per 6 µg DNA, more than 10 times that of the parental RIN-38 line and is capable of processing human proinsulin to insulin [50, 51]. βG I/ 17 expresses transgenic aminoglycoside phosphotransferase, which confers neomycin resistance [50]. Its transgenic phenotype has been maintained after a year of continuous culture [50]. In the present study, we used βG I/ 17 used at the 53rd passage after it received the human proinsulin gene. We maintained it under selection with the neomycin analog, G418 sulfate (Geneticin, Gibco-BRL, Life Technologies, Grand Island, NY, USA), until the 27th passage. Phenotypic expression was stable at the 53rd passage, as demonstrated by background production of human insulin. Immediately prior to use in this study, βG I/ 17 was shown to be free of contamination with mycoplasmata by the bisbenzimide 33258 fluorochrome (Hoechst, Sigma, St. Louis, MO, USA). For this study, near-confluent cultures of βG I/17 on tissue-culture polystyrene were rinsed, trypsinized, re-suspended, counted and plated at a density of 80 000 cells/ cm2 , approximately one fifth of confluence, as described [43]. Cultures proceeded 7 days in a humidified incubator set to 95% air, 5% CO2 at 37◦ C, with daily changes of media. Conditioned media were frozen for later assays of metabolites. Basal and maximal insulin secretion at seven days After one week, constitutive insulin secretion was evaluated under 100 µM diazoxide (DZ; Sigma), a strong inhibitor of regulated insulin secretion [52]. Cultures were subsequently stimulated for 1 h with a ‘Swiss cocktail’ of secretagogues to assess their maximum potential for insulin secretion. As previously described [43], this cocktail included 3-isobutyl-1-methylxanthine or IBMX, Larginine, D-glucose, L-leucine, L-glutamine and carbamylcholine chloride or carbachol (an acetylcholine analogue). Finally, samples were fixed in 10% neutral, buffered formalin, stained with Harris’s hematoxylin and eosin, inverted and mounted on microscope slides. Assays of glucose, lactate and insulin Media conditioned by the βG I/17 cultures were frozen in polypropylene microcentrifuge tubes, thawed and assayed for glucose consumption and lactate production as described [43]. Insulin levels were assessed in the conditioned media from the seventh day of routine culture, the basal incubation under diazoxide and the Swiss-
Insulinoma culture on aminofluorosiloxanes
203
cocktail stimulation, using the radioimmunoassay (RIA) kit from Incstar (Stillwater, MN, USA) [43]. In subsequent studies of surface conditioning (infra), insulin was assayed with a different RIA kit (Coat-A-Count , Diagnostic Products, Los Angeles, CA, USA). Both RIA kits have calibration histories traceable to the World Health Organization insulin standard 66/304. Where necessary, insulin secretion is expressed in molar units using a molecular mass of 5.808 kDa for fully processed human insulin. Surface conditioning by insulinoma cells After conducting the cell-culture studies described above, we investigated whether conditioning of surfaces by an initial culture of insulinoma cells had any effects on the behavior of subsequent cultures. For this second round of studies, the transgenic insulinoma cell line, βG I/17, was cultured on the pure F3 surface for seven days, as described above and then stripped from the surface without the use of enzymes. This was accomplished by rinsing the cultures twice with Hanks’ balanced-salt solution without divalent cations (HBSS; Gibco-BRL) and then incubating them under a proprietary, enzyme-free, chelating solution (Gibco-BRL, number 13151014). Two sequential chelations were performed, each followed by a firm smack of the plate onto the lab bench. This caused most cells to break free into the overlying liquid. Cells so liberated were removed by pipette and discarded. The very few cells remaining on the surface were then lysed with 20 mM NH4 OH (sterilefiltered) for 5 min at room temperature, a procedure that previous workers developed to remove cells while at the same time preserving extracellular matrix (ECM) molecules in a functional state [53, 54]. Finally, plates were rinsed extensively with HBSS, given fresh, dry lids and stored briefly under HBSS at 37◦ C in preparation for re-plating. Some plates were incubated with antibodies prior to re-plating (infra). Rationale for the selection of anti-laminin reagents to challenge re-plating onto conditioned surfaces To gain insight into the mechanisms of any differences observed between cultures growing on fresh F3 surfaces versus βG I/ 17-conditioned F3 surfaces, we sought to challenge the re-plating exercise with immune reagents raised against ECM proteins that might be involved. Which cell-adhesion ligands are used by native islet β cells and the insulinomas derived from them? In vivo, the connective tissue supporting pancreatic islets contains collagens, laminins and fibronectins. In rats and humans, islets and endocrine tumors of the pancreas express laminin and certain laminin receptors (viz., various combinations of integrin subunits α3, α6, β1 and β4 [22, 55– 60]). Taken together, these published observations, along with our preliminary immunofluorescence results (infra), suggested that laminin might be secreted onto F3 surfaces by βG I/ 17 rat insulinoma cells during the first week of culture. We therefore decided to challenge re-plating onto conditioned surfaces
204
J. R. Bain and A. S. Hoffman
with anti-laminin antibodies, rather than with antibodies to other pancreatic ECM proteins, such as fibronectin or collagen IV. Selection of anti-laminin reagents by indirect immunofluorescence First, we searched for antibodies with reactivity or cross-reactivity to rat laminins. Eight antibody preparations were procured and screened. Monoclonals 2E8, 3.1C12, C4, D5, D7 and D18 were obtained from Dr. David R. Soll, Developmental Studies Hybridoma Bank, National Institute of Child Health and Human Development, University of Iowa (Iowa City, IA, USA). All monoclonals were produced in mouse-mouse hybridomas. Two polyclonal antisera were evaluated: Z0097 (Dako, Glostrup, Denmark) and 5620-3019 (Biogenesis, Poole, UK). Negative controls for monoclonals and polyclonals were the mouse Ig1,KAPPA , MOPC-21 (number 03171D, Becton-Dickinson PharMingen, San Diego, CA, USA) and rabbit serum (number R-9133, Sigma), respectively. Phosphate-buffered saline (PBS, Gibco-BRL number 10010-023) with 1% (w /v) RIA-grade bovine-serum albumin (BSA; Sigma number A-7888) served as a diluent for antibodies. BSA was not heat-treated prior to use. BSA-PBS solution was sterilized by filtration. All monoclonals except D5 were evaluated at dilutions of 1 : 100, 1 : 200 and 1 : 300. D5, which had been received as a raw culture supernatant, was used full strength and at dilutions of 1 : 10 and 1 : 100. Dako polyclonal Z0097 was diluted 1 : 25, 1 : 50 and 1 : 100. Rabbit serum and Biogenesis polyclonal 5620-3019 were diluted 1 : 30, 1 : 100 and 1 : 300. βG I/17 cultures grown for one week on 3/3 F3 were rinsed thrice with 1 ml warm PBS, fixed in situ with ice-cold methanol for ten minutes, again rinsed thrice with 1 ml warm PBS, incubated for 1 h at 37◦ C under the primary antibodies, rinsed twice with PBS, then incubated under the appropriate, fluorescently labeled, secondary antibody, for 1 h at 37◦ C. Secondary antibodies were Texas-Red-X® conjugated goat-anti-mouse IgG for monoclonal preparations and Texas-Red-Xconjugated goat-anti-rabbit IgG for polyclonals (Molecular Probes, Eugene, OR, USA). Cultures were rinsed yet again, then examined with a Diaphot 300 inverted microscope (Nikon, Melville, NY, USA) equipped with an epifluorescent apparatus with a Texas Red filter set (Chiu Technical, Kings Park, NY, USA). Antibody challenge to re-plating Results of the indirect immunofluorescence studies (infra) suggested that Biogenesis polyclonal 5620-3019 might be able to block cellular adhesion to laminin deposited on the conditioned surfaces by the initial cultures of βG I/17. Insulinomaconditioned surfaces were incubated for 90 min under antiserum diluted to 1 : 30 and 1 : 300 with BSA-PBS. Controls included conditioned F3 discs incubated under rabbit serum (diluted 1 : 30) and F3 discs incubated under plain BSA-PBS.
Insulinoma culture on aminofluorosiloxanes
205
Re-plating and secondary culture Cells propagated on tissue-culture polystyrene (TCPS) T-flasks were plated into sample wells and fed daily for 7 days, exactly as described above. Note that these cells were exposed to trypsin immediately prior to the plating experiment. Controls included TCPS (Falcon Multiwell 3047, supra) and fresh, unconditioned F3 wells. TCPS was used as received. Unconditioned F3 controls were incubated under BSAPBS prior to use. At the completion of the week of secondary culture, media conditioned during the final 24 h of cell growth were removed and frozen for later analysis of insulin. Terminal cell counting of re-plated cultures Plates were rinsed in HBSS and then trypsinized. Cells were suspended in their individual culture wells by repeatedly pipetting up and down with a 1 ml disposable pipette tip. Because we knew it would take many hours to count all the samples (allowing for the occasional aperture jams that are perhaps inevitable with such preparations), we lightly fixed each individual culture by adding 0.5 ml of 0.05% (v/ v) glutaraldehyde in Gibco-BRL cell-dissociation buffer (number 13151-014), for a total volume of 1 ml per well. Aliquots of the fixed cell suspensions were diluted in Coulter Isoton® for counting in a Coulter Counter (model Z1, Coulter Electronics, Hialeah, FL, USA). Statistical analysis Pair-wise comparisons of data were evaluated post hoc with the Tukey– Kramer honestly significantly different (HSD) test [61], using version 3.0 of the JMP software (SAS Institute, Cary, NC, USA). Differences were considered significant at P < 0.05.
RESULTS AND DISCUSSION
Surface chemistry affects growth form of βG I/ 17: permissive surfaces βG I/ 17 displays a distinctive morphology when grown on planar substrata in vitro. Cells sprout emergent structures in late log phase when grown on such permissive surfaces as glass, as shown by white arrows in Fig. 2. Cell-material interactions appear to dominate early development of such βG I/ 17 cultures, while cell-cell interactions dominate late events. In this study, growth of βG I/ 17 cells to near confluence, followed by formation of emergent foci, was quite similar on alkalineetched borosilicate glass (Figs 2 and 3), the pure aminosiloxane, N2 (Fig. 3), surfaces with a mole fraction of 1/3 F3 (not shown) and the pure fluorosiloxane, F3 (Fig. 3).
206 J. R. Bain and A. S. Hoffman Figure 2. βG I/17 cells growing on alkaline-etched borosilicate glass, a permissive surface. The same morphologic progression is seen in cultures grown on permissive members of the N2-F3 silane series. Arrows show emergent colonies. Phase-contrast microscopy. Each image represents a sample area measuring 1.36 × 0.9 mm.
Insulinoma culture on aminofluorosiloxanes 207
Figure 3. Appearance of fixed, stained cultures of βG I/17 cells after stimulation with the ‘Swiss cocktail’ of secretagogues. One surface (2/3 F3) is inhibitory. The other three are permissive. Arrows show emergent colonies. Hematoxylin and eosin stain. Bright-field microscopy with Köhler illumination. Each image represent an area of 1.36 × 0.9 mm. This figure is published in colour on http://www.ingenta.com.
208
J. R. Bain and A. S. Hoffman
A generic description of the growth of βG I/17 on these permissive surfaces follows. At one day, βG I/ 17 cells attach and spread into irregular bipolar, threepointed, or stellate shapes, with pronounced cytoplasmic processes (Fig. 2). Cells are dispersed. Doublets, triplets and higher aggregates are rare at 24 h. During days three to five, βG I/ 17 cells spread further and grow rapidly (Fig. 2). The underlying surface is never fully covered. Groups of spread cells have scalloped margins, leaving bare lacunae demarcated by cells. By days five to seven, most cells appear to be joined by cytoplasmic processes, large areas of nearly-confluent cells are seen and it becomes increasingly difficult to distinguish individual cells (Figs 2 and 3). A similar sequence of morphologic events has been observed when disaggregated islet cells from normal rats are grown in plate culture [62]. Emergent, multicellular domes then begin to appear, progressing to oblate, tethered spheroids, apparently by combined processes of mitosis, aggregation and local delamination of cell sheets. Incipient supra-confluent structures are visible at seven days along the margins of a delaminating cell sheet on etched borosilicate glass in Fig. 3. Multicellular character of these emergent structures is evident after Swiss-cocktail stimulation, formalin fixation and staining with hematoxylin and eosin (Fig. 3). Occasionally, these large aggregates become lobulate, grow to macroscopic dimensions and develop arms hundreds of microns in length (e.g. Fig. 4, day 5, taken at low magnification). We have seen this formation of three-dimensional structures and a loss of strong adherence to the substratum in maturing cultures of βG I/ 17 cells and the related 832/13 cell line [19] (Fig. 1) grown on other permissive materials, including tissueculture polystyrene (TCPS, data not shown). Fong et al. [63] grew the closely related rat insulinoma line, RINr (Fig. 1), on TCPS. Morphology of their fiveday cultures was strikingly similar to our own. Clark and Chick [64] also noted delamination of maturing RINr clusters. RINr is ancestral to our βG I/17 [50] (Fig. 1). A similar morphologic progression occurs when diverse lines of insulinsecreting cells from mice are grown on permissive surfaces [65– 67]. These findings of adhesion to surfaces by freshly-plated cells, followed by later aggregation and partial delamination, contrast with most studies of two related insulinoma lines from rats, RINm5F and BRIN-BD11 (Fig. 1), in which cultures grow on TCPS as a monolayer of well-spread, stellate cells (Refs [7, 41, 68], but see Ref. [69]). When interpreting published accounts of the behavior of insulinoma cells in vitro, one must keep in mind that most insulin-secreting cell lines are unstable and change their phenotype with time [5, 7, 10, 16, 32, 70, 71]. Also confounding the interpretation of published studies is the apparent genetic heterogeneity of many insulin-secreting lines: βG I/ 17 is clonal [50], but many other insulinoma lines are not [7, 16, 19]. Cells adopt a distinct growth form on inhibitory culture surfaces To our surprise, some aminofluorosiloxane surfaces of mixed character inhibited growth of βG I/ 17 cells. Depression of cell growth was most pronounced at a mole fraction of 2/3 F3, where cells never approached confluence (Fig. 3). Sparse aggregates of poorly spread cells were visible at 24 h. Compared to cells grown on
Insulinoma culture on aminofluorosiloxanes 209
Figure 4. βG I/ 17 cells growing on untreated polystyrene or UPS, the negative control. Note the slow, incomplete coverage of the surface and the early formation of emergent masses (arrows). Phase-contrast microscopy. The five-day image was taken at low magnification to show formation of a macroscopic, emergent colony and represents an area 3.35 × 2.23 mm. The other images represent areas of 1.36 × 0.9 mm.
210
J. R. Bain and A. S. Hoffman
more permissive surfaces, cells on 2/3 F3 displayed an earlier tendency toward the formation of emergent domes and spheroids, often giving isolated small aggregates a ‘sunny-side-up, fried-egg’ appearance (Fig. 3). In all these characteristics, βG I/ 17 cells grown on 2/3 F3 resembled cultures grown on the negative control, untreated polystyrene (Fig. 4). Thivolet et al. [72] made similar observations when they cultured the rat insulinoma line, RINm5F (Fig. 1), on uncoated ‘plastic’ dishes. Phenotype of cells grown on mixed surfaces with a mole fraction of 5/6 F3 (not shown) was intermediate in appearance between the inhibitory 2/3 F3 and the permissive 100% F3. We have grown other types of mammalian cells on silanized surfaces of the N2F3 series and we have not yet found another case where the culture surface has such striking effects on cellular morphology. When grown under serum-containing medium, the human melanoma cell line, A-375 (American Type Culture Collection, Manassas, VA, USA), for example, spreads and grows equally well on all surfaces of the N2-F3 series (data not shown). The culture surface influences glucose consumption and lactate production Metabolic data from assays of conditioned media correlated well with the qualitative observations of cell growth described above. βG I/ 17 insulinoma cells have a highly active metabolism, as measured by their consumption of glucose and production of lactate (Figs 5– 7). Our plate cultures of βG I/17 cells depleted sugar so rapidly that by the end of the first week, they consumed more than half of the available glucose in 24 h (Fig. 5). Related rat insulinoma (RIN) cell lines exhibit a similar voracious appetite for metabolic fuels, which is much higher than that of the native β cells from which they were derived [7]. A plot of raw data of lactate and glucose concentration in one representative well shows the ‘saw-tooth’ profile (Fig. 5) characteristic of intermittently fed plate cultures. Papas et al. [9] reported a similar stoichiometry in glucose-lactate metabolism in cultures of the mouse insulinoma cell line, βTC3. Roughly exponential plots of glucose consumption (Fig. 6) and lactate production (Fig. 7) during the week of culture correlate with observations of morphology (Figs 2– 4). Culture surfaces of glass, pure N2 (0/3 F3), 1/3 mole fraction F3 and pure F3 are permissive. Untreated polystyrene and 2/3 mole fraction F3 are inhibitory and 5/6 mole fraction F3 is intermediate. The molar ratio of lactate produced to glucose consumed provides a useful measure of cell metabolism in vitro [73]. In the present study, the roughly 2 : 1 stoichiometry of production of the triose, lactate, to consumption of the hexose, glucose, remained relatively stable throughout the week of culture (Figs 5– 10). This suggests that catabolism of glucose to pyruvate to lactate was a significant energetic pathway in these rapidly growing cultures of insulinoma cells. Data on consumption of glucose and production of lactate during the seventh day of culture are plotted as a function of surface composition in Fig. 8. Analysis of lactate and glucose data (Fig. 8) with the Tukey– Kramer HSD statistic separates cultures into
Insulinoma culture on aminofluorosiloxanes
211
Figure 5. Characteristic ‘saw tooth’ profile of raw data on glucose consumption and lactate production by a single culture. With appropriate scaling, the lactate profile would nearly mirror the glucose profile. Given the pronounced daily swings in composition of the aqueous phase, life in such a dish more closely resembles life in a tide pool than life in a stream. This culture support, a mixed aminofluorosiloxane surface with a mole fraction 1/3 silane F3, is permissive, allowing growth as rapid as alkaline-etched borosilicate glass, the positive control.
three exclusive groups, based on identity of their culture surfaces: Group A surfaces are permissive, those in Group B are inhibitory and surfaces with a mole fraction of 5/6 F3 are intermediate (Group C). Group separations are significant at P < 0.05. These correspond to the three classes defined by growth morphology (vide supra). βG I/ 17 cells grown on permissive surfaces converted more glucose to lactate than those grown on inhibitory or intermediate surfaces. Molar ratio of glucose consumed to lactate produced was >1.90 for cultures grown on the permissive aminofluorosiloxane surfaces (Fig. 9), indicative of a highly anaerobic metabolism. Cultures on the other permissive surface, alkaline-etched borosilicate glass, were similar and produced an average of 1.84 mol lactate per mol glucose (Fig. 9). Cultures on the inhibitory and intermediate surfaces averaged less than 1.80 mol lactate per mol glucose. Many paired comparisons were significant (Tukey– Kramer HSD test, P < 0.05, Fig. 9). Thus, in the present study, a picture emerges that rapidly growing cultures tended to be more anaerobic in their metabolism than cultures with slower growth. Confirming this, a cross plot of glucose consumption and the lactate/ glucose ratio (Fig. 10) shows a weak, positive correlation between these two variables (r 2 approx. 0.6) and a separation of data into two groups based on the permissive-
212
J. R. Bain and A. S. Hoffman
Figure 6. Glucose consumption by βG I/17 cells cultured on aminofluorosiloxanes and two reference materials. Solid lines show data for cultures grown on permissive surfaces, while broken lines show data for inhibitory or intermediate surfaces.
Figure 7. Lactate production by βG I/ 17 cells cultured on aminofluorosiloxanes and two reference materials. Solid lines show data for cultures grown on permissive surfaces, while broken lines show data for inhibitory or intermediate surfaces.
Insulinoma culture on aminofluorosiloxanes
213
Figure 8. Glucose consumption and lactate production for the final 24-h culture period, plotted as a function of the mole fraction of silane F3 residues in the culture surface. UPS, untreated polystyrene.
Figure 9. Molar ratio of lactate produced to glucose consumed. Complete anaerobic conversion of glucose to lactate via pyruvate would give a ratio of 2.0. Complete aerobic conversion of glucose to carbon dioxide would give a ratio of zero. UPS, untreated polystyrene. ∗ P < 0.05, Tukey– Kramer HSD test.
214
J. R. Bain and A. S. Hoffman
Figure 10. Cross plot of the lactate/glucose ratio as a function of glucose consumption. Rapidly growing cultures on permissive surfaces (ellipse) are more anaerobic than slower, more rounded colonies growing on inhibitory and intermediate surfaces (circle).
ness of the growth surface. Constantinidis et al. [74] observed a similar relationship between growth morphology and glucose/ lactate metabolism when they compared gel-entrapped and plate cultures of the mouse insulinoma cell line, βTC3. Bulk secretion of human insulin is highest on fluorinated surfaces All cultures in this study secreted 17 ng or more of human insulin per well per h on the seventh and final day of culture (Fig. 11). Moreover, all cultures grown on aminofluorosiloxane surfaces showed a higher mean seventh-day and Swisscocktail-stimulated insulin secretion than cultures grown on the two reference materials, etched glass and untreated polystyrene. Insulin secretion was highest when cells were grown on the three surfaces with the highest fluorine content (Fig. 11 and Table 1). Surface composition can favorably influence sensitivity to secretagogue drugs Cell biologists evaluating responses of secretory cells to stimuli often examine the ratio of stimulated to basal secretion. In the present study, the ‘fold stimulation’ was similar for βG I/ 17 cells grown the two reference materials, etched glass and untreated polystyrene (2.3- and 2.4-fold, respectively; Fig. 12). This surprised us, since the insulinoma cells showed strikingly divergent morphology and glucoselactate metabolism on the two materials. In all cases, the mean responsiveness of βG I/17 cells grown on N2-F3 surfaces was about 3-fold or greater. Moreover,
Insulinoma culture on aminofluorosiloxanes
215
Figure 11. Bullk insulin secretion by βG I/17 cells cultured on aminofluorosiloxanes N2-F3 and two reference materials. UPS, untreated polystyrene. ∗ Significantly different from UPS, P < 0.05; #significantly different from glass, P < 0.05.
Table 1. Summary of biological studies of silanized surfaces and two reference materials
Growth rate of βG I/17 insulinoma Growth form of βG I/ 17 insulinoma Formation of emergent cell aggregates Glucose consumed on day 7 Lactate produced on day 7 Mol lactate/ mol glucose on day 7 Insulin secreted on day 7 Secretagogue sensitivity (Swiss/basal) Insulin secreted per glucose consumed UPS, untreated polystyrene.
Etched glass
0/3 F3
2/3 F3
3/3 F3
UPS
rapid spread late high high 1.85 modest <3.5× modest
rapid spread late high high 1.95 higher <3.5× modest
slower round early lower lower 1.76 highest <3.5× highest
rapid spread late high high 1.91 highest >5.0× highest
slower round early lower lower 1.79 modest <3.5× modest
216
J. R. Bain and A. S. Hoffman
Figure 12. Ratio of stimulated to basal insulin secretion. Responsiveness to the Swiss cocktail of secretagogue drugs is greatest when cells are grown on surfaces with the highest content of trifluoropropyl groups (F3). UPS, untreated polystyrene. ∗ Significantly different from both UPS and glass, P < 0.05.
mean responsiveness averaged as high as 5.0-fold for cultures propagated on the pure F3 surface, a significant departure from the two control materials (P < 0.05; Fig. 12). But these data illustrate a limitation of βG I/ 17 as a potential islet surrogate. When subjected to the same regimen of diazoxide inhibition, followed by Swisscocktail stimulation, natural human β cells and the newer rat insulinoma line, βG11/ 3E9, a descendant of βG I/17 (Fig. 1), can show a more than 15-fold stimulation of insulin secretion [16, 20]. Inappropriate glucose response hampers development of human implants Figure 11 illustrates a fundamental limitation of the extant insulinoma cell lines that are candidates for therapeutic use in the artificial endocrine pancreas. In the present study, cells in routine culture secreted maximal levels of insulin. That is, secretion of insulin provoked by the Swiss cocktail of secretagogues was no different than secretion during normal culture (Fig. 11). Fresh media in the present study had glucose concentrations (10.4 mM) well above the diabetic threshold for human plasma (approx. 7.8 mM), so maximal insulin secretion was appropriate in fresh media. But maximal insulin secretion by βG I/17 apparently persisted when glucose was depleted down into the range of plasma from normal humans (approx. 3.9 to 6.1 mM). Ideally, a cell used in an artificial pancreas will cease insulin secretion when normoglycemia is attained. βG I/ 17 and many other rodent insulinoma cell lines exhibit glucose-stimulated insulin secretion. But when insulin secretion is plotted as a function of environmental glucose, one sees that the curve is shifted far to the left of the curve characteristic of the β cells of healthy pancreatic islets. That is, insulin secretion is hyper-stimulated under normoglycemia and insulin secretion is not ‘turned off’ until perilously low glucose concentrations
Insulinoma culture on aminofluorosiloxanes
217
are reached, often <1 mM [5, 6, 14, 32, 38, 51, 75]. A human patient with plasma glucose levels <1 mM would be dangerously hypoglycemic and at risk of convulsions, coma and death. In nature, the pancreata of essentially all healthy mammals cease insulin secretion at plasma glucose levels less than 5 mM [76]. Engineering appropriate glucose-stimulated insulin secretion into transplantable cells has been a stubborn obstacle and remains a central focus of many laboratories, worldwide [5, 7, 16, 17, 19, 20, 23– 25, 27, 28, 30, 31, 51]. Unfortunately, most cellular engineering efforts to date have resulted in β-cell surrogates in which the ‘glucose switch’ is defective: maximal insulin secretion occurs at micromolar, not millimolar concentrations of environmental glucose [27]. Thus, even if one could safely implant allo- or xenogeneic cell lines into humans, one would still need to engineer a stable glucose-insulin response into the cells. Surface chemistry influences insulin/ glucose ratios The ratio of insulin produced to glucose consumed is of interest to those using insulinoma cells to produce human insulin in bioreactors [19, 24, 29, 77]. In the present study, insulin/ glucose ratios were highest when βG I/ 17 cells were cultured on the more-highly-fluorinated members of the N2-F3 series, peaking at 2/3 F3 (Fig. 13). Cultures on untreated polystyrene were intermediate in rank between the aminofluorosiloxanes with a high insulin/ glucose ratio (i.e. those with a mole
Figure 13. Hourly insulin secretion by βG I/ 17 cells normalized to daily glucose consumption during the final day of culture. ∗ Significantly different from UPS, P < 0.05; #significanlty different from glass, P < 0.05.
218
J. R. Bain and A. S. Hoffman
fraction >50% F3) and those with a lower insulin/ glucose ratio (i.e. surfaces relatively rich in N2 residues). Surprisingly, during both routine culture and Swisscocktail stimulation, the ratio of insulin secreted to glucose consumed was lowest in cells grown on the classic permissive material, glass (Fig. 13). Laminin is present on surfaces conditioned by βG I/ 17 cells None of the monoclonal antibodies gave a response above the negative control. Dako polyclonal Z0097 showed only slightly more fluorescence than the negative control. Biogenesis polyclonal anti-laminin, 5620-3019, yielded immunofluorescence that was strong and localized to specific sites on cell surfaces and inside cells that had been disrupted during processing. Fluorescence was strongest at the 1 : 30 dilution. Brightest staining was seen on the undersurfaces of cell sheets at sites of local delamination. There appeared to be a polarization, with the greatest amount of laminin on the surfaces of cells in contact with the F3 surface and on the silanized surface itself, where ‘footprints’ of cells fluoresced brightly. These observations were the basis for the use of Biogenesis 5620-3019 antiserum to challenge the replating exercise. Surface conditioning does not affect morphology of subsequent cultures To our surprise, growth morphology of the secondary cultures never differed appreciably from that seen on permissive surfaces evaluated in the initial cell-culture studies (e.g. etched borosilicate glass, pure F3, or pure N2). Fresh-TCPS and freshF3 controls had much less amorphous biologic debris on their surfaces at the micron scale than the cell-conditioned surfaces, but the morphologic progression of the cell cultures was the same in all cases: solitary cells and small groups grew into sheets fenestrated with lacunae, with formation of emergent dome-like structures in the final days of culture (e.g. Fig. 2). Terminal cell counts (Fig. 14) confirmed the morphologic findings. Populations on all surfaces were statistically equivalent (Tukey– Kramer HSD test, P > 0.05). Conditioned F3 surfaces and unconditioned F3 surfaces were equally as permissive to cell growth as TCPS, the positive control. Surface laminin appears to promote insulin secretion by subsequent cultures Bulk insulin secretion (Fig. 15) and insulin productivity normalized to cell number (Fig. 16) were about 50% greater on conditioned versus unconditioned F3 surfaces. Pre-treatment of conditioned surfaces with high concentrations of the Biogenesis anti-laminin reagent partially abolished this improvement in secretory function (Figs 15 and 16). Laminin deposited onto surfaces by the conditioning cultures of cells might therefore have promoted insulin secretion by the subsequent cultures of βG I/ 17 cells. This correlates with reports that, in vitro, whole islets of Langerhans, purified beta cells and insulinoma cells from mammals, as well as the homologous cells from birds, have been observed to grow and function well on surfaces or
Insulinoma culture on aminofluorosiloxanes
219
Figure 14. Insulinoma cell numbers one week after re-plating. Cells grow at the same rate, regardless of whether the surface has been conditioned by an earlier culture (Tukey– Kramer HSD test, P > 0.05). BSA-PBS = albumin buffer.
inside gels comprising collagen, fibronectin, laminin, or crude ECM. Increasingly, laminin-coated surfaces are identified as desirable culture substrata for such cells and tissue fragments (see Refs [78, 79], review in Ref. [80]). How might surface chemistry influence cell growth and function? Table 1 summarizes the results of this study. Surfaces richest in trifluoropropyl residues were the most hydrophobic (θA 90◦ [44]) and we were surprised that they supported cell growth and function. A priori, we had expected F3 surfaces to be poor culture supports [44]. Not only did βG I/ 17 insulinoma cells grow on F3 surfaces, but the pure F3 surface supported cell growth and glucose-lactate metabolism equivalent to the positive control, alkaline-etched borosilicate glass (Figs 2, 3, 6– 8 and 10). Most surprising of all, cultures grown on surfaces richest in CH2 CH2 CF3 residues showed the highest secretion of human insulin (Figs 11, 13)
220
J. R. Bain and A. S. Hoffman
Figure 15. Bulk insulin production during the seventh day of culture after re-plating. Conditioning of the surface by an earlier culture of insulinoma cells enhances bulk insulin production by the subsequent culture, and laminin anti-serum abolishes this effect, when applied at a dilution of 1 : 30. ∗ Significantly different from cultures on the unconditioned F3 surface, P < 0.05, Tukey– Kramer HSD test. BSA-PBS, albumin buffer.
and, in the case of the pure F3 surface, improved responsiveness to secretagogues (Fig. 12). Mammalian cells that can be grown on artificial materials generally do best when the surface has an advancing water contact angle, θA , between 20 and 60◦ [81], though exceptions are known [82, 83]. The 20– 60◦ range includes tissue-culture polystyrene, the dominant commercial cell-culture material in the world today. Why do βG I/ 17 cells grow and function so well on the hydrophobic, F3 materials (θA 90◦ [44])? Perhaps CH2 CH2 CF3 groups promote tight binding or favorable conformations of physisorbed proteins [84, 85]. Steele, Griesser and their co-workers have demonstrated that relative affinities and binding states of adsorbed cell-adhesion proteins can in turn influence cell function [83, 86]. The ways in which the terminal CF3 groups are tethered to the surface and the chemical nature of the underlying and surrounding groups all appear to be important determinants of protein binding to surfaces bearing the trifluoromethyl group. Horbett, Hoffman and their co-workers observed that CF3 groups created by certain radio-frequency, glow-discharge plasmas favor tight binding by a variety of proteins [84, 85], while the CF3 groups on the surface of perfluorinated poly (ethylene-co-propylene) or FEP do not [87].
Insulinoma culture on aminofluorosiloxanes
221
Figure 16. Insulin secretion on the seventh day, normalized to cell number. Again, conditioning of the surface by a previous population of cells favors insulin secretion by a subsequent culture and laminin antiserum negates the effect. ∗ Significantly different from cultures on the unconditioned F3 surface, P < 0.05, Tukey– Kramer HSD test. BSA-PBS, albumin buffer.
In the present study, tightness of protein binding might also have been affected by the nanometer-scale, insular morphology of the F3-silanized surfaces [44]. Investigators examining protein adsorption to surfaces with silane gradients have commented on the patchy nature of silane residues [88– 92]. Anomalously tight protein binding in transition regions led Gölander et al. [90] to speculate that, ‘such increased adsorption affinity of an amphiphilic molecule can be envisaged if such a molecule can interact with hydrophobic and hydrophilic adsorption sites simultaneously’. Takahara et al. [93] presented experimental evidence that proteins can adsorb preferentially to fluorosilanized domains of mixed surfaces under some experimental conditions. In the present study, we speculate that boundaries between un-silanized glass and islands rich in CF3 groups [44] might have constituted sites that favored strong adsorption of cell-adhesion proteins. Protein ligands bound tightly or in favorable conformations might have in turn influenced the secretory phenotype of βG I/ 17 cells. Favorable adsorption of laminin to trifluoropropyl (F3) domains might account in part for the observed dual benefits of high F3 content and surface conditioning.
222
J. R. Bain and A. S. Hoffman
CONCLUSIONS
In this study, we were surprised to find that βG I/ 17 insulinoma cells grown on surfaces richest in the F3 trifluoropropyl group, CH2 CH2 CF3 , exhibited the greatest insulin secretion and responsiveness to secretagogue drugs (Table 1). Polymers bearing trifluoropropyl residues might therefore be attractive candidates for use in the artificial endocrine pancreas. Silicone polymers that include this pendant group have already been used clinically in contact lenses, voice prostheses and a variety of soft-tissue implants [94– 97]. Conditioning of trifluoropropyl-silanized surfaces by βG I/ 17 cells promoted insulin secretion by subsequent cultures. A polyclonal antiserum against rat laminin largely negated this effect, indicating that laminin might have been an important component of the cell-conditioned surfaces. F3 silicones or F3-silanized surfaces, when coated with laminin, might be useful in bioreactors and in artificial organs that rely on transgenic insulinoma cells. In particular, should anyone ever develop a stable, glucose-sensitive, implantable line of human insulinoma cells, scaffoldings of F3 silicones coated with recombinant human laminin might be useful cell supports inside the artificial endocrine pancreas. Acknowledgements This work was supported by grants from the National Institutes of Health (NIH # 1-T32-GM-08437-01 and 5-T32-HL-07403-15) and by Gore Hybrid Technologies. The cell line βG I/17 was a gift from Dr. C. B. Newgard, Duke University Medical Center (Durham, NC, USA). M. Truini sketched the rodent cartoons. Monoclonal antibodies were gifts from Dr. J. R. Sanes, Washington University Medical School (Saint Louis, MO, USA; clones C4, D18, D5 and D7), Dr. E. Engvall, La Jolla Cancer Research Center (La Jolla, CA, USA; clone 2E8) and Dr. T. M. Jessell and Dr. J. Dodd, Columbia University (New York, NY, USA; clone 3.1C12). D. Ballard, B. D. Becker, B. Benigni, N. Burns, Dr. M. D. Butler, Dr. S. A. Clark, Dr. M. S. Cooper, Dr. W. R. Gombotz, K. Gooby, Dr. H.-E. Hohmeier, Dr. T. A. Horbett, Dr. K. A. Knisley, B. H. Kram, S. L. Mish, Dr. M. J. Muehlbauer, S. C. Newman, Dr. N. P. Robertson, B. Roche, Dr. A. E. Schmierer, Dr. G. T. Schuppin, G. E. Smith and Dr. P. S. Stayton helped in the lab and with the manuscript. Two anonymous referees made helpful suggestions.
REFERENCES 1. P. Onkamo, S. Väänänen, M. Karvonen and J. Tuomilehto, Diabetologia 42, 1395 (1999). 2. M. Silink, Horm. Res. 57 (Suppl. 1), 1 (2002). 3. A. T. Cheung, B. Dayanandan, J. T. Lewis, G. S. Korbutt, R. V. Rajotte, M. Bryer-Ash, M. O. Boylan, M. M. Wolfe and T. J. Kieffer, Science 290, 1959 (2000). 4. A. Green and C. C. Patterson, Diabetologia 44 (Suppl. 3), B3 (2001). 5. N. H. McClenaghan and P. R. Flatt, J. Mol. Med. 77, 235 (1999).
Insulinoma culture on aminofluorosiloxanes 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38.
223
A. Sambanis, Diabetes Technol. Ther. 2, 81 (2000). S. J. Persaud, Adv. Mol. Cell Biol. 29, 21 (1999). P. Soon-Shiong, Adv. Drug Deliv. Rev. 35, 259 (1999). K. K. Papas, R. C. Long, Jr., A. Sambanis and I. Constantinidis, Biotechnol. Bioeng. 66, 219 (1999). B. Soria, E. Roche, G. Berná, T. León-Quinto, J. A. Reig and F. Martin, Diabetes 49, 157 (2000). G. M. Beattie, G. Leibowitz, A. D. Lopez, F. Levine and A. Hayek, Cell Transplant. 9, 431 (2000). G. Luca, M. Calvitti, L. M. Neri, E. Becchetti, S. Capitani, G. Basta, G. Angeletti, C. Fanelli, P. Brunetti and R. Calafiore, J. Invest. Med. 48, 441 (2000). K. Edamura, H. Ohgawara, K. Nasu, Y. Iwami, A. Sato, S. Ishikawa, N. Matsuki, K. Ono, H. Ogawa and N. Sasaki, Cell Transplant. 10, 493 (2001). R. Suzuki, N. Okada, H. Miyamoto, T. Yoshioka, K. Sakamoto, H. Oka, Y. Tsutsumi, S. Nakagawa, J.-I. Miyazaki and T. Mayumi, Life Sci. 71, 1717 (2002). R. Valdés-González, in: 19th International Congress of the Transplantation Society, Miami, FL, abstract number 524 (2002). C. B. Newgard, Diabetes 51, 3141 (2002). A. Aynsley-Green, K. Lindley, K. Docherty, M. Dunne, W. MacFarlane and R. F. L. James, US Patent No. 6,372,501 (2002). L. Yang, S. Li, H. Hatch, K. Ahrens, J. G. Cornelius, B. E. Petersen and A. B. Peck, Proc. Natl. Acad. Sci. USA 99, 8078 (2002). H. E. Hohmeier, H. Mulder, G. Chen, R. Henkel-Rieger, M. Prentki and C. B. Newgard, Diabetes 49, 424 (2000). C. B. Newgard, S. Clark, H. Beltran del Rio, H. E. Hohmeier, C. Quaade and K. Normington, Diabetologia 40, S42 (1997). A. Shiroi, M. Yoshikawa, H. Yokota, H. Fukui, S. Ishizaka, K. Tatsumi and Y. Takahashi, Stem Cells 20, 284 (2002). H. Mashima, S. Yamada, T. Tajima, M. Seno, H. Yamada, J. Takeda and I. Kojima, Diabetes 48, 304 (1999). S. Motoyoshi, T. Shirotani, E. Araki, K. Sakai, K. Kaneko, H. Motoshima, K. Yoshizato, A. Shirakami, H. Kishikawa and M. Shichiri, Diabetologia 41, 1492 (1998). A. Powers and L. Wu, World Intellectual Property Organization, Munich, Germany, International Patent, Publication number WO 99/54451 (1999). S. Bonner-Weir and A. Sharma, J. Pathol. 197, 519 (2002). N. Giannoukakis and P. D. Robbins, BioDrugs 16, 149 (2002). M. Tiedge, M. Elsner, N. H. McClenaghan, H.-J. Hedrich, D. Grube, J. Klempnauer and S. Lenzen, Hum. Gene Ther. 11, 403 (2000). P. M. Thulé, J. Liu and L. S. Phillips, Gene Ther. 7, 205 (2000). B. H. Kram, S. L. Mish, M. J. Muehlbauer and J. R. Bain, US Patent No. 6,306,491 (2001). A. S. Pollock, S. Christakos and D. Reddy, US Patent No. 6,319,495 (2001). S. Efrat, Trends Mol. Med. 8, 334 (2002). D. Zhou, E. Kintsourashvili, S. Mamujee, I. Vacek and A. M. Sun, Ann. NY Acad. Sci. 875, 208 (1999). M. Butler and S. Mish, US Patent No. 6,426,214 (2002). G. H. Algire, M. L. Borders and V. J. Evans, J. Natl. Cancer Inst. 20, 1187 (1958). M. Fakir, A. Penfornis, N. Elian, P.-H. Cugnenc and J. J. Altman, Int. J. Artif. Org. 20, 637 (1997). P. M. Galletti and J. J. Altman, Trans. Am. Soc. Artif. Intern. Org. 30, 675 (1984). J. R. Brooks and G. J. Hill II, Am. J. Surg. 99, 588 (1960). T. A. Desai, Microfabricated biocapsules for the immunoisolation of pancreatic islets of Langerhans, PhD dissertation, University of California, San Francisco and Berkeley, CA (1998).
224
J. R. Bain and A. S. Hoffman
39. L. Leoni and T. A. Desai, Inst. Electrical Electron. Eng., Trans. Biomed. Eng. 48, 1335 (2001). 40. K.-H. Park, M. Goto, J.-I. Miyazaki, C.-S. Cho and T. Akaike, J. Biomater. Sci. Polymer Edn 12, 911 (2001). 41. N. Oturan, H. Serne, G. Reach and M. Jozefowicz, J. Biomed. Mater. Res. 27, 705 (1993). 42. C. Stabler, K. Wilks, A. Sambanis and I. Constantinidis, Biomaterials 22, 1301 (2001). 43. J. R. Bain and A. S. Hoffman, Biomaterials 23, 3347 (2002). 44. J. R. Bain and A. S. Hoffman, J. Biomater. Sci. Polymer Edn 14, 325 (2003). 45. J. R. Bain, A. S. Hoffman and T. A. Horbett, in: 37th International Symposium on Macromolecules, Gold Coast, Queensland, Australia, World Polymer Congress, International Union of Pure Applied Chemistry, preprints, p. 789 (1998). 46. W. L. Chick, S. Warren, R. N. Chute, A. A. Like, V. Lauris and K. C. Kitchen, Proc. Natl. Acad. Sci. USA 74, 628 (1977). 47. A. F. Gazdar, W. L. Chick, H. K. Oie, H. L. Sims, D. L. King, G. C. Weir and V. Lauris, Proc. Natl. Acad. Sci. USA 77, 3519 (1980). 48. J. Philippe, W. L. Chick and J. F. Habener, J. Clin. Invest. 79, 351 (1987). 49. S. Ferber, H. Beltran del Rio, J. H. Johnson, R. J. Noel, L. E. Cassidy, S. Clark, T. C. Becker, S. D. Hughes and C. B. Newgard, J. Biol. Chem. 269, 11523 (1994). 50. S. A. Clark, C. Quaade, H. Constandy, P. Hansen, P. Halban, S. Ferber, C. B. Newgard and K. Normington, Diabetes 46, 958 (1997). 51. H. E. Hohmeier, H. Beltran del Rio, S. A. Clark, R. Henkel-Rieger, K. Normington and C. B. Newgard, Diabetes 46, 968 (1997). 52. P. Petit and M. M. Loubatières-Mariani, Fund. Clin. Pharmacol. 6, 123 (1992). 53. C. Halberstadt and J. J. Grzesiak, US Patent No. 5,672,361 (1997). 54. J. C. R. Jones, US Patent No. 5,541,106 (1996). 55. C. Haglund, S. Nordling, P. J. Roberts and P. Ekblom, Am. J. Surg. Pathol. 8, 669 (1984). 56. G. Pelosi, F. Pasini, E. Bresaola, G. Bogina, P. Pederzoli, S. Biolo, S. Menard and G. Zamboni, J. Pathol. 183, 62 (1997). 57. R. N. Wang, S. Paraskevas and L. Rosenberg, J. Histochem. Cytochem. 47, 499 (1999). 58. D. Bosco, P. Meda, P. A. Halban and D. G. Rouiller, Diabetes 49, 233 (2000). 59. S. Kantengwa, D. Baetens, K. Sadoul, C. A. Buck, P. A. Halban and D. G. Rouiller, Exp. Cell Res. 237, 394 (1997). 60. B. Schmeid, G. Liu, M. P. Moyer, L. S. B. Hernberg, W. Sanger, S. Batra and P. M. Pour, Carcinogenesis 20, 317 (1999). 61. R. R. Sokal and F. J. Rolf, Biometry, 2nd edn, Chapter 9. W. H. Freeman, San Francisco, CA (1981). 62. W. J. Tze and J. Tai, Transplantation 34, 228 (1982). 63. H. K. W. Fong, W. L. Chick and G. H. Sato, Diabetes 32, 1022 (1981). 64. S. A. Clark and W. L. Chick, Endocrinology 126, 1895 (1990). 65. S. Efrat, S. Linde, H. Kofod, D. Spector, M. Delannoy, S. Grant, D. Hanahan and S. Baekkesov, Proc. Natl. Acad. Sci. USA 85, 9037 (1988). 66. J.-I. Miyazaki, K. Araki, E. Yamato, H. Ikegami, T. Asano, Y. Shibasaki, Y. Oka and K.-I. Yamamura, Endocrinology 127, 126 (1990). 67. A. C. Hauge-Evans, P. E. Squires, S. J. Persaud and P. M. Jones, Diabetes 48, 1402 (1999). 68. N. H. McClenaghan, C. R. Barnett, E. Ah-Sing, Y. H. A. Abdel-Wahab, F. P. M. O’Harte, T.-W. Yoon, S. K. Swanston-Flatt and P. R. Flatt, Diabetes 45, 1132 (1996). 69. H. Abrahamsson, P.-O. Berggren and B. Hellman, Am. J. Physiol. 247, E719 (1984). 70. M. Noda, M. Komatsu and G. W. G. Sharp, Diabetes 45, 1766 (1996). 71. M. Asfari, D. Janjic, P. Meda, G. Li, P. A. Halban and C. B. Wollheim, Endocrinology 130, 167 (1992). 72. C. H. Thivolet, P. Chatelain, H. Nicoloso, A. Durand and J. Bertrand, Exp. Cell Res. 159, 313 (1985).
Insulinoma culture on aminofluorosiloxanes
225
73. B. Obradovic, R. L. Carrier, G. Vunjak-Novakovic and L. E. Freed, Biotechnol. Bioeng. 63, 197 (1999). 74. I. Constantinidis, N. E. Mukundan, M. P. Gamcsik and A. Sambanis, Cell. Mol. Biol. 43, 721 (1997). 75. F. Schuit, A. de Vos, S. Farfari, K. Moens, D. Pipeleers, T. Brun and M. Prentki, J. Biol. Chem. 272, 18572 (1997). 76. C. B. Newgard and F. M. Matschinsky, in: Handbook of Physiology. Section 7. The Endocrine System. Volume II. The Endocrine Pancreas and Regulation of Metabolism, L. S. Jefferson, A. D. Cherrington and H. M. Goodman (Eds), Chapter 5. Oxford University Press, New York, NY (2001). 77. G. E. Grampp, M. A. Applegate and G. Stephanopoulos, Biotechnol. Prog. 12, 837 (1996). 78. N. A. Nagata, K. Inoue and Y. Tabata, J. Biomater. Sci. Polymer Edn 13, 579 (2002). 79. F.-X. Jiang, E. G. Stanley, L. J. Gomez and L. C. Harrison, J. Cell Sci. 115, 753 (2002). 80. J. R. Bain, Growth and function of transgenic endocrine cells on silanized surfaces, PhD dissertation, University of Washington, Seattle, WA (2001). 81. K. Webb, V. Hlady and P. A. Tresco, J. Biomed. Mater. Res. 41, 422 (1998). 82. M. Franco, P. F. Nealey, S. Campbell, A. I. Teixeira and C. J. Murphy, J. Biomed. Mater. Res. 52, 261 (2000). 83. H. J. Griesser, R. C. Chatelier, T. R. Gengenbach, G. Johnson and J. G. Steele, J. Biomater. Sci. Polymer Edn 5, 531 (1994). 84. J. L. Bohnert, B. C. Fowler, T. A. Horbett and A. S. Hoffman, J. Biomater. Sci. Polymer Edn 1, 279 (1990). 85. D. Kiaei, A. S. Hoffman, T. A. Horbett and K. R. Lew, J. Biomed. Mater. Res. 29, 729 (1995). 86. J. G. Steele, B. A. Dalton, G. Johnson and P. A. Underwood, Biomaterials 16, 1057 (1995). 87. D. K. Pettit, T. A. Horbett and A. S. Hoffman, J. Biomed. Mater. Res. 26, 1259 (1992). 88. R. L. Williams, J. A. Hunt and P. Tengvall, J. Biomed. Mater. Res. 29, 1545 (1995). 89. P. Tengvall, L. Olsson, B. Wälivaara, A. Askendal, I. Lundström and H. Elwing, Adv. Biomater. 10, 511 (1992). 90. C.-G. Gölander, Y.-S. Lin, V. Hlady and J. D. Andrade, Colloid. Surf. 49, 289 (1990). 91. Y.-S. Lin, V. Hlady and J. Janatova, Biomaterials 13, 497 (1992). 92. L. Liu and H. Elwing, J. Biomed. Mater. Res. 28, 767 (1994). 93. A. Takahara, Y. Hara, K. Kojio and T. Kajiyama, Colloid. Surf. B: Biointerf. 23, 141 (2002). 94. K. Nakamura, M. F. Refojo, D. V. Crabtree, J. Pastor and F.-L. Leong, Invest. Ophthalmol. Vis. Sci. 32, 3007 (1991). 95. S. Bondurant, V. Ernster and R. Herdman (Eds), Safety of Silicone Breast Implants. National Academy Press, Washington, DC (2000). 96. Y. Lakshmanan and S. G. Docimo, J. Long-Term Eff. Med. Impl. 7, 65 (1997). 97. E. P. J. M. Everaert, H. F. Mahieu, B. van de Belt-Gritter, A. J. G. E. Peeters, G. J. Verkerke, H. C. van der Mei and H. J. Busscher, Arch. Otolaryngol. Head Neck Surg. 125, 1329 (1999).
This page intentionally left blank
Part III
Materials
This page intentionally left blank
Elastomeric biodegradable polyurethane blends for soft tissue applications J. D. FROMSTEIN and K. A. WOODHOUSE ∗ Department of Chemical Engineering and Applied Chemistry, Institute of Biomaterials and Biomedical Engineering, Toronto, Ontario, M5S 3E5, Canada Received 29 November 2001; accepted 11 March 2002 Abstract—Four biodegradable polyurethane blends were made from segmented polyurethanes that contain amino acid-based chain extender and diisocyanate groups. The soft segments of these parent polyurethanes were either polyethylene oxide (PEO) or polycaprolactone (PCL) diols. The blends were developed to investigate the effect of varying soft segment compositions on the overall morphological, mechanical, and degradative properties of the materials, with a view to producing a family of materials with a wide range of properties. The highly hydrophilic PEO material was incorporated to increase the blend’s susceptibility to degradation, while the PCL polyurethane was selected to provide higher moduli and percent elongations (strains) than the PEO parent materials can achieve. All four blends were determined to be semi-crystalline, elastomeric materials that possess similarly shaped stress-strain curves to that of the PCL-based parent polyurethane. As the percent composition of PEO polyurethane within the blend increased, the material became weaker and less extensible. The blends demonstrated rapid initial degradation in buffer followed by significantly slower, prolonged degradation, likely corresponding to an initial loss of primarily PEOcontaining polymer, followed by the slower degradation of the PCL polyurethane. All four blends were successfully formed into three-dimensional porous scaffolds utilizing solvent casting/ particulate leaching methods. Since these new blends possess a range of mechanical and degradation properties and can be shaped into three-dimensional objects, these materials may hold potential for use in soft tissue engineering scaffold applications. Key words: Polyurethane; elastomer; biodegradable; scaffold; tissue engineering.
INTRODUCTION
In recent years, much attention has been focused on the development of biomaterials for use in tissue engineering applications. The increasing shortage of tissue and organ donors has driven the development of viable tissue substitutes [1]. A highly attractive approach to tissue engineering involves the use of biodegradable ∗ To
whom correspondence should be addressed. E-mail:
[email protected]
230
J. D. Fromstein and K. A. Woodhouse
scaffolding materials that can be replaced by natural tissue within the body. To date, most of the biodegradable materials available for use in tissue engineering are hard, brittle substances best suited to bone and hard tissue applications [2– 4]. Numerous tissues that need to be replaced or regenerated, however, are soft tissues that require substantial elasticity. Thus, there is a considerable need for synthetic degradable materials that exhibit the properties of elastic recoil and malleability. Skarja and Woodhouse have developed a family of novel biodegradable segmented polyurethanes (PUs) with a wide variety of chemical and mechanical properties suitable for use in soft tissue applications [4– 6]. The hard segments of these materials are composed of a phenylalanine-based diester chain extender (Phe) and a lysine-based diisocyanate (LDI). The soft segments are either a polycaprolactone (PCL) or polyethyleneoxide (PEO) diol. The subset of these materials containing the PCL soft segment possesses particularly high moduli and ultimate tensile stresses, and relatively slow degradation rates in vitro [6]. The PEO-containing polyurethanes, however, are tacky, viscous materials that cannot be formed into structures and degrade significantly more quickly [4]. In this work, these biodegradable segmented polyurethanes have been combined to form four polyurethane blends that take advantage of the mechanical properties provided by the PCL-based polyurethanes and the high degradation rates of the PEO-based materials. The new blends were characterized extensively, and initial investigations into the fabrication of three-dimensional scaffolds were performed.
MATERIALS AND METHODS
Materials Three polyurethanes were used in this study to make the four blends: PCL2000/ Phe (Mw 185 750; Mw /Mn 2.06), PEO600/ Phe (Mw 46 140; Mw /Mn 1.91) and PEO1000/ Phe (Mw 173 720; Mw /Mn 2.37) [4]. The parent polyurethanes were synthesized following the methods of Skarja and Woodhouse [4, 5], using 2,6diisocyanato methyl caproate (LDI), obtained from Kyowa Hakko Kogyo Co. Ltd. (Tokyo, Japan), polycaprolactone diol, molecular weight 2000 (PCL2000) and polyethylene oxide diol, molecular weights 600 and 1000 (PEO 600 and 1000 respectively), from Aldrich Chemicals (Milwaukee, WI), and an L-phenylalanine based diester chain extender (Phe), synthesized as described previously [5]. The nomenclature for the parent polyurethanes indicates the soft segment diol type and molecular weight, followed by the chain extender identifier (Phe). Casting materials included chloroform, obtained from ACP Chemicals Inc. (Montreal, PQ), and confectioner’s sugar (Lantic Sugar Limited, Toronto, ON). Fabrication of the blends For each blend, the parent polyurethanes (PCL2000/ Phe, PEO600/Phe and PEO1000/ Phe) were weighed out according to the mass ratios listed in Table 1.
Degradable elastomeric polyurethane blends
231
Table 1. Nomenclature and compositions of the four degradable polyurethane blends Polyurethane blend
PCL2000/ Phe (% by mass)
PEO600/ Phe (% by mass)
PEO1000/ Phe (% by mass)
Blend 1 Blend 2 Blend 3 Blend 4
75 75 50 50
25 — 50 —
— 25 — 50
The polymers were then dissolved together in chloroform at a concentration of 5% w /v, and filtered using Whatman 42 ashless filter paper (Fisher Scientific, Mississauga, ON). Next, the polyurethane solution was poured into polytetrafluoroethylene (PTFE) casting dishes, and covered to allow for a controlled evaporation rate. Following 48 h evaporation, the films were then placed in a vacuum chamber (0.02 mmHg) at room temperature for another 48 h to remove any residual solvent, and then stored under desiccated conditions prior to use. Polyurethane blend characterization Proton nuclear magnetic resonance spectroscopy (NMR) was performed in order to assess the chemical structures and compositions of the new blends, using d6 −dimethyl sulphoxide (DMSO) as the solvent. The dissolved polyurethane samples (1% w /v) were analyzed by performing 64 scans per sample using a 400 MHz Bruker spectrometer. Control spectra of the individual components (soft segment diols, diisocyanate and chain extender) were run in order to aid in peak assignment and identification. Polyurethane blend molecular weights and distributions were assessed using gel permeation chromatography (GPC). The polymers were dissolved at 0.1% w /v (blends 1 and 2) and 0.2% w /v (blends 3 and 4) in HPLC-grade dimethylformamide (DMF) containing 0.05 M lithium bromide. Samples (200 µl) were injected into an HPLC column bank containing styrenedivinylbenzene copolymer packed columns (HR2, HT3 and HT4, Waters, Milford, MA). Number and weight average molecular weights (polystyrene equivalents) were determined from retention time data, collected using Waters Baseline™ software (Waters Chromatography, Mississauga, ON), and a calibration curve generated with polystyrene standards (Tosoh Corp., Tokyo, Japan). Differential scanning calorimetry (DSC) was performed in order to determine soft segment crystallinities and glass transitions using a Thermal Analyst 2100 thermal analyzer located at the Brockhouse Institute for Materials Research (McMaster University, Hamilton, ON). The samples used for DSC all had identical processing histories prior to analysis. In addition, the blends had been maintained at room temperature at all times. Scans were run from −140◦ C to 250◦ C at a ramping rate of 15◦ C /min. Glass transition (Tg ) values were identified as the midpoints
232
J. D. Fromstein and K. A. Woodhouse
of the inflexion point in each spectrum. Percent crystallinity of the soft segment was calculated using the area under the melt endotherm to determine the heat of fusion, and comparing this value with the enthalpy of fusion of 100% crystalline polycaprolactone (32.4 cal/ g) [7]. In preparation for morphological determination using polarized light microscopy, samples were cut from the pre-cast films into 8 mm discs with a metal punch. These discs were then placed under crossed polarized filters on a Leitz Orthoplan microscope (Leitz, Germany), and the images captured using a passive image capture (PCI) system, Quartz PCI (Quartz, Vancouver, BC). Uniaxial tensile testing was carried out using a SINTECH 1 Instron testing machine (SINTECH, Stoughton, MA) following ASTM standard D638M. Prior to testing, samples were conditioned at 23◦ C and 50% relative humidity for a minimum of 48 h. Samples were stretched until failure at a strain rate of 500 mm /min using a 50 lb load cell. Data was collected using TestWorks II, v. 2.10p software (SINTECH). Stress vs. strain curves were generated from this data in Microsoft® Excel (Microsoft Corp., Seattle, WA). Degradation studies Degradation studies were performed using tris-buffered saline (TBS, 0.05 M Tris, 0.1 M NaCl, pH 7.4). Pre-cast samples were punched into 8 mm discs, and weighed to an accuracy of 0.1 mg using an analytical balance (HR-120, A&D Company, Ltd., Tokyo, Japan). Each sample was placed into an individual vial containing 15 ml TBS, and incubated at 37◦ C. Four samples of each blend were removed from buffer following 3 days, 1 week, 2 weeks, 4 weeks and 6 months. After drying under vacuum at room temperature for 1 week, samples were reweighed to determine total percentage of mass lost. Samples were then analysed using scanning electron microscopy (SEM). The films were mounted on metal stubs, and sputtered with a 5 nm thick layer of titanium. These samples were examined under an S-2500 scanning electron microscope (Hitachi, Japan), and images were captured using the Quartz PCI system. Porous scaffold formation Prior to fabrication of three-dimensional scaffolds, the polyurethane blends were dissolved in chloroform, at a concentration of 5% w /v. A mass of 1.6 g confectioner’s sugar (particles 150 µm in size) was then placed in PTFE vials 18 mm in diameter, and levelled off using a low-frequency shaking platform (Rotomix 50800, Thermolyne, Dubuque, IA) to help achieve even particulate distribution. Next, 3 ml of chloroform-polyurethane blend solution was added to the vials, taking care not to disturb the sugar within the vials, and to distribute the solution evenly around the surface of the sugar. The vials were then covered lightly, and the solvent allowed to evaporate for 48 h. Subsequent to drying, the samples were soaked in
Degradable elastomeric polyurethane blends
233
pyrogen-free water (changed twice daily) for 4 days in order to leach the sugar out of the scaffolds. Finally, the materials were dried in a vacuum chamber at room temperature for 48 h. The scaffolds were prepared for SEM following the same procedure outlined for the degradation samples above, and were observed using the same microscope.
RESULTS AND DISCUSSION
Polyurethane blend characterization The proton NMR spectra for all four polyurethane blends are very similar. Representative NMR spectra of blends 1 and 3 are shown in Figs 1A and B, respectively. Since the only differences between the blends are present in the relative composition and molecular weights of the soft segment diols, the proton environments in all four materials should be identical [8]. Only the ratio of peak areas should vary from one material to the next. Once incorporated into the polyurethane backbone, PEO produces one chemically distinct proton environment under NMR, appearing as a peak at a chemical shift of 3.5 ppm (∗ 1). Polycaprolactone contributes four chemically unique protons represented by: two doublets of triplets (or hextets), one at 1.3 ppm (∗ 2), and a second between 1.5 and 1.6 ppm (∗ 3), and two triplets, one at approximately 2.25 ppm (∗ 4), and the other at 4.0 ppm (∗ 5). These peaks that we have identified correspond to those in the literature for PEO /PCL copolymers [9]. In order to confirm final percent content of each of the two soft segments within the blends, quantitative analysis of the NMR spectra was attempted. However, comparison of the relative peak areas, both as a whole and as intensity per proton, did not provide reproducible results. For each blend, three samples were run on NMR spectra, and the ratios of peak area and proton intensity varied greatly not only from one scan to the next, but also from one PCL-peak to the next within a given scan. These difficulties were expected to be the result, in part, of ‘interference’ from ethyl-derived protons located elsewhere in the polyurethane backbone. Some of the protons, particularly those in the LDI, possess highly similar chemical environments to those in the PCL soft segment. Even taking these protons into account, however, calculated ratios were inconsistent. Given that accurate comparison within scans of the same material were not feasible, quantitative comparison between spectra of one blend to another were deemed impossible at this time. Hence, proton NMR analysis provided qualitative data only. Figure 1 illustrates that the PCL-related peaks (peaks 2 to 5) are significantly larger, relative to the other peaks, in 1A (blend 1) than in 1B (blend 3). Similarly, in relation to the area under the PCL peaks, the PEO resonance at 3.5 ppm is also significantly larger in blend 3 than in blend 1. These results were anticipated since the original blending ratios of PCL to PEO were 1 : 1 and 3 : 1, respectively. GPC results (polystyrene equivalent data) for the polyurethane blends are summarized in Table 2. Blends 1 and 3 have lower number average molecular weights than
234
J. D. Fromstein and K. A. Woodhouse
Figure 1. Proton NMR spectra of (A) blend 1 (75% PCL2000/ Phe : 25% PEO600/ Phe) and (B) blend 3 (50% PCL2000/ Phe : 50% PEO600/ Phe).
their PEO1000-containing counterparts. These results agree with previous work on the parent polyurethanes, which indicated that the PEO600-based polyurethane has a significantly lower average molecular weight than the PEO1000 PU [4]. The molecular weights of blend 3 are notably lower than those of the other three blended materials. This difference is likely caused by the low molecular weight of
Degradable elastomeric polyurethane blends
235
Table 2. Gel permeation chromatography results: molecular weights (polystyrene equivalents) and polydispersity (n = 3) Polyurethane
Mn (×103 )
Mw (×103 )
Polydispersity Index (Mw /Mn )
Blend 1 Blend 2 Blend 3 Blend 4
93 101 31 145
182 181 120 234
2.0 1.8 3.8 1.6
Table 3. Differential scanning calorimetry data, represented as mean ± standard deviation (n = 5). Scans were run from −140◦ C to 250◦ C at a ramping rate of 15◦ C/min Polyurethane
Blend 1 Blend 2 Blend 3 Blend 4
Soft segment Tg (◦ C)
Tm (◦ C)
% Crystallinity
−16 ± 9 −22 ± 11 −20 ± 3 −38 ± 8
48 ± 3 47 ± 2 47 ± 1 47 ± 1
27 ± 7 24 ± 5 18 ± 1 19 ± 1
PEO600/ Phe, which comprises approximately half of the blend. In addition, the highly different molecular weights of the two parent polyurethanes contribute to a wider molecular weight distribution, as represented by Mw /Mn . Differential scanning calorimetry results (Table 3) indicate that as the soft segment molecular weights increase, the glass transition temperature (Tg ) decreases. This finding was only statistically significant for blends 3 and 4, however. This trend was also seen for the parent polyurethanes in a previous study [4]. A second trend indicated a decrease in the glass transition temperature when the amount of PCL2000/ Phe in the blends was decreased and the amount of PEO /Phe increased (e.g. blend 1 vs. 3 and blend 2 vs. 4). These data were significant for the comparison of blends 2 and 4, but not for blends 1 and 3. It is well known that Tg determinations using the inflection point tend to have a large amount of error associated with them [10, 11]. In addition, the chemical natures of all four blends are highly similar. Hence, the lack of statistical significance in the trends related to blend 1 could be due in part to the comparable properties of the materials themselves, and in part to errors inherent in the Tg determination method. All of the polyurethane blends have very similar soft segment melt temperatures (Tm ), ranging from approximately 47◦ C to 48◦ C. This Tm is similar to values obtained by other researchers for various multiblock copolymers containing PEO and PCL [12– 14]. In most copolymers containing PEO, the Tm is lower for the copolymer than for the PCL-based homopolymers [12– 15]. For the polyurethanes used in this study, however, the melt temperature of the parent polyurethane
236
J. D. Fromstein and K. A. Woodhouse
PCL2000/ Phe is lower (45.1◦ C) than the Tm values of the polyurethane blends [4]. Gan et al. [15] showed that the presence of PEO influences the crystallization of polycaprolactone by decreasing the crystallization entropy and providing nucleating sites for PCL block crystallization. Thus, in the polyurethane blends described here, the presence of PEO material may serve to decrease the crystallization entropy, represented by better crystal organization, thereby causing an increase in Tm . Using the area under the melt endotherms, percent crystallinity of the soft segments was calculated for each blend (Table 3). All four blends possessed lower percent crystallinities than PCL2000/ Phe, which is approximately 46% crystalline. The polycaprolactone soft segment parent PU is semi-crystalline, while the PEO polyurethanes are amorphous materials. Hence, only PCL contributes directly to the soft segment crystallinity. However, PEO does have a direct effect on the crystallization of PCL, as discussed above. As expected, the blends containing primarily PCL-based polyurethane (blends 1 and 2) had higher soft segment crystallinity than the blends containing approximately equal amounts of PCL and PEO polyurethanes. In addition to the smaller amount of PCL available to crystallize in blends 3 and 4, the presence of a larger amount of PEO may actually interfere with the PCL-PEO interactions necessary to improve PCL’s crystallization. Polarized light micrographs of blends 1 through 4 are shown in Fig. 2. All four blends exhibit the characteristic Maltese cross pattern indicative of spherulites [16– 18]. These results are consistent with the DSC data indicating that the blends are semi-crystalline in nature. While DSC indicated that the higher PCL-containing blends are more crystalline than the other two, these differences were not significant enough to be observed under crossed polarized filters. In spherulitic patterns, the crosses represent folded crystalline lamellae, surrounded by amorphous material. The PEO polyurethanes have been shown to be completely amorphous materials, and the PCL PU is only semi-crystalline. Hence, in blends 3 and 4, the materials are predominantly amorphous. However, the polyurethane blends still manage to orient themselves into spherulites, indicating that the PEO polyurethane’s presence in the blend does not hinder the formation of PCL crystalline regions significantly. In fact, as discussed above, the PEO may actually encourage PCL crystallization. These results differ from those of Cohn et al., who found that in poly(ether ester urethane)s consisting of triblock copolymers of PEO and PCL chain extended with hexamethylene diisocyanate (HDI), when the PEO segment length became too large, it interfered with the PCL’s ability to crystallize [13]. In Cohn et al.’s study, however, the PEO segment molecular weights went as high as 20,000. In addition, in their material, the PEO and PCL segments are located directly next to each other in the polyurethane backbone. In this study’s blends, however, the PEO and PCL are all in separate chains that will only be bound through weak hydrogen bonding, etc. rather than through a rigid covalent bond. This composition allows for the PCL and PEO domains to shift and organize themselves more easily in order to encourage interactions between the two groups.
Degradable elastomeric polyurethane blends
237
Figure 2. Polarized light micrographs of the four polyurethane blends, at 40× original magnification.
Representative stress strain curves of all four polyurethane blends are provided in Fig. 3, and the tensile properties of the materials are summarized in Table 4. Statistical significance analysis of the data in Table 4 was performed using the Student’s t distribution for a probability value of 0.05. In Fig. 3, all four stressstrain curves follow similar paths, demonstrating steep initial slopes, culminating in a yield point at a relatively low extension of approximately 10%. Following this yield point, the materials can be stretched further, out to an extension of over 500% for the higher PEO-containing blends 3 and 4, and to nearly 800% for blends 1 and 2. As was expected from the known mechanical properties of the parent polyurethanes [4], the blends containing a higher amount of PEO, either 600 or 1000 MW, have lower moduli, ultimate strengths and extensibilities than those containing predominantly PCL2000/ Phe. For a given initial composition of PEO (either 25% or 50%) polyurethane, no significant differences in mechanical properties were observed. Hence, a change of PEO molecular weight from 600 to 1000 does not appear to alter the ultimate tensile behaviour of the polyurethane blends. Frequently, an increase in molecular weight corresponds to an increase in ultimate strength. Kim et al. created blends containing PCL, PEO and cornstarch [19]. They found that for PCL and PEO molecular weights similar to those used in the current study, increasing the molecular weight
238
J. D. Fromstein and K. A. Woodhouse
of the PEO did improve the strength of their materials. It is possible that since their blends contain only the PCL and PEO diols, the two materials can move around more easily, and can therefore form a greater concentration of chemical bonds to one another. In the polyurethane blends, interference and hindrance from the hard segment may prevent the additional PEO present in the higher molecular weight soft segment from forming any additional bonds. Although molecular weight of the overall materials should also have some effect on the mechanical properties, no such trends were evident. For instance, blend 3 has a significantly lower molecular weight than blend 4, but no difference in tensile properties was seen. Again, this may be due largely to the number of bonds formed between the PCL soft segment regions and the PEO regions. Once the maximal number of PCL-PEO interactions has been achieved, the PCL /Phe portion of the material will provide the remainder of the material’s strength. Hence, molecular weight of the overall blend may actually have a minimal effect on the ultimate mechanical properties of the blends. Table 4. Mechanical properties of polyurethane blends, as determined by uniaxial tensile testing. Values indicate mean values, ± standard deviation for an n of 6 Polyurethane blend
Ultimate tensile stress (MPa)
Ultimate tensile strain (% elongation)
Young’s modulus (MPa)
Blend 1 Blend 2 Blend 3 Blend 4
20 ± 3 26 ± 8 10 ± 2 6±1
690 ± 70 726 ± 151 510 ± 64 512 ± 49
96 ± 8 105 ± 19 76 ± 16 49 ± 10
Figure 3. Representative stress vs. strain curves for all four blends, as obtained through uniaxial tensile testing.
Degradable elastomeric polyurethane blends
239
Degradation studies The effects of incubating the blended materials in buffer at 37◦ C were investigated out to 6 months, and assessed using percent mass loss and appearance under SEM. Data comparing the percentages of total mass lost following two weeks of incubation in buffer with those after six months of incubation are given in Fig. 4. Within the first two weeks of the study, a significant percent mass loss was observed. This was on the order of 5– 10% for the predominantly PCL-based blends and approximately 20– 25% for the other two blends. Following this initial rapid loss of material, the subsequent degradation rate was considerably slower, as is evidenced by the six-month percent mass loss values. For the 75/ 25 blends, the same amount of material was lost over the following five and a half months as was lost during the first two weeks. For the PEO-rich blends 3 and 4, minimal additional mass loss was seen after the first two weeks. Li et al. observed similar degradation trends for their PCL /PEO copolymers, where the first two weeks demonstrated sharp mass losses, followed by a significantly slower degradation rate for the following 61 weeks [9]. It is suspected that the initial losses during the first two weeks represent exposed PEO /Phe content, whereas later mass loss occurs much more slowly, since it is dependent primarily on the degradation of the PCL-based polyurethane material, which is much more resistant to degradation [6]. This was the pattern observed by Li et al. [9] as well. Li monitored their degradation using a variety of analytical techniques, including proton NMR [9]. Their analysis of the degraded samples indicated a rapid decrease in the ratio of PEO to PCL as degradation proceeded. These results were attributed to the hydrophilic nature of PEO. The copolymer’s hydrophilicity could be improved by increasing either the length of the PEO segments or the ratio of PEO to PCL. Cerrai et al. [20, 21] showed that
Figure 4. Mass loss results following two weeks and six months of incubation in buffer at 37◦ C. Values represent mean mass loss for an n of 4. Error bars represent 95% confidence intervals.
240
J. D. Fromstein and K. A. Woodhouse
in materials like these, an increase in hydrophilicity can be directly correlated to an increased initial degradation rate. Since the PEO polyurethanes in the blends are hydrophilic [4], blends 3 and 4 are expected to swell more in aqueous solution, thereby enhancing hydrolysis rates. It is therefore suspected that since most of the mass lost by blends 3 and 4 occurred in the first two weeks, the PEO material is likely attracted to the surface of the material due to its hydrophilic nature. Thus, the majority of the PEO content of the blends may have been exposed at the surface, facilitating its degradation within the first two weeks. During the subsequent slow degradation of the predominantly PCL2000/ Phe remaining, small pockets of additional PEO /Phe may have become exposed. These PEO-rich regions will likely have been degraded quickly once released from their PCL /Phe surroundings, leaving a semi-crystalline PCL polyurethane surface exposed once again to degrade very slowly. The molecular weight of PEO diol used also appeared to have an effect on the amount of degradation sustained by the 75% PCL-based blends following two weeks of incubation. Blend 2 demonstrated a significantly higher percent mass loss than blend 1, a phenomenon that may also be attributed to the hydrophilic nature of PEO since the PEO chains are longer in blend 2 than blend 1. In addition to the hydrophilicity effects on mass loss, the presence of crystallinity will also play a role in determining degradation rates. Highly ordered crystalline materials are significantly more resistant to hydrolysis than amorphous materials. Hence, while the PEO content can be used to explain the mass loss profile over time, the morphology of the polyurethane blends must be utilized when considering the relative degradation rates of the four materials. Since blends 3 and 4 contain a lower percent crystallinity, it would be expected that these materials would degrade more quickly than blends 1 and 2. SEM analysis of the polymeric materials following degradation further enhances the speculation that the PEO-containing material is being preferentially lost. Figures 5A and 5B illustrate the degradation of blend 1 and blend 3, respectively, as a function of time. These photographs are representative of the PEO1000 blends as well. In the time 0 images of Fig. 5, it is apparent that the materials possess a partially porous nature prior to degradation. These voids may be caused by the presence of moisture during and following the film casting process. Chen et al. [22] showed that during solvent evaporation from a PCL /PEO block copolymer, the PEO segments migrated to the surface of the material. Therefore, in the polyurethane blends, the PEO regions may have preferentially migrated to the surface, where they could have been exposed to moisture and therefore degraded, resulting in micropore formation. For the 75/ 25 blend, after 3 days incubation in buffer, numerous voids could be seen on the surface of the material. These holes are relatively small, on the order of 0.2 to 0.5 µm, and while they are visible out to 6 months, their overall appearance did not appear to change significantly over time. With the higher PEO content materials, however, a significantly different degradation pattern was seen. For blend 3, as time increased, the voids enlarged, growing until they merged with
Degradable elastomeric polyurethane blends
241
Figure 5. Scanning electron micrographs of (A) blend 1 (75% PCL2000/ Phe : 25% PEO600/ Phe) and (B) blend 3 (50% PCL2000/ Phe : 50% PEO600/ Phe) following 0 days, 3 days, 1 week, 2 weeks and 6 months of degradation. Images were taken at 500× magnification. Bars represent 60 µm.
one another, forming large crevices. In the centres of these crevices, small holes began to form in the next level of material. It appears that as time increased, these holes within the crevices enlarged, thereby eroding the material one level at a time. Some of the crevices grew to the order of approximately 6 µm in size. These degradation patterns are consistent with the percent mass loss results shown earlier. Following one week of incubation time, the materials’ appearances under SEM remained relatively constant, indicating uniform degradation across the surface over time. Porous scaffold formation Scanning electron micrographs of the preliminary three-dimensional matrices created using the four polyurethane blends are shown in Fig. 6. All of these materials have similar appearances, characterized by a highly interconnected network of pores, with pore sizes of approximately 1 to 3 µm in size. No effect of varying blend composition or PEO molecular weight was seen on the ability to create porous scaffolds. In order to confirm that the interconnectivity observed through SEM is not an artefact, a quantitative method such as mercury intrusion porosimetry could be used. Future scaffolds will be characterized using this technique. It is anticipated that through the use of larger size sugar crystals or alternate porogens, larger pore sizes can be achieved without losing the interconnectivity between pores seen in these specimens. In addition, both in vitro and in vivo, as degradation proceeds, the size of the pores and degree of interconnectivity is expected to increase. Hence, upon
242
J. D. Fromstein and K. A. Woodhouse
Figure 6. SEM images of porous scaffolds constructed using the blends. The images in the top row were taken at 250× magnification, and the black bars represent 120 µm. The detail of the scaffolds is shown at 5000× magnification in the bottom row, where the white bars represent 6 µm.
implantation, the porous properties of the polyurethane blend scaffolds will change with time.
