This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
P b ra rb > > : r a r b
assumptions on probabilities, but a different expected profits formulation, first order conditions may not necessarily characterize the optimal solutions. Again, to simplify the exposition (without losing much generality, however), throughout the article i shall assume that the first-order conditions identify the expected profit-maximizing positive solutions. For an illustration of the above analysis, consider p(ra) = 1 eara and q(rb) = 1 ebrb.
with probability pðr a Þqðr b Þ with probability pðr a Þð1 qðr b ÞÞ with probability ð1 pðr a ÞÞqðr b Þ with probability ð1 pðr a ÞÞð1 qðr b ÞÞ
Assuming the firm to pursue expected profit maximization, the R&D investment levels maximizing the firm expected profit EP(ra, rb) is obtained by solving the following problem:
which, rearranging terms, becomes
Because these two functional forms are the same, in this simple case the intrinsic difficulty of the two drug discovery projects is formalized only by the parameters a and b. In particular, the higher their value, the easier the discovery is. And, indeed, conditions for positive R&D investments in project a and b are given, respectively, by a > (1/Pa) and b > (1/Pb). Assuming they are met, (3a) and (3b) now become
Maxra ;rb EPðr a ; r b Þ
aeara ¼
1 ; Pa
(4a)
bebrb ¼
1 Pb
(4b)
Maxra ;rb EPðr a ; r b Þ ¼ Maxra ;r b ½ pðr a Þqðr b ÞðP a þ P b Þ þ P a pðr a Þð1 qðr b ÞÞ þ P b qðr b Þð1 pðr a ÞÞ r a r b
(1)
¼ Maxra ;r b ½ pðr a ÞP a þ qðr b ÞP b r a r b (2) The above shows that because the two projects are independent, the optimal R&D solutions for the portfolio coincide with the optimal solutions of the two projects taken separately. Given the assumptions on success probabilities, the firm would find it profitable to invest a positive amount of resources in project a if p0 (0)Pa > 1 and in project b if q0 (0)Pb > 1. Having supposed that project a is the one qualifying for PRV, if p0 (0)Pa < 1 then no R&D investment would be made in such project and the related disease be neglected. Assume instead that both conditions for positive investment are met, the optimal R&D effort levels for the two projects solve the following (first-order) optimality conditions. p0 ðr a Þ ¼
1 ; Pa
(3a)
q0 ðr b Þ ¼
1 Pb
(3b)
It is worth noticing that the assumptions behind p(ra) and q(rb), as well as the form of the expected profit, guarantee that the firstorder conditions (3a) and (3b) identify the positive R&D solutions to (2). With the same
from which, as optimal R&D investment levels, i obtain ra = (log(a Pa)/a) and rb = (log(b Pb)/b), which, in fact, are positive when the above inequalities hold. R&D will be higher in project a, namely ra > rb, if (log(aPa)/a) > (log(bPb)/b), hence if a > ((bPb)(a/b)/Pa). To summarize, the example suggests that it is the shape of the success probabilities that determines the relation between the intrinsic difficulty of a and b and their prospective profits, which will define which of the two projects receives a higher amount of R&D resources.
The model with PRVs Suppose now that PRVs are available; below, i formalize how, in my view, PRVs mainly operate. I have already discussed that PRVs can be meaningful drivers stimulating R&D if they can act as economic incentives (namely, if they can increase the prospective profits of a drug discovery project for infectious diseases). That is, a firm owning a voucher can use it to raise its profits, but whether this could effectively occur – and, if so, to what extent – might be uncertain
when the company decides its R&D investment level. More specifically, the economic value of a PRV can come from two possible sources. First, the voucher could be sold to another company, but if it will be sold and at what price might be uncertain ex ante. This possibility could also include the case of trading the PRV. Alternatively, the holder can use the voucher to prioritize the review of another compound in its portfolio, as long as there is such compound and if its prioritization can generate additional profits. In the next section, i introduce PRVs in the model discussed in the previous section and, in a simple way, formalize the two possible channels through which PRVs could increase the firm profits and act as an economic incentive.
PRVs as an incentive to R&D effort for NDs Having assumed project a to be the one eligible for a PRV, below i formalize its potential economic value. (i) With no main loss of generality i model the market value of the voucher, and related uncertainty, simply by assuming that it could be sold to another company either at a high price, VH, with probability 0 d 1, or at a low price, VL, with probability (1 d), where VH > VL 0. The case of VL = 0 would capture the possibility of the voucher remaining unsold. Therefore, EV = dVH + (1 d)VL > 0 is the expected market value of the voucher. Notice that these two prices could also be interpreted as the highest and the lowest price, respectively, at which the holder could sell the PRV. (ii) If not sold to another company, the voucher could be used to prioritize project b, for which the associated profits would increase to lPb, with l > 1. It follows that prioritizing the review of b would increase the firm profits by (l 1)Pb. To have a proper understanding of how PRVs could operate as economic incentives, in what follows i consider three cases separately: for convenience, (i) will be called external and (ii) internal incentives to be awarded a voucher.
VH > VL (l – 1)Pb > 0 I start with the case of a weak internal incentive, namely a situation in which the additional prospective returns coming from prioritizing project b, already in the holder portfolio, are below the two possible market prices of the voucher. In this case, the firm profits generated by the portfolio of compounds are www.drugdiscoverytoday.com
889
Features PERSPECTIVE
has an overall budget, B, allocated to R&D, but with no major loss of generality, i shall assume that B will always be high enough to cover the chosen R&D investment levels, namely ra + rb B.
PERSPECTIVE
PERSPECTIVE
8 P a þ Pb þ V H ra rb > > > > > > < P a þ Pb þ V L ra rb Pðr a ; r b Þ ¼ P a þ EV r a r b > > > P b ra rb > > > : r a r b
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
with probability pðr a Þqðr b Þd with probability pðr a Þqðr b Þð1 dÞ with probability pðr a Þð1 qðr b ÞÞ with probability ð1 pðr a ÞÞqðr b Þ with probability ð1 pðr a ÞÞð1 qðr b ÞÞ
and the optimal R&D effort levels obtained by solving
p0 ðr a Þ¼
1 ; pðr a Þð1dÞ½ðl1ÞP b V L þP a þ EV (8a)
Maxra ;rb EPðr a ; r b Þ ¼ Maxra ;r b pðr a Þ½P a þ EV qðr b ÞP b r a rb
(5)
from which the following first-order conditions obtain 1 p0 ðr a Þ ¼ ; P a þ EV
(6a)
1 Pb
(6b)
q0 ðr b Þ ¼
Features PERSPECTIVE
Because EV > 0, from (6a) it follows immediately that with the PRV the R&D investment in project a now, if positive, increases, whereas (6b) coincides with (2b) and so rb remains unaltered. Intuitively, this is because the internal incentive is dominated by the external incentive, which induces changes only in R&D investment for project a. In this case, the projects remain independent, as when PRVs are not available. The above consideration, however, does not guarantee positive expected profits in project a, and so a positive ra, if the related disease was neglected before PRVs became available. Indeed, if p0 (0)Pa < 1, the condition for positive investment would now be p0 (0)(Pa + EV) > 1, which could only be satisfied if EV is high enough (i.e. if the external market incentive is sufficiently strong).
q0 ðr b Þ ¼
1 pðr a Þð1 dÞ½ðl 1ÞP b V L þ P b (8b)
Because [(l 1)Pb VL] 0 from (8a) i obtain that ra solving the equation now is, in general, larger than the solution to (6a). Moreover, for the same reason, the rb solving (8b) is higher than the one solving (6b). To summarize, if when neither the internal nor the external incentive dominates, the R&D level in both projects could potentially increase with respect to when the external, market incentive is dominant.
(l 1)Pb VH > VL > 0 Finally, when the internal incentive is the strongest, prospective profits are as follows: 8 P a þ lP b r a r b > > > > > > < P a þ lP b r a r b Pðr a ; r b Þ ¼ P a þ EV r a r b > > > P b ra rb > > > : r a r b
and the optimal R&D effort obtained by solving Maxra ;rb EPðr a ; r b Þ þ P a þ EVg þ qðr b ÞP b r a r b
In this case, the internal incentive is not dominated by the external incentive, and the firm prospective profits are now 8 P a þ Pb þ V H ra rb > > > > > > < P a þ lP b r a r b Pðr a ; r b Þ ¼ P a þ EV r a r b > > > P b ra rb > > > : r a r b
Maxra ;rb EPðr a ; r b Þ
With positive R&D effort levels, optimality conditions would now become
p0 ðr a Þ ¼
1 ; qðr b Þ½ðl 1ÞP b EV þ P a þ EV (10a)
¼ Maxra ;r b pðr a Þfqðr b Þð1 dÞ½ðl 1ÞP b V L þ P a þ EVg þ qðr b ÞP b r a r b (7) Having assumed first-order conditions to characterize optimality, solutions to (7) are given by 890
www.drugdiscoverytoday.com
(9)
with probability pðr a Þqðr b Þd with probability pðr a Þqðr b Þð1 dÞ with probability pðr a Þð1 qðr b ÞÞ with probability ð1 pðr a ÞÞqðr b Þ with probability ð1 pðr a ÞÞð1 qðr b ÞÞ
so that (5) now becomes
q0 ðr b Þ ¼
Concluding remarks Within a simple economic model, this article is an endeavor toward a better understanding of how effective PRVs can be as incentives, stimulating R&D investments for drugs treating NDs. The analysis suggests that, under some general conditions, PRVs tend to increase R&D efforts, notably if a firm obtaining a voucher has
with probability pðr a Þqðr b Þd with probability pðr a Þqðr b Þð1 dÞ with probability pðr a Þð1 qðr b ÞÞ with probability ð1 pðr a ÞÞqðr b Þ with probability ð1 pðr a ÞÞð1 qðr b ÞÞ
¼ Maxra ;r b pðr a Þfqðr b Þ½ðl 1ÞP b EV
VH > (l 1)Pb VL > 0
being when (l 1)P = VH and d = 1. Moreover, because (l 1)Pb VH implies [(l 1)Pb EV] (1 d)[(l 1)Pb VL], it follows that the R&D solution to (10a) and (10b) are in general larger than the solutions to (8a) and (8b). I summarize the above findings in the following proposition: PRVs tend to increase R&D investment for drugs treating NDs. Such an increase would be larger whenever a firm has as a sufficiently strong internal incentive, as compared to the external, market incentive, to obtain the voucher. When this is so, PRVs will also tend to increase R&D investments for the internal compound prioritized by the voucher, which would not occur when the internal incentive is weak.
1 pðr a Þ½ðl 1ÞP b EV þ P b
(10b)
In this case, too, the solution to both (10a) and (10b) in general contemplates higher R&D effort levels than in (6a) and (6b), the only exception
in its portfolio a particularly valuable compound to prioritize. When this is so the bearer is, in some sense, outperforming the market because the value it can create internally, by prioritizing a drug, is higher than what it could obtain by selling the voucher. Such a finding seems to capture the spirit behind PRV, of linking R&D efforts for NDs with those for non-NDs. As a first attempt to formalize the economics of PRV, the article leaves open to future research few issues. In particular it abstracts from strategic, and asymmetric information, considerations that in more articulated versions of the model might have a role. For example, the strategic use of private information might prevent an efficient exchange from taking place when the bearer assigns a low value to the voucher but tries to sell it at a high price, whereas the interested buyer might assign a high value to the voucher but try to purchase it at a low price. References 1 Amendments, F.D.A. Act of 2007 (2007) (Public Law 110-85)
2 FDA (2008) Guidance for Industry. Tropical Disease Priority Review Vouchers 3 Waltz, E. (2008) FDA launches priority vouchers for neglected disease drugs. Nat. Biotechnol. 26, 1315– 1316 4 Ridley, D.B. et al. (2006) Developing drugs for developing countries. Health Aff. 25, 313–324 5 Flanagan, M. and Writer, S. (2008) Defining the priority review marketplace. In BioCentury. October 27, A12 6 Noor, W. (2009) Placing value on FDA’s priority review vouchers. In Vivo 9, 27 7 Grabowski, H. et al. (2009) Priority review vouchers to encourage innovation for neglected diseases. In Prescribing Cultures and Pharmaceutical Policy in the Asia-Pacific (Eggleston, K., ed.), Brookings Institution Press
8 Kremer, M. and Glennerster, R. (2006) Strong Medicine. Princeton University Press 9 Kesselheim, A.S. (2008) FDA review vouchers. Reply. N. Engl. J. Med. 360, 837–838 10 Kesselheim, A.S. (2008) Drug development for neglected diseases. The trouble with priority review vouchers. N. Engl. J. Med. 359, 1981–1983 11 Kesselheim, A.S. (2009) Priority review vouchers: an inefficient and dangerous way to promote neglecteddisease drug development. Clin. Pharmacol. Ther. 85, 573–575 12 Sonderholm, J. (2009) In defense of priority review vouchers. Bioethics 23, 413–420 13 Carpenter, D. (2008) Drug-review deadlines and safety problems. Reply. N. Engl. J. Med. 359, 95–98 14 Carpenter, D. et al. (2008) Drug-review deadlines and safety problems. N. Engl. J. Med. 358, 1354–1361
PERSPECTIVE
15 Wilson, J.F. (2006) Alterations in processes and priorities needed for new drug development. Ann. Intern. Med. 145, 793–796 16 Grabowski, H. and Wang, R. (2008) Do faster FDA drug reviews adversely affect patient safety? An analysis of the 1992 Prescription Drug User Fee Act. J. Law Econ. 51, 377–406 17 Nardinelli, C. et al. (2008) Drug-review deadlines and safety problems. N. Engl. J. Med. 359, 95–98 18 Moe, J. et al. (2009) FDA review vouchers. N. Engl. J. Med. 360, 837–838
Nicola Dimitri 1,2 1 Faculty of Economics, University of Siena, 53100, Italy 2 Visiting Professor IMT Lucca, 55100, Italy [email protected]
Features PERSPECTIVE
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
www.drugdiscoverytoday.com
891
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
‘Study the past – with a critical eye – if you would define the future.’ Confucius (modified).