CONCLUSIONS
In this work, four segmented elastomeric polyurethane blends were produced. The effects of soft segment diol molecular weight and ratio of PCL polyurethane to PEO polyurethane on the polymer blend characteristics and degradation properties were evaluated. Finally, the blends were formed into three-dimensional scaffolds. There was no clear effect of blending ratio on molecular weight of the polyurethanes. Differential scanning calorimetry indicated that blends 1 and 2 have higher glass transition temperatures than the corresponding blends 3 and 4. Using DSC, it was also shown that the higher PCL-content blends contain more percent crystallinity in the soft segment than the other two blends. These differences in crystallinity were not significant enough, however, to be visible via microscopy under crossed polarized filters. Mechanically, the higher PCL-content blends 1 and 2 had higher Young’s modulus, ultimate tensile stress and ultimate tensile strain values. The higher PEO content blends demonstrated significantly increased percent mass lost over time than the predominantly PCL blends, owing largely to the hydrophilic nature of PEO. Trends based on the PEO diol molecular weight utilized were also observed. Since PEO600/ Phe is a significantly lower molecular weight polyurethane than the
Degradable elastomeric polyurethane blends
243
PEO1000 PU, the blends created using PEO600 had significantly lower molecular weights as observed using GPC. For the PEO600 materials, the polydispersity values were also elevated. Using DSC, it was determined that the Tg values are lower for the PEO1000-based blends than for those containing PEO600. For the 75/ 25 blends, however, the Tg values were not statistically different. Tm remained constant around 47– 48◦ C for all four blends. No clear effect of diol molecular weight was seen on the percent crystallinity. These results are expected since the PEO polyurethanes are both amorphous materials, and should therefore not play an important role in this factor. Changing PEO molecular weight did not appear to affect the mechanical tensile properties of the blends in a consistent manner. In particular, regardless of which molecular weight diol was incorporated into the blend for a given percent PEO composition, the ultimate tensile strains were not statistically different. No distinguishable effects of soft segment molecular weight were seen on prolonged degradation rates. Following two weeks degradation, however, blend 2 demonstrated significantly higher mass loss than blend 1. Hence, the higher molecular weight PEO may improve the initial degradation of the materials. This effect can be attributed to the presence of longer PEO chains allowing for increased water adsorption. The presence of a greater volume of water within the polyurethane can allow for a greater amount of hydrolysis to take place. In this work, we have produced blends of degradable polyurethanes with modifiable degradation and mechanical properties with potential applications as soft tissue engineering scaffolds. Acknowledgements Funding for this work was provided by Materials and Manufacturing Ontario and the Natural Sciences and Engineering Research Council of Canada. The authors would also like to thank Gary Skarja for his guidance and advice.
REFERENCES 1. J. P. Vacanti and C. A. Vacanti, in: Principles of Tissue Engineering, R. Lanza, R. Langer and W. Chick (Eds), p. 1. R. G. Landes Company, Austin, TX (1997). 2. I. Engelberg and K. Joachim, Biomat. 12, 292 (1991). 3. M. Vert, J. Feijen, A. Albertsson et al. (Eds), Biodegradable Polymers and Plastics. The Royal Society of Chemistry, Cambridge (1992). 4. G. A. Skarja and K. A. Woodhouse, J. Appl. Polym. Sci. 75, 1522 (2000). 5. G. A. Skarja and K. A. Woodhouse, J. Biomater. Sci. Polymer Edn 9, 271 (1998). 6. G. A. Skarja and K. A. Woodhouse, J. Biomater. Sci. Polymer Edn 12, 851 (2001). 7. V. Crescenzi, G. Manzini, G. Calzolari et al., Eur. Polym. J. 8, 449 (1972). 8. F. A. Bovey and P. A. Mirau, NMR of Polymers. Academic Press, Inc., San Diego (1996). 9. S. M. Li, X. H. Chen, R. A. Gross et al., J. Mater. Sci.: Mater. Med. 11, 227 (2000). 10. T. Hatakeyama and F. X. Quinn, Thermal Analysis. John Wiley & Sons, New York (1994). 11. G. W. H. Hohne, W. Hemminger and H.-J. Flammersheim, Differential Scanning Calorimetry: An Introduction for Practitioners. Springer-Verlag, Berlin, Germany (1996).
244 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
J. D. Fromstein and K. A. Woodhouse M. L. Arnal, V. Balsamo, F. Lopez-Carrasquero et al., Macromolec. 34, 7973 (2001). D. Cohn, T. Stern, M. F. Gonzalez et al., J. Biomed. Mater. Res. 59, 273 (2002). S. Li, H. Garreau, M. Vert et al., J. Appl. Polym. Sci. 68, 989 (1998). Z. Gan, J. Zhang and B. Jiang, J. Appl. Polym. Sci. 63, 1793 (1997). P. C. Painter, in: Fundamentals of Polymer Science: An Introductory Text, P. C. Painter and M. M. Coleman (Eds), p. 197. Technomic Publishing Co., Pennsylvania (1997). P. Munk, Introduction to Macromolecular Science. Wiley-Interscience, New York (1989). C. J. Spaans, J. H. de Groot, F. G. Dekens et al., Polym. Bull. 41, 131 (1998). C.-H. Kim, E.-J. Choi and J.-K. Park, J. Appl. Polym. Sci. 77, 2049 (2000). P. Cerrai, M. Tricoli, L. Lelli et al., J. Mater. Sci.: Mater. Med. 5, 308 (1994). R. S. Del Guerra, C. Cristallini, N. Rizzi et al., J. Mater. Sci.: Mater. Med. 5, 891 (1994). D. Chen, H. Chen, J. Bei et al., Polym. Int. 49, 269 (2000).
Influence of surface morphology and chemistry on the enzyme catalyzed biodegradation of polycarbonate-urethanes Y. W. TANG 1 , R. S. LABOW 2 , I. REVENKO 3 and J. P. SANTERRE 1,∗ 1 Department
of Biological and Diagnostic Science, Faculty of Dentistry, University of Toronto, 124 Edward St. Toronto, Ontario, Canada, M5G 1G6 2 University of Ottawa Heart Institute, 40 Ruskin Street., Ottawa, Ontario, Canada, K1Y 4W7 3 Digital Instruments, 112 Robin Hill Road, Santa Barbara, CA 93117, USA Received 28 November 2001; accepted 26 February 2002 Abstract—Polycarbonate based polyurethanes were synthesized with varying hard segment content as well as hard segment chemistry based on three different diisocyanates,1,6-hexane diisocyanate (HDI), 4,4 -methylene bisphenyl diisocyanate (MDI) and 4,4-methylene biscyclohexyl diisocyanate (HMDI). The surface chemistry and morphology were characterized using X-ray photoelectron spectroscopy (XPS) and atomic force microscopy (AFM). The polymers were incubated with cholesterol esterase (CE) in a phosphate buffer solution at 37 ◦ C over 10 weeks. XPS results showed that the surface chemistry changed as the size and chemistry of the hard segment varied within the materials. AFM images exhibited distinctive surface morphologies for all polymers, and this was particularly apparent with changes in the hard segment chemistry. The results showed that the surface of HDI polymers consisted of relatively stiff rod-like structures, which corresponded to the soft segment domains. Polymers with a higher HDI content exhibited a dense top layer containing a relatively higher hard segment component, covering the sub-surface matrix of rod like structures. The MDI based polyurethane had large aggregates on its top surface, which corresponded to the aggregation of harder components. The HMDI based polycarbonate-urethane presented a relatively homogeneous surface where no phase separation could be detected. The relative differences in hard and soft segment content in their surface structure was supported by XPS findings. The analysis of the biodegradation results, concluded that enzyme catalyzed biodegradation within these materials was initiated in amorphous soft segment regions located in the region of the interface between hard and soft segments. A higher hard segment content at the surface contributed significantly to an increase in biostability. The findings provided an enhanced understanding for the role of surface molecular structure in the enzyme catalyzed biodegradation of polyurethanes. Key words: AFM; XPS; polyurethanes; enzymes; hard segment; biodegradation; esterase.
∗ To
whom correspondence should be addressed. E-mail:
[email protected]
246
Y. W. Tang et al.
INTRODUCTION
The application of polyurethanes in the manufacture of implant devices and particularly long-term implants have been limited due to their relative instability in the bio-environment [1, 2]. The degradation mechanisms include hydrolysis to which polyester-urethanes are the most susceptible [3] and oxidative degradation which may be more related to polyether-urethanes [4]. The former degradation process has been catalysed by different hydrolytic enzymes, including papain [5], and inflammatory cell-derived enzymes, such as cholesterol esterase (CE), elastase, and carboxyl esterase [3, 6]. The second degradation mechanism includes oxidation processes described by metal ion oxidation (MIO), autooxidation (AO), and environmental stress cracking (ESC) [4]. In the early 90’s, a new class of polycarbonate (PCN) based polyurethanes attracted interest in the biomaterials market due in part to the material having similar mechanical properties to those of conventional polyurethanes, but with relatively enhanced oxidative stability [7, 8]. Tanzi et al. compared the chemical stability of Pellethane 2363 80A and Corethane 80A (commercial polyether and polycarbonate urethanes, respectively) by incubating them in nitric acid (0.5 N) and hydrogen peroxide (20% w /v) [8]. The results showed that molecular reorganization occurred in both polymers, regardless of the aging media. Pellethane 2363 80A showed more chain scission than Corethane 80A when incubated with nitric acid. While Pellethane 2363 80A was degraded by hydrogen peroxide, Corethane 80A did not show any significant change after incubation with the latter medium. Therefore, the authors concluded that Corethane 80A presented a greater chemical stability than did Pellethane 2363 80A. It is accepted that the interactions between the biological systems responsible for degradation and the artificial surfaces occurs at the tissue-biomaterial interface, and it is reasonable to expect that the nature of this surface influences these interactions. Therefore, characterization of the surface is an important factor in defining tissue-biomaterial relationships, as well as providing information on the chemical structure, orientation and mobility of groups within the topmost atomic layers at the interface. It has also been shown that the composition of polymer surfaces often differs from the composition of the bulk material. While a number of techniques have been developed to characterize the surface, at present there is no single analytical method that can provide all the relevant information needed. X-ray photoelectron spectroscopy (XPS) has become an important technique for the surface analysis of biomaterials in the past two decades because it can provide chemical composition in a region closer to the surface than other available methods [9]. The advantages of XPS includes the speed of analysis, the high information content, the low sample damage potential, and the ability to analyze samples with no specimen preparation. The disadvantages, however, include the need for vacuum compatibility (i.e. no outgassing of volatile components), the possibility of sample damage if long analysis times are used, the need for experienced operators, and the cost associated with the analysis [10].
Influence of surface morphology and chemistry on the enzyme
247
Atomic force microscopy (AFM) is one of the most recent techniques applied to the study of biomaterial surfaces. Just a few examples include imaging contact lenses [11], orthodontic wires [12], hydrogels [13], bone [14], and scaffolds [15]. With the development in imaging samples in tapping mode, this technique has allowed for the collection of data by phase contrast imaging [16– 18] (also called phase imaging or phase mode). Aside from providing a better visualization of small features on the surfaces, due to a higher sensitivity, phase mode also allows one to distinguish surface regions based on different mechanical properties [17]. While the differences between polyether and PCN based polyurethanes have been primarily attributed to the chemistry of the soft segments, changes in hard segment structure can alar dramatically alter the phase separation of polyurethanes [19, 20], and hence lead to further changes in their interactions with bio-systems [19]. Tingey and Andrade showed that moderately sized and well phase-separated hard segment domains contributed to low protein adsorption and subsequently low platelet adhesion [21]. Hence, the ability to modulate protein or enzyme interactions at the surface of a polyurethane can be achieved by altering the domain structures and their surface distribution. Of particular interest would be the manner by which enzymes interact with polyurethanes in terms of surface initiated degradation. There is still very little known with regard to the role of surface molecular structure and the enzyme catalyzed biodegradation of polyurethanes although there have been studies with some hypotheses [4, 7]. In this study, six polycarbonate-based polyurethanes were synthesized with varying hard segment size and chemistry, but with the same soft segment. The materials were then degraded by incubation with solutions of CE in phosphate buffer. In order to differentiate the nature of each material surface prior to biodegradation, the polymer surfaces were characterized for surface chemistry and phase morphology using XPS and AFM, respectively. This analysis permitted the investigators to study the relationship between the native surface structure of each material and their respective susceptibility to hydrolysis by CE.
MATERIALS AND METHODS
Synthesis of polycarbonated polyurethanes with different hard segment Six segmented polycarbonated polyurethanes were synthesized with the same reagents, but with varying hard segment component content and chemistry. The diisocyanates used for the synthesis of the polyurethanes were 1,6-hexane diisocyanate (HDI, Aldrich, Milwaukee, WI, USA), 4,4 -methylene bisphenyl diisocyanate (MDI, Aldrich, Milwaukee, WI, USA) and 4,4-methylene biscyclohexyl diisocyanate (HMDI, Aldrich, Milwaukee, WI, USA). The same chain extender and soft segment were used for all the polymers and were 1,4-butanediol (BD, Aldrich, Milwaukee, WI, USA) and poly(1,6-hexyl 1,2-ethyl carbonate) diol (PCN, 1000, received in kind from Corvita Corporation, Miami, FL, USA), respectively. Table 1
248
Y. W. Tang et al. Table 1. The reagent and stoichiometry used in each polymer synthesis Polymer
Reagent and stoichiometry
Hard segment, wt%
HDI-312-41 HDI-211-30 HDI-321-23 HDI-431-20 MDI-321-30 HMDI-321-30
HDI : PCN : BD = 3 : 1 : 2 HDI : PCN : BD = 2 : 1 : 1 HDI : PCN : BD = 3 : 2 : 1 HDI : PCN : BD = 4 : 3 : 1 MDI : PCN : BD = 3 : 2 : 1 HMDI : PCN : BD = 3 : 2 : 1
41% 30% 23% 20% 30% 30%
lists the nomenclature for the polyurethanes, and is reported as DI-XXX-YY, where DI is the diisocyanate, XXX represents the molar ratio of DI : PCN : BD, and YY represents the weight percentage of the hard segments (Table 1). While HDI is less commonly used in the synthesis of commercial biomedical polyurethanes than methylene bis(phenyl isocyanate) (MDI) or methylene bis(cyclo hexamethylene isocyanate), it was selected for this study in part because of its availability in a radiolabelled form. This facilitated the monitoring of degradation by the enzyme. As well, the more recent use of HDI in the synthesis of polyurethanes where biodegradation is desired make it relevant. Prior to the synthesis, the diisocyanates and BD were vacuum distilled, while PCN was degassed under vacuum at 40 ◦ C overnight. The solvent used in the synthesis was N,N-dimethylacetamide (DMAC, Aldrich, Milwaukee, WI, USA). Because of the low reactivity of the diisocyanates with the polyol and BD, dibutyltin dilaurate (DBDA, Aldrich, Milwaukee, WI, USA) was used as a catalyst in the reaction. The synthesis was carried out as a conventional two-step procedure in a controlled atmospheric glove box containing dried nitrogen gas. Details of the reaction was provided elsewhere [22, 23]. The polymers were precipitated in a solution of 30% ether (Aldrich, Milwaukee, MI, USA) in distilled water in order to wash out the residual DBDA and low molecular weight oligomer. Physical properties including molecular weight, differential scanning calorimetery, small angle X-ray scattering, and Fourier transform infra-red spectroscopy were previously reported on for all six materials [22, 23]. Polymer surface characterization XPS. For XPS analysis, the polymers were dissolved in DMAC at a 10% (w /v) concentration and then 5 mL of this solution was poured into clean aluminum weighing dishes. The dishes were wrapped in Kimwipes® in order to prevent any contamination from the air. After being dried in an air oven at 50 ◦ C for 48 hours, the polymer was further dried under vacuum at 50 ◦ C for another 24 h. The polymer surface was then cleaned with HPLC-grade 1,1,2-trichlorotrifluoroethane (TCTF), using a foam applicator and rinsed with distilled water. This step was done to remove silicon contaminants accumulated at the surface. The samples were then
Influence of surface morphology and chemistry on the enzyme
249
wrapped with Kimwipes® to filter contaminants and dried in an oven at 50 ◦ C in order to remove the residual solvent, prior to being sent for analysis. The XPS measurement was carried out by Dr. Rana Sodhi at the Institute for Biomaterials and Biomedical Engineering, Surface Science Unit, University of Toronto, Toronto, Ontario. The polymer samples were mounted and analyzed using a Mg X-ray source at incident angles of 90◦ (upper 8 nm) and 15◦ (upper 1 nm). Low-resolution scans were performed over the entire electron energy spectrum to determine elemental concentrations of carbon, nitrogen, oxygen, tin, and silicon. High-resolution scans, which targeted the C1s bonding energies, were performed between −278 and −292 eV. A Leybold Max 200 instrument (Leybold, Cologne, Germany) was used for the XPS study. Matlab®for Windows (Version 4.2b) was used with an ESCA Toolkit software to analyze and deconvolute the spectra that were obtained. AFM. The polymer samples were cast by placing two drops of a 10% (w /v) polymer solution in DMAC onto a steel disc (15 mm in diameter). The coated disc was then dried in an air oven at 50 ◦ C for 24 h. The disc was put in a petri-dish in order to reduce the accumulation of air contaminants. The experiments were carried out on a Multimode AFM equipped with Nanoscope IIIa controller (Digital Instruments, Santa Barbara, CA) under ambient conditions. Silicon probes (TESP, Digital Instruments) with 125-µm-long cantilevers were used at their fundamental resonance frequencies, which typically varied between 270– 350 kHz. The scan frequency was 1.0 Hz. Images were taken with a free oscillating amplitude A0 ≈ 2 V, and a set-point amplitude A ≈ 1.0 V. The topographic images did not change for a large range of set-points. The phase images were dependent on the A/A0 ratios such that at a relatively high ratio, A/A0 = 0.9, there was no phase contrast and for small values of the ratio (about 0.3) the tip was pushing too hard on the surface and the topographic images appeared flat. For this study an average A/A0 value of 0.5 was selected, where phase contrast was clear and where the topographic data were reproducible. Experiments were repeated with different tips to avoid artifacts due to tip shape. All roughness calculations were performed directly with the commercial software provided with the AFM, and the same surface area was always used, because roughness calculations are size dependent. The system was operated in Tapping Mode™, which consists of an oscillation of the tip normal to the plane of the surface. The height data were complemented with a simultaneous measurement of phase shift. Topography was measured as the tip distance normal to the surface was adjusted to retain the constant amplitude of the oscillation tip, thereby allowing roughness data to be obtained. The phase change was measured as a function of the phase shift or phase lag between the drive signal and the actual tip response signal. Biodegradation experiments The polyurethanes used in the biodegradation experiments were synthesized with 14 C-HDI, or 14 C-BD (custom synthesis from NEN DuPont, Mississauga, Ontario)
250
Y. W. Tang et al.
in order to provide a sensitive measurement of product release in the study. The radiolabelled polyurethanes were cast, as previously described [24], onto hollow glass tubes (2 mm ID, 3 mm OD) using a 10% w /v polymer solution in DMAC. The coated tubes were then dried in a vacuum oven for 48 h at 50 ◦ C after having been dip coated three times with intermittent overnight drying at 50 ◦ C. The coated tubing was sectioned into 10 pieces of 2.55 mm in length (to yield an approximate total surface area of 36 cm2 ) and the sections were placed into a sterile 15 ml vacutainer (Becton-Dickinson). All incubation experiments were prepared in a laminar flow hood, employing sterile technique. As in previous studies [3, 25], CE was selected as the hydrolytic enzyme for the study since it represents an enzyme activity that is present in increasing levels as human monocytes differentiate into macrophages [26]. Hence, it is a representative model enzyme for the chronic inflammatory response to foreign substances. Vacutainers were set up with 0.05 M phosphate buffer, pH 7.0, either with bovine pancreatic CE (E.C. 3.1.1.13, Genzyme Diagnostics, Cambridge, MA, USA) or without (control buffer solution). CE solutions were prepared by dissolving the CE powder in 0.05 M phosphate buffer, pH 7.0, at a concentration of 160 units/ ml. 1 unit/ ml of CE is defined as the required amount of enzyme to generate 1 nmol/min of p-nitrophenol, formed from hydrolyzing p-nitrophenyl acetate in the presence of CE at pH 7.0 and 25 ◦ C. All solutions were sterile filtered using a 0.22 µm pore size filter (Millex-GP filter unit, Millipore Corporation, Bedford, MA, USA). Aliquots were removed every week from the polymer incubation solutions and counted in a liquid scintillation counter (Model LS 6500, Beckman Instrument Canada, Toronto, ON) for radioactivity. The enzyme activity lost between sampling points was calculated based on half-life data and replenished daily. Bacterial cultures were run on samples at the conclusion of all incubation periods in order to determine if sterility was maintained throughout the experiment. Triplicate samples for each incubation condition were run over ten weeks at 37 ◦ C. A Scheffe multiple comparison after one way analysis of variance was applied for each polymer incubation, with the dependent variable being the cumulative CPM /ml. The independent variable was the polymer type. The results were expressed as a mean ± standard error. The confidence interval was set at 95%.
RESULTS
XPS XPS analysis was employed in an attempt to assess the nature of the chemistry at the polymer’s outmost surface. Low-resolution spectra for the lower surface (90◦ incident angle) and the upper surface (15◦ incident angle) were analyzed for their concentrations of C, O, N, and Sn (a possible contaminant from the catalyst used in polyurethane synthesis). High-resolution scans of the C1s region (−278 and −292 eV) at 90◦ and 15◦ incident angles were analyzed to assess the changes in
Influence of surface morphology and chemistry on the enzyme
251
Table 2. Low resolution XPS data for the polyurethanes (atomic %) Polymer
HDI-312-41 HDI-211-30 HDI-321-23 HDI-431-20 MDI-321-30 HMDI-321-30
O
Sn
N
C
90◦
15◦
90◦
15◦
90◦
15◦
90◦
15◦
24.9 25.2 26.3 26.8 25.2 22.5
23.1 21.7 24.5 25.7 23.9 20.0
0.1 0.1 0.0 0.0 0.0 0.0
0.1 0.1 0.0 0.0 0.0 0.0
3.9 3.0 2.2 1.9 2.1 2.4
2.4 2.0 1.4 1.3 1.5 1.4
71.1 71.7 71.5 71.3 72.7 75.1
74.5 76.2 74.0 73.0 74.7 78.6
Table 3. C1s high resolution XPS data for the polyurethanes (atomic %) Polymer
HDI-312-41 HDI-211-30 HDI-321-23 HDI-431-20 MDI-321-30 HMDI-321-30
OCOO
NCOO
CO/ CN
CC
90◦
15◦
90◦
15◦
90◦
15◦
90◦
15◦
6.6 6.6 9.4 9.5 7.6 5.2
5.9 6.2 7.1 8.7 5.4 4.9
4.3 4.0 3.7 2.5 2.4 3.1
2.5 2.5 2.4 2.4 2.3 2.2
33.8 35.2 34.5 40.4 31.5 32.2
31.1 32.3 34.4 35.4 30.3 23.9
55.3 54.2 52.4 47.6 58.5 59.5
60.5 59.0 56.0 53.5 62.1 69.0
chemical groups at the surface. There were four main peaks, located at −290.5 eV, −289.2 eV, −286.6 eV and −285.0 eV. These peaks were assigned to the carbonate (OCOO), urethane (NCOO), ether or nitrile (CN) and carbon-carbon (CC) groups, respectively. The results of the analysis for each polymer are presented in Tables 2 and 3. The tin content was negligible for all six polymers. For each polymer, there was a lower nitrogen content within the deeper surface layer of the polymer when compared to the upper surface. This is a classical observation associated with polyurethanes [27] and has been related to the depletion of hard segment at the surface. The results showed that in conjunction with an increase in polymer hard segment content (see the HDI-polymers, Table 1), there was an increase in nitrogen content at the surface. The low-resolution oxygen and carbon data did not show obvious differences between the polymers with the exception of HMDI-321-30 which exhibited a lower oxygen and higher carbon content when compared to all of the other polymers. The high resolution XPS data showed that, within the HDI-polymers, the amount of urethane linkages increased as the hard segment content increased. This agreed with the elemental nitrogen data in Table 2. As expected, the carbonate linkages decreased as the hard segment content increased (Table 3). Among the 321polymers, HDI-321-23 exhibited the highest carbonate and urethane content at the
252
Y. W. Tang et al.
surface, while HMDI-321-30 showed the lowest carbonate content at the surface. As well, the C C linkage content was the highest in the upper surface of the latter polymer, suggesting that initially this material surface was perhaps more hydrophobic in comparison to the other polymers. AFM The surface morphologies of the polyurethanes were analyzed using AFM in Tapping mode. Using this technique, height and phase information was obtained. The lighter area in the phase images (Figs 1 and 2) corresponded to the region with the greater phase shift. The surface roughness (Table 4) could be obtained from the height images (reported on elsewhere [18]). HDI-312-41 exhibited the roughest surface among the materials, with a root mean square (RMS) roughness value of 24.05 nm. This value would be similar to the RMS value for a pair of new contact lenses [28], therefore indicating that all the polyurethane surfaces for this study were relatively smooth in comparison to biomaterials in general. Figure 1 shows the phase data for these six materials at a relatively low magnification (5×5 µm2 ). These images provided a most interesting comparison between the different materials. The phase image of HDI-312-41 (Fig. 1a) shows that this polymer surface was not homogeneous in terms of the pattern depicted by the tapping probe. Some parts of the surface showed rod like patterns and appeared meshed over each other. There are also small areas where these structures did not appear to be so dense, showing spacing between each other. AFM images of HDI-211-30 (Fig. 1b) also exhibited a heterogeneous surface. However, the formed layer of rods was less densely packed than those found in HDI-312-41 (comparing Figs 1a and 1b). The surface of HDI-321-23 appeared to be quite homogeneous and some small rod like features were evenly distributed throughout the surface (Fig. 1c). The surface of HDI-431-20 showed a phase image that was homogeneous without any clearly defined features at this magnification (Fig. 1d). While the AFM data indicated that MDI-321-30 was a relatively smooth surface (Table 4), it was not homogenous in terms of phase signals (Fig. 1e). There were some distinct phase contrast that resembled aggregate structures, which were clearly observed at this magnification. The edges of the dense aggregate on MDI-321-30 illustrates a crystallized appearance, Table 4. Roughness of polymers from AFM topographic images Polymer
RMS roughness (nm)
HDI-312-41 HDI-211-30 HDI-321-23 HDI-431-20 MDI-321-30 HMDI-321-30
24.05 3.30 3.82 8.09 2.67 0.36
Influence of surface morphology and chemistry on the enzyme
(a)
(b)
(c)
(d)
(e)
253
(f)
Figure 1. AFM phase data collected at a magnification of 5 × 5 (a) HDI-312-41; (b) HDI-21130; (c) HDI-321-23; (d) HDI-431-20; (e) MDI-321-30; (f) HMDI-321-23. µm2 :
254
Y. W. Tang et al.
(a)
(b)
(c)
(d)
(e)
(f)
Figure 2. AFM phase data collected at a magnification of 1 × 1 µm2 . (a) HDI-312-41; (b) HDI-21130; (c) HDI-321-23; (d) HDI-431-20; (e) MDI-321-30; (f) HMDI-321-23.
Influence of surface morphology and chemistry on the enzyme
255
with a sharp distinction from the background. In comparison to the HDI-polymers and MDI-321-30, the HMDI-321-30 material showed a very smooth surface (Table 4) with no defined features (Fig. 1f). Figure 2 shows a higher magnification of the AFM images for all the materials. At the higher magnification, phase images of HDI-312-41 showed more detail and depicted the presence of densely packed rod-like features throughout the surface meshed across each other (Fig. 2a). In general, the interpretation of information obtained from phase images is still not completely understood by experts in the field and therefore one must be careful in the analysis of the data. Magonov et al. proposed that the phase images could provide a map of stiffness variation on the sample surface [29]. His analysis stated that when the ratio of A0 (driving amplitude of the probe) over A (set-point amplitude) is in the region of 0.4 to 0.8, a stiffer region would have a more positive phase shift and appear brighter in the phase image. This was also confirmed by Bar et al. 1997 in their study of poly(ethyleneco-styrene) and poly(2,6-dimethyl-1,4-phenylene oxide) blends [30]. Therefore, in the current work, it is proposed that the rod-like features in Fig. 2a are stiffer than the background. As well, the areas with densely packed rod features (i.e. area with a continuous or ‘cemented appearance’) are stiffer than the background since these areas are brighter. The phase image for HDI-211-30 (Fig. 2b) at higher magnification shows similar rod-like structures to those observed in HDI-312-41. Similarly, the higher magnification AFM images for HDI-321-23 and HDI-431-20 (Figs 2c and 2d) both showed rod-like features. The high magnification images for MDI-321-30 (Fig. 2e) highlights both the bright and the background within Fig. 1e. The phase data show that the bright region has a lace-like structure associated with more rigid components. The area outside of the bright region is not homogeneous and some lace-like structures can be visualized on a 500 nm scan size (Fig. 3). The high magnification of HMDI-321-30 (Fig. 2f) exhibited no large rigid aggregates or rod-like structures as observed for the other materials. There are however, some lighter shade dots evenly distributed in the images, having a size of approximately 7 nm in diameter. Biodegradation The radiolabel counts were normalized within the two groups of polymers that contained 14 C-HDI and 14 C-BD labels using the specific radioactivity of each polymer for the purpose of comparing data. Prior to degradation, the total available normalized radioactivity of a polymer associated with the 14 C-HDI sample vials was 3.6 ± 1 × 105 CPM, while that of the 14 C-BD labelled samples was 2.3 ± 0.2 × 105 CPM. If desired, the fraction of sample degraded for each material can be estimated by dividing the released radiolabel values by the above numbers. Figure 4a shows the biodegradation results for the 14 C-HDI-polymers incubated in buffer solution, plotted as the cumulative radiolabel release versus time. The radiolabel release caused by incubation with buffer, for all four HDI-polymers,
256
Y. W. Tang et al.
Figure 3. AFM phase data for the softer (i.e. darker) region for MDI-321-30 (Fig. 2e), at a magnification of 500 × 500 nm2 .
was lower than 200 CPM /ml over the 10 week incubation period. Despite the differences in chemistry and other characteristic properties of the materials [22, 23], the effect of buffer on the stability of the polymer was not statistically different between the polymers (p < 0.05). The results for radiolabel release from the same polymers following CE incubation (160 units/ ml) are shown in Fig. 4b. It was found that the cumulative radiolabel release from the polymers incubated with enzyme was increased over that of the buffer controls. HDI-431-20 released approximately five times more radiolabeled product as compared to HDI-312-41. Consistently, the polymer with the higher hard segment released the less radiolabel. Figure 5a shows the biodegradation results for polymers with similar stoichiometry but different diisocyanate chemistry (i.e. the 14 C-BD polymers) incubated in buffer solutions and plotted as cumulative radiolabel release versus time. The cumulative radiolabel release from HDI-321-23 and MDI-321-30 was lower than 100 CPM /ml at the end of the 10 week incubation period, however the release from HMDI-321-30 exceeded this value to produce a cumulative release of approximately 300 CPM /ml. This indicated that despite the XPS data which showed that this polymer might be relatively hydrophobic in chemical nature, water itself had an effect on the surface stability for the HMDI based polymer. The results of radiolabel release from the CE incubation (160 units/ ml) are shown in Fig. 5b. HDI-321-23 was the most sensitive polymer to degradation by the enzyme. Despite the effect of buffer alone, HMDI-321-30 exhibited a similar radiolabel release to that of MDI-321-30 over the first nine weeks. However, it should be noted that by the tenth week there was a burst of release from HMDI-321-30 (p < 0.05).
Influence of surface morphology and chemistry on the enzyme
257
(a)
(b) Figure 4. Cumulative radiolabel release (counts per minute (CPM)) for 14 C-HDI-polymers incubated at pH 7.0 and 37 ◦ C with (a) phosphate buffer; (b) CE (160 units/ ml).
258
Y. W. Tang et al.
(a)
(b) Figure 5. Cumulative radiolabel release (counts per minute (CPM)) for 14 C-BD-polymers incubated at pH 7.0 and 37 ◦ C with (a) phosphate buffer; (b) CE (160 units/ ml).
Influence of surface morphology and chemistry on the enzyme
259
DISCUSSION
Analysis of the surface morphology The information that can be provided by AFM is rich in morphological character, despite some issues regarding data processing that remain controversial in the literature [31]. Based on the discussion of current concepts in AFM, the data provided in this study indicate that the surfaces of these polymers do not possess homogenous features. The images demonstrated that the HDI-polymer surfaces were made up of relatively stiff rod-like structures (Figs 2a to 2d), and that the MDI-321-30 had large aggregates on its surface (Fig. 1e), while HMDI-321-30 showed a relatively homogenous surface (Fig. 2f). Among the HDI-polymers, the surfaces were also different. HDI-312-41 exhibited what seemed to be a densely crystallized layer covering a compacted sub-surface matrix of rod like structures (Fig. 1a). HDI-211-30 also had an upper layer on top of the fibre-laced surface, but it appeared to be a ‘transition state’ with some areas not exhibiting the layer and other areas showing similar features to HDI-312-41 (comparing Figs 1a and 1b). A previous report has attributed this layer to be made up of a highly dense hard segment region since it represents a relatively stiffer region of the material’s surface [18]. Being that the XPS results for HDI-312-41 showed less carbonate groups at the upper surface (i.e. 15◦ take-off angle) as compared to HDI-211-30, it is reasonable to suggest that the greater density of the layer at the HDI-312-41 surface was due in part to its higher hard segment content (Table 3). HDI-321-23 and HDI-431-20 are polymers with relatively softer surfaces and no evidence of a densely packed layer covering the rod-like features. Most of the rod-like structures found at the HDI-polymer surfaces were overlapping onto each other, and it was difficult to determine their exact length. Previous analytical work has suggested that these features are crystallized soft segments [18, 22, 23]. Changing the diisocyanate chemistry in the polyurethanes produced dramatic changes in the surface morphologies. Rather than the rod-like crystallized polycarbonate structures found at the surface of HDI-321-23, MDI-321-30 showed some aggregates that appeared to have an organized form at the surface. Upon close examination, these aggregates were seen to be formed from smaller circular structures rather than rods. However, the background did contain smaller rod-like features similar to those found in the HDI-polymers. It is reasonable to postulate that the stiff aggregates found at the MDI-321-30 surface were formed from hard segment domains. This also corresponds well with the XPS data, which showed more carbonate (OCOO) at the HDI-321-23 (equivalent stoichiometry to MDI-321-30) and HDI-211-30 (equivalent hard segment content to MDI-321-30) surfaces than at the surface of MDI-321-30 (Tables 2 and 3). However, it should be noted that the MDI-321-30 aggregates do not only contain hard segments since there was the appearance of soft segment rod-like structures within these phases. Figure 3 showed the existence of a relatively ‘softer’ (darker color) segment ‘background’ in the aggregates.
260
Y. W. Tang et al.
The AFM images of HMDI-321-30 (Fig. 2f) indicated that this polymer was not phase-separated and the data supported previous SAXS and DSC data on the same materials which indicated similar findings [23]. According to Van Bogard et al., HMDI polyurethanes have smaller, more numerous micro-domains than MDI polyurethanes having the same hard segment composition [32]. Hence, it would be reasonable to suggest that the small dots observed in Fig. 2f may be small hard segment dense domains distributed evenly in the polymer matrix. Relationship between surface chemistry/ molecular morphology and biodegradation While neither the AFM or XPS data were provided for the actual aqueous surface exposed to enzyme, they provide us with information on the initial structure of the material. From this, it is possible to formulate a rationale for the progress of degradation that was observed in the aqueous enzyme solution. Polycarbonates themselves are hydrolyzable polymers and have been used in degradable drug delivery systems as carrier materials [33]. Suyama et al. studied the enzymatic degradation of pure poly(tetramethylene carbonate) and found that the materials could be degraded by CE [34]. It was therefore reasonable to suspect that, during the degradation process, the carbonate linkages of the polyurethanes would be the weakest link in the polymer backbone. Carbonates (OCOO) are known to be hydrolytically more stable than esters (COO) due to the two oxygens adjacent to the carbon, which form a pseudo π electron system and stabilize the C O bond. This bond, however, is less stable than the urethane linkage because of the strong electro-negative effect of the oxygen [35]. Hence, it could be proposed that if more carbonate groups than urethane groups were found at the polymer surface, this would indicate a greater potential for hydrolysis to occur in the polymer. Figure 6 shows a plot of the radiolabel released from each HDI-polymer versus the amount of carbonate groups (OCOO) found at the outer-most polymer surface (Table 3, 15◦ take-off angle) for two different incubation periods, i.e. two and ten weeks. It was found that, for both time periods, there was an overall increase in the radiolabel release for the polymers as the amount of carbonate groups at the outermost surface increased. However, the fact that there was not a linear relationship over the whole range of carbonate content indicates the existence of factors other than carbonate content which are dictating the process of degradation. Previous characterization of these materials using FTIR showed that the degree of hydrogen bonding between hard segments and the carbonate groups in the HDI-polymers was significantly different from each other [22]. A hydrogen-bonded carbonate would be anticipated to be more stable than a non-hydrogen bonded carbonate in the hydrolysis reactions because of the following two reasons. First the water molecules may not readily access the hydrogen bonded carbonate groups due to the hindrance effect of organized chemical structures included in the hydrogen bonded network. The second reason is that the C O linkages in a hydrogen-bonded OCOO may be
Influence of surface morphology and chemistry on the enzyme
261
Figure 6. Relationship between biodegradation (counts per minute (CPM) from Fig. 4b) of the HDIpolymers and OCOO content at a 15 take-off angle (Table 3) for two and ten weeks incubation.
more stable because there would be more electrons available for C O to withdraw and distribute over the bond to render the linkage stronger. By comparing the ratio of carbonate (OCOO) over urethane (NHCOO) groups on the polymer surface (see XPS measurements in Tables 3), an estimated measure of the potential for interacting groups can be obtained. These data are reported in Table 5. This represents the potential number of carbonates (OCOO) that could be hydrogen bonded to urethane groups (NHCOO) at the surface. The results showed that, at the surface of HDI-polymers, at both the outer-most surface (15◦ take-off angle) and sub-surface (90◦ take-off angle), the ratio decreased as the hard segment content decreased. This ratio also corresponds well with the rank of extent of biodegradation (Fig. 4b), where more degradation occurred in the polymer with a lower potential to form hydrogen bonded carbonate. What is perhaps more interesting is that for HDI-312-41 and HDI-211-30, there was a substantial increase in the ratio for the deeper section of the surface (i.e. 90◦ takeoff angle), while HDI-321-23 and HDI-431-20 showed a similar ratio between the upper surface and deeper layer (Table 5). The lower potential to form hydrogen bonded carbonates in the deeper zone of the surface could in part contribute to the rapid degradation of the HDI-321-23 and HDI-431-20 films, observed for these two polymers (Fig. 4b). A similar trend was also observed for the three polymers that had the 3 : 2 : 1 stoichiometry and were labelled with 14 C-BD. Within the upper
262
Y. W. Tang et al. Table 5. Ratio of urethane over carbonate groups at polycarbonate-urethane surfaces (based on the XPS data in Table 3) Polymer
HDI-312-41 HDI-211-30 HDI-321-23 HDI-431-20 MDI-321-30 HMDI-321-30
NHCOO/OCOO 15◦ take-off angle
90◦ take-off angle
0.43 0.40 0.34 0.28 0.41 0.46
0.65 0.60 0.39 0.26 0.31 0.61
surface (i.e. 15◦ take-off angle) HDI-321-23 showed less potential to form hydrogen bonded carbonates than did MDI-321-30 and HMDI-321-30. A point worth noting is that the data in Table 5 can be used to partially explain the unique biodegradation character for HMDI-321-30. At 160 units/ ml, the radiolabel release from HMDI-321-30 was similar to that of MDI-321-30, however, near ten weeks, the release did increase above that of MDI-321-30 (Fig. 5b). In this system, there was a long delay before significant product release. One possible explanation for this is that the degradation was initiated at the more stable urethane groups and then later gained access to the carbonates. The XPS results showed that of the three 14 C-BD labelled polymers this material possessed the highest surface ratio of urethane to carbonate groups (Table 5), both at the 15◦ and 90◦ take-off angles. This indicates that HMDI-321-30 has good potential for forming hydrogen bonds among the carbonate groups at the outermost surface. Furthermore, the AFM data (Fig. 2f) indicate that HMDI-321-30 is quite phase mixed which further substantiates the potential for the hard segment to be shielding the carbonate groups. The masking of the carbonates by the urethane groups implies that the degradation will be delayed because urethane bonds are more difficult to degrade than carbonates [35]. The initial increase in degradation which occurred in the 10th week of incubation (Fig. 5b) relative to MDI-321-30 indicates that the cohesive structure for HMDI-321-30 at the surface was not as effective at maintaining stability as that of MDI-321-30. It has also been noted elsewhere that this degradation is even more pronounced at higher enzyme concentrations [23]. Again, the AFM data indicate that the phase features involving hard segments of HMDI-321-30 may have been more numerous, but were very small in comparison to MDI-321-30. In such a structure, access to the soft segment would be anticipated to be easier than with MDI-321-30 and therefore broken down more easily. After a lengthy incubation period, the urethane bonds were hydrolyzed and the degradation rate was accelerated with the breakdown of carbonates since the physical barrier to these latter groups, within the polymer’s upper surface matrix, was no longer present. The ease for which water was able to penetrate into HMDI-321-30 was
Influence of surface morphology and chemistry on the enzyme
263
further highlighted in the buffer control experiment (Fig. 5a), where this polymer was more sensitive to the early extraction of polymer derived material. The AFM surface images of all the HDI-based polymers also supports the role of surface morphology in the biodegradation of the polymers by CE. As discussed earlier, the HDI-polymers exhibited crystallized polycarbonate forms, similar to rod-like structures at the surface. However, the AFM images themselves depicted a background phase, which did not consist of the rod-like features and was less stiffer than the latter. If this background phase was an amorphous soft segment phase then the enzymes could readily initiate the degradation of this component. It has been documented that the presence of crystallized domains near an amorphous phase may even promote the degradation of these latter domains. Abe et al. investigated the enzymatic degradation of poly(R-3-hydroxybutyrate) (PHB) stereoisomers and found that the enzyme preferred to bind to crystallized surfaces and then start the cleaving in the amorphous domains, near the crystals [36]. Focarete et al. mixed a crystallizable polyester into PHB and found that it promoted the enzyme catalyzed hydrolysis of PHB, since the crystallized polyester provided more stable binding sites for the enzyme [37]. Iwata studied the enzyme degradation of poly(L-lactic acid) single crystals and also found that the degradation would not initiate at the center of the crystal, but rather, the degradation started near the edge of the crystal where non-ordered polymer chains were located [38]. The AFM images (Figs 1 and 2) for both HDI-312-41 and HDI-211-30 exhibited a hard segment layer on their upper surfaces. As discussed earlier, this layer is believed to be related to a dense hard segment phase, which has strong potential for H-bonding with the carbonate groups at the top surface (Table 5). However, the observed hard segment layer did not cover the whole polymer surface and hence some polycarbonate segments were exposed. These exposed regions were the likely initiation sites for degradation. Since the layer of the formed structure on the HDI-312-41 covered a larger area than that of HDI-211-30 (Figs 1a and 1b), it can be understood that HDI-312-41 would be more resistant to degradation than HDI-211-30 (Fig. 4b). HDI-321-23 and HDI-431-20 have similar surface morphologies to each other and don’t have the barrier layer covering the soft segment rods. Based on the current analysis, it would be anticipated that they would have a greater level of degradation.
SUMMARY
In summary, six polycarbonate based polyurethanes were characterized at their upper surface using two different methods, XPS and AFM. In turn, the biodegradation results indicated a strong dependence on the surface morphology of the materials. It was noted that soft segment crystallization alone did not provide an effective barrier to biodegradation. Rather, the molecular structure and distribution of the hard segment appeared to be a principal contributor to enhanced surface stabilization. The data represent a novel image of the parameters that govern the biodegradation of
264
Y. W. Tang et al.
polyurethanes in hydrolyic processes that may contribute to the in vivo response to implanted medical devices. Acknowledgements Funding was provided from the Canadian Institute of Health Research and Materials and Manufacturing Ontario.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
21. 22. 23. 24. 25. 26. 27. 28. 29. 30.
K. Stokes and K. Cobian, Biomaterials 3, 225 (1982). M. Szycher, J. Biomat. Appli. 3, 297 (1988). R. S. Labow, D. G. Duguay and J. P. Santerre, J. Biomater. Sci. Polymer. Edn 6, 169 (1994). K. Stokes, J. Biomat. Appli. 3, 228 (1988). B. D. Ratner, K. W. Gladhill and T. A. Horbett, J. Biomed. Mater. Res. 22, 509 (1988). R. S. Labow, D. J. Erfle and J. P. Santerre, Biomaterials 16, 51 (1995). L. Pinchuk, J. Biomat. Sci. Polymer Edn 6, 225 (1994). M. C. Tanzi, D. Mantovani, P. Petrini, R. Guidoin and G. Laroche, J. Biomed. Mater. Res. 36, 550 (1997). B. D. Ratner, S. C. Yoon and N. B. Nates, in: Polymer Surface and Interfaces, W. J. Feast and H. S. Munro (Eds), pp. 231– 251. John Wiley & Sons, Toronto (1987). B. D. Ratner, Cardiovasc. Pathol. 2, 87s (1993). G. L. Grobe, P. L. Valint and D. M. Ammon, J. Biomed. Mater. Res. 32, 45 (1996). C. Bouranel, T. Fries, D. Drescher and R. Plietsch, Euro. J. Orthodon. 20, 79 (1998). S. P. Gorman, M. M. Tunney, P. F. Keane, K. Vanbladel and B. J. Bley, J. Biomed. Mater. Res. 39, 642 (1998). C. Larsson, P. Thomsen, B. O. Aronsson, M. Rodahl, J. Lausmaa, B. Kasomo and L. E. Ericson, Biomaterials 17, 605 (1996). D. K. Han, K. D. Park, J. A. Hubbell and Y. H. Kim, J. Biomater. Sci. Polymer Edn 9, 667 (1998). S. N. Magonov, V. Elings and M. H. Whango, Surf. Sci. 375, L385 (1997). S. N. Magonov and D. H. Reneker, Ann. Rev. Mater. Sci. 27, 175 (1997). I. Revenko, Y. W. Tang and J. P. Santerre, Surf. Sci. 491, 346 (2001). N. M. Lamba, K. A. Woodhouse and S. L. Cooper, Polyurethanes in Biomedical Applications. CRC Press LLC, NY (1998). X. Chen, S. L. McGurk, M. C. Davies, C. J. Roberts, K. M. Shakosheff, S. J. B. Tendler and P. M. Williams, Macromolecules 31, 2278 (1998); K. G. Tingey and J. L. Andrade, Langmuir 7, 2471 (1991). K. G. Tingey and J. D. Andrade, in: Protein at Interfaces, T. Horbett and J. L. Brash (Eds), p. 66. Ser. 602 Am. Chem. Soc. Symp. (1995). Y. W. Tang, R. S. Labow and J. P. Santerre, J. Biomed. Mater. Res. 56, 516 (2001). Y. W. Tang, R. S. Labow and J. P. Santerre, J. Biomed. Mater. Res. 57, 597 (2001). J. P. Santerre, R. S. Labow, D. G. Duguay, D. J. Erfle and G. A. Adams, J. Biomed. Mater. Res. 28, 1187 (1994). J. P. Santerre, R. S. Labow and G. A. Adams, J. Biomed. Mater. Res. 27, 97 (1993). R. S. Labow, E. Meek and J. P. Santerre, J. Biomed. Mater. Res. 39, 469 (1998). M. J. Hearn, B. D. Ratner and D. Briggs, Macromolecules 21, 2950 (1988). F. M. Serry, I. Revenko and M. J. Allen, Biomaterials Forum 20, 11 (1998). S. N. Magonov, V. Elings and M. H. Whangbo, Surface Science 375, L385 (1997). G. Bar, Y. Thomann, R. Brandsch, H. J. Cantow and M. H. Whangbo, Langmuir 13, 3807 (1997).
Influence of surface morphology and chemistry on the enzyme
265
31. R. S. McLean and B. B. Sauer, Macromolecules 30, 8314 (1997). 32. J. W. C. Van Bogart, A. Lilaonitkul, L. E. Lerner and S. L. Cooper, J. Macromol. Sci. Phys. B17, 267 (1980). 33. A. Tabak, E. Hoffer and E. Taitelman, J. Pharmacy and Pharmacology 46, 789 (1994). 34. T. Suyama and Y. Tokiwa, Enzy. and Microbial Technol. 20, 122 (1997). 35. T. M. Chapman, J. Polym. Sci., Part A: Polym. Chem. 27, 1993 (1989). 36. H. Abe, I. Matsubara and Y. Doi, Macromolecules 28, 844 (1995). 37. M. L. Focarete, G. Ceccorulli and M. Scandol, Macromolecules 31, 8485 (1998). 38. T. Iwata and Y. Doi, Macromolecules 31, 2461 (1998).
This page intentionally left blank
Perspective Cardiopulmonary bypass technology transfer: musings of a cardiac surgeon F. D. RUBENS ∗ Room H211, Ottawa Heart Institute, 40 Ruskin St., Ottawa, Ontario K1Y 4W7, Canada Received 11 November 2001; accepted 29 January 2002 Abstract—The development of cardiopulmonary bypass (CPB) has been one of the greatest technical advancements in cardiovascular medicine. With heparin anticoagulation, this device can safely replace the circulatory and gas-exchanging functions of the heart and lung, facilitating complex cardiac operations. Limitations still exist however, related to blood reactions at the biomaterial surface, such as cell activation, inflammation and low-grade thrombosis. In this brief review, the thought processes which paralleled the development of CPB biocompatible surfaces such as heparin-coating, will be explored, as well as current theories on the suspected mechanisms by which heparin-coated surfaces act as an anti-inflammatory device during CPB. Results with new surfaces for CPB designed to capitalize on superior protein adsorption properties, such as surface modifying additive (SMA) and poly (2-methoxyethylacrylate) (PMEA), will also be described. Finally, the significance of biomaterial-independent blood activation will be discussed, emphasizing the current need to develop strategies utilizing optimal biomaterials, modified surgical technique and pharmacologic therapy to minimize the systemic complications of CPB. Key words: Biocompatibility; heparin-coated circuits; SMA; PMEA; biomembrane mimicry; thrombosis; inflammation; cardiopulmonary bypass; review.
Since its clinical introduction in 1953 [1], cardiopulmonary bypass (CPB) has shaped surgical strategies for the correction of congenital and acquired cardiac disease for millions of patients worldwide [2]. Cardiac surgeons have been the interface between new CPB technology and the patient. With each new development, it was essential for surgeons to provide information about CPB technology to impact practice choices. As this science progresses, cardiac surgeonscientists and basic scientists (engineer/ biologist) must continue to work together to facilitate this technology transfer. ∗ E-mail:
[email protected]
268
F. D. Rubens
THE BYPASS PARADIGM. IDENTIFYING THE QUESTION
From a historical perspective, both surgeon-scientists and basic scientists played key roles in the conceptualization and realization of the science of CPB. Surgeonscientists, such as Gibbon and De Wall [2– 4] primarily used their knowledge of applied physiology to address the significant clinical demand for bypass, with little attention being given to the molecular and cellular events at the bloodbiomaterial interface. Despite this, the technology and its clinical application still advanced. Although we now have an aging, sicker population, the parallel evolution of technology and surgical skill has resulted in declining rates of morbidity and mortality in those undergoing cardiac surgery. I propose that by the elucidation of the perceived thought processes that eventually lead to the current clinical success of CPB, we may be able to identify how we can proceed to newer technologies that will further enhance biocompatibility in CPB. At least two questions arise when we consider applying the principles of biomaterial science to CPB development. First, can we synthesize a biomaterial interface that allows us to reliably predict improved biocompatibility? By definition, this would imply optimal functioning of CPB in the presence of an appropriate host response [5]. Second, if we do succeed in developing a truly biocompatible surface, how much will this improvement actually contribute to increased patient safety during CPB?
OVERVIEW OF EVENTS RELATED TO BLOOD-BIOMATERIAL SURFACE INTERACTION
When blood contacts the synthetic surfaces of the CPB circuit, plasma proteins quickly become adsorbed to the biomaterial surface. Further exposure of blood with the surface results in protein activation (contact activation) and blood cell activation. Several of the key steps are illustrated in Fig. 1. This diagram is not inclusive but reflects mechanisms and interactions pertinent to the current discussion. The contact activation system can be thought of as a primitive host defense mechanism able to isolate and destroy a foreign substance or surface that the blood ‘sees’. Contact activation involves 4 primary plasma proteins: factors XII and XI, prekallikrein and high-molecular-weight kininogen (HMWK). In the presence of the negatively charged surface of the biomaterial, a conformational change occurs in factor XII. This permits the activation of factor XII in the presence of prekallikrein and HMWK. Factor XIIa activates factor XI and initiates the intrinsic coagulation pathway to generate thrombin and subsequently fibrin. Factor XIIa also activates prekallikrein to form kallikrein. The generation of kallikrein is extremely rapid, and it can be detected almost instantaneously at the start of bypass [6, 7]. Two other pathways directly activated by kallikrein include the kinin system and the fibrinolytic system. Kallikrein catalyzes the conversion of HMWK to bradykinin [8]. Bradykinin has a very
Biocompatibility and cardiopulmonary bypass
269
Figure 1. Schematic illustrating some of the key reactions related to blood activation at the biomaterial surface (CPB surface), the wound surface and the vasculature (blood vessel) during CPB. Thrombin and fibrin generation (1) primarily result from activation in the pericardial cavity (2) and contact activation through factor XII (3). Contact activation also initiates the inflammatory pathway through the conversion of prekallikrein to kallikrein (4). Products released from the inflammatory pathway contribute to the initiation of fibrinolysis (5). Finally, the CPB surface is one of the key initiators of complement generation (6). CD — cluster of differentiation antigen, TF — tissue factor, F1.2 — Prothrombin fragment, TAT — thrombin anti-thrombin III (ATIII) complex, C — complement factor (C1q, C3, C3a, C3b, C5a, C5b-9), HMWK — high molecular weight kininogen, tPA — tissue plasminogen activator, VIIa — activated factor VII, XIIa — activated factor XII, XIa — activated factor XI, IgG — immunoglobulin G.
short half-life in the plasma. This may be because of the rapid metabolism by angiotensin-converting enzyme in the pulmonary circulation [9, 10] and by the vascular endothelium [11]. Bradykinin is believed to be a mediator of increased capillary permeability and the development of tissue edema. Bradykinin mediates vasodilation by stimulating the release of endothelial nitric oxide [11]. Kallikreinrelated activation of the fibrinolytic system includes its role in catalyzing the conversion of plasminogen to plasmin and the activation of pro-urokinase. Further, bradykinin is a major stimulus for endothelial tissue plasminogen activator (tPA) release [12, 13]. The complement system consists of some 20 plasma proteins and forms part of the body’s defence mechanism. It is activated through several mechanisms during CPB of which two of the mechanisms are directly related to contact activation and the biomaterial surface [14, 15]. First, the third component of complement (C3) binds to the CPB circuit surface and releases C3a (a potent chemoattractant).
270
F. D. Rubens
The fragment of C3 remaining on the surface is joined by the inactive factor B and properdin to produce the active proteolytic enzyme C3 convertase. The C3 convertase will cleave C3, resulting in the formation of C5 convertase from C3b. This latter enzyme complex subsequently cleaves the fifth component C5, to generate the active fragment C5a and to continue the recruitment of the terminal complement complex (TCC) C5b-9. This C5b-9 sequence is able then to perform and generate the vasoactive, chemotactic, immunoregulatory, and cytolytic activities of complement. The second mechanism for complement generation is related to contact activation. Kallikrein is able to cleave C5 directly to produce C5a [16] and C3 may be cleaved by plasmin [17].
THE QUEST FOR THE PERFECT SURFACE FOR CPB
In the earliest CPB devices (bubble and sheet oxygenators), both in vitro and in vivo circuits were prepared with stainless steel and glass [18]. Polymers were first introduced for the pragmatic rationale of creating a disposable circuit so that cleaning and re-sterilizing was not necessary [19, 20] rather than for considerations of improved biocompatibility. Similarly, with the introduction of membrane oxygenators, the material characteristics were preferentially chosen, again not on the basis of biocompatibility, but on the physical characteristics that enhanced gas exchange within the lowest surface area, such as the Bramson lung [21] and the Landé-Edwards device [22]. These actions do not imply that there was no understanding at the time of the importance of surface biocompatibility. Information has existed for more than a century showing that different surfaces have different thrombogenicity [23, 24]. (Although we now recognize that thrombogenicity is merely one facet of biocompatibility related to the blood-biomaterial interface, in the pioneering days of CPB it was the only detectable marker of bioincompatibility.) Biocompatible biomaterials have subsequently been introduced into cardiovascular devices as a priority. In some cases, they were chosen after in vitro screening of materials having defined surface properties/ characteristics. However, even when tests suggested improvement in comparison to previously used surfaces, some of these materials could not be used for CPB for a variety of reasons. For example, surface energy was studied as a potential candidate for the identification of suitable surfaces. Baier et al. [25] proposed a hypothetical zone of biocompatibility related to surface energy, however, this has not proven to uniformly predict thrombogenicity. Hydrophilic surfaces, such as albumin-coated surfaces, were found to enhance biocompatibility by decreasing platelet adhesion when compared to hydrophobic surfaces [26]. The overall benefit of this property is not yet defined as modern oxygenators with microporous membranes depend on the property of hydrophobicity to prevent infiltration of plasma or plasma water into the membrane pore structure [18]. Attempts have also been made to try to choose surfaces based upon the application of the principles of the cellular and molecular processes of blood activation, thus
Biocompatibility and cardiopulmonary bypass
271
allowing one to try to predict a priori which surfaces would be biocompatible. This approach has been criticized in the past due to applications of theories based upon unproven hypotheses [27] however, in some instances, the approach has been successful and has resulted in clinical products that have demonstrated clinical efficacy and improved biocompatibility. This ‘approach’ was utilized in the development of heparin-coated circuits, albeit not initially related to theories of heparin’s role as an inhibitor of thrombosis. From reviews by Dr. Vincent L. Gott [28, 29], it appears that serendipity played a key role in the development of heparin-coated circuits. The initial goal by the investigators was to develop a technique to create a negative charge on an implanted biomaterial surface, as it was believed that this was the property most responsible for the nonthrombogenic behavior of the intact endothelium. A relatively severe screening test was utilized in which ring-shaped test materials coated in conductive carbon paint were inserted into the canine inferior vena cava and connected to an electric current. Prior to insertion, the rings were sterilized in a common hospital solution (benzalkonium chloride) and then rinsed in heparin. It was soon recognized that even in the absence of the electric current, this processing significantly decreased the formation of thrombus. Further evaluation demonstrated that the graphite surface was firmly bonding the cationic surfactant benzalkonium chloride, which in turn bonded the negatively charged heparin. The non-thrombogenic behavior of this surface was confirmed with a series of in vitro tests of clot formation in the presence of heparin-graphite surfaces compared to silicone and polycarbonate [29] and heparin-coated surfaces were born. These findings set into motion a separate industrial directive to develop novel means by which heparin could be coated onto biomaterials, including those utilized for CPB. At present, there are three primary ways in which heparin may be bound to polymer surfaces. Heparin-releasing surfaces In the first group, heparin is bound such that it may be slowly released into the circulation directly from the surface. The original ionic binding pioneered by Gott has its current form in the DurofloII™ surface (Baxter Healthcare Corporation, Irvine CA) in which the heparin is ionically bound to benzalkonium that is attached to the substrate polymer. In 1968, Hufnagel developed a method to disperse heparin in silicone rubber, and this mixture was coated on a substrate surface. The heparin could then be demonstrated to slowly leech from the surface [30]. The third heparinreleasing surface is the thermosensitive heparin-containing hydrogel that contracts at body temperature, releasing heparin. Heparin-immobilized surfaces In the second group, heparin is immobilized permanently upon the biomaterial surface. Larm et al. [31] first described the preparation of covalently bound
272
F. D. Rubens
surface heparin. Polyethylene oxide (PEO) was utilized as a spacer group as its hydrophilicity and its dynamic motion would further inhibit platelet interactions with this surface. In vitro proof of the inhibition of coagulation of covalently bound heparin has been provided [32]. The Carmeda product (Carmeda™, Medtronics, Minneapolis, MN) is a commercial example of this technology for CPB in which heparin is covalently bound via an end-point immobilization technique. The Trillium™ Biopassive Surface (Medtronics, Minneapolis, MN) is a similar coating process involving two polymers on the substrate surface. The first polymer functions as a primer, strongly bonding to the surface of the substrate. The second polymer, which is bound to the primer coat, is composed of sulfate/ sulfonate groups, the latter providing a negative surface charge, as well as polyethylene oxide chains covalently bound to heparin [33]. The Corline™ system utilizes multiple heparin molecules covalently bound by specific linkers to an inert polyamine chain, thus preserving the key active pentasaccharide sequence of heparin. There is little information on the use of this system for CPB, but the technology has been applied for heparin-coated stents as the Jostent™ (Jomed, Netherlands). BioLine™ (Jostra, Germany) is a hybrid surface (i.e. combination heparinreleasing and heparin-immobilized) in which heparin is adsorbed onto a layer of immobilized polypeptides through a combination of ionic interactions and covalent bonds. Heparin-grafted polymers In the third group, substrate biomaterial surfaces are coated with copolymers containing heparin. This preparation is really a triblock polymer with a hydrophobic substrate as a base, and then a hydrophilic moiety (e.g. PEO) subsequently bound to heparin. There are no commercial products for CPB in this group. CLINICAL RESULTS WITH HEPARIN-COATED CIRCUITS
Large doses of intravenous heparin (300 units/ kg) are usually required during CPB. The convincing in vitro evidence of the thromboresistant behavior of heparin-coated surfaces led to the hypothesis that a less intense regime of anti-coagulation would be required for CPB when heparin-coated circuits were utilized. Many of the early clinical trials compared outcomes after CPB in patients where heparin-coated circuits were combined with lower dosages of heparin (one-third to one-half) versus control (un-coated) circuits and standard heparin doses (300 units/ kg). Despite the methodologic flaws of this approach, there have been consistent findings supportive of clinical advantages associated with the combination of heparin-coated circuits and low intensity heparinization (100 units/ kg). These advantages have included decreased perioperative bleeding and decreased transfusion requirements [34– 36]. In a large clinical trial reported by Aldea et al. [37], an integrated blood conservation strategy was utilized including heparin-coated circuits, decreased heparin dosing, maximal blood salvage from the operative field, low pump prime and dilution,
Biocompatibility and cardiopulmonary bypass
273
closed venous reservoirs, normothermia, precise heparin and protamine titration and routine use of anti-fibrinolytics. Compared to patients treated with standard circuits and heparin, this approach resulted in decreased blood transfusion requirements and significantly improved clinical outcomes as measured by the length of stay in both the intensive care unit and hospital, as well as the duration of required ventilatory support. This was translated into considerable potential cost-savings of approximately $1700 (US) per patient. A recently reported meta-analysis of trials comparing heparin-coated circuits and low heparin versus standard circuits with standard heparin, showed an estimated cost benefit of $3231 (US) per patient with covalently bonded circuits and $1068 (US) per patient with ionically bonded circuits [38]. Compelling as these clinical results are, there are several factors specifically related to the transfer of this technology to the commercial spectrum that must not be overlooked. First, neither thrombin generation nor fibrinolysis during CPB, have been consistently demonstrated to be effectively decreased with heparin-coated circuits in human trials, either with standard or decreased heparin doses as compared to standard (non-heparin-coated) circuits [39– 41]. Therefore, if there is no evidence that the process of thrombosis is inhibited, it is difficult to justify decreasing the heparin dose due to the potential risk of low-grade thrombosis [42]. Second, comparing the combined use of heparin-coated circuits and low heparin with standard circuits with standard heparin is somewhat akin to comparing apples and oranges. Any measured differences in patient clinical and biochemical outcomes may solely be related to differences in the amounts of administered heparin and protamine and not to the surface. Heparin itself can induce a myriad of cellular and biochemical changes such as fibrinolysis [43] and platelet dysfunction [43, 44]. As the dose of the heparin-neutralizing drug protamine must also be increased in the control group in these trials, it is anticipated that changes such as complement activation must be proportionally increased. In fact, in the majority of clinical trials in which the same doses of heparin are used with both heparin-coated circuits and standard circuits, little if any clinical benefit has been consistently demonstrated [45]. A European multi-center randomized controlled trial of heparincoated circuit (DurofloII™) versus standard circuits with identical heparin in low risk patients undergoing primary coronary surgery, failed to show any significant clinical benefit [46]. There may be some benefit in high-risk patients however, as an identically-designed European multi-center randomized controlled trial with these patients, did demonstrate improvement in intensive care unit and hospital length of stay as well as a drop in the incidence of postoperative lung dysfunction, in the DurofloII™ group [47]. This is not to say that heparin-coated circuits do not represent an improvement in biocompatibility. However, it would appear that the beneficial clinical impact of this surface is almost entirely related to its intrinsic anti-inflammatory effect. What has been consistently demonstrated is the capacity of heparin-coated surfaces to decrease complement activation [40] with consistent evidence of decreased generation of terminal complement complex (C5b-9) [48– 50]. (C5b-9 has been demonstrated
274
F. D. Rubens
to have a pro-coagulant effect, likely related to platelet vesiculation and exposure of prothrombinase complex sites [51, 52], however the clinical relevance of this during CPB has not been established). This reduction in complement activity does not appear to be related to the decreased heparin dose used in some studies in the heparincoated circuit group [53]. Other demonstrated anti-inflammatory effects of heparincoated surfaces include decreased leukocyte surface activation markers/ receptors [54] decreased cytokine production [55] and decreased monocyte tissue factor [56]. Therefore, how does one explain how this surface effectively inhibits the inflammatory pathway without parallel decrease in coagulation? Wendel et al. [57] incubated heparinized (1 u/ ml) whole human blood in closed-loop tubing with and without covalently bound heparin. The adsorbed proteins were then eluted and analyzed, and this demonstrated a marked increase of adsorbed high molecular weight kininogen (HMWK) on the surface of the heparin-coated tube. This suggests that heparin-coated surfaces are acting as a selective protein sink adsorbing HMWK. The primary inhibitor of contact activation is C1-esterase inhibitor, however its action is not facilitated by heparin [58] whereas surface adsorbed HMWK may interact with anti-thrombin III in the presence of surface heparin to potentiate heparin-mediated inhibition of kallikrein. [59]. Decreased kallikrein activity with heparin-coated circuits has been further supported by the demonstration of decreased generation of C1-inhibitor-kallikrein complexes in the presence of heparin-coated circuits [39] (Generation of this complex reflects increased kallikrein generation). An alternative hypothesis for the mechanism of inhibition of complement generation is related to augmented inactivation of the complement system by inhibitory factors. Kazatchkine et al. demonstrated that heparin coupled to zymosan or Sepharose inhibits complement by augmenting C3b inactivation through factors H and I [60]. Mollnes et al. also showed that solid phase heparin in vitro has specific complement inhibitory activity not due to adsorption to surface of activation products [61]. The effect of inhibition of complement formation has not been tested in clinical trials. In summary, heparin-coated surfaces have been demonstrated to exhibit potent anti-thrombotic behavior in in vitro testing. When utilized clinically with a diminished intensity of anti-coagulation, although there is no evidence of decreased thrombogenicity, there is consistent evidence that inflammation related to complement activation is decreased and this may be responsible for the improvement seen in some clinical outcomes after CPB. ADDRESSING THE DISCORDANCE BETWEEN THEORY AND PRACTICE
The discordance between the predicted theory (thrombin inhibition related to surface heparin) and the actual molecular outcome at the surface in vivo (complement inhibition) illustrates that it may be naive to think we can develop a surface that would primarily inhibit contact activation in the presence of the complex milieu of human blood. It probably further emphasizes that it is the composition of the ad-
Biocompatibility and cardiopulmonary bypass
275
sorbed protein layer, a ‘signature’ composition for that particular biomaterial that determines its biocompatibility. Surfaces have been introduced, whose biocompatibility has been predicted based upon their protein adsorption characteristics and there are some promising early clinical results. An example of this approach includes a new generation of biomaterials developed for CPB into which a surface-modifying additive (SMA) has been incorporated into the polymer used to prepare the CPB circuit [62, 63]. The additive is a triblock-copolymer with polar and nonpolar polymer chains of the general formula polycaprolactone-polydimethylsiloxane-polycaprolactone [63, 64]. During the manufacturing process, the SMA migrates to the surface of the base polymer yielding a stable microdomain-like configuration. Although the mechanism of action by which SMA decreases blood protein and cellular activation is not precisely known, it is hypothesized that the alternating hydrophobic and hydrophilic regions of the microdomain surface lead to uniform adhesion of fibrinogen, such that all of the sites for potential platelet interaction with surfacebound fibrinogen are occupied [65, 66]. We recently completed a randomized controlled trial in which patients underwent CPB using a standard control circuit, or a circuit prepared ‘tip-to-tip’ with the SMA copolymer (SMA-CPB) [12]. Striking changes were found in the effect of this surface on markers of coagulation, fibrinolysis, platelet function and number. In the control group, the platelet count fell significantly throughout the period of CPB as compared to the SMA-CPB group (p < 0.05). There was also evidence for decreased platelet activation in the SMA-CPB group as demonstrated by decreased expression of the platelet surface marker P-selectin (ANOVA p = 0.0132) and decreased release of β-thromboglobulin (BTG) (ANOVA p = 0.017). Thrombin generation, as measured by thrombin anti-thrombin III (TAT) (ANOVA p = 0.011) was significantly decreased and there was a trend to a decrease in the prothrombin fragment F1.2 (ANOVA p = 0.158) with SMA-CPB compared to controls, despite carefully matched heparin doses in the two groups. This trial was not intentionally designed to detect differences in bleeding and the need for blood transfusion; however, Defraigne et al. [67] in a larger clinical trial confirmed the same degree of platelet preservation with SMA-CPB and a similar decreased release of BTG. In addition, in Defraigne’s study, there was evidence of a decreased need for the transfusion of platelets and fresh frozen plasma in the group with SMA-CPB circuits. Terumo Corporation (Tokyo, Japan) has recently developed a surface for CPB circuits with a similar biocompatibility profile to SMA in that it was engineered to positively influence protein adsorption. Surfaces are coated with poly (2-methoxyethylacrylate) (PMEA) which has a hydrophobic polyethylene backbone, and its residue has mild hydrophilicity with no chemical functional groups such as OH or NH2 . It was predicted that as the outer side of the PMEA molecule is inactive chemically, the surface would have little tendency to react with blood components [68].
276
F. D. Rubens
Although clinical studies in humans are not available, there are promising data from in vitro and in vivo animal models supporting the potential improved biocompatibility of PMEA-coated surfaces. In a porcine model, there is a significant decrease in the total amount of adsorbed protein on the PMEA-coated oxygenator (Capiox SX18, Terumo, Tokyo, Japan) as compared to the uncoated oxygenator (uncoated Capiox SX18) [68]. Analysis of the composition of the adsorbed protein layer demonstrated a striking decrease in the amount of adsorbed IgG: 480 ± 70 ng/ cm2 for the uncoated circuits and 7 ± 6 ng/ cm2 for the PMEA-coated circuits. This was correlated to platelet count preservation controlled to the control group (p < 0.05) likely related to the fact that IgG-coated surfaces have long been known to be potent platelet activators [69]. Another marker measured during these experiments was the monocyte receptor for complement component C3b, CD35 [70]. An increase in the expression of CD35 on monocytes indicates the activation of C3, namely, the production of C3b and anaphylatoxin C3a. The use of the PMEA circuit was associated with a significant decrease in CD35-positive monocytes during CPB (p < 0.05 at 180 and 240 min after the start of bypass). As surface-adsorbed IgG is a potent initiator of the classical complement pathway (formation of C1q-IgG complexes) [71] decreased adsorption of immunoglobulin is probably the mechanism of the resulting decreased stimulation of the complement system. Plasma bradykinin levels are also decreased with PMEA [68, 72] as are thrombin anti-thrombin III levels [68]. No convincing arguments have been presented to support the mechanism for these findings. A limitation of these studies relates to the fact that as porcine platelet adhesion also appears to be suppressed by the PMEA surface, the authors have not taken into account the differences in protein composition on the surface that may solely be due to platelet-associated proteins (such as IgG) — not adsorbed proteins. Another novel approach that has been introduced for clinical CPB is the preparation of biomaterials that mimic the non-thrombogenic nature of erythrocytes (biomembrane mimicry). The chosen substrate biomaterial is coated with a derivative of phosphorylcholine, which is the major lipid head group component found on the outer surface of biologic cell membranes [73]. In vitro data has demonstrated the efficacy of this surface in inhibiting fibrinogen adsorption and platelet deposition [73]. Clinical experience in CPB with this device (Memsys™, Sorin Biomedica) is limited, but in a recent trial of pediatric cardiac surgical cases by De Somer et al. [74], there was suggestive evidence that the test surface limited platelet activation as shown by decreased thromboxane B2 and β-thromboglobulin release, as compared to patients treated with control circuits. The evidence supporting minimized complement activation (C5b-9) was less convincing as there was a difference between the groups prior to CPB. Finally, there was no difference in thrombin generation in the two groups. Therefore in summary, SMA and PMEA represent biomaterial surfaces for CPB, in which the surfaces (alternating hydrophobic and hydrophilic regions) were
Biocompatibility and cardiopulmonary bypass
277
selected for their effect on protein adsorption. There has subsequently been a concordance between the theory and the in vivo results with a correct predictions of decreased thrombogenicity of the surface. Biomembrane mimicry is another promising approach in which the beneficial effects such as platelet preservation, may be totally related to altered fibrinogen adsorption on the surface.