Reviews KEYNOTE REVIEW
The role of Ca2+ in ultrasound-elicited bioeffects: progress, perspectives and prospects Mariame A. Hassan1,2, Paul Campbell3 and Takashi Kondo1 1 Department of Radiological Sciences, Graduate School of Medicine and Pharmaceutical Sciences, University of Toyama, Sugitani 2630, Toyama 930-0194, Japan 2 Department of Pharmaceutics and Industrial Pharmacy, Faculty of Pharmacy, Cairo University, Kasr Al-Aini Street, Cairo 11562, Egypt 3 Carnegie Physics Laboratory and Division of Molecular Medicine, University of Dundee, Dundee DD1 4HN, Scotland, United Kingdom
Intracellular calcium (Ca2+) transients have been observed in association with exposure to therapeutic ultrasound and correlated to both early- and late-onset bioeffects. For example, it has been suggested that early ‘ultrashort’ Ca2+ transients recorded during sonoporation can mediate Ca2+dependent exocytosis and endocytosis processes as complementary mechanisms for membrane self-sealing. Moreover, apoptosis induction has been reported to occur through a partial mediation of a Ca2+dependent pathway. In this review, we attempt to assemble the salient facts into a cogent whole, with special attention given to the relationships arising through altered Ca2+ levels, which underscore its crucial role during ultrasonic interactions with biological systems and its consequent implications in the context of therapeutics. Since the first report on the biological effects of ultrasound in 1927 [1], the concept of therapeutic ultrasound (TUS) has been consolidated considerably and extended in versatility across a wide range of applications. Recent exciting demonstrations – for example, in accelerated bone fracture healing and wound healing – underscore this [2–4]. Moreover, the facility for direct lysing of cells and induction of apoptosis [5,6], together with adjuvant qualities gained by the combination with chemotherapeutic [7,8] and thrombolytic [9,10] agents, also serve to highlight the promise of this approach once its full potential is reached. Such successes paint TUS in a positive light; however, this belies the fact that our understanding of the fundamental processes giving rise to both real-time and downstream effects is somewhat incomplete and, in some instances, rather confused. That the interaction between TUS and biological systems has been shown to produce a wide range of often contrary effects is perhaps not surprising given the complexity and the multitude of parallel processes involved [11,12]. The dominance of any individual effect in relation to another depends on many variables, both acoustic and non-acoustic, which makes inter-comparisons difficult and might actually stymie translational research [13]. Therefore, to
Dr Paul Campbell read Physics as an undergraduate at the University of London – Queen Mary College, before taking a PhD in Experimental Physics at Queen’s University Belfast. He is presently a reader in Physics at the Carnegie Physics Laboratory at the University of Dundee, and also deputy head of the Division of Molecular Medicine. His input to the present research collaboration was facilitated through a Japan Society for the Promotion of Science (JSPS) Fellowship, and a Royal Society International Collaboration Award. His current research programme is centred on understanding the fundamental microscopic interactions of ultrasound waves with biological cells and tissues, with the ultimate goal of achieving reliable non-invasive drug delivery. His work is supported by a Medical Research Council Fellowship, EPSRC Translational Technology Award, and a MRC Milstein Award. Dr Campbell is a fellow of the UK Institute of Physics (FInstP) and also holds a Royal Society Industry Fellowship.
Takashi Kondo received his PhD from Hokkaido University, Sapporo, Japan in 1980. He became a member of the Department of Experimental Radiology and Health Physics, Fukui Medical School since 1981. He was an International Research Fellow in the Radiation Oncology Branch, NCI, NIH, USA from 1986 to 1989. He joined the Department of Radiation Biophysics, Kobe University, School of Medicine since 1993 as an assistant professor and, finally moved to the Department of Radiological Sciences, Graduate School of Medicine and Pharmaceutical Sciences, University of Toyama, Japan, as a professor and chairman in 1997. His current research interest includes mechanistic studies related to biological effects of ultrasound and ionizing radiation. His scientific contributions are estimated to be over 100 original papers and review articles.
Corresponding author:. Kondo, T. ([email protected])
892
www.drugdiscoverytoday.com
1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.08.005
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
TABLE 1
Cell line
Exposure system
Transducer
Transducer position in relation to cells
Central frequency (MHz)
Acoustic pressure (MPa)
Sonication period(s)
Contrast agent
Refs
Xenopus oocyte
Special chamber
Planar
Beneath (2 cm away)
0.96
0.24–1.2
0.1–1.0
Optison
[20]a
U937
Rotating tube (30 rpm)
Planar
Aside (20 cm away)
1.0
0.6
60
Levovist
[21,22]
CHO
Culture dish
Planar
Above (4 mm away)
1.0
0.30–0.45
0.2
Optison
[25]a
DU 145
Special chamber
Cylindrical planar
Beneath and sides (the axial and radial centre)
24 kHz
0.7
20
_
[26]
MCF-7
Rotating tube (60 rpm)
Focused
Beneath (the focal length)
1.0
0.19–0.48
40
+
[27]
BAEC
OptiCell
Planar
Beneath at an angle of 458
1.0
0.22
30
SonoVue
[28]
Xenopus oocyte
Culture dish
Planar
Beneath
1.075
0.3
0.2
Definity
[33]a
MAT B III
Rotating tube (60 rpm)
(a) Air-backed (b) Focused
Beneath (7.6 cm away)
(a) 1.15 (b) 2.25
(a) 402 kPa (b) 570 kPa
10
+
[34]
BHK-21
Culture plates
Planar
Above
1.0
0.045–0.159
10–40 min
Optison
[36]
H9c2
OptiCell
Planar
Beneath at an angle of 458 (7 mm away)
1.0
0.27
10–15 ms
Definity
[42]a
RPMI 1788
Rotating tube (200 rpm)
Planar
Beneath (7.5 cm away)
1.0
0.5
30
–
[60]
DU 145
Special chamber
Focused
Beneath (the focal length)
500 kHz
0.6, 1.6, 2.4, 3.0
2, 9 and 34
Optison
[61]
HUVEC
Culture dishes
Focused
Above (5 cm away)
1.6
0.6–1.515
30 min
–
[66]
a
Real-time analysis. Abbreviations: BAEC, bovine aortic endothelial cells; BHK-21, baby hamster kidney cells; CHO, Chinese hamster ovary cells; DU 145, human prostate cancer cells; HUVEC, human umbilical vein endothelial cells; H9c2, rat cardiomyoblast cells; MAT B III, rat mammary carcinoma cells; MCF-7, human breast cancer cells; RPMI 1788, human peripheral lymphocytes; U937, human myelo-monocytic lymphoma cells.
promote TUS and assist progress towards clinical acceptance and application, it has become a necessary priority to understand the detailed interactions at a fundamental level. In addressing this, two distinct routes seem to be favoured: investigating post-sonication effects via a retrospective analysis, usually on a population of cells or extended tissue, or conducting real-time imaging at the single-cell level for recording in situ changes, which can then be correlated with physical cause. Studies using the former approach have given rise to the ‘sonoporation’ hypothesis [14]. The vast technological strides that have been taken in the past few decades, however, have seen the emergence of real-time sensing systems, such as the patch-clamp technique, affordable (ultra) high-speed cameras and other innovations that have facilitated a significantly enhanced capability to probe TUS-driven bioeffects. The main objective of this review is to assemble a cogent and comparable snapshot of the existing hypotheses, especially in relation to their time-dependent transients in intracellular Ca2+ concentrations, in an attempt to draw these into a logically consistent and more unified view. Calcium ions (Ca2+) have pivotal roles in living cells and are key regulators of cell proliferation and cell death [15]. A key requirement for the regulation of cellular functions by cytosolic Ca2+ is to maintain a steep concentration gradient between the extracellular and intracellular environments [16]. In addition, within the intracellular space, a further Ca2+ gradient is established between the cytosol and other organelles such as the endoplasmic reticulum (ER) and mitochondria [17,18]. Any change in this balance affects Ca2+ homeostasis and can ultimately affect the fate of the cell [19].
Table 1 summarizes – and, indeed, underscores – several key studies that highlighted some specific roles for Ca2+ dynamics in the context of TUS-induced bioeffects.
Pioneering studies on Ca2+-dependent TUS bioeffects A seminal early report highlighting the involvement of Ca2+ in TUS bioeffects occurred in 2004 when Deng et al. achieved the first parallel electrophysiology measurements on a patch-clamped Xenopus oocyte during irradiation with tone-burst ultrasound [20]. There, it was observed that a slightly delayed inward electric current developed, which proceeded to increase in magnitude in a step-wise manner during the sonication procedure, whereupon it returned to control levels upon the termination of the ultrasound irradiation (Fig. 1a). This observation was suggestive of an enhanced cell permeability facilitating a transmembrane ion flux. Moreover, the effect was strongly related to bubble activity because it occurred only in the presence of Optison (5%). Whether acoustic streaming – leading to threshold local shear stresses with the potential to disrupt the membrane – or an alternative mechanism was the cause of permeability could not be discriminated with confidence. It was certainly indicated, however, that the increased ionic permeability was due to the opening of non-specific pores rather than endogenous voltage- or ligand-gated ion channels. Interestingly, the sonoporation process, as inferred from the enhanced electrical current, was irreversible if Ca2+-free buffer was used (Fig. 1b) or if the direction of current was reversed so that little Ca2+ could enter the cell from outside (Fig. 1c). This lack www.drugdiscoverytoday.com
893
Reviews KEYNOTE REVIEW
Summary of TUS setups used in major studies included in the review
REVIEWS
[(Figure_1)TD$IG]
Reviews KEYNOTE REVIEW FIGURE 1
Ionic currents occurring across the membrane of a single Xenopus oocyte during and after ultrasound exposure (tone burst, 0.29 MPa, 0.5 s duration) in the presence of 5% Optison (109 microbubbles/ml). The oocyte bathed in (a) ND96 solution (in mM: 96 NaCl, 2 KCl, 1.8 CaCl2, 1 MgCl2, 5 HEPES, pH 7.6) and (b) Ca2+-free ND96 (in mM: 96 NaCl, 2 KCl, 1 MgCl2, 5 EGTA, 5 HEPES, pH 7.6) and the membrane potential was clamped at 50 mV. Repetitive lines indicate current traces recorded from the same oocyte under repeated exposures. The numbers from 1 to 4 in (b) give the sequence of recordings. (c) Ionic currents when the membrane potential was clamped at either 50 or +50 mV (tone burst, 0.29 MPa, 1 s duration). The numbers 1–5 indicate the sequence of recordings. The membrane potential at each recording was: 1, 50 mV; 2 and 3, consecutive recordings at +50 mV; and 4 and 5, consecutive recordings at 50 mV. Inward current decay was only observed in the presence of extracellular Ca2+, whereas the current decay, and hence membrane sealing, failed when the oocyte was bathed in Ca2+-free ND96 or when the membrane potential was clamped at +50 mV to reverse current direction and prevent Ca2+ entry to the cell through pores. Reproduced, with permission, from Ref. [20].
of current decay was suggestive of a role for Ca2+ in membrane resealing immediately after TUS exposure, a very early bioeffect. Cases in which the pressure amplitude was higher (>1 MPa) or the insonation period was longer (>0.5 s) also led to prolonged currents, and this was assumed to be due to irreversible damage to the cell membrane leading to cell death even in the presence of Ca2+. Around this time, Honda et al. [21] published a study showing that TUS was able to induce apoptosis in myelo-monocytic lymphoma U937 cells. Apoptosis induction was shown to be strongly related to inertial cavitation and synergistically increased in the presence of microbubbles; however, there seemed to be no correlation between apoptosis and the amount of free radicals generated extracellularly during sonication [22]. Here, a transient increase in intracellular Ca2+ was observed during four hours of immediate post-exposure monitoring, after which the Ca2+ levels returned to normal levels. In experiments conducted without extracellular Ca2+, no such increase was observed. Based upon these findings, the authors concluded that the TUS-induced apoptosis was regulated, in part, through a Ca2+-dependent pathway and the transient increase in intracellular Ca2+ was caused by an extracellular Ca2+ influx occurring again through non-specific membrane pores created during sonication. This was further supported by the inability of Verapamil, a well-known Ca2+ channel 894
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
blocker, to suppress increases in intracellular Ca2+ concentrations when added before TUS exposure. After these two studies, a flurry of further reports were published supporting the occurrence of TUS-associated Ca2+ changes [23–25]; however, these particular studies offered little conceptual insight into the fundamental interactions arising between TUS and living cells in terms of the consequent biological processes that were triggered and/or controlled by Ca2+ levels. The first study embracing this objective was conducted by Schlicher et al. [26] using a combination of real-time imaging and advanced microscopies. The group reported that TUS induces ‘micron-scale wounds in plasma membrane that reseal using intracellular vesicles by an energy-intensive process requiring Ca2+’ [26]. Subsequently, in late 2008 and early 2009, two very complementary studies appeared, which provided a clearer perspective about the Ca2+-mediated responses occurring in the earliest stages of insonation and their implication in molecular delivery [27,28]. Before discussing these important contributions, however, it is prudent to consider the state of salient knowledge in the context of mammalian cells.
What happens when cells are exposed to TUS? Ultrasound is a mechanical wave that manifests as a series of pressure fluctuations that transmit through a body under freefield conditions or can simply establish a spatial standing wave pattern if reflecting constraints are present [29]. The effects of ultrasound are well known to be amplified when microbubbles are present in solution, either in the form of shelled micrometer-scale contrast agents (as is often the case with modern sonoporation studies) or through the formation of natural bubbles (cavitation). Any bioeffects arising, therefore, will have a direct correlation to the ultrasound parameters used, the location of the affected cells relative to the energy source (transducer) or to any nodal structures (when standing waves are present), and the proximity to acoustically active bubbles, amongst other possible contributants. The possibility of membrane disruptions occurring during insonation has been inferred from many observations of cellular deformation, which also correlate with the uptake of otherwise impermeable dyes within the deformed cells [30], and through measured changes in membrane electrophysiology [20,31–33]. The occurrence of actual physical pores has been confirmed in several other studies that used scanning electron microscopy and atomic force microscopy [34–37]. Pores are thought to arise via the occurrence of excessive shear stresses on the membrane caused by acoustic streaming. Microjetting and other more exotic events can also contribute to the overall state of permeabilization [38]. Perhaps the most important aspect of these pores is their ability to pass, non-specifically, extracellular molecules that would be otherwise impermeable to the cell under normal conditions. Logically, larger pores would be expected to lead to a more ready passage of species across the membrane compared with smaller counterparts; however, this figure of merit (from a drug delivery standpoint) would become moot if the cell were not able to repair itself within some finite timescale whereupon damage were seen as lethal. Interestingly, mammalian cells are known to tolerate pores of the order of 1000 mm2 [39]. If we consider that gross cellular ‘wear and tear’ – for example, on load-bearing tissue structures – is a daily occurrence in humans, it might be anticipated that specialized pathways
for the repair of damaged cells would exist and the mechanisms would operate without prejudice or consideration of the actual physical cause of the damage as (fast) repair becomes a biological imperative for cell survival. This hypothesis was first proven by Saito et al. [40] and studied further by Schlicher et al. [26], who showed that after TUS exposure, ‘cells actively reseal these holes using a native healing response’ within a short period ‘similar to the kinetics of membrane repair after mechanical wounding’ with indication to the role of Ca2+. In membrane repair processes, the pore size dictates the mechanism of resealing [41]. Thus, small pores might be sealed passively, an inherent quality of lipid bilayers that is favoured energetically, whereas the sealing of larger pores requires more complex processes, provided that the extracellular fluid contains Ca2+ at near-physiological levels. In such conditions, Ca2+ is driven into the cell – amongst other molecules – by the concentration gradient [42]. Proximal bubble activity, stimulated by ultrasound, has been implicated in this process. For instance, the intracellular Ca2+ increase has been shown to start at the cell side closer to a bubble located just before irradiation and was presumed either to dislocate or ‘to collapse’ after exposure [25,42] (Fig. 2a). The influx of Ca2+ would be expected to trigger Ca2+-responsive proteins within the cell and to react with the underlying cytoskeleton, perhaps causing depolymerization of the filaments. This depolymerization reaction would be important to free the way for internal vesicles (lysosomes, which are normally subcortical organelles) to approach the disruption site. Lysosomes can be stimulated by a high Ca2+ concentration to fuse with the membrane [43], as well as with each other (homotypic fusion) [44] so they can adequately patch the site of disruption (the ‘patch hypothesis’). This Ca2+-regulated exocytosis mediated by lysosomes was shown to be prevalent in many cell types [45]. Furthermore, its role was clearly evidenced through immunostaining with antibodies against the luminal domain of the lysosomespecific protein, LAMP-1, in mechanically injured cells [43], as well as TUS-treated cells [27] (Fig. 2b,c).
Alternative pathways For the sake of completeness, it is worth mentioning the alternative scenarios for membrane self-sealing. It has been suggested that the cytoplasm can build immediate barriers by itself in the presence of elevated Ca2+ concentrations. These barriers act to hinder diffusion of extracellular fluid and/or to guard against the loss of cytoplasm through the disrupted membrane, and they are mostly regulated by non-exocytotic Ca2+-dependent pathways [46]. Moreover, it has been hypothesized that elevated Ca2+ concentrations can activate a family of cytoplasmic enzymes known as tissue transglutaminases (TGase) that have the capability to crosslink proteins – especially those of the extracellular matrix, a pathway implicated in wound healing [47–50] – thus forming intracellular ‘clots’. However, whether these pathways are implicated in TUS has not yet been investigated.
Types of Ca2+ transients Once the membrane disruption is patched, it is expected that Ca2+ can no longer enter the cell non-specifically. Researchers have been able to show that the Ca2+ transients observed in cells exposed to mechanical injury or TUS application are terminated within approximately 3 min [25,51–53]. This time period might be
REVIEWS
sufficient to kill the cell if the increased permeability persisted throughout [41]. Rather, the cell could patch the membrane disruptions in much shorter periods (5–20 s is the reported time range taken for the transmembrane potential to return to its preexposure value [33], and 3–8 s is the interval during which the Ca2+ influx peaks before recovery starts [25]). To be clear, the figure of 3 min represents the time required for the re-establishment of Ca2+ homeostasis, and this Ca2+ influx defines the first ‘ultra-short’ Ca2+ transient experienced by cells exposed to TUS, which is different from the ‘short’ transient noticed by Honda et al. [21] during the first six hours post-sonication. At this juncture, it is natural to ask why the increase in intracellular Ca2+ noticed by Kumon et al. [25], despite being localized in the beginning, later transpired to be in a diffuse state within the cytoplasm (Fig. 2a). Outwardly, logic might suggest that the existing large Ca2+ concentration gradient between the extracellular and intracellular environments will facilitate an appreciable ion flux as long as the disruption is sustained during and immediately after TUS exposure, a unique feature of ultrasound treatment. Acoustic streaming, either owing to the acoustic beam or from cavitational activity, could also augment the ion flux through convection. In the present context – that is, with direct comparison to the paper by Deng’s group [20] – this seems not to be the case because the acoustic pulse took only 0.2 s, after which the localized transient could still be discernible. It was not until 1.8 s post-sonication that intracellular homogenous fluorescence was observed. For this particular cell experiencing an immediate transient (i.e. immediate membrane disruption), this phenomenon might be justified by the persistence of an open pore for a short period until complete patching was achieved [26,30,54]; however, this justification cannot work for the surrounding cells, which also encountered delayed Ca2+ transients after TUS termination. According to the authors, the transmission of an inter- and intracellular messenger in response to TUS exposure and/or the ultra-short Ca2+ transient might offer a more plausible explanation. In such a case, the persistence of localization for finite time periods might be in support of the formation of a cytoplasmic barrier [46]. Interestingly, cells after TUS irradiation experienced two modes of ultra-short Ca2+ transients: a Ca2+ transient followed by a monotonous recovery and Ca2+ oscillations [24,25,42]. Ca2+ oscillations – which are responsible, in part, for Ca2+ signalling and the subsequent release of ER Ca2+ stores – are usually produced by the generation of inositol phosphate (IP3) [19], reflecting that the observed Ca2+ transients might be, in part, of an intracellular origin (as discussed below). Because the authors in this study did not identify the fate of these cells in the longer term, the justification of these different transient patterns and their real correlation to TUS-induced effects, such as apoptosis, still warrants dedicated attention. On one hand, there is evidence that TUSinduced Ca2+ oscillations are not essentially indicative of cell death, owing to the diversity of Ca2+ oscillations and the complex processes associated with their encoding [55]. On the other hand, these Ca2+ oscillations could simply reflect sub-threshold responses to trigger apoptosis [56]. In all cases, the presence of intra- and intercellular Ca2+ waves indicates that sealing is necessary but not sufficient to retain viability and that Ca2+ will be a mediator and/or stimulator in the end-stage TUS-induced ultimate responses. www.drugdiscoverytoday.com
895
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_2)TD$IG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews KEYNOTE REVIEW FIGURE 2
(a) Time-lapse ratio images showing calcium waves induced by ultrasound pulse for Chinese hamster ovary (CHO) cells in phosphate buffer saline (PBS) containing Ca2+ ions at a concentration of 0.9 mM. The colour bar indicates the local fura-2 fluorescence ratio R = F340/F380. The labels in the upper right corner of each image list the time since the first image. The images before ultrasound application show variation in baseline fluorescence intensity within different cells. (i), ratio image before ultrasound pulse. The arrow points to the cell that will be immediately affected by the ultrasound pulse. (ii), ratio image at first frame after ultrasound pulse
896
www.drugdiscoverytoday.com
TUS-induced endocytosis If clear evidence on the involvement of lysosomes in pore selfsealing after TUS exists, an obvious question at this point is: ‘where does the lytic cargo of lysosomes go?’ In fact, the lytic enzymes enclosed within the lysosomes are released to the external environment upon the fusion of the vesicle with the membrane. Momentarily, these enzymes are suggested to play a part in facilitating the fusion process. For secretory lysosomes, the release of their luminal contents could participate in several effector functions [57]; however, the unfavourable external pH (7.4) strongly curtails their activity. Moreover, Ca2+-stimulated endocytosis after lysosomal exocytosis helps clear the area from residuals [58,59]. The stimulation of endocytosis in membrane-repair mechanism is better described by Idone et al. [58] as that ‘the discovery of injuryinduced endocytosis provides a new framework for understanding plasma-membrane-repair pathways in mammalian cells’ and that ‘it is no longer clear whether new membrane addition by exocytosis directly mediates plasma membrane resealing, or whether it functions indirectly, by triggering a subsequent endocytotic response that represents the true resealing event’. Following on from this statement, and setting in the context of TUS, the first experiments to prove the concept were conducted by Brayman et al. [60] who noticed the removal of CD19 receptors from the cell surface after sonication. Although the authors did not refer to this phenomenon as endocytosis-mediated, a similar removal of the stable pores formed by the bacterial protein streptolysin O (SLO) from the cell surface was shown to be mediated through endocytosis [58]. Recently, Meijering et al. [28] reported the occurrence of endocytosis post-TUS exposure in endothelial cells. The group found that low molecular weight species (4.4 and 70 kDa fluorescein isothiocyanate [FITC]-dextran) could be pushed into the cell through pores, thus appearing diffuse in the cytoplasm, whereas high molecular weight species (155 and 500 kDa FITC-dextran) were able to traverse the cell membrane only through endocytosis (Fig. 3a). Upon inhibition of the endocytotic pathways, they found that endocytosis played a part, even in the traversal of low molecular weight dextrans. The latter finding is consistent with the sealing scenario because the stimulation of endocytosis during the operation of the sealing machinery will lead to the engulfment of part of the extracellular solution, including all its constituents, regardless of their molecular weights. In fact, inspection of the literature with respect to the dependence of delivery on molecular size reveals that several other reports have found enhanced TUS uptake for species as large as 2000 kDa. The internal dispersion manifested size dependence in terms of the number of molecules taken up [34], the number of cells internalizing the molecules [7,61] and the intracellular distribution [26,61]. It is notable that these studies used different
REVIEWS
acoustic arrangements; thus, exposure to focused beams [34,61] could not eliminate the presence of different pore sizes, including those larger than the diameters of these molecules. Yet according to Guzman et al. [61], the lack of intermediate pore sizes internalizing intermediate molecules, such as bovine serum albumin (66 kDa), suggests that the uptake of larger sizes (probably >60 kDa) is accomplished by a mechanism other than sonoporation that could be a form of post-insonation-induced endocytosis. Careful examination of the confocal fluorescence micrographs (Fig. 3b) reveals spots of increased fluorescence that could represent localization. In addition, the still obvious diffuse pattern seen in the images and the statement by the authors that ‘among HUP (high uptake subpopulation) cells, intracellular macromolecule uptake appears to have reached thermodynamic equilibrium with the extracellular solution for all the molecules studied’ indicates that the internalization in this subpopulation is due to extensive damage of the membrane that, despite being fixed in the period immediately after sonication and therefore endowing the cell with an apparent short-term viability, leads, in the longer term, to cell death, possibly caused by a lack of full re-establishment of homeostasis [62] or perhaps in response to the detection of DNA damage. In conclusion, these recent studies imply that the two hypotheses employed in the interpretation of TUS-enhanced delivery (i.e. sonoporation and endocytosis-mediated delivery) might be occurring as sequential processes at least in some cases. The generalization of this concept still requires further attention.
Is post-insonation endocytosis a reality? In 2010, Cheri Deng and co-workers discovered that not all Ca2+ ultra-short transients were associated with uptake of extracellular matter [42]. Only those that occurred immediately during TUS exposure showed concomitant delivery of the extracellular dye (propidium iodide, or PI), whereas the delayed transients did not (Fig. 4a). They also found that these delivery-associated transients were only correlated to bubble activity in close proximity to respective cells, despite the exposure of the whole field of view to acoustic irradiation. They further negated the presence of any localization of the dye in the permealized cells and thus concluded that endocytotic delivery post-TUS was an invalid route. We believe that although the authors succeeded in proving that Ca2+ transient measurements are not reliable indicators for delivery (i.e. transient permealization) – a noteworthy finding (see later) – their argument about endocytosis seems incomplete. This is because PI cannot be considered as a high or as an intermediate molecular weight species as previously classified [61] (molecular weight 668.4 Da) and because PI is not excluded by viable cells owing to its size but because it is being actively pumped out of the cells with intact membranes [63]. Again, the more proximal any bubble activity, the
(duration 0.2 s). The first significant change is seen in only one cell (arrow). Subsequent ratio images show (iii–v) propagation of an intracellular wave and (vi–x) propagation of intercellular calcium waves originating from this immediately affected cell and locations likely from outside the field of view. Arrows in (ix) indicate examples of locations showing cell-to-cell propagation of the calcium wave, connecting the two calcium transient regions in the image. (xi), dissipation of the calcium waves; (xii), full recovery of all the cells in the field of view. Note that the immediately affected cell also recovers (arrow). Reproduced, with permission, from Ref. [25]. Sections (b) and (c) show immunostaining with antibodies against the luminal domain of lysosome-specific protein, LAMP-1. (b) Control monolayer of human osteosarcoma (U2OS) cells that were not insonated (left) and monolayer subjected to ultrasound (1000 1 ms pulses at 3.2 MPa) (right). Insonated cells show a clearance zone (CZ; dark area to the left): cells near the periphery of the CZ exhibit LAMP-1 staining, whereas those further to the right are much less affected, suggesting that the occurrence of a jetting event that flowed over the cells with sufficient shear force to permealize the cells. (c) Confocal fluorescent imaging of human breast cancer (MCF-7) cells fixed within 60 s of insonation (1000 1 ms pulses at 3.2 MPa) showing the anti-LAMP-1 stain extending within the interior of the cell. Photomicrographs courtesy of Paul Campbell, University of Dundee. www.drugdiscoverytoday.com
897
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_3)TD$IG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews KEYNOTE REVIEW FIGURE 3
Cellular distribution of different fluorescent markers with different molecular weights after ultrasound microbubbles targeted delivery (UTMD) exposure. (a) Confocal laser microscopy images showing the uptake of fluorescein isothiocyanate labelled (FITC)-dextrans (m.wt. 4.4 kDa (ii), 70 kDa (iii), 155 kDa (iv) and 500 kDa (v)) in bovine aortic endothelial cells immediately post-sonication. Homogenous distribution in the cytosol and nucleus can be seen with 4.4 kDa dextran, whereas the 70 kDa dextran was excluded from the nucleus. Larger dextrans were localized in vesicle like structures (arrows) in the cytosol only. Reproduced, with permission, from Ref. [28]. (b) Confocal fluorescence micrographs showing intracellular uptake of calcein (m.wt. 623 Da, i), bovine serum albumin (BSA, m.wt. 66 kDa, ii), FITC-dextran 42 kDa (iii) and FITC-dextran 464 kDa (iv) in human prostate cancer (DU145) cells immediately after TUS exposure. Calcein is distributed throughout the whole cell, whereas BSA and dextrans molecules are excluded from the nucleus. Hoechst nuclear stain (not shown) was used to identify cell nuclei. Figure A1–A3 show the simultaneous presence of three cells having different levels of calcein uptake. A1, the brightly fluorescent cell is indicative of cells in the high uptake subpopulation (HUP); A2, the dimmer fluorescent cell is indicative of low uptake (LUP); and A3, the dimmest cell is indicative of nominal uptake (NUP). Reproduced, with permission, from Ref. [61].
more prone the cells become to poration of the membrane and, thus, the more probable the above-mentioned scenario will be. Similarly, the opposition to the TUS-induced endocytosis-mediated delivery by Schlicher et al. [26] can be argued by the smaller molecular weight of calcein molecules (623 Da) used in the experiments compared to the 150 and 500 kDa dextrans observed in 898
www.drugdiscoverytoday.com
endocytotic vesicles [28], as well as the reliability of the FM1-43 staining test in acoustic treatments, which requires further confirmation [64,65]. It could be, however, that the endocytotic role in the uptake of low molecular weight species was overestimated upon inhibiting the endocytotic activity by Flipin and chlorpromazine [28] owing to their effects on membrane structure, the factor that
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Which endocytotic pathway is involved? Whereas Meijering et al. [28] revealed that the TUS-enhanced delivery for FITC-dextrans stimulated clathrin-mediated endocytosis more than caveolae-mediated endocytosis and macropinocytosis in primary bovine aortic endothelial cells (BAEC), Lionetti et al. [66] found that the exposure to diagnostic-level ultrasound enhanced the uptake of the fluorescent probe for caveolaemediated endocytosis more than that for the clathrin-mediated endocytosis in human umbilical vein endothelial cells (HUVEC). It is still unknown whether this discrepancy is due to the dependence on the delivered substance [67] or due to the duration of sonication, which was 30 s in the former and 30 min in the latter. Furthermore, Basta et al. [68], who conducted their work in the same laboratory under similar conditions, showed that the same later cell line (namely, HUVEC cells) exhibited a DNA ladder after exposure for 15 and 30 min. Do these findings reflect a dependence of the type of endocytosis involved on the cellular response or vice versa, or is it merely the effect of incubation time before ultrasound application? Basta et al. [68] performed TUS immediately after the addition of the fluorescent markers, whereas Lionetti et al. [66] allowed the cells to incubate for 90 min before different protocols of exposure were applied. These questions still require rigorous investigation to achieve a complete clarification.