HOW MUCH WILL SURFACES DESIGNED FOR IMPROVED BIOCOMPATIBILITY CONTRIBUTE TO INCREASED PATIENT SAFETY DURING CPB?
Our knowledge has been shaped by the universal acceptance to date of the primary role of the contact activation system as the instigator of blood activation during CPB. Virtually simultaneous to the clinical introduction of CPB was the discovery of the role of factor XII (Hageman factor) and the intrinsic coagulation pathway in the generation of thrombin when blood contacts a non-endothelial surface [75]. It now appears that biomaterial-independent blood activation may also play a more important role in the systemic changes we equate with CPB. For example, thrombin generation during CPB does not appear to be temporally related to measured peaks of activation of the intrinsic coagulation pathway (activated factor XII) [76] suggesting alternative pathways must be playing a role in thrombin generation. Exposure of blood to wound surfaces may be an important source of thrombin generation during CPB. Blood collected in the wound around the heart during CPB, is returned via a sucker to the CPB apparatus and this blood is referred to as ‘cardiotomy blood’ [12]. Tissue factor on the wound surface [77] and on the surface of activated monocytes contributes to coagulation [56, 78]. In addition, monocytes adherent to oxygenator fibres after CPB can be shown to have increased CD 11b expression by fluorescent image analysis [79]; this surface glycoprotein can directly activate factor X [79] and thus augment thrombin generation [79]. Other factors contribute further to thrombin generation in cardiotomy blood. Heparin has a limited efficacy as an anticoagulant in the wound as the heparin levels are well below those found in the systemic blood [80]. Heparin can bind to nonplasma components such as platelets or debris [81] and it can also be inactivated by platelet factor 4 (a heparin-neutralizing protein released from platelets) [82]. Further, more heparin is not capable of inhibiting thrombin completely; particularly thrombin that is bound to fibrin where it becomes inaccessible to the heparinantithrombin III complex [83]. Once thrombin is generated, other hematologic changes are inevitable. Thrombin contributes to fibrinolysis through the direct activation of endothelial cells and the release of tPA [13]. Total tPA antigen is increased in patients receiving cardiotomy blood [84] as are the concentrations of thrombin-antithrombin III, fibrin monomers and fibrin degradation peptides [84, 85]. tPA may contribute directly to fibrinolysis in blood that collects in the pericardial cavity [77] and may contribute to systemic fibrinolysis after reinfusion [80, 85, 86].
278
F. D. Rubens
Thrombin is also a powerful stimulus for platelet activation [87] through the thrombin receptors [88, 89], resulting in platelet shape change and release of platelet granule contents [90]. This may contribute to the almost universal increase in the bleeding time seen postoperatively. Activated platelets are more rapidly cleared from the circulation, both in the body and in the extracorporeal device. Edmunds et al demonstrated that the postoperative platelet count was inversely proportional to the percentage of total blood that was returned by the cardiotomy system [91]. Other biomaterial-independent factors contributing to platelet activation (and subsequent dysfunction) after CPB include heparin at the doses used for CPB [92], activated complement [93], plasmin [94] and a variety of other soluble blood elements. Platelets adhere to binding sites located on surface-adsorbed fibrinogen and form aggregates [95]. Platelets also express P-selectin and form aggregates with circulating monocytes and to a lesser extent, neutrophils [96]. In summary, CPB leads to activation of multiple components of the hematologic system. Although some of this activation is related to contact with the biomaterial, it is now becoming increasingly evident that biomaterial-independent sources of blood activation must be considered as potential major contributors.
PREDICTIONS FOR FUTURE DEVELOPMENTS
‘The prospects for advances in development of artificial organs are clouded by ignorance of the critical events in blood-surface interaction’ [97]. Thirty years later, these comments still ring true. Our attempts to synthesize biocompatible surfaces for CPB have evolved to the point whereby we are better able to predict some degree of clinical improvement and patient benefit (Table 1). Table 1. Surface modification
In Vitro prediction
Clinical outcome
Heparin-coated Circuits
Thrombin Inhibition
Anti-inflammatory (complement inhibition)
Biomembrane Mimicry
(1) Modification fibrinogen adsorption (2) Platelet preservation
Anti-thrombotic (platelet preservation)
Surface Modifying Additive
(1) Modification fibrinogen adsorption (2) Platelet preservation
Anti-thrombotic (inhibition thrombin generation, platelet preservation, inhibition fibrinolysis)
PMEA
(1) Modification fibrinogen adsorption (2) Platelet preservation
Anti-thrombotic (platelet preservation) Anti-inflammatory (complement inhibition, bradykinin inhibition)
Biocompatibility and cardiopulmonary bypass
279
Further, although we do know slightly more about the key events at the bloodbiomaterial interface with CPB, we also are aware that there are many other previously unrecognized biomaterial independent factors that we can potentially modify during surgery that could minimize the deleterious consequences of CPB. We are now at a stage whereby we can capitalize on this knowledge in order to optimize patient outcome after CPB. Tomorrow’s cardiac surgery patient should be approached with a combined strategy involving the most blood compatible biomaterials and adjunctive pharmacotherapy (e.g. anti-proteases, anti-inflammatory drugs). This approach has been demonstrated to be feasible and attractive [98, 99]. Clearly, advances in improving patient outcomes after cardiac surgery will be dependent upon the continuing productive collaboration among surgeon-scientists and basic scientists. Acknowledgements I would like to extend my thanks to Drs. Marc Voorhees, Marian Packham and Raelene Kinlough-Rathbone for their helpful constructive comments in the formulation of this paper.
REFERENCES 1. J. H. Gibbon, Minnesota Med. 37, 171 (1954). 2. H. B. Shumacker, in: Cardiopulmonary Bypass. Principles and Practice, G. P. Gravlee, R. F. Davis, M. Kurusz and J. R. Utley (Eds), p. 22. Lippincott Williams & Wilkins, Philadelphia (2000). 3. R. A. DeWall, T. B. Grage, A. S. McFee and M. A. Chiechi, Surgery 50, 931 (1961). 4. R. A. DeWall, T. B. Grage, A. S. McFee and M. A. Chiechi, Surgery 51, 251 (1962). 5. D. Williams, Concise Encyclopedia of Medical and Dental Materials. Pergamon Press, Oxford (1990). 6. W. Heller et al., Adv. Exp. Med. Biol. 247B, 43 (1989). 7. G. Fuhrer, M. J. Gallimore, W. Heller and H. E. Hoffmeister, Blood Coag. Fibrinol. 3, 99 (1992). 8. R. W. Colman, J. Clin. Invest. 73, 1249 (1984). 9. Y. S. Bakhle, Br. J. Pharmacol. 59, 123 (1977). 10. A. F. Junod, Am. Rev. Respir. Dis. 115, 51 (1977). 11. J. V. Mombouli and P. M. Vanhoutte, J. Mol. Cell Cardiol. 31, 61 (1999). 12. F. D. Rubens et al., Ann. Thorac. Surg. 67, 689 (1999). 13. H. Kitaguchi, A. Hijikata and M. Hirata, Thromb. Res. 16, 407 (1979). 14. J. B. Steinberg, D. P. Kapelanski, J. D. Olson and J. M. Weiler, J. Thorac. Cardiovasc. Surg. 106, 1008 (1993). 15. N. C. Cavarocchi et al., J. Thorac. Cardiovasc. Surg. 91, 252 (1986). 16. R. C. Wiggins, P. C. Giclas and P. M. Henson, J. Exp. Med. 153, 1391 (1981). 17. P. A. Ward, J. Exp. Med. 126, 189 (1967). 18. M. E. Voorhees and B. F. Brian, International Anesthesiology Clinics 34, 29 (1996). 19. R. A. DeWall, H. E. Warden and R. C. Read, Surg. Clin. N. A. 36, 1025 (1956). 20. I. H. Rygg and E. Kyvsgaard, Acta Chir. Scand. 112, 443 (1956). 21. M. L. Bramson, J. J. Osborne and F. B. Main, J. Thorac. Cardiovasc. Surg. 50, 391 (1965). 22. A. J. Landé, L. Edwards and J. H. Block, Trans. Am. Soc. Artif. Intern. Organs 16, 352 (1970).
280 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.
F. D. Rubens J. Lister, Proc. Roy. Soc. London 12, 580 (1863). J. Bordet and B. Gengou, Ann. Inst. Pasteur (Paris) 15, 129 (1901). R. E. Baier, Bull. New York Acad. Med. 48, 257 (1972). J. H. Lee and H. B. Lee, J. Biomed. Mater. Res. 41, 304 (1998). E. W. Salzman, Blood 38, 509 (1971). V. L. Gott and R. L. Daggett, Ann. Thorac. Surg. 68, S19 (1999). V. L. Gott, J. D. Whiffen and R. C. Dutton, Science 14, 1297 (1963). C. A. Hufnagel et al., Ann NY Acad. Sci. 146, 262 (1968). O. Larm, R. Larsson and P. Olsson, Biomater. Med. Devices Artif. Organs 11, 161 (1983). G. Elgue, M. Blomback, P. Olsson and J. Riesenfeld, Thromb. Haemost. 70, 289 (1993). Trillium™ Affinity NT Oxygenator. Product monograph (2000). E. Ovrum et al., Ann. Thorac. Surg. 60, 365 (1995). L. K. von Segesser et al., Ann. Thorac. Surg. 58, 391 (1994). O. M. Shapira et al., J. Card. Surg. 11, 307 (1996). G. S. Aldea et al., Ann. Thorac. Surg. 62, 410 (1996). C. B. Mahoney, Perfusion 13, 192 (1988). H. te Velthuis et al., J. Thorac. Cardiovasc. Surg. 114, 117 (1997). C. Baufreton et al., Ann. Thorac. Surg. 63, 50 (1997). R. C. Gorman et al., J. Thorac. Cardiovasc. Surg. 111, 1 (1996). L. H. Edmunds, Ann. Thorac. Surg. 58, 285 (1994). S. Khuri et al., Ann. Thorac. Surg. 60, 1008 (1995). E. W. Muriithi et al., Ann. Thorac. Surg. 69, 1827 (2000). E. Ovrum, E. AmHolen, G. Tangen and M.-A. L. Ringdal, Ann. Thorac. Surg. 62, 1128 (1996). C. R. Wildevuur et al., Eur. J. Cardiothorac. Surg. 11, 616 (1997). M. Ranucci et al., Ann. Thorac. Surg. 67, 994 (1999). V. Videm et al., J. Thorac. Cardiovasc. Surg. 117, 794 (1999). V. Videm et al., J. Thorac. Cardiovasc. Surg. 117, 803 (1999). M. Pekna et al., Ann. Thorac. Surg. 58, 421 (1994). T. Wiedmer et al., Blood 82, 1192 (1993). P. J. Sims and T. Wiedmer, Semin. Cell Biol. 6, 275 (1995). E. Ovrum et al., Ann. Thorac. Surg. 60, 1755 (1995). O. Moen et al., Ann. Thorac. Surg. 63, 105 (1997). B. M. Steinberg et al., Ann. Thorac. Surg. 60, 525 (1995). R. M. Barstad et al., British Journal of Haematology 94, 517 (1996). H. P. Wendel, N. Weber and G. Ziemer, Immunopharmacology 43, 149 (1999). T. Nilsson, Thromb. Haemost. 49, 193 (1983). S. T. Olson, R. Sheffer and A. M. Francis, Biochem. 32, 12136 (1993). M. D. Kazatchkine, D. T. Fearon, J. E. Silbert and K. F. Austen, J. Exp. Med. 150, 1202 (1979). T. E. Mollnes et al., Artif. Organs 19, 909 (1995). C.-C. Tsai et al., ASAIO J. 40, M619 (1994). R. S. Ward and J. S. Riffle, US Patent No. 4,663,413 (1987). A. J. Lovinger, B. J. Han, F. J. Padden and P. A. Mirau, J. Polymer. Sci. 31, 115 (1993). T. Okano et al., J Biomed. Mater. Res. 15, 393 (1981). T. Okano et al., J. Biomed. Mater. Res. 20, 919 (1986). J. O. Defraigne et al., Ann. Thorac. Surg. 70, 2075 (2000). H. Suhara et al., Ann. Thorac. Surg. 71, 1603 (2001). M. A. Packham, G. Evans, M. F. Glynn and J. F. Mustard, J. Lab. Clin. Med. 73, 686 (1969). D. T. Fearon, J. Exp. Med. 152, 20 (1980). P. Tengvall, A. Askendal and I. Lundstrom, J. Biomed. Mater. Res. 35, 81 (1997). N. Saito, S. Motoyama and J. Sawamoto, Artif. Organs 24, 547 (2000). E. J. Campbell et al., ASAIO J. 40, M853 (1994).
Biocompatibility and cardiopulmonary bypass 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99.
281
F. De Somer et al., Eur. J. Cardiothorac. Surg. 18, 602 (2000). O. D. Ratnoff and J. E. Colopy, J. Clin. Invest. 34, 602 (1955). M. D. Boisclair et al., Blood 82, 3350 (1993). J. H. Chung et al., Circulation 93, 2014 (1996). M. Ernofsson, S. Thelin and A. Siegbahn, J. Thorac. Cardiovasc. Surg. 113, 576 (1997). R. Parratt and B. J. Hunt, Br. J. Haemat. 101, 40 (1998). N. Tabuchi, J. de Haan, P. W. Boonstra and W. van Oeveren, J. Thorac. Cardiovasc. Surg. 106, 828 (1993). L. D. Brace and J. Fareed, Semin. Thromb. Hemost. 11, 190 (1985). B. Rucinski et al., Blood 53, 47 (1979). J. I. Weitz et al., J. Clin. Invest. 86, 385 (1980). J. de Haan et al., Ann. Thorac. Surg. 59, 901 (1995). J. de Haan et al., J. Thorac. Cardiovasc. Surg. 106, 1017 (1993). J. de Haan et al., J. Thorac. Cardiovasc. Surg. 111, 272 (1996). V. A. Ferraris et al., Ann. Thorac. Surg. 65, 352 (1998). R. W. Colman, Hemat./ Onc. Clin. NA 4, 27 (1990). P. B. Tracy, Semin. Thromb. Hemost. 14, 227 (1988). R. L. Kinlough-Rathbone and D. W. Perry, Thromb. Haemostas. 74, 958 (1995). L. H. Edmunds, N. C. Saxena, P. Hillyer and T. J. Wilson, Ann. Thorac. Surg. 25, 306 (1978). N. Ellison, L. H. Edmunds, Jr. and R. W. Colman, Anesthesiology 48, 65 (1978). C. S. Rinder et al., J. Clin. Invest. 96, 1564 (1995). S. Rabhi-Sabile and D. Pidard, Thromb. Haemostas. 73, 693 (1995). F. D. Rubens et al., Thromb. Haemostas. 73, 850 (1995). C. S. Rinder et al., Blood 79, 1201 (1992). M. G. Davies et al., J. Vasc. Surg. 23, 410 (1996). P. G. M. Jansen et al., Ann. Thorac. Surg. 61, 1363 (1996). S. Bannan and P. G. Martin, Br. J. Haematol. 101, 455 (1998).
This page intentionally left blank
Review Surface analysis methods for characterizing polymeric biomaterials K. MERRETT 1 , R. M. CORNELIUS 2 , W. G. MCCLUNG 2 , L. D. UNSWORTH 2 and H. SHEARDOWN 2,∗ 1 Department of Chemical Engineering, University of Ottawa, Ottawa, 2 Departments of Chemical Engineering and Pathology and Molecular
ON, K1N 6N5, Canada Medicine, McMaster
University, Hamilton, Ontario, L8S 4L7, Canada Received 7 December 2001; accepted 1 April 2002 Abstract—Surface properties have an enormous effect on the success or failure of a biomaterial device, thus signifying the considerable importance of and the need for adequate characterization of the biomaterial surface. Microscopy techniques used in the analysis of biomaterial surfaces include scanning electron microscopy, transmission electron microscopy, atomic force microscopy, and confocal microscopy. Spectroscopic techniques include X-ray photoelectron spectroscopy, Fourier Transform infrared attenuated total reflection and secondary ion mass spectrometry. The measurement of contact angles, although one of the earlier techniques developed remains a very useful tool in the evaluation of surface hydrophobicity/ hydrophilicity. This paper provides a brief, easy to understand synopsis of these and other techniques including emerging techniques, which are proving useful in the analysis of the surface properties of polymeric biomaterials. Cautionary statements have been made, numerous authors referenced and examples used to show the specific type of information that can be acquired from the different techniques used in the characterization of polymeric biomaterials surfaces. Key words: Surface characterization; SEM; AFM; XPS; SIMS; contact angle; ellipsometry.
INTRODUCTION
Surface properties can have an enormous effect on the success or failure of a biomaterial device. It is widely accepted that such factors as surface preparation and the subsequent characterization are central issues in biomaterials research [1]. Characterization of biomaterial surface properties should be thorough. While acknowledging that individual research groups may be limited, due to a variety of factors, in the availability of surface characterization techniques, it must be recognized that such ∗ To
whom correspondence should be addressed. Phone: 905 525-9140 ext. 24794. Fax: 905 521 1350. E-mail:
[email protected]
284
K. Merrett et al.
characterization is essential for correlating any surface modifications with changes in biological performance. In addition, information gained may allow for tailoring future surface modifications to favour specific biological responses [1– 14]. The properties that are of interest in the characterization of biomaterial surfaces include the chemical structure, the hydrophilicity or hydrophobicity, the presence of ionic groups, the morphology (i.e. the domain structure), and the topography (i.e. the surface roughness, planarity, and feature dimensions) [15– 18]. Varying degrees of information about these properties can be obtained using different analysis methods [8, 17] including microscopic, spectroscopic and thermodynamic as well as other methods. Choice of the surface characterization method used can be influenced by a plethora of considerations including the type of measurement required, the extent of the analyzed surface region, the required precision and accuracy, the influence of the technique on the surface (i.e. does the probe, electron beam, ion beam, X-ray, required sample preparation, or the analysis environment induce undesirable effects on the surface of interest), the influence of the sample on the instrument, limitations imposed by the surface, as well as the ease of use and availability of equipment [19, 20]. In addition, a very realistic factor or constraint that may influence the choice of surface analysis techniques used is that many surface analysis facilities have become centralized, and there are significant costs associated with sample processing and preparation, the timed usage of equipment, and technician time. The aim of this paper is to provide a brief, easy to understand, synopsis of some of the specific techniques frequently used in the characterization of biomaterial, particularly polymeric biomaterial, surfaces. In addition some discussion and reference is made to emerging methods that have recently come to the forefront of biomaterials research.
SURFACE ANALYSIS METHODS
Microscopic methods The success of using microscopic methods to characterize biomaterial surfaces is well established [11, 13]. The microscopy techniques often found in the biomaterials literature include [1] scanning electron microscopy (SEM), transmission electron spectroscopy (TEM), scanning tunneling microscopy (STM) and atomic force microscopy (AFM). More recent developments in the area of light microscopy include the optical near-field principle, and the confocal laser scanning technique [13, 21]. Scanning electron microscopy and transmission electron microscopy. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are commonly used for studying both the surface morphology of and the cellular response to biomaterials [11, 22]. These techniques make use of a primary beam of electrons that interact with the specimen of interest, in a vacuum environment, resulting in different types of electrons and electromagnetic waves being emitted.
Surface characterization of biomaterials
285
Figure 1. Scanning electron micrograph of the inside polymer surface of a fibre obtained from a bleach/ formaldehyde reprocessed Gambro 21S Polyamide hemodialyzer. Bar = 1 µm. SEM image was obtained using a Phillips SEM 502 microscope. Sample was air dried and sputter coated with gold prior to insertion into the SEM. Figure courtesy of R. M. Cornelius and J. L. Brash.
The secondary electrons ejected from the specimen surface are collected and displayed to provide a high-resolution micrograph. SEM sample preparation involves fixation (if proteins, cells, or tissue are present), followed by drying, attachment to a metallic stub, and then coating with a metal prior to data collection. The thin metallic coating, usually applied by sputter coating, is typically 20 to 30 nm in thickness. Common conductive metals used include gold, platinum, or gold/ palladium alloy. It should be noted that the drying and metal coating processes used in the preparation of some polymeric materials might alter surface morphology, particularly those surfaces that may undergo changes in a hydrated environment. Upon insertion of the sample into the SEM, acquisition of the micrographs can usually be done fairly quickly allowing for a large number of images to be obtained with varying magnifications. Newer SEM models are reported to resolve in the nm range with magnifications in excess of 200 000×. Photographic prints, or computerized image acquisition, provide a permanent record. Micrographs typically included in the biomaterial literature depict images of 10 to 300 µm in length (see Fig. 1). In addition to imaging the surface morphology of polymeric biomaterials, the SEM can be combined with other analysis methods such as energy dispersive X-ray analysis (EDX) to determine elemental distribution (11) and IR and Raman spectroscopy to monitor surface modification procedures [23]. EDX results are typically obtained from a sampling
286
K. Merrett et al.
depth on the order of micrometers, thus are more representative of the bulk material rather than the surface [11]. TEM sample preparation involves fixation (if proteins, cells, or tissue are present), processing, embedding and sectioning. Embedding media can include methacrylates, polyester and acrylic resins, although epoxy resins are now commonly used. Specimens are typically sectioned using a microtome and need to be very thin since electrons with an accelerating voltage of 100 kV will not penetrate specimens more than 1 µm thick. Good resolution and clarity of detail can normally be obtained with sample thicknesses on the order of 50– 90 nm. A modest TEM can resolve in the sub nm range with magnifications considerably higher than 200 000×. It should however be noted that the embedding and sectioning processes used in the preparation of some polymeric materials may alter the polymeric material itself or the quality of the image obtained due to factors such as drying, thickness variations, wrinkling or compression. Micrographs typically included in the biomaterial literature depict images of 5 to 300 µm in length (see Fig. 2). Scanning tunneling microscopy. In scanning tunneling microscopy (STM), 3-dimensional images of surface topography of samples are obtained by monitoring the tunneling current flowing between an extremely sharp conductive probe and the sample surface. As the probe scans the surface, the magnitude of this current is inversely proportional to the probe/ surface separation, with a change in the separation producing an order of magnitude change in the tunneling current. If the probe scans a raised area on the surface, the current increases. To compensate, a piezoscanner tube moves the probe tip, returning the tunneling current to its original value [12]. Scan sizes typically range from 10 nm × 10 nm to 15 µm × 15 µm, and detect less than nm vertical changes in topography not possible with electron microscopes. This technique also has advantages over other imaging techniques, such as SEM and TEM, in that no special sample preparation procedures are required. However, the application of this method in the characterization of polymeric biomaterials is somewhat limited in that most polymers are not sufficiently conductive to allow STM images to be generated [12, 24]. A way around this limitation is to coat the polymeric materials of interest with an extremely thin conductive metallic coating and then image the surface. However the resolution obtained is limited by the nature of the coating. Atomic force microscopy, discussed next, overcomes the problem of the required sample conductivity needed for STM. Atomic force microscopy. Atomic force microscopy (AFM) has become the most common type of scanning microscopy used for polymeric biomaterials [25]. A three-dimensional image of the surface is created by scanning a tip attached to the end of a cantilever across the surface and monitoring the minute forces of interaction between the sample surface and probe [12, 25]. The forces of interaction may be repulsive or attractive and this gives rise to the different modes of operation of the AFM. A very high resolution of surface topography can be obtained, with
Surface characterization of biomaterials
287
(a)
(b) Figure 2. Transmission electron micrographs of a cross section of a fibre obtained from a bleach/ formaldehyde reprocessed Gambro 21S Polyamide hemodialyzer. Arrow indicates inside polymer surface of the fibre. (a) Bar = 5 µm; (b) bar = 1 µm. TEM images were obtained using a JEOL 1200EX Biosystem. Fibres were embedded in epoxy and 60 nm thick cross sections of the fibre were cut. Figure courtesy of R. M. Cornelius and J. L. Brash.
288
K. Merrett et al.
dimensions on the nanometer scale [12, 13], although it should be noted that properties and dimensions of the cantilever and tip, as well as the selected mode of operation, play an important role in determining the sensitivity and resolution of the acquired image. Unlike electron microscopes, a significant advantage of AFM is that sample topographies, as well as surfaces roughness values, can be obtained without surface treatment or coating which may damage or alter the material surface under investigation [12, 13, 16, 26, 27]. Furthermore, AFM images can be acquired under vacuum, air, or liquid conditions. The ability to image polymeric materials within an aqueous environment is extremely useful in the biomaterials field, as it allows for the examination of the surface of biomaterials in an environment similar to one that would be found in an implant situation, thus allowing for the examination of dynamic processes such as erosion, hydration, and adsorption at interfaces. For example, it is possible to visualize individual plasma protein molecules under aqueous environments using phase imaging AFM [12, 13, 28]. Although AFM is proving to be an extremely useful technique in providing a 3D visualization of the biomaterial surfaces being studied, the time required, dependant on such factors as scan size and scan rate, to obtain quality images can be significant (i.e. in excess of 20 min/ image). In addition, since typically scan sizes are small (i.e. ranging from 500 nm × 500 nm to 15 µm × 15 µm), variations in the surface may be missed. AFM — contact mode. In contact mode, the AFM tip scans across a surface at very low force and is deflected by repulsive forces acting between the tip and the surface atoms. A photodiode detector monitors the deflections of a laser light reflected from the tip of a cantilever. A feedback loop maintains constant deflection of the cantilever, by vertically moving the scanner as it scans laterally across the surface. A computer stores the information and a topographic image with potentially atomic-scale resolution is generated. The forces at the tip are very small (0.01 to 1.0 N /m in air) and metal or hard polymeric surfaces are not generally damaged [7]. However, the lateral shear forces caused by the scanning motion may alter soft materials, thus distorting measurement data and causing damage to the sample [29]. Obtaining images of hydrated polymeric materials in fluid may be further hampered by the fact that some hydrated polymers are softer than dried samples leading to an increase in sample deformation and damage and a reduced image quality resulting from the dragging motion of the tip. AFM — non contact mode. In non-contact mode, attractive rather than repulsive forces are measured. The scanning tip is oscillated perpendicular to, and just above the sample surface with an amplitude typically less than 10 nm. As with contact mode, a photodiode detector monitors the deflections of the laser light reflected from the tip of a cantilever. A feedback loop maintains constant oscillation amplitude or frequency, as the scanner moves laterally. A computer stores the information and a topographic image, with a lower resolution than in contact mode, is generated. Noncontact mode may work well with hydrophobic polymers. However hydrophilic
Surface characterization of biomaterials
289
Figure 3. AFM image of the inside polymer surface of a fibre obtained from a bleach/ formaldehyde reprocessed Gambro 21S Polyamide hemodialyzer. Scan size is 1 µm × 1 µm. AFM image was obtained using a Nanoscope III scanning probe microscope (Digital Instruments Inc.) operated in tapping mode in air. Figure courtesy of R. M. Cornelius and J. L. Brash.
polymers or imaging in a liquid environment in non-contact mode is not typically used due to the low resolution obtained. AFM — tapping mode. In tapping mode, the cantilever is vibrated at or near its resonance frequency, and lightly taps the sample surface with an amplitude typically ranging from 20 to 100 nm. A split photodiode detector monitors the deflections of the laser light reflected from the tip of the cantilever. A feedback loop maintains constant oscillation amplitude or frequency, as the scanner moves laterally across the surface. This mode therefore maintains the high-resolution capabilities of contact mode but is not destructive since there are no lateral frictional forces applied to the sample that can distort or damage the material [12]. Tapping mode AFM has proved to be very successful for high-resolution studies of polymeric biomaterials allowing for characterization of nanometer scale features not visible by other microscopic techniques (see Fig. 3). Tapping mode AFM images typically included in the biomaterial literature have a scan size ranging from 500 nm × 500 nm, to 15 µm × 15 µm. AFM — phase imaging mode. The phase imaging mode can be used to map the surface composition of a sample. In this mode, the cantilever is vibrated at or near its resonance frequency, and lightly taps the sample surface. A feedback loop maintains constant oscillation amplitude, as the scanner moves laterally across the surface. A phase image shows the phase difference between oscillation of the piezoelectric
290
K. Merrett et al.
crystal that drives the cantilever and oscillation of the cantilever itself as it interacts with the surface. Phase imaging can reveal fine features that are obscured by a rough surface topography. However factors such as surface hardness, elasticity, adhesive properties, and surface charge may affect the phase images obtained. Phase and tapping mode (height) images can be obtained simultaneously (see Fig. 4), so the position of the surface features observed in a phase image can be correlated directly with the surface topography. This mode can therefore be used to determine the size, shape and spacing of different material domains that could not otherwise be discerned from height alone [27, 30]. For example, phase imaging AFM can successfully detect adsorbed proteins that are not observable in conventional topographic images, since proteins imaged on surfaces with roughness near the dimensions of the protein cannot be distinguished from the material topography with conventional AFM [27], which is limited to imaging proteins only on smooth surfaces (1 µm2 roughness of <0.5 nm) [30, 34]. AFM using coated tips. Recently the power of AFM has been coupled with appropriate biomolecules to enable the study of interactions of the surface with various proteins and lipids. While this technique has been applied more often to study specific interactions with biological systems including polysaccharides of living microbial cells [31], as well as extracellular ATP on living cells [32], it has also been applied to the characterization of polymeric [33, 34] and metallic [35] materials. The use of these techniques has added an additional dimension to AFM in addition to that used conventionally for analysis of morphology and topography of a biomaterial surface. Often a variety of characterization techniques are used in the evaluation of biomaterial surfaces. However, it should be noted that the use of even different microscopic techniques might give rise to a different view of the surfaces obtained. Figures 1, 2, and 3 show SEM, TEM and AFM images respectively of the same polyamide polymeric material. The results obtained by SEM show a fairly smooth surface, with some very distinct porous regions. TEM results show a smooth surface, and in addition show that the bulk polymer contains a thick support layer consisting of interconnected highly porous domains. AFM results show significant surface morphology, not visible by the other techniques used. Thus, the combination of microscopic techniques used may play a valuable role in the characterization of surface features and morphology of polymer biomaterials. Confocal scanning microscopy. The major disadvantage with conventional light microscopy (LM) lies in illumination: because the entire specimen is illuminated, in-focus and out-of-focus information points contribute equally to the image resulting in blurring and poor contrast. There is also a sharp decline in image quality with increasing sample thickness, resulting in the need for thinly sectioned samples. Confocal scanning light microscopy (CSLM) provides blur-free optical sectioning of a specimen by eliminating out-of-focus information through spatial filtering using a point source of light for excitation. The technique is convenient, in that no special
Surface characterization of biomaterials 291
Figure 4. AFM height and phase images of a polycarbonate urethane. Phase mode image shows a densely crystallized top layer believed to contain a high hard segment component. Images were obtained using a Multimode AFM (Digital Instruments Inc.) operated in tapping mode in air. Figure courtesy of J. P. Santerre, Y. Tang and I. Revenko, Surface Science 491, 346 (2001). Copyright © 2001 Elsevier Science BV. Reprinted by permission of Elsevier Science BV.
292
K. Merrett et al.
sample preparation is required and features high-resolution images as compared to LM. Confocal microscopy is capable of resolving details at interfaces that could previously only be seen using electron microscopy of processed specimens [36]. Image resolution in CSLM is generally given as 0.2 µm in the xy-plane (transverse plane) and about 0.6 µm in the z-plane (the in-focus optical section) [37]. The actual useful depth of CSLM is limited to approximately 100– 200 µm, even with the use of near infrared illumination, which is more penetrating than visible light [38, 39]. The truly outstanding feature of CLSM is that two-dimensional pictures can be generated by scanning points across the focal plane of the specimen [36, 38] and subsequently compiled to give detailed three-dimensional images [40, 41]. Three imaging modes available with confocal microscopes, each named according to the origin of the light that is imaged [38] include confocal epitransmission with backscattered light from within the tissue, confocal reflectance, with light reflected from an opaque surface and confocal epifluorescence with light emitted by fluorescence. Confocal microscopy is sensitive to chromatic aberrations, which can be a problem in work employing white light or laser excitation of multiple fluorophores, which need to be imaged simultaneously, but which fluoresce with and are excited by different wavelengths [37, 38, 40]. Despite this drawback, photobleaching and phototoxicity to the surrounding specimen is minimal [38]. Also, the confocal image may be degraded by unwanted reflections of light from the surface of the objective lens or the specimen surface. However, this problem can be minimized using suitable immersion optics [37]. In the field of biomaterials, CSLM has been used in a number of applications. By conjugating a fluorophore to aqueous hydrogel solutions, CLSM in fluorescence mode has been used to investigate the bulk structure of hydrogels [42]. In this manner, hydrogel samples may be imaged in situ, a distinct advantage over techniques such as SEM where critical point drying and freeze etching may alter the original gel structure. Detailed morphological characterization of the hydrogel may be accomplished by assembling images collected at successive depth intervals from the surface into the bulk of the polymer. Pore sizes in the micrometer range and the polymer’s three-dimensional structure may thus be obtained. While there are problems with tissue accessibility in live animal work due to equipment design, real time (video rate) confocal imaging is achievable using the tandem scanning confocal microscope [43]. Confocal microscopes also possess the ability to view individual cells, nuclei and nucleoli without staining [44] and to view hollow organs via intraluminal endoscopy [38, 45]. Spectroscopic methods Spectroscopic methods are widely used to reveal valuable information regarding the constituent elements and chemical structure near the surface region of a sample [46]. Two new characterization techniques utilized in the biomaterials field are scanning transmission X-ray microscopy (STXM), and photoelectron emission microscopy (PEEM). Both techniques require synchrotron light sources
Surface characterization of biomaterials
293
and advanced instrumentation. While STXM has the advantages of not being affected by sample charging or topography, as well as the ability to image the sample in solution, the depth of sampling is full film often on the order of 100 nm. PEEM has a sampling depth of typically 5– 10 nm for polymers. However, it requires ultra high vacuum, no sample charging and a sample roughness of <30 nm. Both techniques, STXM and PEEM, have a high lateral spatial resolution on the order of 50 nm, and excellent chemical sensitivity. Another paper in this special issue will discuss the relevance of these two emerging techniques (STXM, PEEM) in the biomaterials field. Much more commonly utilized in the surface characterization of biomaterials, the traditional spectroscopic techniques used include: • Auger electron spectroscopy (AES). • X-ray photon spectroscopy (XPS). • Secondary ion mass spectrometry (SIMS). • Attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR). Auger electron spectroscopy (AES). Auger electron spectroscopy (AES) has been used for investigating surface morphology [47], and in particular elemental analysis. A focused beam of electrons is used to excite Auger electrons from the surface, which are then detected and analyzed. No special sample preparation is required and data collection is rapid (i.e. a few minutes) and reproducible. This technique has proven very useful in some engineering fields for elemental analysis and composition depth profiling. However, in general AES is limited in its use in the analysis of polymeric biomaterials as it is generally considered unsuitable for studying organic matter. Organic samples literally burn up in the electron beam, radically altering the chemistry and morphology of the sample being measured [20]. Because the risk of artifact and misinterpretation is large, the use of AES for characterizing organic, polymeric or biological specimens must be approached with caution [20]. X-ray photoelectron spectroscopy. X-ray photoelectron spectroscopy (XPS), also called ESCA (electron spectroscopy for chemical analysis) is widely used in biomaterial applications to determine the elemental composition of solid surfaces [48]. Special sample preparation is generally not required for XPS, although surface contamination upon storage or during transport from the research laboratory to the XPS facilities may certainly have an adverse effect on the XPS results obtained. In addition, additives, such as catalysts that may be used during the polymerization of polymeric biomaterials, or impurities can be present at the surface of the polymers and thus contribute significantly to the XPS results obtained. The principle of XPS is based upon the emission of electrons from matter in response to irradiation of the surface by a beam of monochromatic X-rays. The kinetic energy of the emitted photoelectrons (KE ) is unique for the different elements as well as being sensitive to the chemical state of the atoms [8, 48]. The
294
K. Merrett et al.
two energies of X-rays commonly used in experimental practice are 1253.6 eV (MgKα ) and 1486.6 eV (AlKα ). Since the emitted electrons have little ability to penetrate matter, only those electrons emitted near from the outermost 100 Å of a polymeric sample can escape from the surface and be quantified [8, 49]. All elements, except hydrogen and helium, can be detected by the characteristic binding energies of the electrons with a sensitivity of about 0.1 atom percent [8, 11, 48, 49]. XPS survey (low resolution) scans (see Fig. 5a) are typically 1000 eV wide and are often used to identify, as well as quantify in terms of atomic %, the elements (see Table 1) present at the biomaterial surface. High resolution scans are typically 20 eV wide and are used to obtain information about chemical shifts which can provide additional details about the chemical environment of the detected elements [8, 17, 48, 49] as well as shake-up transitions which provide details about aromatic or unsaturated structures [8]. High resolution scans (Fig. 5b) are referred to as such, Table 1. Oxygen Content (atomic %) of a PU control surface, and a PU surface modified with PEO, determined from an XPS survey (low resolution) scan. The expected increase in oxygen content of the PEO modified surface was observed. Data courtesy of J. Tan and J. L. Brash Polymer
Oxygen content (%)
PU Control Surface (theoretical) PU Control Surface PU surface modified with PEO
17 17.1 26.2
Figure 5a. Survey (low resolution) XPS scans of a polyurethane (PU) surface, and a PU surface modified with polyethylene oxide (PEO). XPS signals for carbon (C), nitrogen (N) and oxygen (O) are indicated. Scans show an increase in the O peak. Table 1 shows the corresponding atomic % of oxygen present clearly indicating that the surface has been successfully modified with PEO. XPS spectra were obtained using a Leybold MAX200 X-ray photoelectron spectrometer, and analyzed using ESCATOOLS (Surface/ Interface, Mountain View, CA). Figure courtesy of J. Tan and J. L. Brash.
Surface characterization of biomaterials
295
because typically more time is spent acquiring data over a narrower energy range as compared to a survey scan (Fig. 5a) resulting in a better signal to noise ratio. The XPS technique is highly versatile, with a sampling depth related to the inelastic mean free path of photoelectrons in the surface region, which is typically greater that 30 Å for polymeric materials, although this depends on both the photoelectron energy and on the material studied [49]. Typically, the signal, under ultra-high vacuum, is gathered from a large sample area (i.e. ∼6 mm in diameter)
Figure 5b. High resolution XPS C1s detailed spectra of a polyurethane (PU) surface, and a PU surface modified with polyethylene oxide (PEO). Representative 4 peak curve fits of the XPS C1s signal are also shown. Table 2 shows the corresponding Carbon Atom Bonding (%) of C O C determined from best fit of C1s High Resolution XPS Spectra clearly indicating that the surface has been successfully modified with PEO. XPS spectrum was obtained using a Leybold MAX200 X-ray photoelectron spectrometer, and analyzed using ESCATOOLS (Surface/ Interface, Mountain View, CA). Figure courtesy of J. Tan and J. L. Brash.
296
K. Merrett et al.
Table 2. Carbon Atom Bonding (%) of C O C determined from best fit of C1s High Resolution XPS Spectra of a PU control surface, and a PU surface modified with PEO. The expected increase in C O C content of the PEO modified surface was observed. Data courtesy of J. Tan and J. L. Brash Polymer
C O C content (%)
PU control surface PU surface modified with PEO
33.5 66.1
in order to minimize collection time, and possible damage to some polymeric materials. However, lateral resolution using a smaller sample area, in the range of square microns, has been reported [48]. Refined instrumentation simplifies sampling handling and data collection [8]. Well developed theory and computer programs are available to assist in the interpretation of the data obtained. XPS is considered to be a relatively non-destructive technique, although some care is needed to ensure that the X-ray does not alter the surface chemistry [20]. Problems have been reported to occur with polymers, especially flouropolymers, with the effects being minimized by use of a monochromatic source in conjunction with an effective compensation method to avoid surface charging problems or by simply minimizing the X-ray exposure to the surface and optimizing spectral acquisition times [11]. Information about functional groups can be obtained by using derivitization reactions [8, 50]. Compositional variation as a function of sample depth can be obtained by measurement of the photoelectron intensities at different emission angles, termed angle resolved XPS [8, 51]. Angle resolved XPS can also be used to examine an overlayer that may not be uniform, to investigate a surface where coverage is believed to be patchy [11], or to examine the change or transition between bulk and surface of a ‘surface modified’ material. To obtain elemental information several thousand angstroms into the sample, argon etching can be used in conjunction with XPS although this method is sample destructive [8]. While the signal is typically gathered from a large sample area (i.e. 25 mm2 ), it is possible, as mentioned earlier, to use XPS to determine lateral variations in surface composition [8] and to estimate the thickness of both organic and inorganic layers [52]. In addition, hydrated freeze dried surfaces can be examined using XPS [8] which may be of considerable interest when characterizing polymer surfaces that may reorient upon exposure to an air, vacuum or aqueous environment. There is considerable literature available on the principles of XPS, analytical procedures, instrumentation and approaches to quantitative analyses [3, 8, 49, 53– 55]. In the study and characterization of biomaterial surfaces, XPS is possibly the most extensively used technique. It has been used for a variety of surfaces and surface modifications, including studies of adsorption and retention of chemicals such as antibiotics and bonding agents [11], for the detection of immobilized proteins [20], understanding the chemistry of the structure, formation and stability
Surface characterization of biomaterials
297
of plasma treated surfaces, [7, 52, 56– 59] as well as for the characterization of the steps of formation of thin coatings [60]. Secondary ion mass spectrometry (SIMS) and time-of-flight secondary ion mass spectrometry (ToF-SIMS). Secondary ion mass spectrometry (SIMS) is effective for providing detailed molecular surface information [7] and increased usage of this technique in the surface characterization of polymeric biomaterials has been observed in recent years. Typically, no special sample preparation is required for SIMS. However as with XPS, the surface must be clean and free of any contamination that may occur upon storage or transport. Surfaces are bombarded with a focused beam of ions or atoms and the energy from the incident beam (approximately 5– 25 keV) is transferred to the surface zone of the material resulting in the emission of secondary particles, some of which are ionized, at and around the impact site [8]. These ionized particles are separated as a function of the ration of mass per electric charge and positively and negatively charged species are detected in two different acquisitions [61]. Low flux or static SIMS results in minimal damage to the sample and the fragments emitted are characteristic of the surface molecular structure [8, 62]. High flux, or dynamic SIMS results in rapid etching of the surface during analysis and can be used to monitor changes in the elemental composition with depth [8, 61]. Time-of-flight SIMS (ToF-SIMS) is a very efficient method for characterizing the elemental (including H and isotopes) and molecular composition of the top surface of biomaterials [61]. Developments in the ToF-SIMS analyzer have increased the capabilities of static SIMS due to significantly improved mass transmission (independent of the ion mass), mass resolution, mass range and sensitivity [61]. In this case, the sample surface is bombarded with very short-pulsed ion beams. Between two consecutive pulses, all of the secondary ions are extracted and electrostatically accelerated into a field free drift region. Lighter ions will have higher velocities and hence will reach the detector at the end of the drift region earlier than the heavier masses [61]. From the time of flight, the mass to charge ratio can be determined [61]. Under static conditions, there is minimal surface damage with ToF SIMS [61]. This technique is also associated with a high mass range, that is theoretically unlimited but has a practical limit of m/z = 15 000 [7, 63]. While there is a high surface sensitivity of 107 to 1011 atoms/ cm2 ) [61, 64] and mass and lateral resolutions [60], quantification of data is difficult [61] and information depth can be different from the sampling depth [65]. SIMS can be used for the identification of all elements including hydrogen [8] as well as atomic and molecular ions [61] and extremely high mass fragments at very low concentrations. Dynamic SIMS permits depth profiling from one to two atomic layers (1 nm in thickness) up to 1 µm into the sample with a high mass resolution [8, 17]. Although the analysis destroys the surface, information generated is directly related to the initial surface [20]. While static SIMS can be applied successfully to the analysis of both organic and inorganic surfaces, the accuracy
298
K. Merrett et al.
of the information with dynamic SIMS is considerably reduced for organic and biological surfaces [20]. These techniques can be used to obtain direct proof of covalent binding of molecules to a surface [66], as well as to predict other surface properties including wettability, adhesiveness, and biological reactivity [61]. SIMS spectra can provide characteristic ‘fingerprints’ for biomaterial surfaces [67], and has been used to correlate surface chemistry with cell growth [56] as well as for monitoring the degradation kinetics of biodegradable polymers using molecular weight distributions of the oligomeric hydrolytic reaction products [68– 70]. Temperature-programmed SIMS offers information on adsorption energy and thus helps to distinguish between physisorption and chemisorption phenomena [71]. It is often used in conjunction with XPS to examine the integrity, mean thickness and chemical state of multi-layer biomaterial coatings [72], corroborating contact angle and XPS measurements and complementing data obtained from less surface sensitive techniques [61]. Infrared spectroscopy (IR) and attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR). Infrared spectroscopy is used to obtain information about molecular structure by measuring the frequency of IR radiation needed to excite vibrations in molecular bonds [24]. Sample preparation is minimal involving application of the polymeric material of interest, in film form, onto a crystal element. Instrumentation is relatively inexpensive, and the resulting spectra provide chemical bonding information [7]. Infrared spectroscopy in attenuated total reflection (ATR-FTIR) couples the analytical method of infrared spectroscopy with the physical phenomena of total internal reflection (i.e. reflection and refraction of electromagnetic radiation at an interface of two media having different indices of refraction) to restrict the analyzed volume on the surface region of the sample [73, 74]. For this technique, the incident electromagnetic waves are entirely reflected back into the initial medium. The electromagnetic field is established in the second medium as represented by an evanescent wave due to diffraction at the edges of the incident radiation at the interface [73]. In attenuated total reflectance (ATR) sampling mode the second medium is the material to be studied, with the first medium acting as the internal reflection element [24, 73]. Information about the molecular structure of the material, inter- and intra-molecular interactions, crystallinity, conformation (e.g. proteins) and orientation of molecules can be obtained through analysis of the infrared spectra [74, 75]. Although depth profiles can be obtained using this technique, the XPS and SIMS techniques discussed earlier are considered much more surface sensitive. Surface matrix-assisted laser desorption ionization mass spectrometry (Surface MALDI). Surface Matrix-Assisted Laser Desorption Ionization Mass Spectrometry (Surface-MALDI), also known as MALDI-time of flight mass spectrometry (MALDI-TOFMS) is a novel surface analytical method that is an extension of conventional MALDI-MS, offering extremely high mass resolution and very low detec-
Surface characterization of biomaterials
299
tion limits for macromolecular analytes [76]. Compared to traditional MALDI-MS which is used to record the molecular weights and purities of known solution-based analytes, MALDI-TOFMS can be used to analyze adsorbed multicomponent bimolecular layers directly on the biomaterial surface using the charge to mass ratio of expelled ions [77– 79]. Macromolecule adsorbed surfaces are coated with matrix molecules. The matrix molecules that form on the surface are subsequently irradiated by a pulsed UV laser with various wavelengths dependent on the applications for 3– 15 ns, resulting in the volatilization of surface adsorbed entities which are detected. Analysis area is a spot with a diameter of approximately 10 µm. For good mass resolution, ions of equal masses but different energies must be detected simultaneously [80], meaning that the time from ionization to detection should be less than a nanosecond. Analyte ejection into the vapour phase, while not fully understood, is thought to be affected by the ability of the matrix molecules to mediate UV energy transfer to organic molecules that do not absorb in the UV wavelength range [80]. Nicotinic acid, dihydroxybenzoic acid and sinapinic acid have been found to be suitable matrix molecules. Due to the low pHs involved (pH < 4.5), it is however not unlikely that the matrix molecules will affect the structure of the bioactive adsorbents, resulting in crosslinking, denaturation, or degradation. It is also essential that the samples under study be pure, that the matrix be present at a 100– 1000 fold excess (vol/ vol) and that the layer under study be between 0.1 and 1 µm thick. MALDI-TOFMS has a theoretically unlimited mass detection range that is limited in practice by the fact that larger ions with higher mass are usually moving too slowly to be registered by the detector which has a limited threshold detection energy [80]. Furthermore, while the technique has a relatively high sensitivity for low adsorbent concentrations [76], the mass resolving power decreases as the molecular weight of the macromolecule increases. For example, the mass resolving power of small molecules is 300– 500, while molecules larger than 200 kDa have a mass resolving power of 50– 100 [80]. This is clearly evident as the peaks become broader with increases in molecular weight, amplifying the error in molecular weight determination, which is dependent on finding the center of the peak [80, 81]. Other limitations of MALDI-TOFMS include the requirement for vacuum operation and the ions measured can be products of side reactions between adsorbents that can decay with time [80, 82]. The main sources of error in this technique are time errors, where ions of equal mass reach the detector at different times and metastable ion decay during flight to the detector [80]. At present, the technique is not quantitative and requires control experiments and parallel XPS analyses for quantification of adsorbed amounts [83]. This technique has been applied to a variety of systems. Polymer MALDITOFMS studies are extensive and the technique has been used to determine surface coupling reaction yields [84], and polyethylene glycol molecular weight distributions [85] for example. The technique is said to provide very high mass resolution and thus has the ability to separate proteins that may give overlapping signals in
300
K. Merrett et al.
PAGE analysis with very low detection limits [80] and can detect physiosorbed from covalently bound proteins on synthetic polymers carrying thin layers of biological molecules [80]. It has been successfully applied to detect molecular ions of a number of proteins adsorbed from single and complex solutions onto various synthetic biomaterials even though the adsorbed proteins are bound strongly enough to resist removal by standard aqueous rinsing protocols [78, 79]. However, there can exist a residence time effect. With increasing time following adsorption onto the synthetic biomaterial, some proteins are increasingly difficult to desorb and analyze by surface-MALDI-MS. This is believed to be a consequence of protein denaturation, which increases the binding strength of physiosorbed proteins [83]. MALDITOFMS also allows detection of minor and major proteinaceous constituents of biofouled layers at substantially below monolayer coverage [78] and has been used for the direct analysis of biofilms produced in vivo on synthetic biomaterials [79]. Thermodynamic methods Several methods can be used to obtain surface parameters related to the interfacial free energy and other thermodynamic interaction measures such as the enthalpy of adsorption or displacement [74]. This type of analysis has been shown to be an efficient method for obtaining preliminary information on the biocompatibility of biomaterials [4]. The major thermodynamic method used to characterize biomaterial surfaces are wetting or contact angle experiments. Contact angle methods. Measurement of the contact angle of a liquid test droplet on a solid surface is a straightforward technique revealing surface energetic information inaccessible by the surface spectroscopies [7]. Although, this method represents perhaps one of the earliest methods used to investigate surface structure, it still yields very useful information [8]. Contact angle measurements at biomaterials surfaces can be carried out by several different methods [17] including: • The Wilhelmy plate method. • The sessile drop method [86]. • The captive bubble method [87]. A drop of fluid is placed on the biomaterial surface of interest, an equilibrium position is achieved, and the contact angle determined from the tangent associated with the drop/ polymer surface. Contact angles, for immobile surfaces, are believed to be sensitive to the outermost 3– 10 Angstroms of a surface [7]. Typically, in the biomaterials literature, contact angles are measured in air, utilizing a goniometer, using the sessile drop method with water as the test fluid. Contact angle analysis of control and modified polymeric materials can quickly provide valuable information about the relative hydrophilicity/ hydrophobicity of the surfaces. In addition, information about the hysteresis can be obtained when comparing contact angles obtained during increase, referred to as the advancing contact angle, and decrease, referred to as the receding contact angle, in the test droplet volume [74]. Comparison of contact
Surface characterization of biomaterials
301
Figure 6. Advancing and receding water contact angles measured by a goniometer using the sessile drop method for low density polyethylene (LDPE), LDPE surface modified with allylamine, LDPE surface modified with allylamine and reacted with polyethylene oxide (PEO). Figure clearly shows a reduction in advancing and receding contact angles due to the allylamine surface modification of the control surface, as well as a further reduction upon reaction with PEO. Surfaces were rinsed with methanol and dried prior to contact angle determination. Figure courtesy of Z. Jia and J. L. Brash.
angle measurements obtained for control and surface modified polymeric materials can be used to confirm the successful alteration of the control surface (see Fig. 6). Acid base contact angles can also give information about surface chemistry [88] and underwater contact angles, using the captive bubble technique, can be used to give information about hydrated surfaces [8, 17, 74]. Typically no special procedures are required in the preparation of polymeric materials for contact angle analysis. However, the surfaces should be clean, smooth, homogeneous and not swell or dissolve in the test fluid [74]. It should also be recognized that factors such as the choice and purity of the organic liquids and water, droplet size, and time required to obtain readings may affect the contact angle results obtained. Although the application of contact angles in evaluating polymeric biomaterial surface properties is a simple and straight forward approach, the use of this technique assumes thermodynamic equilibrium as well as a smooth, homogeneous surface which does not swell or dissolve in the test liquid [74]. Contact angle hysteresis has been reported to occur with surfaces that are rough, chemically heterogeneous, or contaminated with surface-active agents [89], as well as for surfaces that reorient upon exposure to different fluids.
302
K. Merrett et al.
Other methods Ellipsometry. Ellipsometry is an optical technique that uses the properties of polarized light to non-destructively examine interfacial phenomena [90]. Film characteristics are determined through the change in polarization of an electromagnetic wave upon reflection from the interface between two media with dissimilar complex refractive indices. Ellipsometry is particularly attractive in that it can be used to non-destructively monitor the growth and evolution of thin films in situ. Film thicknesses of the order of nanometers can be measured, with errors on the order of 2 Å [91]. It has been used to measure the optical properties of many materials including both metal and ‘soft’ thin films. However, in order for accurate measurements to be obtained, a significant refractive index discontinuity must exist at the interface under study. Surfaces should also be free from scratches and as flat as possible. All self-nulling ellipsometric systems have the same basic components shown in Fig. 7. The operational setup includes a quasi-monochromatic (λ) light source that passes through a polarizer prism, faraday rod, and quarter wave plate. These optical components produce elliptically polarized light that interacts with the sample. The interaction changes the state of polarization, which is determined through the position of the analyzer prism required to stop the beam passage. The photodetector is used in a feedback loop configuration so as to change the analyzer angle to achieve the null position. The ratio of the ratio of the amplitude of the outgoing resultant wave to that of the incoming wave, defined as the total reflection coefficient [92, 93] can then be used to determine the film thickness and the optical constants of the substrate material, including the refractive index (n2 ) and extinction coefficient (k2 ). Adsorbed polymer studies have shown that macromolecular density normal to the interface decreases with increasing distance from the interface [94– 97]. Therefore, the measurements represent averages over the change in macromolecular density at the interface and average refractive index and film thickness are returned. From
Figure 7. Self-nulling ellipsometer setup. A monochromatic, linearly polarized, electromagnetic wave becomes elliptically polarized upon passing through several components. This polarization changes following reflection from the sample, and is determined by the analyzer prism position that nullifies the wave.
Surface characterization of biomaterials
303
these parameters an estimate of the macromolecular amount of adsorbed polymer, (), can also be determined [98]. Although the ellipsometric data for thick films can be used to calculate both the film thickness and the complex refractive index simultaneously, this same convenience does not apply to thin films due to the relatively small changes in the azimuth angle with these surfaces as shown in Fig. 8. Therefore, when analyzing thin films (d < 50 Å), both the film complex refractive index and film adsorption characteristics, particularly the dielectric properties, must be estimated. The film refractive index is usually based upon the refractive index for similar molecules. A value of 1.50 is usually assumed for any solid, dry, organic monolayer [99], based on crystalline polyethylene, which has a refractive index in the range of 1.49– 1.55 [100]. Alkyltrichlorosilane monolayers however have been found by X-ray studies to be more akin to bulk paraffin rather than crystalline paraffin and a suggested refractive index is closer to 1.45 [101]. Differences in the layer thickness corresponding to these differences in refractive index values are not great. Tillman et al. estimated that an increase in film refractive index of 0.05, resulted in a decrease in thickness of ∼1 Å [102]. A similar estimate by Porter et al. gave a thickness change of 1.4 Å [103]. Therefore, for simple alkyl chains of C10 and longer, nf = 1.5 should be used. For any film containing metal ions nf > 1.5 should be used and for simple alkyl monolayers of C9 or less nf < 1.5 is recommended [104]. For all other materials the refractive index of the corresponding pure material should be used.
Figure 8. Illustration of refractive index values for several models as a function of layer thickness. For thin layers (<50 Å) it is impossible to differentiate between refractive index values. The solution RI is 1.3355 ± i0.00, the substrate constants are n = 0.18 ± i3.1, the wavelength is 6328 Å, and the angle of incidence is 60.0◦ .
304
K. Merrett et al.
For in situ studies, interactions of the polymer with the surrounding fluid greatly affects the resulting thin film thickness, refractive index, and density profile normal to the interface. For example, poly (ethylene imine)- poly (ethylene glycol) copolymers have been shown to have a layer refractive index of 1.352 [105], which is quite different from the solid PEG refractive index of ∼1.45. In our lab, in situ ellipsometry results for adsorbed thiolated PEG give a refractive index of 1.359. Given that the inherent error of an ellipsometer is approximately 2 Å, nf values within 0.15 of the actual value are sufficiently accurate without affecting the thickness measurement significantly and a knowledge of the absolute value of the refractive index is not critical as long as the estimate is reasonable Ellipsometry has been used to measure the optical properties of many materials. In biomaterials applications, ellipsometry has been used to study blood protein adsorption [106, 107] and cell interactions with [108] solid surfaces. It has, for example, been used to examine the role of IgM and IgG adsorption on complement activation [109], the adsorption of fibrinogen, serum and plasma onto oligo (ethylene glycol) terminated alkane thiols self assembled on gold [110] and protein interactions with polymeric substrates modified with RGD [111] as well as for quantification of surface modification with alkanethiolates [112]. In situ wet cell design is a significant factor in obtaining reproducible, accurate kinetic ellipsometric information. The major design factor is the construction of the laser window. Based on the need for appropriate window material and the requirement for perpendicularity with respect to the incident polarized beam, common designs include either a fixed angle or a cylindrical system. The main advantage of a cylindrical cell is the ability to take readings at many angles without having to change cells. However, it can be difficult to set up the cell such that the sample is parallel to, and located at the center plane of the cylinder. The fixed angle system is therefore much easier to operate, and can be easy to construct with minimal birefringence errors. Common designs using a fixed angle setup include the use of glass microscope slides fixed using an epoxy and cured [113] or the use of cuvettes with edges constructed at an appropriate angle [114]. Regardless of the design selected however, several important aspects must be considered for accurate measurement of film thickness. The desired angle of incidence based on the Brewster’s Angle must be determined. Fused quartz and fused silica are commonly used as window materials, with fused silica being preferred due to its lower thermal expansion coefficient. In vacuum systems or wet environments it has been suggested that a quartz disc sandwiched between an o-ring and a Teflon restrainer provides a suitable seal [115]. The viability can be confirmed through the polarizer and analyzer readings in Zones 1 and 3, which should conform to the following: (P3 − P1 ) − 90◦ 0.3, A 0.3.
(1) (2)
Surface characterization of biomaterials
305
Specular neutron reflectivity. Specular neutron reflectivity (SNR) has the potential to be a powerful tool for elucidating biomaterial surface information and protein interfacial mechanisms. SNR measures the change in scattering length density at the interface between two different neutron refractive index materials [116]. It is suited to evaluating the adsorbed protein conformation for all three interface types, and is perhaps the best technique for studying adsorption at the liquid/ liquid interface. This technique offers the benefit of spatial resolution of ∼1.0 nm and penetration depths for some systems of several hundred nanometers. By reflecting neutrons from the sample, it is possible to independently and simultaneously determine both the composition and concentration distribution normal to the interface in question from the resulting interference fringes as shown in Fig. 9 [117– 119]. The high penetrating power of neutrons allows them to pass through a medium before reaching the interface under study. Other advantages include enhanced contrast through isotope substitution like deuterium labeling [120], minimal surface damage allowing for successive measurements on a single sample and neutron sensitivity to magnetic interactions [121]. The sensitivity of neutron reflection is ∼2 Å, or two methylene groups. Contrast is achieved by using the difference in the coherent scattering length densities (β) of D and H. However, the specialized nature of the equipment result-
Figure 9. Illustration of interference fringes for a gold coated silicon wafer, with chemisorbed PEGthiol. This can be used to determine the thin film thickness, and composition profile both normal and parallel to the interface. Figure courtesy of L. Unsworth, H. Sheardown and J. L. Brash.
306
K. Merrett et al.
ing in limited access. An additional limitation is the requirement of a large surface (20 cm2 ) for thin film experiments. SNR has been used to identify the surface composition for complex polymer mixtures [122], surface and interfacial segregation [123, 124], time dependant processes [125], liquid mixture adsorptions [126], surfactant (ethylene glycol) adsorption used in colloidal dispersions [127], copolymer adsorption at the liquid/ liquid interface [128] and protein adsorption at various interfaces [129– 134]. Studies have also been done on the adsorption of star and deuterated polyethylene oxide (PEO) at the solid liquid interface [135, 136] and lipidated PEO and peptides at the air water interface [137]. In our lab, SNR studies have been used to understand the adsorbed film characteristics for thiolated PEG on gold-coated silicon wafers. An example of the data that can be obtained is shown in Fig. 9. Once analyzed, these interference fringes will reveal the thickness of the adsorbed PEG layer and that change in the macromolecular density normal to the interface.
CONCLUSIONS
Since the properties of the surface can impact enormously on the success or failure of a biomaterial, the application of appropriate surface analysis techniques for correlating surface properties, including chemical structure, hydrophobicity, morphology and topography, and material performance is of paramount importance. Surface analysis methods can be used for verification of an intended surface modification procedure as well as for correlating surface properties to the biological performance. They also play key roles in understanding the effects of model surfaces with systematically varied surface properties. Characterization of biomaterials surface properties should therefore be as thorough as possible given technical and practical limitations. A combination of surface characterization methods is usually necessary to provide the comprehensive information necessary for correlation with performance. Spectroscopic techniques such as XPS are often used in combination to broaden knowledge about the surface. The ease and rapidity with which water contact angles can be applied means that this technique is often used in combination with others. Spectroscopic methods are also frequently used in combination with the microscopic methods. The choice of surface characterization method can be influenced by numerous factors that must be considered. In this paper we have provided an overview of techniques commonly used in biomaterials, particularly polymeric biomaterials, characterization, as well as some discussion of emerging methods that have recently come to the forefront of biomaterials research, including the applications and limitations of each. Acknowledgements The authors would like to thank the students of Dr. John Brash, J. Tan and Z. Jia, and former student J. P. Santerre, as well as Y. Tang and I. Revenko for allowing us
Surface characterization of biomaterials
307
to present some of their experimental results that clearly demonstrate the suitability of the different techniques discussed for the characterization of biomaterials. The authors would also like to thank Dr. Farid Bensebaa for critically reviewing portions of the manuscript.
REFERENCES 1. B. Kasemo and J. Lausmaa, Biomaterials, in: Surface Characterization: A User’s Sourcebook, D. Brune, R. Helborg, O. Hunderi and H. Whitlow (Eds). Scandinavian Science Publ., Weinheim (1997). 2. A. D. Cook, J. S. Hrkach, N. N. Gao, I. M. Johnson, U. B. Pajvani, S. M. Cannizzaro and R. Langer, J. Biomed. Mat. Res. 35, 513 (1997). 3. J. D. Andrade (Ed.), Surface and Interfacial Aspects of Biomedical Polymers I: Surface Chemistry and Physics. Plenum Press, New York (1985). 4. A. A. A de Queiroz, E. R. Barrak and S. C. de Castro, Journal of Molecular Structure 394, 271 (1997). 5. G. L. Grobe, III, J. A. Gardella, Jr., R. L. Chin and L. Salvati, Research Society Symposia Proceedings 55, 401 (1982). 6. N. Ignjatovic, V. Savic, S. Najman, M. Plavsic and D. Uskokovic, Biomaterials 22, 571 (2001). 7. E. E. Johnston and B. D. Ratner, J. Elec. Spectroscopy and Related Phenomena 81, 303 (1996). 8. B. D. Ratner, B. J. Tyler and A. Chilkoti, Clinical Materials 13, 71 (1993). 9. B. D. Ratner, A. S. Hoffman, J. E. Lemmons and F. J. Schoen, Biomaterials Science — An Introduction to Materials in Medicine. Academic Press, New York (1996). 10. L. Sabbatini and P. G. Zambonin, J. Elec. Spectroscopy and Related Phenomena 81, 285 (1996). 11. R. N. S. Sodhi, J. Elec. Spectroscopy and Related Phenomena 81, 269 (1996). 12. S. O. Vansteenkiste, M. C. Davies, C. J. Roberts, S. J. B. Tentfler and P. M. Williams, Progress in Surface Science 57, 95 (1998). 13. C. Werner and H. J. Jacobasch, The International Journal of Artificial Organs 22, 160 (1999). 14. R. H. West, A. J. Paul, S. Hibbert, P. Cahalan, L. Cahalan, M. Verhoeven, M. Hendriks and B. Fouache, Journal of Materials Science: Materials in Medicine 6, 63 (1995). 15. J. D. Andrade (Ed.), Surface and Interfacial Aspects of Biomedical Polymers I: Surface Chemistry and Physics. Plenum Press, New York (1985). 16. T. Cunningham, F. M. Serry, L. M. Ge, D. Gotthard and D. J. Dawson, Surface Engineering 16, 295 (2000). 17. D. Klee and H. Hocker, Advances in Polymer Science 149, 1 (2000). 18. Y. F. Missirlis and W. Lemm, Modern Aspects of Protein Adsorption on Biomaterials. KluwerAcademic Publishers, Dordrecht (1991). 19. N. L. Burns, K. Emoto, K. Holmberg, J. M. Van Alstine and J. M. Harris, Biomaterials 19, 423 (1998). 20. B. D. Ratner, Surf. Interface Anal. 23, 521 (1995). 21. J. Hwang, L. K. Tamm, Bohm, T. S. Ramalingam, E. Betzig and M. Eddin, Science 270, 610 (1995). 22. C. Battistioni, M. P. Casaletto, G. M. Ingo, S. Kaciulis, G. Mattogno and L. Pandolfi, Surf. Interface Anal. 29, 773 (2000). 23. R. Born, D. Scharnweber, S. Robler, M. Stolzel, M. Thieme, C. Wolf and H. Worch, Surface analysis of titanium based biomaterials, Fresenius J. Anal. Chem. 361, 697 (1998). 24. B. D. Ratner, A. Chilkoti and D. G. Castner, Contemporary methods for characterizing complex biomaterial surfaces, Clinical Materials 11, 25 (1992). 25. I. Amato, Science 276, 1982 (1997).
308
K. Merrett et al.
26. I. Revenko, International Laboratory (July), 32 (2000). 27. N. B. Holland and R. E. Marchant, J. Biomed. Mater. Res. 51, 307 (2000). 28. R. E. Marchant, M. D. Barb, J. R. Shainoff, S. J. Eppell, D. L. Wilson and C. A. Siedlecki, Thromb Haemostasis 77, 1048 (1997). 29. E. Meyer, L. Howald, R. Overney, D. Brodbeck, R. Luthi, H. Haefke, J. Frommer and H. Guntherodt, Ultramicroscopy 274, 42 (1992). 30. W. Stocker, S. N. Magonov, H. J. Cantow, J. C. Wittman and B. Lotz, Macromolecules 26, 5915 (1993). 31. M. Gad, A. Itoh and A. Ikai, Cell Biology International 21, 697 (1997). 32. S. W. Schneider, M. A. Egan, B. P. Jena, W. B. Guggino, H. Oberleithner and J. P. Geibel, Proc. Natl Acad. Sci. USA 96, 12180 (1999). 33. J. Y. Wong, T. L. Kuhl, J. N. Israelachvili, N. Mullah and S. Zalipsky, Science 275, 820 (1997). 34. C. Jeppesen, J. Y. Wong, T. L. Kuhl, J. N. Israelachvili, N. Mullah, S. Zalipsky and C. M. Marques, Science 293, 465 (2001). 35. D. J. Wold and C. D. Frisbie, J. Am. Chem. Soc. 123, 5549 (2001). 36. T. Appenzeller and C. Norman, Science 276, 1981 (1997). 37. D. M. Shotten, Histochem. Cell Biol. 104, 97 (1995). 38. P. J. Tadrous, Journal of Pathology 191, 345 (2000). 39. C. Cullander, J. Invest. Dermatol. Symp. Proc. 32, 166 (1998). 40. O. Akinyemi, A. Boyde, M. A. Browne, M. Hadravsky and M. Petran, Scanning 14, 136 (1992). 41. M. A. Browne, O. Akinyemi and A. Boyde, Scanning 14, 145 (1992). 42. F. Fergg, F. J. Keil and H. Quader, Colloid Polym. Sci. 279, 61 (2001). 43. W. M. Petroll, J. V. Jester and H. D. Cavenagh, Int. Rev. Exp. Pathol. 36, 93 (1996). 44. J. V. Jester, P. M. Andrews, W. M. Petroll, M. A. Lemp and H. D. Cavanagh, Journal of Electron Microscopy Technique 18, 50 (1991). 45. P. M. Delaney, R. G. King, J. R. Lambert and M. R. Harris, J. Anatomy 184, 157 (1994). 46. Y. Uyama, K. Kato and Y. Ikada, Adv. Polym. Sci. 137, 1 (1998). 47. C. Battistoni, M. P. Casaletto, G. M. Ingo, S. Kaciulis, G. Mattogno and L. Pandolfi, Surf. Interface Anal. 29, 773 (2000). 48. C. Werner, H. J. Jacobasch and G. Reichelt, J. Biomater. Sci. Polymer Edn 7, 61 (1995). 49. A. Jablonski, Polish J. Chem. 74, 1533 (2000). 50. C. D. Batich, Appl. Surf. Sci. 32, 57 (1988). 51. P. J. Cumpson, J. Elec. Spectroscopy and Related Phenomena 72, 25 (1995). 52. G. Iucci, A. Frontini, G. Infante, C. Furlani, L. Avigliano, N. Rosato, D. Del Principe, G. Palumbo and G. Polzonetti, Surf. Interface Anal. 30, 40 (2000). 53. A. Dilks, in: Electron Spectroscopy: Theory, Techniques and Applications, A. D. Bruker and C. R. Brundle (Eds), Vol. 4. Academic Press, London (1981). 54. D. R. Miller and N. A. Peppas, Journal of Macromolecular Science Reviews of Macromolecular Chemistry and Physics C26, 33 (1986). 55. B. D. Ratner and B. J. McElroy, in: Spectroscopy in the Biomedical Sciences, R. Gendreau (Ed). CRC Press, Boca Raton, FL (1986). 56. A. Chilkoti, B. Ratner and D. Briggs, Chem. Mater. 3, 51 (1991). 57. R. Daw, S. Candan, A. J. Beck, A. J. Devlin, I. M. Brook, S. MacNeil, R. A. Dawson and R. D. Short, Biomaterials 19, 1717 (1998). 58. J.-C. Lin and C.-H. Wu, Biomaterials 20, 1613 (1999). 59. R. W. Paynter, Surf. Interface Anal. 29, 56 (2000). 60. L. Dai, H. S. W. St. John, J. Bi, P. Zientek, R. C. Chatelier and H. J. Griesser, Surf. Interface Anal. 29, 46 (2000). 61. D. Leonard and H. J. Mathieu, Fresnius J. Anal. Chem. 365, 3 (1999). 62. J. C. Vickerman, A. Brown and N. M. Reed, Secondary Ion Mass Spectrometry, Principles and Applications. Clarendon Press, Oxford, UK (1989).
Surface characterization of biomaterials
309
63. B. Hagenhoff, Biosensors Bioelectronics 10, 885 (1995). 64. A. Benninghoven, E. Niehuis, T. Friese, D. Greifendorf, P. Steffens, Org. Mass Spectrom. 19, 346 (1984). 65. B. Hagenhoff, M. Diemel, A. Benninghoven, H.-U. Siegmund and D. J. Holtkamp, J. Phys. D: Appl. Phys. 25, 818 (1992). 66. D. Leonard, Y. Chevolot, O. Bucher, H. Sigrist and H. J. Mathieu, Surf. Interface Anal. 26, 783 (1998). 67. V. H. Perez-Luna, K. A. Hooper, J. Kohn and B. D. Ratner, Journal of Applied Polymer Science 63, 1467 (1997). 68. J. Chen and J. A. Gardella, Jr., Macromolecules 32, 7380 (1999). 69. J. Chen, J. W. Lee, N. L. Hernandez de Gatica, C. A. Burkhardt, D. M. Hercules and J. A. Gardella, Jr., Macromolecules 33, 4276 (2000). 70. J. W. Lee and J. A. Gardella, Jr., Macromolecules 34, 3928 (2001). 71. G. Egbers, K. Hackling and A. Benninghoven, in: Secondary Ion Mass Spectrometry, A. Benninghoven, B. Hagenhoff and H. W. Werner (Eds). Academic Press, London (1981). 72. R. H. West, A. J. Paul, S. Hibbert, P. Cahalan, L. Cahalan, M. Verhoeven, M. Hendriks and B. Fouache, Journal of Materials Science, Materials in Medicine 6, 63 (1995). 73. R. Barbucci, M. Casolaro and A. Magnini, Clinical Materials II, 37 (1992). 74. C. Werner and H. J. Jacobasch, International Journal of Artificial Organs 22, 160 (1999). 75. K. K. Chittur, Biomaterials 19, 301 (1998). 76. S. Jespersen, W. M. A. Niessen, U. R. Tjaden, J. van der Greef, E. Litborn, U. Lindberg and J. Roeraade, Rapid Commun. Mass Spectrom. 8, 581 (1994). 77. P. Kingshott, H. A. W. St. John, R. C. Chatelier and H. J. Griesser, Polym. Mater. Sci. Eng. 76, 81 (1997). 78. P. Kingshott, H. A. W. St. John, R. C. Chatelier and H. J. Griesser, J. Biomed. Mater. Res. 49, 36 (2000). 79. H. A. W. St. John, P. Kingshott and H. J. Griesser, Polym. Mater. Sci. Eng. 76, 83 (1997). 80. M. L. Gross, Mass Spectrometry in the Biological Sciences: A Tutorial, Series C: Mathematical and Physical Sciences, 353. Kluwer Academic Publishers, Boston (1992). 81. R. D. Oleschuk, M. E. McComb, A. Chow, W. Ens, K. G. Standing, H. Perreault, Y. Marois and M. King, Biomaterials 21, 1701 (2001). 82. B. Spengler, D. Kirsch and R. Kauffman, Rapid Commun. Mass Spectrom. 5, 198 (1989). 83. K. M. McLean, S. L. McArthur, R. C. Chatelier, P. Kingshott and H. J. Griesser, Coll. Surf. B: Biointerfaces 17, 23 (2000). 84. J. Liu, R. S. Loewe and R. D. McCullough, Macromolecules 32, 5777 (1999). 85. T. Kinumi, T. Saisu, M. Takayama and N. Haruki, J. Mass Spect. 35, 417 (2000). 86. T. Matsunaga and Y. Ikada, J. Colloid Interface Sci. 84, 8 (1981). 87. D. Li and A. W. J. Neumann, J. Colloid Interface Sci. 148, 190 (1992). 88. G. M. Whitesides and P. E. Laibinis, Langmuir 6, 87 (1990). 89. J. F. Joanny and P. G. de Gennes, J. Chem. Phys. 81, 552 (1984). 90. P. Drude, Ann. Phys. Chem. 36, 865 (1989). 91. J. P. Rabe, V. Novonty, J. D. Swalen and J. F. Rabolt, Thin Solid Films 159, 359 (1988). 92. R. M. A. Azzam and N. M. Bashara, Ellipsometry and Polarized Light. North-Holland Publishing Company, New York (1977). 93. O. S. Heavens, Optical Properties of Thin Solid Films. Dover Publications, New York (1965). 94. C. A. J. Hoeve, J. Polym. Sci. C30, 361 (1970). 95. C. A. J. Hoeve, J. Polym. Sci. C34, 1 (1971). 96. A. Silberberg, J. Phys. Chem. 66, 1884 (1962). 97. A. Silberberg, Pure Appl. Chem. 26, 583 (1971). 98. J. A. De Feijter, J. Benjamins and F. A. Veer, Biopolymers 17, 1759 (1978). 99. D. L. Allara and R. G. Nuzzo, Langmuir 1, 45 (1985).
310
K. Merrett et al.
100. J. Brandrup and E. H. Immergut, Polymer Handbook. John Wiley, New York (1975). 101. S. R. Wasserman, G. M. Whitesides, I. M. Tidswell, B. M. Ocko, P. S. Pershan and J. D. Axe, J. Am. Chem. Soc. 111, 5852 (1989). 102. N. Tillman, A. Ulman, J. S. Schildkraut and T. L. Penner, J. Am. Chem. Soc. 110, 6136 (1988). 103. M. D. Porter, T. B. Bright, D. L. Allara and C. E. D. Chidsey, J. Am. Chem. Soc. 109, 3559 (1987). 104. A. Ulman, An Introduction to Ultrathin Organic Films. Academic Press, San Diego (1991). 105. M. Malmsten, K. Emoto and J. M. Van Alstine, J. Coll Interfacial Sci. 202, 507 (1988). 106. L. Vroman and A. C. Adams, Ellipsometry in the Measurement of Surfaces and Thin Films. Nat. Bur. Std. Misc. Publ. 335ff (1964). 107. L. Vroman and A. C. Adams, Proceedings of the Symposium on Recent Developments in Ellipsometry. North-Holland, Amsterdam (1969). 108. G. Poste and C. Moss, Progress in Surface Science. North-Holland, Amsterdam (1972). 109. P. Tengvall, A. Askendal and I. Lundstrom, Colloids and Surfaces B: Biointerfaces 20, 51 (2001). 110. J. Benesch, S. Svedhem, S. C. Svensson, R. Valiokas, B. Liedberg and P. Tengvall, J. Biomater. Sci. Polymer Edn 12, 581 (2001). 111. M. P. Olivieri and K. S. Tweden, J. Biomed. Mater. Res. 46, 355 (1999). 112. V. A. Tegoulia and S. L. Cooper, J. Biomed. Mater. Res. 50, 291 (2000). 113. H. Brunner, T. Vallant, U. Mayer and H. Hoffmann, J. Coll. Interfacial Sci. 212, 545 (1999). 114. C. A. C. Karlsson, M. C. Wahlgren and A. C. Tragardh, Coll. Surf. B: Biointerfaces 20, 9 (2001). 115. K. A. Conrad, R. K. Sampson and H. Z. Massoud, Rev. Sci. Instrum. 67, 3954 (1996). 116. M. L. Fernandez, J. S. Higgins, J. Penfold, R. C. Ward, C. Shackleton and D. Walsh, Polymer 29, 1923 (1988). 117. J. Penfold and R. K. Thomas, J. Phys.: Condens. Matter. 2, 1369 (1990). 118. T. P. Russell, Mater. Sci. Rep. 5, 171 (1990). 119. X. L. Zhou and S. H. Chen, Phys. Rep. 257, 223 (1995). 120. J. Penfold, Physica B 173, 1 (1991). 121. J. R. Lu, Z. X. Li, J. Smallwood, R. K. Thomas and J. Penfold, J. Phys. Chem. 99, 8233 (1995). 122. W. A. Goedel, C. Luap, R. Oeser, P. Lang, C. Braun and R. Steitz, Macromolecules 32, 7599 (1999). 123. I. Hopkinson, K. T. Kiff and R. W. Richards, Macromolecules 28, 627 (1995). 124. J. Strutwolf, A. L. Barker, M. Gonsalves, D. J. Caruana, P. R. Unwin, D. E. Williams and J. R. P. Webster, J. Electroanalytical Chem. 483, 163 (2000). 125. I. R. Gentle, P. M. Saville, J. W. White and J. Penfold, Langmuir 9, 646 (1993). 126. J. R. Howse, J. Bowers, E. Manzanares-Papayanopoulos, I. A. McLure and R. Steitz, Phys. Rev. E 59, 5577 (1999). 127. E. J. Staples and J. R. Lu, Langmuir 15, 250 (1999). 128. B. J. Clifton, T. Cosgrove, R. M. Richardson, A. Zarbakhsh and J. R. P. Webster, Physica B 248, 289 (1998). 129. J. R. Lu and T. J. Su, Langmuir 15, 6975 (1999). 130. B. Harzallah, V. Aguie-Beghin, R. Douillard and L. Bosio, Int. J. Biological Macromolecules 23, 73 (1998). 131. J. R. Lu, T. J. Su, R. K. Thomas, J. Penfold and J. R. P. Webster, J. Chem. Soc. Faraday Trans. 94, 3279 (1998). 132. L.-T. Lee, B. K. Jha, M. Malmsten and K. Holmberg, J. Phys. Chem. B 103, 7489 (1999). 133. T. J. Su, J. R. Lu, R. K. Thomas, Z. F. Cui and Z. F. J. Penfold, Langmuir 14, 438, (1998). 134. A. P. Maierhofer, M. Lieb, D. G. Bucknall and T. M. Bayerl, Biophysical Journal 76, A211 (1999). 135. E. M. Lee, R. K. Thomas and A. R. Rennie, Europhys. Lett. 13, 135 (1990).
Surface characterization of biomaterials
311
136. D. J. Irvine, A. M. Mayes, S. M. Satija, J. G. Barker, S. J. Sofia-Allgor and L. G. Griffith, J. Biomed. Mater. Res. 40, 498 (1998). 137. H. Bianco-Peled, Y. Dori, J. Schneider, L. P. Sung, S. Satija and M. Tirrell, Langmuir 17, 6931 (2001).
This page intentionally left blank
Glucose binding to molecularly imprinted polymers HASOO SEONG 1 , HAI-BANG LEE 2 and KINAM PARK 1,∗ 1 Purdue
University, Departments of Pharmaceutics and Biomedical Engineering, West Lafayette, IN 47907, USA 2 Korea Research Institute of Chemical Technology, Biomaterials Laboratory, Taejon, Korea Received 18 December 2001; accepted 22 March 2002 Abstract—The main goal of this study was to prepare molecularly imprinted polymers (MIPs) with glucose recognition sites and to evaluate their glucose-binding properties for potential applications in glucose sensing and self-regulating insulin delivery devices. To mimic glucose-binding sites of natural proteins, monomers possessing functional groups similar to amino acids were used. Vinyl acetic acid (VAA), acrylamide (AAm), 4-pentenoic acid (PA), and allyl benzene (AB) were copolymerized with a cross-linking agent (N,N -methylenebisacrylamide, BIS) in the presence of glucose as a template. The binding affinity of glucose to MIPs was examined by using an equilibrium dialysis technique. The dissociation constants of the MIPs were determined by Scatchard analysis. MIPs showed glucosebinding affinity, while polymers synthesized in the absence of glucose template did not show a glucose-binding property. MIPs composed of VAA, AAm, PA, and AB at optimized mole ratios of monomers and cross-linker showed the highest glucose-binding affinity, KD = 1.66 mM, which is comparable to that of a well-known glucose binding protein, concanavalin A (KD = 1.84 mM). The affinity between monomer and glucose was in the order VAA > AAm > AB > PA. Key words: Glucose; glucose affinity; glucose-binding protein; glucose sensor; molecularly imprinted polymer; glucose imprinting.
INTRODUCTION
A self-regulated insulin delivery system is defined as a system that is capable of releasing insulin in response to changing blood glucose levels [1]. The increase of the glucose level triggers the system to modulate the release of insulin, i.e. release of insulin at the right time in the right amount. This requires glucose-sensitive or glucosesensing ability in the system. A number of glucose-sensitive polymer and hydrogel systems have been developed. For convenience, they can be divided into systems utilizing glucose-sensitive swelling membranes [2, 3], glucose-sensitive erodible ∗ To
whom correspondence should be addressed. Purdue University, School of Pharmacy, West Lafayette, IN 47907, USA. Phone: (765) 494-7759. Fax: (765) 496-1903. E-mail:
[email protected]
314
H. Seong et al.
matrices [4– 6], immobilized insulin [7], a glycosylated insulin– concanavalin A (Con-A) complex [8, 9], and glucose-sensitive phase-reversible hydrogels [10– 12]. All of these systems utilize glucose oxygenase or Con-A as a glucose-sensing moiety. Since they are proteins, their stability is rather poor for long-term in vivo applications. Con-A is also known to be immunotoxic [13, 14] and for this reason, routine application in implantable devices may be prevented. Since at present there are no glucose-specific molecules other than several proteins (enzymes and lectins), we were interested in the synthesis of glucose-recognizing molecules that are biocompatible enough for repeated in vivo applications. For the rational design of glucose-specific molecules, we conducted a comparative stereochemical analysis of specific interactions between glucose and five glucose-binding proteins, such as human β-cell glucokinase, d-xylose isomerase, lectins (Lathyrus ochrus isolectin I and Con-A), and glucose/ galactose-binding protein [15]. The analysis revealed that the main interaction providing glucose specificity was hydrogen bonding between amino acids of the proteins and the hydroxyl groups of glucose. The most common amino acid residues involved in the hydrogen bonds were Asp, Glu, and Asn. Almost every hydroxyl group of the glucose molecule had at least one hydrogen bond with amino acid residues or water molecules. It was very common for each hydroxyl group to form multiple hydrogen bonds with many amino acid residues. Certain amino acid residues could form multiple hydrogen bonds with different hydroxyl groups of the glucose molecule. This network of hydrogen bonds in the right spatial arrangement was expected to provide glucose specificity. Hydrophobic interaction between the pyranose ring of glucose and aromatic rings of hydrophobic amino acid residues, such as Phe and Trp, also played an important role in the glucose specificity. This indicated that a certain spatial arrangement of those amino acid residues or their derivatives would result in glucose-specific binding sites. One approach to make such a spatially oriented amino acid derivatives is molecular imprinting. Molecularly imprinted polymers (MIPs) are probably the most promising materials in the field of artificially generated molecular recognition [16– 18]. The most common method of molecular imprinting is the polymerization of functional monomers and a cross-linking agent in the presence of a target molecule used as a template [19, 20]. Thermally or photochemically initiated polymerization results in a highly cross-linked insoluble polymer. Subsequent removal of the print molecules by extraction or hydrolysis leaves recognition sites that are complementary in size, shape, and chemical functionality to the template molecule. Molecular imprinting was also performed in aqueous solution using a polymer and cross-linkers rather than functional monomers [21]. MIPs have already been applied in several fields, such as ligand-binding assays [22], chromatographic separations of stereoisomers by using columns packed with MIPs [23, 24], selective sample enrichment by solid-phase extraction [25], and biomimetic sensors [26, 27]. In this work, we have prepared MIPs that recognize glucose as the first step towards preparing a modulated insulin delivery system that utilizes a synthetic glucose-binding mole-
Glucose binding to molecularly imprinted polymers
315
cule. According to the analysis of glucose-binding sites of several proteins [15], the most common and essential amino acid residues involved in interactions with glucose are Asp and Asn for hydrogen bonding and Phe for hydrophobic interaction. For this reason, MIP systems were prepared using functional monomers that have functional similarities to the amino acid residues. EXPERIMENTAL
Materials The monomers used were vinyl acetic acid (VAA), acrylamide (AAm), allyl benzene (AB), and 4-pentenoic acid (PA). All of the monomers were purchased from Aldrich Chemical Co. (Milwaukee, WI, USA) and purified before use by sublimation under vacuum. N,N -Methylenebisacrylamide (MBAAm), a crosslinking agent purchased from Aldrich Chemical Co., was used as received. 2,2 Azobisisobutyronitrile (AIBN) was purchased from Junsei Chemical Co., Ltd. (Tokyo, Japan). All solvents were of analytical grade quality. Polymer preparations Glucose MIPs were synthesized by either free radical solution polymerization or UV polymerization. The composition of the monomers, cross-linker (MBAAm), template molecule (glucose), and AIBN for polymers prepared by free radical polymerization (polymers P1– P3) is shown in Table 1. The predetermined amounts of glucose and monomers were dissolved in 20 ml of dimethyl sulfoxide (DMSO) until a homogeneous solution was obtained. MBAAm and AIBN were dissolved in this solution. The solution was purged with nitrogen to remove oxygen, which acts as a free radical scavenger, and polymerized under a nitrogen atmosphere at 60 ◦ C for 4 h. For UV polymerization, the predetermined amounts of glucose and monomers were dissolved in 20 ml of DMSO. After MBAAm and AIBN were added, the solution was purged with nitrogen and polymerized under a UV source (366 nm) at 0 ◦ C for 12 h. The bulk polymers were washed with excess DMSO to remove any unreacted monomers, filtered, and dried in vacuum at 40 ◦ C for 18 h. The dried polymers were ground to particles of 50 µm diameter or smaller and the particles were separated by sedimentation from ethanol. Control polymers without glucose imprinting were prepared at the same time under identical conditions. Glucose extraction Glucose molecules used in imprinting were extracted by washing the MIPs with deionized distilled water (DDW). MIPs (0.5 g) were added to 10 ml of DDW and the solution was gently stirred at 25 ◦ C for 24 h. After decanting the DDW, the MIPs were dried under vacuum at 40 ◦ C for 18 h. Glucose extracted into DDW was assayed by the phenol– sulfuric acid assay method [28]. Glucose extractions were repeated until at least 97% of the imprinted glucose was removed from the MIPs.
316
H. Seong et al.
Swelling studies The dried MIPs (0.1 g) were placed into 20 ml of phosphate-buffered saline (PBS, pH 7.4) and the solution was kept at 25 ◦ C. Polymer samples were taken out of the buffer at timed intervals and the excess of buffer on the samples was removed. The weights of the swollen polymers were measured. The swelling ratio (S) was calculated by dividing the weight of the wet polymer by that of the dried polymer. Evaluation of glucose-binding affinity to MIPs Equilibrium dialysis between MIPs and glucose was run at 25 ± 0.5 ◦ C at pH 7.4, using a multi-sample microvolume dialyzer (EDM101B Equilibrium Dialyzer, Hoefer Pharmacia Biotech Inc., San Francisco, CA, USA). The dialyzer system consists of an eight-well dialysis module and a dialysis membrane with a molecular weight cut-off of 6000– 8000. The membrane divided each well into two chambers of equal volume (0.5 ml). The module was attached to a dialysis mixer. Under gentle mixing, the glucose diffused into the chamber containing MIPs or non-imprinted control polymer samples. 0.5 ml aliquots of glucose solution (at concentrations of Table 1. Polymerization conditions of molecularly imprinted polymersa Polymers
Monomers (mM) VA
AAm
PA
AB
Cross-linker (mM) MBAAm
Template (mM) α-D-glucose
P1
1– 4 5– 8 9 10 11
2.0 2.0 4.0 1.0 1.0
2.0 2.0 1.0 4.0 1.0
— — — — —
2.0 2.0 1.0 1.0 4.0
15, 30, 45, 60 60 60 60 60
2.0 0.5, 1.0, 1.5, 2.0 2.0 2.0 2.0
P2
1– 4 5– 8 9 10 11
2.0 2.0 4.0 1.0 1.0
2.0 2.0 1.0 4.0 1.0
2.0 2.0 1.0 1.0 4.0
— — — — —
15, 30, 45, 60 60 60 60 60
2.0 0.5, 1.0, 1.5, 2.0 2.0 2.0 2.0
P3
1– 4 5– 8 9 10 11 12
2.0 2.0 3.5 1.5 1.5 1.5
2.0 2.0 1.5 3.5 1.5 1.5
2.0 2.0 1.5 1.5 3.5 1.5
2.0 2.0 1.5 1.5 1.5 3.5
20, 40, 60, 80 80 80 80 80 80
2.7 0.7, 1.4, 2.0, 2.7 2.7 2.7 2.7 2.7
a Polymerization was initiated by heating at 60 ◦ C for 4 h. The initiator azobis(isobutyronitrile) was added to make 1 wt% with respect to the total amount of monomers and cross-linker. VAA = vinyl acetic acid; AAm = acrylamide; PA = 4-pentenoic acid; AB = allyl benzene; MBAAm = N,N -methylene bisacrylamide.
Glucose binding to molecularly imprinted polymers
317
25– 200 µg/ ml PBS) were placed in the donor chambers and 10 mg of the MIPs or non-imprinted control polymer samples were loaded into the receptor chambers. After loading PBS into the receptor chambers, the dialysis module was gently rotated at a speed of 20 rpm at 25 ± 0.5 ◦ C. After reaching equilibrium (after 12 h), polymer samples were removed from the receptor chamber and dried in vacuum. Dried polymer samples were added to 2 ml of DDW and the solution was stirred at 25 ◦ C for 24 h to extract polymer-bound glucose molecules. The extracted glucose was assayed by the phenol– sulfuric acid assay method at 485 nm using a Beckerman DU-7 spectrophotometer [28]. RESULTS AND DISCUSSION
The binding affinity of MIPs is dependent on the stability of complexes between the template molecule and the functional monomers in the reaction mixture, as well as on preservation of the stability in the resulting polymers [29, 30]. Fixation of the template molecules within the selective cavities could be achieved either by utilizing non-covalent interactions, such as hydrogen bonding and ion-pair interaction (for non-covalent imprinting), or by reversible covalent interactions (for covalent imprinting) between the template molecule and the functional monomers. In the case of covalent imprinting, there is a stoichiometric relationship between the template molecule and the imprinted binding sites. In general, covalent imprinting is considered to be a less flexible method, since the interactions between the template molecule and functional monomers are limited to rapidly reversible covalent interactions. Non-covalent imprinting has no such restrictions, so the range of compounds that can be imprinted is much larger. However, since the association constant is relatively low, an excess of functional monomer is required to saturate the recognition sites and the removal of the template leaves a heterogeneous population of binding sites (Fig. 1). Though non-covalent imprinting has
Figure 1. Schematic representation of molecular imprinting by non-covalent interaction.
318
H. Seong et al. Table 2. Structural comparison of amino acids with monomers used in the preparation of MIPs Amino acids
Structure
Monomers
Aspartic acid (Asp)
Vinyl acetic acid (VAA)
Glutamic acid (Glu)
4-Pentenoic acid (PA)
Asparagine (Asn)
Acrylamide (Aam)
Phenyl alanine (Phe)
Allyl benzene (AB)
Structure
such a limitation, it can still mimic interactions between glucose and the amino acid residues in glucose-binding proteins, which are non-covalent. Four types of interactions involved in glucose binding are hydrogen bonding, hydrophobic interaction, van der Waals interaction, and ionic coordination [31]. In the design of synthetic glucose-binding polymers, monomers that can interact with glucose by either hydrogen bonding or hydrophobic interaction were the first choices. Monomers that have the same functional groups as those of amino acids were chosen to prepare the MIPs (Table 2). The molar ratio of cross-linker (MBAAm) to total functional monomers was varied from 2.5 : 1 to 10 : 1 for optimization of the physical stability of MIPs. The ratio of total functional monomers to glucose was in the range of 3.0 : 1 to 12.0 : 1, as determined by the solubility of glucose. Properties of MIPs Complete extraction of template molecules from MIPs is a prerequisite for the precise evaluation of the binding affinity of MIPs. To investigate the effect of the degree of cross-linking on the glucose extraction, the amounts of glucose extracted from polymer 1 or polymer 4 of the P1, P2, and P3 series (see Table 1) were measured by the phenol– sulfuric acid assay method. The amounts of glucose remaining in different MIPs after each extraction and drying process are shown in Fig. 2. Although more than 97% of bound glucose molecules were extracted after five consecutive extractions, the amount of glucose extracted from highly
Glucose binding to molecularly imprinted polymers
319
Figure 2. Amount of glucose remaining in MIPs with different cross-linking densities after repeated extraction in deionized distilled water at 25 ◦ C. Each extraction lasted for 24 h (n = 3).
cross-linked polymers (P1-4, P2-4, and P3-4) after each extraction in the early cycles was much smaller than the amounts from the less cross-linked polymers (P1-1, P2-1, and P3-1). This may be because the high degree of cross-linking can restrict the mobility of glucose and/ or of functional monomers in MIPs [20]. When the polymers prepared with the same amount of cross-linker were compared, the amount of glucose remaining in MIPs was in the order P3-4 > P1-4 > P2-4 or P3-1 > P1-1 > P2-1. This indicates that the interaction between glucose and functional monomers is influenced by the composition of the functional monomers. Figure 3 shows the amount of glucose remaining in MIPs after each extraction. MIPs were synthesized using the same cross-linkers, but in the presence of different amounts of glucose. The amount of glucose used during the imprinting process does not appear to affect the glucose binding to the MIPs. For example, no significant differences were observed between P1-5 and P1-8. The amount of glucose remaining in MIPs was in the order P3-8 > P1-8 > P2-8 or P3-5 > P1-5 > P2-5. These results suggest that monomers in P3 interact with glucose with a higher affinity than those in P1 or P2. The swelling behavior of MIPs was examined by measuring the swelling ratios (S) in PBS. The swelling behavior of MIPs prepared with different cross-linker amounts is shown in Fig. 4. While there were no significant differences in the S values among the highly cross-linked MIPs (P1-4, P2-4, and P3-4) irrespective of their monomer compositions, there was a pronounced difference in the S values among MIPs with
320
H. Seong et al.
Figure 3. Amount of glucose remaining in MIPs after repeated extraction in deionized distilled water at 25 ◦ C. Each extraction lasted for 24 h. MIPs were synthesized in the presence of different amounts of glucose as template. The cross-linking densities of the MIPs were the same (n = 3).
Figure 4. Swelling ratios (S) of MIPs having different cross-linking densities. Swelling was measured at 25 ◦ C in PBS (n = 3).
Glucose binding to molecularly imprinted polymers
321
lower cross-linking densities ((P1-1, P2-1, and P3-1). In addition, the degree of swelling was in the order P2 > P1 > P3, indicating that the physical stability of MIPs in aqueous solution was dependent on the degree of cross-linking [30]. Glucose-binding affinity to MIPs The binding affinities between glucose and MIPs were determined by equilibrium dialysis techniques. Each MIP was dialyzed against glucose and the amount of glucose bound to the MIP was calculated using the predetermined calibration curve of the glucose solution. Information on the equilibrium, glucose + MIP ↔ glucose– MIP, was obtained using the following Scatchard plot, a tool already applied in MIP work [25, 32]: B/[F] = Bmax /KD − B/KD , where B is the amount of glucose bound to the polymer, [F] is the concentration of free glucose (approximated by the analytical concentration of glucose), KD is the dissociation constant of the glucose– MIP complex, and Bmax is the apparent maximum number of binding sites. Figure 5 shows the amount of glucose bound to P1-4 polymer against free glucose (a) and the Scatchard plot of the data (b) after the equilibrium dialysis test. Linear regression of the Scatchard plot (R 2 = 0.997) gave a KD value of 1.94 mM [association constant (KA ) = 1/KD = 5.13 × 102 M−1 ]. The linearity of the Scatchard plot indicates that the binding sites are identical and independent. Con-A was also dialyzed against glucose to determine the KD value, which was then compared with those of the MIPs. The KD value of Con-A was found to be 1.57 mM (KA = 6.35 × 102 M−1 ), which is in agreement with the value reported in the literature [33]. The KD or KA value of P1-4 suggests that MIP has glucose-binding affinity and thus, MIPs having higher glucose-binding affinity than Con-A could be achieved by an optimized composition of monomers and cross-linker. On the other hand, the KD values of glucose-non-imprinted polymers were in the range of 32.6– 49.1 m M (KA = 30.7– 20.4 M−1 ) and no relationship between KD and the polymerization conditions was found. The results of Scatchard analyses for MIPs prepared using different amounts of cross-linker or glucose are summarized in Table 3. With an increase in the cross-linker amount (1– 4 of P1, P2, and P3 polymers), the KD value decreased in all series of polymers, indicating that glucose bound with higher affinity to MIPs with a higher cross-linking density at the fixed ratio of monomers. Once glucose-binding sites are formed in MIPs by non-covalent interactions between glucose and functional monomers, the stability of the binding sites would determine the subsequent binding affinity of MIPs. The stability of a binding site is dependent on the degree of cross-linking of the functional monomers. Thus, the higher glucose-binding affinity of highly crosslinked MIPs may be due to the increased stability of glucose-binding sites in the MIPs. In addition, the KD values of the MIPs decreased in all series of polymers as the amount of glucose for imprinting was increased (5– 8 of P1, P2, and P3
322
H. Seong et al.
Figure 5. Amount of glucose bound to P1-4 at pH 7.4 (a) and the Scatchard plot for the bound glucose (b). B is the amount of glucose bound to polymer and [F] is the concentration of free glucose (n = 3).
Glucose binding to molecularly imprinted polymers
323
Table 3. Dissociation constants (KD ) of MIPs having different amounts of cross-linkera or glucose as a template moleculeb Polymer
KD ± SD (mM)
Polymer
KD ± SD (mM)
Polymer
KD ± SD (mM)
P1-1 P1-2 P1-3 P1-4 P1-5 P1-6 P1-7 P1-8
6.48 ± 0.07 4.21 ± 0.09 2.93 ± 0.11 1.94 ± 0.21 11.59 ± 0.01 8.24 ± 0.09 2.49 ± 0.01 1.94 ± 0.21
P2-1 P2-2 P2-3 P2-4 P2-5 P2-6 P2-7 P2-8
8.09 ± 0.38 5.93 ± 0.17 3.37 ± 0.04 2.32 ± 0.29 13.05 ± 0.05 10.37 ± 0.22 5.29 ± 0.06 2.32 ± 0.29
P3-1 P3-2 P3-3 P3-4 P3-5 P3-6 P3-7 P3-8
4.27 ± 1.31 2.84 ± 0.55 2.14 ± 0.19 1.66 ± 0.03 9.31 ± 0.02 6.64 ± 0.06 2.05 ± 0.16 1.66 ± 0.03
a 1– 4 b 5– 8
of P1, P2, and P3. of P1, P2, and P3.
polymers). In these polymers, the same amounts of functional monomers and crosslinker were used. Therefore, the higher glucose-binding affinity by a higher glucose concentration is most likely due to the fact that more functional monomers would participate in the formation of a binding site. When the KD values were compared with respect to the monomer composition, P1 polymers composed of VA, AAm, and AB had lower KD values than those of P2 polymers composed of VA, AAm, and PA. In addition, P3 polymers composed of VA, AAm, PA, and AB had lower KD values than those of P1 polymers. These results support the finding that interactions between glucose and functional monomers can be achieved by hydrogen bonding (P2) and that additional hydrophobic interaction can increase the glucose-binding affinity of MIPs. To compare the contribution of each monomer to glucose binding, MIPs having different monomer ratios (9– 11 of P1 and P2 polymers, or 9– 12 of P3 polymers) were dialyzed against glucose and the data were analyzed by a Scatchard plot. The results are summarized in Table 4. When the data of 9– 11 in P1 polymers were compared, the contribution of functional monomers to the glucose binding affinity was in the order VAA > AAm > AB. This implies that hydrogen bonding of the carboxyl group in VAA is stronger than that of the amine group in AAm and also that the hydrophobic interaction of AB is weaker than the hydrogen bonding of VAA or AAm. The KD values of 9– 11 in P2 polymers were in the range of 2.3– 4.1 mM, lower than those of P1 polymers. In addition, the highest KD value of P2-11 polymer, representing the lowest glucose-binding affinity, indicates that PA has weaker hydrogen bonding than that of VAA or AAm. As can be seen in Table 4, the KD values of 9– 12 of P3 polymers were in the range of 2.0– 2.9 m M. The lowest KD value of P3-9 indicates that the carboxyl group of VAA serves as the main functional group for glucose binding. In addition, the relatively low KD value of P3-12 suggests that hydrophobic interaction of AB with glucose can increase the glucose-binding affinity of MIPs which already have glucose-binding affinity by hydrogen bonding [34].
324
H. Seong et al.
Table 4. Dissociation constants (KD ) of MIPs having different molar ratios of monomers Polymer
KD ± SD (mM)
Polymer
KD ± SD (mM)
Polymer
KD ± SD (mM)
P1-9 P1-10 P1-11 —
1.99 ± 0.03 2.24 ± 0.01 3.58 ± 0.22 —
P2-9 P2-10 P2-11 —
2.26 ± 0.09 2.95 ± 0.02 4.11 ± 0.06 —
P3-9 P3-10 P3-11 P3-12
1.99 ± 0.11 2.23 ± 0.06 2.84 ± 0.01 2.35 ± 0.05
Table 5. Dissociation constants (KD ) of MIPs prepared by UV polymerization and thermal polymerization Polymer
KD ± SD (mM)a
KD ± SD (mM)b
P1-8 P2-9 P3-8
2.49 ± 0.06 3.83 ± 0.09 2.72 ± 0.01
1.94 ± 0.21 2.26 ± 0.09 1.66 ± 0.03
aK bK
D D
of UV-polymerized polymers. of thermally polymerized polymers.
UV polymerization is preferred since it has been demonstrated that polymers made at low temperature exhibit higher recognition abilities [35]. Therefore, the polymers having the highest glucose-binding affinity in each polymer series (P1-8, P2-9, and P3-8) were also prepared by UV polymerization at 0 ◦ C. The KD values of UV-polymerized polymers and those of thermally polymerized polymers are summarized in Table 5. In general, it is believed that weak non-covalent interactions, such as hydrogen bonding, essential for imprint formation and subsequent recognition, are stronger at lower temperatures due to a favorable entropy term [36]. However, the KD values of UV-polymerized polymers were higher than those of thermally polymerized polymers. The lower binding affinity of UV-polymerized MIPs may be explained by the lower degree of cross-linking, because of less initiation of AIBN at low temperature.
CONCLUSIONS
MIPs having glucose-binding affinity can be synthesized by copolymerization of mixtures of amino acid-mimicking functional monomers, excess cross-linker, and glucose as a template. Glucose-binding affinity evaluated by the Scatchard analysis was dependent on the composition of the functional monomers, as well as on the degree of cross-linking. The MIPs composed of vinyl acetic acid (VAA), acryl amide (AAm), 4-pentenoic acid (PA), and allyl benzene (AB) had the lowest KD value of 1.66 mM, which is comparable to the KD value of Con-A (1.84 mM). Furthermore, the results of the glucose-binding affinity of MIPs having different
Glucose binding to molecularly imprinted polymers
325
molar ratios of functional monomers demonstrate that the strength of interaction with glucose is in the order VAA > AAm > AB > PA. Acknowledgement This study was supported in part by the National Institute of Health through grant DK54164.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.
J. Heller, J. Control. Rel. 8, 111 (1988). K. Ishihara, M. Kobayashi and I. Shinohara, Makromol. Chem., Rapid Commun. 4, 327 (1983). M. Casolaro and R. Barbucci, Int. J. Artif. Organs 14, 732 (1991). F. Fischel-Ghodsian and J. M. Newton, J. Membr. Sci. 48, 33 (1990). S. Kitano, Y. Koyama, K. Kataoka, T. Okano and Y. Sakurai, J. Control. Rel. 19, 162 (1992). D. Shiino, Y. Murata, K. Kataoka, Y. Koyama, M. Yokoyama, T. Okano and Y. Sakurai, Biomaterials 15, 121 (1994). Y. Ito, D. J. Chung and Y. Imanishi, Bioconjugate Chem. 5, 84 (1994). M. Brownlee and A. Cerami, Science 206, 1190 (1979). C. M. Pai, Y. H. Bae, E. J. Mack, D. E. Wilson and S. W. Kim, J. Pharm. Sci. 81, 532 (1992). S. J. Lee and K. Park, J. Mol. Recognition 9, 549 (1996). A. A. Obaidat and K. Park, Pharm. Res. 13, 989 (1996). A. A. Obaidat and K. Park, Biomaterials 18, 801 (1997). W. H. Beckert and M. M. Sharkey, Int. Arch. Allergy Appl. Immunol. 39, 337 (1970). A. E. Powell and M. A. Leon, Exp. Cell Res. 62, 315 (1970). T. Li, H. Lee and K. Park, J. Biomater. Sci. Polymer Edn 9, 327 (1998). D. Kriz, C. B. Kriz, L. I. Anderson and K. Mosbach, Anal. Chem. 66, 2636 (1994). K. Mosbach, Trends Biochem. Sci. 19, 9 (1994). K. J. Shea, D. A. Spivak and B. Sellergren, J. Am. Chem. Soc. 115, 3368 (1993). G. Wulff and H. Haarer, Makromol. Chem. 192, 1329 (1991). G. Wulff and S. Schauhoff, J. Org. Chem. 56, 395 (1991). W. J. Wizeman and P. Kofinas, Biomaterials 22, 1485 (2001). G. Vlatakis, L. I. Anderson, R. Muller and K. Mosbach, Nature 361, 645 (1993). J. M. Lin, K. Uchiyama and T. Hobo, Chromatographia 47, 625 (1998). M. Kempe, Anal. Chem. 68, 1948 (1996). M. T. Muldoon and L. H. Stanker, Anal. Chem. 69, 803 (1997). C. Malitesta, I. Losito and P. G. Zambonin, Anal. Chem. 71, 1366 (1999). D. Kriz, O. Ramström and K. Mosbach, Anal. Chem. 69, 345A (1997). M. Dubois, K. A. Gilles, J. K. Hamilton, P. A. Rebers and F. Smith, Anal. Chem. 28, 350 (1956). B. Sellergren, Makromol. Chem. 190, 2703 (1989). S. A. Piletsky, E. V. Piletskaya, T. L. Pannasyuk, A. V. El’skaya, R. Levi, I. Karubo and G. Wulff, Macromolecules 31, 2137 (1998). D. J. Cram, Angew. Chem. Int. Ed. Engl. 25, 1039 (1986). J. Matsui, Y. Miyoshi, O. Doblhoff-Dier and T. Takeuchi, Anal. Chem. 67, 4404 (1995). L. L. So and I. J. Goldstein, Biochim. Biophys. Acta 165, 398 (1968). R. D. Whitaker, Y. Cho, J. Cha, H. L. Carrell, J. P. Glusker, P. A. Karplus and C. A. Batt, J. Biol. Chem. 270, 22895 (1995). D. J. O’Shannessy, B. Ekberg, L. I. Anderson and K. Mosbach, J. Chromatogr. 470, 391 (1989). P. Turkewitsch, B. Wandelt, G. D. Darling and W. S. Powell, Anal. Chem. 70, 2025 (1998).
This page intentionally left blank
The effect of oxidation on the enzyme-catalyzed hydrolytic biodegradation of poly(urethane)s ROSALIND S. LABOW 1,∗ , YIWEN TANG 2 , CHRISTOPHER B. MCCLOSKEY 2 and J. PAUL SANTERRE 2 1 University of Ottawa Heart Institute, 40 Ruskin Street, Ottawa, ON, K1Y 4W7, Canada 2 Faculty of Dentistry, University of Toronto, 124 Edward Street, Toronto, ON, M5G 1G6,
Canada
Received 30 November 2001; accepted 1 April 2002 Abstract—Although the biodegradation of polyurethanes (PU) by oxidative and hydrolytic agents has been studied extensively, few investigations have reported on the combination of their effects. Since neutrophils (PMN) arrive at an implanted device first and release HOCl, followed by monocytederived macrophages (MDM) which have potent esterase activities and oxidants of their own, the combined effect of oxidative and hydrolytic degradation on radiolabeled polycarbonate-polyurethanes (PCNU)s was investigated and compared to that of a polyester-PU (PESU) and a polyether-PU (PEU). The PCNUs were synthesized with PCN (MW = 1000), and butanediol (14 C-BD) and one of two diisocyanates, hexane-1,6-diisocyanate (14 C-HDI) or methylene bis-p-phenyl diisocyanate (MDI). The PESU and PEU were synthesized using toluene-diisocyanate (14 C-TDI), with polycaprolactone and polytetramethylene oxide as soft segments respectively, and ethylene diamine as the chain extender. The effect of pre-treatment with 0.1 mM HOCl for 1 week on the HDI-based PCNUs and both TDI-based PUs resulted in a significant inhibition of radiolabel release (RR) elicited by cholesterol esterase (CE), when compared to buffer alone, whereas the MDI-based PCNU showed a small but significant increase. When PMN were activated on the HDI-based PCNU surface with phorbol myristate acetate (PMA), HOCl was released for 3 h, and was almost completely abolished by sodium azide (AZ). Simultaneously, the PMN-elicited RR, shown previously to be due to the esterolytic cleavage by serine proteases, was inhibited ∼75% by PMA-activation of the cells, but significantly increased relative to the latter when AZ was added. Both in vitro oxidation by HOCl and the release of HOCl by PMN were associated with the inhibition of RR and suggest perturbations between oxidative and hydrolytic mechanisms of biodegradation. Key words: Polyurethanes; biodegradation; oxidation; hydrolysis; neutrophils.
∗ To
whom correspondence should be addressed. Phone: (613) 761-4010. Fax: (613) 761-5035. E-mail:
[email protected]
328
R. S. Labow et al.
INTRODUCTION
Polycarbonate-based polyurethanes (PCNUs) are materials that are being promoted as having a lower susceptibility to hydrolysis relative to polyester-urethanes (PESU) and lower oxidation relative to polyether-urethanes (PEU) [1]. Biodegradation, which includes environmental stress cracking (ESC), is still observed clinically on long-term implanted polyurethane (PU) devices. The process is postulated to involve multiple components, of which the chemical interactions between the biomaterial and biological components are not clearly understood. Oxidation is hypothesized to be an important parameter in the process of ESC in vivo, since the end result of the latter process can be simulated in vitro on the surface of materials using oxidizing reagents such as hydrogen peroxide and hypochlorous acid (HOCl) [1]. Although monocyte-derived macrophages (MDM) eventually become the most abundant cell type at the material interface of implanted devices and release reactive oxygen species when activated during the inflammatory response, neutrophils (PMN) are the first to arrive during the early hours following implantation and may be recruited for the first few days following surgery [2]. PMN release one of the strongest oxidative biological compounds yet identified in these cells, HOCl [3]. Although the steroid dexamethasone, which inhibits the respiratory burst, was able to reduce the effect of MDMs on a polyetherurethane [4] and thus imply a role for oxidants in ESC, this agent will also shut down the remaining events in the inflammatory response, i.e. the release of lysosomal and other intracellular enzymes [5] which may be responsible for hydrolytic degradation. Hence, the MDM agents that are effectively responsible for the chemical breakdown of the PUs are still being elucidated. Previous studies have suggested that MDM are capable of promoting hydrolytic degradation of PU in addition to the potential degradation due to the release of reactive oxygen species [1, 6, 7]. Using a trypsinized activated mature MDM cell system, degradation of a 14 C-labeled-PESU [8] as well as a 14 C-labeled PCNU [9], has been demonstrated by measuring radiolabel release into the cell supernatant. In parallel to these findings, it was found that a significant increase in esterase activity was measured not only in the MDM after differentiation [10], but continued to increase while the activated MDM remained in culture on the PESU [9] and PCNU [10] surfaces. In several enzymatic and cell studies involving the hydrolytic degradation of PUs, esterolytic activities have been shown to be involved. Thus far, when commercial cholesterol esterase (CE) was incubated with several polyurethanes it has been shown to be the most destructive [11]. These include polyester- [12– 14] and polyether-PUs [15], and dental restorative composite materials [16]. The polyester PUs were primarily cleaved at the ester bonds in the soft segment and the products were identified as oligomers of TDI [17]. CE also cleaved polyether PUs at the most probable bond susceptible to hydrolytic cleavage, which is the urethane bond, resulting in the release of the free amine (toluene diamine) [18, 19]. Recently, PCNUs have been shown to be susceptible to cleavage by CE as well [20, 21]. Both
The effect of oxidation on hydrolysis of polyurethanes
329
the carbonate and urethane bonds were cleaved resulting in many products ranging in molecular weights from ∼150– 850 which were identified by GC-MS [22]. The degree of susceptibility varied considerably with the ratio of hard to soft segment as well as with different diisocyanates [20, 21]. It was possible to determine the susceptibility of each of the PCNUs to hydrolytic cleavage by CE and relate the extent of degradation to the amount of hard segment interaction within the polymer and at the surface. More specifically, the degree of phase separation and soft segment crystallinity were found to be less important in comparison to the hydrogen bonding among the carbonate and urethane linkages [21]. The rank of the different chemical groups’ susceptibility to hydrolysis was as follows: nonhydrogen bonded carbonate > non-hydrogen bonded urethane > hydrogen bonded carbonate > hydrogen bonded urethane. There is a strong belief that it is a synergistic effect between hydrolytic and oxidative activation pathways which are involved in the biodegradation processes leading to the clinically observed phenomenon of ESC [23]. Since PCNUs have been reported to be more stable to oxidation than PEU and more stable to hydrolysis than PESU, the combined effect of oxidation on the subsequent CE-mediated hydrolysis was of interest with respect to these PUs. This study attempted to partly simulate the biochemical effects of the in vivo condition, by pre-treating PUs with HOCl, the oxidative compound released from PMN, followed by incubation with CE which is secreted by MDM [24]. Although PMN have been shown previously to release serine proteases which are capable of hydrolytic cleavage of PUs, their degradative potential is much less than that of MDM [8]. In this study, an in vitro PMN cell system was used to measure HOCl and radiolabel release simultaneously, in order to determine if the oxidative effect of PMN had an impact on the subsequent hydrolytic biodegradation.
MATERIALS AND METHODS
Materials Unless otherwise specified all reagents were purchased from the Sigma Chemical Company, St. Louis, MO. Polymer synthesis The diisocyanates used for the synthesis of the PUs were toluene diisocyanate (TDI), 1,6-hexane diisocyanate (HDI, Aldrich, Milwaukee, WI, USA) and 4,4 methylene bis-phenyl diisocyanate (MDI, Aldrich, Milwaukee, WI, USA). The same chain extender and soft segment were used for all the PCNUs and were 1,4-butanediol (BD, Aldrich, Milwaukee, WI, USA) and poly(1,6-hexyl 1,2-ethyl carbonate) diol (PCN, MW = 1000, received in kind from Corvita Corporation, Miami, FL, USA), respectively. The chain extender for the polycaprolactone
330
R. S. Labow et al.
Table 1. Polyurethanes synthesized Reagent
Stoichiometry
Acronym
MW
CPM/ 100 mg
MDI/ PCN/BD HDI/ PCN/BD HDI/ PCN/BD HDI/ PCN/BD HDI/ PCN/BD HDI/ PCN/BD TDI/PCL/ED TDI/PTMO/ ED
3:2:1 3:2:1 3:2:1 2:1:1 3:1:2 4:3:1 2 : 1 : 2.2 2:1:2
a MDI-321B
10 × 104 10 × 104 10 × 104 3.9 × 103 3.8 × 103 11 × 104 11 × 104 18 × 104
2.8 × 105 4.7 × 105 11 × 105 10 × 105 9.4 × 105 12 × 105 7.0 × 105 9.1 × 105
a HDI-321B b HDI-321 b HDI-211 b HDI-312 b HDI-431 c TDI/PCL/ED c TDI/PTMO/ ED
Abbreviations: Diisocyanate: Methylene bis-p-phenyl diisocyanate (MDI), 1,6-Hexane diisocyanate (HDI), toluene diisocyanate (TDI), Macroglycoldiol soft segment: Poly(1,6-hexyl 1,2-ethyl carbonate) (PCN) (MW 1000); polyether diol (PTMO) (MW 1000), polycaprolactone diol (PCL) (MW 1250) Diol chain extender: 1,4-butanediol (BD) (MW 90). a 14 C-BD used in the synthesis. b 14 C-HDI used in the synthesis. c 14 C-TDI used in the synthesis.
(PCL) (polyester) and polytetramethylene oxide (PTMO) (polyether) based PUs was ethylene diamine (ED). Table 1 lists the stoichiometry of the reagents used in the PU synthesis, the specific radioactivities and the location of the radiolabel, as well as the acronyms to which the polymers are referred. The 14 C-labeled PUs used in the biodegradation experiments were synthesized with either 14 C-BD, 14 C-HDI, 14 C-TDI (custom synthesis from NEN DuPont, Mississauga, Ontario). The details of the synthesis and characterization of the polymers were previously described [13, 14, 20, 21]. The specific radioactivity for each polymer was calculated by dissolving 1.0 mg of each polymer in dimethylacetamide (DMAC) (BDH Chemical, Mississauga, ON) (1.0 ml) and counting the solution in 10 ml of Formula 989 liquid scintillation cocktail (Packard Instruments, Inc., Meriden, CT) in a liquid scintillation conter (LKB Rackbeta, Gaithersburg, PA). The release of radioactivity (CPM) into incubation solutions from the PU substrate has proven to be a reliable and sensitive method for measuring the degree of material degradation [9, 13– 15, 18– 20, 25]. The radioactive products have been isolated and identified and related to the surface damage due to chemical bond cleavage [17]. The preparation of polymer samples was carried out using solutions of 10% (wt/vol) PU in DMAC. Any non-soluble material in the solution was removed from the polymer solutions by passing them through a 0.45 µm Teflon® filter (Chromatographic Specialties, Toronto, ON). Samples incubated in cell culture plates consisted of round glass coverslips (15 mm diameter) (Fisher Scientific, Ottawa, ON) coated with the PCNU solutions (100 µl) under sterile conditions in a laminar flow hood and dried overnight at 50 ◦ C, followed by purging at 50 ◦ C without vacuum for 24 h and then under vacuum for 72 h as described in detail previously [13]. The experiments with the polyester and polyether PUs were
The effect of oxidation on hydrolysis of polyurethanes
331
carried out using hollow glass tubes coated with polymer and incubated in sealed vacutainers using an established method previously reported on [13– 15, 20, 21]. Assay of HOCl Hypochlorous acid (HOCl) (0.5 mM) solutions were prepared by adding 168 µl of sodium hypochorite (NaOCl) (6%) (VWR Scientific, VW 3248-1) to 25 ml of 0.05 M sodium phosphate buffer, pH 7.0 and diluting 1 : 10, giving a concentration of 0.5 mM. In order to determine the amount of HOCl consumed by the PUs or that was produced by the PMN, a standard curve was prepared by adding 0.01– 0.5 ml of the 0.5 mM solution of HOCl, with a final volume of 1.0 ml in water. The colorimetric assay was a modification of the procedure of Weiss et al. [26]. The following reagents were added to these different concentrations of HOCl: 0.5 ml of 0.05 M sodium phosphate buffer, pH 7.4, 0.15 ml of a taurine solution (0.2 g of taurine (f.w. 125.1) (Sigma T0625) into 25 ml of distilled deionized water) and 0.25 ml of a potassium iodide solution (1.99 g of potassium iodide (f.w. 166) (Sigma P8256) in 100 ml of distilled deionized water). Each sample was read at 350 nm in a spectrophotometer (Beckman Instruments). Assay of cholesterol esterase activity CE (Genzyme Diagnostics, Cambridge, MA) (0.003 g) was dissolved in 10.0 ml of 0.05 M sodium phosphate buffer and diluted before assaying according to the procedure of Moore et al. [27]. A 0.05 M sodium phosphate buffer, pH 7.0 (1.4 ml) and p-nitrophenylbutyrate (PNB) solution (4 mM in acetonitrile, 50 µl) were mixed with 25 µl of CE solution diluted 1 : 100. This solution was incubated at 37 ◦ C for 30 min and read at 400 nm. One unit of activity was defined as the release of 1 nmol of p-nitrophenol (ε = 16 300 cm−1 M−1 ) per minute at 37 ◦ C. Polyurethane degradation experiments All biodegradation experiments using a pre-incubation of HOCl utilized a 0.1 mM solution of the oxidative agent. The PU samples were pre-treated with either 0.1 mM HOCl solution or 0.05 M sodium phosphate buffer, pH 7.0 and incubated at 37 ◦ C. The HOCl solution (prepared fresh daily) was replaced every day for 5 days (a total of 5.0 ml). After 1 week, the pre-incubation solution was removed and replaced with either CE solution (100 units/ ml in sodium phosphate buffer, pH 7.0) or buffer and incubated for either 24 h or extended periods up to 18 weeks as described previously [12, 13]. The radiolabel release (0.8 ml of solution) was counted in 10 ml of Formula 989 liquid scintillation cocktail and counted in a liquid scintillation counter. The effect of HOCl or buffer pre-treatment was assessed by comparing data to PU-coated samples which had not been pre-treated but were incubated with buffer and CE solutions for the 24 h period. For comparative purposes the radiolabel release data among the PCNUs were normalized to HDI431 based on the
332
R. S. Labow et al.
specific radioactivity values for each polymer in Table 1 (e.g. HDI431 had a specific radioactivity of 12 × 105 CPM /100 mg and MDI 321’s was 2.8 × 105 CPM /100 mg. The value for the CPM for the latter polymer was multiplied by 4.8). Isolation of human neutrophils Human PMNs were isolated from whole blood using a density gradient centrifugation procedure as described in detail previously [25]. The freshly isolated PMNs were resuspended in DPBS (Dulbecco’s phosphate buffered saline) and seeded onto 14 C-HDI-431-coated glass slips (4 × 106 cells/ ml) or HDI-431 without a radiolabel for measurement of HOCl (2 × 106 cells/ ml) and allowed to adhere for 1 h. Following 1 h the DPBS was removed and the following reagents were dissolved in DPBS and added (phorbol myristate acetate (PMA), 10−7 M or sodium azide (AZ) 5×10−3 M or both) to the adherent cells which were then incubated at 37 ◦ C for 24 h. The use of PMA and AZ at these defined concentrations was established based on a previous report [28]. The influence of the cells on radiolabel release, i.e. material breakdown and release, was determined by counting the cell supernatant in a liquid scintillation counter. The HOCl released into the cell supernatant was determined using the assay described above with the following change, 0.18 g/ 100 ml of glucose was included with the DPBS according to the procedure of Schacter et al. [3]. Both the radiolabel release data and the HOCl production were related to the number of live adherent cells, determined by lysing the cells and assaying the lysate for lactic acid dehydrogenase (LDH) activity as described in detail previously [29]. Briefly, the lysate was added to pyruvate and the remaining pyruvate determined colorimetrically by the absorbance of the 2,4-dinitrophenylhydrazone derivative using a kit (Sigma Chemical Co., St. Louis, MO). A standard curve was generated by plotting the number of PMN seeded (103 – 106 cells, determined by counting the cells seeded in each well using a Baker cell counter (Serono Instruments)) against the LDH activity in cells following lysis with 0.05% Triton X-100. The correlation coefficient for the curve was r 2 = 0.993. Statistical analysis The data were analyzed by a 1 way ANOVA using a program written in Lotus 1-2-3. A significant difference was defined at p < 0.05.
RESULTS
Following contact with the PU samples, the disappearance of HOCl was monitored spectrophotometrically as described above and compared to samples not coated with PUs. Approximately 10% of the HOCl remained in the solutions in contact with glass slip controls after 24 h, however, no HOCl was detected in the PU-containing incubation solutions for the same time frame (data not shown). Based on these
The effect of oxidation on hydrolysis of polyurethanes
333
Figure 1. Effect of HOCl on the cholesterol esterase (CE) hydrolysis of polycarbonate-urethanes (PCNUs): PCNUs-coated slips (described in Table 1) were treated with either 0.1 mM HOCl (closed bars) or 0.05 M sodium phosphate buffer, pH 7.0 (open bars) for 1 week, followed by treatment with CE for 24 h according to the procedure described in MATERIALS and METHODS. The radiolabel release in CPM/ml was plotted for each polymer and each condition.
findings, fresh HOCl was prepared daily and added to the PU samples in order to maintain oxidant activity levels. At the end of the one week incubation period, no HOCl could be measured prior to the addition of CE. From these observations it was concluded that the polymer was continuously consuming the HOCl via oxidation reactions. In this study a relatively simplistic in vitro model for the initial oxidative burst and the subsequent synthesis of esterase activity from the MDM system was carried out by first treating the PCNU surface with 0.1 mM HOCl for a period of 1 week and then with CE (100 units) for 24 h. No difference in radiolabel release after the treatment with either HOCl or buffer (341 ± 16 CPM) was detected prior to the addition of CE. The latter radiolabel count was subtracted from the data presented in Fig. 1 in order to facilitate quantifying the specific contribution to biodegradation from the subsequent CE treatment. The radiolabel release from the incubation of all the PCNU samples for 24 h in CE solution without any pre-treatment was not significantly different from the buffer pre-treated/ CE samples that were recently reported for HDI-431 [9]. When the PCNU-coated slips were treated with HOCl prior to incubation with CE, there was a significant reduction in radiolabel release for all the HDI-based polymers in comparison to control samples exposed to CE after only pre-treatment with buffer (Fig. 1). The CPM released were normalized based on the specific radioactivity values (Table 1) in order to compare the effect of HOCl and CE on each PCNU. There was no significant difference in the radiolabel release or the effect of HOCl when the position of the radiolabel in HDI-321 was changed between the HDI
334
R. S. Labow et al.
or the BD moiety (Fig. 1). The effect of diisocyanate chemistry (i.e. HDI vs MDI) had an important influence on the effect of HOCl treatment and the subsequent radiolabel release generated by CE. The MDI-321B was degraded to a lesser extent than HDI-321B which had a similar stoichiometry (Table 1) (Fig. 1). Moreover, the effect of the pre-treatment by HOCl enhanced the hydrolysis by CE rather than decreasing it as was observed for the HDI polymers. This was the only polymer to show this small but significant increase in degradation following oxidation (Fig. 1) (p < 0.05). The inhibitory effect of HOCl on PU degradation (Fig. 1) was not only observed on the PCNU materials. A biodegradation study using a similar concept to that used for acquiring data in Fig. 1, examined the effect of HOCl on the CE catalyzed degradation for TDI/PTMO /ED and TDI/PCL /ED, polyether and polyester-PUs respectively (Fig. 2). Despite the fact that the degradation study was extended to 18 weeks and that the samples had completely different chemistries than that of the PCNU based materials, it was observed that the HOCl had a significant influence on the polyester based PU (Fig. 2A) (p < 0.05). The effect was less pronounced for the polyether based PU (Fig. 2B) (p < 0.05). Again, there was no significant difference in radiolabel release during the first week of incubation when these PUs were treated with only HOCl or buffer (Fig. 2A). Similar results were obtained with the polyester-PU (Data not shown) (Fig. 2B). In a separate series of experiments, to assess if HOCl generation from PMN influenced the hydrolytic potential of these cells to release radiolabel from the polymers, PMN were seeded on HDI-431-coated glass slips in 4 groups: DPBS (control), PMA (protein kinase C activator), AZ (myeloperoxidase inhibitor) or PMA with AZ. HOCl production (Fig. 3) by the PMN was measured simultaneously with radiolabel release (Fig. 4) as well as cell viability by LDH activity (Fig. 5). Only a very small amount of HOCl could be measured if PMA was not included in the DPBS. At 1 h there was no difference in HOCl production between PMA and PMA with AZ. The rate of HOCl release increased with PMA up until 2 h (∼18 nmol/ h), but the generation of HOCl from the PMN with AZ present along with PMA decreased significantly by that time (∼2 nmol/ h). Between 3 and 24 h the rate of HOCl production from the PMA group had decreased to 0.5 nmol/ h (Fig. 3). In the absence of cells, the reaction of DPBS with HDI-431 gave a reading of zero in the assay. Radiolabel release data (Fig. 4) showed that although PMA elicited the greatest release of HOCl compared to the other 3 groups of PMN, at 24 h there was ∼75% inhibition of radiolabel release observed when compared to the PMN in DPBS. Moreover, at the same time point there was a significant increase of radiolabel release for the AZ treated group when compared to the DPBS control (Fig. 4). As was observed in Fig. 3, the effect of PMA over AZ, when combined together with the PMN, was dominant and produced an overall reduction in radiolabel release when compared with the DPBS control. At 2 h, where HOCl release was peaking
The effect of oxidation on hydrolysis of polyurethanes
335
Figure 2. Effect of HOCl on the cholesterol esterase (CE) hydrolysis of polyether- and polyester urea urethanes. The polyester-urethane (PU) (Fig. 2A) and polyether-PU (Fig. 2B) were treated with either 0.1 mM HOCl (closed circles) or 0.05 M sodium phosphate buffer, pH 7.0 (open circles) for 7 days, followed by treatment with CE for a further 126 days (18 weeks) according to the procedure described in MATERIALS and METHODS. Control samples were treated with either 0.1 m M HOCl (closed triangles) or 0.05 m sodium phosphate buffer, pH 7.0 (open triangles) only. The cumulative radiolabel release in CPM/ml was plotted for each polymer and each condition.
336
R. S. Labow et al.
Figure 3. Effect of phorbol myristate acetate (PMA) and sodium azide (AZ) on neutrophil (PMN)mediated HOCl release: PMNs (2 × 106 ) were seeded onto HDI-431 coated slips in DPBS (closed circles), PMA (open circles), AZ (closed triangles), or PMA plus AZ (open triangles). HOCl production was followed from 1 h to 24 h as described in MATERIALS and METHODS.
(Fig. 3), there were no significant differences in radiolabel release (Fig. 4) between any of the groups. Both radiolabel release and HOCl production were normalized to the number of adherent PMN as determined by LDH activity. The freshly isolated PMN, at the time of seeding onto the HDI-coated glass slips were >95% viable. However, the cells in the 4 groups died at different rates over the course of the 24 h incubation. At 2 h, when HOCl production was at its maximum (see Fig. 3), there was no difference in the viability in the 4 groups as was previously reported [9]. At 24 h, the cells with PMA were almost completely dead whereas the cells with AZ and PMA were >20% viable. The cells in DPBS and AZ were not significantly different at 24 h than from time = 0 (Fig. 5).
DISCUSSION
Since previous studies have implied that esterase activities are most likely involved in the degradation of PU by MDM [9, 10], it was of interest to assess the combined contribution of both hydrolytic and oxidative processes in this degradation. In earlier work with the PCNU polymers, significant differences in their susceptibility to CE hydrolysis was found [20, 21]. These changes were quite pronounced after a
The effect of oxidation on hydrolysis of polyurethanes
337
Figure 4. Effect of phorbol myristate acetate (PMA) and sodium azide (AZ) on neutrophil-mediated radiolabel release: Neutrophils (4×106 ) were seeded on to HDI-431 coated slips and radiolabel release was measured as described in MATERIALS and METHODS at 2 h and 24 h in DPBS (closed bars), PMA (hatched lines right), AZ (hatched lines left), PMA with AZ (crossed lines).
few weeks of incubation with HDI-312 showing high stability early on. Here again, HDI-312 established itself as the most stable of the HDI-based PCNUs. As well, it shows the least amount of change following oxidation (Fig. 1). The availability of CE sensitive cleavage groups in the PCNU polymers was previously shown to be dependent on the chemical nature of the diisocyanate contained in the PCNU [20]. Although phase separation of the polymers may play a role in the biostability, it is the chemical nature of cleavage sites and their immediate structural environment which was reported to allow the diisocyanate components to shield the polycarbonate soft segment from hydrolytic degradation. Among PCNUs with similar reagent stoichiometry, but different diisocyanates, MDI-321 was the least susceptible to hydrolytic cleavage. The most influential factor on the biodegradation of the PCNUs, catalyzed by CE, was demonstrated to be the degree of hydrogen bonding among the carbonate and urethane groups and inter-urethane groups [20, 21]. The enhanced stability of MDI-321 was again demonstrated in this study (Fig. 1). There are two oxidation mechanisms that are suspected of being involved in the degradation process of the PUs by HOCl. These include chlorination of the nitrogen in the urethane bond which would be similar to the chlorination of taurine, a reaction which has been well characterized [26]. The other reaction would be related to the oxidation of the soft segment hydrocarbon chain which may contribute to the
338
R. S. Labow et al.
Figure 5. Effect of phorbol myristate acetate (PMA) and sodium azide (AZ) on neutrophil viability. Neutrophils (2×106 ) were seeded onto HDI-431 coated slips and viability was measured by lactic acid dehydrogenase activity in the Triton X-100 lysate of the adherent cells in DPBS at 24 h as described in MATERIALS and METHODS.
cross-linking of the polymer chains [23]. There has been some documentation in the literature which specifically reports on the susceptibility of diisocyanates and their urethane linkages. Particularly, it is noted that aromatic systems can readily undergo oxidative degradation to yield quinone-imide formation and this results in the loss of the urethane hydrogen on the nitrogen [30]. The latter product is in part responsible for color changes in these PUs and can ultimately lead to chain scission of the urethane bond. Loss of the urethane bonds has been observed to be more prominent in phase mixed polymer domains than in segregated hard segment domains [31]. Furthermore, this breakdown can result in the loss of H bonding in these systems [32]. The above considerations may provide some very important insight towards explaining the dissimilar behaviour among the PCNU materials in the current study (Fig. 1). Very recent information [20, 21] has indicated that HDI-based PCNUs are much more phase separated than is the MDI-321 polymer. Furthermore, the surface of the MDI-321 material is made up primarily of the H-bonded urethane/ polycarbonate moities as opposed to segregated hard segment domains [20, 33]. These data were corroborated by ATR-FTIR and FTIR reported recently [20, 21]. Meijs et al. [34] have shown that oxidation is particularly prominant at hard/ soft segment interphases. If H-bonding can be lost through oxidation processes, as discussed above, then the protective shield of the carbonate groups against hydrolysis would be effectively compromised at the soft/ hard segment
The effect of oxidation on hydrolysis of polyurethanes
339
interphase, and hydrolytic degradation of carbonate groups would be expected to increase. This would have an end result of more labeled products being released. On the other hand, the structural character of the HDI-based PCNUs is such that the surfaces are highly phase separated and the polycarbonate is reported to be in a crystalline state [21, 33]. Therefore, oxidation would be expected to be low in the phase separated hard segment domains. Oxidation of the PCNUs may possibly lead to crosslinking of the oligomeric soft segment, as well as chain scission. If the former is occurring, then there could be a reduction in the amount of released product from the surface. The alternative explanation is that the change in carbonate chemistry may have rendered the crystalline soft segment less vulnerable to the particular enzyme studied in this paper. However, these considerations remain to be further elucidated. The variation in hydrolytic behaviour of CE with respect to oxidation pretreatment is not restricted to the PCNU materials, because as seen in Fig. 2, both the polyether- and polyester-PUs have unique responses. Hence, it is concluded that if synergistic mechanisms are at play between hydrolytic and oxidative pathways of the biodegradation process then consideration must be given to both polymer chemistry and structure in order to access the potential direction that such mechanisms will take for a given implant material. The whole question of whether a synergistic mechanism does exist in vivo is further elaborated on in the PMN experiment. PMA stimulates the respiratory burst and should enhance the oxidative effect of PMN on the polymer, whereas AZ inhibits myeloperoxidase which catalyzes the formation of HOCl and therefore should minimize the oxidative effect of PMN on the polymer degradative process [26]. In the current study, the PMN cell system (Fig. 3) yielded a similar conclusion to that observed for the in vitro biodegradation of HDI-431 (Fig. 1). The PMA which caused the release of measurable amounts of HOCl (Fig. 3), also inhibited radiolabel release (Fig. 4). Although a similar mechanism may be operating in vivo, these findings should be taken as circumstantial evidence and not proof that such synergistic events are at play. In fact, it was previously found that although PMA significantly inhibited radiolabel release elicited by PMN, it was possible that this was due to the effect of PMA on cell survival and its ability to synthesize and/ or secrete the hydrolytic agents acting on the polymer [8]. In the latter study by Labow et al., the rate of radiolabel release was found to be directly proportional to cell viability and that as the PMN died, radiolabel release decreased [8]. Similar results were found in the current study. PMA greatly accelerated the rate of cell death (Fig. 5). The additional finding in this study was that AZ, which decreased the rate of cell death due to PMA, significantly increased the amount of radiolabel release, although there was no significant difference in the LDH activity or HOCl production by PMN treated with AZ and no PMA when compared to DPBS controls. The above results are interesting because they suggest that by inhibiting specific cellular activities, the cells’ hydrolytic potential may be increased. PMA stimulates the respiratory burst by activating protein kinase C [35]. This increases the
340
R. S. Labow et al.
production of hydrogen peroxide which is cytotoxic itself, in addition to being a substrate for myeloperoxidase which catalyzes the production of HOCl, another very cytotoxic agent, from chloride and hydrogen peroxide. AZ inhibits the production of hydrogen peroxide, thereby inhibiting HOCl production, and so decreases the rate of cell death [35], although not significantly different from that of the DPBS control in the current study. Future studies will need to probe the level of hydrolytic enzyme activity between these different conditions in order to gain a deeper understanding of the mechanisms at play. In summary, this study showed that rather than accelerating biodegradation, oxidation significantly decreased the release of hydrolysis degradation products from several PUs. The effect of surface and bulk chemistries on the formation of stabilizing forces such as inter- and intra-molecular hydrogen bonding, is believed to be important in terms of defining the effect of the oxidation on the PUs susceptibility to subsequent hydrolysis. The DSC and SAXS data along with the ATR-FTIR reported for the materials in other recent publications [20, 21, 33] support the hypothesis that hydrogen bonding is more important and possibly more dominant than crystallinity in stabilizing the PCNUs. The effect of oxidation by the release of HOCl from PMN may be a perturbation factor on hydrolytic cleavage, but cell survival may also be a factor. Therefore, further work remains to be done with regard to elucidating the biochemical reactions involved in the process of in vivo ESC. Acknowledgements The assistance of Janet Malowany, Samir Hazra, Sanjay Jacob, Erin Meek and Girija Waghray is gratefully acknowledged. This study was funded by grants from the Canadian Institutes of Health Research (CIHR) and Materials Manufacturing Ontario (MMO).
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
L. Pinchuk, J. Biomater. Sci. Polymer Edn 6 (3), 225 (1994). J. M. Anderson, Cardiovasc. Pathol. 4 (3), 33S (1993). E. Schacter, R. L. Lopez and S. Pati, Biochem. Pharmacol. 41, 975 (1991). J. Casas, Q. Zhao, M. Donovan, P. Schroeder, K. Stokes and K. Untereker, J. Biomed. Mater. Res. 46, 475 (1999). N. R. Ackerman and J. R. Beebe, J. Pharmacol. Exp. Ther. 193, 603 (1975). W. J. Kao, A. Hiltner, J. M. Anderson and G. A. Lodoen, J. Biomed. Mater. Res. 28, 819 (1994). J. A. Hunt, B. F. Flanagan, P. J. McLaughlin, I. Stickland and D. F. Williams, J. Biomed. Mater. Res. 31, 139 (1996). R. S. Labow, E. Meek and J. P. Santerre, J. Biomed. Mater. Res. 54, 189 (2001). R. S. Labow, E. Meek and J. P. Santerre, Biomaterials 22, 3025 (2001). R. S. Labow, E. Meek and J. P. Santerre, J. Biomed. Mater. Res. 39, 469 (1998). R. S. Labow, E. Meek and J. P. Santerre, J. Biomater. Sci. Polymer Edn 10, 790 (1999). J. P. Santerre, R. S. Labow, D. Erfle, D. Duguay and G. A. Adams, J. Biomed. Mater. Res. 28, 1187 (1994).
The effect of oxidation on hydrolysis of polyurethanes 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35.
341
R. S. Labow, D. G. Duguay and J. P. Santerre, J. Biomater. Sci. Polymer Edn 6, 169 (1994). J. P. Santerre, R. S. Labow and G. A. Adams, J. Biomed. Mater. Res. 27, 97 (1993). J. P. Santerre and R. S. Labow, J. Biomed. Mater. Res. 36 (2), 223 (1997). J. P. Santerre, L. Shajii and H. Tsang, J. Dent. Res. 78, 1459 (1999). F. G. B. Wang, R. S. Labow and J. P. Santerre, J. Biomed. Mater. Res. 35, 207 (1997). F. G. B. Wang, R. S. Labow and J. P. Santerre, in: 23rd Amer. Soc. Biomater. Mtg., p. 451 (1997). F. G. B. Wang, J. P. Santerre and K. Solovey, in: 24th Amer. Soc. Biomater. Mtg., p. 393 (1997). Y. W. Tang, J. P. Santerre and R. S. Labow, J. Biomed. Mater. Res. 56, 516 (2001). Y. W. Tang, J. P. Santerre and R. S. Labow, J. Biomed. Mater. Res. 57, 597 (2001). Y. W. Tang, PhD Thesis, University of Toronto (2000). Y. Wu, C. Sellitti, J. M. Anderson, A. Hiltner, G. A. Lodoen and C. R. Payet, J. Appl. Polym. Sci. 46, 201 (1992). F. Li and D. Y. Hui, J. Biol. Chem. 272, 28666 (1997). R. S. Labow, D. J. Erfle and J. P. Santerre, Biomaterials 16, 51 (1995). S. J. Weiss, R. Klein, A. Slivka and M. Wei, J. Clin. Invest. 70, 598 (1982). A. N. J. Moore, H. Vakos, T. A. M. Neville, P. Tonge, P. Arya and G. W. Burton, Biotech. Tech. 10, 523 (1996). R. S. Labow, E. Meek and J. P. Santerre, J. Cell. Physiol. 186, 95 (2001). D. J. Erfle, J. P. Santerre and R. S. Labow, Cardiovasc. Path. 6, 333 (1997). V. Gajewski, Rubber World 15 (1990). M. A. Schubert, M. J. Wiggins, M. P. Schaefer, A. Hiltner and J. M. Anderson, J. Biomed. Mater. Res. 29, 337 (1995). S. Khatua and Y. L. Hsieh, J. Polym. Sci., Part A: Polym. Chem. 35, 3263 (1997). I. Revenko, Y. W. Tang and J. P. Santerre, Surf. Sci. 491, 346, (2001). G. Meijs, S. J. McCarthy, E. Rizzardo, Y.-C. and R. Chatelier, J. Biomed. Mater. Res. 27, 345 (1993). H. Bergstrand, T. Eriksson, A. Hallberg, B. Johansson, K. Karabelas, P. Michelsen and A. Nybom, J. Pharmacol. Exp. Ther. 262, 1334 (1992).
This page intentionally left blank
Novel dendrimer based polyurethanes for PEO incorporation X. DUAN 1 , C. M. GRIFFITH 2 , M. A. DUBÉ 1 and H. SHEARDOWN 3,∗ 1 Department
of Chemical Engineering, University of Ottawa, 161 Louis Pasteur St., Ottawa ON, K1N 6N5, Canada 2 University of Ottawa Eye Institute, 501 Smyth Rd., Ottawa ON, K1H 8L6, Canada 3 Departments of Chemical Engineering and Pathology and Molecular Medicine, McMaster University, 1280 Main St. W., Hamilton ON, L8S 4L7, Canada Received 2 January 2002; accepted 3 April 2002 Abstract—A series of segmented polyurethanes based on methylene diisocyanate/ poly (tetramethylene oxide) and chain extended with either ethylene diamine or butane diol in combination with a generation 2 polypropylenimine octaamine dendrimer were synthesized. For polymer synthesis, the dendrimers were protected with either t-boc or Fmoc groups and were incorporated into the polyurethane microstructure to permit further functionalization with biologically active groups. Following deprotection, the dendrimers were reacted with succinimidyl propionate polyethylene oxide (SPA-PEO) to improve the protein resistance of the polymers and to examine the potential of this technique for polymer functionalization. Different synthesis techniques were examined to optimize the incorporation of the PEO into the polymer microstructure. Incorporation of the dendrimers and the PEO were confirmed by NMR and FTIR. Gel permeation chromatography was used to examine the molecular weights of the various polyurethanes. The dendrimer incorporated polymers had significantly lower molecular weights than the ED or BDO chain extended controls, likely due to lower reactivity of the dendrimers as a result of steric factors. Following PEO reaction, the molecular weights of the resultant polymers were consistent with the levels of PEO incorporation noted by comparison of peak intensities in the NMR spectra. Due to the highly hydrophilic nature of the PEO, some migration to the polymer surface was expected. Water contact angles and XPS, used to characterize the surfaces, suggest that there was some PEO enrichment at the surface of the polymers. Adsorption of radiolabeled fibrinogen to the polymer surfaces was decreased by a factor of approximately 40% in some of the PEO incorporated polymers. There were also differences in the patterns of plasma protein adsorption on the various surfaces as evaluated by SDS PAGE and immunoblotting. Therefore, the use of dendrimers in biomaterials for incorporation of a large number of functional groups seems to be promising. Key words: Polyurethanes; PEO; dendrimers; protein adsorption.
∗ To
whom correspondence should be addressed.