Waypoints along the route to self-sealing As mentioned earlier, the cell membrane returns to its pre-exposure transmembrane potential in less than 30 s after the insonation burst. In addition, the time in which the Ca2+ influx continues to increase is of a comparable period, after which a recovery process starts to occur; thus, this period is the first waypoint that the cell reaches on the way to resealing. The second point represents the establishment of Ca2+ homeostasis. This step might comprise different processes including Ca2+ efflux, Ca2+ uptake by intracellular stores and so on. This step has been shown to be accomplished within 3 min. Studies using scanning electron microscopy and atomic force microscopy showed that altered membrane roughness and pit-like structures were observed in the cell membrane post-TUS exposure and, interestingly, did not recover until 24 hours later [35,36]. What is notable about these studies is that both of them used Optison microbubbles (heat-denatured human albumin shell with perfluorocarbon [PFC] gas core). In a recent study, another 24-hour recovery was reported with PFC-filled phospholipid-shelled microbubbles [37]. Whether these long-term membrane changes are a part of the normal sealing procedure after the patching of the membrane disruption (in which case, there should be discrimination between membrane ‘patching (sealing)’, which can be defined as the closure of membrane disruptions, and membrane ‘recovery’, which indicates the return to a pre-disruption condition; Fig. 6) or a specific response to TUS exposure in the presence of PFC-filled microbubbles is unknown at present [69]. It should be noted, however, that some controversy still pervades the use of perflutren-based contrast agent microbubbles with the FDA raising concerns over the safety of such products since 2007, whereas others have argued strongly for their efficacy [70,71].
Cell morphology after TUS exposure Cells exposed to TUS were observed to shrink in size and acquire a smoother surface [34]. This phenomenon, termed the ‘shaving effect’, might be due to the mechanical stripping of microvilli, for example, by radial flow from collapsing bubbles in the vicinity of cells (Fig. 5). Other studies have shown sonicated cells with smaller cross-section diameters but with irregular surfaces and villiform structures [37]. In a recent study of immediate morphological changes occurring post-sonication, cells were found to exhibit ‘balloon’-like membrane blebs consisting of intracellular lipids sprouting outward and finally shedding off into the extracellular environment and ‘blisters’ that can eventually re-integrate with the plasma membrane [72]. These membrane blebs were observed in cells recovering from membrane wounds caused by acoustic exposure, as well as laser ablation, mechanical shear and poreforming polypeptides (SLO) [73], reflecting their role in membrane sealing. How these altered surface topographies are correlated, what the exact mechanisms behind them are and the role of membrane blebbings in self-sealing (Is it mere occlusion of the ‘hot spots of Ca2+ entry’ or intended shedding of plasma membranes?) all require further investigation to clarify.
Having sealed: will cells always survive? Cells that fail to seal their membranes because of deficiency in Ca2+ in the surrounding environment or because of extensive membrane trauma seem to die immediately. However, those that have their membranes sealed still retain a facility to die in a selective manner mediated via other pathways [50,74]. For instance, apoptosis was shown to be induced under certain acoustic conditions [22,69,75–77]. The evidence for the integrity of membranes at this point is the externalization of phosphatidylserine, a marker of early apoptosis, on the outer leaflet of the cell membrane with a simultaneous exclusion of dyes indicative of disrupted cell membrane permeability, such as PI, as detected by double staining technique performed by flow cytometry [78]. Thus, ‘sealing’ is not simply sufficient for ‘survival’, but rather a necessary step for cells to regain decision-making ability. Fig. 6 summarizes the possible survival–death pathways in the post-insonation period.
What are the stimuli that trigger the apoptotic cell death after TUS? It is doubtless that Ca2+ contributes to the apoptosis signalling in mammalian cells [19,79,80]. In TUS-induced apoptosis, there is evidence that the molecular pathway of apoptosis proceeds via a Ca2+-dependent pathway together with the intrinsic mitochondrial pathway [21], during which an increase in intracellular Ca2+ was noticed over a period of four hours and then returned to the normal levels, probably as a result of impaired membrane permeability and the consequent leakage of Ca2+ at later stages of apoptosis. This inference is supported by the occurrence of larger percentages of secondary necrotic cells after six hours incubation compared with early apoptotic cells. Although the authors of this study concluded that the rise was, in part, due to the extracellular Ca2+ influx through non-specific pores because no increase was observed under sonication in Ca2+-free buffer, it is now clear that this conclusion cannot be drawn from such an experimental protocol because the lack of extracellular Ca2+ prevents the resealing of the membrane and in such a case, the cells die immediately, www.drugdiscoverytoday.com
899
Reviews KEYNOTE REVIEW
was eliminated when the inhibition was carried out by potassium depletion and FM1-43 stain [26].
REVIEWS
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Figure_4)TD$IG]
(a)
1a
(i)
(?)
Reviews KEYNOTE REVIEW
Long-term (hr) 3 min
(ii)
Time
Immediate necrosis
1b (?)
Ca2+rise
2b
2c
Long-term (hr)
Cell before the application of TUS
3 min
Time
(iii)
Two - component transient (?)
2c
Ca2+rise
2b
Sealed membrane
(?)
Nucleus
Long-term (hr)
Time
Ca2+
(iv)
Extracellular matter
(?) 4a Ca2+rise
Intracellular matter
(b)
TUS
Microbubbles
Time
Ca2+rise
Long-term (hr)
Single - component transient (?)
Intact cell membrane
Ca2+rise
3a
Induction of physical pores (sonoporation)
Cell
Mechano-transduction (mediated by intracellularly generated free radicals)
Ca2+rise
Time Two–component , immediate Ca2+ transient.
Time Single–component , delayed Ca2+ transient. Drug Discovery Today
FIGURE 4
(a) A representation of the different types of Ca2+ ‘ultra-short’ transients in response to TUS exposure in relation to increased permeability. (i) and (ii) represent the immediate transients, and (iii) and (iv) represent the delayed transients. The vertical dashed lines indicate the sonication period; the solid black lines indicate the Ca2+ transients, whereas the grey dashed lines indicate the uptake of extracellular matter (implication in delivery). Dashed squares refer to the temporal condition of the exposed cells. (i) Extensive cell disruption occurs, resulting in immediate and sustained influx of Ca2+ and extracellular matter simultaneous with efflux of intracellular matter (necrosis). (ii) The cell succeeds in patching the disrupted membrane, yet, its long-term fate is unknown (indicated by a question mark). Delayed transients, which can be monotonous (iii) or in the form of oscillations of different frequencies (iv), possess lower amplitude than immediate transients. 900
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
[(Figure_5)TD$IG]
FIGURE 5
Scanning probe microscopy representative image of an unexposed human prostate cancer (DU145) cell cultured as a monolayer. The cell surface is dominated by microvilli and shallow (circa 20–30 nm) depressions (left). Exposure to 2 60 ms ultrasound bursts at peak negative pressure of 2.1 MPa in the presence of Optison (1.7%) resulted in microvilli clearance (shaving effect) with the appearance of a micron-radius ‘sonopore’ extending 450 nm into the cell (right). Notably, the inner leaflet of the cell membrane seems to have flapped out over the cell surface, suggestive of a ‘bursting’ response entrance wound. Courtesy of Paul Campbell, University of Dundee.
as was discussed earlier. The observed increase in the intracellular Ca2+ could, rather, be due to the mobilization of intracellular Ca2+ stores and a possibly delayed influx [19]. Whether the initial ultrashort Ca2+ transient is a stimulus, partially or wholly, to these events is still unknown. Thus, in all cases, a role for the extracellular Ca2+ cannot be excluded in apoptosis induction [81]. Ca2+ has been shown to activate a diverse range of Ca2+-sensitive factors that are compartmentalized in cellular organelles such as the ER and mitochondria and even in the cytoplasm [82,83]. Several hypothetical possibilities can operate. If Ca2+ entering the cell during sonication succeeds in accessing the mitochondrial matrix, activation of the mitochondrial efflux mechanisms and matrix-Ca2+ buffering occurs. One might expect that success in these mechanisms could lead to recovery of homeostasis and that failure would result in Ca2+ overload in mitochondria, leading to a decrease in its membrane potential and, thus, opening the permeability transition pores, releasing cytochrome C, which, in turn, regulates the mitochondrial downstream events for apoptosis. Although this sequence might hold true for sonically disrupted cells, it cannot work for those cells showing post-irradiation delayed ultra-short transients. These cells experience different ‘frequency-modulated oscillations’, which are believed to be a cellular language to affect Ca2+-sensitive targets and, thus, the encoding of these oscillations (and waves) might contribute to the modulation of cellular responses to TUS [84]. Generally, the release of cytochrome C is known to activate the IP3 receptor leading to Ca2+ efflux from ER [85]. Recently, the existence of cross-talk between ER and mitochondria during apoptosis signalling and regulation has been reported [80]. Moreover, ER has been shown to be able to initiate the apoptotic signalling before mitochondrial involvement through caspase 4. Other reports have shown that the delocalization of Bak and Bax on
the ER because of ER stress resulted in Ca2+ release to the cytosol, which can then be taken up by juxtaposed mitochondria. The involvement of ER in TUS-induced apoptosis is also to be expected, especially when heme oxygenase-1 increased expression after TUS application is considered [86]. Although heme oxygenase-1 overexpression is anti-apoptotic, it has been found to be induced by ER stress, which simultaneously initiates apoptosis [87,88]. The involvement of ER stress was further evidenced to contribute to TUS-induced apoptosis through the twofold increase in the expression of GRP78/Bip protein protein after TUS application [89]. The study by Honda et al. [21] stated that the mitochondrial pathway is only partially rather than wholly affected by the intracellular Ca2+ increase, which indicates the presence of other initiators, at least in leukaemia cells [90]. In support of this study, the occurrence of delayed apoptotic effects with increasing post-sonication incubation time augments the presence of another initiator stimuli [68,69,83,89]. Nevertheless, the exact interplay between organelles mediating apoptosis and whether the extracellular Ca2+ influx is an initiator or a delayed consequence of TUS-induced apoptosis remain to be explored. In addition, apoptosis might not be the sole programmed response to TUS exposure. Future research might expose alternative processes, especially when TUS-induced DNA damage [91–93] is considered (under study).
Is inertial cavitation important for TUS-mediated bioeffects? Much of the recent research ascribes sonoporation to the presence of artificial microbubbles during sonication. In such cases, two possibilities can operate: that these microbubbles collapse (inertial cavitation) or that they merely oscillate upon exposure to ultrasound. Some researchers not only claim that the bubble collapse is a prerequisite for enhanced delivery but also suggest a role for the
The extracellular matter uptake is solely controlled by normal transport mechanisms that are molecule dependent. (b) A hypothetical diagram of the possible origins of the observed ultra-short Ca2+ transients. www.drugdiscoverytoday.com
901
REVIEWS
[(Figure_6)TD$IG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews KEYNOTE REVIEW FIGURE 6
A schematic representation of the evolutionary possibilities post-insonation in relation to the extent of cell membrane trauma. The validity of thermodynamic resealing in which pores are sealed passively has been argued recently and replaced by the ‘patch hypothesis’, at least for larger wounds. Endocytosis has been shown to occur in the course of membrane self-sealing, together with exocytosis, yet the exact details are still unresolved (shaded square). It is still unknown whether the pits observed by atomic force microscopy correspond to a sort of endocytotic activity or a specific response to PFC-filled microbubbles (see the section ‘Waypoints along the route to self-sealing’).
free radicals generated upon the collapse, namely the hydroxyl radicals (OH) [53]. In fact, the notion that inertial cavitation is a prerequisite for induction of pores might not be true [94]. van Wamel et al. have shown that vibrating bubbles are also capable of internalizing normally impermeable molecules and that internalization was accompanied by membrane deformation and poration that was followed by self-sealing within a few minutes [30]. Another report by Mortimer and Dyson [95] showed that Ca2+ uptake was increased in fibroblasts exposed to TUS in the absence of transient cavitation. Moreover, the presence of a microbubble near the cell observed by Kumon et al. [25] and its disappearance upon ultrasound application is not indicative of inertial cavitation, for several reasons. (i) The authors did not perform a test for free radical formation accompanied by microbubble collapse or other relevant tests. (ii) The disappearance of 98% of the microbubbles after sonication can be justified by their dissolution in the large volume 902
www.drugdiscoverytoday.com
of sonicating medium (7–10 ml). (iii) The 4 mm distance between the emitting transducer surface and the cells might not induce instantaneous in situ collapse; rather, it could be the violent dislocation of the bubble that caused shearing of the membrane. As for the role of free radicals, the literature has conflicting reports as to their involvement in permealization and TUSinduced apoptosis. For example, the presence of free radical scavengers was shown to decrease the selective enhancement of caveolar-dependent permeability and the oxidative stress in endothelial cells treated with ultrasound [66,68]. It has been reported by Burlando and Viarengo [96] that free radicals can stimulate Ca2+ increase intracellularly, which is a net of both extra- and intracellular Ca2+ contributions. Their observations revealed that this action is selectively carried out by OH radicals rather than superoxide radicals. Hydroxyl radicals are the major free radicals generated from water sonolysis upon the collapse of
microbubbles. According to the authors, however, the Ca2+ increase took more than 8 min incubation with the free-radicalreleasing mixture to be discernible; therefore, it is unlikely that the short-lived free radicals formed within the short periods of sonication could contribute directly to the Ca2+ transients observed. Yet their being localized by TUS still revives the hypothesis [53], and an indirect role cannot be excluded except by further detailed investigations. For instance, although the cavitation-induced free radical formation occurs only during sonication, which – in most cases – extends to very short periods, it could still be that these radicals can initiate reactions with the medium contents that extend for longer times and thus exert some effect downstream of their initial generation. This conjecture is supported by the observation that transferring the sonicated media to untreated cells resulted in a similar oxidative stress to that found with sonicated cells [68]. In addition, free radicals that could be produced intracellularly might be responsible. If we consider the work done by Juffermans et al. [53], we find that the authors, using two different TUS conditions with different mechanical indices, namely 0.1 and 0.5, stated that at these conditions, inertial cavitation was not prone to occur – even at the higher mechanical index (MI), which is already beyond the reported MI threshold for inertial cavitation, the thing that was argued elsewhere [69]. Despite that, they showed a close correlation between OH formation and transient permealization as indicated by Ca2+ transients. Their results suggest that the OH radicals might be responsible for the Ca2+ influx as indicated by the blockade of entry in the presence of catalase. This work demands closer attention because when combined with two other studies [42,52], it might provide new insights into TUS interactions. First, the authors detected hydrogen peroxide (H2O2) formation by using a specific intracellular dye; it is thus suggested that the radicals might be produced intracellularly. Second, the authors used an extracellular scavenger, catalase, to check the effect of H2O2 on Ca2+ transients. The catalase present during sonication blocked the Ca2+ transients at MI 0.1 and partially suppressed them at MI 0.5. It is notable that the cells used (H9c2 rat cardiomyoblasts) – being muscle cells – responded with a separate transient at each pulse compared to human umbilical vein endothelial cells under similar sonication conditions [52]. In the presence of catalase, the suppressive effect increased with each pulse, implying that catalase was being internalized with successive pulses, thus further augmenting the possible intracellular origin of H2O2. Moreover, if we recall this group’s statement: ‘It is suggested that superoxide is formed by high shear stress, especially by the vortex like microstreaming around oscillating microbubbles. Superoxide is rapidly dismutated to H2O2,’ together with the fact that mechanical stress was also shown to result in Ca2+ oscillations [23] and the observation that Ca2+ transients were not detected except in the presence of microbubbles in most of the respective studies cited here, we can infer that the Ca2+ transients detected could be two-component transients with one component related to pore formation from which extracellular Ca2+ fluxes into the cell (leading to immediate transients) and the other component related to mechanotransduction that results in a comparatively delayed Ca2+ transient initially of intracellular origin. The former component has been shown to be correlated to delivery [42] and could mask the later component. In summary, the ultra-short Ca2+ transients might be either
REVIEWS
immediate two-component or delayed single-component transients (Fig. 4b). It should be noted that these are merely analytical conclusions, however; more directly evident practical data are still required to present a logically consistent overview. Finally, the exact interplay between free radicals, transient permealization and Ca2+ transients is not yet clear in terms of which one stimulates the other and what the outcomes are [97].