344
X. Duan et al.
INTRODUCTION
Polyurethanes have seen widespread medical and biomedical use, particularly in blood contacting applications due to their superior mechanical properties and acceptable biocompatibility. Nevertheless, blood compatibility remains a significant limitation in the widespread application of these materials. A number of different methods for improving the blood and biocompatibility of these materials have been used, including functionalization of the bulk material [1– 5] and surface modification with various moieties [6– 12]. One of the more promising methods for improving blood compatibility of biomedical polymers involves modification with polyethylene oxide (PEO). The protein resistant properties of PEO are well known e.g. [13, 14] and as such, surface modification with PEO has been widely used to generate materials with superior blood compatibility. The protein resistant properties of PEO have been demonstrated in a number of studies to be dependent on the density of the polymer. As a result, star PEO [15] and comb-like PEO [16] have been examined as a means of improving the protein resistance of various surfaces. PEO incorporation into the polymer microstructure has also been used to improve biocompatibility. For example, Corneille et al. [17] incorporated PEO into the microstructure of a polyurethane backbone and demonstrated by surface plasmon resonance (SPR) that these polymers showed significantly reduced albumin adsorption. Santerre et al. [11] also demonstrated that PEO containing polyurethanes showed reduced protein adsorption. However, as expected, incorporation of PEO into a polyurethane backbone has been shown to decrease the mechanical properties of the polymer [18]. To overcome such limitations, Tan and Brash have developed novel polyurethane based pluronic equivalents, PEO-PU-PEO, which are coated onto the surface of polyurethanes. They have demonstrated significant reductions in the adsorption of plasma proteins with these coated polyurethanes, due to the presence of PEO at the polymer– biological interface [19]. Dendrimers are three-dimensional, highly ordered oligomeric and polymeric compounds formed by reiteration reaction sequences starting from smaller molecules or initiator cores [20]. These highly branched spherical polymers have well defined microstructures and a set number of terminal amine groups, which doubles (or triples or quadruples depending on the monomer selected) with each subsequent level of monomer addition. These molecules combine characteristics typical of small organic molecules, including monodispersity and definite composition with the attributes of traditional polymers including high molecular weight. These molecules have been examined in biomedical applications as in vitro gene transfer agents [21– 24], and in various therapeutic applications [25– 27]. Polyamidoamine dendrimers having poly(ethylene glycol) grafts were designed as a novel drug carrier with an interior suitable for the encapsulation of drugs and a biocompatible surface [28]. Dendrimers have also been shown to be useful in providing strong signal amplification in diagnostic applications [29]. Recently, fructose modified dendrimers have been used in biomaterials to generate surfaces with a high density of ligands for improving interactions between hepatocytes and surfaces [30]. There-
Novel dendrimer based polyurethanes for PEO incorporation
345
fore, as a result of their regular and multifunctional nature, these molecules seem promising in biomaterials applications. These molecules have not been widely examined for generating surfaces with higher levels of functional groups. Here we report on the development of methods for the incorporation of dendrimers into a polyurethane backbone via the chain extension step. This resulted in a polymer with a potentially large number of sites for further functionalization. Generation 2 poly (propylene imine) dendrimers were used in all studies. In the current work, these polymers were subsequently reacted with PEO and examined for plasma protein interactions. These results as well as extensive polymer characterizations are presented. MATERIALS AND METHODS
Polymer synthesis Control polyurethane synthesis. A conventional two step procedure was used in the synthesis of all polyurethanes used in the study [11]. To minimize the introduction of water, and the initiation of undesired side reactions, where possible, reactants were dried overnight under vacuum. The control polyurethanes were synthesized by reacting methylene diphenyl p isocyanate (MDI, Acros Organics, Nepean ON) with poly tetramethylene oxide (PTMO, Aldrich Chemical Co. Milwaukee WI, MW 650) dissolved in dry dimethyl sulfoxide (DMSO) (10%, w / v) in a 2 : 1 ratio in a dry nitrogen environment for a period of 90 min at 50◦ C. The polymers were chain extended with a 1 : 1 molar ratio (prepolymer : chain extender) of either ethylene diamine (ED, Adrich Chemical Co., Milwaukee WI) or butanediol (BDO, Aldrich Chemical Co., Milwaukee WI) in DMSO (2% w /v) for 2 h at room temperature in the case of ED or at 40◦ C in the case of BDO. Two to 3 drops of dibutyl tin dilaurate in tetrahydrofuran were added as catalyst when BDO was used as the chain extender. All polymers were purified by precipitation in water and extensive water washing prior to characterization. Several different techniques were used in an attempt to optimize the incorporation of the dendrimers into the polyurethane microstructure. The various methods are described below. PEO functionalization prior to dendrimer incorporation into polyurethane. A solution of polypropyleneimine octaamine dendrimer generation 2.0 (G2 ) (Aldrich Chemical Co., Milwaukee WI) and SPA-PEO with a reported molecular weight of 2000 (Shearwater Polymers, Huntsville, AL) in CH2 Cl2 in a 1 : 6 (partially modified) or 1 : 8 (fully modified) molar ratio was stirred at room temperature for 2 h to generate a statistical distribution of PEO modified dendrimers. Dendrimer modification was confirmed by GPC. These PEO modified dendrimers were then incorporated into the polyurethane backbone during the chain extension step by reacting the prepolymer in a 1 : 1 molar ratio with a mixture of standard chain extender and PEO modified dendrimer in a 9 : 1 molar ratio.
346
X. Duan et al.
Simultaneous chain extension of polyurethanes with dendrimers and reaction with PEO. Using a 6 : 1 molar ratio of SPA-PEO 2000 and dendrimers, and a 9 : 1 molar ratio of standard chain extender and dendrimer PEO mixture, the SPA-PEO 2000 and dendrimers were introduced into the polyurethane during the chain extension step in the following sequence : standard chain extender, PEO and dendrimer. This sequence was selected to minimize polymer crosslinking by the multifunctional dendrimers.
Dendrimer incorporation and subsequent PEO functionalization. In order to facilitate the incorporation of the dendrimers into the polymer backbone and their subsequent functionalization by PEO, the amine functional groups were first protected by either t-boc [31] or Fmoc chemistry. Briefly, for t-boc protection, the G2 polypropyleneimine dendrimer was reacted with N-t-Boc-L-Ala hydroxy succinimide ester (1.01 equivalent per NH2 end group at a molar ratio of 6 : 1 protecting group: dendrimer) overnight in CH2 Cl2 with triethylamine. Following the reaction, the protected dendrimers were washed with distilled water and saturated Na2 CO3 . The product was subsequently dried with H2 SO4 and the solvent removed by rotoevaporation to yield the protected dendrimers. Fmoc protection of the NH2 groups of the dendrimer was achieved by reaction of a G2 polypropyleneimine dendrimer with 9-fluorenylmethyloxycarbonyl chlorocarbonate (9-Fmoc) in CH2 Cl2 at a molar ratio of 6 : 1. The Fmoc was added slowly in 1– 2 ml aliquots and the mixture was reacted at 4◦ C for 4 h, followed by 8 h of reaction at room temperature. Following removal of the solvent by roto-evaporation, the product was dried under vacuum at 60◦ C to yield the Fmoc protected dendrimers. The t-Boc or Fmoc protected dendrimers were then incorporated into the polyurethane during the chain extension step in combination with either ED or BDO in a 9 : 1 molar ratio of standard chain extender to dendrimer. Following precipitation of the polymers in water and purification by extensive water washing and methanol extraction, the protection groups were removed. In the case of the t-boc protected dendrimer incorporated polyurethanes, stirring overnight in 99% formic acid facilitated deprotection. In the case of the Fmoc protected dendrimer incorporated polymers, deprotection occurred by reaction with diethylamine for a period of two hours. The deprotection step was followed by PEO incorporation. The deprotected dendrimer polyurethane was dissolved in DMF, and reacted with the theoretical maximum amount of SPAPEO 2000 for full dendrimer modification overnight. Following solvent removal, the final product was purified extensively prior to characterization. Films of the final polymers were cast from 10% solutions of the polymers in DMF for characterization. Several of the PEO incorporated polymers had poor mechanical properties. Films of these polymers were therefore cast onto the control polyurethanes for characterization. A summary of the various polyurethanes synthesized is presented in Table 1.
Novel dendrimer based polyurethanes for PEO incorporation
347
Table 1. Summary of polyurethanes synthesized and nomenclature Nomenclature
Chain extender
Dendrimer
PU-ED PU-G2 -ED2a PU-ED-PEO2a PU-ED-PEO2b PU-BDO PU-G2 -BDO2a PU-G2 -BDO2b PU-BDO-PEO2b PU-ED-PEO1
ED ED ED ED BDO BDO BDO BDO ED
No Yes Yes Yes No Yes Yes Yes Yes
1 Synthesized 2 Synthesized
Protection
Deprotection
t-Boc t-Boc Fmoc
Formic acid Diethylamine
t-Boc Fmoc Fmoc
Diethylamine
by serial addition of chain extender, followed by dendrimer and PEO. by reaction with protected dendrimers followed by deprotection and reaction with
PEO. a t-boc used for dendrimer protection. b Fmoc used for dendrimer protection.
Polymer characterization Gel permeation chromatography. Gel permeation chromatography was used to obtain molecular weights of the various polymers. The polyurethanes were dissolved in dimethyl formamide (DMF) at a concentration of 0.1% (w /v). The samples were analyzed using a Waters GPC system equipped with three Waters Styragel® HR4 GPC columns at room temperature. A solution of 0.1% (w /v) LiBr in DMF was used as the mobile phase. Polystyrene standards were used for calibration. Nuclear magnetic resonance. 1 H-NMR spectroscopy was performed to confirm the polymer microstructure using an AV-2000 spectrometer after dendrimer and using a Bruker AMX-500 after PEO incorporation. This was done using the peaks at 6.13 ppm and 8.37 ppm which are assigned to the protons in the urea groups formed by the reaction of NH2 with NCO of the prepolymer and the peak at 3.51 ppm ( CH2 O ) for PEO. All spectra were recorded in a mixture of deuterated DMSO and chloroform (v : v = 1 : 1). Fourier transform infrared spectroscopy. All polymers were analyzed using transmission FTIR spectroscopy with a Bio-Rad FTS-10 FTIR. Films were cast on a NaCl crystal window and dried under vacuum at 60◦ C for 24 h prior to recording. Water contact angles. Surface hydrophilicity was examined by measuring advancing water contact angles using a Rank Scherr Tumico 22-2000 series 14 inch horizontal beam bench comparator. A 10 µl water bubble was placed on a previ-
348
X. Duan et al.
ously dried and methanol rinsed polymer surface using a microsyringe. For each sample, 3– 5 separate drops were examined. X-ray photoelectron spectroscopy. XPS analysis was performed at the National Research Council of Canada. The surface of the samples was analyzed using a KRATOS AXIS HS X-ray photoelectron spectrophotometer (Kratos, Manchester UK). The size of the analyzed area was approximately 1 mm2 . Monochromatized AlKα radiation was used for excitation and a 180◦ hemispherical analyzer with a three channel detector was employed. The X-ray gun was operated at 15 kV and 20 mA. The spectrophotometer was operated in Fixed Analyzer Transmission (FAT) mode throughout the study using electrostatic magnification. Surface and highresolution spectra were collected using a 160 and 20 eV pass energy respectively. The pressure in the analyzer chamber was 10−8 to 10−9 torr. An electron flood gun was used to neutralize the charge during the experiment. Binding energies were referenced to the carbon-carbon bond that was assigned a binding energy of 285 eV. Atomic composition was estimated using standard software provided with the instrument using the following sensitivity factors: 0.25 for C1s, 0.66 for O1s and 0.42 for N1s relative to F1s at 1.00. Peak deconvolution was performed using the software provided with the instrument. Protein adsorption Adsorption of radiolabeled fibrinogen from buffer. Human fibrinogen (Enzyme Research Labs, South Bend IN) was labeled with 125I (Na125 I, 0.5 mCi, Amersham, Arlington Heights IL) using the iodogen method. Free iodide was removed by overnight dialysis at 4◦ C with three changes of the phosphate buffered saline (PBS) dialysate. Free iodide was determined by trichloroacetic acid precipitation of the protein. Fibrinogen concentration in the final solution was determined spectrophotometrically. A 0.25% labeled 1 mg/ ml fibrinogen solution in PBS was prepared and diluted for adsorption experiments. 6 mm diameter polyurethane discs were preincubated overnight in PBS at 4◦ C. The surfaces were then incubated with 250 µl of the radiolabeled fibrinogen solution for three hours. Following the adsorption, the surfaces were dip-rinsed three times in PBS, and the radioactivity determined using a gamma counter. The amount of adsorbed fibrinogen was determined by comparison of the results to known standards. SDS PAGE and immunoblotting. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting were used to examine the patterns of protein adsorption from plasma to the various surfaces. The surfaces were incubated in citrated pooled normal plasma (>25 donors) for two hours. Following 3 dip rinses in PBS to remove loosely bound proteins, the surfaces were incubated overnight in 2% (wt) SDS at 4◦ C to elute the adsorbed proteins. SDS-PAGE and
Novel dendrimer based polyurethanes for PEO incorporation
349
immunoblotting were performed as previously described [32]. Total protein adsorption was determined based on bovine serum albumin (BSA, Sigma Chemical Co., St. Louis MO) standards using a Bio-Rad DC total protein microassay.
RESULTS
Polymer synthesis In the current work, two approaches were used to incorporate dendrimers and additional functionality via PEO into a polyurethane microstructure. A summary of the polymers synthesized and the nomenclature used to describe them is shown in Table 1. In the first approach, a star PEO dendrimer was synthesized by reacting SPA-PEO 2000 with poly (propylene imine octaamine) G2 dendrimer at a molar ratio of 6 : 1 (PEO : dendrimer). This approach should leave, on average, two free amine groups for polyurethane chain extension. To overcome the steric hindrance problem, a modification was made to the method. SPA-PEO2000 and dendrimers were introduced in sequence during the chain extension step, with the standard chain extender (i.e. ED or BDO) first, SPAPEO2000 second, and the G2 dendrimer last. The sequence was selected in order to minimize crosslinking by the multifunctional dendrimers. A number of parameters, including the reaction temperature, and speed of reactant addition, affected the chain extension reaction. However, at the determined optimal conditions, summarized in Table 2, a DMSO and DMF soluble polyurethane was formed (denoted PU-EDPEO1). While actual tensile testing was not performed due to the need for large and regular films, this polymer appeared to be similar if not superior mechanically to the control polyurethanes, likely the result of some crosslinking during the reaction. An alternative approach, involving protection/ deprotection of the amine groups in the dendrimer for chain extension and subsequent PEO attachment was also examined. The dendrimers were partially protected with the amine protecting groups t-Boc or Fmoc at a molar ratio of 1 : 6, leaving on average two free Table 2. Optimal conditions for the synthesis of dendrimer incorporated polyurethane with PEO attached (PUED-PEO1 ) Prepolymerization temperature (◦ C) Prepolymerization duration (min) Chain extension temperature (◦ C) Chain extension duration (min) Standard chain extender concentration (%, w/v) Dendrimer G2 concentration (%, w/ v) Chain extension sequence Dropping speed of dendrimer solution (drops/ s) Post chain extension temperature (◦ C) Post chain extension duration (min)
50 120 25 60 2.0 2.0 ED/ BDO→SPA-PEO→dendrimer 1 90 60
350
X. Duan et al.
Table 3. Stability of ED chain extended polyurethanes to formic acid treatment overnight: GPC results Sample ED chain extended control BDO chain extended control Formic acid treated ED PU Formic acid treated BDO PU
Mn
Mw
82 800 49 400 68 900 30 300
133 100 64 700 110 000 48 900
amine groups for the chain extension reaction. The protected dendrimers were incorporated during the chain extension step with either ED or BDO at a ratio of 9 moles of standard chain extender to one mole dendrimer and the mixture was reacted as in the case of the controls. The resulting polyurethanes (denoted PU-G2 ED2a for the t-Boc protected polymers, which were ED chain extended and PU-G2 BDO2b for the Fmoc protected polymers which were BDO chain extended) showed no macroscopically apparent signs of crosslinking, suggesting that there were no significant amounts of unprotected dendrimer present in the chain extension reaction mixture. The PEO incorporated polyurethanes were synthesized following the appropriate deprotection technique. Upon precipitation in water, the previously t-Boc protected PEO — polyurethanes formed a stable emulsion, necessitating solvent removal by roto-evaporation. Since the acid treatment used for deprotection breaks urea bonds between the protecting group and the dendrimer, the effect of the acid treatment on ED and BDO chain extended polymer controls was examined. GPC analysis results for the untreated polymers and those treated with acid are summarized in Table 3. It is clear from these results that there is degradation of the polymer by the acid treatment, with a 17% decreased noted in the molecular weight of the ED chain extended polymers and a 39% decrease noted in the molecular weight of the BDO chain extended polymers. However, the molecular weights of the polymers remained significant, suggesting that a combination of factors, including polymer degradation and increased hydrophilicity as a result of the PEO incorporation likely resulted in the formation of the emulsion. No emulsion formed following PEO incorporation into the Fmoc protected polymers, providing additional evidence that the degradation of the polyurethane chains contributed to the formation of the emulsion. The final PEO polyurethanes did not possess the superior mechanical properties traditionally associated with polyurethanes, and it was difficult to obtain workable films for subsequent tests with these polymers. This hindered the absolute measurement of the mechanical strength of these polymers. Polymer characterization 1
H-NMR and FTIR. 1 H-NMR spectra were used to confirm the microstructure of the polymers based on peak assignments. Spectra for the ED chain extended polyurethanes before and after dendrimer and PEO modification are shown in Fig. 1.
Novel dendrimer based polyurethanes for PEO incorporation
351
Table 4. Amount of PEO incorporated into polyurethanes as determined by 1 H-NMR Polymer
PEO amount (g/ g of G2 dendrimer)
Theoretical maximum PEO amount (g/ g of G2 dendrimer)
PU-ED-PEO2a PU-ED-PEO1 PU-BDO-PEO2b
1.57 12.36 10.02
15.52 15.52 15.52
Of specific interest are the peaks at 6.13 and 8.37 ppm, which are assigned to protons in the urea groups formed by the reaction of NH2 (from ED or dendrimer) with NCO of the prepolymer. These were found to increase slightly with incorporation of the dendrimer prior to deprotection, as expected, due to the generation of a urea linkage by the protection reaction. However, the presence of existing urea groups in the polyurethane urea structure and the relatively low amount of dendrimer incorporated make it difficult to evaluate the level of incorporation. A PEO peak is clearly seen in Fig. 1c at 3.51 ppm following deprotection and reaction with PEO. The incorporation of the dendrimers into the polyurethane microstructure was shown more clearly with the BDO chain extended polymers due to the lack of urea groups in the base polymers. Quantitative analysis of the spectra, summarized in Table 4, show that relative to the dendrimer peaks at 6.13 and 8.37 ppm, the PU-ED-PEO2a had the least amount of incorporated PEO, while PU-ED-PEO1 had the most PEO chains. Similar results were noted in the FTIR spectra of the various polymers. Compared with the ED chain extended polyurethane control, the dendrimer incorporated polyurethane, had increases in peaks at 3315, 1708 and 1647 cm−1 , which are indicative of increases in hydrogen bonding of N H and C O of urethane and urea, probably due to the incorporation of dendrimers. Following PEO attachment, the FTIR spectrum for the ED chain extended polymer (PU-ED-PEO1), shows decreased peaks at 3315, 1708 and 1647cm−1 , indicative of the lower hydrogen bonding in these polymers. Gel permeation chromatography. GPC analysis of the modified dendrimers and the polymers, shown in Table 5, demonstrated that star PEO dendrimers had been formed and that the dendrimers modified at a 6 : 1 molar ratio were not completely modified compared to those modified at an 8 : 1 ratio. However, these measured molecular weights were lower than those expected based on complete reaction, likely due to a combination of steric factors, the use of polystyrene rather than PEO standards in the analysis, the fact that for branched polymers such as those in the current study, the measured molecular weights will be lower than the actual values or simple changes in the hydrodynamic volume based on the new chemical structure. Similarly, the polyurethane molecular weights were lower what would be expected for full PEO modification. Nonetheless, these measurements are useful to reveal important trends in the data. As expected, based on the reactivity of
352
X. Duan et al.
Table 5. Polymer molecular weights determined by GPC Polymer
Mn
Mw
PU-ED PU-G2 -ED2a PU-ED-PEO2a PU-BDO PU-G2 -BDO2a PU-G2 -BDO2b PU-BDO-PEO2b PU-ED-PEO1
82 800 67 800 62 600 49 400 30 400 76 400 86 000 138 700
133 100 106 300 100 600 64 700 48 800 117 000 135 900 214 800
Figure 1. Typical 1 H-NMR spectrum of the synthesized polymers. Shown is (a) the ED chain extended polyurethane control (PU-ED), (b) dendrimer modified ED chain extended polyurethane (PU-G2 -ED2a ) and (c) the PEO modified ED chain extended PU (PU-ED-PEO2a). Of interest are the peaks at 6.13 ppm and 8.37 ppm (as shown), indicative of the formation of urea bonds in both polymers but to a greater extent in the dendrimer modified polymer as expected and the presence of the PEO at 3.51 ppm as shown.
Novel dendrimer based polyurethanes for PEO incorporation
353
isocyanates and amine groups relative to hydroxyl groups, the molecular weights of the ED chain extended polymers were significantly greater than those of the BDO chain extended polymers. The dendrimer incorporated polyurethanes had significantly lower molecular weights than control polyurethanes when t-Boc was used (PU-G2 -ED2a compared to PU-ED, and PU-G2 -BDO2a to PU-BDO), which is likely the result of the lower reactivity of the dendrimers due to steric factors. The polymers synthesized using Fmoc protected dendrimers had higher molecular weights than those synthesized using the t-Boc protected dendrimers (PU-G2 BDO2b compared to PU-G2 -BDO2a ). This, we hypothesize, was due to slight crosslinking in the polymer by unprotected dendrimer when Fmoc was used, as the t-Boc protected dendrimers were extensively purified prior to chain extension. The GPC results also demonstrate that no unreacted dendrimers remained in the polyurethanes following reaction and purification, as there were no peaks other than the polyurethane peak present. Following PEO attachment, the t-Boc protected dendrimer modified polyurethanes showed a slight but unlikely significant decrease in molecular weight. (PU-EDPEO2a compared to PU-G2 -ED2a ). While the incorporation of PEO would be expected to result in an increase in the molecular weight, it seems likely that degradation of the polymer backbone by the acid treatment used for deprotection affected the resulting polymer molecular weights. When the Fmoc protection was used, the molecular weight increased after PEO attachment, as expected. Taken together, the NMR, FTIR and GPC results provide strong evidence that the dendrimers and the PEO have been incorporated into the polymer microstructure and that differences in the attachment methods and protection methods resulted in differences in the amount of PEO incorporated in the polymer microstructure.
Water contact angles. Advancing water contact angles on the various polymers are summarized in Fig. 2. It can be seen that water contact angles measured on the PU-ED and PU-BDO controls were similar at approximately 70◦ . Following incorporation of the dendrimer, there was a small decrease in the water contact angle relative to the respective control. In the case of the ED polymer (PU-G2 -ED2a compared to PU-ED), this decrease was not significant, as might be expected due to the similarity in the microstructure of the dendrimer and ED. However, in the case of the BDO chain extended polymers (PU-G2 -BDO2b compared to PU-BDO), this decrease was significant and suggests surface differences between these polymers. Following incorporation of the PEO into the polymer microstructure (i.e. PU-EDPEO2a and PU-BDO- PEO2b), the contact angles showed a significant (p > 0.99) decrease to approximately 45◦ with both chain extenders. While these advancing water contact angles are higher than those noted with PEO surface modification, which average between 15 and 30◦ depending on the underlying substrate, they are similar to the results obtained by others for PEO grafted polyurethanes [33].
354
X. Duan et al.
X-Ray photoelectron spectroscopy (XPS). XPS results, summarized in Table 6, provide additional evidence for dendrimer incorporation into the polymer microstructure and presence at the surface. Specifically, there was an increase in the N1s signal on the dendrimer incorporated polyurethanes relative to the control polymers. For the ED chain extended polymers, the high resolution C1s envelope showed a significant increase in the contribution at 289.2 eV, consistent with an
Figure 2. Advancing water contact angles measured on the various polyurethanes. The PEO polymers in this case were PU-ED-PEO2a and PU-BDO-PEO2b. Significant decreases in the water contact angles were noted on the surfaces following PEO incorporation. These differences did not seem to be a function of the amount of PEO incorporated based on the NMR data. Table 6. Summary of XPS results on modified polyurethanes Sample
PU-ED PU-G2 -ED2a PU-ED-PEO2a (TFA) PU-ED-PEO2a (FA) PU-BDO PU-G2 -BDO2b PU-BDO-PEO2b
C1s Total
284.8 C C
286.3 C N, C O
289.2 urea
289.6 COOR
76.3 76.3 78.0 75.7 77.2 77.8 76.8
50.6 50.5 51.8 46.1 57.9 56.2 47.7
24.6 24.5 22.1 26.8 17.6 20.1 28.1
0.4 0.62 4.02 2.02 1.65 0.90 0.48
0.68 0.72 0 0.73 0 0.66 0.98
O1s
N1s
21.4 21.1 18.2 20.3 21.3 19.4 19.7
2.28 2.65 3.86 4.04 1.50 2.83 3.44
Novel dendrimer based polyurethanes for PEO incorporation
355
increase in the relative atomic concentration of C-OR, which would be expected following PEO incorporation. However, a similar increase was not noted on the BDO chain extended polymers. Furthermore, there was a decrease in the O1s signal on both the ED and BDO chain extended polymers following PEO incorporation. As well, a consistent increase in the N1s signal was noted. These results suggest surface enrichment of the dendrimers and the hard segment of the polyurethane in these polymers, likely due to the highly hydrophilic PEO chains being buried in the bulk polyurethane matrix when in the hydrophobic XPS environment. It is difficult however, to obtain an accurate picture of the surface composition in a biologically relevant aqueous environment however. Protein adsorption Adsorption of 125 I fibrinogen from buffer. Fibrinogen adsorption results to the various surfaces are summarized in Figs 3 and 4. For all experiments, the free iodide was determined to be less than 0.5%, and therefore free iodide is not expected to contribute to the measured amount of adsorbed fibrinogen on these polymers. A slight decrease in the amount of adsorbed fibrinogen was consistently noted following incorporation of the dendrimer. Following PEO incorporation, the decrease in fibrinogen adsorption was more significant. Furthermore, the effect of dendrimer and PEO incorporation on fibrinogen adsorption seemed more apparent in the BDO chain extended polymers. This result correlates well with the water contact angle measurements, which show a significant decrease in the hydrophilicity of the surfaces following dendrimer and PEO incorporation. On the t-Boc protected, ED chain extended polyurethanes, there was no significant difference between the amounts of fibrinogen adsorbed with the different acids used for deprotection purposes, i.e. formic acid and trifluoroacetic acid. Consistent with the NMR results, which showed that the level of PEO incorporation was significantly less when
Figure 3. Fibrinogen adsorption from buffer to the various ED chain extended polyurethanes. A decrease in fibrinogen adsorption was noted following dendrimer incorporation into the polymer microstructure and again following PEO incorporation. Furthermore, the trend follows the NMR results for level of PEO incorporation into these polymers.
356
X. Duan et al.
Figure 4. Fibrinogen adsorption from buffer to the various BDO chain extended polymers. As with the ED chain extended polymers, a decrease in fibrinogen adsorption was noted following dendrimer incorporation and again following PEO incorporation. However, the reduction in fibrinogen adsorption was not as great as with the ED chain extended polymers.
(a) Figure 5. Immunoblot results for plasma protein adsorption to the various polymers. With both chain extenders, there was a slight decrease in protein adsorption overall with the incorporation of the dendrimers and an even greater decrease with the incorporation of PEO chains into the polymer with the qualitative trends again following the measured levels of PEO incorporation. (a) PU-ED, (b) PUG2 -ED, (c) PU-ED-PEO1, (d) PU-ED-PEO2a, (e) PU-ED-PEO2b, (f) PU-BDO, (g) PU-G2 -BDO, (h) PU-BDO-PEO2b. The two right hand lanes on the blots represent molecular weight standards and the acrylamide gel of all adsorbed proteins.
Novel dendrimer based polyurethanes for PEO incorporation
357
t-Boc was used for dendrimer protection (PU-ED-PEO2a), there was no significant difference in the level of protein adsorption between the dendrimer incorporated and the PEO incorporated polymers. However, the polyurethane synthesized by sequential addition of the various reagents (PU-ED-PEO1) showed the greatest decrease in fibrinogen adsorption relative to the control, with an almost 40% decrease in the amount of adsorbed fibrinogen. This result is again consistent with the GPC and NMR results that suggest higher levels of PEO incorporation in these polymers. SDS PAGE and immunoblotting. SDS PAGE and immunoblotting analyses of the eluates following protein adsorption from plasma were performed to determine the patterns of plasma protein adsorption on the various surfaces. Immunoblots are presented in Fig. 5. Consistent with the fibrinogen adsorption results, there was a decrease in the adsorption of most of the proteins on the surfaces following incorporation of the dendrimers and following PEO attachment. On both the ED and BDO chain extended polyurethanes, decreases were noted in the amounts of adsorbed fibrinogen, C3, transferrin, albumin, vitronectin, apolipoprotein A1 as well as the complement proteins. However, the banding patterns indicated that there were no changes in adsorption of some of the other proteins, possibly the result of the relatively low amounts of adsorption noted on the control polymers. These low levels of adsorption to the base polyurethanes are consistent with the results of others [19] and are therefore not unexpected. While optical scans of the gold stained
(b) Figure 5. (Continued).
358
X. Duan et al.
(c)
(d) Figure 5. (Continued).
Novel dendrimer based polyurethanes for PEO incorporation
(e)
(f) Figure 5. (Continued).
359
360
X. Duan et al.
(g)
(h) Figure 5. (Continued).
Novel dendrimer based polyurethanes for PEO incorporation
361
gels and immunoblots were obtained and subsequently analyzed using Whole Band Analysis software (BioImage) to determine band molecular weights and intensities, gel/ blot data are at best semiquantitative and the intensities of the bands depend on the characteristics of the different antigen-antibody responses and colour reactions.
DISCUSSION
In the current work, we examined the potential of using dendrimers for incorporating higher levels of functional groups into the microstructure of polyurethanes for biomaterials applications. Since surface modification with PEO has been shown in numerous studies to significantly reduce protein adsorption, PEO has been widely used in biomaterials applications and its effects have been shown to be density dependent, we examined the potential of dendrimer incorporation into polymers and subsequent PEO modification to increase the protein resistance of the polymers. Several synthesis methods were used in order to maximize the conversion of reactive groups in both the dendrimers and in the polymers. Not surprisingly, modification of the dendrimers with the PEO and subsequent incorporation into the polyurethane microstructure during the chain extension step failed to result in high molecular weight products, likely a result of steric factors. However, sequential addition of the various reagents in an order derived to minimize polyurethane crosslinking while maximizing PEO incorporation resulted in mechanically strong yet solvent soluble polyurethanes. A variety of protection/ deprotection strategies were also used to develop PEO incorporated polyurethanes. With the protection/ deprotection approach, high molecular weight polyurethanes were synthesized in all cases. However, when t-Boc was used to protect the amine groups in the dendrimer, the acid deprotection step apparently affected the polyurethane backbone, decreasing the molecular weight of the resultant polymer. Based on these results, it is therefore preferable to use Fmoc for dendrimer protection. In all cases, the mechanical properties of the polyurethanes synthesized using the protection/ deprotection strategy followed by PEO reaction were poorer than the control polymers, despite the fact that a significant amount of the base chain extender was used in the synthesis of these polymers. While dendrimer incorporation did result in polyurethanes with marginally lower mechanical strength than that controls, the PEO incorporated polyurethanes could not be cast into workable films and coating with these polymers was necessary. It would therefore be expected that the incorporation of either higher relative amounts of dendrimer or higher dendrimer generations would have a further negative impact on these properties. However, based on the results of Tan and Brash [19], it is possible that these polyurethanes could be used with standard ED or BDO chain extended polyurethanes either as coatings or as blended films, thereby improving their mechanical integrity. NMR, FTIR and GPC confirmed incorporation of both the dendrimer and the PEO into the polyurethane microstructure. Furthermore, comparison of the amounts of PEO incorporated into the polyurethane chains as determined by peak ratios in
362
X. Duan et al.
NMR and by molecular weight changes determined by GPC demonstrates that the greatest amounts of PEO were incorporated by simultaneous reaction with the chain extender, PEO and dendrimer. Consistent with the degradation problems during the deprotection step in the t-Boc protected polymers, the BDO chain extended polymers reacted with the Fmoc protected dendrimers and subsequently modified with PEO showed significantly greater amounts of PEO incorporation than those using the t-Boc protected dendrimers. The amount of PEO incorporated obtained from the NMR spectra and that obtained from the GPC results were surprisingly consistent. Advancing water contact angles measured on the various polymer surfaces decreased significantly following PEO incorporation, although the measured angles remained somewhat higher than those noted in other studies following surface modification with PEO, suggesting that the mobility of the PEO is not adequate to generate surfaces with high PEO concentrations. However, it is also possible that with higher PEO concentrations obtained by using a higher dendrimer : standard chain extender ratio or higher dendrimer generations could also result in surfaces with higher PEO concentrations and therefore lower advancing water contact angles. XPS analysis of the various polyurethane surfaces demonstrates that some dendrimer and PEO were present on the surfaces. Specifically, relative to the control polyurethanes, there was an increase in the N1s signal following dendrimer incorporation, indicative of dendrimer enrichment at the polymer surface. A further increase in the N1s signal was noted following PEO incorporation, which may suggest enrichment of the dendrimer at the surface since this contribution was found to be predominantly from amine rather than isocyanate groups. For all of the ED chain extended polymers after PEO incorporation, there was an increase in the contribution of COR in the C1s envelope, which may be indicative of the presence of PEO. However, this was not accompanied by an increase in the O1s signal. It is likely that the relatively complex chemistry of these polymers, relatively low amounts of PEO incorporated and the hydrophobicity of the XPS environment contributed to the lack of clear evidence of PEO surface enrichment in these polymers. However, in agreement with the water contact angle measurements, it is clear that there is a significant amount of standard polyurethane present on the surfaces, suggesting that incorporation of higher amounts of PEO may be possible. The literature suggests that the levels of protein adsorption to PEO modified surfaces are dependent on such factors as the PEO density and molecular weight e.g. [15, 18]. Relative to the control polyurethanes, the levels of fibrinogen adsorption decreased following incorporation of the dendrimer in the ED series of polyurethanes. However, relative to the dendrimer incorporated polyurethanes, there was no decrease in the adsorption of fibrinogen following PEO incorporation. This is not surprising given that with the t-Boc protection method, both GPC and NMR analysis suggested that very low levels of PEO were present in these polymers. However, on the surface that showed the highest level of PEO incorporation, there was a significant decrease (approximately 40% relative to the control) in the
Novel dendrimer based polyurethanes for PEO incorporation
363
adsorption of fibrinogen from buffer. On the BDO series of polyurethanes, using Fmoc protection for the dendrimers, a similar decrease in the adsorption relative to the controls was noted following incorporation of the dendrimers. Furthermore, there was a small but significant decrease in fibrinogen adsorption following PEO incorporation. However, not surprisingly, the levels of fibrinogen adsorption remained higher than those noted on the ED-PEO-1 polymers. Therefore, the protein adsorption results correlate well with the NMR and GPC results, with lower levels of fibrinogen adsorption noted on the surfaces that showed higher amounts of attached PEO. SDS-PAGE and immunoblotting results were also consistent with the levels of PEO incorporation into the polymers, with decreases in plasma protein adsorption noted following dendrimer and PEO incorporation. While it would obviously be desirable to see a greater reduction in the fibrinogen adsorption since the levels noted in the current work would not likely significantly alter the response of the system to blood, the relatively low levels of dendrimer used and the low generation of dendrimer selected could potentially be altered to increase the protein resistance of these surfaces. It seems likely increasing the dendrimer and PEO content of these polymers, which in the current work was quite low, can further decrease that protein adsorption. Furthermore, only molecular weight of PEO was examined in the current study and it is possible that the mechanical and/ or biological properties of the polymers can be further enhanced through the use of different chain lengths of PEO.
CONCLUSIONS
In the current work, dendrimer incorporation into the microstructure of polyurethanes was examined, based on the hypothesis that these multifunctional polymers could be used to further functionalize with the polymers with large numbers of biologically relevant molecules. Modification of the dendrimers with PEO for improving protein resistance was examined in the current work. A number of strategies were used to incorporate the dendrimers and PEO into the polymer microstructure, including simultaneous reaction during polyurethane synthesis and the use of different amine protection groups. Relatively low levels of dendrimer to standard chain extender were examined in these studies in order to develop polymers with reasonable mechanical properties. NMR and FTIR confirmed dendrimer and PEO incorporation into the polymer microstructure. Gel permeation chromatography was used to confirm the molecular weights of the various polymers. The results suggested that a highest level of PEO in the polymers was achieved using the simultaneous reaction approach. Furthermore, physical observations and GPC analysis demonstrated that t-Boc protection of the dendrimers was not optimal as the deprotection step resulted in degradation of the polymer chains and relatively low levels of PEO incorporation. Surface analysis of the polymers using advancing water contact angles and XPS suggested some PEO enrichment at the polymer surfaces, with the potential for improvement through the use of greater amounts of dendrimers in the chain
364
X. Duan et al.
extension step. On the polymers with the highest levels of PEO incorporated, a 40% decrease in the adsorption of fibrinogen was noted. Furthermore, immunoblotting results suggest that there were decreases in the adsorption of a number of plasma proteins on these polymers. Therefore, the results of the current study demonstrate that dendrimers can be used in biomaterials applications in order to generate surfaces or bulk polymers with higher levels of functional groups, including PEO and other biologically relevant molecules such as cell adhesion peptides. Acknowledgements The authors gratefully acknowledge the technical assistance of J. Tan, S. Yonson and M. Bergeron. As well, we would like to thank Dr. Farid Bensebaa for XPS analysis. Funding from the Natural Sciences and Engineering Research Council and the University of Ottawa is gratefully acknowledged.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24.
L. Y. Yung and S. L. Cooper, Biomaterials 19, 31 (1998). J. N. Baumgartner, C. Z. Yang and S. L. Cooper, Biomaterials 18, 831 (1997). F. Lim, X. H. Yu and S. L. Cooper, Biomaterials 14, 537 (1993). J. H. Silver, H. B. Lin and S. L. Cooper, Biomaterials 14, 834 (1993). J. P. Santerre, P. ten Hove, N. H. VanderKamp and J. L. Brash, J. Biomed. Mater. Res. 26, 39 (1992). W. G. McClung, D. L. Clapper, S. P. Hu and J. L. Brash, Biomaterials 22, 1919 (2001). G. A. Skarja and J. L. Brash, J. Biomed. Mater. Res. 34, 439 (1997). H. B. Lin, W. Sun, D. F Mosher, C. Garcia-Echeverria, K. Schaufelberger, P. I. Lelkes and S. L. Cooper, J. Biomed. Mater. Res. 28, 329 (1994). H. B. Lin, K. B. Lewis, D. Leach-Scampavia, B. D. Ratner and S. L. Cooper, J. Biomater. Sci. Polymer Edn 4, 183 (1993). K. A. Woodhouse and J. L. Brash, Biomaterials 13, 1103 (1992). J. P. Santerre, P. ten Hove and J. L. Brash, J. Biomed. Mater. Res. 26, 1003 (1992). T. G. Grasel, D. G. Castner, B. D. Ratner and S. L. Cooper, J. Biomed. Mater. Res. 24, 605 (1990). J. H. Lee, J. Kopecek and J. D. Andrade, J. Biomed. Mater. Res. 23, 351 (1989). J. H. Lee, H. B. Lee and J. D. Andrade, Prog. Polym. Sci. 20, 1043 (1995). E. W. Merrill, J. Biomater. Sci. Polymer Edn 5, 1 (1993). P. Banerjee, D. J. Irvine, A. M. Mayes and L. G. Griffith, J. Biomed. Mater. Res. 50, 331 (2000). S. Corneillie, P. N. Lan, E. Schacht, M. Davies, A. Shard, R. Green, S. Denyer, M. Wassall, H. Whitfield and S. Choong, Polymer International 46, 251 (1998). J. H. Silver, C. W. Myers, F. Lim and S. L. Cooper, Biomaterials 15, 695 (1994). J. Tan and J. L. Brash, Trans. Soc. Biomaterials. 23, 499 (2001). D. Tomalia, M. Naylor and W. Goddard, Angew. Chem. Inter. Ed. 29, 138 (1990). J. Haensler and F. C. Szoka, Jr., Bioconj. Chem. 4, 372 (1993). A. Bielinska, J. F. Lakowska-Latallo, J. Johnson, D. A. Tomalia and J. R. J. Baker, Nucleic Acid Res. 24, 2176 (1996). J. F. Kukowska-Latallo, A. U. Belinska, J. Johnson, R. Splindler, D. A. Tomalia and J. R. J. Baker, Proc. Natl. Acad. Sci. USA 93, 4897 (1996). M. X. Tang, C. T. Refemann and F. C. Szoka, Jr., Bioconj. Chem. 7, 703 (1996).
Novel dendrimer based polyurethanes for PEO incorporation
365
25. J. P. Defoort, B. Nardelli, W. Huang, D. D. Ho and J. P. Tam, Proc. Natl. Acad. Sci. USA 89, 3879 (1992). 26. V. V. Martin, W. H. Ralston, M. R. Hynes and J. F. W. Keana, Bioconj. Chem. 6, 616 (1995). 27. R. Roy, D. Zanini, S. J. Meunier and A. Romanowska, J. Chem. Soc. Series Chem. Commun. 24, 1869 (1993). 28. C. Kojima, K. Kono, K. Maruyama and T. Takagishi, Bioconj. Chem. 11, 910 (2000). 29. S. Capaldi, R. C. Getts and S. D. Jayasena, Nucleic Acids Research 28, E21 (2000). 30. M. Kawase, T. Shiomi, H. Matsui, Y. Ouji, S. Higashiyama, T. Tsutsui and K. Yagi, J. Biomed. Mater. Res. 54, 519 (2001). 31. J. F. G. A. Jansen, E. M. M. de Brabander-van den Berg and E. W. Meijer, Science 266, 1226 (1994). 32. R. Cornelius and J. L. Brash, J. Biomater. Sci. Polymer Edn 4, 291 (1993). 33. K. D. Park, A. Z. Piao, H. Jacobs, T. Okano and S. W. Kim, J. Polymer Sci. 29, 1725 (1991).
This page intentionally left blank
Identification of biodegradation products formed by L -phenylalanine based segmented polyurethaneureas S. L. ELLIOTT 1 , J. D. FROMSTEIN 1 , J. P. SANTERRE 2 and K. A. WOODHOUSE 1,∗ 1 Department
of Chemical Engineering and Applied Chemistry, University of Toronto, 200 College Street, Ontario, M5S 3E5, Canada 2 Faculty of Dentistry, University of Toronto, 12H Edward St., Ontario M56 166, Canada Received 29 November 2001; accepted 2 April 2002 Abstract—The degradation of novel biodegradable segmented polyurethanes was investigated with a view to determining the cleavage points within the polymer backbones targeted by the enzyme chymotrypsin. While the materials were developed with specific enzyme cleavage sites designed into the polymer chains, the nature of their degradation had not yet been determined. In this work, two segmented polyurethaneureas containing L -phenylalanine residues in the chain extender and two control polymers were subjected to degradation in the presence of chymotrypsin. Samples were collected for analysis over a time period from 1 day to 8 weeks. The degradation products from these materials were isolated using solid phase extraction and reversed phase high pressure liquid chromatography, and identified using mass and tandem mass spectrometry. Three hard segment related degradation products were identified and provide important insight into the polyurethane backbone cleavage sites. Cleavage of urea, ester and urethane bonds were observed. The results confirmed that chymotrypsin was able to cleave ester bonds adjacent to phenylalanine residues contained within the novel chain extender. Key words: Polyurethaneurea; biodegradation; chymotrypsin; L -phenylalanine; mass spectrometry.
INTRODUCTION
Biodegradable polymeric materials are currently of great interest for use as medical implants and drug delivery vehicles. These materials, designed to degrade into non-toxic components over a pre-determined period can either degrade once they have completed their intended function, or may achieve their goal via degradation itself [1– 6]. Synthetic bioadsorbable sutures, invented in the 1960s are an example of a commonly used degradable polymeric material [7]. The biodegradable ∗ To
whom correspondence should be addressed. E-mail:
[email protected]
368
S. L. Elliott et al.
approach has the potential to alleviate a number of the problems associated with non-degradable implants, such as long-term safety and implant removal [1, 3, 5, 6]. Several recent candidates for use in biodegradable polymeric materials include poly(ortho esters), polyanhydrides, polyphosphazenes, poly-ε-caprolactone and its copolymers, polyglycolic acid (PGA), polylactic acid (PLA), polydioxane, and degradable segmented polyurethanes [8– 13]. Segmented polyurethanes are an important class of biomaterials with a long history of use as medical implants due to their excellent physical properties and relatively good biocompatibility [14]. A broad range of physical properties can be achieved by varying the chemistries and molecular weights of the various components, and through manipulation of the ratios in which they are reacted [14]. Material properties can range from very brittle and hard materials to soft, tacky, viscous ones [14]. While relatively few biodegradable segmented polyurethanes exist, due in part to concerns about the toxicity of potential degradation products, the development of an isocyanate based on the amino acid lysine opened up the possibility of synthesizing polyurethanes whose ultimate degradation products include lysine [4, 15, 16]. The residence time of a polymeric implant in an organism can be controlled by changing the quantity of cleavable links present in the polymeric chain [17]. The most common method of introducing hydrolysable linkages into polyurethanes has been to incorporate soft segments such as polylactides and ε-polycaprolactone into the backbone [18]. An alternative is to utilize hydrolysable groups in the hard segment, particularly in the chain extender. If hydrolysable hard segments are used, a wider variety of soft segment chemistries can be introduced [15]. This increases the range of material properties attainable without altering the degradable portion of the polyurethane [15]. In addition to simple hydrolysis, enzyme mediated hydrolysis has long been known to play a role in polyurethane biodegradation [19– 21]. Such enzymatic degradation may be encouraged through the incorporation of suitable amino acids into the polymer’s backbone [22]. Kartvelishvili et al. [23] synthesized degradable non-segmented poly(ester urethane)s containing L-phenylalanine, and found that these materials exhibited an enhanced susceptibility to chymotrypsin mediated degradation. Chymotrypsin-like serine proteinases are implicated in a wide variety of pathological states including inflammation (cathepsin G) and cardiomyopathy (mast cell chymase) [24, 25]. Chymase is also released in response to allergens or other challenges [26]. Chymotrypsin hydrolyzes peptide and ester bonds after phenylalanine residues [27, 28]. Pkhakadze et al. [29] synthesized polyurethanes containing a chain extender based on symmetric esters of phenylalanine and glycols. They found that chymotrypsin enhanced the degradation of their materials. Huang et al. [27] and Katsarava et al. [30] also found that chymotrypsin improved the degradation of L-phenylalanine containing polymers. Recently, we synthesized a family of segmented polyurethanes containing an L-phenylalanine based chain extender [15]. In these studies, the phenylalanine
Phenylalanine polyurethane degradation products
369
containing materials demonstrated what appeared to be preferential degradation in the presence of chymotrypsin [15]. In addition, these polyurethanes and their degradation products were not found to elicit a significant cytotoxic response, thereby demonstrating potential for use in biomedical applications [31]. Knowledge of the backbone cleavage sites will allow for the design of materials with controllable residence time in the body, based on the number of these sites present in the polyurethane. In the current work, two of the polyurethaneureas developed by Skarja and Woodhouse [32] were degraded in the presence of chymotrypsin. Degradation products were subsequently isolated using solid phase extraction and reversed-phase high pressure liquid chromatography (HPLC), and identified using tandem mass spectrometry (MS /MS). The identities of the polyurethane fragments were then used to determine the suspected cleavage sites in the polyurethane backbone.
MATERIALS AND METHODS
Materials Polycaprolactone diol (molecular weight 530), polyethylene oxide (molecular weight 600) and 1,4 cyclohexane dimethanol were obtained from Aldrich, Milwaukee, WI, USA. The soft segment diols were placed in a vacuum oven at 60◦ C for 48 h to remove residual water prior to reaction. 2,6-diisocyanato methylcaproate (LDI) was supplied by Kyowa Hakko Kogyo Co. Ltd., Tokyo, Japan and was distilled under vacuum prior to use. Stannous 2-ethylhexanoate was obtained from Sigma, St. Louis, MO, USA. Polyurethane solvents N,N-dimethyl formamide (anhydrous grade DMF) and chloroform were obtained from Aldrich and ACP Chemicals Inc., Montreal, PQ, Canada, respectively. Synthesis of polyurethanes Polyurethanes were synthesized via the two-step reaction procedure of Skarja and Woodhouse [32], using LDI as the diisocyanate. One of two soft segments was used for each polyurethane: polycaprolactone diol of molecular weight 530 (PCL) or polyethylene oxide of molecular weight 600 (PEO). An L-phenylalanine based chain extender (PHE), developed and characterized by Skarja and Woodhouse [32], was also used. This PHE compound is comprised of a 1,4 cyclohexane dimethanol (CDM) group covalently linked on either side by an L-phenylalanine residue. CDM was used as a phenylalanine-free control chain extender. All reactions were carried out under nitrogen with a stoichiometry of 2 : 1 : 1 of LDI : soft segment : chain extender. The prepolymer reaction was catalyzed by stannous-2-ethylhexanoate and allowed to proceed for 2 h at 85◦ C. The chain extension reaction was allowed to continue for 18 h at <60◦ C for PHE and 85◦ C for CDM. The polymers were precipitated in saturated aqueous KCl, immersed in 37◦ C distilled water for 48 h,
370
S. L. Elliott et al.
and subsequently dried under vacuum at 60◦ C for 48 h. Finally, the polyurethanes were stored in a dessiccator until use. The nomenclature used to identify the polyurethanes is based on the reactants. Since all four polyurethanes contain LDI, this was not incorporated into the nomenclature. Each nomenclature sequence is comprised of three letters representing the soft segment followed by three letters referring to the chain extender. For example, PCL /PHE refers to a polyurethane synthesized from LDI, 530 molecular weight PCL, and chain extended with the L-phenylalanine based chain extender. A total of four polyurethanes were synthesized: PCL /PHE, PEO /PHE, PCL /CDM, and PEO /CDM. Polyurethane characterization Gel permeation chromatography (GPC) was employed to determine polyurethane molecular weight and polydispersity. The mobile phase consisted of 0.05 M lithium bromide (LiBr) (Aldrich) in DMF at a flowrate of 1 ml/ min and the column bank was heated to 80◦ C. Samples were prepared at a concentration of 0.2% w /v in mobile phase. The number and weight average molecular weights were determined from the retention time data with Waters Baseline™ software (Waters Chromatography, Mississauga, ON, Canada) using a calibration curve generated with polystyrene standards (TSOH Corporation, Tokyo, Japan). Polymer samples were sent to Galbraith Laboratories, Inc. (Knoxville, TN) for bulk carbon, nitrogen, and hydrogen elemental analysis. Biodegradation experiments An in vitro biodegradation study was carried out over 8 weeks to determine the degradation products of the synthesized polyurethanes. Polyurethane solutions were prepared by dissolving the polymers in chloroform at 10% w / v. The polyurethanes were then coated onto hollow glass tubes (∼130 mm × 2.0 mm I.D. × 4.0 mm O.D.) using a dip-coating technique previously reported by Santerre et al. [20]. Next, the coated tubes were dried in a convection oven at 60◦ C for 18 h. This coating procedure was repeated a total of three times, with the final coat dried under vacuum at 60◦ C for 24 h. The coated tubes were snapped into 20 mm long pieces. Subsequently, the coated tube segments (or uncoated glass tube controls) were placed into separate 3 ml autoclaved glass screw cap vials under sterile conditions. The samples were incubated at 37◦ C in 2 ml of tris buffer (0.036 M tris (BioBasic Inc., Scarborough, ON), 0.045 M calcium chloride (ACP Chemicals Inc.), and 0.02% w /v sodium azide (Sigma) in deionized water, adjusted to a pH of 8 using 1 N HCL or 1 N NaOH (Sigma)) containing 500 U /ml chymotrypsin (Sigma). Uncoated glass tubes were also incubated in the same chymotrypsin solution and used as a control for HPLC purposes (see below). Enzyme was replenished daily. Incubation solution samples were withdrawn after 1 day, 3 days, 2 weeks, 5 weeks,
Phenylalanine polyurethane degradation products
371
and 8 weeks. Samples were stored at −80◦ C until processing for HPLC and MS analysis. Degradation product isolation Prior to product separation and identification, residual enzyme (chymotrypsin) was removed from the incubation solution samples. This process was necessary since the chymotrypsin could interfere with the accurate detection of the degradation products in the high performance liquid chromatography (HPLC) columns, and because proteins have a tendency to aggregate and then later precipitate during the gradient run, thereby causing additional difficulties in data acquisition [33, 34]. Enzyme was removed from the incubation solution using solid phase extraction (SPE) with 3 ml Waters HLB Oasis extraction cartridges (Waters Corporation, Bedford, MA). During SPE, the cartridge was conditioned with a methanol (HPLC grade) rinse followed by a water rinse. The sample was then loaded into the cartridge, allowing the degradation products to bind to the sorbent bed. Next, the bed was rinsed with a 5% methanol in water solution, and finally, the degradation products were eluted in a methanol fraction. The eluted product was dried under nitrogen, reconstituted in the HPLC mobile phase, and filtered using a 5000 MWCO centrifuge filter. High performance liquid chromatography (HPLC) A Waters HPLC system was used for chromatographic separation of degradation products. It consisted of a 600E multisolvent delivery system, a U6K injector (Waters, Milford, MA) and a 996 photodiode array (PDA) detector. A Millennium 2010 chromatography manager system (Waters) was used to acquire and process the data. A Phenomonex Luna 5µ C18 (2) steel cartridge column (Phenomonex, Torrance, CA) was used to run all the chromatographic separation experiments. All the mobile phases were filtered and degassed by ultrasonication (prior to run) and a helium sparge (during HPLC). The mobile phase consisted of methanol (solvent A) and 2 mM ammonium acetate (BDH Inc., Toronto, ON) adjusted to pH 3.0 (solvent B). HPLC grade water was used to flush the column of the buffer salts between injections. The flowrate was maintained at 1 ml/min. HPLC fractions of interest were collected and freeze dried overnight. Samples were reconstituted with 200 µl of methanol (HPLC grade) prior to injection into the mass spectrometer. Mass spectrometry Ion spray mass spectra were acquired on an API III+ triple mass spectrometer (PE Sciex, Thornhill, Canada) through the Carbohydrate Laboratory, University of Toronto. The third quadrupole was used as a mass analyzer to obtain the molecular mass analytes in MS and MS /MS analysis. In MS /MS analysis, the
372
S. L. Elliott et al.
second quadrupole was used as a reaction region for collision-induced dissociation of a selected parent ion by the first quadrupole and the fragmentation of the fragment ions are assigned by the third quadrupole. Using this technique, the MS /MS spectrum of a selected parent ion was obtained. The degree of fragmentation was adjusted by the pressure of the collision gas (argon) in the second quadrupole and was set such that the collision gas target thickness was 1.3 × 1015 molecules/ cm2 . The HPLC fractions (20 µl) were injected through a 7125 injector (Rheodyne, CA). Samples were introduced into the electrospray ionization source with HPLC grade methanol at a flow rate of 30 µl/ min using an HPLC solvent delivery pump. The electrospray needle was operated at 4.8 kV and the orifice voltage was 45 V. Full scan mass spectra were acquired over the mass/charge (m/z) ratio range 100– 800. The data were acquired at a 0.002 s dwell time and 0.2 atomic mass unit (amu) step size.
RESULTS AND DISCUSSION
Polyurethanes Gel permeation chromatography (GPC) data, provided in Table 1, are reported as polystyrene equivalent molecular weights. Generally, the polycaprolactone based polyurethanes exhibited higher number and weight average molecular weights than the polyethylene oxide based polyurethanes. This difference is likely related to differences in the hydrodynamic volumes of PCL versus PEO since the individual molecular weights of the diols, 530 and 600 respectively, were quite close. PCL is crystalline and relatively non-polar while PEO is non-crystalline and relatively hydrophilic [35, 36]. The GPC analysis was carried out in DMF, a polar solvent; hence the chain mobility of PEO would likely be greater than that of PCL. For each polyurethane, the polydispersity values were near 2. These values are consistent with those reported in the literature for segmented polyurethanes [14, 20, 21]. Elemental analysis for carbon, hydrogen, and nitrogen was also conducted on the polyurethanes, and this data is also summarized in Table 1. The measured Table 1. Polyurethane molecular weight and compositional data, obtained using GPC and elemental analysis. Values reported in polystyrene equivalent weights for the GPC Polyurethane Mn
PCL/PHE PCL/CDM PEO/PHE PEO/CDM
Mw
(×103 )
(×103 )
37.5 31 18 18.5
71.5 62 40.5 41
Polydispersity % Carbon
% Hydrogen
% Nitrogen
Actual Theory Actual Theory Actual Theory 2.0 2.0 2.2 2.2
59.9 57.0 56.0 52.5
62.6 59.7 57.9 54.0
7.9 8.1 7.8 8.2
7.6 8.0 7.6 8.0
5.9 5.0 5.8 5.1
6.1 5.1 5.8 4.8
Phenylalanine polyurethane degradation products
373
elemental compositions were all within 5% of the theoretical predicted atomic weight percents. High performance liquid chromatography (HPLC) separation There were several complications in analyzing the HPLC results. First, since there were no UV absorbing groups in the polyurethane backbone other than Lphenylalanine, the only degradation products that could be observed using the PDA detector contained the L-phenylalanine residue. Since the latter group was of particular interest in this study, the limitations of the PDA were not prohibitive to the investigation. Second, it was often difficult to determine whether the HPLC peaks were related to polyurethane degradation products or were the result of residual protein material derived from the enzyme. HPLC chromatograms showed that not all the residual protein was removed during the purification process. Chymotrypsin contains L-phenylalanine and tyrosine residues, which have maximum absorbances at 260 nm and 275 nm respectively. Thus, it is likely that HPLC peaks with absorbance spectra having double peaks for maximum absorbance at 260 nm and 275 nm or a single peak at 275 nm, are most likely due to the enzyme. Polyurethane degradation products will have a single peak for maximum absorbance of 260 nm. Unfortunately, HPLC peaks with a single maximum at 260 nm do not necessarily indicate a polyurethane degradation product since some enzyme fragments may contain only phenylalanine and not tyrosine residues. When deciding which peaks may have been polyurethane degradation products, the HPLC chromatograms for the different polyurethanes and the glass control were compared. The peaks that appeared in chromatograms for the polyurethanes containing PHE, and not for either the polyurethanes containing CDM or the incubation solutions from enzyme/ glass controls, were considered to contain degradation products of the PHE based polyurethane. Figure 1 shows HPLC chromatograms plotted at 260 nm for 1 day samples. Many of the peaks appear to be enzyme related since the glass control (Fig. 1e) shows peaks similar to the chromatograms for the polyurethanes. Both the PHE based polyurethanes (chromatograms 1a and b in Fig. 1) show small peaks at 29– 30 min with a maximum absorbance of 260 nm. Since these peaks were not seen for the polyurethanes synthesized with CDM (Fig. 1c and d) or the glass control (Fig. 1e), it was concluded that these peaks represented degradation products associated with the PHE containing polyurethanes. The 29– 30 min fraction was therefore analyzed further using MS techniques. Identification of products by mass spectrometry The possible presence of cyclohexane dimethanol (CDM) as a degradation product was of interest since the ester bond adjacent to the L-phenylalanine residue would need to be cleaved in the PHE based polyurethanes to create this fragment. These bonds were incorporated into the backbone with the expectation that they would
374
S. L. Elliott et al.
Figure 1. HPLC chromatogram for degradation solutions at 260 nm following 1 day of incubation in the presence of chymotrypsin (500 units/ ml, pH 8, 37◦ C). (a) PCL/PHE, (b) PEO/PHE, (c) PCL/CDM, (d) PEO/CDM, (e) glass control. On the absorbance axis, each division represents 0.01 absorbance units. The region of interest indicates peaks that are unique to the PHE containing polyurethanes.
be susceptible to enzymatic cleavage by chymotrypsin [27, 31]. The cleavage sites required for the production of CDM as a degradation product for both the PHE and CDM based materials are shown in Fig. 2. For the control CDM extended polyurethanes, urethane bonds would need to be cleaved for CDM to appear as a degradation product. Pure CDM was run through the MS system. Its protonated molecular ion ([M+H]+1 ) was found to have a mass to charge ratio (m/z) of 145. In order to assist with the identification of degradation product structure, tandem mass spectrometry (MS /MS) was conducted. The MS /MS spectrum of the protonated CDM peak at m/z 145 is shown in Fig. 3. The strongest signal is at m/z 67. This peak and the smaller ones at m/z 65, 55, 54.4, 42.8, 41, and 29 result from the breakdown of the cyclohexane ring. There are also strong signals at m/z 127, 109, and 81 which
Phenylalanine polyurethane degradation products
375
Figure 2. Diagram indicating the bonds that would have to be cleaved in order to produce CDM as a degradation product for both the PHE and CDM extended polyurethanes.
Figure 3. MS/ MS spectrum of CDM, parent ion m/z 145. The inset table indicates the proposed structures of the resultant fragments.
provide a fingerprint pattern for CDM. The proposed structures of the fragments are summarized in the inset table in Fig. 3. Since CDM does not strongly absorb energy in the UV range, liquid chromatography/ mass spectrometry (LC /MS) was used to determine the elution time of this compound. From CDM standards run on LC /MS, it was determined that the CDM should appear within the retention time range of 14– 20 min on the HPLC system [37].
376
S. L. Elliott et al.
HPLC fractions were collected between 14 and 20 min and analyzed for the presence of CDM using MS. Those spectra with a m/z of 145 were further analyzed using MS /MS to confirm the compound’s structure. The MS /MS spectra were compared with the MS /MS spectrum of CDM (Fig. 3). PEO /PHE and PEO /CDM samples after three days incubation did not contain CDM. Since it was previously found that the PEO based polyurethanes degraded faster than the PCL based polyurethanes [31], it was not considered necessary to run MS on the PCL samples for the one day timepoint. Eight-week samples for all four of the polyurethanes contained CDM. The parent ion at m/z 145 for these materials was analyzed using MS /MS, and all four polyurethanes resulted in the same peaks. Figure 4 shows representative spectra, obtained from PCL /PHE (Fig. 4a) and PCL /CDM (Fig. 4b) following 8 weeks of degradation in the chymotrypsin solution (500 units/ ml, 37◦ C). The mass ion peaks labelled with an * indicate m/z values in common with those seen in Fig. 3, the MS /MS spectrum of the pure CDM parent ion. The presence of pure CDM in the incubation solutions for all four polyurethanes indicates that the bonds shown in Fig. 2 were cleaved, either by simple chemical hydrolysis or through enzyme mediated cleavage. The CDM containing polyurethanes were intended to serve as a control against the chymotrypsin labile L-phenylalanine chain extender for this research; however in this case the control did not aid in the interpretation of the data since the bonds cleaved are different for the CDM (urethane) and the PHE (ester) based polyurethanes. Urethane bonds are considered to be relatively stable towards hydrolysis in aqueous conditions at 37◦ C [38]. Thus, the urethane cleavage seen in this study may be a result of enzyme-mediated hydrolysis, as chymotrypsin is known to recognize many substrates other than Lphenylalanine including aliphatic rings [28]. Previous degradation studies using these polyurethanes demonstrated that both the CDM and PHE based materials demonstrated significant mass loss in the presence of chymotrypsin [31]. The PHE polyurethanes exhibited a higher degree of mass loss than their corresponding controls [31]. These higher mass loss values were not seen to the same degree when the materials were incubated in enzyme-free buffer, indicating that the presence of PHE in the polyurethanes increases their susceptibility to enzymatic attack preferentially over general hydrolysis [31]. In future work, mass spectrometry and MS /MS analysis of the polyurethanes following degradation in buffer could be used to determine whether CDM is created as a degradation product when hydrolysis alone is responsible for degradation of the polyurethanes. As noted earlier (Fig. 1) the HPLC fraction at 29 to 30 min was also of interest and was analyzed using MS and MS /MS. Fractions from degradation samples of PCL /PHE, PEO /PHE, and the glass control after one day of incubation were studied using MS. The spectra for PCL /PHE (Fig. 5a) and PEO /PHE contained the same peaks, while the spectrum for the glass control was significantly different (Fig. 5b). Four peaks with m/z values of 511, 528, 543, and 560 were found for both PCL /PHE and PEO /PHE. The molecular ion with a m/z of 511 was related to the molecular ion with m/z 528 by the substitution of the H+ with a NH+ 4 ion.
Phenylalanine polyurethane degradation products
377
Figure 4. MS/ MS spectra of (a) PCL/PHE and (b) PCL/CDM, parent ion m/z 145. HPLC fraction 14– 20 min, incubated in chymotrypsin (500 units/ ml, pH 8, 37◦ C) for 8 weeks. Peaks marked with a * represent peaks in common with CDM (Fig. 3). These spectra are characteristic of those obtained for PEO/CDM and PEO/PHE as well.
The same relationship existed between the peaks with m/z values of 543 and 560. The NH+ 4 group is present due to the ammonium acetate buffer used in the HPLC gradient method.
378
S. L. Elliott et al.
Figure 5. MS spectra of HPLC fraction 29– 30 min following 1 day incubation in chymotrypsin (500 units/ ml, pH 8, 37◦ C). (a) PCL/PHE, representative of PEO/PHE as well, (b) Glass control.
It was expected that the m/z 511 and 543 (Fig. 5a) fragments contain L-phenylalanine since they absorbed UV at 260 nm. Thus, the MS /MS of these molecules should contain fragments in common with PHE. The MS analysis for pure PHE gave a m/z value of 439.2 for [M + H]+1 . MS /MS was undertaken on this molecular ion in order to generate a fingerprint analysis that could be used to identify the
Phenylalanine polyurethane degradation products
379
Table 2. Proposed structure of the MS/ MS fragments of m/z 439 for PHE Fragment m/z
Proposed structure associated with the mass ion
439 (Parent ion)
292
274
166
149
120
103 67
components in the polyurethane derived products (Fig. 5a). Table 2 summarizes the proposed structures of the fragments for pure PHE. While structures were not determined for m/z values of 119, 109, 107, 81, 55, and 43, it is suspected that they were related to fragments of the cyclohexane and benzene rings. MS /MS was performed on the products common to PCL /PHE and PEO /PHE (i.e. m/z 511 and m/z 543). Figure 6a shows the MS /MS of m/z 511 for PCL /PHE while Fig. 6b shows the MS /MS of m/z 511 for PEO /PHE. They clearly have the
380
S. L. Elliott et al.
Table 3. Summary of MS/MS fragments for parent ions m/z 511 and 543 for the HPLC fraction between 29– 30 min collected from PCL/PHE and PEO/PHE following 1day of incubation in chymotrypsin (500 units/ ml, 37◦ C). Fragments in common with the PHE chain extender are shown in bold m/z 511 PCL/PHE
m/z 543 PEO/PHE
PCL/PHE
84.0
84.2
102.8 120.2
103.4 120.2
102.8 120.2 127.2
130.8 149.2 166.2
131.0 149.2 166.0
274.4 300.2
273.8 300.0
320.0 328.2
320.0 328.2
346.0
346.0
447.8 465.2 493.6
448.2 464.8 493.6
511.2
510.8
PEO/PHE 84.0 100.0 120.2 127.0
149.0 166.2 187.2 274.0 300.2
149.2 166.2 187.2 274.5 300.2 306.2
332.0
332.4
352.0 378.2
352.4 378.0
497.6 511 542.8
497.6 511 543.4
same fragmentation pattern and therefore the same product was found for the two different polyurethanes. Figure 7 shows the MS /MS of m/z 543 for (a) PCL /PHE and (b) PEO /PHE. Again, the fragmentation pattern was identical for the products isolated from the two polyurethanes, indicating that these products were derived from the hard segment of the polyurethane, which was common to both polymers. There were many MS /MS fragments in common between the two m/z parent ions (m/z = 511 and 543), indicating that these products could be related. Table 3 summarizes the dominant fragments from the MS /MS spectra in Figs 6 and 7. Fragments in common with PHE are shown in bold. It should be noted that while products similar to the breakdown fragments of PHE were seen, pure PHE (m/z 439.2) did not appear among the isolated products. Based on the MS /MS fragments of the pure PHE (Table 2) it was possible to determine almost all of the fragments for the m/z 511 peak. The bonds that were cleaved in the polyurethane backbone to produce the product with a m/z of 511 were urea bonds, as indicated
Phenylalanine polyurethane degradation products
381
Figure 6. MS/ MS spectra for parent ion m/z 511, HPLC fraction 29– 30 min, following incubation for 1 day in chymotrypsin (500 units/ ml, pH 8, 37◦ C). (a) PCL/PHE, (b) PEO/PHE.
in Fig. 8. Proposed structures for the fragments from the product with the m/z of 511 are shown in Fig. 9. The fragments with m/z 166, 149, and 120 correspond to characteristic fragments from PHE (Table 2). Based on this observation, it can be stated that chymotrypsin catalyzes the cleavage of urea bonds adjacent to
382
S. L. Elliott et al.
Figure 7. MS/ MS spectra for parent ion m/z 543, HPLC fraction 29– 30 min, following incubation for 1 day in chymotrypsin (500 units/ ml, pH 8, 37◦ C) of degradation. (a) PCL/PHE, (b) PEO/PHE. L-phenylalanine residues (shown in Fig. 8). It is believed that the cleavage of urea bonds adjacent to L-phenylalanine residues has not been observed previously [27]. Taking into consideration the fragments in common with the m/z 511 molecule (Table 3) and the difference in mass between the m/z of 543 and m/z 511, it
Phenylalanine polyurethane degradation products
383
Figure 8. Bond cleavage necessary to produce the proposed structure for the degradation product with m/z 511.
was proposed that the m/z 543 consisted of the m/z 511 molecule complexed with methanol (molecular weight 32). The methanol complex is a by-product of the HPLC /MS solvent. The addition of a methanol group is enough to produce a different fragmentation pattern for the m/z 543 although the MS fragmentation pattern contains many of the same fragments to those found for m/z 511. As with m/z 511, the fragment of m/z 166 is the most dominant peak and m/z 120 is also a strong peak. The proposed fragmentation pattern for m/z 543 is shown in Fig. 10. Many of the fragments (m/z 497, 378, 352, and 332) are related to those proposed for m/z 511 with the attachment of a methanol ion [CH3 OH+ 2 ]. The cleavage sites for the degradation product with m/z 543 are the same as those for the degradation product with m/z 511 as shown in Fig. 8. In future work, the proposed structure of the m/z 543 peak could be verified by running the MS analysis using ethanol rather than methanol as the solvent. Since the product collected between 29– 30 min was found to be a polyurethane degradation product associated with the L-phenylalanine portion of the hard segment, the same HPLC fraction was collected and analyzed for different samples and different time points using MS. None of the glass control samples or CDM based samples showed the peaks of interest (i.e. m/z 511, 528, 543, and 560). All the PHE based samples tested contained at least one of these peaks although the intensity of these peaks at later time points (8 weeks) was greatly diminished. Thus, the products at m/z 511 and m/z 543 are produced early in the incubation and are most likely degraded further upon continued exposure to chymotrypsin. The limited presence of this product may also be the result of a surface limiting effect as enzymatic degradation is well known to preferentially progress from the surface of polymeric materials [39].
CONCLUSIONS
The degradation of four biodegradable segmented polyurethanes was evaluated in order to confirm the cleavage of the hard segment during enzymatic degradation. Two degradation products were identified in addition to the presence of cyclohexane dimethanol (CDM). CDM was identified both for the polyurethanes containing
384
S. L. Elliott et al.
Figure 9. Proposed fragmentation pattern for the degradation product with m/z 511 for PCL/PHE and PEO/PHE, HPLC fraction 29– 30 min, incubated for 1 day in chymotrypsin (500 units/ ml, pH 8, 37◦ C).
a phenylalanine (PHE) based chain extender and for the control polymers that only contain CDM. These results were expected for the PHE based materials since chymotrypsin has been reported to cleave ester bonds after L-phenylalanine residues [27]. However, the urethane bonds of the CDM based polyurethanes
Phenylalanine polyurethane degradation products
385
Figure 10. Proposed fragmentation pattern for the degradation product with m/z 543 for PCL/PHE and PEO/PHE, HPLC fraction 29– 30 min, incubated for 1 day in chymotrypsin (500 units/ ml, pH 8, 37◦ C).
were not expected to be susceptible to hydrolysis under standard physiological conditions [38]. Therefore it was concluded that this cleavage may be the result of chymotrypsin mediated hydrolysis as this enzyme recognizes many substrates including non-aromatic rings [28].
386
S. L. Elliott et al.
The results of the MS /MS analysis indicated that chymotrypsin may act to cleave urea bonds adjacent to L-phenylalanine residues. This is a significant finding since it confirms that the polyurethanes are susceptible to selective enzymatic degradation in the hard segment. Traditionally, this domain of the polyurethane has been considered a relatively stable group. The materials used in this study, however, were specifically developed to encourage degradation of the hard segment rather than relying solely on degradation of the soft segment. Hence, the results of this study confirm that this design goal was achieved. The cleavage of urea bonds by chymotrypsin is an important finding as it contradicts results of a previous study with similar chemistry that found that urea bonds adjacent to L-phenylalanine residues were not cleaved [27]. However, since the level of chymotrypsin activity was not stated in the other study, it may be possible that the right conditions were not presented in order to degrade the urea bond. Future work on enzyme dose response will need to be carried out in order to quantify the extent of degradation associated with various enzyme concentrations. In addition, degradation studies in buffer alone should also be performed to determine which degradation products were formed by simple hydrolysis rather than by enzyme mediated degradation. MS analysis should also be repeated using ethanol as the solvent in order to confirm the results reported in this study. Acknowledgements The authors would like to acknowledge the support of the Natural Sciences and Engineering Research Council of Canada, Materials and Manufacturing Ontario, and the National Institutes of Health (USA).
REFERENCES 1. T. D. Hirt, P. Neuenschwander and U. W. Suter, Macromol. Chem. Phys. 197, 4253 (1996). 2. S. J. Huang, M. S. Roby, C. A. Macri et al., in: Biodegradable polymers and plastics, M. Vert, J. Feijen, A. Albertson et al. (Eds), p. 149. Royal Society of Chemistry, Montpellier, France (1992). 3. S. I. Ertel and J. Kohn, J. Biomed. Mater. Res. 28, 919 (1994). 4. R. Saad, T. Hirt, M. Welti et al., J. Biomed. Mater. Res. 36, 65 (1997). 5. R. Smith, C. Oliver and D. F. Williams, J. Biomed. Mater. Res. 21, 991 (1987). 6. D. F. Williams, Clin. Mater. 10, 9 (1992). 7. T. Nakamura, Y. Shimizu, T. Matsui et al., in: Degradation Phenomena on Polymeric Biomaterials, H. Planck, M. Dauner and M. Renardy (Eds), p. 153. Springer-Verlag, New York (1991). 8. J. Heller, R. Sparer and G. Zenther, in: Biodegradable Polymers as Drug Delivery Systems, M. Chasin and R. Langer (Eds), p. 121. Marcell Dekker, New York (1990). 9. M. Chasin, A. Domb, E. Ron et al., in: Biodegradable Polymers as Drug Delivery Systems, M. Chasin and R. Langer (Eds), p. 43. Marcell Dekker, New York (1990). 10. H. Allcock, in: Biodegradable Polymers as Drug Delivery Systems, M. Chasin and R. Langer (Eds), p. 163. Marcell Dekker, New York (1990). 11. C. Pitt, in: Biodegradable Polymers as Drug Delivery Systems, M. Chasin and R. Langer (Eds), p. 71. Marcell Dekker, New York (1990).