Can Ca2+ contribute to other cellular responses? Since 2005, interest in the role of TUS-enhanced delivery in nuclear localization has continued apace since it was proved by DuvshaniEshet et al. [36,98] that increased plasmid DNA (pDNA) localization occurs in cell nuclei after acoustic application. What was particular to their studies is that they employed a long sonication schedule, reaching 30 min, whilst retaining high viability in the cell population through sonicating the cells from above, thus minimizing the detachment of cells as when sonicated from beneath, which seems to play a part in cell death [99]. They found that increasing the sonication duration could increase the percentage of cells localizing pDNA in their nuclei. Furthermore, they succeeded in specifying the percentage of cells localizing the pDNA at different cellular sites after different treatment protocols. By contrast, the uptake of high molecular weight species, though enhanced by short sonication periods, the molecules were commonly excluded from the nucleus despite the attainment of a thermodynamic equilibrium at least in the HUP cells as previously mentioned [26,61]. Taken together, it might be concluded that prolonged sonication increased the nuclear access of these molecules, despite their reported poor diffusion through the cytoplasm under normal conditions. It has been found that the intermediate-sized DNA molecules (10–70 kDa) traverse the nucleus through a nuclear pore complex (NPC) that spans both layers of the nuclear membrane. The regulation of NPC is achieved by the cisternal Ca2+ release and storage, where its release causes the NPC to close, and vice versa [100]. The intracellular Ca2+ stores depletion through IP3 mediation results in Ca2+ release from the cisternae of the nuclear membrane as well, leading to the closure of the NPC. In such a case, the DNA cannot access the nucleus. In calcium phosphate nanoparticles-mediated transfection, the presence of excessive free intracellular Ca2+ in the cytosol inactivates the IP3 through complexation, preventing the IP3 receptor-assisted drainage of cisternal Ca2+, and the NPC will be open for pDNA to traverse and transfect the cells successfully. Similarly, if we suppose that prolonged sonication leads to an increase in the intracellular Ca2+ owing to the sustained opening of the pores and consequently increasing the cytosolic free Ca2+, the same scenario might operate. Whether this is true is still unknown and requires further efforts to ascertain in a definitive fashion, but the low transfection levels obtained in the U937 cell line despite the successful cytoplasmic delivery produced by TUS might provide a proof of concept if studied carefully. U937 cells have been shown to be a very sensitive cell line in which TUS can induce apoptosis that is accompanied by intracellular Ca2+ increase, as mentioned previously. Thus, if this increase in Ca2+ is due to intracellular depletion of Ca2+ stores, then it is probable that the cisternal Ca2+ will also be depleted, with a consequent closure of the NPC. This postulation was implicitly embraced in a recent study conducted by Miller and Dou [11], in which the authors compared leukaemiawww.drugdiscoverytoday.com
903
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
derived HL-60 cells – a cell line with an inherent capacity for apoptosis – and epithelium-like CHO-K1 cells, with ‘a lower tendency for apoptosis’, with respect to simultaneous apoptosis induction and gene transfection by ultrasound. The study showed that although sonoporation occurred successfully in both cell lines, more cells of the loaded population showed apoptosis in the former cell line, rendering the percentage of the viable geneexpressing cells lower. Interestingly, the authors referred to extracellular Ca2+ influx in their discussion, although they did not clarify their views concerning its role. Moreover, the attractiveness of muscle cells as targets for gene delivery [101], compared to tumour cells, might involve a role of Ca2+ in muscle physiology that one might expect to be more adapted to intracellular Ca2+ transients and to their rapid buffering as a part of their normal function as ‘contractile cells’ [73]. Again, this postulation requires verification and, at least until now, does not preclude the involvement of other mechanisms. For instance, prolonged sonication might deactivate some degrading enzymes or promote the diffusion of large molecules through the cytoplasm. Finally, it is worth mentioning that there are other responses to TUS exposure that involve the Ca2+ signalling system, including the stimulation of normal physiological processes, such as accelerated healing of bone fractures by TUS [51,102,103] and wound healing [104]; however, a discussion of these responses and their mechanisms is beyond the scope of this review.
Concluding remarks If intracellular Ca2+ is involved in both cell proliferation and cell death, what specific stimulus is there for activating either pathway? The answer to this question might be the key for progressing TUS ‘from bench to bedside’ and into a fully realized clinical application. Could it be that the sonication conditions and their associated chemical and mechanical effects prevail? Or that the cell type and its inherent response to external stimuli dictate the outcome? Is it the encoding of Ca2+ oscillations? Could it be a threshold dictated by the intracellular Ca2+ concentration of sonically porated cells? It has been shown, for instance, that
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
membrane self-sealing can operate at elevated intracellular Ca2+ concentrations between 5 and 10 mM, above which the cell reaches the point of no return, undergoing immediate lysis [73]. If we can extend these results to sonically porated cells in the light of, for example, the study conducted by Hutcheson et al. [62], could it be that the cells previously designated as ‘HUP’ showing initial shortterm viability are the same cells having the maximum tolerated intracellular Ca2+ elevation at which sealing can be achieved but meanwhile seen as toxic and apt to trigger apoptosis? If yes, why did only a small percentage of cells in this regime undergo apoptosis? And what was their intracellular Ca2+ profile within the period from sonication until apoptosis occurrence? The success of lithotripsy and more recently, high-intensity focused ultrasound, as interventional tools, underscores the power and versatility of ultrasound as a minimally invasive therapy. However, the full and complete therapeutic potential of ultrasound, especially for the compelling scenarios of molecular delivery, tissue sensitization, and for the subtle control of biochemical pathways, remains untapped at a clinical level. We conclude here that one important route to addressing this lies in developing a concerted effort that is focussed on achieving a more complete understanding of the role of Ca2+ transport during insonation. This will consolidate the present knowledge base, and in the longer term, assist in paving the way towards a much broader and powerful implementation of ultrasound for therapeutic purposes. Evidently, the intervention to save cells from TUS-induced apoptosis through post-sonication Ca2+ chelation would not have been studied successfully but for the knowledge of the role of Ca2+ in membrane resealing. Although this approach is suggested to enhance delivery in vitro only, it offers some suggestion as to how the achievement of detailed fundamental knowledge can initiate fast strides likewise in vivo.
Acknowledgements Assistance with aspects of the microscopy and images is gratefully appreciated from Alan Prescott, Amyn Teja, Barry Marshall and Luke Howse.
References 1 Wood, R.W. and Loomis, A.L. (1927) The physical and biological effects of highfrequency sound waves of great intensity. Philos. Mag. 7, 417–436 2 Warden, S.J. (2003) A new direction for ultrasound therapy in sports medicine. Sports Med. 33, 95–107 3 Griffin, X.L. et al. (2008) The role of low intensity pulsed ultrasound therapy in the management of acute fractures: a systematic review. J. Trauma 65, 1446–1452 4 Warden, S.J. et al. (2006) Ultrasound produced by a conventional therapeutic ultrasound unit accelerates fracture repair. Phys. Ther. 86, 1118–1127 5 Feril, L.B., Jr and Kondo, T. (2004) Biological effects of low intensity ultrasound: the mechanism involved, and its implications on therapy and on biosafety of ultrasound. J. Radiat. Res. (Tokyo) 45, 479–489 6 Feril, L.B., Jr et al. (2003) Enhancement of ultrasound-induced apoptosis and cell lysis by echo-contrast agents. Ultrasound Med. Biol. 29, 331–337 7 Yoshida, T. et al. (2008) Combination of doxorubicin and low-intensity ultrasound causes a synergistic enhancement in cell killing and an additive enhancement in apoptosis induction in human lymphoma U937 cells. Cancer Chemother. Pharmacol. 61, 559–567 8 Yu, T. et al. (2009) Ultrasound increases DNA damage attributable to cisplatin in cisplatin-resistant human ovarian cancer cells. Ultrasound Obstet. Gynecol. 33, 355–359 9 Devcic-Kuhar, B. et al. (2002) In vitro thrombolysis enhanced by standing and travelling ultrasound wave fields. Ultrasound Med. Biol. 28, 1181–1187
904
www.drugdiscoverytoday.com
10 Pfaffenberger, S. et al. (2003) 2 MHz ultrasound enhances t-PA-mediated thrombolysis: comparison of continuous versus pulsed ultrasound and standing versus travelling acoustic waves. Thromb. Haemost. 89, 583–589 11 Miller, D.L. and Dou, C. (2009) Induction of apoptosis in sonoporation and ultrasonic gene transfer. Ultrasound Med. Biol. 35, 144–154 12 Hassan, M.A. et al. (2010) Modulation control over ultrasound-mediated gene delivery: evaluating the importance of standing waves. J. Control. Release 141, 70–76 13 Campbell, P. and Prausnitz, M.R. (2007) Future directions for therapeutic ultrasound. Ultrasound Med. Biol. 33, 657 14 Fechheimer, M. et al. (1987) Transfection of mammalian cells with plasmid DNA by scrape loading and sonication loading. Proc. Natl. Acad. Sci. U. S. A. 84, 8463–8467 15 Giacomello, M. et al. (2007) Mitochondrial Ca2+ as a key regulator of cell life and death. Cell Death Differ. 14, 1267–1274 16 Lange, K. and Gartzke, J. (2001) Microvillar cell surface as a natural defense system against xenobiotics: a new interpretation of multidrug resistance. Am. J. Physiol. Cell Physiol. 281, C369–C385 17 Pietrobon, D. et al. (1990) Structural and functional aspects of calcium homeostasis in eukaryotic cells. Eur. J. Biochem. 193, 599–622 18 Monteith, G.R. (2000) Seeing is believing: recent trends in the measurement of Ca2+ in subcellular domains and intracellular organelles. Immunol. Cell Biol. 78, 403–407
19 Preston, G.A. et al. (1997) Effects of alterations in calcium homeostasis on apoptosis during neoplastic progression. Cancer Res. 57, 537–542 20 Deng, C.X. et al. (2004) Ultrasound-induced cell membrane porosity. Ultrasound Med. Biol. 30, 519–526 21 Honda, H. et al. (2004) Role of intracellular calcium ions and reactive oxygen species in apoptosis induced by ultrasound. Ultrasound Med. Biol. 30, 683–692 22 Honda, H. et al. (2002) Effects of dissolved gases and an echo contrast agent on apoptosis induced by ultrasound and its mechanism via the mitochondria–caspase pathway. Ultrasound Med. Biol. 28, 673–682 23 Kono, T. et al. (2006) Spontaneous oscillation and mechanically induced calcium waves in chondrocytes. Cell Biochem. Funct. 24, 103–111 24 Kumon, R.E. et al. (2007) Ultrasound-induced calcium oscillations and waves in Chinese hamster ovary cells in the presence of microbubbles. Biophys. J. 93, L29–L31 25 Kumon, R.E. et al. (2009) Spatiotemporal effects of sonoporation measured by real-time calcium imaging. Ultrasound Med. Biol. 35, 494–506 26 Schlicher, R.K. et al. (2006) Mechanism of intracellular delivery by acoustic cavitation. Ultrasound Med. Biol. 32, 915–924 27 Yang, F. et al. (2008) Experimental study on cell self-sealing during sonoporation. J. Control. Release 131, 205–210 28 Meijering, B.D. et al. (2009) Ultrasound and microbubble-targeted delivery of macromolecules is regulated by induction of endocytosis and pore formation. Circ. Res. 104, 679–687 29 NCRP, (1983) NCRP Report No. 74. Biological Effects of Ultrasound: Mechanisms and Clinical Implications. National Council on Radiation Protection and Measurements 30 van Wamel, A. et al. (2006) Vibrating microbubbles poking individual cells: drug transfer into cells via sonoporation. J. Control. Release 112, 149–155 31 Pan, H. et al. (2005) Study of sonoporation dynamics affected by ultrasound duty cycle. Ultrasound Med. Biol. 31, 849–856 32 Tran, T.A. et al. (2007) Effect of ultrasound-activated microbubbles on the cell electrophysiological properties. Ultrasound Med. Biol. 33, 158–163 33 Zhou, Y. et al. (2009) The size of sonoporation pores on the cell membrane. Ultrasound Med. Biol. 35, 1756–1760 34 Mehier-Humbert, S. et al. (2005) Plasma membrane poration induced by ultrasound exposure: implication for drug delivery. J. Control. Release 104, 213–222 35 Taniyama, Y. et al. (2002) Development of safe and efficient novel nonviral gene transfer using ultrasound: enhancement of transfection efficiency of naked plasmid DNA in skeletal muscle. Gene Ther. 9, 372–380 36 Duvshani-Eshet, M. et al. (2006) Therapeutic ultrasound-mediated DNA to cell and nucleus: bioeffects revealed by confocal and atomic force microscopy. Gene Ther. 13, 163–172 37 Zhao, Y.Z. et al. (2008) Phospholipids-based microbubbles sonoporation pore size and reseal of cell membrane cultured in vitro. J. Drug Target. 16, 18–25 38 Prentice, P. et al. (2005) Membrane disruption by optically controlled microbubble cavitation. Nat. Phys. 1, 107–110 39 McNeil, P.L. (2002) Repairing a torn cell surface: make way, lysosomes to the rescue. J. Cell Sci. 115, 873–879 40 Saito, K. et al. (1999) Plasma membrane disruption underlies injury of the corneal endothelium by ultrasound. Exp. Eye Res. 68, 431–437 41 McNeil, P.L. and Steinhardt, R.A. (1997) Loss, restoration, and maintenance of plasma membrane integrity. J. Cell Biol. 137, 1–4 42 Fan, Z. et al. (2010) Intracellular delivery and calcium transients generated in sonoporation facilitated by microbubbles. J. Control. Release 142, 31–39 43 Reddy, A. et al. (2001) Plasma membrane repair is mediated by Ca(2+)-regulated exocytosis of lysosomes. Cell 106, 157–169 44 Bakker, A.C. et al. (1997) Homotypic fusion between aggregated lysosomes triggered by elevated [Ca2+]i in fibroblasts. J. Cell Sci. 110, 2227–2238 45 Coorssen, J.R. et al. (1996) Ca2+ triggers massive exocytosis in Chinese hamster ovary cells. EMBO J. 15, 3787–3791 46 Terasaki, M. et al. (1997) Large plasma membrane disruptions are rapidly resealed by Ca2+-dependent vesicle–vesicle fusion events. J. Cell Biol. 139, 63–74 47 Gross, S.R. et al. (2003) Importance of tissue transglutaminase in repair of extracellular matrices and cell death of dermal fibroblasts after exposure to a solarium ultraviolet A source. J. Invest. Dermatol. 121, 412–423 48 Haroon, Z.A. et al. (1999) Tissue transglutaminase is expressed, active, and directly involved in rat dermal wound healing and angiogenesis. FASEB J. 13, 1787–1795 49 Nardacci, R. et al. (2003) Transglutaminase type II plays a protective role in hepatic injury. Am. J. Pathol. 162, 1293–1303 50 Kawai, Y. et al. (2008) Transglutaminase 2 activity promotes membrane resealing after mechanical damage in the lung cancer cell line A549. Cell Biol. Int. 32, 928–934 51 Parvizi, J. et al. (2002) Calcium signaling is required for ultrasound-stimulated aggrecan synthesis by rat chondrocytes. J. Orthop. Res. 20, 51–57