Phenylalanine polyurethane degradation products
387
12. D. Gilding and A. Reed, Polym. 20 (1979). 13. J. Ray, N. Doddi, D. Regula et al., Surg., Gynecol., Obstet. 153, 497 (1981). 14. N. M. K. Lamba, K. A. Woodhouse and S. L. Cooper, Polyurethanes in Biomedical Applications. CRC Press, New York (1998). 15. G. A. Skarja and K. A. Woodhouse, J. Appl. Polym. Sci. 75, 1522 (2000). 16. P. Bruin, G. J. Veenstra, A. J. Nijenhuis et al., Macromol. Chem., Rapid Commun. 9, 589 (1988). 17. B. Masar, P. Cefelin, T. Lipatova et al., J. Polym. Sci., Polym. Symp. 66 (1979). 18. R. F. Storey, J. S. Wiggins and A. D. Puckett, J. Polym. Sci. A: Polym. Chem. 32, 3245 (1994). 19. Y. W. Tang, R. S. Labow and J. P. Santerre, J. Biomed. Mater. Res. 56, 516 (2001). 20. J. P. Santerre, R. A. Labow, D. G. Duguay et al., J. Biomed. Mater. Res. 28, 1187 (1994). 21. J. P. Santerre, R. S. Labow and G. A. Adams, J. Biomed. Mater. Res. 27, 97 (1993). 22. Z. Wirpsza, Polyurethanes: Chemistry, Technology and Applications. Ellis Horwood Limited, Toronto (1993). 23. T. Kartvelishvili, G. Tsitlanadze, L. Edilashvili et al., Macromol. Chem. Phys. 198, 1921 (1997). 24. D. Lomas, S. Stone, C. Llewellyn-Jones et al., J. Biol. Chem. 270, 23437 (1995). 25. N. Shiota, A. Fukamizu, S. Takai et al., J. Hypertension 15, 431 (1997). 26. A. Walls, in: Inflammatory Mechanisms in Asthma, S. Holgate and W. Busse (Eds), p. 89. Marcel Dekker, Inc., New York (1998). 27. S. Huang, D. Bansleven and J. Knox, J. Appl. Polym. Sci. 23, 429 (1979). 28. A. L. Lehninger, Biochemistry: The Molecular Basis of Cell Structure and Function. Worth Publishers, Inc., New York (1970). 29. G. Pkhakadze, M. Grigorieva, I. Gladir et al., J. Mater. Sci.: Mater. Med. 7, 265 (1996). 30. R. Katsarava, V. Beridze, N. Arabuli et al., J. Polym. Sci. A: Polym. Chem. 37, 391 (1999). 31. G. A. Skarja, The development and characterization of degradable, segmented polyurethanes containing amino acid-based chain extenders, PhD Thesis, Department of Chemical Engineering and Applied Chemistry, University of Toronto (2001). 32. G. A. Skarja and K. A. Woodhouse, J. Biomater. Sci. Polymer Edn 9, 271 (1998). 33. T. Hanai (Ed.), Liquid Chromatography in Biomedical Analysis. Elsevier, New York (1991). 34. W. Hancock and J. Sparrow (Eds), HPLC Analysis of Biological Compounds. A Laboratory Guide. Marcel Dekker, Inc., New York (1984). 35. F. Li, J. Hou, W. Zhu et al., J. Appl. Polym. Sci. 62, 631 (1996). 36. C. Hepburn, Polyurethane elastomers. Elsevier Science Publishing Co., Inc., New York (1992). 37. J. G. Yu, Oxidation of ultra high molecular weight polyethylene (UHMWPE) containing traces of cobalt ion, MASc Thesis, Department of Chemical Engineering and Applied Chemistry, University of Toronto (2000). 38. M. Lelah and S. Cooper, Polyurethanes in Medicine. CRC Press, Inc., Boca Raton, Florida (1986). 39. M. Vert, J. Feijen, A. Albertsson et al. (Eds), Biodegradable Polymers and Plastics. The Royal Society of Chemistry, Cambridge (1992).
This page intentionally left blank
Review Bioresorbable polymeric stents: current status and future promise ROBERT C. EBERHART 1,2,∗ , SHIH-HORNG SU 1,2 , KYTAI TRUONG NGUYEN 1 , MEITAL ZILBERMAN 1 , LIPING TANG 2 , KEVIN D. NELSON 2 and PETER FRENKEL 3 1 Department
of Surgery, University of Texas Southwestern Medical Center at Dallas, 5323 Harry Hines Boulevard, Dallas, TX 75390-9130, USA 2 Biomedical Engineering Program, University of Texas Southwestern Medical Center at Dallas and University of Texas at Arlington, Arlington, TX, USA 3 Department of Internal Medicine, University of Texas Southwestern Medical Center at Dallas, Dallas, TX, USA Received 2 January 2002; accepted 23 July 2002 Abstract—Metal stents and, more recently, polymer-coated metal stents are used to stabilize dissections, eliminate vessel recoil, and guide remodeling after balloon angioplasty and other treatments for arterial disease. Bioresorbable polymeric stents are being developed to improve the biocompatibility and the drug reservoir capacity of metal stents, and to offer a transient alternative to the permanent metallic stent implant. Following a brief review of metal stent technology, the emerging class of expandable, bioresorbable polymeric stents is described, with emphasis on developments in the authors’ laboratory. Key words: Stent; bioresorbable polymer; drug delivery; gene therapy; in-stent restenosis.
INTRODUCTION
Percutaneous transluminal coronary angioplasty (PTCA) is a standard treatment for focal arterial stenosis. Use of this ‘noninvasive’ treatment has rapidly expanded, since its introduction in 1977, to more than 500 000 cases per year in the United States alone. There has been a high restenosis rate of the treated segment following PTCA, up to 30%, with an estimated cost of $3.5 billion per year in the United ∗ To
whom correspondence should be addressed. Tel.: (1-214) 648-2052; e-mail:
[email protected]
390
R. C. Eberhart et al.
States alone in 1997 [1]. Metallic intracoronary stents were introduced to prevent arterial dissection, elastic recoil, and intimal hyperplasia associated with PTCA treatment. However, metal stents themselves induce an inflammatory response which can contribute to intimal hyperplasia [2, 3]. This problem has led to the intensive development of radiation-emitting stents and polymer- and ceramic-coated metal stents that can be loaded with various drugs, both strategies intended to reduce the inflammatory response to the metal stent.
METAL STENTS
Metal stents have evolved from relatively stiff, difficult to deploy structures designed to prevent wall dissection and collapse, to more flexible, open architectures which can negotiate tortuous channels and also overlay vessel branches whilst maintaining their patency. Early designs, including the Wallstent (Schneider), PalmazSchatz (Johnson & Johnson), Wiktor (Medtronic) and Gianturco-Roubin (Cook) stents, have given way to the Micro (AVE), Multilink (ACS) and other designs [4– 7]. The metals from which these stents are made are selected for strength, elasticity, and malleability or shape memory. Commonly used materials include stainless steel, tantalum and nitinol alloys [6, 8, 9]. Nitinol offers superelastic and thermal shape memory properties, which allow stent self-expansion at deployment, and thermally-induced collapse for theoretical removal procedures [10]. Several clinical trials have shown the benefit of stenting compared with PTCA alone. BENESTENT and STRESS, two major trials, were designed to assess the clinical outcome following PTCA with de novo (primary) stenting [6, 11, 12]. The BENESTENT trial found a reduction in the 6-month restenosis rate for the stent group compared with the control PTCA group (22% vs. 32%, respectively). The STRESS trial demonstrated a similar reduction; the rate of restenosis was 30% in the stent group and 42% in the control group. More recent trials, analyzing newer stents with those used in BENESTENT and STRESS, such as WEST, MUSIC and FINESS, have demonstrated lower restenosis rates [13]. Despite the improved results, stent-induced intimal hyperplasia and restenosis remain problematical, especially in complex procedures, with long lesions and multiple stent deployments [3]. Furthermore, the long-term (> 10 years) effects of metallic stents in humans are still unknown. Thus, other measures are required to resolve the restenosis issue. Two approaches to the management of residual stent-induced restenosis have emerged: stent polymer or ceramic coatings loaded with various pharmacologic agents [14– 18] and beta- or gamma-emitting radioisotopes, delivered via the stent itself or at stent implantation. Early clinical results suggest that pactitaxel, sirolimus, GPIIb/IIIa inhibitors, and other agents reduce short-term stent restenosis rates almost to zero (Table 1) [3, 13, 19]. These encouraging early results must be verified in longer-term trials. Low-dose radioisotope treatments with metal stents loaded with beta (90 Sr, 90 Y, 32 P) or gamma (192 Ir) emitters have also improved the
Bioresorbable polymeric stents
391
Table 1. Clinical trials and animal studies via coated stents and use of drugs with stenting Agents
TRIALS/studies
References
Paclitaxel, microtubular inhibitor Sirolimus, immunosuppressive agent Actinomycin D Metalloproteinase inhibitora
TAXUS I, ASPECT RAVEL, US trial Pig coronary arteries Porcine coronary arteries, porcine femoral arteries Porcine coronary model
[81, 82] [83, 84] [85] [86, 87]
Pig coronary arteries
[89, 90]
Porcine coronary model
[91]
TARGET, ADMIRAL CRUISE, ESPRIT Porcine coronary model Pig coronary arteries
[92, 93] [94– 96] [97] [98]
Pig coronary arteries NUGGET
[99] [100, 101]
Cytochalasin D, inhibitor of the actin microfilament formation Oligodeoxynucleotides to human transcription factor EGR-1 Antisense morpholino compound (AVI-4126) GPIIb/IIIa inhibitors: Abciximab Eptifibatide Methylprednisolone, anti-inflammatory Dextrose albumin microbubles containing c-Myc antisenses Cross-linked hyaluronan or chitosan Gold film coating a Opposing
[88]
responses were observed.
restenosis rate [19– 21]. Such stents have also been proposed for the local treatment of tumors and the prevention of excessive granulation tissue formation [22, 23]. Improvements in the restenosis rate notwithstanding, metal stents have other important limitations, including thrombogenicity, permanence, a limited potential for local drug delivery, and, for isotope-loaded stents, continuing radiation-induced damage [3, 12, 24]. Metal stent surfaces are moderately thrombogenic, requiring short-term antiplatelet or anticoagulant therapy. Metal stents are permanent implants. It is practically impossible to remove a metal stent, despite claims to the contrary for shape memory alloy stents that, in theory, can be narrowed in situ by application of heat or cold. Surgical revision of a stented vessel is also a practical impossibility, due to the difficulty of freeing the metal fiber impacted in the neointima. Coated metal stents have been introduced recently to provide controlled drug release, with very good short-term results (Table 1). Current practice is to use a bioresorbable phosphoryl choline polymer, or other polymer coating. However, the small reservoir capacity of the polymer film limits the amount of drug that can be loaded and eluted. Radiation-emitting stents have also been effective in reducing stent-induced restenosis; these, however, may induce radiation damage to the vessel wall. Finally, although not reported to date, there is the theoretical possibility of erosion of the arterial wall, due to compliance mismatch between the stent and arterial tissue. Given the difficulties with further development of metal stents, consideration of bioresorbable polymeric stents is attractive, as they may avoid the cited limitations of metal stents and offer other advantages as well.
392
R. C. Eberhart et al.
POLYMERIC STENTS
Several reports of resorbable and nonresorbable polymeric stents have recently appeared [12, 24– 29]. The rationale for the nondegradable stent is improved biocompatibility over the metal stent and convenient drug loading. Nonresorbable polymers being investigated for stent use include polyethylene terephthalate, polyurethane urea, and polydimethyl siloxane. The rationale for the bioresorbable stent is support of the arterial wall only during vessel healing, with gradual transfer of the mechanical load to the tissue as the stent mass and strength decrease over time, longer-term delivery of drug and/ or gene therapy to the vessel wall from an internal reservoir, and no need for a second surgery to remove the device. Bioresorbable polymers under investigation include aliphatic polyesters, polyorthoesters, and polyanhydrides. Recently, bioresorbable, linear, multiblock copolymers with shape memory capability have been introduced [30]. Controlled incremental heating of this thermoplastic material has been used to shrink sutures, making graded tissue approximations feasible in minimally invasive surgery applications. The same concept is also valuable for bioresorbable stent applications. Following balloon expansion, heat (approx. 5 ◦ C temperature change) applied to shape memory elements in the stent could reinforce designs that might not otherwise have sufficient recoil resistance. Poly-L-lactic acid (PLLA), poly-D,L-lactic acid (PDLA), poly ε-caprolactone (PCL) and polyglycolic acid (PGA), all aliphatic polyesters, are the most frequently used materials for bioresorbable stents [12, 24, 25]. PLLA and PDLA have a high tensile strength, permitting robust mechanical design, but requiring long degradation times. PGA and PCL have less strength, but faster degradation rates. Useful combinations of these materials (copolymers and blends) can be made to improve flexibility. These materials degrade principally by simple hydrolysis of the ester bond in the polymer backbone. Partial chain scission degrades the polymer to 10– 40 µm particles, capable of being phagocytosed and metabolized to carbon dioxide and water, which are of course fully resorbed. The degradation time is a function of the chemical structure of the polymer and its molecular weight. In typical formulations, PGA degrades over a time period of 6– 12 months, while PLLA degrades over months to years (Table 2). The long-term behavior of biodegradable polymers in blood vessels has not been well established. Van der Giessen et al. [31], testing strips of five different biodegradable polymers, PGA /PLA, PCL, polyhydroxybutyrate valerate, polyorthoester and polyethylene oxide/ polybutylene terephthalate, found extensive inflammatory responses within the coronary arterial wall. The observed tissue responses might be due the parent polymer compound, additives to the polymer, intermediate biodegradation products, the implant geometry, or combinations thereof. On the other hand, the authors noted that the implants were cleaned but not sterilized; therefore, bacterial or nonbacterial contamination might also have accounted for the inflammatory response. We have also observed a similar inflammatory response to sterilized PLLA stents implanted in the porcine femoral artery [32]. This
Bioresorbable polymeric stents
393
Table 2. Characteristics of typical bioresorbable polymersa Polymer
Melting point ( ◦ C)
Glass transition temperature ( ◦ C)
Modulus (Gpa)
Degradation time (months)
PGA PLLA PDLA PCL
225– 230 173– 178 Amorphous 58– 63
35– 40 60– 65 55– 60 −65 to −60
7.0 2.7 1.9 0.4
6– 12 >24 12– 16 >24
PGA = poly(glycolic acid); PLLA = poly(L -lactic acid); PDLA = poly(D, L -lactic acid); PCL = poly(ε-caprolactone). a Adapted from Ref. [102].
may have been due to the original polymer formulation, which was not intended for medical applications and contained an epoxide functionality of unknown quantity. More recent, purified formulations appear to induce a much less intense inflammatory response, as determined both by studies in the progress of PLLA fiber implantation into the rat aortic wall over 1– 4 weeks and by PLLA stent implantation in the pig femoral artery for 2 weeks. In the same vein, long-term study of polylactide copolymer, PLLA /PDLA (PLA96) stents in a rabbit abdominal aorta model found that the stents degraded with minimal tissue response within 24 months, with suitable encapsulation of polymer fragments in a thin neointima, leading the authors to suggest PLA96 as a promising stent core material [33]. Several early calls for expandable bioabsorbable stents have been published as alternatives for metallic stents [12, 24, 25]. These led to bioresorbable stents from Duke University [34], Tianjin/ Beijing University [35], Kyoto University [28], Igaki/ Tamai [36] and the University of Texas at Arlington/ Southwestern Medical School (UTA /SW) [32, 37]. The first biodegradable stent was developed and investigated at Duke University. This PLLA stent, based on a slotted polymer fiber design, was reported to withstand up to 1000 mmHg compression pressure; in vivo studies demonstrated minimal thrombosis and inflammatory responses, and moderate neointimal growth. The Tianjin/ Beijing stent, made of PDLA /PCL copolymer with an inner heparin layer, was deployed with a balloon catheter, employing heating and pressurization. This stent produced mild neointimal proliferation in swine carotid artery models at 2 months. The Kyoto University PGA coil stent exhibited thrombus deposition in canine implant studies, but no subacute closure. The Igaki/ Tamai stent, a bioresorbable PLLA zigzag coil thought to be derived from the Kyoto design, was studied in the first clinical report of a bioresorbable stent in the human coronary artery. This stent also required a combination of heating and pressurization for expansion. The preliminary (6-month) results suggest that this stent is safe and effective for human use. Long-term studies are anticipated.
394
R. C. Eberhart et al.
THE MULTIPLE LOBE STENT
This PLLA stent is designed using a linear, continuous coil array principle, by which multiple furled lobes (four in the present design) convert to a single large lobe upon balloon expansion (Fig. 1). Melt-extruded PLLA fibers (drawn 6 : 1) with a diameter of 0.14 mm and an ultimate stress of 350 ± 40 MPa are woven continuously around a four-mandrel array (one central, three peripheral) into a fourlobe configuration. Three longitudinal fibers are interwoven and glued to the coil for mechanical support. After fabrication, a conventional angioplasty balloon catheter is inserted in the central lobe and the stent can be deployed at the target site. The structure of the fully expanded stent is that of a helical coil with three longitudinal reinforcing fibers. The initial and final diameters of stents are adjustable by various combinations in sizes of central and peripheral rod mandrels. Stents with furled diameters ranging from 1.6 to 2.4 mm were fully expanded by 3 atm pressure, to 2.3– 4.7 mm: the corresponding expansion ratios ranged from 1.4 to 1.9. Collapse pressure under radial compression was adequately high, ranging from 0.4 to 2.4 atm, depending on the fiber ply and other design parameters. Preliminary results from various in vitro and in vivo studies of this expandable bioresorbable stent suggested that the design principles and fabrication technique were sufficiently robust and versatile, thus meriting further investigation [37, 38].
Figure 1. Schematic diagram of the helical coil polymeric stent design (external version). The fiber is wound continuously over four mandrels to obtain the multiple lobes. The side lobes are flattened passively, or by use of a sheath during delivery. These lobes can also be wound inside the primary coil to reduce the profile during delivery. Both designs open readily with balloon expansion.
Bioresorbable polymeric stents
395
In 1- and 2-week implant studies in the porcine common femoral artery, stents did not migrate; however, the vessel lumen was markedly reduced at 2 weeks, due to a strong inflammatory response. Stents, like other implants, elicit a range of host responses, which interfere with the patency of the device [31, 39]. Various approaches have been investigated to improve the biocompatibility of these stents, including surface plasma treatment and drug incorporation. Pulsed RF plasma treatment with di(ethylene glycol) vinyl ether significantly reduced platelet adhesion in a 1 h porcine arteriovenous shunt model to less than 10% of untreated control values [37]. Curcumin (diferoyl methane), a non-steroidal anti-inflammatory drug, was melt-extruded with PLLA to generate curcumin-loaded PLLA fiber (C-PLLA). The curcumin was uniformly distributed within the fibers and a stable curcumin release rate for 36 days was observed. In vitro studies of mouse peritoneal phagocytes indicated significant reductions in the adhesion of these cells to C-PLLA compared with PLLA controls. These results suggested that C-PLLA has antiinflammatory properties, which may benefit the implants. Other non-steroidal antiinflammatory agents with sufficiently high melting points can be introduced into the polymer bulk in the same way. We have also investigated the bulk loading of aqueous drugs that cannot tolerate melt extrusion, using a wet spinning technique that permits the incorporation of large amounts of drug (up to 20 wt%) in the PLLA fiber. In addition, hollow PLLA fiber spinning processes that allow loading drugs, genetic vectors, or radioisotopes into PLLA accessories have been examined.
STENTS AS RESERVOIRS FOR LOCAL DRUG AND GENE THERAPY
Efforts have been directed towards developing stents coated with a biodegradable drug-impregnated polymer, capable of gradually releasing therapeutic agents into the vessel wall to reduce thrombosis and restenosis [2, 3, 39, 40]. The use of antithrombotic drugs such as heparin and hirudin is one strategy [41]. Other agents include prostacyclin analogy Iloprost [42], glycoprotein IIb/ IIIa receptor antibodies or inhibitors [43], and antiproliferative agents such as nitric oxide donors, corticosteroids and taxanes that inhibit neointima and local tumor proliferation [18, 39, 40]. Drugs or peptides contained within polymers can be in a nonchemically bonded configuration (physical entrapment) or chemically bonded to the polymer side-chains [26]. Stents coated with drug-eluting polymers such as hirudin, prostacyclin, and nitrosylated albumin were shown to reduce neointima formation [39, 42]. Decreased early thrombosis and neointima formations were also observed in stents loaded with glycoprotein IIb– IIIa inhibitors [44, 45] and with nitric oxide donors [46, 47]. Furthermore, intramural delivery of an antiproliferative agent, a specific tyrosine kinase inhibitor, using biodegradable stents has suppressed the restenotic changes of coronary arteries of treated pigs [28]. In addition to local drug delivery, stents can also serve as carriers for gene therapy delivery. Stents seeded with cells transfected with the desired gene, stents loaded
396
R. C. Eberhart et al.
with recombinant adenovirus gene transfer vectors, and stents loaded with naked DNA impregnated in various matrices have been proposed [2, 29, 48– 50]. The introduction of an interested gene into the arterial wall can be achieved either by in vitro genetic manipulation of cells before their seeding onto stents or by direct in vivo gene transfer. Cell-based gene transfer using stents as platforms has been shown a major advantage in terms of site-specific gene expression. However, cell-based gene delivery has several limitations, including removal or injury of cells from the stent after balloon expansion and a significant time delay required for cell harvest, expansion, gene transfer, and subsequent selection prior to stent seeding. Yet seeding the stents with genetically engineered endothelial cells (ECs) to produce agents such as tPA has been shown to inhibit smooth muscle cell (SMC) proliferation [51]. A recent study has shown that a mesh stent coated with fibronectin is an excellent platform for adherent gene transduced SMCs [52]. Similar to the advantage of cell-based gene transfer, site-specific gene delivery, gene-stent therapy has been applied to reduce thrombosis and in-stent restenosis. Genes that encode enzymes of the prostacyclin synthetic pathway, nitric oxide synthase, the thrombin inhibitor, and thrombomodulin have been studied and demonstrated a significant reduction in thrombosis and restenosis in animal models [29, 50, 53– 55]. We successfully demonstrated local gene transfer and expression from PLLA stents impregnated with a recombinant adenovirus carrying a nuclear localizing βGal reporter gene into the carotid and renal arteries in the rabbit. Liver transfection was negligible in both cases, suggesting that gene delivery was local, not convected to remote sites to a significant degree [48]. In spite of promising results in animal models, to date no effective human gene therapy has been found to prevent restenosis [29, 50, 56]. In addition, potential side effects of the gene therapy approach such as subsequent malignant transformation due to oncogene activation with utilization of retroviral gene vectors and subsequent expression in other organs need to be further evaluated. In order to prevent potentially dangerous distal spread of viral vectors, a recent study has developed stent-based anti-viral antibody tethering of vectors onto the collagen coating surface of stents as a suitable platform for local gene delivery [57]. Another promising strategy for gene therapy delivery involves the introduction of antisense oligonucleotides into cells in order to inactivate the mRNA encoding proteins important in the restenotic process [58]. Uses of synthetic oligonucleotides to suppress proto-oncogenes including c-myb and c-mbc, proliferating cell nuclear antigen, and cell cycle-specific proteins cdc2 and cdk2 kinases were reduced protein expression and cell proliferation [12, 58].
NON-CORONARY USES OF STENTS
The range of stent applications has expanded with increases in experience and encouraging results in the treatment of vascular diseases. Stents have been used for the treatment of urethral obstruction from benign prostatic hyperplasia; for the treatment of tracheobronchial obstruction of benign or malignant origin; for the
Bioresorbable polymeric stents
397
treatment of benign and malignant strictures of the esophagus, the gastrointestinal (GI) tract, and the bile duct; and for the treatment (stents and stent-grafts) of arterial dissections, aneurysms and various neurovascular diseases.
STENTS IN UROLOGY
Stents have been used to prevent postoperative urine retention following thermal treatment of benign prostatic hyperplasia (BPH) by various means, including direct vision laser ablation of the prostate and transurethral microwave therapy. Several stent designs, including the Nissenkorn, Barnes, Finnish biodegradable selfreinforced polyglycolic acid (SR-PGA) spiral and Trestle, were shown to prevent obstruction of the prostatic urethra and restricture of the anterior urethra [27, 59, 60]. Biodegradable stents have been studied clinically in the treatment of BPH and are claimed to provide superior results to suprapubic catheters [61– 65]. Self-reinforced PLLA bioresorbable spiral stents are also undergoing evaluation for use in the anterior urethra, posterior urethra and upper urinary tract, to prevent urinary retention and repair of local ureteral trauma or defects [66, 67]. Surface modification of these biodegradable stents, by grafting with hydroxyethylmethacrylate or by incorporation of biologically active compounds, is claimed to be an efficient approach to improve biocompatibility and cell adhesion properties [68, 69].
STENTS FOR THE MANAGEMENT OF TRACHEOBRONCHIAL OBSTRUCTION
Tracheobronchial obstruction from either benign or malignant disease causes significant morbidity and mortality. Metal stents, developed originally for the vascular system, have been adapted for lesions involving the tracheobronchial tree. These include the Palmaz (Johnson & Johnson), Strecker (Boston Scientific), Gianturco-Z (William Cook Europe), Wallstent (Boston Scientific) and Ultraflex (Boston Scientific) stents [70]. These stents were successfully used to treat patients with inoperable bronchogenic cancer, esophageal tumors, primary tracheal tumors, and metastatic malignancy. Bioresorbable tracheal stents have been investigated in the setting of pediatric tracheal malacia, to solve the problem of limited tracheal growth in children with rigid external fixation and to avoid the necessity of a second procedure to remove the synthetic material [70– 72]. The general results from these studies suggest that stenting is a promising method to treat tracheal obstruction.
STENTS IN THE ESOPHAGUS AND GASTROINTESTINAL (GI) TRACT
Many malignant and benign strictures in the esophagus and GI tract can be treated by minimally invasive alternatives to surgery, including the use of stents. Most
398
R. C. Eberhart et al.
commonly used in the esophagus and GI tract are the Wallstent (Boston Scientific), Ultraflex (Boston Scientific), Gianturco-Z (William Cook Europe), Esophacoil (Instent) and Flamingo stents (Boston Scientific). In general, these stents have been shown to be effective in relieving esophageal dysphagia [22, 73, 74]. This success has led to the employment of stents to manage lesions of the GI tract, including the stomach, pylorus, upper small intestine, duodenum, and colon [22, 73]. The use of bioresorbable material is currently being explored for the esophageal stent. First results in the placement of a PLLA stent (Instent) for the management of benign esophageal stricture suggest that a bioresorbable stent offers a new treatment modality [75]. In addition, bioresorbable stents were recently used in pancreaticojejunal anastomoses [76].
STENTS IN THE MANAGEMENT OF NEUROVASCULAR DISEASE
Stents and stent-grafts have been used for the management of arterial and venous sinus stenosis, arterial dissection, arterial aneurysm and arteriovenous fistulae [77, 78]. A number of case reports have been published describing the significant reduction in carotid stenosis with the use of stents in the treatment of carotid stenosis, recurrent carotid stenosis, vertebral artery stenosis, and venous sinus stenosis [78]. As cited in this review, Shawl et al. [79] reported a series of 124 stented vessels in which carotid stenosis was reduced from 86 ± 7% to 2 ± 2%; the major postprocedural stroke rate was 0.8% and the minor stroke rate was 2%. Three cases of basilar artery stenosis have been successfully treated with coronary stents at our institution [80]. Other unpublished reports from our institution have demonstrated the effectiveness of stents in bridging side-wall aneurysmal ostia, suggesting that stents are a promising means for the management of arterial dissection and pseudoaneurysm. Unfortunately, no large studies have yet been published, so the effectiveness of stents in this application remains to be determined.
CONCLUSION
Stents play an increasingly important role in percutaneous coronary interventions. Various metal stents have been shown to reduce the restenosis rate compared with angioplasty alone. This success has prompted the expansion of stent usage to peripheral arteries, the urethra, trachea, esophagus and GI tract. Stents do not eliminate the problem of coronary arterial restenosis and may contribute to it by inducing neointimal hyperplasia. Thus, isotope-loaded metal stents and polymer- and ceramic-coated metal stents, using the coating as a vehicle for local anti-inflammatory drug delivery, have been introduced, with promising results. Several bioresorbable stent designs are in development for temporary mechanical support combined, in some cases, with drug and/ or gene therapy delivery. Such temporary, bioresorbable stents that match the expandability and recoil resistance
Bioresorbable polymeric stents
399
of metal stents in the coronary arterial setting have been reported. These stents are theoretically superior for arterial wall healing, but face challenges in their application to smaller, more tortuous channels. Radioisotope-loaded metal or polymeric stents are also appealing for the local treatment of tumors and for the prevention of excessive granulation tissue formation in non-coronary settings. Stent design and development is currently a very active area of bioengineering practice. The expanding range of applications and new designs, materials, and surface treatments suggest that more effective, less invasive therapies may be anticipated in the near future. Acknowledgement This work was supported in part by USPHS grants RO1 HL /DE 53225 and F32HL010380.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.
G. Belli, S. G. Ellis and E. J. Topol, Prog. Cardiovasc. Dis. 40, 159 (1997). H. Bult, Trends Pharmacol. Sci. 21, 274 (2000). R. Hoffmann and G. S. Mintz, Eur. Heart J. 21, 1739 (2000). A. J. Carter, D. Scott, D. Rahdert, L. Bailey, J. De Vries, K. Ayerdi, T. Turnlund, R. Jones, R. Virmani and T. A. Fischell, J Invasive Cardiol. 11, 127 (1999). J. M. Dumonceau and J. Deviere, Baillieres Best Pract. Res. Clin. 13, 109 (1999). Y. Ozaki, A. G. Violaris and P. W. Serruys, Prog. Cardiovasc. Dis. 39, 129(1996). A. G. Violaris, Y. Ozaki and P. W. Serruys, Int. J. Cardiac Imaging 13, 3 (1997). T. Nicholson, Hosp. Med. 60, 571 (1999). T. J. Cleveland and P. Gaines, Hosp. Med. 60, 630 (1999). C. D. J. Barras and K. A. Myers, Eur. J. Vasc. Endovasc. Surg. 19, 564 (2000). D. R. Ramsdale, Hosp. Med. 60, 624 (1999). J. F. Tanguay, J. P. Zidar, H. R. Phillips and R. S. Stack, Cardiol. Clin. 12, 699 (1994). A. H. Gershlick, Heart 86, 104 (2001). C. von Birgelen, M. Haude, J. Herrmann, C. Altmann, W. Klinkhart, D. Welge, H. Wieneke, D. Baumgart, S. Sack and R. Erbel, Catheter Cardiovasc. Interv. 47, 496 (1999). C. Stefanandis, E. Tsiamis, C. Vlachopoulos, K. Toutouzas, C. Stratos, I. Kallikazaros, M. Vavuranakis and P. Toutouzas, Cathet. Cardiovasc. Diagn. 40, 217 (1997). C. Herdeg, M. Oberhoff and K. R. Karsch, Semin. Interv. Cardiol. 3, 197 (1998). W. J. van der Giessen, H. M. van Beusekom, M. H. Eijgelshoven, M. A. Morel and P. W. Serruys, Semin. Interv. Cardiol. 3, 173 (1998). J. Gunn and D. Cumberland, Eur. Heart J. 20, 1693 (1999). A. W. Chan and D. J. Moliterno, Cleve. Clin. J. Med. 68, 796 (2001). J. Kotzerke, H. Hanke and M. Hoher, Eur. J. Nucl. Med. 27, 223 (2000). C. Hehrlein and W. Kubler, Semin. Interv. Cardiol. 2, 109 (1997). R. Lamber, Endoscopy 32, 322 (2000). S. Balter, Cathet. Cardiovasc. Diagn. 45, 292 (1998). M. Labinaz, J. P. Zidar, R. S. Stack and H. R. Phillips, J. Int. Cardiol. 8, 395 (1995). A. Colombo and E. Karvouni, Circulation 102, 371 (2000). T. Peng, P. Gibula, K. Yao and M. F. A. Goosen, Biomaterials 17, 685 (1996). M. Talja, T. Valimaa, T. Tammela, A. Petas and P. Tormala, J. Urol. 11, 391 (1997).
400
R. C. Eberhart et al.
28. T. Yamawaki, H. Shimokawa, T. Kozai, K. Miyata, T. Higo, E. Tanaka, K. Egashira, T. Shiraishi, H. Tamai, K. Igaki and A. Takeshita J. Am. Coll. Cardiol. 32, 780 (1998). 29. C. Indolfi, C. Coppola, D. Torella, O. Arcucci and M. Chiariello, Cardiol. Rev. 7, 324 (1999). 30. A. Lendlein and R. Langer, Science 296, 1673 (2002). 31. W. J. van der Giessen, M. Lincoff, R. S. Schwartz, H. M. M. van Beusekom, P. W. Serruys, D. R. Holmes, S. G. Ellis and E. J. Topol, Circulation 94, 1690 (1996). 32. S. Su, Biomedical engineering, Doctoral dissertation, University of Texas at Arlington, Arlington (2000). 33. E. Hietala, U. Salminen, A. Stähls, T. Välimaa, P. Maasilta, P. Törmälä, M. S. Nieminen and A. L. J. Harjula, J. Vasc. Res. 38, 361 (2001). 34. C. M. Agrawal and H. G. Clark, Invest. Radiol. 27, 1020 (1992). 35. R. Gao, R. Shi, S. Qiao, L. Song and Y. Li, J. Am. Coll. Cardiol. 27, 85A (1996). 36. H. Tamai, K. Igaki, E. Kyo, K. Kosuga, A. Kawashima, S. Matsui, H. Komori, T. Tsuji, S. Motohara and H. Uehata, Circulation 102, 399 (2000). 37. S. Su, R. Y. Chao, C. L. Landau, K. D. Nelson, R. B. Timmons, R. S. Meidel and R. C. Eberhart, Ann. Biomed. Eng. (2003) (in press). 38. S. Su, C. L. Landau, R. Y. N. Chao, R. B. Timmons, R. S. Meidel, L. Tang and R. C. Eberhart, Circulation 104, II-507 (2001). 39. O. F. Bertrand, R. Sipehia, R. Mongrain, J. Rodés, J. Tardif, L. Bilodeau, G. Côté and M. G. Bourassa, J. Am. Coll. Cardiol. 32, 562 (1998). 40. D. Brieger and E. Topol, Cardiovasc. Res. 35, 405 (1997). 41. R. Herrmann, G. Schmidmaler, B. Märkl, A. Resch, I. Hähnel, A. Stemberger and E. Alt, Thromb. Haemost. 82, 51 (1999). 42. E. Alt, I. Haehnel, C. Beiharz, K. Prietzel, D. Preter, A. Stemberger, T. Fliedner, W. Erhardt and A. Schömig, Circulation 101, 1453 (2000). 43. A. Colombo and C. Briguori, Am. Heart J. 138, S153 (1999). 44. R. Aggarwal, D. Ireland, M. Azrin, D. De Bono and A. Gershlick, J. Am. Coll. Cardiol. 29, 353A (1997). 45. R. Aggarwal, D. Ireland, M. Azrin, M. Ezekowitz, D. De Bono and A. Gershlick, Circulation 94, 3311 (1996). 46. J. Folts, N. Maalej, J. Keaney and J. Loscalzo, Circulation 92, I-670 (1995). 47. J. Folts, N. Maalej, J. Keaney and J. Loscalzo, J. Am. Coll. Cardiol. 27, 86A (1996). 48. Y. Ye, C. Landau, J. E. Willard, G. Rajasubramanian, A. Moskowitz, S. Aziz, R. S. Meidell and R. C. Eberhart, Ann. Biomed. Eng. 26, 398 (1998). 49. B. D. Klugherz, P. L. Jones, X. Cui, W. Chen, N. F. Meneveau, S. DeFelice, J. Connolly, R. L. Wilensky and R. J. Levy, Nature Biotechnol. 18, 1181 (2000). 50. M. D. Feldman, B. Sun, B. J. Koci, C. C. Wu, J. R. Kneller, H. S. Borovetz, S. Watkins, A. Nadeem, L. E. Weiss, M. L. Reed, A. J. Smith and W. D. Rosenblum, J. Long Term Eff. Med. Implants 10, 47 (2000). 51. D. A. Dichek, R. F. Neville, J. A. Zwiebel, S. M. Freeman, M. B. Leon and W. F. Anderson, Circulation 80, 1347 (1989). 52. C. J. Panetta, K. Miyauchi, D. Berry, R. D. Simari, D. R. Holmes, R. S. Schwartz and N. M. Caplice, Hum. Gene Ther. 13, 433 (2002). 53. P. Vermeersch, Z. Nong, E. Stabile, O. Varenne, H. Gillijns, M. Pellens, N. Van Pelt, M. Hoylaerts, I. De Scheerder, D. Collen and S. Janssens, Arterioscler. Thromb. Vasc. Biol. 21, 1604 (2001). 54. L. M. Akyurek, S. Nallamshetty, K. Aoki, H. San, Z. Y. Yang, G. J. Nabel and E. G. Nabel, Mol. Ther. 3, 779 (2001). 55. O. Varenne and P. Sinnaeve, Curr. Interv. Cardiol. Rep. 2, 309 (2000). 56. L. Gruberg, R. Waksman, L. F. Satler, A. D. Pichard and K. M. Kent, Exp. Opin. Invest. Drugs 9, 2555 (2000).
Bioresorbable polymeric stents
401
57. B. D. Klugherz, C. Song, S. DeFelice, X. Cui, Z. Lu, J. Connolly, J. T. Hinson, R. L. Wilensky and R. J. Levy, Hum. Gene Ther. 13, 443 (2002). 58. N. Kipshidze, J. Moses, L. R. Shankar and M. Leon, Curr. Opin. Mol. Ther. 3, 265 (2001). 59. R. Kapoor, E. N. Liatsikos and G. Badlani, Curr. Opin. Urol. 10, 19 (2000). 60. B. A. Kletscher and J. E. Oesterling, Urol. Clin. 22, 423 (1995). 61. J. J. M. C. H. de la Rosette, H. P. Beerlage and F. M. J. Debruyne, J. Endourol. 11, 467 (1997). 62. E. Kemppainen, M. Talija, M. Riihelä, T. Pohjonen, P. Törmälä and O. Alfthan, Urol. Res. 21, 235 (1993). 63. A. Pétas, M. Talja, T. Tammela, K. Taari, K. Lehtoranta, T. Välimaa and P. Törmälä, J. Urol. 157, 173 (1997). 64. A. Pétas, T. Isotalo, M. Talja, T. L. J. Tammela, T. Välimaa and P. Törmälä, Scand. J. Urol. Nephrol. 34, 262 (2000). 65. A. Pétas, P. Kärkkäinen, M. Talja, K. Taari, M. Laato, T. Välimaa and P. Törmälä, Br. J. Urol. 80, 903 (1997). 66. J. Lumiaho, A. Heino, T. Pietiläinen, M. Ala-Opas, M. Talja, T. Välimaa and P. Törmälä, J. Urol. 164, 1360 (2000). 67. J. Lumiaho, A. Heino, V. Tunninen, M. Ala-Opas, M. Talja, T. Välimaa and P. Törmälä, J. Endourol. 13, 107 (1999). 68. A. Brauers, P. K. Jung, H. Thissen, O. Pfannschmidt, W. Michaeli, H. Hoecker and G. Jakse, Tech. Urol. 4, 214 (1998). 69. A. Brauers, H. Thissen, O. Pfannschmidt, H. Bienert, A. Foerster, D. Klee, W. Michaeli, H. Höcker and G. Jakse, J. Endourol. 11, 399 (1997). 70. A. L. Rafanan and A. C. Mehta, Int. Chest Radiol. 38, 395 (2000). 71. T. C. Robey, P. M. Eiselt, H. S. Murphy, D. J. Mooney and R. A. Weatherly, Laryngoscope 110, 1936 (2000). 72. H. Lochbihler, J. Hoelzi and H. Dietz, J. Pediatr. Surg. 32, 717 (1997). 73. R. Morgan and A. Adam, J. Vasc. Interv. Radiol. 12, 283 (2001). 74. T. Nicholson, Hosp. Med. 61, 97 (2000). 75. S. W. Fry and D. E. Fleischer, Gastrointest. Endosc. 45, 179 (1997). 76. M. Parviainen, J. Sand, A. Harmoinen, H. Kainulainen, T. Välimaa, P. Törmälä and I. Nordback, Pancreas 21, 14 (2000). 77. H. A. Gray and A. W. Crane, Am. J. Roentgenol. 175, 289 (2000). 78. M. B. Horowitz and P. D. Purdy, Neurosurgery 46, 1335 (2000). 79. F. Shawl, W. Kadro, M. J. Domanski, F. L. Lapetina, A. A. Iqbal, K. G. Dougherty, D. D. Weisher, J. F. Marquez and S. T. Shahab, J. Am. Coll. Cardiol. 35, 1721 (2000). 80. M. B. Horowitz, G. L. Pride, D. F. Graybeal and P. D. Purdy, Neurosurgery 45, 925 (1999). 81. E. Grube, S. M. Silber and K. E. Hauptmann, Circulation 104, II-463 (2001). 82. S. Park, W. H. Shim, D. S. Ho, A. E. Raizner, S. W. Park, J. J. Kim, M. K. Hong, C. W. Lee, S. Y. Cho, Y. S. Jang, D. H. Choi, C. P. Lau, R. Lam and Y. Wang, Circulation 104, II-464 (2001). 83. J. W. Moses, M. B. Leon, J. J. Popma and R. E. Kuntz, Circulation 104, II-464 (2001). 84. J. E. Sousa, M. Morice, P. W. Serruys, J. Fajadet, M. Perin, E. B. Hayashi, A. Colombo, G. Schuler, P. Barragan and C. Bode, Circulation 104, II-463 (2001). 85. K. A. Robinson, N. A. Chronos, J. Royal, L. Suh, G. D. Cipolla, R. Virmani and R. S. Stack, Circulation 104, II-506 (2001). 86. F. Gobeil, M. Laflamme, M. Bouchard, Campbell E., M. Wood, P. St-Jacques, J. Vincent and G. Leclerc, Circulation 104, II-388 (2001). 87. H. M. Van Beusekom, M. J. Post, B. J. De Smet, M. G. Put, M. S. Harteveld and W. J. Van der Giessen, Circulation 104, II-506 (2001). 88. K. J. Salu, H. Yanming, J. M. Bosmans, L. Xiaoshun, I. K. Scheerder and C. J. Vrints, Circulation 104, II-506 (2001).
402
R. C. Eberhart et al.
89. H. C. Lowe, R. Fahmy, C. N. Chesterman and L. M. Khachigian, Circulation 104, II-265 (2001). 90. H. C. Lowe, R. G. Fahmy, M. M. Kavurma, A. Baker, C. N. Chesterman and L. M. Khachigian, Circ. Res. 89, 670 (2001). 91. H. Kim, H. A. Yazdi, R. Seaborn, P. Iversen, G. S. Roubin, S. S. Lyer, M. B. Leon, J. W. Moses and N. Kipshidze, Circulation 104, II-545 (2001). 92. T. A. Swierkosz, S. Kapoor, D. C. Tardiff, J. W. Hirshfeld, B. D. Klugherz, D. M. Kolansky, K. Magness, N. Valettas, R. L. Wilensky and H. C. Herrmann, Circulation 104, II-385 (2001). 93. G. Montalescot, P. Barragan, O. Wittemberg, P. Pinton, S. Elhadad, X. Deboisgelin, A. Akesbi, B. Jouve, M. Amor and F. Funck, Circulation 104, II-386 (2001). 94. G. Amoroso, A. J. van Boven, D. J. van Veldhuisen, R. A. Tio, C. P. Balje-Volkers, A. S. Petronio and W. van Oeveren, J. Cardiovasc. Pharmacol. 38, 633 (2001). 95. D. L. Bhatt and A. M. Lincoff, Circulation 104, II-384 (2001). 96. A. G. Rebeiz, K. S. Pieper, J. C. O’Shea, I. C. Gilchrist, B. Chandler, J. Slater, J. B. Muhlestein, T. J. Loenz, D. Joseph, M. M. Kitt and J. E. Tcheng, Circulation 104, II-386 (2001). 97. Y. Huang, L. Wang, I. Vermeire, E. Verbeken, E. Schacht and I. D. Scheerder, Circulation 104, II-665 (2001). 98. T. R. Porter, P. Iversen, F. Xie, D. Kricsfeld, S. J. Radio, J. Lof and S. Flynn, Circulation 104, II-589 (2001). 99. R. Rohde, B. Heublein, S. Ohse, E. Evagorou, K. Pethig and A. Haverich, Circulation 104, II-666 (2001). 100. G. B. Danzi, C. Capuano, M. Sesana, A. D. Blasi and D. Antoniucci, Circulation 104, II-625 (2001). 101. S. Silber, N. Reifart, M. Morice, J. V. Dahl, E. Benit, K. Hauptmann, J. E. Sousa, J. G. Webb, U. Kaul and C. Chan, Circulation 104, II-623 (2001). 102. J. C. Middleton and A. J. Tipton, Med. Plast. Biomater 30 (March) (1998).
Tissue-culture surfaces with mixtures of aminated and fluorinated functional groups. Part 1. Synthesis and characterization JAMES R. BAIN 1,2,∗ and ALLAN S. HOFFMAN 3 1 Sarah
W. Stedman Center for Nutritional Studies, Duke University Medical Center, Durham, NC 27710, USA 2 Department of Pharmacology and Cancer Biology, Duke University Medical Center, Mail Stop DUMC 3813, LSRC building, room C348, Durham, NC 27710, USA 3 Department of Bioengineering, Box 352255, University of Washington, Seattle, WA 98195, USA Received 10 January 2002; accepted 4 December 2002 Abstract—Surface chemistry of culture dishes can have profound effects on the phenotype of cultured cells. In the present study, chemisorption from aqueous, binary mixtures of organosilanes onto borosilicate glass created surfaces bearing diamine groups (N2), trifluoropropyl groups (F3) and mixtures of the two. Composition of N2-F3 surfaces was controlled by the ratio of monomers in the silanization bath, as confirmed by electron spectroscopy for chemical analysis and by conjugation of surface amines with fluorescein-5-isothiocyanate. Atomic-force microscopy revealed that silanized surfaces are patchy, though their root-mean-square roughnesses do not differ significantly from that of smooth glass (0.3 nm). Surfaces richest in diamine residues were the most hydrophilic, with advancing water-contact angles 90◦ . The accompanying paper (the next article in this issue) describes the effects of these surface chemistries on the phenotype of transgenic insulinoma cells in vitro. We conclude that chemisorption from the N2-F3 system provides a simple, one-pot method for tailoring the chemistry of glass culture surfaces. Key words: Silanes; surface modification; tissue culture; electron spectroscopy for chemical analysis; atomic-force microscopy; wettability.
INTRODUCTION
Chemistry of culture surfaces can affect the behavior and phenotype of cultured mammalian cells [1– 4]. In the present study, we sought to create novel culture substrates on borosilicate glass by chemisorption of organosilanes. Resulting ∗ To
whom correspondence should be addressed. Tel.: (1-919) 613-8652; Fax: (1-919) 668-6044; e-mail:
[email protected]
404
J. R. Bain and A. S. Hoffman
surfaces bear diamine groups, trifluoropropyl groups and mixtures of the two. We chose silanized borosilicate glass as a model system because we were able to finetune the surface chemistry by competitive chemisorption from binary, aqueous solutions of silane monomers. Silanization of glass surfaces results from the hydrolysis and subsequent condensation of chloro- or alkoxysilanes with surface silanols (SiOH, Fig. 1). Lateral condensation polymerization of silanes on the surface creates thin deposits of polysiloxanes [5, 6]. Numerous investigators have employed silanes and the related silazanes to create chemically defined surfaces for the study of cell-material interactions [4, 7– 34]. Swalec [20] created cell-culture surfaces of mixed character by competitive, vapor-phase deposition of methyl- and chloropropyl silanes. We hoped that competitive chemisorption from aqueous solutions of mixed silanes would offer a practical, controllable and inexpensive alternative to vapor-phase deposition. Competitive chemisorption of organosilanes from water and other solvents has previously been used to create surfaces of mixed character [5, 6, 35– 37], but to our knowledge such surfaces have not yet been evaluated in cell culture. Portions of this report are reproduced from a preliminary presentation [38].
MATERIALS AND METHODS
Selection of silane monomers for competitive chemisorption A priori, we expected that amine-rich surfaces would favor cell growth and that fluorinated surfaces would be less suitable for anchorage-dependent mammalian cells, so we selected silane monomers bearing these functional groups. We sought a competitive pair of monomers that are stable for a reasonable time in a common
Figure 1. Silanization of etched borosilicate glass with monomers N2 and F3. A monolayer is shown for simplicity. These and most other silanized surfaces are in fact patchy (see text).
Aminofluorosiloxane culture surfaces
405
aqueous solvent and in which the pendant chains of both moieties are grossly similar in length. Diamine monomer N2: 3-(2-aminoethylamino) propyltrimethoxysilane Many anchorage-dependent vertebrate cells spread and grow well on surfaces rich in amines, amides and other nitrogen-containing groups, including surfaces rendered nitrogenous by organosilanization [8– 10, 13, 16, 17, 21– 25, 28, 32– 34, 39, 40]. A principle focus of the present investigation was to create novel surfaces for culture of insulinoma cells [1]. Limited data indicate that some insulin-secreting endocrine cell lines will grow and function on amide- and amine-rich surfaces [41– 44]. We chose silane N2, with its primary and secondary amine groups (Fig. 1), because it has been used in previous studies to create surfaces acceptable to a wide variety of mammalian cells, including such fastidious cells as neurons and endothelial cells [10– 14, 18, 19, 26, 27, 29– 31]. For the promotion of cell attachment, spreading and growth, N2’s diamine structure might be preferable to monoamines and other common aminosilane structures [29]. The residue remaining after aminosilanization with N2 is shown in Fig. 1. Several hypotheses have been put forth to account for the ‘cell-friendly’ nature of N2 residues (i.e. 3-(2-aminoethylamino) propylpendant groups) and other solid-phase amines, including direct association of cellsurface proteoglycans with synthetic cations by electrostatic or hydrogen-bonding interactions [12, 19, 21, 24, 26, 29, 31, 39, 45– 48] and the preferential adsorption of soluble cell-attachment proteins, such as fibronectin and vitronectin, to nitrogenrich surfaces [13, 30, 31, 39, 40, 49]. Trifluorinated monomer F3: (3,3,3-trifluoropropyl) trimethoxysilane We chose F3 because we expected fluorinated surfaces to be relatively poor growth substrata for our insulinoma cells [1]. In general, fluorinated materials have large advancing contact angles with water (θA 85◦ ), low critical surface tensions (Zisman’s γc 25 dyne/ cm) and do not support vigorous cell growth [50– 55]. Poor cell growth seen on many fluorinated materials might be due to low binding strengths or unfavorable binding conformations adopted by physisorbed celladhesion proteins [56]. We were unable to find evidence that surfaces modified with silane F3 have previously been evaluated in cell culture. The C3 residue resulting from its use is shown in Fig. 1. Silanized surfaces bearing longer, partially fluorinated, aliphatic side chains have been shown to inhibit growth of mammalian cells. Side-chain lengths in past studies have included C8 [19], C10 [34] and C18 [23]. Silanes with such long fluoroalkyl side chains were not considered for use in the present study, because they are insoluble in the aqueous solutions we wished to use for competitive deposition with the diamine co-monomer, N2. In selecting the hydrophobic member of the monomer pair, we chose the fluoroalkyl silane F3 over nonfluorinated alkylsilanes because it bears an element (F) not present in N2. Extreme electronegativity of fluorine makes it especially easy to detect at low degrees
406
J. R. Bain and A. S. Hoffman
of incorporation; F3’s fluorines provided a strongly reporting atomic marker for surface characterization by electron spectroscopy for chemical analysis, infra [57]. Glass selection, cleaning and etching Conventional borosilicate cover glasses for microscopy were purchased from Carolina Biological Supply (Burlington, NC, USA). According to the manufacturer (Glaswarenfabrik Karl Hecht, Sondheim / Rhön, Germany), the bulk formulation, in mass percentages, is 65% SiO2 , 12% (Na2 O + K2 O), 7.5% B2 O3 , 6.5% ZnO, 5.5% TiO2 , 3% Al2 O3 and 0.5% Sb2 O3 . This formulation is expressed in terms of atomic percentages in Table 1. Squares (12 × 12 mm) provided a constant circumference for Wilhelmy-plate, contact-angle goniometry (infra). Circles (15 mm diameter) fit snugly in 24-well plates for cell culture. Before silanization, glass samples were cleaned and etched while mounted in polytetrafluoroethylene racks, as previously described [15]. Silanes, silanization and annealing (3,3,3-Trifluoropropyl) trimethoxysilane (F3) and 3-(2-aminoethylamino) propyltrimethoxysilane (N2) were used as received from PCR (Gainesville, FL, USA). Organic solvents were from J.T. Baker (Phillipsburg, NJ, USA) and were ‘analyzedreagent’ grade or better. In all cases, silanization mixtures had a total silane concen-
Acetone-rinsed glass
Detergent-washed glass
Alkaline-etched glass
Silanized with 0/3 F3 (100% N2)
Silanized with 3/3 F3 (100% F3)
O Si Na B K Al Zn Ti Sb S N F C
Glassmaker’s nominal formulation*
Table 1. ESCA characterization of borosilicate glass surfaces after cleaning, etching and two different organosilanizations
61.79 22.60 4.04** 4.50 2.66** 1.23 1.67 1.44 0.07 0.00 0.00 0.00 0.00
53.2 22.8 4.5 3.0 2.4 1.5 1.5 0.6 0.0 0.0 0.0 0.0 10.5
49.4 21.8 3.0 3.4 1.3 2.0 0.6 0.6 0.0 0.8 1.4 0.0 15.8
57.0 24.4 2.6 2.5 1.5 1.6 1.0 0.6 0.0 0.0 0.0 0.0 8.9
48.8 24.3 2.3 2.8 0.6 2.1 0.6 0.4 0.0 0.0 2.5 0.0 15.6
49.4 23.9 1.9 1.9 1.0 1.9 0.6 0.6 0.0 0.0 0.4 8.1 10.4
Values are given as atomic percentages. ∗ Theoretical values, ∗∗ approximate values.
Aminofluorosiloxane culture surfaces
407
tration of 0.1 M in 50 : 50 ddH2 O /isopropanol, apparent pH 4.75. Extensive experimentation showed that departures from this solvent condition led to instability of mixtures of N2 and F3, with rapid polymerization or precipitation of the monomers, particularly those richest in the F3 monomer. Solutions were used immediately after adjustment of pH. Silanization proceeded 2 h at approx. 20◦ C, with stirring, followed by extensive rinsing in 50 : 50 ddH2 O /isopropanol, apparent pH 4.75, followed by a final rinse in 70% (v/ v) ethanol in ddH2 O to disinfect samples in preparation for cell culture. Annealing by heat is necessary to give stable deposits of siloxanes on glass surfaces, but can destroy amines if carried to excess [6, 58, 59]. To assess the effects of annealing regimens on amine reactivity, we conjugated fluorescein-5isothiocyanate (FITC, Aldrich, Milwaukee, WI, USA) to surfaces, using 1 mg/ ml FITC in N,N-dimethylformamide (DMF), with 0.712 µl/ ml triethylamine as a base catalyst (Fig. 2). Thiourea conjugation proceeded for 2 h at approx. 20◦ C, unstirred. After a 30-min rinse in DMF, samples were rinsed thrice in ddH2 O and dried in vacuo at approx. 20◦ C, about 0.5 mmHg. Bound FITC residues were then stripped in 0.1 M NaOH for 2 h, unstirred, at approx. 20◦ C. Fluorescence of the acid-neutralized stripping solution was determined in a Perkin-Elmer LS5B luminescence spectrometer, using a xenon source (excitation 495 nm, emission 519 nm; Perkin-Elmer, Norwalk, CT, USA).
Figure 2. Reactivity of aminosiloxane (N2) residues with fluorescein-5-isothiocyanate.
408
J. R. Bain and A. S. Hoffman
Electron spectroscopy for chemical analysis (ESCA) Analyses were performed with an X-Probe ESCA instrument (Surface Science Instruments, Mountain View, CA, USA). This instrument permits analysis of the outermost approx. 50 Å of a sample in an elliptical area whose short axis is aobut 600 µm (D. Leach-Scampavia, University Washington, Seattle, WA, USA, personal communication). An aluminum Kα1,2 monochromatized X-ray source was used to stimulate photoemission. Energy of the emitted electrons was measured in a hemispherical energy analyzer at pass energies ranging from 25 to 150 eV. Spectral data were collected with the analyzer at 55◦ with respect to the surface normal of the sample. SSI data-analysis software was used to calculate elemental compositions from the peak areas and to peak-fit the high-resolution spectra. An electron flood gun set at 5 eV was used to minimize surface charging. Binding energy was referenced by setting the CHx peak maximum in the C1s spectrum to 285.0 eV. Typical pressures in the analysis chamber during spectral acquisition were 10−9 torr. Contact-angle goniometry Advancing (θA ) and receding contact angles (θR ) with ddH2 O (pH approx. 7.4), were determined by the Wilhelmy-plate method [60], using a Cahn DCA 312 Contact Angle Analyzer (Cahn Instruments, Cerritos, CA, USA). Purity of ddH2 O was monitored by daily measurements of the surface tension of water (γLV ddH2 O), using freshly-flamed cover glasses, which were used as soon as they cooled to room temperature. Crosshead velocity was 10 mm /min. Three independent samples were assayed, and θA and θR were measured during the first three immersion-emergence cycles. Data presented below are the means of these nine values. Throughout this report, error bars show ±1 standard deviation. Atomic-force microscopy (AFM) Annealed, silanized surfaces were imaged in air in tapping mode with the NanoScope Dimension 3100 Scanning Probe Microscope (Digital Instruments, Santa Barbara, CA, USA). Root-mean-square (RMS) roughness, the standard deviation of height measurements (z), was calculated over a square sample area 3 µm on a side (512 × 512 pixels). Heights, widths and lengths of randomly selected surface features were measured in the same area. Images were recorded with a false-color scheme showing a 5-nm full z-axis in all cases. Statistical analysis Pair-wise comparisons of data were evaluated post hoc with the Tukey– Kramer honestly significantly different (HSD) test [61], using version 3.0 of the JMP software (SAS Institute, Cary, NC, USA). Differences were considered significant at P < 0.05.
Aminofluorosiloxane culture surfaces
409
RESULTS AND DISCUSSION
Surface composition by electron spectroscopy for chemical analysis Silicon and oxygen dominate the surface of borosilicate cover glasses cleaned with acetone (Table 1). Theoretical values for composition are based on formulation data from the glassmaker. Presence of carbon is the most striking departure of the observed composition from theoretical values. This optical glass does not contain carbonates; carbon is probably introduced by physisorption of hydrocarbons and other organic contaminants onto the surface from the atmosphere. This is perhaps inevitable on such high-energy surfaces as clean metals and glasses [35, 62– 64]. Though the surface can be cleaned by flaming, as was done with the standards used in the Wilhelmy-plate experiments (supra), it rapidly fouls upon standing in the laboratory air (in approx. 20 min). Carbon is further enriched in the alkalinedetergent-washed sample (Table 1), possibly by deposition of surfactants from the Micro cleaning solution [15]. Detergent residue would also account for the appearance of small amounts of sulfur and nitrogen at this stage (Table 1). Nitrogen, sulfur and nearly half the carbon are subsequently removed by etching in 0.5 M NaOH on the day of silanization (Table 1). Consistent with published reports that describe etching of glass surfaces at extremes of pH, the two alkaline treatments (Micro detergent, followed by the 0.5 M NaOH etchant) depleted the surface of most minor elements (Na, B, K, Zn and Ti) and the surface was concomitantly enriched with silicon and oxygen [62, 63]. Antimony (Sb) was not detected by ESCA (Table 1), probably because it is present at less than 0.1 atomic percent and because of strong interference from the 2s and 2p lines of silicon. Silanization with pure, 0.1-M solutions of the diaminosilane (N2) gave 2.5 atomic percent nitrogen, while the pure, 0.1-M trifluoropropyl silane (F3) gave 8.1 atomic percent fluorine (Table 1). The trace amount of nitrogen present in the surface silanized with a pure solution of F3 might represent residual soap. Normalized data on atomic ratios of nitrogen and fluorine show good agreement between observed surface compositions and those predicted from stoichiometry of the reactive silane monomers (Fig. 3). A slight excess of nitrogen was observed in every case except the pure aminosiloxane deposits. Reactivity of amine groups with fluorescein-5-isothiocyanate Organosiloxane deposits condensed on glass are unstable in water, a property that is often overlooked by biologists who culture cells on silanized surfaces. Hydrolysis of organosiloxanes can be greatly slowed if surfaces are heat-annealed prior to use [6, 36, 64]. An interactive series of short studies on annealing, contact-angle goniometry and reactivity of FITC toward surface amine groups (Fig. 2) demonstrated that an air-oven cure of 2 h at 100◦ C offered an acceptable compromise between annealing and loss of amine reactivity, producing silanized surfaces resistant to hydrolysis, while at the same time preserving approx. 85% of the reactive amines of the uncured surface (data not shown). Gradual loss of
410
J. R. Bain and A. S. Hoffman
amine reactivity upon heating aminosiloxanes in air might be due to reactions with CO2 , conversion of amines to amides, or the formation of internal Zwitterions when amines ‘bite back’ with free silanol anions or other anions on the surface (e.g. borates, metal oxides) [65, 66]. Figure 4 shows a monotonic decrease of binding and subsequent alkaline hydrolysis of FITC residues with increasing mole fraction
Figure 3. Normalized ESCA data on atomic percentages of nitrogen (N) and fluorine (F) in mixed aminofluorosiloxane surfaces, expressed as the ratio F/(F + N).
Figure 4. Decrease in binding and subsequent lysis of fluorescein-5-isothiocyanate residues in silanized surfaces with increasing mole fractions of the trifluoropropyl residues (r 2 = 0.977).
Aminofluorosiloxane culture surfaces
411
of silane monomer F3 in surfaces of mixed composition. These surfaces had been annealed at 100◦ C for 2 h prior to FITC conjugation. Contact angles All pure and mixed surfaces of N2 and F3 showed advancing contact angles, θA , greater than 80◦ and receding contact angles, θR , greater than 55◦ (Fig. 5). The relatively high contact angles observed on many aminosilanized surfaces have been ascribed to the aforementioned formation of internal Zwitterions and to concealment of high-energy head groups among alkane and siloxane groups [67]. Heat-annealed surfaces were stable during repeated wetting, showing only a slight decrease in θA and θR as the polysiloxanes hydrated during the first three immersion-emergence cycles. Surprisingly, both θA and θR reached their maxima at a mole fraction of 2/3 F3 (Fig. 5) and not at 100% F3, as expected. Surface roughness, hydrogen F2 CF· · ·HNH ), or lateral heterogeneity in surface chemistry bonding (e.g. might account for the peak in θA and θR at intermediate mixtures of N2 and F3. Nanometer-scale roughness, as studied with the AFM, did not vary significantly among the glass and silanized surfaces (P > 0.05, infra). In view of the insular
Figure 5. Advancing and receding water-contact angles (θA and θR , respectively) for silanized surfaces. Data enclosed in dashed ellipses do not differ from one another (Tukey– Kramer HSD test, P 0.05).
412
J. R. Bain and A. S. Hoffman
Figure 6. Contact-angle hysteresis (θH ) for silanized surfaces. Data enclosed in dashed ellipses do not differ from one another (Tukey– Kramer HSD test, P 0.05).
nature of the silanized surfaces (infra), patchiness in surface chemistry [51] seems to be the most likely explanation for the peak in θA and θR at 2/3 silane F3. Chemical heterogeneity of surfaces is often invoked to explain contact-angle hysteresis, θH , defined as θA minus θR [53]. Though we had expected θH to reach its maximum in mixed surfaces of N2 and F3, θH was in fact greatest in surfaces of pure N2 (26.6◦ ) and declined gradually with increasing mole fraction of F3, reaching a minimum of 17.9◦ in surfaces of pure F3 (Fig. 6). Declining θH with increasing fluorine content was significant at P < 0.05 (Tukey– Kramer HSD test; Fig. 6). This might have been due to greater hydration of the diamino functional groups on surfaces richer in N2 residues. Atomic-force microscopy (AFM) As expected, solvent-cleaned and alkaline-etched borosilicate glass surfaces were extremely smooth. Four randomly selected spots were evaluated on both materials and every area sampled had a root-mean-square (RMS) roughness of less than 0.5 nm (Figs 7 and 8). This is consistent with previous reports of roughness values 0.5 nm for the smoothest known surfaces of glass, muscovite mica, silica and silicon [67– 70]. On all seven silanized surfaces studied by AFM, the RMS roughness averaged more than 0.5 nm (Figs 7 and 8). Four of the seven silanized surfaces had average RMS roughness values greater than 0.9 nm (viz., the pure diaminosilanized surface, 0/3 F3, along with three mixed surfaces, 1/2 F3, 2/3 F3 and 5/6 F3; Fig. 8). These same four surfaces exhibited the most striking decoration with ellipsoid, elevated features in the false-color portraits (Fig. 7). Such features, which Brunner et al. [71] called ‘submonolayer islands’, probably represent domains of polymerized silanes on a background of flat, unmodified glass. This is consistent with the strong reporting by minor constituents of the underlying borosilicate glass in all samples evaluated by ESCA (e.g. Na, B, K, Al, Zn and Ti; Table 1). Lateral
Aminofluorosiloxane culture surfaces
413
Figure 7. False-color portraits of glass, etched-glass and silanized-glass surfaces under atomic-force microscopy. Each image represents a square sample area of 3 µm × 3 µm. This figure is published in colour on http://www.ingenta.com.
Figure 8. Root-mean-square (RMS) roughness of silanized and glass surfaces. Note the logarithmic scaling. Average RMS roughness values for all materials are statistically indistinguishable (P 0.05, Tukey– Kramer HSD test).
414
J. R. Bain and A. S. Hoffman
variability in surface roughness was most pronounced in the 0/3 F3 and 1/2 F3 surfaces (Fig. 8). One randomly selected area on the 1/2 F3 surface had an RMS roughness in excess of 7 nm (Fig. 8). Roughness and its variability (Figs 7 and 8) showed no clear relationship with contact angle and its hysteresis (Figs 5 and 6). Section analysis was performed to profile individual, randomly selected surface asperities on all materials (data not shown). Ellipsoid islands on silanized surfaces generally had lengths of 140 to 240 nm and aspect ratios (greatest length over narrowest width) of 1.3 to 1.9. Heights generally ranged from 1.5 to 5 nm. Previous investigators have noted asperities of similar size emerging from flat backgrounds of silanized surfaces. Numerous studies suggest that the size and shape of features created during silanization of flat surfaces, along with numerous other physicochemical characteristics of such surfaces, are influenced by experimental variables, including cleaning, etching and hydration prior to surface modification, as well as silanization conditions and any subsequent annealing by heat [7, 37, 67– 75]. In the most heavily studied silanized surface, residues of octadecyltrichlorosilane or OTS, with their long (C18 ), pendant alkyl chains, are known to form orderly structures, including densely packed, self-assembled monolayers, under some conditions. Independent observers have measured the height of condensed phases of OTS and similar C18 silane residues as 2.2 to 2.5 nm or so, which corresponds well to the theoretical extended-chain length of the octadecyl residue [7, 69, 71– 77]. In contrast to these orderly OTS systems, we speculate that the short, chemically disparate, trialkoxy silane monomers used in the present study (viz., N2, F3) form more random or dispersed networks, without discernible nanostructure [67]. Falsecolor images such as those shown in Fig. 7 might give the reader an exaggerated impression of surface roughness. Please note that while each image covers an area of 3 µm × 3 µm, the vertical scale is in nanometers. Though slightly rougher than clean glass, all culture surfaces evaluated in this study were quite smooth on the scale of mammalian cells.
CONCLUSIONS
Competitive chemisorption creates mixed surfaces in a rational manner Table 2 summarizes the results of this study. As we had hoped, we were able to control the surface chemistry by adjusting the reaction mixture. ESCA and FITC binding confirmed this (Figs 3 and 4). Topography of silanized surfaces was insular and complex, though their root-mean-square roughness did not differ significantly from that of smooth glass (Figs 7 and 8). Surfaces richest in diamine residues were the most hydrophilic (θA 90◦ , Fig. 5). We conclude that chemisorption from the N2-F3 organosilane system provides a simple and inexpensive, one-pot method for tailoring the chemistry of glass culture surfaces. Cell-culture data are presented in the accompanying paper [1].
Aminofluorosiloxane culture surfaces
415
Table 2. Summary of physical studies of etched glass and silanized surfaces Atomic ratio (F/F + N) by ESCA (%) Reactive amines by FITC binding Advancing water-contact angle θA (◦ ) Receding water-contact angle, θR (◦ ) Contact-angle hysteresis, θH (◦ ) Roughness by AFM, root-mean-square (nm) Morphology at nanometer scale
Etched glass 0 nd <10 <10
0/3 F3 0 high 83 57 26 1.1 islands
2/3 F3 72 low 99 76 23 1.0 islands
3/3 F3 96 trace 90 72 18 0.58 islets
Contact angles on glass refer to freshly flamed material. ESCA, electron spectroscopy for chemical analysis; nd, experiment not done.
Acknowledgements This work was supported by grants from the National Institutes of Health (NIH Nos. 1-T32-GM-08437-01 and 5-T32-HL-07403-15) and by Gore Hybrid Technologies (Flagstaff, AZ, USA). M. E. Blaylock helped with silane chemistry. D. LeachScampavia of the National ESCA and Surface Analysis Center for Biomedical Problems, University of Washington (Seattle, WA, USA) (NIH No. RR01296), assisted with ESCA. Dr. S. N. Magonov, Digital Instruments, Veeco Metrology Group (Santa Barbara, CA, USA) and P. Spevack, W. L. Gore and Associates (Elkton, MD, USA), performed the AFM. The staff at Glaswarenfabrik Karl Hecht (Sondheim/Rhön, Germany), gave freely of their knowledge of glass chemistry. B. D. Becker and D. Ballard helped with literature study. Drs. T. A. Horbett, W. R. Gombotz, P. S. Stayton and M. S. Cooper, along with three anonymous referees, made helpful suggestions on the manuscript. REFERENCES 1. J. R. Bain and A. S. Hoffman, J. Biomater. Sci. Polymer Edn 14, 341 (2003). 2. M. Shen and T. A. Horbett, J. Biomed. Mater. Res. 57, 336 (2001). 3. B. A. Dalton, C. D. McFarland, T. R. Gengenbach, H. J. Griesser and J. G. Steele, J. Biomater. Sci. Polymer Edn 9, 781 (1998). 4. C. H. Thomas, J. B. Lhoest, D. G. Castner, C. D. McFarland and K. E. Healy, J. Biomech. Eng. 121, 40 (1999). 5. B. Arkles, J. R. Steinmetz, J. Zazyczny and P. Mehta, in: Silicon Compounds: Register and Review, R. Anderson, G. L. Larson and C. Smith (Eds), 5th edn, p. 65. Hüls America, Piscataway, NJ (1991). 6. P. G. Pape and E. P. Plueddemann, J. Adhesion Sci. Technol. 5, 831 (1991). 7. J. D. Cox, M. S. Curry, S. K. Skirboll, P. L. Gourley and D. Y. Sasaki, Biomaterials 23, 929 (2002). 8. M. E. Hasenbein, T. T. Anderson and R. Bizios, Biomaterials 23, 3937 (2002). 9. A. S. Köhler, P. J. Parks, D. L. Mooradian, G. H. R. Rao and L. T. Furcht, J. Biomed. Mater. Res. 32, 237 (1996). 10. S. Saneinejad and M. S. Shoichet, J. Biomed. Mater. Res. 42, 13 (1998). 11. T. O. Collier, C. H. Thomas, J. M. Anderson and K. E. Healy, J. Biomed. Mater. Res. 49, 141 (2000).
416
J. R. Bain and A. S. Hoffman
12. K. E. Healy, B. Lom and P. E. Hockberger, Biotechnol. Bioeng. 43, 792 (1994). 13. C. D. McFarland, C. H. Thomas, C. DeFilippis, J. G. Steele and K. E. Healy, J. Biomed. Mater. Res. 49, 200 (2000). 14. W. Shain, H. Ancin, H. C. Craighead, M. Isaacson, L. C. Kam, B. Roysam, D. Szarowski, J. N. Turner and S. W. Turner, in: Proceedings of the Fifth World Congress on Biomaterials, Toronto, Canada, p. 310 (1996). 15. J. R. Bain and A. S. Hoffman, Biomaterials 23, 3347 (2002). 16. L. A. Culp and C. N. Sukenik, J. Biomater. Sci. Polymer Edn 9, 1161 (1998). 17. D. C. Giliberti, K. A. Anderson and K. C. Dee, J. Biomed. Mater. Res. 62, 422 (2002). 18. M. Matsuzawa, R. S. Potember, D. A. Stenger and V. Krauthamer, J. Neurosci. Methods 50, 253 (1993). 19. D. A. Stenger, J. H. Georger, C. S. Dulcey, J. J. Hickman, A. S. Rudolph, T. B. Nielsen, S. M. McCort and J. M. Calvert, J. Am. Chem. Soc. 114, 8435 (1992). 20. J. J. Swalec, Jr., The vapor phase deposition, polymerization and surface characterization of methyltrimethoxysilane and 3-chloropropyl trimethoxysilane on a glass substrate, Master’s thesis, University of Washington, Seattle, WA (1981). 21. K. Webb, V. Hlady and P. A. Tresco, J. Biomed. Mater. Res. 41, 422 (1998). 22. P. M. Saint John, L. Kam, S. W. Turner, H. G. Craighead, M. Isaacson, J. N. Turner and W. Shain, J. Neurosci. Methods 75, 171 (1997). 23. B. M. Cooke, S. Usami, I. Perry and G. B. Nash, Microvasc. Res. 45, 33 (1993). 24. J. A. Hubbell, C. P. Pathak, A. S. Sawhney, N. P. Desai, J. L. Hill and S. F. A. Hossainy, US Patent No. 5,858,746 (1999). 25. J. P. Ranieri, R. Bellamkonda, J. Jacob, T. G. Vargo, J. A. Gardella and P. Aebischer, J. Biomed. Mater. Res. 27, 917 (1993). 26. A. Rezania, C. H. Thomas and K. E. Healy, Ann. Biomed. Eng. 25, 190 (1997). 27. S. Britland, P. Clark, P. Connolly and G. Moores, Exp. Cell Res. 198, 124 (1992). 28. B. C. Wheeler, J. M. Corey, G. J. Brewer and D. W. Branch, J. Biomech. Eng. 121, 73 (1999). 29. D. Kleinfeld, K. H. Kahler and P. E. Hockberger, J. Neurosci. 8, 4098 (1988). 30. R. Kapur and A. S. Rudolph, Exp. Cell Res. 244, 275 (1999). 31. T. O. Acarturk, M. M. Peel, P. Petrosko, W. LaFramboise, P. C. Johnson and P. A. DiMilla, J. Biomed. Mater. Res. 44, 355 (1999). 32. K. Webb, V. Hlady and P. A. Tresco, J. Biomed. Mater. Res. 49, 362 (2000). 33. D. H. Davis, C. S. Giannoulis, R. W. Johnson and T. A. Desai, Biomaterials 23, 4019 (2002). 34. M. Franco, P. F. Nealey, S. Campbell, A. I. Teixeira and C. J. Murphy, J. Biomed. Mater. Res. 52, 261 (2000). 35. 35. R. G. Dillingham and F. J. Boerio, J. Adhesion Sci. Technol. 6, 207 (1992). 36. M. J. Wirth and H. O. Fatunmbi, Anal. Chem. 65, 822 (1993). 37. K. Kojio, S. Ge, A. Takahara and T. Kajiyama, Langmuir 14, 971 (1998). 38. J. R. Bain, A. S. Hoffman and T. A. Horbett, in: 37th International Symposium on Macromolecules, Gold Coast, Queensland, Australia, World Polymer Congress, Int. Union Pure Appl. Chem., preprints, p. 789 (1998). 39. W. L. McKeehan and R. G. Ham, J. Cell Biol. 71, 727 (1976). 40. H. J. Griesser, R. C. Chatelier, T. R. Gengenbach, G. Johnson and J. G. Steele, J. Biomater. Sci. Polymer Edn 5, 531 (1994). 41. N. Oturan, H. Serne, G. Reach and M. Jozefowicz, J. Biomed. Mater. Res. 27, 705 (1993). 42. B. Ziegler, M. Schlosser, B. Fürll, S. Witt and M. Ziegler, Diabet. Res. 16, 41 (1991). 43. M. Kriat, J. Fantini, J. Vion-Dury, S. Confort-Gouny, J. P. Galons and P. J. Cozzone, Biochemie 74, 949 (1992). 44. G. E. Grampp, A. Sambanis and G. N. Stephanopoulos, Adv. Biochem. Eng. Biotechnol. 46, 35 (1992). 45. J. G. Aunins and D. I. C. Wang, Biotechnol. Bioeng. 34, 629 (1989).
Aminofluorosiloxane culture surfaces
417
46. P. C. Letourneau, Dev. Biol. 44, 77 (1975). 47. A. Bruil, J. I. Sheppard, J. Feijen and I. A. Feuerstein, J. Biomater. Sci. Polymer Edn 5, 263 (1994). 48. U. T. Rüegg and F. Hefti, Neurosci. Lett. 49, 319 (1984). 49. J. G. Steele, B. A. Dalton, G. Johnson and P. A. Underwood, Biomaterials 16, 1057 (1995). 50. T. G. van Kooten, J. M. Schakenraad, H. C. van der Mei and H. J. Busscher, Biomaterials 13, 897 (1992). 51. J. D. Andrade, L. M. Smith and D. E. Gregonis, in: Surface and Interfacial Aspects of Biomedical Polymers. Volume 1. Surface Chemistry and Physics, J. D. Andrade (Ed.), p. 249. Plenum Press, New York, NY (1985). 52. R. J. Good, Adv. Chem. Ser. 43, 74 (1964). 53. R. E. Johnson, Jr. and R. H. Dettre, Surfact. Sci. Ser. 49, 1 (1993). 54. P. B. van Wachem, T. Beugeling, J. Feijen, A. Bantjes, J. P. Detmers and W. G. van Aken, Biomaterials 6, 403 (1985). 55. W. A. Zisman, Adv. Chem. Ser. 43, 1 (1964). 56. D. K. Pettit, T. A. Horbett and A. S. Hoffman, J. Biomed. Mater. Res. 26, 1259 (1992). 57. D. E. Williams and T. J. Tangney, in: Silanes, Surfaces and Interfaces, D. E. Leyden (Ed.), p. 471. Gordon and Breach, New York, NY (1986). 58. J. M. Harris, J. M. Dust, R. A. McGill, P. A. Harris, M. J. Edgell, R. M. Sedaghat-Herati, L. J. Karr and D. L. Donnelly, Am. Chem. Soc. Symp. Ser. 467, 420 (1991). 59. S. M. Wojcik and D. A. Puleo, Trans. Soc. Biomater. 21, 487 (1998). 60. E. A. Vogler, Surfact. Sci. Ser. 49, 183 (1993). 61. R. R. Sokal and F. J. Rolf, in: Biometry, 2nd edn, Chapter 9. W. H. Freeman, San Francisco, CA (1981). 62. B. D. Ratner, J. J. Rosen, A. S. Hoffman and L. H. Scharpen, in: Surface Contamination, K. L. Mittal (Ed.), Vol. 2, p. 669. Plenum Press, New York, NY (1979). 63. T. H. Elmer, in: Silylated Surfaces, D. E. Leyden and W. T. Collins (Eds), p. 1. Gordon and Breach, New York, NY (1980). 64. A. Menawat, J. Henry, Jr. and R. Siriwardae, J. Colloid Interf. Sci. 101, 110 (1984). 65. K. P. Battjes, A. M. Barolo and P. Dreyfuss, J. Adhesion Sci. Technol. 5, 785 (1991). 66. E. P. Plueddemann, in: Silylated Surfaces, D. E. Leyden and W. T. Collins (Eds), p. 31. Gordon and Breach, New York, NY (1980). 67. D. J. Dunaway and R. L. McCarley, Langmuir 10, 3598 (1994). 68. A. Barrat, P. Silberzan, L. Bourdieu and D. Chatenay, Europhys. Lett. 20, 633 (1992). 69. C. A. Siedlecki and R. E. Marchant, Biomaterials 19, 441 (1998). 70. N. Ishida, N. Kinoshita, M. Miyahara and K. Higashitani, J. Colloid Interf. Sci. 216, 387 (1999). 71. H. Brunner, T. Vallant, U. Mayer, H. Hoffmann, B. Basnar, M. Vallant and G. Friedbacher, Langmuir 15, 1899 (1999). 72. N. Kumar, C. Maldarelli, C. Steiner and A. Couzis, Langmuir 17, 7789 (2001). 73. D. W. Britt and V. Hlady, J. Colloid Interf. Sci. 178, 775 (1996). 74. J. W. Davidovits, V. Pho, P. Silberzan and M. Goldmann, Surf. Sci. 352-354, 369 (1996). 75. S. Ge, K. Kojio, A. Takahara and T. Kajiyama, J. Biomater. Sci. Polymer Edn 9, 131 (1998). 76. R. Resch, S. Meltzer, T. Vallant, H. Hoffmann, B. E. Koel, A. Madhukar, A. A. G. Requicha and P. Will, Langmuir 17, 5666 (2001). 77. K. Kojio, A. Takahara and T. Kajiyama, Colloids Surf. A. Physicochem. Eng. Asp. 169, 295 (2000).
This page intentionally left blank
Deterioration of polyamino acid-coated alginate microcapsules in vivo J. M. VAN RAAMSDONK 1 , R. M. CORNELIUS 2 , J. L. BRASH 2 and P. L. CHANG 1,3,∗ 1 Department
of Medical Sciences, Chemical Engineering, McMaster University, Hamilton, Ontario, Canada 2 Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada 3 Department of Pediatrics, McMaster University, Hamilton, Ontario, Canada Received 13 February 2002; accepted 22 May 2002 Abstract—The implantation of immuno-isolated recombinant cell lines secreting a therapeutic protein in alginate microcapsules presents an alternative approach to gene therapy. Its clinical efficacy has recently been demonstrated in treating several genetic diseases in murine models. However, its application to humans will depend on the long-term structural stability of the microcapsules. Based on previous implantations in canines, it appears that survival of alginate– poly- L -lysine– alginate microcapsules in such large animals is short-lived. This article reports on the biological factors that may have contributed to the degradation of these microcapsules after implantation in dogs. Alginate microcapsules coated with poly-L -lysine or poly- L -arginine were implanted in subcutaneous or intraperitoneal sites. The retrieved microcapsules showed a loss of mechanical stability, as measured by resistance to osmotic stress. The polyamino acid coats were rendered fragile and easily lost, particularly when poly-L -lysine was used for coating and the intraperitoneal site was used for implantation. Various plasma proteins were associated with the retrieved microcapsules and identified with western blotting to include Factor XI, Factor XII, prekallikrein, HMWK, fibrinogen, plasminogen, ATIII, transferrin, alpha-1-antitrypsin, fibronectin, IgG, alpha-2-macroglobulin, vitronectin, prothrombin, apolipoprotein A1, and particularly albumin, a major Ca-transporting plasma protein. Complement proteins (C3, Factor B, Factor H, Factor I) and C3 activation fragments were detected. Release of the amino acids from the microcapsule polyamino acid coats was observed after incubation with plasma, indicating the occurrence of proteolytic degradation. Hence, the loss of long-term stability of the polyamino acid-coated alginate microcapsules is associated with activation of the complement system, degradation of the polyamino acid coating, and destabilization of the alginate core matrix, probably through loss of calcium-mediated ionic cross-linking of the guluronic acid polymers in the alginate. These destructive forces may be slightly mitigated by using poly- L -arginine instead of polyL -lysine for coating and by implanting in a subcutaneous instead of an intraperitoneal site. However, the long-term stability of such devices may require significant improvements in the microcapsule polymer chemistry to withstand such biological impediments.
∗ To whom correspondence should be addressed. Department of Pediatrics, Health Sciences Centre Room 3N18, McMaster University, 1200 Main St. West, Hamilton, Ontario, Canada L8N 3Z5. Phone: (905)525-9140, ext. 22278. Fax: (905)524-5707. E-mail: [email protected]
420
J. M. Van Raamsdonk et al.
Key words: Alginate microcapsules; serum proteins; SDS-PAGE; immunoblots.
INTRODUCTION
Microencapsulation of living cells has been investigated as a means of providing a host with an adequate mass of non-autologous cells to ameliorate a disease due to deficient biomolecule or defective organ function [1, 2]. Microcapsules made of alginate, have been commonly used for such applications. Alginate is a polysaccaride extracted from seaweed composed of homopolymeric regions of 1,4-linked β-D-mannuronic acid (M-blocks) and α-L-guluronic acid (G-blocks) residues in varying proportions and arrangements interspersed with MG blocks [3]. Different types of alginate microcapsules and the behavior of the encapsulated cells have been studied in previous work. It is believed that the membrane formed by the ionic interaction between polyamino acids such as poly-L-lysine (PLL) may prevent or restrict the passage of cytotoxic components of the host’s immune system so as to protect the encapsulated cells from immune rejection, while permitting the transport of products secreted by the encapsulated cells [4]. This strategy has been extended to the use of recombinant cells so that the scope of application is no longer limited to naturally occurring cells or tissues (see reviews in refs 5 and 6). Its proofof-principle as a method for gene therapy has been demonstrated successfully in murine models of several diseases such as dwarfism, lysosomal storage disease, hemophilia, and cancer [7– 11]. However, when scale-up implantations into canine recipients were investigated [12, 13], the stability of the microcapsules was much diminished compared with implantations in the murine hosts. While microcapsules have been successfully implanted both intraperitoneally and subcutaneously into both mice and dogs, the survival of hollow alginate– poly-L-lysine microcapsules in dogs appears to be quite limited. In mice, they can be retrieved for longer than 152 days after intraperitoneal implantation, but none is retrieved after just 14 days in the peritoneal cavity of dogs. However, when the alginate was cross-linked with barium cations instead of calcium, these microcapsules could be recovered from the peritoneal cavity of dogs beyond 42 days [12]. However, since ionized barium is a neurotoxic agent, it is unlikely that such soluble barium compounds are acceptable for clinical application if it participates in ionic exchange with other cations present in the physiological fluid. Other alginate-based microcapsules that have been studied for immuno-isolation include uncoated alginate microspheres [14, 15], alginate– protamine– heparin capsules [16], alginate– poly-L-ornithine capsules [17, 18], and alginate– cellulose sulfate– poly-methylene-co-guanidine capsules [19]. To assess the long-term stability of these microcapsules, the retrieval of intact capsules from peritoneal cavities of dogs should provide a good in vivo measure of capsule durability. While it is possible that developing microcapsules that survive intraperitoneal implantation in canines may not be necessary for the microcapsules to survive in humans, testing
Protein adsorption in microcapsules
421
the microcapsules in the harsher environment of the canine peritoneal cavity should facilitate the development of stronger microcapsules than would be possible in mice. In this study, alginate microcapsules laminated with various polyamino acids and further coated with alginate were examined and possible causes for their biological instability were investigated both in vitro and after implantation in canines. MATERIALS AND METHODS
Microcapsule preparation Alginate microcapsules were fabricated based on the method described [8]. Briefly, alginate droplets were formed by extrusion through a 27-gauge needle and shearing with concentric airflow. They were cross-linked with divalent calcium ions to form the alginate core, and coated with a polyamino acid. The three polyamino acids used — poly-L-lysine hydrobromide (PLL) (MW 15 000– 30 000 D), polyL-arginine hydrochloride (PLArg) (MW 15 000– 70 000 D), and poly- L-ornithine hydrobromide (PLO) (MW 15 000– 30 000 D) — were purchased from Sigma (St. Louis, MO, USA). Finally, the microcapsules were further coated with alginate. The specific coating and washing times and conditions were according to standard procedures [20] and the citrate wash was omitted to make solid APA microcapsules. PLArg–APA and PLO –APA microcapsules were made according to the standard encapsulation protocol, except that PLArg (0.05% in 0.9% NaCl) and PLO (0.05% in 0.9% NaCl) solutions, respectively, were used instead of PLL solution. Osmotic pressure test The procedure used to test the mechanical stability of the microcapsules to osmotic stress was as described previously [21]. Briefly, microcapsules were equilibrated in serum-free media (SFM), stained with Trypan blue, and then exposed to a series of hypotonic solutions of decreasing osmotic pressure (SFM diluted to various ratios with deionized water). The percentage of broken microcapsules in each dilution was scored under the microscope. Analysis of proteins adsorbed on the microcapsules in vitro Microcapsules were incubated in vitro with saline, SFM, media (SFM supplemented with 10% fetal bovine serum), or plasma (mouse, human or canine) for 16 h at 37 ◦ C. The capsules were then rinsed extensively with phosphate-buffered saline (PBS), and exposed to 2% sodium dodecyl sulfate (SDS) solution (24 h, 4 ◦ C) to extract the proteins associated with the capsules. The samples were stored at −70 ◦ C until used for electrophoresis and western blots. Polyacrylamide gel electrophoresis (PAGE) and immunoblotting Reduced SDS-PAGE (12% gels) was carried out on the extracted samples according to published protocols [22]. The proteins were transferred from the gels to
422
J. M. Van Raamsdonk et al.
Immobilon PVDF membranes (Millipore, Bedford, MA, USA) and stained using a stabilized gold sol (Protogold, Cedarlane Laboratories, Hornby, Ontario, Canada). Immunoblot analysis was similarly performed according to published procedures [23]. The majority of the primary antibodies (dilution of 1 : 1000) utilized for this study were in the form of fractionated antisera developed in goat. Affinitypurified secondary antibodies conjugated to alkaline phosphatase were used at a dilution of 1 : 1000. Digital images of the gold-stained gels and immunoblots were analyzed using whole-band analysis software (BioImage, Ann Arbor, MI, USA). Amino acid analysis Amino acid analysis of plasma samples was performed in the McMaster University– Chedoke Hospital clinical laboratory (Dr. R. Hill) with HPLC. Canine implantation Implantation of solid PLL –APA and PLArg– APA microcapsules and retrieval in dogs were performed by CAF (central animal facility) staff in accordance with the Canadian Animal Care Guidelines. Prior to implantation, the capsules were kept in physiologic saline (0.9% NaCl) and loaded into 20 cm3 syringes. For both intraperitoneal and subcutaneous implantations, the site of implantation was shaved and sterilized immediately before implantation. The dogs were anaesthetized for both types of implantations. Subcutaneous implantations (N = 4 dogs, 2 sites per dog) were performed by injecting about 12– 15 ml of microcapsules through an IV catheter (Angiocath 16 G) into sites on the back of the dog. Intraperitoneal implantations (N = 4 dogs) were performed by injecting approximately 50 ml of microcapsules through an IV catheter into the peritoneal cavity. Implanted dogs were euthanized 2 weeks after implantation and capsules were retrieved. The subcutaneous capsules were retrieved either by putting a catheter into the subcutaneous pocket where the capsules were implanted and drawing up the capsules in a 20 cm3 syringe, or by simply cutting open the subcutaneous site and scooping out the microcapsules. The retrieved capsules were placed in physiologic saline on ice. Capsules were retrieved from the peritoneal cavity by making an opening in the intraperitoneal cavity and flushing out the capsules with 60– 150 ml of physiologic saline. It is possible that some intact microcapsules remained in the peritoneal cavity after washing, thereby reducing the apparent recovery percentage. Retrieved microcapsules were washed with physiologic saline and stored in physiologic saline at 4 ◦ C.
RESULTS
The efficacy of using PLArg or PLO instead of PLL for laminating the microcapsules was investigated first by comparing the strengths of microcapsules made
Protein adsorption in microcapsules
423
Figure 1. Mechanical stability of microcapsules coated with different poly-amino acids. On days 1, 8, 14, and 28 post-encapsulation, PLL– APA, PLA– APA, and PLO– APA microcapsules made with the same alginate core were incubated in hypotonie solutions (0.78% SFM, 0% SFM) and scored for the percentage of intact capsules. Three batches of microcapsules were tested in duplicate at each time point for each type of microcapsule. Data are averages ± SD. Significant differences between PLA and PLL microcapsules occurred in 0% SFM and in 0.78% SFM on all days tested (p < 0.05). No significant differences occurred between PLL and PLO microcapsules on any of the days in either hypotonic solution.
with the respective polyamino acids. At various time points after encapsulation, the structural stability of each type of microcapsule was determined by the osmotic pressure test (Fig. 1). The PLArg–APA microcapsules were found to be stronger than the other two types of microcapsules at each time point tested. After the osmotic pressure test, there was a significantly higher percentage of intact PLArg– APA microcapsules than PLL –APA microcapsules in 0% SFM on days 8, 14, and 28, and in 0.78% SFM on days 1, 14, and 21. In contrast, there were no significant differences between the strengths of the PLL –APA and the PLO –APA microcapsules. It was interesting to note that with the exception of the samples on day 8, the percentage of intact capsules after the osmotic pressure test remained fairly constant for all three types of capsules in all solutions tested throughout the 28 days of the experiment. It is uncertain why all the capsules appeared more resistant to breakage on day 8 when tested with the 0.78% SFM. However, it is clear that solid APA microcapsules do not become detectably weaker over time during a month-long in vitro incubation and that the arginine-coated microcapsules are more resistant to the tensile stress of the osmotic pressure test than the PLL –APA or PLO –APA microcapsules. In vivo characterization of microcapsules While in vitro studies can be used to characterize microcapsules in a controlled environment, the microcapsules must inevitably be successfully tested in vivo
424
J. M. Van Raamsdonk et al.
before considering their use in clinical trials. Hence, both the PLL –APA and the PLArg– APA types of microcapsules were implanted into dogs to assess the stability of these microcapsules in vivo. Four dogs were implanted each with approximately 50 ml of microcapsules intraperitoneally and 15 ml of microcapsules subcutaneously, with two dogs receiving solid PLL –APA microcapsules and two dogs receiving solid PLArg–APA microcapsules. Two weeks after implantation, the dogs were euthanized and the microcapsules were retrieved from both sites of implantation. While both types of microcapsules implanted subcutaneously could be recovered quantitatively, only 10% of the PLL –APA and 10– 50% of the PLArg– APA microcapsules implanted intraperitoneally could be recovered. Morphologically, both types of capsules retrieved from the intraperitoneal cavity were in much worse condition than those retrieved from the subcutaneous sites. Some of these capsules recovered from the intraperitoneal cavity were covered with cells and most appeared poorly defined and difficult to visualize under the microscope, with no defined border and unstainable with Trypan blue. The more severely affected PLL –APA microcapsules were stained and examined to gain further insight into the degradative changes that had occurred in these microcapsules (Fig. 2). Based on the observation that alginate beads alone without the polyamino acid coating are not stained by Trypan blue, while alginate beads coated with PLL are darkly stained (Fig. 2A vs. 2B — Tryrpan blue), it seems that the ghost-like microcapsules retrieved from intraperitoneal implantation have lost their polyamino acid coating, leaving just the alginate core (Fig. 2C). Moreover, since simple alginate beads showed a defined border under dark-field microscopy, the alginate core of the retrieved microcapsules may be partially degraded as the border was no longer distinct (Fig. 2A vs. 2C — dark field). Further evidence that suggests the loss of the polyamino acid layer comes from the fact that the intraperitoneally retrieved capsules stained darkly with Alcian blue, which is specific for polysaccharides such as alginate. For comparison, alginate beads without the polyamino acid coating were shown to stain darkly with Alcian blue (Fig. 2A), while microcapsules coated with PLL only showed light Alcian blue staining (Fig. 2B). However, the intraperitoneally retrieved capsules, though stained darkly initially, completely dissolved in the phosphate-buffered saline used in the rinsing after the Alcian blue staining. Hence, no photograph could be taken, thus indicating a dramatic destabilization of the microcapsules after the implantation (Fig. 2C). In contrast to the intraperitoneally retrieved capsules, all of the capsules retrieved from the subcutaneous sites appeared morphologically normal after the 2-week implantation (Fig. 2D). On comparing the intraperitoneally retrieved and subcutaneously retrieved microcapsules, it is clear that only the capsules retrieved from subcutaneous sites still had a defined capsular membrane, as observed under dark-field microscopy (Fig. 2C vs. 2D). However, this membrane was visibly thinner than the capsular membranes of the preimplantation APA microcapsules (Fig. 2B) and some of the subcutaneously retrieved capsules were seen to have lost their capsular
Protein adsorption in microcapsules
425
Figure 2. Staining properties of PLL–alginate microcapsules. Various types of microcapsules from preparations kept in vitro or retrieved from intraperitoneal (IP) or subcutaneous (SC) implantations in dogs were photographed under dark-field illumination (first column). They were then stained with Trypan blue to reveal the polyamino acid coat (second column) and Alcian blue to indicate polysaccharides (third column) and photographed under light-field illumination. Alginate beads: microcapsules without polyamino acid coating. PLL–alginate beads: microcapsules with poly- L lysine coating. IP-retrieved capsules: microcapsules with poly- L -lysine coating retrieved after 14 days from intraperitoneal implantation in dogs. SC-retrieved capsules: microcapsules with poly- L -lysine coating retrieved after 14 days from subcutaneous implantation in dogs. Plasma-incubated capsules: microcapsules with poly-L -lysine coating after incubation in dog plasma in vitro. (This figure is published in colour on http://www.vsppub.com/jconts/JBS)
membranes upon staining and washing with PBS during the osmotic pressure test (Fig. 3). For both types of microcapsules, the capsules retrieved from the subcutaneous implantation sites were detectably weaker than the preimplantation microcapsules (Fig. 4). Unfortunately, it was not possible to measure the strength of the intraperitoneally retrieved ghost-like microcapsules because the lack of Trypan blue staining precluded visualization of capsule breakage necessary for scoring the percent-
426
J. M. Van Raamsdonk et al.
Figure 3. Loss of the membranes in microcapsules retrieved from canine implantations. On day 14 post-subcutaneous implantation in canines, PLA–APA or PLL–APA microcapsules were retrieved for the osmotic pressure test. During the test, it was observed that the membranes on the surface of the microcapsules started to peel off. The capsular membrane stained well with Trypan blue, indicating the presence of the polyamino acid coating layer. (This figure is published in colour on http://www.vsppub.com/jconts/JBS)
Figure 4. Mechanical stability of microcapsules after implantation in canines. PLA–APA (upper panel) and PLL–APA (lower panel) microcapsules were retrieved on day 14 post-implantation from the subcutaneous (SC) sites in dogs and compared with the osmotic pressure test to control microcapsules kept in vitro. The percentage of intact capsules was determined. The data are averaged from triplicate determinations ± SD.
Protein adsorption in microcapsules
427
age of broken microcapsules. At all of the hypo-osmolarities tested (0%, 0.78%, 3.25% SFM), the percentage of intact capsules was always higher in the control preimplantation than in the subcutaneously retrieved microcapsules (Fig. 4). The difference became more dramatic as the osmolarity of the test solution was decreased (0% > 0.78% > 3.25%). The retrieval of ghost microcapsules from the intraperitoneal cavity of dogs suggested that the microcapsules can lose their capsular membrane. It is possible that the breakdown of the polyamino acid layer by the fibrinolytic system or other mechanisms contributes to the loss of the outer layers. If the PLL or PLArg was in fact being broken down, it should be possible to detect PLL or PLArg being released from the capsules. This hypothesis was tested by incubating the microcapsules in plasma (overnight at 37 ◦ C) and then assessing the amino acid composition of the plasma before and after incubation. Macroscopically, the overnight incubation in plasma resulted in the microcapsules appearing more transparent and shiny (Fig. 2E). When the amino acid concentration of the plasma was determined with HPLC and corrected for dilution attributed to the hydrogel composition of the microcapsules (∼90% water), it was found that the levels of lysine in plasma incubated with PLL –APA microcapsules were 21% greater than control plasma and the plasma incubated with PLArg–APA microcapsules had 22% more arginine than the control (Fig. 5). In addition, the plasma incubated with PLArg– APA microcapsules showed an increase in several amino acids other than arginine,
Figure 5. Loss of polyamino acid from microcapsules during plasma incubations. One milliliter of PLL–APA and PLA–APA microcapsules, respectively, was incubated overnight in 5 ml of dog plasma. A control sample of plasma was incubated without microcapsules. The amino acid content of these three plasma samples was analysed with HPLC and the increases in amino acid content over the control samples were calculated after correction for dilution by the microcapsules.
428
J. M. Van Raamsdonk et al.
including threonine, serine, glycine, alanine, tyrosine, phenylalanine, ornithine, and histidine, while the plasma incubated with PLL –APA capsules only showed increases in serine and lysine (data not shown). Analysis of proteins adhering to microcapsules after implantation in canines or in vitro incubation in murine, canine or human plasma Since microcapsules retrieved after implantation were weaker than control microcapsules, it is possible that biological causes in the implantation environment may have adversely affected the stability of the microcapsule. To identify which proteins may have played a role in capsule weakening, the proteins bound to the PLL –APA
Figure 6. Proteins associated with PLL or PLA microcapsules. Proteins bound to microcapsules were compared with reduced SDS-PAGE gel. Molecular mass in kDa are shown on the left-hand side of the figure. (A) Proteins from PLL–APA microcapsules incubated in mouse plasma (mouse), dog plasma (dog) or human plasma (human). One milliliter of capsules was incubated overnight in 5 ml of plasma for the canine and human plasmas. For the mouse plasma, 150 µl of capsules was incubated overnight in 500 µl of plasma. All incubated microcapsules were washed four times with PBS before 100 µl samples were taken for protein extraction with 2% SDS. The first three columns show 1 µl of protein extract. The second three columns show 2 µl of protein extract. (B) Proteins from PLL–APA microcapsules (PLL) incubated overnight in saline, serum-free media (SFM), media containing fetal bovine serum (media), dog plasma (dog) or human plasma (human). Results for PLA– APA microcapsules (PLA) incubated in saline, dog plasma, and human plasma are also shown. The experimental conditions are as described in A. (C) Comparison of the intensity profiles obtained for protein associated with PLL capsules and PLA capsules following exposure to dog plasma. The gold-stained gels and immunoblots were scanned and their digital images were analyzed using wholeband analysis software (BioImage, Ann Arbor, MI, USA) to determine band molecular weights and intensities (Y -axis). The relative migration value (Rf ) on the X-axis is the ratio of the distance migrated by the protein from the origin to that migrated by the dye front.
Protein adsorption in microcapsules
429
and PLArg–APA microcapsules retrieved after overnight incubation in murine, canine or human plasma in vitro (Fig. 6), or retrieved from the canine implantations in vivo (Fig. 7), were extracted with SDS for analysis on reduced SDS-PAGE gels. The proteins eluted from PLL –APA microcapsules incubated in murine, canine, and human plasma showed qualitatively similar banding patterns in human and murine plasma (Fig. 6A). Interestingly, the pattern and intensity of the protein bands for the proteins isolated from capsules incubated in canine plasma were different from the results from the other two plasmas. This suggests that either different proteins bind to microcapsules in canines compared with mice or humans, or the same proteins in canines have different electrophoretic mobilities from those in mice and humans. The proteins eluted from microcapsules incubated in canine and human plasma showed that the same proteins bind both PLL –APA and PLArg– APA microcapsules (Fig. 6B). This result was expected, since the outer surface of both types of
Figure 6. (Continued).
430
J. M. Van Raamsdonk et al.
Figure 7. Proteins bound to microcapsules in plasma in vitro and after implantation in vivo. Proteins bound to microcapsules either implanted for 14 days intraperitoneally (IP) or subcutaneously (SC) in canines, or incubated in canine plasma in vitro were compared with reduced SDS-PAGE gel. For the in vitro incubation, 1 ml of capsules was incubated overnight in 5 ml of plasma. Both in vivo and in vitro samples were washed four times with PBS before proteins were extracted from the surface of 100 µl samples of microcapsules using 2% SDS for 24 h. Control microcapsules were kept in physiologic saline solution. Lanes are identified by the type of microcapsules and the type of incubation (PLL = PLL– APA microcapsules; PLA = PLA–APA microcapsules; control = capsules kept in saline; plasma = plasma-incubated capsules). Note that bands appearing in the PLA control lane are likely spill-over from the adjacent lane.
microcapsules was coated with alginate. A number of clear but faint bands were noted in microcapsules incubated in media, indicating that there were some proteins associated with the capsules following exposure to media containing fetal bovine serum. Control microcapsules incubated in serum-free medium or saline did not show any significant protein binding (Fig. 6B). The similarity in banding patterns obtained for the different capsule types may suggest that the binding, activation, and degradation of proteins due to contact with the capsules are alike for the two polyamino acids (Fig. 6C). Comparison of the banding patterns obtained for the proteins associated with the PLL and PLArg capsules following exposure to dog plasma showed that they were very similar. The molecular weights of the bands suggest that transferrin (80 kD), albumin (66 kD), and IgG (55, 27 kD), as well as other proteins, are present (Fig. 6B). On the other hand, the banding patterns for the proteins eluted from the intraperitoneally and subcutaneously implanted microcapsules appeared somewhat different for the PLL –APA and PLArg–APA microcapsules (Fig. 7). With the exception of one or two bands, the protein bands and their relative intensities were similar
Protein adsorption in microcapsules
431
for the proteins eluted from intraperitoneally implanted, subcutaneously implanted, and plasma-incubated PLArg–APA microcapsules. However, for the PLL –APA microcapsules, the intraperitoneally implanted and subcutaneously implanted microcapsules showed fewer protein bands than the plasma-incubated microcapsules and it was also noted that the relative intensities of bands within a sample were also different between all three samples. The antigenic origin of the proteins that bind to the microcapsules was further analyzed by performing western blots on the proteins isolated from microcapsules incubated in human plasma (Fig. 8). Immunoblots were run on the proteins associated with the capsules (PLL, PLArg) following exposure to human plasma. The extreme right and left lanes show the blots after gold staining, and give a rough indication of the molecular weights of the more abundant proteins in the sample. Both PLL –APA and PLArg–APA microcapsules had bound varying amounts of Factor XI, Factor XII, prekallikrein, high-molecular-weight kininogen (HMWK), fibrinogen, plasminogen, ATIII, C3, transferrin, alpha-1-antitrypsin, fibronection, albumin, IgG, beta-lipoprotein, alpha-2-macroglobulin, vitronectin, protein C, prothrombin, haemoglobin, Factor B, Factor I, protein S, and apolipoprotein A1 (Figs 8A and 8B). Large amounts of C3, transferrin, albumin, IgG, Factor B, Factor H, and apolipoprotein A1 were bound to both types of microcapsules. These data indicated that the contact phase coagulation proteins (Factor XI, Factor XII, prekallikrein) were detected. A clear, strong band was noted for plasminogen. Activation of plasminogen to plasmin does not appear to have occurred, as noted by the absence of a band at 60 kD. This suggests that activation of the fibrinolytic system, due to contact with the capsules and the subsequent activation of the intrinsic coagulation system, has not occurred. A strong band at 57 kD was noted for ATIII. C3 is an abundant complement protein. The control blot (not shown here) shows that the C3 molecule is intact at a molecular weight of 110 kD (heavy α-chain) and 70 kD (light β-chain). However, the protein associated with the PLL capsules showed a decrease in the intensity of the 110 kD band and the appearance of numerous lower-molecular-weight bands, suggesting that the complement system has been activated. A strong band at 80 kD was noted for transferrin. Fibronectin, an adhesive protein that binds cells to other cells as well as solid surfaces, is a dimeric glycoprotein with subunit polypeptides of 200 kD. A clear, thin, distinct band depicting the intact protein was present in the protein samples obtained following exposure of PLL capsules to human plasma. A very strong response to albumin was noted. Strong bands at 55 and 27 kD depicting the heavy and light chains were noted for IgG. A faint response was noted for alpha-2-macroglobulin. Vitronectin is an adhesive protein and a strong clear response depicting the intact molecule was noted at 70 kD. Strong positive responses were also noted for the complement proteins Factor B, Factor H, and Factor I. A very strong response for apolipoprotein A1, at a molecular weight of 27 kD, was noted. The relative binding of fibrinogen, Factor XI, Factor XII, and alpha-2-macroglobulin was greater for the PLArg– APA microcapsules than for the PLL –APA microcapsules.
432
J. M. Van Raamsdonk et al.
Figure 8. Western blots of proteins bound to microcapsules after overnight incubation in human plasma. Western blots for PLL–APA (A) and PLA–APA (B) microcapsules are shown. One milliliter of each type of microcapsule was incubated overnight in 5 ml of human plasma. After washing four times with PBS, the proteins from a 100 µl sample of these microcapsules were extracted with 2% SDS. After electrophoresis and transfer, human antibodies were used to identify the specific proteins. Molecular size in kD are shown on the left-hand side of the figure. Proteins targeted by antibody are identified at the top of the figure. Lanes on the extreme left and right were stained with total protein gold stain. The remaining lanes are the immunostained patterns for specific antisera as indicated.
For comparison, human antibodies were also used to identify proteins isolated from microcapsules incubated in canine and mouse plasma (data not shown). The resulting western blots were qualitatively similar to those of the human plasma. However, the response was weaker, or in some cases absent, likely due to the fact that the antibodies are directed against human proteins rather than dog or mouse. Nevertheless, these data confirmed that a contact phase of coagulation has been activated, as suggested by the degradation of prekallikrein. The fibrinolytic protein, plasminogen, is intact. Some activation of the complement system has occurred, as evident by the appearance of lower-molecular-weight fragments for C3. The results of these western blot analyses are summarized in Tables 1 and 2.
Protein adsorption in microcapsules
433
Figure 8. (Continued).
DISCUSSION
Microcapsules are formed through the gelation of a biopolymer around cells or tissues which are to be encapsulated. While several different types of microcapsules have been developed, the most studied microcapsules are alginate– poly-Llysine– alginate (APA) microcapsules, originally developed for the encapsulation of pancreatic islets [24, 25]. These capsules are primarily composed of alginate, a naturally produced polysaccharide composed of D-mannuronic acid (M) residues and α-L-guluronic acid (G) residues [3]. The restoration of normoglycemia in diabetic animals was achieved by the implantation of pancreatic islets encapsulated in APA microcapsules in mice, rats, dogs, monkeys and humans [26– 30]. While insulin was delivered for up to 2 years in these studies, normal regulation of glucose levels could not be achieved. APA microcapsules have also been used to treat hepatic failure through the encapsulation of hepatocytes, and to treat kidney failure through the encapsulation of urease-secreting bacteria [31, 32].
434
J. M. Van Raamsdonk et al. Table 1. Relative levels of protein adherence to PLL–APA and PLA–APA microcapsules after overnight incubation in human plasma Protein
PLL–APA
PLA– APA
Factor XI Factor XII Prekallikrein HMWK Fibrinogen Plasminogen ATIII C3 Transferrin Alpha-1-antitrypsin Fibronectin Albumin IgG Beta-lipoprotein Alpha-2-macroglobulin Vitronectin Protein C Prothrombin Haemoglobin Factor B Factor H Factor I Protein S Apolipoprotein Al
± − + + ++ ++ ++ + + ++ +++ + ++ + + ++ +++ ± + + + + ± +++ +++ ++ ± +++
+ + + + + + ++ ++ ++ +++ ++ + ++ + + ++ +++ ± ++ + + + ± +++ +++ ++ ± +++
The intensities of the bands in the western blots from Figs 8A and 8B were compared qualitatively by visual scoring. The levels of protein adherence are indicated by − for no binding, ± for trace amounts, or +’s according to the intensity and number of bands present on western blots.
While alginate microcapsules are appealing because they are well characterized and biocompatible, their usefulness may be limited by their structural instability. It has recently been shown that the fibrinolytic system can recognize the poly-Llysine in APA microcapsules, leading to plasmin-mediated degradation of the PLL and subsequent weakening of the microcapsules [33]. The efficacy of using PLArg or PLO instead of PLL for microencapsulation was investigated first by comparing the strength of microcapsules made with PLL, PLArg or PLO. It was found that the PLArg–APA microcapsules were stronger than the other two types of microcapsules at each time point tested. After the osmotic pressure test, there was a significantly higher percentage of intact PLArg– APA microcapsules than PLL –APA and no significant differences were detected between the strengths of the PLL –APA and the PLO –APA microcapsules (Fig. 1).
Protein adsorption in microcapsules
435
Table 2. Relative protein binding to capsules incubated in mouse, dog, and human plasmas Protein
Mouse plasma
Dog plasma
Human plasma
Factor XI Factor XII Prekallikrein HMWK Fibrinogen Plasminogen ATIII C3 Transferrin Alpha-1-antitrypsin Fibronectin Albumin IgG Beta-lipoprotein Alpha-2-macroglobulin Vitronectin Protein C Prothrombin Haemoglobin Factor B Factor H Factor I Protein S Apolipoprotein Al
± ± + + + + ++ ++ ++ ± − +++ ++ + − ± ± + + ++ ++ + ± ++
+ ± ++ + + ++ ++ ++ ++ ± ++ + + ++ + + ± ± ± + + ++ ++ ++ ± + + ++
± − + + ++ ++ ++ + + ++ +++ + ++ + + ++ +++ ± + + + + ± +++ +++ ++ ± +++
PLL–APA microcapsules wtere incubated overnight in mouse, dog, and human plasmas and western blots were performed as described for Fig. 8 (data not shown). The intensities of the bands in the western blots were compared qualitatively by visual scoring. Levels of protein adherence are indicated by − for no binding, ± for trace amounts, or +’s according to the intensity and number of bands present on western blots.
The present work was designed to study the durability of solid microcapsules implanted in canines. This study supported previous findings that subcutaneous pockets provide a better site for implantation than the peritoneal cavity in terms of microcapsule stability [13]. Both the yield and the condition of the retrieved microcapsules were significantly better for capsules implanted subcutaneously than for capsules implanted intraperitoneally. While the yield from the peritoneal cavity may not represent all of the intact microcapsules present after 2 weeks, unretrieved microcapsules cannot account for the large difference in microcapsule recovery and the condition of the peritoneally retrieved microcapsules suggests that microcapsules at this site are indeed broken down. This difference in microcapsule condition may be the result of the chemical and physical environment at the site of implantation, but it seems more likely to be caused by specific processes
436
J. M. Van Raamsdonk et al.
more prevalent in the peritoneal cavity acting to weaken the microcapsules. This is supported by the discovery of ghost-like microcapsules only in the peritoneal cavity. For these microcapsules, it appears that the capsular membrane is being totally degraded or that it is separated from the alginate core (Fig. 2). Based on the shedding of the capsular membrane during the osmotic pressure test in microcapsules retrieved from subcutaneous implantation (Fig. 3), it seems likely that the capsular membrane is vulnerable to shedding in the intraperitoneal cavity. Furthermore, since the ghost-like microcapsules do not have a defined border, it appears that the alginate core itself may be subject to chemical or biological degradation. The observed loss of the capsular membrane (Fig. 3) and the disintegration of the alginate core during washes in PBS after retrieval (Fig. 2) provide some insight into why the majority of the implanted microcapsules could not be retrieved from the intraperitoneal cavity. It is possible that the alginate core is readily depolymerized after the microcapsule has lost the protective polyamino acid outer membrane. The strong adsorption of albumin to the microcapsules (Fig. 8) is consistent with depolymerization of the alginate due to loss of Ca2+ , as albumin is a major plasma protein responsible for the transport of Ca2+ in the blood. It has also been shown that the fibrinolytic system can recognize the PLL in these microcapsules, leading to plasmin-mediated degradation of the PLL and the subsequent weakening of the microcapsules [33]. While it is uncertain to what extent this occurs in vivo and how it may affect the stability of the microcapsules, it is possible that this potentially degradative process can be avoided by replacing PLL with another positively charged polymer. Since the structural properties of other polyamino acids such as polyarginine (PLArg) and polyornithine (PLO) are predicted to be similar to PLL, it follows that replacing PLL with either PLArg or PLO should have very little effect on the properties of the microcapsule, but may serve to avoid degradation by the fibrinolytic system. Overall, it was found that despite improved mechanical stability from coating with PLArg, the microcapsules are still lost over time when they are implanted intraperitoneally into canines. On the other hand, capsules implanted subcutaneously are well maintained over time. It has recently been shown that it is possible to deliver human growth hormone from microcapsules implanted subcutaneously in dogs [13]. Hence, capsules implanted subcutaneously may achieve a more prolonged delivery compared with the intraperitoneal site, due to an increased survival of the microcapsules. It is of interest to note that other groups have reported the survival of intraperitoneally implanted APA microcapsules for more than 2 years in both canines and monkeys [29, 30]. The success in these canines may be related to the administration of sub-therapeutic levels of immunosuppressant to inhibit the mannuronic acid-triggered cytokine response as well as complement activation, which may otherwise have contributed to a more immediate demise of the microcapsules. While it was shown that microcapsules are weaker after an intraperitoneal implantation in canines, it is uncertain what forces are acting to weaken the microcapsules. Undoubtedly, there are physical forces, most notably compressive and
Protein adsorption in microcapsules
437
shearing forces, that act on the microcapsules, but there must also be chemical and biological forces contributing to the breakdown of the microcapsules. Moreover, since there was a large difference in the condition of microcapsules retrieved from the peritoneal cavity and subcutaneous sites, it is likely that either the chemical environment or some biological agent has a greater effect in the peritoneal cavity than at subcutaneous sites. In order to identify biological factors that may be contributing to microcapsule weakening, proteins adsorbed to retrieved microcapsules were extracted and compared for capsules retrieved from the peritoneal cavity and from subcutaneous sites. Reduced SDS-PAGE gels of proteins isolated from both the PLL –APA and the PLArg– APA microcapsules showed that similar proteins were adsorbed to microcapsules implanted at both sites with small variations in band intensities. There were less distinct protein bands observed for the PLL –APA capsules than for the PLArg– APA capsules retrieved from both sites. It is likely that the reduced band intensity is related to the hypothesized loss of the capsular membrane (Fig. 2). Since both types of microcapsules have an outer coating of alginate, it is expected that they should show similar protein binding activities under normal conditions. This is supported by the fact that proteins isolated from both types of capsules incubated overnight in vitro in canine plasma showed similar protein banding patterns and relative intensities. It was also shown that the protein banding patterns for the PLArg–APA microcapsules incubated in plasma were the same as those for the PLArg–APA microcapsules implanted intraperitoneally or subcutaneously. The fact that the banding patterns were different for the PLL –APA microcapsules suggests the possibility that these implanted microcapsules have a different surface from capsules incubated in plasma, due to the loss of the capsular membrane or degradation of the capsule surface. Since it appeared that the same proteins were binding to capsules incubated in plasma that were binding to the microcapsules in vivo, the specific proteins involved were identified by western blotting after PLL –APA and PLArg– APA microcapsules were incubated in human plasma overnight (Fig. 8). It should perhaps be acknowledged at this point that while SDS is generally found to be a powerful eluant, it is possible that not all of the adsorbed proteins are eluted from the capsule surface. Furthermore, it is recognized that the capsules allow the exchange of small molecules, such as oxygen and nutrients, as well as the diffusion of other molecules ranging from 22 to 300 kD. Thus, proteins that are existing within the capsules, but not necessarily adsorbed to the capsule surface, may be present in the SDS-protein samples obtained. All of the proteins tested adsorbed to both types of microcapsules in similar amounts, either because both types of capsules presented an identical surface for protein adherence or because the proteins bind different surfaces equally, based on their prevalence in the plasma. The proteins tested included cell adhesive proteins (fibronectin, vitronectin, and fibrinogen), coagulation factors (Factor XI, Factor XII, prekallikrein, HMWK, fibrinogen, antithrombin III, protein C, and prothrombin), fibrinolytic factors (plasminogen, α1 -antitrypsin, α2 -macroglobulin, and protein C),
438
J. M. Van Raamsdonk et al.
complement proteins (C3, Factor B, Factor H, and Factor I), transport proteins (transferrin, albumin, and haemoglobin), and an immune mediator (IgG). Binding of components of the fibrinolytic system may contribute to the weakening of the microcapsules through plasmin-mediated cleavage of PLL in the capsular membrane. Since albumin is known to carry calcium ions in the blood, the large amounts of albumin present on the surface of the microcapsules may be able to bind cross-linking calcium ions, thereby depolymerizing the alginate core. The large amounts of complement proteins (C3, Factor B, and Factor H) bound to the microcapsules suggests the possibility of a complement-mediated attack on the microcapsules. Finally, the activation of an immune response can be mediated by leukocyte activation through attachment to an adhesive protein or activation of bound complement [34]. The weakening of microcapsules implanted intraperitoneally is much more rapid and significant in dogs than it is in mice. Reduced SDS-PAGE gels comparing proteins isolated from microcapsules incubated in mouse, canine or human plasma showed that while similar proteins bind to the microcapsules in human and mouse plasma, it appears that some different proteins bind to the microcapsules incubated in canine plasma (Fig. 6A). It was also observed that the capsules incubated in the canine plasma bind more low-molecular-weight fragments, which may correspond to activated cleavage products. The observed differences in protein binding between capsules incubated in mouse or dog plasma may explain the increased weakening of the microcapsules observed in canine implantations compared with mouse implantations. It is possible that one of the proteins responsible for weakening the microcapsules implanted intraperitoneally in dogs either does not bind or binds to a different extent in mice. Since the proteins binding to microcapsules in human plasma are more similar to those in mouse plasma than in canine plasma, it may be more appropriate to test the microcapsules in mice than in dogs. Western blots were used to compare the levels of specific proteins bound to the microcapsules after incubation in human, canine, and mouse plasma, and to possibly identify proteins responsible for weakening the microcapsules (Table 2). These blots showed that more prekallikrein, plasminogen, fibronectin, albumin, and apolipoprotein A1 were bound to capsules incubated in canine plasma than to capsules incubated in mouse plasma. While these results should be taken with caution since human antibodies were used, it is possible that the increased binding of plasminogen and albumin contribute to the more rapid weakening of the microcapsules in canines for reasons mentioned earlier. In conclusion, the stability of the alginate microcapsules after implantation in mammals is dependent on several biological factors: the type of polyamino acid coating, the site of implantation, and the degradative host response. In general, polyarginine-coated alginate microcapsules may offer greater mechanical stability, and subcutaneous implantation leads to a more long-term survival of the microcapsules than intraperitoneal implantation. However, it was also clear that the host biological responses, such as complement activation and destabilization of
Protein adsorption in microcapsules
439
the alginate hydrogel through loss of ionic cross-linking, may lead to the eventual demise of the microcapsules. Hence, to ensure the long-term survival of the implanted microcapsules, the polymer chemistry of the microcapsules must be improved to withstand these environmental influences in vivo. Acknowledgements We would like to thank the CIHR for support of this research, Dr. Feng Shen for advice and reading the manuscript, Drs K. Delaney and J. Kwiecin for veterinary help in the canine surgeries, Dr. R. Hill for amino acid analysis, and Ms. S. Barsoum for organizing the surgeries. J. van Raamsdonk was the recipient of an NSERC postgraduate student scholarship.
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
T. M. S. Chang, Science 146, 524 (1964). F. Lim and A. M. Sun, Science 210, 908 (1980). O. Smidsrod and G. Skjak-Braek, Trends Biotechnol. 8, 71 (1990). M. F. A. Goosen, G. M. O’Shea, H. Gharapetian, S. Chou and A. M. Sun, Biotechnol. Bioeng. 27, 146 (1985). P. L. Chang, in: Somatic Gene Therapy, P. L. Chang (Ed.), pp. 203– 223. CRC Press, Boca Raton, FL (1995). P. L. Chang, J. M. Van Raamsdonk, G. Hortelano, S. C. Barsoum, N. C. MacDonald and T. L. Stockley, TIBTECH 17, 78 (1999). A. Al-Hendy, G. Hortelano, G. S. Tannenbaum and P. L. Chang, Hum. Gene Ther. 6, 165 (1995). C. J. D. Ross, M. Ralph and P. L. Chang, Exp. Neurol. 166, 276 (2000). C. J. D. Ross, L. Bastedo, S. A. Maier, M. S. Sands and P. L. Chang, Hum. Gene Ther. 11, 2117 (2000). J. M. Van Raamsdonk, C. J. D. Ross, M. A. Potter, S. Kurachi, K. Kurachi, D. W. Stafford and P. L. Chang, J. Lab. Clin. Med. 139, 35 (2000). P. Cirone, J. Bourgeois, R. C. Austin and P. L. Chang, Hum. Gene Ther. 13, 1157 (2002). M. A. Peirone, K. Delaney, J. Kwiecin, A. Fletch and P. L. Chang, Hum. Gene Ther. 9, 195 (1998). T. L. Stockley, K. Robinson, K. Delaney, F. A. Ofosu and P. L. Chang, J. Lab. Clin. Med. 135, 484 (2000). R. P. Lanza, W. M. Kuhtreiber, D. M. Ecker and W. L. Chick, Trans. Proc. 27, 3211 (1995). R. P. Lanza, W. M. Kuhtreiber, D. Ecker, J. E. Staruk and W. L. Chick, Trans. Proc. 28, 835 (1996). K. Tatarkiewicz, E. Sitarek, M. Sabat and T. Orlowski, Trans. Proc. 27, 617 (1995). R. Calafiore, G. Basta, L. Osticioli, G. Luca, C. Tortoioli and P. Brunetti, Trans. Proc. 28, 812 (1996). R. Calafiore, G. Basta, C. Boselli, A. Bufalari, G. M. Giustozzi, G. Luca, C. Tortoioli and P. Brunetti, Trans. Proc. 29, 2126 (1997). T. Wang, I. Lacik, M. Brissova, A. V. Anilkumar, A. Prokop, D. Hunkeler, R. Green and K. Shahrokhi, Nature Biotech. 15, 358 (1997). J. Van Raamsdonk, M.Sc. Thesis, McMaster University, Hamilton, Ontario (1999). J. Van Raamsdonk and P. L. Chang, J. Biomed. Mater. Res. 54, 264 (2001). R. M. Cornelius and J. L. Brash, J. Biomater. Sci. Polym. Ed. 4, 291 (1993).
440
J. M. Van Raamsdonk et al.
23. R. M. Cornelius, J. G. Archambault, L. Berry, A. K. C. Chan and J. L. Brash, J. Biomed. Mater. Res. 60, 622 (2002). 24. F. Lim and R. D. Moss, J. Pharm. Sci. 70, 351 (1981). 25. A. M. Sun, Methods Enzymol. 137, 575 (1988). 26. Z.-P. Lum, M. Krestow, I. T. Tai, I. Vacek and A. M. Sun, Transplant 53, 1180 (1992). 27. P. De Vos, B. De Haan and R. Van Schilfgaarde, Biomaterials 18, 273 (1997). 28. P. Soon-Shiong, E. Feldman, R. Nelson, R. Heintz, Q. Yao, Z. Yao, T. Zheng, N. Merideth, G. Skjak-Braek, T. Espevik, O. Smidsrod and P. Sandford, Proc. Natl. Acad. Sci. USA 90, 5843 (1993). 29. Y. Sun, X. Ma, D. Zhou, I. Vacek and A. M. Sun, J. Clin. Invest. 98, 1417 (1996). 30. P. Soon-Shiong, R. E. Heintz, N. Merideth, Q. X. Yao, Z. Yao, T. Zheng, M. Murphy, M. K. Moloney, M. Schmehl, M. Harris, Ro. Mendez, Ra. Mendez and P. A. Sandford, Lancet 343, 950 (1994). 31. P. A. Rivas-Vetencourt, E. D. Aranda, L. Sorio, Z. Quero, A. Martinez, A. M. Vegas and M. J. Zerpa, Trans. Proc. 29, 920 (1997). 32. S. Prakash and T. M. Chang, Nature Med. 2, 883 (1996). 33. B. R. S. Hsu, S. H. Fu, J. S. Tsai, Y. Y. Huang, H. S. Huang and K. S. S. Chang, Trans. Proc. 29, 1877 (1997). 34. J. E. Babensee, R. M. Cornelius, J. L. Brash and M. V. Sefton, Biomaterials 19 (7– 9), 839 (1998).
Water structure around enkephalin near a GeO2 surface: a molecular dynamics study AARON M. BUJNOWSKI ∗ and WILLIAM G. PITT † Chemical Engineering Department, Brigham Young University Provo, UT 84602, USA Received 3 December 2001; accepted 8 April 2002 Abstract—A molecular model was created consisting of leucine enkephalin (a pentapeptide protein) near a germanium dioxide (GeO2 ) surface surrounded by water molecules. A molecular dynamic (MD) simulation of the system was conducted, and forces exerted by the water on both the surface and the protein were calculated. Orientational and spatial distribution functions of the water were calculated and were examined to determine if structured water existed and how it contributed to the intermolecular forces. Results from the study demonstrated that there is a strong spatial and orientational structuring of water near the GeO2 surface and that it is dependent on the proximity of the protein to the surface. Only minimal structuring occurred near the protein. A linear correlation was observed between the force of water on the protein and the angular distribution of water molecules in the region with the highest spatial density. When the protein is oriented with its polar side toward the GeO2 surface, the water structure near the protein disrupts the typical structure of the water near the surface, causing a force that pushes the protein away from the surface. When the protein is oriented with its non-polar side toward the surface, there is only minimal disruption of the typical water structure, and the resulting net forces between surface and protein are attractive. Key words: Water structure; germanium dioxide; enkephalin; molecular dynamics; protein adsorption.
INTRODUCTION
A common problem associated with the use of any biomaterial in the body is protein adsorption, which can lead to several adverse effects. Among these are thrombus formation [1], bacterial adhesion [2], protein adsorption and denaturation [3], and surface fouling [4]. Unfortunately, most macroscopic experimental methods used to study protein adsorption do not reveal the molecular mechanisms underlying the association of a protein with a surface. ∗ Current
address: Dow Chemical Company, Freeport, TX, 77541, USA. should be addressed. 350 Clyde Bldg, Brigham Young University, Provo, UT 84602, Phone: 801-422-2589 office. Fax: 801-422-0151. E-mail: [email protected] † To whom correspondence
442
A. M. Bujnowski and W. G. Pitt
The objective of this research was to study the structure of water near the protein and the adsorbant surface and to determine the effect of water structure upon the forces exerted on the protein. The spatial and orientation distribution functions (SDF and ODF) of the water were the focus of the examination. There are some previous studies of the SDF and ODF of water around a protein using molecular dynamic (MD) simulations [5– 7]. The data from the research carried out by Komeiji et al. suggested that the environment of the protein atoms influenced the structure of the surrounding solvent [6]. It should be noted that although water structure at solid interfaces and around a protein has been studied independently, little has been done regarding water structure around a protein located near a surface. Our previous research in this area studied water structure around enkephalin near a hydrophobic and organic polyethylene (PE) surface [8, 9], while this report will focus on water structure around enkephalin near a hydrophilic metal oxide surface.
SIMULATION PROCEDURE AND PARAMETERS
This research used MD simulations to determine the influence of solvent structure by calculating the spatial and orientation distribution functions of water in a system comprising a germanium oxide (GeO2 ) surface, water and leucine enkephalin [10]. These distribution functions revealed the average structure of the water at the interface between the protein molecule and the GeO2 surface. Then this average water structure was compared to the calculated intermolecular forces to seek correlations between them. Many of the same MD simulation procedures developed previously were used in this research, and the details are reported elsewhere [9, 10]. A constant number, volume, and temperature ensemble was used in the MD simulation with a consistent valence force field potential function. The temperature of the system was 298 K, and the simulation ran with time steps of 1.5 fs and a cut-off distance of 8 Å. The molecular system consisted of a protein, a solid surface, and water at a density of 1.00 g/ cm3 . The protein leucine enkephalin (YGGFL) was selected because it is a relatively simple molecule and its equilibrium configuration in water presents all of the oxygens on one side and the non-polar side groups on the other side [9], thus easily accommodating distinct protein orientations of protein interaction with the surface. The surface was constructed of a 3D array of germanium dioxide molecules placed in the crystal structure of GeO2 [11, 12]. Unfortunately, the consistent valence force field used did not contain the non-bond force constants for germanium and for oxygen bound to germanium. Because of the similar electronegativities of the silicon and germanium, the silicon-oxygen force constants were used for the atoms in the germanium oxide crystal structure [13]. The open face of the GeO2 was in the x, y plane of the coordinate system, and the z axis extended towards the protein. The x, y, and z periodic boundary conditions for this system were 30.7825 Å, 30.7583 Å, and 50.0000 Å, respectively. When
Water structure around enkephalin near a GeO2 surface
443
Figure 1. Molecular system consisting of a GeO2 surface, the protein in Rotation 1, and the solvent. The lightest shaded atoms in the protein are H, the darkest gray atoms are O, and the moderately shaded atoms are C. A few N atoms are shown in black. The darkest gray atoms in the surface are O and the lightest shaded atoms in the surface are Ge.
the protein was oriented with its oxygens toward the solid surface, the orientation was defined as Rotation 1; when it was oriented with its nonpolar side towards the solid surface it was defined as Rotation 2. A model of the molecular system with the protein in Rotation 1 is shown in Fig. 1. MD simulations were also done on systems of protein-water and water only, which served as controls or comparison systems, and are termed ‘basal’ systems. Initially the system was relaxed during an energy minimization for 450 fs (300 time steps), equilibrated for another 1.5 ps (1000 time steps), and was allowed to run for a final 12 000 time steps which simulated a real time of 18 ps. During the simulation, the protein’s five α carbons and the GeO2 surface were held immobile, but the water surrounding the protein and the remainder of the protein’s atoms were dynamic. Locking the α carbons in place precluded any gross translations or rotations of the protein.
444
A. M. Bujnowski and W. G. Pitt
During the simulation, the spatial and orientation distribution functions (SDF and ODF) were calculated for the surface-water and protein-water interactions. The GeO2 -water interactions were calculated in the volume extending between the surface and a cut-off distance of 15 Å. The protein-water calculations were performed for the volume between the protein and the solid surface, and the volume between the protein and the far end of the water box. The net force on the protein from the GeO2 and the water was also calculated. The simulations were repeated six times (with different initial solvent structures) with the protein at 21 distances from the surface. The resulting spatial and orientation distribution functions from the six runs at each distance were averaged and the standard deviation (SD) recorded. Two angles were calculated to accurately give the angular orientation of each water molecule with respect to the GeO2 surface or the plane of the protein. The angles phi (φ) and theta (θ) were used to define the relationships between the surface and the water dipole as shown in Fig. 2. The dipole vector of water originated at the oxygen atom and bisected the angle between the two hydrogen atoms of the water molecule. The φ angle is the component of the direction of the dipole vector pointing away from (0◦ ) or toward (180◦ ) the GeO2 surface. The θ angle is the component of the vector parallel to the GeO2 surface and ranges from 0◦ to 360◦ . The distribution cut-off distance, the distance beyond which no distribution calculations were performed, was 15 Å and is different from that used in the MD code for force calculations, which was 8 Å. The distribution cut-off distance was divided into 200 bins of equal spacing. After every 10 time steps of the simulation, the positions of the water molecules were analyzed, and for each of the bins, the number of molecules and the water dipole φ and θ angles with respect to GeO2 or the protein plane were recorded. After completing each run, the averages and standard deviations for the number density and for φ and θ were calculated. The spatial, orientation, and standard deviation distributions of water relative to the protein were calculated with respect to a plane containing the polar side of the protein. This plane was parallel to the x, y plane of the coordinate system and located at the z coordinate of the protein oxygen atom closest to the surface. To distinguish between random water and structured water, a standard deviation was calculated for the angular data in each bin. If water within a bin were randomly
Figure 2. Relationship between water dipole vector and surface.
Water structure around enkephalin near a GeO2 surface
445
oriented, with all angle values equally probable over the respective ranges of the two angles, the average angles would be, respectively, 90◦ and 180◦ for φ and θ. However, it is also possible for ‘structured’ water to have average angles of φ = 90◦ and θ = 180◦ , though in such a case all angles would not be equally probable. By calculating the standard deviation it is possible to distinguish between ‘structured’ water or ‘random’ water having angles of φ = 90◦ and θ = 180◦ . Furthermore, random water has specific standard deviation values of 39.8◦ for φ and 104.1◦ for θ. If the standard deviation were statistically different than these numbers, the water in that bin was considered to be non-random or ‘structured’ water. Standard deviations lower than these values would suggest that water is structured in a monomodal distribution, whereas standard deviations greater than those specified above indicate a multimodal distribution of the water structure.
RESULTS
The results for the GeO2 surface alone and the protein alone (termed ‘basal’ cases) will be presented, followed by the results of the complete molecular system. Basal results for the spatial distribution function The SDF for water near the GeO2 surface (Fig. 3) shows a first coordination shell of water at 3.1 Å and that a second, smaller shell centered at 6.1 Å. The number density value at long distances approaches 1, indicating the absence of higher density away from the surface. The SDF for water on the polar and nonpolar sides of the peptide are shown in Fig. 4. The SDF increases slightly above 1 at approximately 2.5 Å from the protein then approaches 1 at longer distances on both sides. Close to the polar side of the protein is a distinct region of low density (Fig. 4A). The protein topography is flat in this region, with the protein having its backbone oxygens exposed to the solvent. Outward from the opposite nonpolar side, the low density region is less prominent.
Figure 3. Surface-water spatial distribution function for GeO2 without protein. The edge of the GeO2 surface is at 0 Å.
446
A. M. Bujnowski and W. G. Pitt
Figure 4. Distribution outward from the polar side (A) and the non-polar side (B) of the protein (0 Å is the plane of the oxygen atom in the protein as defined in the text).
For the polar side it is hypothesized that few water molecules have center positions immediately adjacent to this flat side of the protein, and thus there is a drop in water distribution density adjacent to the protein. Adjacent to the nonpolar side, the decrease in density is less prominent, perhaps due to the irregular and dynamic topography of the hydrophobic side of the protein. It is hypothesized that because the large side groups move freely in the region near the protein, this movement may allow water molecules to penetrate close to the plane from which the protein-water SDF is calculated. Basal results for the ODF Figure 5 shows that there is a very distinct structuring of φ near the GeO2 surface. The value for φ is approximately 100◦ at approximately 2.3 Å from the surface. The structure extends to 5 Å from the surface. Unlike the hydrophobic PE surface studied previously [9], structuring of θ is observed near the surface of GeO2 : the value approaches a value of 100◦ with a standard deviation near 40◦ . Evidence of further structuring at 7.0 Å is demonstrated by the value for θ approaching 240◦ and by its standard deviation approaching 80◦ . At longer distances, the standard deviation approaches the basal value of 104.1◦ . However, the values for θ exhibit cyclical behavior with diminishing amplitude as
Water structure around enkephalin near a GeO2 surface
447
Figure 5. Surface-water orientation distribution functions for φ and θ angles as a function of distance for GeO2 without a protein. (A) Mean φ ODF (F) and φ standard deviation distribution function (2). (B) Mean θ ODF (F) and θ standard deviation distribution function (2).
one moves away from the surface. The values oscillate around an average value of 180◦ , suggesting that this is the true value approached by θ, an outcome consistent with the random behavior expected at long distances from the surface. The protein-water ODF and standard deviation functions revealed no evidence of structuring for either φ or θ on either side of the protein at any distance (data not shown) [10]. This may be due to the ability of the protein’s side groups to dynamically adjust to any forces the water molecules exert upon them. Calculations performed with nonmobile water at fixed angles Several calculations were performed to analyze the effect of φ on the force on the protein and the solid surface. The θ angle was not tested because preliminary studies indicated that little change in force would be observed for any changes in θ. The calculations were carried out by first making a ten-by-ten lattice of water molecules in an x, y plane. All of these molecules were fixed at some specified angle, and the lattice was then placed near the component of interest: the protein or the GeO2 surface. The three different φ values examined were 135◦ (with the hydrogen atoms towards the component), 90◦ , and 45◦ (the hydrogen atoms away from the specific component). The force from this lattice of water molecules was calculated, and the
448
A. M. Bujnowski and W. G. Pitt
Figure 6. Force (as a function of distance) between a lattice of fixed water and (A) the GeO2 surface, (B) the protein with the polar side toward the water, and (C) the protein with the non-polar side toward the water. In all figures, (F) represents data from water dipoles at 45◦ to the surface, (Q) represents dipoles at 90◦ , and (2) represents dipoles at 135◦ .
lattice was moved further away in incremental steps, creating a correlation between force and distance. The resulting graphs for the GeO2 surface and both sides of the protein are shown in Fig. 6. These plots show that the angle of the water and the orientation of the protein have an effect on the force. In all cases the force calculated when φ of the water is 90◦ always falls between the other two force curves, indicating a trend in the force as φ is changed.
Water structure around enkephalin near a GeO2 surface
449
The force of water on the GeO2 follows a characteristic force curve for dispersion forces: repulsive near the surface, decreasing to an attractive well, and ending with the force curves approaching zero. The force on the GeO2 surface is most attractive when the hydrogen atoms are pointing toward the surface (φ = 135◦ ) and least attractive when the hydrogen atoms point away from the surface (φ = 45◦ ). Therefore, these results show that as φ increases the GeO2 -water interactions force becomes less repulsive. For the polar side of the protein interacting with water (Fig. 6B), the force curves rise above zero after the attractive force well, an indication that significant electrostatic forces exist. The protein force curves follow the trend of greater repulsive forces when the hydrogen atoms point away from the protein (φ = 45◦ ) and less repulsive forces when the hydrogen atoms are towards the protein (φ = 135◦ ). This indicates that as φ increases, the force becomes more repulsive, a result opposite to that obtained for the GeO2 surface. For the nonpolar side of the protein (Fig. 6C), the force curves lie very close together, suggesting that there is a minimal effect of water angle on the force curves on this side of the protein. Analysis of SDF in the complete system The previous data establish that water is structured in the interfacial volume between the protein and the surface and that the effect of structured solvent on either the surface or the protein is unique. Next, the effect of moving the protein towards the surface can be analyzed. Fig. 7A shows the change in the number density in the volume between the surface and protein for five representative distances of protein placement in Rotation 1. At 10.63 Å, 8.63 Å, and 7.03 Å from the surface, two distinct peaks of high density are observed. The very high peak near the surface is caused by the hydrophilic nature of the surface: the water molecules appear to gather in a very high density near the surface due to the forces they experience in this region. This high density peak located at approximately 3.1 Å from the surface is independent of protein distance and is at a distance equal to the basal value shown in Fig. 3. Because the polar and hydrophilic side of the protein faces the surface, water molecules may be attracted near the protein, leaving the low density region seen between the two peaks. Very near the protein, another low density region is observed suggesting that water molecules do not overlap atoms on the hydrophilic side of the protein. The two peaks merge together as the protein is moved closer to the surface. A change in the forces on the protein due to the water was observed as the two peaks merged, suggesting that this water structure does directly affect the protein in some fashion. The number density for Rotation 2 is shown in Fig. 7B for five representative distances of the protein. The distribution of densities for Rotation 2 were much flatter near the protein than for Rotation 1, although a large peak near the solid surface was still present due to surface hydrophilicity. Like Rotation 1, this peak
450
A. M. Bujnowski and W. G. Pitt
Figure 7. Spatial distribution functions of water molecules between the GeO2 surface and the protein with the polar side towards the surface (A) and the nonpolar side towards the surface (B). The surface is at 0 Å, and the closest protein O atom is at the distance indicated in the legend.
is located at approximately 3 Å, independent of protein distance, showing the dominance of GeO2 -water interactions in this region. Analysis of the ODF in complete system Analysis of the orientational structure of water for the GeO2 surface revealed that water structure exists in the volume between the protein and the surface. Figure 8 shows angular and standard deviation (SD) distributions for the GeO2 system with the protein in Rotation 1 for three distinct distances of the protein from the surface. The figures show that the SD of φ is about 20◦ near the surface and
Water structure around enkephalin near a GeO2 surface
451
Figure 8. Angular and standard deviation distributions for water in the GeO2 system with the plane of the protein at (A) 4.63 Å, (B) 7.03 Å, and (C) 11.03 Å from the surface for Rotation 1. The edge of the surface is at 0 Å. The filled symbols represent the average values of θ (Q) or φ (F). The lines represent the standard deviations of θ and φ. The horizontal lines indicate the average values expected for random values of θ, φ and their standard deviations.
452
A. M. Bujnowski and W. G. Pitt
that it approaches the value of random water of 39.8◦ at long distances from the surface. The lower SD for φ near the surface indicates a tighter distribution of angles, thereby demonstrating that structure exists near the surface for Rotation 1. This distribution is a bit tighter than that observed for a polyethylene system [9], suggesting that the GeO2 surface exerts a greater effect on the adjacent solvent molecules. Another confirmation of this more pronounced surface influence is that for most all of the distances of the protein from the GeO2 , the φ structure is the same as the basal structure with φ values of 100◦ near the surface. The θ angle approaches a value of 140◦ with a SD which approaches 40◦ near the surface for nearly all distances of the protein from the surface. This indicates a high degree of θ structure in this region. Since θ near the surface shows only slight dependence on protein distance, the interactions appear to be dominated by structured water adjacent to the GeO2 . Additionally, the behavior of θ is oscillatory in nature away from the surface. At 6.0 Å, a large peak in θ and in its standard deviation with values of 200◦ and 120◦ , respectively, is observed for the larger separation distance of the protein. This large standard deviation indicates structure that approximates a multimodal distribution of θ in this area. For the smaller separation distances, the values approach 230◦ and 90◦ for θ and its standard deviation, respectively, indicating structure with a monomodal distribution. Other peaks with lower amplitudes are observed at longer distances from the solid surface. It can be seen that these peaks cycle about a value of 180◦ with a standard deviation of approximately 103◦ . There is some change observed in the water structure in the region immediately adjacent to the protein, in contrast to the results found with the polyethylene surface [9]. This may be attributed to the effect of the hydrophilic GeO2 surface on the surrounding water structure and the ability of the protein to perturb those effects. It should be noted that the structure near the solid surface for φ does not extend beyond 6.0 Å into the volume between the protein and the surface when the protein is at long distances from the surface. This is attributed to sufficient thermal energy in the system to allow the water molecules to overcome any structuring imposed by the surface. Therefore, any effect from the surface or the protein on the water molecules immediately near them is not propagated far through the volume. The results for Rotation 2 showed a similar structuring of the water for φ and θ near the surface, as shown in Fig. 9, for three distances of the protein. As in Rotation 1, φ and θ near the surface exhibit structuring as shown by their low standard deviations. The φ value adjacent to the GeO2 is similar to its basal value (100◦ ) for nearly all distances of the protein from the surface. The θ angle, however, approaches a value much lower than the value shown in Fig. 5B; it approaches approximately 80◦ near the surface for nearly all distances of the protein from the surface. This is even much lower than the 140◦ seen for Rotation 1. Also, the fact that the value of these angles did not change with protein movement shows that near the surface, the surface-water interactions dominate over protein-water interactions. The φ distribution did not change in the volume between the surface and the protein
Water structure around enkephalin near a GeO2 surface
453
Figure 9. Angular and standard deviation distributions for water in the GeO2 system with the protein at (A) 7.65 Å, (B) 10.05 Å, and (C) 14.05 Å from the surface for Rotation 2. The edge of the surface is at 0 Å. The filled symbols represent the average values of θ (Q) or φ (F). The lines represent the standard deviations of θ and φ. The horizontal lines indicate the average values expected for random values of θ, φ and their standard deviations.
454
A. M. Bujnowski and W. G. Pitt
beyond 6 Å due to the thermal energy of the system. However, the same oscillatory behavior in θ observed in Rotation 1 was also present in Rotation 2. Although the amplitudes were much more damped, a distinct peak in the θ ODF with a value of 210◦ and a standard deviation of 65◦ was still observed at 6 Å. It should be noted that this indicates a monomodal structuring of θ at this distance, a result opposite to that found for Rotation 1. While φ and its standard deviation approached its random values at long distances from the surface, θ continued its cyclic behavior. As for Rotation 1, the θ standard deviation for Rotation 2 approached the random value of 104◦ and the θ distribution oscillated around 180◦ , suggesting that this distribution was random at long distances from the solid surface. Finally, no change in structure is observed in the region adjacent to the protein for any protein distance, a result consistent with the findings for water adjacent to a polyethylene surface [9]. Force plots Graphs of the force on the protein versus distance were prepared for the GeO2 system. These graphs are shown in Fig. 10 for the protein in Rotation 1 (backbone oxygens towards the surface) and in Rotation 2 (backbone oxygens away from the surface). The solid lines on these plots represent the average force for interactions of the protein-surface, the protein-water, and the their sum. All six replicates for the protein-surface and protein-water interactions are also shown on the graphs. There is much more scatter in the protein-water data than in the protein-surface data. Both figures show that the surface-protein interaction follows a standard dispersion force potential: repulsive at close distances with an attractive region which is followed by a region that approaches zero force. It can be seen that for the protein with its oxygens towards from the surface in Fig. 10A, the water-protein interaction forces are generally repulsive, resulting in the total protein-water force curve being generally repulsive in nature. For the protein with its oxygens away from the surface (Fig. 10B), the proteinwater forces is primarily attractive, resulting in an attractive trend in the total force. These results are similar to those seen for hydrophobic polyethylene surface [9], particularly for the protein in Rotation 1. A difference observed between results for the GeO2 surface versus PE is that the repulsive potential for the GeO2 system extends farther away from the surface than for the PE system [9]. Again, based on the difference in the protein-water force fields for the two orientations of the protein, it is evident that the key to understanding the protein behavior at the interface lies with an analysis of the protein-water interactions. Analysis of H2 O structure versus force Having established the presence of structured water and having shown that changes in this structure occur as the protein is moved towards the surface, an analysis of the force distribution and the angular distributions were made to determine if any relationship existed between them. The angles near the surface, near the protein,
Water structure around enkephalin near a GeO2 surface
455
Figure 10. Force versus distance plot for Rotation 1 (A) and Rotation 2 (B). The surface is at 0 Å. The (P) represents the average force between the water and protein for six different simulations, and the (!) represents the average force between the water and protein. The solid symbols (Q and ") represent the average of the six simulations, and the (2) represents the sum of the surface and water averages.
and at the distance of peak number density for each distance and each orientation of the protein were correlated with the force distribution. The only significant correlation obtained in this analysis was a negative linear trend observed when the force versus φ was plotted for water molecules located at the peak of the density distribution when only one peak was observed and when the protein was in Rotation 1 (see Fig. 11), a result similar to that of the PE surface [9]. The correlation produced a slope and an intercept of −1.5 ± 1.2 and 130.2 ± 98.2, respectively, and a correlation coefficient of R 2 = 0.54.
456
A. M. Bujnowski and W. G. Pitt
Figure 11. Total protein-water force versus φ for distance 2.5 Å– 5.7 Å of the protein in Rotation 1. Only one peak in water density was observed at these distances.
This result is consistent with the trend shown earlier for the GeO2 surface: the force becomes more repulsive as φ decreases or as hydrogen atoms point away from the GeO2 surface (Fig. 6A). In addition, recall that these ‘fixed angle’ studies showed that for the protein, the force becomes more repulsive as the hydrogen atoms point away from the protein. It was also shown earlier that for the protein, the force becomes more repulsive as the angle increases or as the hydrogen atoms point away from the protein and towards the GeO2 , a positive linear trend (Fig. 6B). Therefore, the GeO2 forces dominate the water structure as the protein is brought close to the solid surface, causing the solvent near the protein to be in a conformation which is not favored with protein-water interactions. This may be one contribution to the repulsive forces observed for the water-protein force in Rotation 1. This result is qualitatively the same for GeO2 and PE [9], indicating that both surfaces have similar effects on water structure. Previously it was shown that a change in the water structure near the solid surface was observed for Rotation 1 as the protein position was changed. It is clear from these results, however, that the positional change of the protein did not significantly perturb the solvent structure caused by the GeO2 -water interactions in the volume immediately adjacent to the GeO2 surface. No other correlation between the angles of the water and the force on the protein due to the water was found. This was not surprising for two reasons. First, when the protein is in Rotation 2, with its backbone oxygens away from the solid surface, the part of the protein facing the surface is very dynamic. The influence on the water solvent by the surface may not affect the protein even at close distances because the large side-groups on the protein can move and adjust accordingly. Second, it is believed that due to the thermal energy of the system, the water molecules are able to assume their most energetically favorable position in the volume between the protein and the surface. Although the water can be affected very near the surface, this effect is not propagated throughout the volume when the protein and the surface are separated by longer distances.