REVIEWS
52 Juffermans, L.J. et al. (2009) Ultrasound and microbubble-induced intra- and intercellular bioeffects in primary endothelial cells. Ultrasound Med. Biol. 35, 1917–1927 53 Juffermans, L.J. et al. (2006) Transient permeabilization of cell membranes by ultrasound-exposed microbubbles is related to formation of hydrogen peroxide. Am. J. Physiol. Heart Circ. Physiol. 291, H1595–H1601 54 Kudo, N. et al. (2009) Sonoporation by single-shot pulsed ultrasound with microbubbles adjacent to cells. Biophys. J. 96, 4866–4876 55 Schuster, S. et al. (2002) Modelling of simple and complex calcium oscillations. From single-cell responses to intercellular signalling. Eur. J. Biochem. 269, 1333–1355 56 Kang, M. and Othmer, H.G. (2007) The variety of cytosolic calcium responses and possible roles of PLC and PKC. Phys. Biol. 4, 325–343 57 Blott, E.J. and Griffiths, G.M. (2002) Secretory lysosomes. Nat. Rev. Mol. Cell Biol. 3, 122–131 58 Idone, V. et al. (2008) Two-way traffic on the road to plasma membrane repair. Trends Cell Biol. 18, 552–559 59 Idone, V. et al. (2008) Repair of injured plasma membrane by rapid Ca2+dependent endocytosis. J. Cell Biol. 180, 905–914 60 Brayman, A.A. et al. (1999) Transient poration and cell surface receptor removal from human lymphocytes in vitro by 1 MHz ultrasound. Ultrasound Med. Biol. 25, 999–1008 61 Guzman, H.R. et al. (2002) Equilibrium loading of cells with macromolecules by ultrasound: effects of molecular size and acoustic energy. J. Pharm. Sci. 91, 1693–1701 62 Hutcheson, J.D. et al. (2010) Saving cells from ultrasound-induced apoptosis: quantification of cell death and uptake following sonication and effects of targeted calcium chelation. Ultrasound Med. Biol. 36, 1008–1021 63 Vanni, A. et al. (1998) DNA damage and cytotoxicity induced by beta-lapachone: relation to poly(ADP-ribose) polymerase inhibition. Mutat. Res. 401, 55–63 64 Togo, T. et al. (2000) A decrease in membrane tension precedes successful cellmembrane repair. Mol. Biol. Cell 11, 4339–4346 65 Kilic, G. (2002) Exocytosis in bovine chromaffin cells: studies with patch-clamp capacitance and FM1-43 fluorescence. Biophys. J. 83, 849–857 66 Lionetti, V. et al. (2009) Enhanced caveolae-mediated endocytosis by diagnostic ultrasound in vitro. Ultrasound Med. Biol. 35, 136–143 67 Fittipaldi, A. et al. (2003) Cell membrane lipid rafts mediate caveolar endocytosis of HIV-1 Tat fusion proteins. J. Biol. Chem. 278, 34141–34149 68 Basta, G. et al. (2003) In vitro modulation of intracellular oxidative stress of endothelial cells by diagnostic cardiac ultrasound. Cardiovasc. Res. 58, 156–161 69 Hassan, M.A. et al. (2009) Evaluation and comparison of three novel microbubbles: enhancement of ultrasound-induced cell death and free radicals production. Ultrason. Sonochem. 16, 372–378 70 Main, M.L. et al. (2009) Ultrasound contrast agents: balancing safety versus efficacy. Expert Opin. Drug Saf. 8, 49–56 71 Main, M.L. et al. (2007) Thinking outside the ‘‘box’’ – the ultrasound contrast controversy. J. Am. Coll. Cardiol. 50, 2434–2437 72 Schlicher, R.K. et al. (2010) Changes in cell morphology due to plasma membrane wounding by acoustic cavitation. Ultrasound Med. Biol. 36, 677–692 73 Babiychuk, E.B. et al. (2009) Intracellular Ca(2+) operates a switch between repair and lysis of streptolysin O-perforated cells. Cell Death Differ. 16, 1126–1134 74 Majno, G. and Joris, I. (1995) Apoptosis, oncosis, and necrosis. An overview of cell death. Am. J. Pathol. 146, 3–15 75 Feril, L.B., Jr et al. (2005) Apoptosis induced by the sonomechanical effects of low intensity pulsed ultrasound in a human leukemia cell line. Cancer Lett. 221, 145–152 76 Takeuchi, S. et al. (2006) Basic study on apoptosis induction into cancer cells U-937 and EL-4 by ultrasound exposure. Ultrasonics 44 (Suppl. 1), e345–e348 77 Ando, H. et al. (2006) An echo-contrast agent, Levovist, lowers the ultrasound intensity required to induce apoptosis of human leukemia cells. Cancer Lett. 242, 37–45 78 Vermes, I. et al. (2000) Flow cytometry of apoptotic cell death. J. Immunol. Methods 243, 167–190 79 Smaili, S.S. et al. (2003) Mitochondria, calcium and pro-apoptotic proteins as mediators in cell death signaling. Braz. J. Med. Biol. Res. 36, 183–190 80 Jeong, S.Y. and Seol, D.W. (2008) The role of mitochondria in apoptosis. BMB Rep. 41, 11–22 81 Kim, J.A. et al. (1998) Role of Ca2+ influx in the tert-butyl hydroperoxideinduced apoptosis of HepG2 human hepatoblastoma cells. Exp. Mol. Med. 30, 137–144 82 Nicotera, P. and Orrenius, S. (1998) The role of calcium in apoptosis. Cell Calcium 23, 173–180
www.drugdiscoverytoday.com
905
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
83 Mirnikjoo, B. et al. (2009) Mobilization of lysosomal calcium regulates the externalization of phosphatidylserine during apoptosis. J. Biol. Chem. 284, 6918–6923 84 Hajnoczky, G. et al. (1995) Decoding of cytosolic calcium oscillations in the mitochondria. Cell 82, 415–424 85 Liao, P.C. et al. (2008) Involvement of endoplasmic reticulum in paclitaxelinduced apoptosis. J. Cell. Biochem. 104, 1509–1523 86 Kagiya, G. et al. (2006) Expression of heme oxygenase-1 due to intracellular reactive oxygen species induced by ultrasound. Ultrason. Sonochem. 13, 388–396 87 Liu, X.M. et al. (2005) Endoplasmic reticulum stress stimulates heme oxygenase-1 gene expression in vascular smooth muscle. Role in cell survival. J. Biol. Chem. 280, 872–877 88 Gissel, C. et al. (1997) Activation of heme oxygenase-1 expression by disturbance of endoplasmic reticulum calcium homeostasis in rat neuronal cell culture. Neurosci. Lett. 231, 75–78 89 Feng, Y. et al. (2008) Low intensity ultrasound-induced apoptosis in human gastric carcinoma cells. World J. Gastroenterol. 14, 4873–4879 90 Lennon, S.V. et al. (1992) Elevations in cytosolic free Ca2+ are not required to trigger apoptosis in human leukaemia cells. Clin. Exp. Immunol. 87, 465–471 91 Miller, D.L. and Thomas, R.M. (1996) The role of cavitation in the induction of cellular DNA damage by ultrasound and lithotripter shock waves in vitro. Ultrasound Med. Biol. 22, 681–687 92 Milowska, K. and Gabryelak, T. (2007) Reactive oxygen species and DNA damage after ultrasound exposure. Biomol. Eng. 24, 263–267 93 Miller, D.L. et al. (1995) Comet assay reveals DNA strand breaks induced by ultrasonic cavitation in vitro. Ultrasound Med. Biol. 21, 841–848
906
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
94 Marmottant, P. and Hilgenfeldt, S. (2003) Controlled vesicle deformation and lysis by single oscillating bubbles. Nature 423, 153–156 95 Mortimer, A.J. and Dyson, M. (1988) The effect of therapeutic ultrasound on calcium uptake in fibroblasts. Ultrasound Med. Biol. 14, 499–506 96 Burlando, B. and Viarengo, A. (2005) Ca2+ is mobilized by hydroxyl radical but not by superoxide in RTH-149 cells: the oxidative switching-on of Ca2+ signaling. Cell Calcium 38, 507–513 97 Gordeeva, A.V. et al. (2003) Cross-talk between reactive oxygen species and calcium in living cells. Biochemistry (Mosc.) 68, 1077–1080 98 Duvshani-Eshet, M. and Machluf, M. (2005) Therapeutic ultrasound optimization for gene delivery: a key factor achieving nuclear DNA localization. J. Control. Release 108, 513–528 99 Liang, H.D. et al. (2004) Optimisation of ultrasound-mediated gene transfer (sonoporation) in skeletal muscle cells. Ultrasound Med. Biol. 30, 1523–1529 100 Maitra, A. (2005) Calcium phosphate nanoparticles: second-generation nonviral vectors in gene therapy. Expert Rev. Mol. Diagn. 5, 893–905 101 Tsai, K.C. et al. (2009) Differences in gene expression between sonoporation in tumor and in muscle. J. Gene Med. 11, 933–940 102 Li, J.K. et al. (2006) Comparison of ultrasound and electromagnetic field effects on osteoblast growth. Ultrasound Med. Biol. 32, 769–775 103 Hsu, S.H. and Huang, T.B. (2004) Bioeffect of ultrasound on endothelial cells in vitro. Biomol. Eng. 21, 99–104 104 Ikai, H. et al. (2008) Low-intensity pulsed ultrasound accelerates periodontal wound healing after flap surgery. J. Periodontal Res. 43, 212–216
REVIEWS
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Flavonoids and genomic instability induced by ionizing radiation Seyed Jalal Hosseinimehr1,2 1 Department of Radiopharmacy, Faculty of Pharmacy, Traditional and Complementary Medicine Research Center, Mazandaran University of Medical Sciences, Sari, 48175-861, Iran 2 Unit of Biomedical Radiation Sciences, Rudbeck Laboratory, Uppsala University, Uppsala 751-85, Sweden
DNA is the cellular target that has the most damage induced by ionizing radiation (IR). If genomic instability resulting from this DNA damage is not correctly repaired, it leads to mutation, cancer and cell death. Flavonoids are a family of natural products that affect oxidative stress and enhance genomic stability through DNA interaction. Although flavonoids exert protective effects against IR in normal cells, they enhance genotoxicity effects of this radiation in cancer cells, a beneficial effect that is of interest in the design of new anticancer pharmaceuticals. This review describes the molecular effects of IR on DNA structure and mechanisms by which flavonoids exert their effect on ionizing-radiation-induced genomic instability.
Dr Seyed Jalal Hosseinimehr received his Doctorate in Radiopharmacy, from Tehran University of Medical Sciences, Iran in 2002. Immediately, he began as a member of Faculty of Pharmacy of Mazandaran University of Medical Sciences, Iran. His research focused on radiation biology mainly radioprotective compounds. He joined as a guest researcher in two universities including, laboratory of Radiation Biology, Hokkaido University, Japan for six months in 2001, and Unit of Biomedical Radiation Sciences, Rudbeck Laboratory, Uppsala, Sweden for eight months in 2010. At the moment, he is as associated professor of Faculty of Pharmacy, Mazandaran, Iran. His research interests include evaluating of radioprotective agents, DNA and tumor targeted radiopharmaceutical agents.
Introduction In normal cellular metabolism, reactive oxygen species (ROS) are produced, but their effects are balanced by the endogenous repair system. ROS levels can be elevated by exposure to oxidative stress such as ionizing radiation (IR; e.g. X-ray, gamma ray, b, a or proton) or by deficiencies in the cellular repair process. Exposure to ionizing irradiation results in the production of free radicals and toxic substances that can damage crucial macromolecules – including DNA, cell membranes and enzymes – and can cause cell death. DNA damage includes genotoxicity, chromosomal abnormalities, gene mutations and cell death [1]. Double-strand breaks (DSBs) induced by IR are considered the most dangerous lesion for cell survival and induction of genomic instability. Single-strand breaks (SSBs), which appear not only as a direct result of radiation but also during nucleotide excision repair of DNA or during normal replication, can be converted into DSBs if they are not rejoined. Flavonoids, a family of natural products present in many types of fruits and herbs, have many pharmacological properties including anti-inflammatory, hepatoprotective and antioxidant activities [2] and are able to interact with free radicals and substances produced by oxidative stress. These interactions are facilitated by their structure of polyphenol rings with hydroxyl and/or methoxyl groups, which can specifically bind with a base or other groups in the DNA backbone, as well as trap ROS. These polyphenolic compounds considerably mitigate the effects of ionizing irradiation at the molecular, cellular and/or tissue level [3–5].
E-mail address: Hosseinimehr, S.J. ([email protected], [email protected].) 1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.09.005
www.drugdiscoverytoday.com
907
REVIEWS
Throughout life, humans are exposed to IR from numerous sources including natural background sources, diagnostic procedures and the treatment of cancer by radiotherapy. Humans also consume daily high amounts of flavonoids through fruit and vegetable intake. This review focuses on how the chemical structure of flavonoids interacts with free radicals and stabilizes DNA structure. Reviews KEYNOTE REVIEW
Reactive substances produced by IR IR can remove electrons from atoms to form ions. The extent of these ionizations is dependent on the type of radiation and its energy. IR affects cellular crucial macromolecules directly and indirectly. Although the direct effect observed during exposure to IR is very small, the effect is more likely to happen with highenergy radiation (mostly particles) than photon. Direct radiation, rarely from IR but more commonly from high-energy particle radiation, can interact directly with the atoms and break bonds in a molecule such as DNA. Because the body and its cells are composed primarily of water, there is a much higher probability of radiation interacting with this molecule. When this occurs, it can break the intramolecular bonds to form ionized water and produce highly reactive free radicals such as hydrogen (H) and hydroxyl (HO). The combination of these free radicals can form toxic substances such as hydrogen peroxide (H2O2), which can contribute to the destruction of the cell. Whereas H and HO have short life spans, the longer life span of H2O2 enables it to migrate to sites distant from the point of IR exposure [6,7]. IR can also generate reactive nitrogen oxide species as toxic products, which are involved in radiation-induced signalling mechanisms [7]. These toxic substances produced by indirect effects are the main reason for the side-effects of IR. Interactions of free radicals with organic molecules can produce free radical residues that subsequently transfer damage to macromolecules and thus contribute to cumulative damage. The phenyl ring of tyrosine is a common target for free radical and oxidative substance attack, and this results in a long-living tyrosine phenoxyl [8]. Radiation-stimulated ROS and reactive nitrogen oxide species have been shown to activate signal transduction cascades such as mitogen-activated protein kinase to initiate through events in the plasma membrane, the cytoplasm and the nucleus, resulting in toxic and protective responses in cells [1,6].