Water structure around enkephalin near a GeO2 surface
457
DISCUSSION
The objectives of this research were threefold: to calculate the spatial and orientation distributions of water, to make force versus distance plots for both protein orientations, and to compare the distribution functions and the force functions in an effort to correlate the water structure with the protein-water forces. This section discusses the molecular interactions giving rise to the angular distribution and force functions. In each section, statistics based on a 95% confidence interval of the six repetitions of each protein distance will be presented. The phi angle distribution The basal tests showed structuring of φ near the GeO2 surface with the φ value approaching 99.5 ± 0.3◦ and the standard deviation approaching 20.7 ± 0.3◦ . This distribution was very tight due to the strong interactions of the GeO2 surface with the solvent. This same structuring of φ near the surface was seen for all distances of the protein and both orientations during the experimental simulations. The φ distribution adjacent to GeO2 did not change significantly as the protein was brought near the surface, again demonstrating the strong interactions of the GeO2 surface with the surrounding water. The φ value indicates that the hydrogen atoms of the waters near the solid surface were pointing preferentially toward the surface. This result was consistent with that found when a lattice of frozen water at fixed angles was placed near the surface: the φ value which resulted in the lowest repulsion force occurred when the hydrogen atoms pointed towards the surface. The φ structure may have been caused by repulsive interactions between the surface oxygen and water oxygen atoms. The oxygen atom has the largest non-bond interaction constants and, therefore, experiences the largest repulsive force. Therefore, in the dynamic system, the most energetically favorable position for the water molecules is one in which the hydrogen atoms would be pointing toward the oxide surface. This orientation is also consistent with hydrogen bonding between the water and the surface oxygens. In addition to orientation of the water near the surfaces, the water surrounding the protein is somewhat oriented by the protein when the protein is at distances longer than 4.2 Å from the surface. Deviations from randomness were noted in the φ distribution for those water molecules in the volume immediately surrounding the protein (see Figs 8 and 9). For basal conditions (no GeO2 present) little or no structuring was observed near the protein for the basal conditions. However, the presence of the GeO2 may have structured the surrounding solvent molecules sufficiently enough to allow the protein to disturb this structure. The theta angle distribution The basal tests of water near the germanium oxide surface revealed that the θ distribution was highly structured. Very near the surface, the θ value approached 98.2 ± 0.7◦ while its standard deviation approached 40.9 ± 5.3◦ , and this θ value oscillated with distance from the surface.
458
A. M. Bujnowski and W. G. Pitt
It is postulated that the oscillatory behavior was caused by the charge distribution on the GeO2 surface. This charge lattice is very regular, and the solvent atoms may respond to the dipole moments of the atoms on the surface by orienting themselves in the θ direction. The interaction of the charged GeO2 with the solvent molecules very near the surface was strong enough to cause them to conform to a non-random θ structure. The subsequent layers of water molecules then responded to the surface and intervening water by orienting themselves in an average θ-direction which minimized the energy of the dynamic system. The same oscillatory behavior in θ observed for the basal simulations of the GeO2 surface was also seen in simulations including the protein. However, there were three obvious differences. First, the θ values near the surface approached different values for the cases with or without (basal case) protein. The value of the θ distribution approached 141.0 ± 2.2◦ with a standard deviation of approximately 37.8 ± 4.3◦ near the surface for all distances of the protein when the protein was in Rotation 1. While the value for θ was higher than the basal value of 98.2 ± 0.7◦ , the standard deviation value differed very little from the basal case. For Rotation 2, the θ value near the surface approached 80.7 ± 2.8◦ while its standard deviation leveled off at 82.0 ± 2.6◦ . It is unclear why the two orientations would cause the average θ value near the surface to change. It is believed that the mobility of the side groups was the cause of the higher standard deviation for Rotation 2. The second difference between the basal and experimental cases was that the first peak in the θ distribution for each orientation had a very different structure than the basal case. The basal simulations produced values of 243.3±0.2◦ and 80.7±5.8◦ for θ and its standard deviation in the first peak. Whereas with protein in Rotation 1, the average θ value and the standard deviation at the first peak in the θ distribution had values of 210.8 ± 2.5◦ and 106.7 ± 13.8◦ , respectively, when the protein was farther than 5 Å from the surface. The θ angle and its standard deviation had values of 211.0 ± 5.3◦ and 96.3 ± 6.3◦ , respectively, when the protein was closer than 5 Å. The standard deviations indicate that the peak in Rotation 1 had a multimodal structure for protein distances greater than 5 Å. At distances shorter than this value, the first peak in the θ distribution was located on the side opposite of the hydrophilic backbone oxygens. However, at 11.03 Å, a very definite multimodal structuring was observed in the first θ distribution peak (Fig. 8C). This multimodal distribution at 11.03 Å in Rotation 1 could have been caused by the hydrophilic interactions of the backbone oxygens with the water molecules. For the experimental run with the protein in Rotation 2, θ and its standard deviation at the first peak had values of 216.0 ± 6.9◦ and 87.1 ± 17.0◦ , respectively, indicating a monomodal distribution of angles. No remarkable differences were observed in the θ standard deviation at any protein distance. The large side groups of the protein in this orientation may have been sufficiently mobile to allow the water molecules to assume the same structure independent of protein position. It can be seen that the structure in this peak was not the same as the basal case, indicating that the protein caused some disruption of water structure.
Water structure around enkephalin near a GeO2 surface
459
The final difference between the experimental and basal tests was that the oscillating behavior in the θ distribution in simulations with protein dampened out much more quickly than the basal case. Again, the interaction of the solvent with the protein may have disrupted the alternating θ layers, causing them to more quickly disorganize. The force correlation The basal tests using fixed water indicated that for the surface and the protein in Rotation 1, the most energetically favorable position for the water molecules was with the hydrogen atoms pointing towards the GeO2 surface or toward the protein. Because the force on the water molecules from the surface is a strong function of distance, those atoms or molecules nearest to the surface experience the largest force. Because oxygen atom has the largest non-bond repulsive force constants, it is more energetically favorable to have this water oxygen atom farther away from the GeO2 surface. This orientation is also consistent with hydrogen bonding between the water and surface. For the protein in Rotation 2, there may have been less preference for φ angle due to the irregular topography of the protein. It should be noted that due to the relative positions of the surface and the protein, water molecules with their hydrogen atoms pointing towards the protein have φ values less than 90◦ and water molecules with hydrogen atoms pointing towards the surface have φ values greater than 90◦ . This signifies that for the surface, a larger φ means a less repulsive force, and for the protein, a smaller φ indicates a less repulsive force. During the simulations with the GeO2 surface and the protein, the force was mapped out for the protein-water and protein-surface force interactions. For both orientations of the protein, the surface-protein interaction force curve followed the behavior of a standard force curve. While the results of the protein-surface interactions were not surprising, the results for the protein-water interactions were unexpected. In general, the force for the protein in Rotation 1 was mainly repulsive, while that for Rotation 2 was mainly attractive. Calculations performed to correlate these forces with water structure revealed that the only significant correlation was for the φ structure in the region when only one high density peak was observed, and the force decreased with increasing φ angle. It is believed that although the protein may perturb the water structure near the GeO2 surface to some extent, the surface-water interactions still dominate. Therefore, the water molecules are in a conformation which is energetically favorable in terms of the GeO2 -water interactions. As was explained earlier, the energetically favorable positions of water molecules for the surface are not favorable for the protein in Rotation 1. Therefore, when the water is in a position favorable to the surface, it is unfavorable to the protein. This may have been the cause for the increased repulsiveness of the force for the protein in Rotation 1 when the protein was closer to the surface. This correlation still does not explain, however, the repulsive forces observed when the protein is farther away. The small influence on the solvent farther away from the surface
460
A. M. Bujnowski and W. G. Pitt
may have been sufficient for the protein in Rotation 1 to be unfavorably affected. However, this conclusion can not be confirmed by this work. Rotation 2 did not have the same behavior with the surface-water interactions possibly due to the ability of the large side groups to move and adjust to system behavior. The observation that the water is responsible for the overall repulsive force interaction when the polar side of the peptide is toward the polar surface may be explained using the statistical thermodynamic model of Besseling for orientationdependent molecules such as water at interfaces [14]. In his model, surfaces with an unequal density of hydrogen bonding acceptor atoms (like oxygen) and donor atoms (like hydroxyl or amine hydrogen) create orientation in the structure of water adjacent to the surfaces. If two such surfaces approach each other, the orientation or the water is in opposite directions, which creates a hydrogen bond donor-acceptor mismatch in the volume between the surfaces, and produces a ‘repulsive hydration force’. In our model, the GeO2 surface has only hydrogen bonding acceptor atoms (oxygen) and no hydrogen atoms, and the orientation distribution function shows that the water immediately adjacent to the surface has the hydrogens pointing to the GeO2 (φ > 90◦ ), and the standard deviation of φ indicates structure in the water for about 5 Å (Fig. 5A). Less structure is evident near the polar side of the protein, but we speculate that even a small amount of structure could produce a donor-acceptor mismatch. In addition, the simulations with non-dynamic water at fixed angles showed that for both the GeO2 surface and the polar side of the protein, the interaction force with water was most attractive when the hydrogens were toward each respective surface. Bringing two such surfaces together, each with hydrogen dipoles toward each surface, would cause a donor-acceptor mismatch in the water between these two surfaces, and the repulsive hydration force.
SUMMARY
This study of MD simulations of water and a protein near a GeO2 surface has shown that the hydrophilic GeO2 interacts with the water molecules in such a way as to cause the water molecules to gather in a very high density near the surface. This high density peak located at approximately 3.1 Å from the surface is independent of the distance of the protein. The water adjacent to the GeO2 is structured in the φ direction with the H atoms pointing slightly (on the average) toward the surface. When the enkephalin protein approaches the surface with the polar side toward the GeO2 , the water exerts a repulsive force on the protein; i.e., the protein experiences a push away from the surface. Near the surface this repulsive force correlates with the average φ angle of the water at the peak of maximum water density. These water molecules are in an orientation that causes repulsive forces between the water and protein. It is postulated that this unfavorable orientation causes the repulsive force between the water and protein.
Water structure around enkephalin near a GeO2 surface
461
The approach of the protein toward the GeO2 also disrupts the structure of water adjacent to the GeO2 . When the protein approaches the GeO2 with its nonpolar side toward the surface, no repulsive forces are observed, and the structure of water adjacent to the GeO2 is minimally perturbed, if at all. The observed behavior was similar for both the hydrophilic GeO2 surface and a previously studied hydrophobic PE surface [9]. However, there are differences in the correlations for each system which show that while the φ values near the surface may be close, the overall structure near the surface is different. These results do not explain the similarity of the protein-water force for both the PE and the GeO2 system. The fact that they are similar may have been only a fortuitous artifact of the scatter in the protein-water force data. Consequently, further testing may be needed to test this hypothesis. Acknowledgements This research was supported by a grant from the Chemical Engineering Department of Brigham Young University.
REFERENCES 1. G. A. Skarja and J. L. Brash, Physicochemical properties and platelet interactions of segmented polyurethanes containing sulfonate groups in the hard segment, J. Biomed. Mater. Res. 34, 439– 455 (1997). 2. H. J. Busscher, G. I. Geertsema-Doornbusch and H. C. van der Mei, On mechanisms of oral microbial adhesion, J. Appl. Bacteriol. 74, 136S – 142S (1993). 3. J. L. Brash and P. Ten Hove, Protein adsorption studies on ‘standard’ polymeric materials, J. Biomater. Sci. Polymer Ed. 4, 591– 599 (1993). 4. W. G. McClung, D. L. Clapper, S. P. Hu and J. L. Brash, Adsorption of plasminogen from human plasma to lysine-containing surfaces, J. Biomed. Mater. Res. 49, 409– 414 (2000). 5. C. L. Brooks and M. Karplus, Solvent effects on protein motion and protein effects on solvent motion — dynamics of the active site region of lysozyme, J. Molec. Biol. 208, 159– 181 (1989). 6. Y. Komeiji, M. Uebayasi, J.-I. Someya and I. Yamato, A molecular dynamics study of solvent behavior around a protein, Proteins: Sturcture, Function, and Genetics 16, 268– 277 (1993). 7. M. Levitt and R. Sharon, Accurate simulation of protein dynamics in solution, Proc. Nat. Acad. Sci. 85, 7557– 7561 (1998). 8. W. G. Pitt and D. R. Weaver, Calculation of Protein-Polymer Force Fields using Molecular Dynamics, J. Colloid Interface Sci. 185, 258– 264 (1997). 9. A. M. Bujnowski and W. G. Pitt, Water structure around enkephalin near a PE surface: a molecular dynamics study, J. Colloid Interface Sci. 203, 47– 58 (1998). 10. A. M. Bujnowski, Spatial and orientation distribution functions of water in protein-solid systems: a molecular dynamics study, Master Thesis, Brigham Young University (1996). 11. G. S. Smith and P. B. Isaacs, The crystal structure of quartz-like GeO2 , Acta Crystallographica 17, 842– 846 (1964). 12. W. H. Baur and A. A. Khan, Rutile-type compounds. IV. SiO2 , GeO2 , and a comparison with other rutile-type structures, Acta Crystallographica B27, 2133– 2139 (1971).
462
A. M. Bujnowski and W. G. Pitt
13. C. Lang, Telephone interview, Biosym Technical Support (1996). 14. N. A. M. Besseling, Statistical thermodynamics of fluids with orientation-dependent properties and its application to water at interfaces, in: Water in Biomaterials Surface Science, M. Morra (Ed.) pp. 91– 125. Wiley, Chichester (2001).
Towards practical soft X-ray spectromicroscopy of biomaterials A. P. HITCHCOCK ∗ , C. MORIN, Y. M. HENG † , R. M. CORNELIUS and J. L. BRASH Brockhouse Institute for Materials Research, McMaster University, Hamilton, ON, Canada, L8S 4MI Received 21 December 2001; accepted 11 April 2002 Abstract—Scanning transmission X-ray microscopy (STXM) is being developed as a new tool to study the surface chemical morphology and biointeractions of candidate biomaterials with emphasis on blood compatible polymers. STXM is a synchrotron based technique which provides quantitative chemical mapping at a spatial resolution of 50 nm. Chemical speciation is provided by the near edge X-ray absorption spectral (NEXAFS) signal. We show that STXM can detect proteins on soft X-ray transparent polymer thin films with monolayer sensitivity. Of great significance is the fact that measurements can be made in situ, i.e. in the presence of an overlayer of the protein solution. The strengths, limitations and future potential of STXM for studies of biomaterials are discussed.
1. INTRODUCTION
The adverse effects of implanted biomaterials (regardless of intended anatomic location) begin with the selective interactions of blood proteins with the surface of the biomaterial [1, 2], typically a polymer. In the work reported here, soft X-ray spectromicroscopy is being developed to investigate a number of issues related to selectivity in the first contact of biological systems with polymers which are heterogeneous at the surface through patterning or intrinsic surface or bulk phase segregation. Demonstration of the ability to detect and map adsorbed protein at the monolayer level on surfaces with lateral chemical differentiation is our initial target. Eventually we seek to track site selectivity in the adsorption of specific proteins from mixtures, although this is likely to require labeling techniques since soft X-ray spectroscopy cannot readily distinguish different proteins. ∗ To
whom correspondence should be addressed. E-mail: [email protected] address: The Hospital for Sick Children, 555 University Avenue, Toronto, ON Canada M5G 1X8. † Present
464
A. P. Hitchcock et al.
Soft X-ray spectromicroscopy (also known as near edge X-ray absorption fine structure or NEXAFS microscopy) is finding increasing use in the analysis of soft materials, on account of its ability to probe chemical complexity quantitatively on spatial scales of 50 nm or better, as well as the versatility with which it can be adapted to a wide range of problems. Reviews of soft X-ray spectromicroscopy techniques, instrumentation, and a broad survey of results have been presented recently [3– 6]. Inner-shell excitation or NEXAFS spectroscopy [7] is used as the chemically sensitive image contrast mechanism. For the past few years we have been exploring two techniques of soft X-ray spectromicroscopy for the study of biomaterials and the interaction of biomaterial surfaces with biological subsystems, particularly proteins. Scanning Transmission X-ray Microscopy (STXM) uses a focused X-ray probe with sample scanning and synchronized detection of transmitted X-rays to measure the wavelength dependent optical density through a column of material. Although not intrinsically surface sensitive, it is possible to detect, and thus quantitatively map the distributions of surface species such as proteins if the NEXAFS spectrum of a surface species is sufficiently different from that of the bulk biomaterial constituents. STXM in the water window energy range (200– 520 eV) can be applied to samples in vacuum, in air or He at atmospheric pressure, and, of greatest importance to biomaterials, to wet samples enclosed in a cell equipped with X-ray transparent windows. In this paper we describe only STXM results. We are also engaged in a parallel effort for developing X-ray photoelectron emission microscopy (XPEEM) for biomaterial studies. XPEEM uses large area X-ray illumination (50 µm × 50 µm, at the facility we use) and an electron lens system to record images of the spatial distribution of electrons ejected from a surface following X-ray absorption. Because the electron yield is proportional to the X-ray absorption coefficient, chemical identification and mapping can be derived from the wavelength dependence of the image contrast. The electron lens systems used are most sensitive to the numerous low energy secondary electrons rather than the relatively few primary photoelectrons. Thus the sampling depth is of the order of 5– 10 nm, considerably larger than techniques based on energy resolved electron analysis such as X-ray photoelectron spectroscopy (XPS). In addition to chemical sensitivity, XPEEM signals are also very sensitive to topography, local work function variations, and charging, and thus deriving quantitative maps from XPEEM data is much more complicated than with STXM. Finally XPEEM requires a ultrahigh vacuum sample environment, which is not compatible with studies of biomaterials under wet conditions. 2. EXPERIMENTAL
2.1. X-ray microscopy instrumentation and techniques Soft X-ray scanning transmission X-ray microscopy (STXM), developed by Kirz, Jacobsen, Ade and co-workers at the National Synchrotron Light Source (NSLS)
Towards practical soft X-ray spectromicroscopy of biomaterials
465
[3, 8, 9], is now implemented at several other synchrotron facilities (Advanced Light Source (ALS), Pohang Light Source), and instruments are under construction elsewhere (BESSY, Swiss Light Source (SLS), Canadian Light Source). Panel (a) of Fig. 1 is a sketch of the STXM experiment at the undulator beamline 7.0 at the Advanced Light Source, where this study was performed. The undulator produces high brightness X-rays which are passed through a spherical grating monochromator to select a narrow photon energy range (typically ∼150 meV). The beam of monochromated soft X-rays is then focused to 50 nm or less by a Fresnel zone plate, then passed through an order sorting aperture to select only the first order diffraction of the zone plate. X-rays transmitted through the sample are detected by conversion to visible light by a fast phosphor, followed by photon detection in single photon counting mode. To obtain an image at a given photon energy, the sample is raster scanned through the focal point while recording the intensity of transmitted X-rays (Fig. 1b). Alternatively, the photon energy can be scanned while sitting at a fixed spot on the sample to acquire the NEXAFS spectra of features of interest (Fig. 1c). The most useful mode is to acquire the NEXAFS signal over a whole field of view, by recording image sequences or stacks [10] (Fig. 1d). Post acquisition analysis of image sequence data is used to correct for image misalignment, and to generate chemical maps, as discussed further below. Typical incident intensities in the 50 nm focused spot of existing STXMs range from 106 Hz (NSLS X1A, ALS 5.3.2) to 108 Hz (ALS 7.0.1) [11, 12]. High brightness third generation light sources are particularly useful to achieve high intensities on the sample and thus rapid scan rates. With sufficient flux, and suitable control and acquisition interfacing, rapid scan rates (currently, 0.2 to 1 ms per pixel at the ALS), and thus high efficiency analytical microscopy, can be achieved. Radiation damage is a concern, but the damage rate relative to the signal acquisition rate is much smaller than in electron microscopy [13]. In these studies we characterize damage rates, we check for extent of damage after key measurements, and we discard data acquired where excessive damage has occurred. STXM is used analytically by acquiring NEXAFS spectra at one location (point mode), along a line (line mode), or through collection of full image sequences (image mode). All of these modes were employed in these studies and are illustrated below. The transmitted X-ray intensity (I) is converted to absorbance (optical density) by using the Beer-Lambert law, A = − ln(I /I0) = µρt = σ t, where I0 is the incident flux, I is the transmitted flux, µ is the mass absorption coefficient, ρ is the density, t is the thickness and σ is the linear absorption coefficient. The incident flux (I0 ) is recorded independently with the sample removed (single beam mode of optical spectroscopy). The measured signal averages over a column of the sample and thus it is generally considered ‘bulk’-sensitive. However, thin samples (50– 200 nm of organic matter at a carbon density of ∼1 g/cm3 ) are needed to achieve adequate
466
A. P. Hitchcock et al.
transmission in the 250– 1250 eV soft X-ray range. When the sample is this thin, the surface region (outer ∼10 nm) contributes significantly to the transmission mode signal and thus large surface species such as adsorbed proteins can be detected, as illustrated below. In this study, the sample thickness is such that absorption saturation is avoided and there is a linear relationship between absorbance and the thickness-density product. 2.2. Data analysis methods Point spectra, linescan spectra, images, or image sequences are converted to quantitative chemical information (point, line or area compositional maps) by spectral fitting on a pixel-by-pixel basis using linear curve fitting procedures [14– 16]. These methodologies, along with many other image and spectral data processing procedures used in this work were accessed by the aXis2000 program [14]. In order to generate quantitative chemical information we fit the spatially resolved NEXAFS signals with reference spectra on a pixel-by-pixel basis. The reference spectra are recorded separately, generally from pure materials. They are placed on absolute linear absorption scales so that the fit coefficient for a given component at a given pixel is the thickness of that component at that position. The ratio of that thickness to the sum of thicknesses of all components fitted is then a measure of the local composition. The array of fitting coefficients for a given component, derived by fitting the individual pixel spectra to linear combinations of reference spectra, is a quantitative chemical component map. The fits can be carried out using singular value decomposition [15, 16], linear regression, or a conjugate gradient algorithm. The fit coefficients can be constrained to be positive or they can be treated as freely adjustable parameters. Each methodology provides maps of residuals and a statistical analysis of the errors. Comparison of the results from different algorithms helps build confidence in the significance of weak signals such as those associated with proteins on polymer surfaces. Where vertical scales are provided for the component maps in the figures shown below, these indicate the estimated thickness in nm of that component, if it was pure. Typically a given pixel can have contributions from 3 or more components. Estimated uncertainties are 10– 15%. Systematic errors are considerably larger than statistical errors (the latter are typically 1– 3% for the majority components). The uncertainty in the quantitation of the weak signal from surface adsorbed proteins is considerably higher — perhaps as much as 50%. In addition to grayscale maps of individual components, we present color maps where the intensity of the red, green or blue component gives the spatial distribution of that component over the region mapped. In all color maps presented here we have individually byte-scaled each component and thus the intensity of one component relative to the other is not properly represented. This approach provides clearer information of spatial localization of the components, since it makes the weak protein signals visible against the much stronger polymer components. With byte-rescaling one must be careful in interpreting intermediate colors, although they do indicate regions where multiple components are present in the column sampled.
Towards practical soft X-ray spectromicroscopy of biomaterials 467
Figure 1. (a) Schematic of undulator beamline 7.0 and the STXM endstation at the Advanced Light Source (ALS) BL 7.0. The inset shows details of the STXM optics. (b) Image of sample #355, a compressed polyurethane foam with styrene acrylonitrile (SAN) and poly-isocyanate poly-addition product ∗ transition at which the SAN particles selectively absorb. (PIPA) filler particles. This image was recorded at 287.2 eV, the energy of the C 1s → πCN (c) Spectra from spatially selected regions compared to a defocused sample average spectrum. These spectra were actually obtained from an image sequence, but equivalent point spectra can be acquired with the STXM. (d) Illustration of the image sequence (stack) concept. Four of 120 images are displayed, along with a color coded chemical constituent map derived by pixel-by-pixel curve fitting (R = PIPA, G = SAN, B = matrix).
468
A. P. Hitchcock et al.
2.3. Substrates and protein exposure methods 2.3.1. Materials. 2.3.1.1. Substrates. Three different substrates were used, all provided by Dow Chemical. All three consist of a compressed polyurethane foam, with a TDI (toluene diisocyanate) hard segment and a butane oxide (BO) soft segment. The three differ with regard to types of filler particles. One substrate (code #355) contains poly(styrenecoacrylonitrile (SAN) and poly-isocyanate poly-addition product (PIPA, a methylene diphenyl diisocyanate (MDI)-based hard segment-like material), both referred to as copolymer polyol (CPP) filler particles. The second substrate (code #530) contains only SAN particles. The third substrate (code #529) contains only PIPA particles. The synthesized foams were cryo microtomed to ∼100 nm thickness and multiple sections were placed either on TEM grids or on 4-pane Si3 N4 windows (2 × 2 array 1.25 × 1.25 mm membrane, 75 nm thick; frame size is 7.5 mm × 7.5 mm, 200 µm thick Si). The silicon nitride (Si3 N4 ) windows were obtained from Silson Ltd. [17] and were rigorously cleaned to semiconductor industry standards by the manufacturer. They were removed from plastic storage capsules and used without further surface preparation. 2.3.1.2. Deionized water. <0.0003%.
From Sigma, HPLC water, residue after evaporation
2.3.1.3. Protein. Human serum albumin (HSA) was from Behringwerke AG, Marburg, Germany, and is reported to be homogeneous as judged by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE). Fibrinogen (Fg) was from Calbiochem, San Diego, California, USA. It is prepared from human plasma and is plasminogen depleted. It is reported to be >95% clottable by thrombin, and is homogeneous as judged by SDS-PAGE. 2.3.1.4. Buffers. IL, USA.
The phosphate buffer saline packs were from Pierce, Rockford,
2.3.2. Sample preparation and mounting methods. Dry sections of the polyurethanes (#355, #529 or #530) deposited on a TEM grid were exposed for 20 min to 1 ml of 0.1 mg/ ml protein solution (albumin or fibrinogen) in deionized H2 O or phosphate buffer. For the sample in Fig. 3, the droplet of solution was allowed to dry out on the surface, thus depositing 0.1 mg over perhaps 0.5 mm2 . This corresponds to an average thickness of 200 nm (assuming ρ = 1 g/ cm3 ). For the sample in Fig. 4, a 5 µl drop of 0.1 mg/ml Fg in saline phosphate buffer was deposited over a #355 section (deposited on a Si3 N4 window). Such a drop covers an area greater than that of the section itself. At the end of the 20 minute exposure time, the drop had not evaporated. Three consecutive rinses of the remaining drop were done by sucking in and out the drop with 50 µl volumes of deionized water. The sample was then loaded into the STXM. For the sample in Fig. 5 a #355 section was exposed
Towards practical soft X-ray spectromicroscopy of biomaterials
469
to fibrinogen as above. It was then extensively washed to remove excess protein, dried and then rehydrated by adding a drop of water on top of the section. A second window was then placed on top of the first one, thereby creating a water layer over the section between the two windows. For the samples in Figs 6 and 7, a 5 µl drop of 0.01 mg/ ml Fg in freshly prepared phosphate buffer was deposited over a #355 section on a S3 N4 window. A second window was immediately placed over it and the sample was placed in the STXM about 10 min after fabrication. Due to instrumental difficulties, the sample was not measured until 2 h after fabrication.
3. RESULTS
3.1. STXM of CPP containing polyurethanes: substrate characterization The polyurethane substrates used in this study were supplied by Dow Chemical as cryo-microtomed thin sections placed on TEM grids or silicon nitride windows. They consist of a polyurethane matrix with a TDI hard segment and a butane oxide soft segment, in which one or more types of copolymer polyol (CPP) particles are embedded [18]. These CPP particles provide chemically differentiated domains at the surface with sizes in the 0.1 to 2 micron range, which is well suited to the spatial resolution of current X-ray microscopes. In contrast the ‘natural’ phase segregation in most polyurethanes is at a 10– 30 nm scale [18], which is too small for current X-ray microscopy capabilities. Both particle materials — SAN and PIPA — are aromatic and hydrophobic in character, while the matrix of the polyurethane is an aliphatic polyether which is more hydrophilic. In other studies we have shown that C 1s NEXAFS spectroscopy can readily distinguish the urea and urethane linkages present in polyurethanes; determine polyol content; and identify the types of R and R groups in a given polyurethane [19– 21]. NEXAFS signals in STXM have been used to map key functional groups quantitatively — urea, urethane and polyol in native polyurethanes [22], and the types of filler particles [6, 23]. STXM spectromicroscopy is being used to help understand how filler materials affect mechanical properties such as elastic modulus, tear strength and resiliency, and to aid the development of improved fillers [23]. Figure 1 shows the spectral signatures of the SAN, PIPA and polyurethane matrix, and their and spatial distributions in the (#355) substrate. The #530 SAN polyurethane substrate was similar except it did not have any of the small PIPA particles and the SAN particles were somewhat larger than in #355. The #529 polyurethane substrate has only PIPA particles and again, the particle size distribution includes larger PIPA particles than found in #355.
470
A. P. Hitchcock et al.
Figure 2. (top) Comparison of the C 1s NEXAFS spectra of albumin and fibrinogen recorded in the STXM. (lower) Illustration of monolayer sensitivity of STXM spectroscopy to pure albumin. The central panel displays an image (288.2 eV) of a deposit of pure albumin on a silicon nitride window (dry), and the signal from a linescan spectrum across the dotted line. The spectra shown in the lower panel were extracted from the linescan by adding signal over less than 1 micron in the areas indicated by dashed lines, labeled A and B. Spectrum B is offset vertically by 0.02 units. The OD of only 0.01 in the C 1s continuum is equivalent to ∼3 nm protein. An albumin molecule in its standard conformation is ∼3 × 8 nm, indicating monolayer sensitivity.
Towards practical soft X-ray spectromicroscopy of biomaterials
471
3.2. Protein-polymer interaction studies Eventually we intend to apply our technique to studies of protein decorated polymer samples while they are fully covered with a solution layer that has a composition of that of the relevant biological system, such as blood plasma. At present we are exploring STXM studies of both dry and wet samples to understand and optimize our data acquisition and analysis methods. Results to date indicate that while STXM is essentially a ‘bulk’ technique, protein monolayer detection is possible, although close to the current limit of sensitivity [6]. In particular, the sensitivity to an adsorbed layer depends on how different the adsorbate spectrum is relative to that of the underlying substrate. We find the most reliable detection of weak protein deposits when we use the full spectral signature in analysis of a full image sequence (40– 80 energies) rather than simply imaging at one or a few photon energies. The upper panel of Fig. 2 presents the C 1s spectra of albumin and fibrinogen. The spectra are rather similar, with each being dominated by the strong peak at 288.2 eV, which is the C 1s(C O) → πC∗ O transition at the amide group of the peptide bond. Each also shows a weak peak at 285.1 eV, associated with C 1s(C C) → πC∗ C transitions at the phenyl groups of aromatic amino acids. One small difference between the spectra of albumin and fibrinogen is that albumin has a weak peak at 289 eV which is not seen in fibrinogen. Since each protein has some of each of the 20 amino acids, it is expected that averaging over such distributions leads to very little differentiation. This is in sharp contrast to the isolated amino acids where it is straightforward to identify the amino acid from the C 1s spectrum [24]. The lower section of Fig. 2 explores the STXM detection limits for pure protein in the absence of any other organic material. The deposit of albumin from a very dilute solution on a clean silicon nitride window shows a readily differentiated C 1s spectrum typical of protein which is three to four times more intense than the noise in a blank (the spectrum from an equivalent length of the adjacent window which does not have deposited protein). The optical density associated with the protein signal in the circled region is equivalent to a sample thickness of 3 nm, approximately the expected thickness of an albumin monolayer. Figure 3 reports the result of a measurement of a sample consisting of human serum albumin deposited onto the #529 PIPA polyurethane substrate. This sample was prepared by placing a drop of a 1 mg/ ml solution onto the microtomed section supported on a 3 mm TEM grid. An uneven distribution of the albumin was left on the surface after solvent evaporation. A region at the edge of a thick deposit was chosen to explore the detection sensitivity. The strength of the protein-like C 1s NEXAFS signal, in the regions where the map indicates the weakest signal which can be attributed to protein (circle in Fig. 3), is approximately equivalent to a monolayer. This sample was not prepared as a controlled biomaterial — protein interaction but it is of historical interest since it was our first demonstration of the surface sensitivity of STXM and thus the viability of the technique. Figures 4– 7 show results from mapping fibrinogen (Fg) adsorbed on a microtomed section of the 2-filler #355 polyurethane system presented in Fig. 1. These
472 A. P. Hitchcock et al. Figure 3. Reference spectroscopy (left), transmission images at 285.1 eV (highlighting PIPA) and 288.2 eV (highlighting protein) (center); and (right) component maps of albumin and PIPA derived from a C 1s image sequence recorded from a thin section sample of #529 (PIPA-in-TDI polyurethane) with deposited albumin, measured in the dry state. In the weakest regions of the albumin ‘smear’ (e.g. the circled region), the signal is at the monolayer level.
Towards practical soft X-ray spectromicroscopy of biomaterials
473
figures exemplify four different steps in the evolution of our ability to detect proteins on polymer surfaces. For the sample shown in Fig. 4, the fibrinogen was adsorbed on the sample by immersion for 20 min in a dilute solution (0.1 mg/ml) in buffer. Excess protein solution was removed by through rinsing without taking the sample through the air water interface. The sample was then dried. The microtomed sample has three classes of chemical constituents exposed at the surface — polyether-rich matrix, SAN and PIPA. The left part of Fig. 4 plots the reference spectra used in the data analysis. The labeled component maps indicate the spatial distribution of the identified components. The residual shows a map of the difference between the fit and the actual image sequence, integrated over the full energy range (54 energies between 282 and 292 eV). The residual map represents deviations between the fit and the measured signal of less than 5%. The spatial distribution of the residuals indicates that the SAN reference spectrum is not a perfect match to the SAN in this sample. While there is some mis-identification in these maps associated with limitations of the reference spectra and statistical noise, the high signals in the Fg map indicate positions of preferred adsorption on the surface. The registry of the protein relative to the other components is indicated in the final part of Fig. 4 namely the color coded composite component map in the lower right part. Here the individual SAN, PIPA and protein maps have been combined by assigning red to SAN, green to PIPA and blue to fibrinogen. The intensities within each color have been byte scaled, so that the very weak protein signal can be located relative to the strong SAN and PIPA signals. In this preparation, we find that fibrinogen has a preference to be attached to the matrix beside the SAN particles, with also indication from the orange color of some adsorption on top of the SAN particles. The sample examined in Fig. 5 was prepared as that for Fig. 4 but, just prior to STXM analysis it was rehydrated with a small drop of deionized water then capped by a second silicon nitride window to form a wet cell. In this measurement we wished to explore the ability of STXM to detect protein adsorbed on a polymer in an environment of reduced contrast caused by the X-ray absorption of an aqueous overlayer of a few micrometers in thickness. The results indicate that STXM can detect adsorbed proteins on polymers in the presence of thin aqueous overlayers. Furthermore the amount of fibrinogen detected is in the monolayer range in some regions. In addition to the color coded composite component map, Fig. 5 shows a fit to the C 1s NEXAFS spectrum extracted from pixels which have high fibrinogen content. These pixels were identified by generating a binary mask based on an intensity threshold for the fibrinogen map (threshold set to 60 nm), and using that mask to define the region of interest for spectral extraction from the full image sequence. The spectrum in the high-Fg region is primarily that of the matrix rather than SAN. We have compared this analysis to that of the adjacent SAN particles (not shown). The quality of the fit is equally good. The amount of Fg determined on top of the SAN particles is only 10 nm, a signal level similar to the residual of the fit.
474 A. P. Hitchcock et al. Figure 4. Analysis of STXM of fibrinogen adsorbed on #355 (TDI polyurethane with PIPA and SAN filler particles) measured in the dry state. The sample was prepared by a true adsorption process from buffer at low concentration (0.1 mg/ ml) followed by carefully washing off the excess protein then drying, (left) reference spectra, placed on absolute linear absorption scales. Four of the six ‘image’ panels are the quantitative component maps of the SAN, PIPA, fibrinogen and matrix components; the vertical gray scale of each is the approximate thickness of that component in nm. The other two are the map of the residuals averaged over the full energy range measured and a color coded component map (red = SAN, green = PIPA, blue = Fg).
Towards practical soft X-ray spectromicroscopy of biomaterials 475
Figure 5. (left) Color-coded composite image derived from a C 1s STXM image sequence of fibrinogen adsorbed from a 0.1 mg/ ml buffer solution onto #355 substrate. The measurements were made with the sample rehydrated by a ∼1 µm thick layer of deionized water in a silicon nitride wet cell, in order to explore the masking effect of overlayer water. The combined byte-scaled SAN (red), PIPA (green) and protein (blue) component maps displays the spatial relationship of the protein relative to the two types of filler particles and the matrix. (right) Spectrum of the blue highlighted regions in the insert image (pixels where the Fg signal indicates more than 60 nm) along with decomposition into the four fitted components. The points are data, the thickest solid line the fit with all components, the thinnest line the fit without fibrinogen, and the other lines the individual components.
476
A. P. Hitchcock et al.
These results indicate that fibrinogen has a strong preference to adsorb at the edges of the SAN particles but still to be attached to polyurethane matrix rather than SAN. Similarly, on the #530 substrate the albumin is seen to adsorb preferentially at the edges of the PIPA particles (see Fig. 3). Several factors might explain this preference. First there could be special aspects of this substrate at the interface of the CPP particles and the polyurethane matrix that enhance bonding to proteins. Second, adsorption at edges of harder, aromatic filler particles could reflect ‘mechanical trapping’. The SAN particles protrude significantly from the matrix as much as 50 nm based on the total sample thickness derived in the analysis. A simple, perhaps overly naïve, model of such trapping would be entanglement of protein molecules (possibly partially denaturing in the process) on the protruding filler particles. A more sophisticated mechanism could be entropically driven deposition [25] in which the protein in solution would be considered to play a role. At this point it is not possible to define the mechanism of preferential attachment, simply to note there is a clear preference for fibrinogen to adsorb on the matrix side of the boundary of the CPP particles (particularly SAN) and the polyurethane matrix, and that the STXM technique is capable of detecting that fibrinogen in a quantitative manner. Figure 6 presents the color coded component map derived from a C 1s image sequence measured from a filler-polyurethane sample (#355) with an overlayer of solution containing 0.01 mg/ml fibrinogen in buffer. In this presentation the (R = SAN, G = PIPA, B = fibrinogen) map has been superimposed on the gray scale map of the polyurethane matrix component. This is essentially the type of ‘in situ’ system which is the experiment of ultimate interest, since the solution overlayer contains both buffer salts and protein. These additional solution components might be expected to reduce the contrast of the adsorbed protein relative to the other components. At first we were concerned that the C 1s signal of the protein in the solution could mask the adsorbed protein. Relative to monolayer surface levels, the amount of protein present in a 10 µm high, 1 µm diameter column of a 0.01 mg/ ml fibrinogen solution is about the same as that adsorbed at a monolayer coverage on a 1 µm diameter circle of a surface. Thus one might expect a background signal from the free protein over the whole surface, which could mask detection of surface adsorption sites. However, this background signal has not been detected. We speculate that local heating of the solution from the X-ray beam could induce increased thermal motion which might tend to move protein in the solution away from the impinging beam. Experiments to seek the threshold for interference from solution protein will be performed. So far, these results indicate that it is possible to map protein on surfaces from 0.01 mg/ ml and even considerably more concentrated solutions since we have measured surface adsorbed protein with protein solution overlayers of up to 0.1 mg/ml concentration. Figure 7 shows the same Fg/#355 sample investigated at the N 1s edge. This figure plots the N 1s reference spectra of the matrix, PIPA, SAN and fibrinogen, along with the color coded composite map of the SAN, PIPA and fibrinogen
Towards practical soft X-ray spectromicroscopy of biomaterials
477
components. The N 1s edge is particularly favorable for studies of protein relative ∗ transition at to polymer substrates since proteins have a strong N 1s → πamide 402 eV, [26] whereas many polymers do not contain unsaturated N environments. ∗ transition at 399.8 eV which is readily The SAN also has a strong N 1s → πCN distinguishable from the higher energy protein signal. However the PIPA and matrix have very similar N 1s signals and thus the PIPA component is less well detected than at the C 1s edge. The quantitative chemical analysis at the N 1s edge gives similar results to that at the C 1s edge (note that the areas measured for Figs 6 and 7 are different). The interesting aspects of the results shown in Figs 6 and 7 are: (1) There was no interference from the protein or buffer salts present in the overlayer solution. (2) The mapping of the fibrinogen derived from the C 1s and N 1s edges is very similar, indicating that detection is not an artifact of a mismatch of model spectra and the unknown. (3) The location of the protein is consistently at the sides of the SAN particles but on the matrix, not on the SAN, in all studies of the protein exposed sample #355. Together these results constitute clear evidence that STXM can detect proteins at polymer surfaces under aqueous layers containing inorganic buffer salts and protein. We do note that the quantitative accuracy is limited and that the enclosed results may be affected to some degree by residual misalignment. In the as-recorded image sequences there are drifts as large as a few microns in the field of view associated with poor tracking of the zone plate along the X-ray axis (it is necessary to move the ZP with photon energy to maintain focus). Software procedures [10, 14] remove most of this misalignment but there is residual jitter of as much as a hundred nanometers. Recently we have measured the dry Fg/#355 system using a new STXM at the ALS, one which uses a two dimensional interferometer system [12] to maintain a constant field of view. The results were very similar to those presented here.
4. DISCUSSION AND SUMMARY
These results demonstrate that STXM is a viable technique to address questions of chemical differentiation of substrates and protein localization at the surfaces of biomaterials under conditions which are relatively close to their actual use. This is in distinct contrast to vacuum based techniques such as TOF-SIMS, XPS or XPEEM where there is always a question whether the absence of water or buffer changes the character of the interface, e.g. by inducing surface segregation of hydrophobic components. STXM provides information on lateral morphology for both dry and wet samples. This is potentially useful for studying artificially patterned biomaterials, chemical mapping of biomaterial substrates, and quantitative mapping of adsorbed protein relative to the substrate. Perhaps the weakest aspect of the STXM method is the lack of intrinsic depth resolution since the signal at any location is a column average. It is very unlikely that the protein signal arises from anywhere but the polymer surface. However the same cannot be said for the signal from the underlying polymer. In that respect XPEEM should be a
478
A. P. Hitchcock et al.
Figure 6. Color-coded composite image derived from a C 1s STXM image sequence of a #355 sample covered with a protein solution (fibrinogen in buffer, 0.01 mg/ ml). In this case a bytescaled combination of the SAN (red), PIPA (green) and protein (blue) component maps has been superimposed on a gray scale image of the polyurethane matrix map. This result was derived from an image sequence measurement of a sample with the fibrinogen buffer solution left over the surface. The overlayer of buffer is estimated from the pre-edge signal to be ∼5 microns thick.
useful complement to STXM since it has the potential to provide better surface characterization of polymer substrates at similar spatial resolution. Indeed recently we have made a detailed AFM – STXM – XPEEM comparison of the surfaces of polystyrene– polymethylmethacrylate blends [27] which showed very considerable differences between composition and morphology of the bulk as sampled by STXM, and that of the surface as sampled by XPEEM. The XPEEM and AFM images showed similar morphology, but the conventional interpretation of the AFM based on comparing bulk sample composition with relative areas of the continuous and discontinuous domains, was found to give an inverted assignment of the chemical identity of the domains. On the other hand the much higher spatial resolution of the AFM revealed micro-domain features which were not detected by XPEEM, and were just barely detectible by STXM. As in many other fields, the most effective way to solve complex problems such as biomaterial interfaces is to use multiple techniques with proper recognition of the strengths and weaknesses of each technique. This philosophy has been a hallmark of the scientific career of Dr. Brash, to whom this article is dedicated. Dramatic improvements are currently underway in scanning transmission X-ray microscopy instrumentation, performance and analysis methodology. Cryo-STXM has been implemented recently at NSLS [28]. Two microscopes with interferometric control of sample-zone plate position have recently been commissioned at the
Towards practical soft X-ray spectromicroscopy of biomaterials
479
Figure 7. Color-coded composite image derived from a N 1s STXM image sequence of fibrinogen adsorbed from a 0.01 mg/ ml buffer solution onto #355 substrate. This is a different area of the same sample for which the C 1s results are plotted in Fig. 6. The N 1s spectra of the pure reference materials are also plotted. The protein spectrum is from albumin recorded at a different synchrotron facility [26], while the N 1s spectra of SAN, PIPA and the matrix were those of the pure #355 material [23].
480
A. P. Hitchcock et al.
ALS [11]. The interferometric signal provides a precise solution to a major problem in earlier instruments, namely drift in the field of view as the photon energy is scanned. These new instruments will allow detailed studies of many biomaterials problems. In a few years it will be possible to readily carry out orientation contrast studies. Early studies [29] showed that orientation contrast can be an important aspect of polymer X-ray microscopy. However those measurements were extremely tedious since the sample had to be removed, rotated, and the same region found in the new orientation. In a few years time there will be STXM microscopes on beam lines at the ALS, Swiss Light Source, and the Canadian Light Source, which will be illuminated by elliptically polarized undulators (EPU). When proper control of the phase shifting of the separate sections is employed, EPUs can produce linearly polarized light with user selectable, arbitrary orientation [30]. This will allow routine exploitation of polarization contrast which will be important for studying polymeric biomaterials with aligned molecules. Finally, a significant portion of the polymer and biomaterial soft X-ray microscopy research carried out to date has been performed by, or in collaboration with industrial researchers. The rapid recognition by industry of the remarkable value of soft X-ray microscopy techniques attests to the added value NEXAFS microscopy brings to practical problem solving relative to other, more accessible, lab-based analytical microscopy techniques. It may be anticipated that the biomaterials industry will recognize the benefits of STXM for investigating and thus advancing their materials. Acknowledgements This work is funded by the Natural Science and Engineering Research Council (Canada) and the Canada Research Chair program. The ALS STXM was developed by T. Warwick (ALS), B. P. Tonner (U Wisconsin Milwaukee) and collaborators, with support from the US DOE under contract DE-AC03-76SF00098. Zone plates used at the ALS were provided by Eric Anderson of CXRO, LBNL. We thank ALS staff for much assistance and expert operation. We thank especially Ed Rightor, Werner Lidy and Gary Mitchell (Dow Chemical) for providing the polyurethane substrates, and Rick Steele, George Meigs, Eli Rotenberg and Tony Warwick (Advanced Light Source, Berkeley) for their capable work in developing and maintaining the ALS 7.0 beamline and STXM. REFERENCES 1. R. M. Cornelius and J. L. Brash, Biomaterials 20, 341 (1999). 2. A. M. Botelho do Rego, O. Pellegrino, J. M. G. Martinho and J. Lopes da Silva, Langmuir 16, 2385 (2000). 3. J. Kirz, C. Jacobsen and M. Howells, Q. Rev. Biophys. 28, 33 (1995). 4. H. Ade, in: Experimental Methods In The Physical Sciences, J. A. R. Samson and D. L. Ederer (Ed.), Vol. 32, p. 225. Academic Press (1998). 5. H. Ade and S. G. Urquhart, in: Chemical Applications of Synchrotron Radiation, T. K. Sham (Ed.). World Scientific Publishing (2002).
Towards practical soft X-ray spectromicroscopy of biomaterials 6. 7. 8. 9. 10. 11. 12.
13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.
24. 25. 26. 27.
28. 29. 30.
481
A. P. Hitchcock, J. Synchrotron Radiation 8, 66 (2001). J. Stöhr, NEXAFS Spectroscopy, Springer Tracts in Surface Science, Vol. 25 (1992). H. Ade, X. Zhang, S. Cameron, C. Costello, J. Kirz and S. Williams, Science 258, 971 (1992). H. Chapman, C. Jacobsen and S. Williams, Ultramicroscopy 62, 191 (1996). C. Jacobsen, S. Wirick, G. Flynn and C. Zimba, J. Microscopy 197, 173 (2000). T. Warwick, H. Ade, D. Kilcoyne, M. Kritscher, T. Tyliszczak, S. Fakra, A. P. Hitchcock and H. Padmore, J. Synchrotron Rad. 9, 254 (2002). A. L. D. Kilcoyne, T. Tyliszczak, R. Steele, A. P. Hitchcock, S. Fakra, K. Frank, C. Zimba, E. Rightor, G. Mitchell, I. Koprinarov, E. Anderson, B. Harteneck, A. P. Hitchcock, T. Warwick and H. Ade, J. Synchrotron Rad. (2002) (in press). E. G. Rightor, A. P. Hitchcock, H. Ade, R. D. Leapman, S. G. Urquhart, A. P. Smith, G. Mitchell, D. Fischer, H. J. Shin and T. Warwick, J. Phys. Chem. B 101, 1950 (1997). aXis2000 is an IDL ‘widget’ (application), which is available from http://unicorn.mcmaster.ca/ aXis2000.html. IDL is a Kodak company — www.rsinc.com. X. Zhang, R. Balhorn, J. Mazrimas and J. Kirz, J. Struc. Biol. 116, 335 (1996). I. Koprinarov, A. P. Hitchcock, C. T. McCrory and R. F. Childs, J. Phys. Chem. B 106, 5358 (2002). Silson Ltd., JBJ Business Park, Northampton Road, Blisworth, Northampton England, NN7 3DW. R. Herrington, Flexible Polyurethane Foams, 2 edn. The DOW Chemical Company (1997). S. G. Urquhart, A. P. Hitchcock, R. D. Leapman, R. D. Priester and E. G. Rightor, J. Polymer Science B: Polymer Physics 33, 1593 (1995). S. G. Urquhart, H. Ade, A. P. Smith, A. P. Hitchcock, E. G. Rightor and W. Lidy, J. Physical Chemistry B 103, 4603 (1999). S. G. Urquhart, A. P. Hitchcock, A. P. Smith, H. Ade, W. Lidy, E. G. Rightor and G. E. Mitchell, J. Electron Spectrosc. 100, 119 (1999). E. G. Rightor, G. E. Mitchell, S. G. Urquhart, A. P. Smith, H. Ade, A. P. Hitchcock, A. Aneja and R. J. Wilkes, Macromolecules 35, 5873 (2002). A. P. Hitchcock, I. Koprinarov, T. Tyliszczak, E. G. Rightor, G. E. Mitchell, M. T. Dineen, F. Hayes, W. Lidy, R. D. Priester, S. G. Urquhart, A. P. Smith, H. Ade, Ultramicroscopy 88, 33 (2001). K. Kaznacheyev, A. Osanna, C. Jacobsen, O. Plashkevych, O. Vahtras, H. Ågren, V. Carravetta and A. P. Hitchcock, J. Chem. Phys. 106, 3153 (2002). K. H. Lin, J. C. Cracker, V. Prasad, A. Schofield, D. A. Weitz, T. C. Lubensky and A. G. Yodh, Phys. Rev. Lett. 85, 1770 (2000). B. W. Loo, Jr., I. M. Sauerwald, A. P. Hitchcock and S. S. Rothman, J. Microscopy 204, 69 (2001). C. Morin, H. Ikeura-Sekiguchi, T. Tyliszczak, R. Cornelius, J. L. Brash, A. P. Hitchcock, A. Scholl, F. Nolting, G. Appel, A. D. Winesett, K. Kaznacheyev and H. Ade, J. Electron Spectroscopy 121, 203 (2001). J. Maser, A. Osanna, Y. Wang, C. Jacobsen, J. Kirz, S. Spector, B. Winn and D. Tennant, J. Microscopy 197, 68 (2000). A. P. Smith and H. Ade, Appl. Phys. Lett. 69, 3833 (1996). A. T. Young, E. Arenholz, S. Marks, R. Schlueter, C. Steier, H. A. Padmore, A. P. Hitchcock and D. G. Castner, J. Synchrotron Rad. 9, 270 (2002).
This page intentionally left blank
A new vascular polyester prosthesis impregnated with cross-linked dextran DELPHINE MACHY 1,∗ , PATRICK CARTERON 2 and JACQUELINE JOZEFONVICZ 1 1 Laboratoire
de Recherches sur les Macromolécules, FRE 2314 CNRS, Université Paris 13, Avenue J.-B. Clément, 93430 Villetaneuse, France 2 Cardial S.A., a Subsidiary of C. R. Bard, 28 Rue de la Télématique, B.P. 746, 42950 Saint-Etienne Cedex 9, France Received 30 November 2001; accepted 29 April 2002 Abstract—It is essential that a synthetic vascular graft is preclotting prior to implantation in order to prevent blood leaking through the graft wall. We have impregnated a knitted polyester prosthesis with cross-linked dextran. The aim of this study was to develop a process for obtaining an impervious prosthesis and to compare the characteristics of this dextran-impregnated graft with those of a commercially available collagen-impregnated graft. This new vascular prosthesis was coated with dextran; sodium trimetaphosphate was utilized as the cross-linking agent. In an attempt to determine the optimal conditions for impregnation, the dynamic viscosity of the dextran solution was measured during the cross-linking reaction. The results suggest that the dynamic viscosity is correlated with the concentrations of dextran, sodium hydroxide, and sodium trimetaphosphate. The effect of temperature on the dynamic viscosity was also investigated. The water permeability, the coating weight, and the structure of the dextran-impregnated graft were compared with those of a collagen-impregnated prosthesis. The water permeability of the vascular grafts was reduced by dextran impregnation, from 1010 ml/ min per cm2 for the control to 0.04 ml/ min per cm2 under standard testing conditions. The dextran coating is capable of rendering the graft impervious to water. The coating weight of the graft treated with dextran was approximately the same as the weight of the collagen-impregnated graft. Finally, the morphology of the prosthetic wall was analyzed using scanning electron microscopy. The promotion of endothelial cell recovery was only observed for the polyester grafts treated with dextran or collagen. Key words: Vascular graft; cross-linked dextran; dynamic viscosity; water permeability; endothelial cell.
∗ To
whom correspondence should be addressed. E-mail: [email protected]
484
D. Machy et al.
INTRODUCTION
Atherosclerosis is the most widely occurring vascular disease and is characterized by smooth muscle cell proliferation, cholesterol and lipoprotein deposition, and thrombus formation [1]. The only treatment available at an advanced stage of the disease is to bypass or replace the damaged blood vessels. When neither autologous veins nor arteries can be used as substitutes, surgeons employ a synthetic vascular graft made of either polyethylene terephthalate (PET), expanded polytetrafluoroethylene (ePTFE) or polyurethane (PU) [2, 3]. PET, ePTFE, and porous PU prostheses all present the same disadvantage of being highly permeable, which may result in severe bleeding. At present, the majority of grafts used for surgical treatment are impregnated either with proteins such as collagen [4, 5] and albumin [6, 7], or with gelatin [8, 9] cross-linked with glutaraldehyde, formaldehyde or carbodiimide cross-linkers. These manufacturing procedures for the preparation of prostheses prevent blood leakage during implantation. However, these procedures are both complicated and expensive, because it is necessary to use proteins with a high degree of purity. The use of such proteins also presents a risk of infectious contamination. Polysaccharide-based materials are a potential solution to avoid this risk of contamination and can be used as a coating for a vascular prosthesis. Dextran is a D-glucopyranose, synthesized from sucrose by bacterial species such as the Leuconostoc mesentoroides strain B 512. This polysaccharide and some of its derivatives are non-toxic and are widely used for medical applications such as plasma expanders, blood substitutes, bonehealing promoters, wound dressings, and drug delivery [10– 15]. In the present work, the conditions to seal the graft with cross-linked dextran, using sodium trimetaphosphate (STMP) as the cross-linking agent, were investigated to produce an impervious prosthesis. Some of the properties of the dextran-impregnated grafts were compared with those of a commercially available collagen-impregnated graft. MATERIALS AND METHODS
Materials Dextran T500 (Mw = 482 000 g/ mol, Mn = 166 200 g/ mol, batch 286 753) was supplied by Pharmacia (St Quentin-en-Yvelynes, France). All ‘analytical grade’ chemicals and solvents were purchased from Aldrich (St Quentin Fallavier, France), Acros (Noisy-le-Grand, France), and Sigma (St Quentin Fallavier, France). Sodium trimetaphosphate (Sigma), lactic acid and glycerol were purchased from Prolabo (Briare Le Canal, France). The knitted PET prostheses (8 mm in diameter) were provided by Cardial S.A. (St Etienne, France). Dextran-impregnated prosthesis The uncoated knitted PET graft used for the collagen-impregnated vascular graft (Dialine II® ) was sealed as follows: the required amount of dextran was dissolved
Dextran-impregnated vascular prosthesis
485
in an aqueous solution of sodium hydroxide, at 40◦ C, for approximately 20 min. Then STMP was added to the stirred solution, which was left to stand at 20◦ C, ready to use for impregnation. The impregnation solution was infused into the graft fibers by a pump (Minipuls 2, Gilson Medical Electronics S.A., Villiersle-Bel, France). This procedure resulted in the cross-linking of the molecular dextran chains via covalent phosphate diester bonds. After washing the graft three times with a water solution which included the plasticizers and glycerol and lactic acid at concentrations ranging from 0.2 to 20% (v/ v) and from 0.07 to 7% (v/ v), respectively, the impregnated grafts were dried. The drying process was stopped when the water content was approximately 80% of the weight of the graft before coating. A non-impregnated knitted PET graft was selected as the control. A Dialine II® prosthesis impregnated with bovine type I collagen cross-linked with vapors of formalin was also used for comparison [4]. Water permeability The water permeability was determined using a standard procedure described by Guidoin et al. [2]. It was measured as the amount of water leaked per unit area and time under a physiological pressure of 120 mmHg. The water permeabilities of the control without a coating and the dextran-impregnated and collagen-impregnated grafts were compared. Each measurement was taken three times. Coating weight The grafts were weighed before impregnation (W1 ). After impregnation, the grafts were washed as previously described, dried, and weighed (W2 ). The coating weight of the dextran-impregnated graft was determined as follows: Coating weight = (W2 − W1 )/W1 . The coating weight is expressed in terms of the weight of sealant in milligrams per gram of prosthesis. Dynamic viscosity Dextran was dissolved in an aqueous sodium hydroxide solution at 40◦ C. After 20 min, STMP was added and the dynamic viscosity of the solution was measured immediately in a volumetric flask with a viscosimeter (Rheovisco L2, Rheo, Shannon, Ireland). The viscosimeter was submerged in a thermostatic bath at various temperatures. Cell culture Human endothelial cells, EA hy 926, were utilized as permanent cells and were provided by Dr. C. J. S. Edgell [16]. The cultures were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Life Technologies, Cergy-Pontoise,
486
D. Machy et al.
France) containing 10% fetal calf serum (FCS; Eurobio, Les Ulis, France), 2 mM L -glutamine, 100 m M hypoxanthine, 0.4 m M aminopterine, and 16 m M thymidine (Gibco) at 37◦ C, in a 5% CO2 humidified atmosphere in tissue culture flasks (Costar, Issy les Moulineaux, France). Before passage and seeding, the cells were detached from the tissue culture flasks using a trypsin– EDTA solution. The absence of mycoplasma contamination was verified by employing an enzymatic method: the AdoP assay of mycoplasma contamination of biological media [17]. Scanning electron microscopy Endothelial cells were seeded by inoculating 10 000 cells/ cm2 in 24-well plates, in 1 ml of culture medium, as described above. After 7 days in the culture medium, the PET samples were withdrawn, rinsed twice in PBS, and then fixed in a solution (6.75 mM NaOH, 10 mM NaH2 PO4 , 4% formaldehyde, 1% glutaraldehyde) for 30 min at room temperature. After rinsing with PBS, the PET samples were dehydrated in a range of water/ ethanol mixtures which were increasingly rich in ethanol and finally in absolute ethanol. The samples were critical point-dried, coated with gold, and then examined using a LEICA S440 scanning electron microscope (LEO, Reuil-Malmaison, France) operating at a 15 kV accelerating voltage.
RESULTS AND DISCUSSION
The reaction of dextran with STMP occurs via the mechanism shown in Fig. 1 [18, 19]. In the presence of STMP as a cross-linking agent and in alkaline conditions, dextran forms a hydrogel. The dynamic viscosity was measured immediately after adding STMP to the dextran solution. The high dynamic viscosity produced by the formation of the hydrogel rapidly inhibits further measurements of viscosity. Effect of the concentration of sodium hydroxide on the dynamic viscosity The influence of the concentration of sodium hydroxide (NaOH) on the dynamic viscosity behavior of dextran (10% w /v) cross-linked with STMP (8 mol of STMP for 100 mol of glucosidic unit, 8% STMP), at 25◦ C, versus the time is shown in Fig. 2. The dynamic viscosity increases with increasing NaOH concentration except for the 0.2 M NaOH solution, where complete gelation was achieved only after a very long period of time. The time necessary for complete gelation was greatly reduced when a 1 M solution of NaOH was employed. This result can be attributed to the increased hydrolysis of STMP at higher concentrations of NaOH. This hydrolysis leads to a greater number of cross-links between the hydroxyl functions of dextran and STMP, and consequently augments the dynamic viscosity of the solution.
Dextran-impregnated vascular prosthesis
487
Figure 1. Schematic synthetic pathway for cross-linking dextran with STMP.
Figure 2. Effect of the NaOH concentration on the dynamic viscosity. Solution: dextran 10% (w/v), STMP 8 mol for 100 mol of glucosidic unit (8% STMP), 20◦ C.
488
D. Machy et al.
Figure 3. Influence of the concentration of the cross-linking agent on the dynamic viscosity. Solution: dextran 10% (w/ v), NaOH 0.7 M, 20◦ C.
Effect of the concentration of the cross-linking agent on the dynamic viscosity The results in Fig. 3 demonstrate the relationship between 10% (w / v) dextran and various concentrations of STMP employing a 0.7 M NaOH solution at 25◦ C. It can be observed that the dynamic viscosity is dependent on the concentration of STMP and increases with increasing concentration of cross-links. This result corresponds to an increase in the formation of diester phosphate linkages when a greater quantity of STMP is available. The time necessary to form the hydrogel is directly dependent on the concentration of the cross-linking agent. Figure 3 also shows that without STMP, the dynamic viscosity of the solution is stable. This result indicates that the dextran solution exhibits Newtonian behavior, as described by Carrasco et al. [20]. The thermodynamic analysis presented by these authors suggests that in an aqueous solution dextran macromolecules behave as flexible polymeric chains. The Newtonian behavior can be explained by the high density of the dextran macromolecular chains produced by their extensive molecular branching, as demonstrated by the experimental viscosity values presented here. Effect of the concentration of the substrate on the dynamic viscosity The influence of the concentration of dextran on the dynamic viscosity versus the time is presented in Fig. 4. The solution used for evaluating the effect of the concentration of the substrate contained 8% STMP and 0.2 M NaOH at 25◦ C. The
Dextran-impregnated vascular prosthesis
489
Figure 4. Effect of the concentration of the dextran on the dynamic viscosity. Solution: NaOH 0.5 M, 8% STMP, 20◦ C.
dextran concentration clearly influenced the dynamic viscosity. The gelation period was very short for a high concentration of dextran, 20% (w / v). The time required to obtain a hydrogel is about 35 min for a dextran concentration of 20% (w / v). This period was approximately twice as long as that for a 15% (w / v) dextran solution. From these data, it is possible to develop a relationship between the dynamic viscosity and the concentration of dextran. A relationship between the dynamic viscosity and the molecular weight was also established. These observations are in agreement with other studies [20]. Effect of the temperature on the dynamic viscosity Figure 5 presents the effect of the temperature on the dynamic viscosity. As expected, an increase in the temperature promotes the cross-linking process between the STMP and the polysaccharide. Consequently, the dynamic viscosity increased more rapidly at a high temperature. This behavior can be explained by an increasing rate of phosphate diester bond formation at higher temperatures. This systematic study clearly shows that the dynamic viscosity is dependent on the four parameters considered: the concentrations of sodium hydroxide, sodium trimetaphosphate, and dextran, and the temperature. We observed that the rate of gelation is also governed by these same key parameters. The various hydrogels obtained present different rheological characteristics, such as swelling capacity, cross-linking homogeneity, water permeability, and dextran release. On the basis
490
D. Machy et al.
Figure 5. Effect of the temperature on the dynamic viscosity. Solution: NaOH 0.2 20% (w/v), 8% STMP.
M,
dextran
of this information, and hence by controlling the relationship between the dynamic viscosity and rheological properties, it is possible to coat a prosthesis with gel by judiciously choosing the production parameters. Water permeability and coating weight It has been clearly shown that the four parameters studied influence the crosslinking between dextran and STMP. The time necessary to obtain complete gelation depends on the experimental conditions. The impregnation during the cross-linking reaction must be sufficient to obtain a good coating of the prosthesis, rendering it impervious. We have investigated a range of conditions for impregnating the vascular graft with dextran, taking into account the results described above. Then the water permeability of the prosthesis was determined. The water permeability and the coating weights of the untreated control, the collagen-impregnated graft (Dialine II® ) and the dextran-impregnated vascular grafts are presented in Table 1. The results were obtained for prostheses impregnated 30 min after the addition of STMP to the dextran solution. In all cases, the dextran-impregnated prostheses exhibited a reduced water permeability compared with the untreated control. The water permeability can be varied as a function of the conditions of impregnation. Increasing the concentration of STMP has no effect on the water permeability of the dextran-impregnated graft. However, an increase in the concentration of NaOH from 0.2 to 1 M greatly
Dextran-impregnated vascular prosthesis
491
Table 1. Water permeability and coating weight of the grafts
Graft
Water permeability (ml/min per cm2 )
Coating weight (mg coating/ g graft)
Control Collagen-impregnated [4] Dextran-impregnateda Dextran-impregnatedb Dextran-impregnatedc Dextran-impregnatedd Dextran-impregnatede
1010 0 14 ± 0.6 16.3 ± 1 16.9 ± 0.5 0.045 ± 0.05 9.8 ± 1
419.4 ± 77.1 151 ± 35 144 ± 5 161 ± 32 873 ± 30 488 ± 82
Water permeability measured at 120 mmHg. a Dextran 10% (w/v), STMP 8 mol for 100 mol of glucosidyl unit, NaOH 0.2 mol/ l, 25◦ C. b Dextran 10% (w/v), STMP 15 mol for 100 mol of glucosidyl unit, NaOH 0.2 mol/ l, 25◦ C. c Dextran 10% (w/ v), STMP 20 mol for 100 mol of glucosidyl unit, NaOH 0.2 mol/ l, 25◦ C. d Dextran 10% (w/v), STMP 8 mol for 100 mol of glucosidyl unit, NaOH 1 mol/ l, 25◦ C. e Dextran 20% (w/ v), STMP 8 mol for 100 mol of glucosidyl unit, NaOH 0.2 mol/ l, 25◦ C.
improved the water permeability of the prosthesis. It is possible to obtain a completely impervious dextran-impregnated graft (0.04 ml/ min per cm2 ) with a 1 M NaOH aqueous solution. By increasing the concentration of dextran in the impregnation solution, the water permeability was decreased two-fold. The weight of the coating on the dextran-impregnated graft was dependent on the impregnation conditions and was of the same order as that of the collagen-impregnated graft. Scanning electron microscopy studies The surface morphologies of the control and the dextran- and collagen-treated grafts were observed by scanning electron microscopy (SEM). The control graft has a warp-knitted structure with void spaces or porosity between the fibers and yarns, which explains the high water permeability obtained (Fig. 6a). This prosthesis was made from a mixture of flat and textured yarns in a two-bar structure with a lockknit lapping sequence. In the case of the graft treated with dextran, the surface wall revealed that the cross-linked dextran sealant had completely impregnated the wall of the prosthesis (Fig. 6c). The same observation was made for the collagenimpregnated Dialine II® graft (Fig. 6b). In order to evaluate the in vitro prosthetic endothelialization, the samples seeded with endothelial cells were analysed by SEM. The SEM images show colonization of the grafts after 7 days in the culture medium. The uncoated graft is not covered by cells (Fig. 7a), while endothelial cells were detected for both the collagen- and the dextran-impregnated grafts (Figs 7b and 7c). The capacity of proliferation on the impregnated vascular graft proved that cells are viable on this coating.
492
D. Machy et al.
(a)
(b)
(c) Figure 6. Scanning electron photomicrographs of the surface of the vascular prosthesis: (a) untreated control, interior; (b) collagen-impregnated, interior; (c) dextran-impregnated, interior.
Dextran-impregnated vascular prosthesis
493
(a)
(b)
(c) Figure 7. Scanning electron photomicrographs of the surface of the vascular prosthesis 7 days after seeding: (a) untreated control, interior; (b) collagen-impregnated, interior; (c) dextran-impregnated, interior.
494
D. Machy et al.
CONCLUSIONS
In the present study, dextran has been used to prepare a novel type of artificial vascular prosthesis. In an attempt to develop a process for obtaining an impervious knitted dextran-impregnated vascular graft, we have studied the dynamic viscosity of the solution used for impregnation. Thus, an appropriate cross-linked dextran for the prosthesis coating was selected which was capable of rendering the vascular prosthesis impervious. Our results show that cross-linked dextran is a good alternative to proteins, such as collagen, for coating vascular prostheses. Furthermore, dextran is known to be non-toxic, is widely used for biomedical applications, and is less expensive than collagen. Finally, it is also known to have relatively low binding properties with protein [21– 24]. The consequent reduction in protein adhesion to the surface of a biomaterial may result in low non-specific binding of blood compounds, such as fibrinogen or thrombin, which are detrimental to graft patency. Dextran needs to be cross-linked with sodium trimetaphosphate, in order to prevent the leaching of dextran. Owing to the hydrophilic nature of dextran without crosslinking, the molecular chains of dextran are not stable within the graft and water permeability is greatly increased (22.54 ± 0.37 ml/ min per cm2 ). This latter property further increased with time, because of the continual release of dextran. As STMP has a low toxicity with no adverse effects in humans, its employment in this process is preferable to the use of glutaraldehyde or other cross-linking agents [25]. Many studies have indicated that coated grafts randomly cross-linked with glutaraldehyde create a surface that is subject to calcification [26] and may induce potential cytotoxicity caused by the presence of degradation products [27]. Another advantage of using STMP is that the cross-linking reaction can be performed in an aqueous solution, which simplifies the experimental procedure. This study clearly demonstrates that dextran-impregnated grafts cross-linked with STMP possess satisfactory handling characteristics. Vascular prostheses manufactured utilizing this procedure are impervious and present a suitable level of softness and flexibility comparable to the characteristics of the Dialine II® prosthesis. This process can also be used to coat other types of prostheses. We have conducted preliminary investigations on the response of endothelial cells to the vascular prosthesis impregnated with dextran cross-linked with STMP. In this preliminary study, we investigated the behavior of endothelial cells on polyester prostheses impregnated with cross-linked dextran. These grafts appear to promote the growth of endothelial cells, as demonstrated by SEM observations. Further detailed and extensive studies must now be carried out to evaluate the performance of the dextran-impregnated vascular graft and to assess cell growth when employing adult human endothelial cells. Acknowledgements The Centre National de la Recherche Scientifique (CNRS), the Agence Nationale de la Valorisation de la Recherche (ANVAR), and Therapol S.A. supported this work. We would like to thank Mr. O. Berteau for his assistance with the computing.
Dextran-impregnated vascular prosthesis
495
REFERENCES 1. A. J. Lusis, Nature 407, 233 (2000). 2. R. Guidoin, M. King, D. Marceau and A. Cardou, J. Biomed. Mater. Res. 21, 65 (1987). 3. A. Eberhart, Z. Zhang, R. Guidoin, G. Laroche, L. Guay, D. De La Faye, M. Batt and M. W. King, J. Biomed. Mater. Res. (Appl. Biomater.) 48, 546 (1999). 4. Y. Marois, R. Guidoin, X. Deng, M. W. King, L. Martin and R. Roy, Ann. Vasc. Surg. 11, 133 (1997). 5. R. Guidoin, D. Marceau, J. Couture, T. J. Rao, Y. Merhi, P. E. Roy and D. D. la Faye, Biomaterials 10, 156 (1989). 6. D. Domurado, R. Guidoin, M. Marois, L. Martin, C. Gosselin and J. Awad, J. Bioeng. 2, 79 (1978). 7. K. Kottke-Marchant, J. M. Anderson, Y. Umemura and R. E. Marchant, Biomaterials 10, 147 (1989). 8. J. K. Drury, T. R. Ashton, J. D. Cunningham, R. Maini and J. G. Pollock, Ann. Vasc. Surg. 1, 542 (1987). 9. Y. Marois, N. Chakfé, X. Deng, M. Marois, T. How, M. W. King and R. Guidoin, Biomaterials 16, 1131 (1995). 10. A. N. de Belder, in: Polysaccharides in Medical Applications, S. Dimitriu (Ed.), p. 505. Marcel Dekker, New York (1996). 11. F. Blanquaert, J. L. Saffar, M. L. Colombier, G. Carpentier, D. Barritault and J. P. Caruelle, Bone 17, 499 (1995). 12. J. P. Draye, B. Delaey, A. Van de Voorde, A. Van Den Bulcke, B. Bogdanov and E. Schacht, Biomaterials 19, 99 (1998). 13. J. P. Draye, B. Delaey, A. Van de Voorde, A. Van Den Bulcke, B. De Reu and E. Schacht, Biomaterials 19, 1677 (1998). 14. H. C. Chiu, G. H. Hsiue, Y. P. Lee and L. W. Huang, J. Biomater. Sci. Polymer Edn 10, 591 (1999). 15. W. E. Hennink, O. Franssen, W. N. E. Van Dijk-Wolthius and H. Talsma, J. Control. Release 48, 107 (1997). 16. C. J. S. Edgell, C. C. McDonald and J. B. Graham, Proc. Natl. Acad. Sci. USA 80, 373 (1983). 17. C. Bonissol and B. Stoiljkovic, Res. Virol. 140, 241 (1989). 18. W. Nierle, Starch/ Stärke 21, 13 (1969). 19. W. M. Kulicke, Y. A. Aggour and M. Z. Elsabee, Starch/ Stärke 42, 134 (1990). 20. F. Carrasco, E. Chornet, R. P. Overend and J. Costa, J. Appl. Polym. Sci. 37, 2087 (1989). 21. S. P. Massia, J. Stark and D. S. Letbetter, Biomaterials 21, 2253 (2000). 22. S. P. Massia and J. Stark, J. Biomed. Mater. Res. 56, 390 (2001). 23. R. E. Marchant, S. Yuan and G. Szakalas-Gratzl, J. Biomater. Sci. Polymer Edn 6, 549 (1994). 24. E. Osterberg, K. Bergstrom, K. Holmberg, J. A. Riggs, J. M. Vanalstine, T. P. Schuman, N. L. Burns and J. M. Harris, Colloids Surfaces A — Physicochem. Eng. Aspects 77, 159 (1993). 25. K. Woo and P. A. Seib, Carbohyd. Polym. 33, 263 (1997). 26. M. Grabenwoger, P. Bock, F. Fitzal, A. Schmidberger, M. Grimm, G. Laufer, H. Bergmeister and E. Wolner, J. Heart Valve Dis. 7, 229 (1998). 27. L. L. H. Hung-Lee, D. T. Cheung and M. E. Nimmi, J. Biomed. Mater. Res. 24, 1185 (1991).
This page intentionally left blank