Effect of ionizing irradiation on macromolecules and DNA DNA, RNA, enzymes, membrane and cellular proteins can suffer considerable damage from IR. Hydroxyl radicals produced by IR [9] oxidize amino acids, resulting in the hydroxylation of aromatic and aliphatic side chains and the addition of other groups on aromatic rings. This can induce progressive alterations in the structure of peptides, including cross-linkage and fragmentation or even alteration to individual amino acids, which change protein functions such as enzymatic activity. Oxidative reactive substances can produce reactive moieties in the protein structure that can interact with other peptides or molecules such as fatty acids and carbohydrate derivatives and lead to cross-linking of the protein with other macromolecules. These protein–macromolecule links are inactive biologically [10,11]. Hydroxyl radicals seem to play an important part in the formation of DNA–protein crosslinks in chromatin [12]. Moreover, reactive moieties of proteins 908
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
can be coupled to the damage of other biomolecules, such as lipids, leading to effects on physiological cell activities [13,14]. There is considerable evidence – primarily from the lethal effects of radioactive compounds that accumulate in the nucleus rather than cytoplasm or plasma membrane – suggesting that DNA is the primary target for cell damage from IR [15,16]. IR and ROS interactions can result in changes in the deoxyribose ring and base structures [17], intra- and interstrand DNA–DNA cross-links, DNA SSBs and DSBs, and DNA–protein cross-links [15]. Each of these changes is discussed in more detail below. Changes to DNA bases and sugars. Hydroxyl radicals react with nucleotide bases such as thymine and cytosine by either adding a double bond or abstracting a hydrogen atom. Both reactions lead to the production of reactive intermediates (Fig. 1). Damage to DNA bases can destabilize the bridge between the base and sugar, with loss of the base moiety and the formation of basic deoxyribose residues. Hydroxyl radicals can abstract hydrogen atoms from the sugar–phosphate backbone of DNA, which generates 2-deoxyribose radicals that attack molecular oxygen or thiols, leading to strand damage. Sugar radicals (Fig. 1) might result in the release of an unchanged base from DNA [18] and can also react with purine or pyrimidine residues on the same nucleotide to yield an interstrand nucleotide cross-link (Fig. 1) [19]. Free radicals can also attack the guanine base to produce a purine ring radical (Fig. 1). This process can lead to an open ring pyrimidine, resulting in the destruction of the guanine base [19,20]. Intra- and interstrand DNA–DNA cross-links. The coupling of methylcytosine or thymine radicals to a neighbouring purine base produces an intrastrand cross-link lesion [21]. Pyrimidine-derived radicals have a more important role than other secondary radicals in inducing intrastrand cross-links [20,22]. Furthermore, there is evidence that the repair of interstrand cross-links involves the generation of a DSB during the unhooking process. These lesions, therefore, are extremely cytotoxic and mutagenic. Finally, a nucleotide radical produced by IR can be covalently bound to the deoxyribose of the same nucleotide or to its neighbouring base to yield cyclonucleosides and nucleobase–nucleobase inter- or intrastrand cross-links, respectively. These alkyl radicals react with other bases, resulting in interstrand cross-linked DNA [23]. Biochemical studies have demonstrated that these lesions markedly block DNA replication and transcription if not repaired. The most cytotoxic DNA lesion is a DSB, and more specifically, a single unrepaired DSB. The religation of a DSB is crucial for cell survival, and SSBs and DSBs are mostly repaired by cellular repair machinery. If repaired improperly, a DSB can potentially result in chromosome aberration and might lead to genetic instability, mutation and chromosome rearrangements [24,25]. If DNA damage is effectively repaired and the cell is capable of proliferation, then the progeny of such an irradiated cell is normal. If the damage is not lethal and is misrepaired or unrepaired, however, then the progeny of a single irradiated cell would be expected to show radiation-induced genetic changes in all descendant cells. This can result in mutation and cell death. These lethal, mutagenic or clastogenic effects are outcomes of radiation-induced genetic instability (Fig. 2). DNA–protein cross-links. These are harmful to cells because protein immobilized on DNA strands can block normal DNA transcription and replication in cells.
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[()TD$FIG]
NH2
O
N
HN
HN
OH
OH
N H
O
H
CH 2
CH 3
H
N H
O
H
H
N H
O
Reviews KEYNOTE REVIEW
O
cytosine radical
thymine radicals
O P
O
base
H 2C
O
N H
NH
HO
H
N H
H
H
NH 2
N
O
P sugar radical
O
guanine radical
H
N NH
N P
O
NH2
N
HC O H
H
H O
H
O
P
NH 2
N NH
interstrand cross link guanine and sugar N N
O
H
N
N
NH2
Sugar O O
Sugar O
P
OH intrastand cross link cytosine-guanine Drug Discovery Today
FIGURE 1
Base and sugar radicals produced by IR.
Cellular defence system DNA damage occurs naturally by several means. Normal cellular metabolic processes generate ROS that can attack DNA to produce a variety of potential DNA lesions. The body is equipped with
several means of mitigating and removing the effects of free radicals on crucial molecules. These mechanisms constitute the cellular defence system, which acts before free radicals can attack molecules and which is involved in the repair of SSB and DSBs www.drugdiscoverytoday.com
909
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[()TD$FIG]
Ionizing radiation
reactive substances
Reviews KEYNOTE REVIEW
DNA-protein cross link DNA inter- and intra-cross link
SSB
DSB
(if it was not normally repaired) genomic instability
gene mutation chromosomal aberration sister chromatide exchange micronuclei formation
(if it was not normally repaired) outcomes
cancer
degenerative diseases death Drug Discovery Today
FIGURE 2
Different types of genomic instability induced by ionizing radiation and their outcomes.
through enzymatic pathways. The antioxidant and antioxidantrelated enzymes superoxide dismutase (SOD), glutathione peroxidase (GSH-Px) and catalase comprise the cellular defence mechanism responsible for the inhibition and mitigation of ROS attacks on biomolecules. SOD is the most effective of these enzymes in protecting cells against IR-induced damage [26]. Animal treatment with SOD [27], overexpression of SOD [28] or use of compounds that mimic SOD (e.g. Mn-porphyrin) [29] resulted in considerably radioprotective effects and reduced radiation-induced DNA damage. The most abundant cellular antioxidant is the tripeptide L-g-glutamyl-L-cysteinylglycine (GSH). GSH prevents oxidation either directly, by scavenging reactive oxygen radicals through its thiol group, or indirectly, by utilizing g-glutamyl transferase enzymes. The cellular GSH pool helps maintain the redox balance to protect against oxidative stress [30]. This plays a crucial part in the inhibition and recovery of cross-linked proteins and other macromolecules under oxidative stress. The radioprotective effects of exogenous antioxidants, such as melatonin and lycopene, have been attributed to increased GSH levels and increased g-glutamyl transferase activity within cells [31,32]. Chromatin DNA is tightly packaged in a complex with histone and non-histone proteins. This DNA–protein complex protects DNA from free radical attack [18,33]. The removal of DNA-bound protein from chromatin results in an increase of radiation-induced DSBs [34]. Histone lysine methylation has been linked to the activation of a series of DNA repair enzymes [33]. Endogenous repair systems are involved in the chromatin reorganization and repair of damaged bases and strand breaks. Damaged bases are excised by a combination of DNA glycosylases, 910
www.drugdiscoverytoday.com
which catalyse the hydrolysis of N-glycosylic bonds linked to chemically altered bases, and endonucleases, deoxyribosephosphodiesterases and exonucleases. The gap is finally filled by DNA polymerase and the strand is sealed by DNA ligase [35]. Major DNA repair pathways include the DNA non-homologous end-joining pathway, the homologous recombination pathway and response pathways such as the ataxia telangiectasia and Rad3related pathway. IR induces response pathways in which enzymes such as Ku, Xrcc4 and Rad51 are involved in DSB repair [36,37].
Exogenous radiation protective agents Several different mechanisms have been proposed for radioprotection through exogenous compounds, including direct scavenging of ROS, hydrogen donation to reactive free radicals, inducing and/ or altering the levels of endogenous enzymes for detoxifying ROS, increasing DNA stability, lowering the production of ROS by the induction of hypoxia with consumption of local oxygen, and enhancing the DNA damage repair pathway. The primary mechanism of action of radioprotective agents is ROS scavenging. Because genomic instability is mainly induced by ROS produced by IR, a compound with sufficient reactivity toward a ROS can intercept the ROS free radical before it has an opportunity to attack crucial molecules such as DNA. Thiol-containing compounds, in particular, have an excellent radioprotective effect. Cysteamine and dimethyl sulfoxide (DMSO) both reduce genomic instability, an effect mainly related to the inhibition of ROS. Cysteamine has a greater protective effect than DMSO because of its positive charge [38], which enables it to interact electrostatically with the negatively charged sugar–phosphate backbone
of the DNA. This interaction acts to position the thiol in a location where it can react with ROS before they can attack the DNA [39]. Other radioprotective thiol-containing synthetic compounds include Amifostine, an FDA-approved drug for the protection of patients against radiation-induced xerostomia, and WR1065, an active metabolite that protected cells from chromosomal damage and cell death [40]. Like the synthetic antioxidants and radioprotectors (e.g. amifostine and cysteamine), antioxidants derived from natural sources also protect against DNA damage and cell death induced by IR. Natural compounds such as vitamin E [41], melatonin [42] and herbal medicine (e.g. citrus and hawthorn extracts) [43,44] have also been shown to exert radioprotective effects by free radical scavenging. Radioprotective agents are categorized based on their mechanism of action (i.e. antioxidant and immunostimulator), origin (i.e. synthetic and natural, compounds) and/or chemical structure (i.e. thiol and bisbenzimidazole ring-containing compounds). This categorization is not exact, and radioprotectors can have overlapping mechanisms and structures. Because the use of radioprotective agents can be crucial in reducing side-effects induced by IR in patients undergoing radiotherapy or in personnel in radiological workplaces, many compounds have been evaluated as radioprotective agents in cell culture, animal and human experiments over the past 60 years. There are several excellent reviews to describe the history, mechanisms and types of these compounds with potential radioprotective properties [3–5,45,46].
Flavonoids Flavonoids are a family of natural products with a polyphenolic structure that are found in plants. More than 4000 flavonoids have been identified, many from fruits, vegetables and beverages (tea, coffee and fruit drinks). For example, apple contains the flavanols quercetin, rutin, epicatechin and catechin [47]. The daily intake of flavonoids is different in each country, reflecting differences in diet, especially the consumption of tea. High consumption of tea might be the most influential factor in total flavonoid intake in certain groups of people. This intake is high compared to the average daily intake of other dietary antioxidants such as vitamin C, vitamin E or carotenoids. Flavonoids have a unique structure based on three phenyl rings, A, B and C (Fig. 3). Ring B can bind to position 3 at fused ring C to form isoflavone. The preferred glycosylation site on flavonoids is the 3 position, and less frequently the 7 position, with glucose the most common sugar residue. Flavonoids are categorized into the following groups based on their chemical structure: flavonols (e.g. quercetin), flavones (e.g. apigenin), flavanones (e.g. hesperetin), isoflavones (e.g. genestein) and flavanols (e.g. catechin). The chemical structure of flavonoids are dependent on hydroxy group at position C2, C2 C3 double bond, 4-carbonyl group and steric conformation of B ring (Fig. 3). The biological effects of flavonoids depend upon their chemical structure. The position of hydroxyl groups and other features are important for their antioxidant and free-radical-scavenging effects. Because of their considerable healthrelated effects and their unique chemical structure, these bioactive compounds are the subjects of much medical and biological research as investigators try to find more biological properties and mechanisms. Extensive studies of flavonoids have shown that these compounds not only have antioxidant [48] and radioprotec-
REVIEWS
tive properties but also have antibacterial [49], inflammatory [50], antioxidant [48], anticancer [51] and antigenotoxic [52] activities. These biological effects of flavonoids are attributed mainly to their antioxidant effects. Whereas several human studies have shown ingestion of flavonoids to have health benefits in preventing cancer [53,54], other studies have not found this to be the case [55,56]. In contrast to their beneficial effects, some flavonoids have also been found to be mutagenic and genotoxic in vitro. Several studies reported that quercetin has genotoxic effects in cell culture and microbial assays [57–59]. By contrast, in vitro studies using Salmonella as a microbial assay did not indicate that quercetin had any mutagenic effects [60]. In addition, quercetin did not show any genotoxicity in an in vivo model [59]. Short- and long-term administration of a quercetin-supplemented diet in animals did not result in any carcinogenicity and genotoxicity [59,61]. Harwood et al. [62] recently reviewed in vitro and in vivo studies related to probable genotoxicity induced by quercetin and concluded that no toxicologically and mutagenically considerable adverse effects were reported when food-grade quercetin was added to foods at levels approximating intakes of naturally occurring quercetin in the diet of consumers with a high fruit and vegetable intake (i.e. 200–500 mg/day). Quercetin, therefore, has been classified as noncarcinogenic to humans and is safe for human usage, with quercetin supplements commercially available [63]. Genistein, a phyto-oestrogen present in high levels in soybean, is a potent inhibitor of type II topoisomerases [64]. Type II topoisomerases are enzymes that unwind DNA and generate DNA strand breaks that have the potential to fragment the genome at every cycle of their action. Although topoisomerases are essential for the survival and proliferation of cells, they also have considerable genotoxic effects. Genistein binds to and stabilizes the topoisomerase– DNA complex, inhibiting religation and resulting in DNA strand breaks [65]. The potentially DNA-damaging or mutagenic effects of genistein that have been suggested in some studies in vitro [64,66] are not likely to occur in vivo, even at high dietary levels of genistein [67]. This paradoxical effect of genistein is partly related to its dosage level. In 1999, the US FDA recommended the daily ingestion of 25 g of total soy protein [67]. Whereas flavonoids generally demonstrate antioxidant activity, they can shift to prooxidant activity in the presence of transition metals. In addition, although flavonoids directly react with free radicals as antioxidants, chelation of metal ions might involve the production of ROS. Flavonoids can bind to transitional metals ions such as Cu(II) and Fe(III), which are the most redox-active metals in living cells. All types of flavonoids possess three domains able to react with metal ions: the 3,4-dihydroxy system located on ring B, the 3-hydroxy or 5-hydroxy groups of ring C and the 4-carbonyl group in ring C [68] (Fig. 3). The reduction of Cu2+ and Fe3+ to Cu+ and Fe2+, respectively, by phenolic compounds can form superoxide radical anions by a single electron reduction of the oxygen molecule. The superoxide radical is, in turn, converted to hydrogen peroxide and a hydroxyl radical, causing the formation of a DNA base adduct [69]. This prooxidation activity is responsible for the ability of flavonoids to promote cellular toxicity mainly through DNA damage [70]. Copper and zinc are the major metals naturally associated closely with chromosomes, so the presence of Cu2+ rather than that of Fe3+ is more considerable for DNA damage produced by the prooxidation of flavonoids [71]. The location of www.drugdiscoverytoday.com
911
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[()TD$FIG]
O
5
4
6
A
C
7
3 2
catechol
2` 1`
O 1
8
Reviews KEYNOTE REVIEW
benzopyran
3`
B
4`
6` 5`
OH
OH
Flavonoid class
base structure
example
O
Flavonol
O OH
OH
O
O
OH
quercetin O OH
Flavone
OH
O
O O
HO
apigenin OH
O
OH
O
Flavanone
OH
O O
OH
rutinose (sugar) hesperidin
OCH 3
O OH
Isoflavone
OH
O
O O HO
genistein OH OH
Flavanol
OH O HO
O OH
epicatechin OH Drug Discovery Today
FIGURE 3
Classification, structure and example of the main classes of flavonoids.
redox-active metals is important because they generate hydroxyl radicals and other ROS with very short life spans and penetration through the medium. To have a DNA-damaging effect, these reactive substances must be produced in close proximity to 912
www.drugdiscoverytoday.com
DNA. Both the antioxidant and the copper-mediated prooxidant activities of polyphenolic compounds depend on the number and position of hydroxyl groups [70]. Quercetin undergoes more autoxidation and causes more DNA damage because of the reduction of
Cu(II) to Cu(I) compared to kaempferol and morin. Quercetin binds to Cu(II) in close proximity to DNA to form a DNA–Cu(II)– quercetin complex. Cu(I) and superoxide anions are generated through the autoxidation of the quercetin–Cu(II) complex, resulting in the formation of hydroxyl radicals and hydrogen peroxide that cause oxidation and DNA damage. In the presence of quercetin or Cu(II) alone, no DNA cleavage was observed [72]. The flavonoid–Cu(II)–DNA complex is proposed to be responsible for the carcinogenic and mutagenic effects of flavonoids owing to its ability to generate oxidative DNA damage. These experiments were mainly conducted using isolated cellular DNA, however, and care should be taken in the extrapolation of these results to humans. Because cancer cells have a higher intracellular copper level, it has been suggested that the flavonoid–Cu(II) complex has more cytotoxic effects in cancer cells than in normal cells [71]. Genistein–metal (Zn2+, Cu2+, Mn2+, Ni2+ and Co2+) complexes also had anti-proliferative effects against cancer cells. These complexes inhibited cell growth more than free isoflavone and corresponding metal ions and exhibited a statistically G2/M-phase arrest for human cancer cell lines [73]. The paradox of the opposing antioxidant and DNA-damaging activities of these flavonoid–metal complexes is further discussed later in this review.
Structure–activity relationships of flavonoids and antioxidant activity Flavonoids have low oxidation potential (+0.125 V and +0.235 V for quercetin and morin, respectively), which means they are easily reduced by ROS and so have greater free-radical-scavenging abilities than high oxidation potential value [74]. The antioxidant activities of flavonoids vary considerably depending on different backbone structures and functional groups and, likewise, the difference in their ROS scavenging can be accounted for by the variation in the number and type of functional groups attached to the main nucleus. There are several functional groups that contribute to increased ROS scavenging. These include the C2 C3 double bond, the adjacent conjugate with the 4-carbonyl group in ring C and the catechol moiety formed by the 40 ,50 -dihydroxy structure of ring B (Fig. 3). The additional presence of a 3- and/or 5hydroxyl group on rings C and A also increases the efficacy of ROS scavenging [75]. The position and number of hydroxyl groups have an important role in antioxidant activity. It is generally accepted that an increase in the number of hydroxyl groups increases the antioxidant activity of flavonoids; for some compounds, however, the presence of a methoxyl group contributes to negative antioxidant activity and is more important for reducing this effect than increasing the number of single hydroxyl groups on ring B [76]. The ROS-scavenging activities might be closely related to the position rather than the numbers of phenolic hydroxyl groups because the activity of an ortho-dihydroxyl group is much higher than that of a meta-dihydroxyl group [73]. The antioxidant activity of flavonoids depends, in part, on their ability to delocalize electron distribution, resulting in a more stable phenoxyl radical. The stability of this radical is related to the structural planarity of flavonoids between ring B and the benzopyrane nucleolus, which enables resonance effects [75]. When flavonoids react with free radicals, the phenoxyl radicals produced are stabilized by resonance effects of aromatic nucleus
REVIEWS
[75]. The presence of a hydroxyl group at the 3 position on ring C increases the planarity, whereas the presence of a methoxyl group at the 3 position inhibits this structural planarity owing to steric hindrance, resulting in lower antioxidant activity [76]. The unsaturation of ring C and subsequent conjugation with the 4-OXO group enables electron delocalization across the molecule that stabilizes the aryloxyl radical [75]. Retention of the catechol type structure in ring B and removal of the C2 C3 double bond in ring C eliminates delocalization of electrons from the aryloxyl radical on ring B to ring A. The antioxidant activity of flavonoids is also affected by glycosylation, which reduces their activity. For example, rutin, a glycone quercetin, has lower antioxidant activity than unglycosylated quercetin [77]. Qurecetin contains almost every functional group required for antioxidant activity and so is more potent than other flavonoids such as catechin and hesperidin, which lack some of these functional groups in their structures [75].
In vitro and in vivo experiments on the effects of flavonoids on radiation-induced genetic damage Two different aspects of the biological effects of flavonoids have been assessed: increased tumour cell death and preventive effects in normal cells during co-administration of flavonoids and IR. Several studies have shown different effects of flavonoids on normal and cancer cells during radiation exposure. Genistein and quercetin are known to have multiple cellular effects such as protection of normal cells from DNA damage, arrest of the cell cycle at the G2/M phase in human cancer cell lines, induced apoptosis and cell death [78,79]. Quercetin-induced apoptosis in tumour cells is associated with caspase activation and suppression of survivin and Bcl-2 [80,81]. In animal models, genistein has immunomodulating and radioprotective properties. Administration of genistein considerably increases the survival rate in gamma-irradiated normal mice by accelerating neutrophil and platelet recovery and hematopoietic progenitor cell reconstitution [82,83]. In different cancer cells, low concentrations of genisteininduced expression of the major cell cycle inhibitory proteins (p53, p21, Bcl-2 and survivin) lead to an increase in radiationinduced genotoxicity. Genistein also stimulates irradiationinduced intracellular ROS production in cancer cells and induced additional apoptosis when cells were subjected to IR [84–86]. This radiosensitivity of genistein is beneficial for the protection of normal cells yet detrimental to cancer cells in patients undergoing radiotherapy. Epigallocatechin-3-gallate (EGCG), a major antioxidant polyphenol found in green tea, has anticarcinogenic and antigenotoxic properties. This polyphenol sensitized the glioblastoma cancer cell response to IR. EGCG antagonized the IR-induced expression of survivin, a cell survival protein, resulting in the arrest of the cell cycle at the G2/M phase and inhibition of cell growth [87]. The concentration of EGCG is probably a key factor in determining whether EGCG serves to protect or to damage tumour cells during cancer therapies because it induces apoptosis at lower concentrations, whereas it does not accelerate ROS formation. EGCG is able to protect salivary gland cells from damage caused by radiation therapy; however, EGCG might also protect tumour cells during radiotherapy when EGCG concentrations are at physiological levels [88]. In summary, flavonoids can induce an increase in tumour cell toxicity during exposure to IR. This effect is not related to ROS scavenging but is associated with the supwww.drugdiscoverytoday.com
913
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
pression of cellular proteins such as survivin and with the promotion of cell cycle arrest at the G2/M phase. These effects cumulatively enhance cell apoptosis induced by IR. Flavonoids similar to nutritional antioxidants such as vitamin E have beneficial effects on radiation-induced cell damage [89]. Flavonoids have been widely shown to have a protective effect on genomic stability. Several chromosomal and genomic methods have been used to assess these protective effects and are discussed below. Evidence for the formation of IR-induced strand breaks in plasmid DNA comes from evaluation of the degree of DNA supercoiling assessed by gel electrophoresis. Pretreatment of plasmid DNA with quercetin [90] or epicatechin [91] decreases the toxic effects of gamma irradiation by inhibiting DNA strand breaks. The comet assay, or single-cell gel electrophoresis technique, is a sensitive method that measures SSBs of DNA for the evaluation of genotoxicity induced by IR or toxic substances. Treatment of irradiated mice with flavonoids considerably decreases the level of primary DNA damage (SSBs). This protective effect is observed for naringin, chrysin [92], epicatechin [91] and troxerutin [93]. Micronuclei (Mn) formation, which assesses the degree of DSBs in DNA, is one of the best assays for the evaluation of genome instability induced by IR. This method is performed in vitro (e.g. in human cultured lymphocytes) and in vivo (e.g. in mouse bone marrow cells and peripheral blood leucocytes). Higher genome instability results in a higher frequency of Mn. Pretreatment of human culture lymphocytes with flavonoids statistically significant reduced the Mn frequencies that were increased by IR. Quercetin [90], hesperidin [94,95], swertisin [96] and troxerutin [93] statistically significant decrease the frequency of Mn induced by IR in human binucleated lymphocytes, an effect potentially leading to reduced genomic damage. These flavonoids did not increase the frequency of Mn in non-irradiated human lymphocyte cultures treated at IR protective doses. Flavonoids including troxerutin [93], naringin [97] and hesperidin [98] have protective effects against genotoxicity induced by IR in mice, as assessed by Mn assay. It is clear that flavonoids reduce cell death owing to their antioxidant effects and protective effects on biomolecules such as DNA. In animal studies, administration of several flavonoids such as morin [99] and genistein [100] statistically significant reduced mortality rates induced by IR. These effects are thought to be related to the protective effects of flavonoids and maintenance or recovery of levels of crucial cell populations such as platelets, white blood cells and hematopoietic progenitor cells, which are reduced by IR [5].
Mechanisms related to flavonoids and genomic stability Free radical scavenging Because approximately 65% of DNA damage is caused by the indirect effect of free radicals such as HO that are produced from the radiolysis of surrounding water molecules, a major role of flavonoids is scavenging these free radicals. Their protective effect in cells is attributed to the inhibition of ROS before they are able to attack DNA and other macromolecules. Because ROS are mostly short-lived, they have to react with DNA immediately after their production by IR [101]. Flavonoids can donate a proton from their hydroxyl groups to the free radical, resulting in free radical repair and an inert molecule. The phenoxy radicals produced are stabi914
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
lized by delocalization of the unpaired electron within the aromatic structure. The o-dihydroxy group in ring B and 3-hydroxy along with 4-OXO and C2 C3 double bond on benzopyrane ring attribute particularly for stabilization of phenoxy radical induced by ROS. Therefore, these polyphenols are excellent scavengers of free radicals because of the high reactivity of their hydroxyl substituents and the resonance effects of their unique polyphenyl [102]. It is estimated that high in vitro antioxidant activity is also related to high radioprotective effects for polyphenols. Grape seed extract, which contains high levels of poyphenolic compounds, has a greater ABTS (2,20 -azinobis(3-ethylbenzothiazoline-6-sulphonic acid) diammonium salt) radical-scavenging capacity and more radioprotective effects than rutin, DMSO or vitamin C [103]. The stable free radical diphenyl picrylhydrazyl (DPPH) is used for the in vitro evaluation of free radical scavenging. The free-radicalscavenging abilities and radioprotective effects of flavonoids do not necessarily seem to be linked. Swertosin, a flavonoid that lacks O-dihydroxy groups in ring B but has c-glycosylation and a 3hydroxyl group in ring C, had the lowest antioxidant activity in vitro by the DPPH method but had greater radioprotective effects in human culture lymphocyte than other compounds with higher free-radical-scavenging abilities [96]. Interestingly, flavonoids can pass through the plasma membrane to chelate and remove intercellular redox-active iron. This inhibits the H2O2-induced oxidative process mediated by iron [104] and protects against DNA damage induced by oxidative stress. This mechanism might contribute to the different antioxidant activities of flavonoids in in vitro assays such as the DPPH method and in cell culture.
Endogenous enzymes Because IR induces a reduction in the level of cellular enzymatic antioxidants, the increase in the antioxidant status during flavonoid pretreatment of cells might further decrease the attack of free radicals on biomolecules such as DNA and membrane lipids, thereby decreasing the deleterious effects of radiation on cells and tissues. Administration of grape seed extracts rich in proanthocyanidins attenuated radiation-induced oxidative stress in mouse tissue by statistically significant increasing the activity of SOD, catalase and GPx [105]. Likewise, administration of quercetin, hesperidin and morin before the IR resulted in an increase in non-enzymatic and enzymatic antioxidant status, indicating that these polyphenolic pretreatments restored the endogenous antioxidant capacity to near normal levels [90,99,106]. SOD contains metal cofactors such as copper and zinc. Some ligand–metal complexes can mimic SOD activity and have radioprotective effects [29]. Although it is established that flavonoid– metal complexes can cause DNA damage through the prooxidative process, flavonoid–metal complexes can also scavenge superoxide radicals [107]. The flavonoid–copper complexes studied mimicked SOD activity and scavenged superoxide radicals produced by nitro blue tetrazolium in vitro more effectively than the flavonoid–iron complex [108]. There is little known about this SOD-mimicking activity, and more biological studies are needed to elucidate the exact mechanism by which this occurs.
Flavonoid–DNA interaction The complex between histones and DNA results in condensation that protects mammalian DNA from strand breaks mostly induced
by the indirect, rather than the direct, effects of IR [34]. DNA strand breaks are increased by removing histones from chromatin because this gives free radicals increased access to DNA [34,109]. Replicating DNA results open chromatin structure with fewer attached histone proteins, which results in more DBS and DNA damage. The bisbenzimidazole derivative Hoechst 33342 [20 -(4methoxyphenyl)-5-(4-methyl-1-piperazinyl)-2,50 -bi-benzimidazole] and its analogs are DNA minor groove binding ligands that, when administered before IR, have protective effects on DNA lesions [110,111]. Hoechst compound contains a planar structure formed by its aromatic groups (two benzimidazole groups and one phenyl group), is positively charged and has an arc shape that ‘fits’ in the minor groove of DNA. The potent radioprotective effects of Hoechst compounds are attributable to this DNA–Hoechst interaction. A methylproamine analog of Hoechst had radioprotective effects approximately 100-fold greater than the classical aminothiol radioprotector, WR1065, possibly because of a stronger interaction with DNA through its positively charged amine group [112]. The interaction of flavonoids with DNA is one of the main mechanisms for protecting DNA from oxidative stress such as IR. Flavonoid intercalation into DNA, which has been established by spectroscopic and electrochemical methods [113–116], stabilizes the DNA structure and enables the flavonoid to react with nearby free radicals. Binding of flavonoids to DNA has been reported for quercetin, rutin, EGCG, morin and kaempferol. Free-energy calculations of binding thermodynamics revealed that binding occurs spontaneously with binding constants on the order of 103 M1 and does not require energy [117]. Flavonoid intercalation also results in the deformation of the DNA double-helical structure [113]. Flavonoids bind to the major (adenine and guanine) and minor (thymine) grooves of DNA bases and to the backbone phosphate groups. Positively charged flavonoids such as delphinidin have a more stabilizing effect on the DNA duplex than neutral flavonoids such as quercetin or kaempferol owing to a stronger electrostatic interaction with the DNA backbone phosphates [118,119]. When this interaction was enhanced by the addition of a tertiary amine at position 7, binding was increased 1000-fold [117]. Flavonoids lacking positively charged side groups interact with the phosphate backbone of DNA primarily by hydrogen bonding through their hydroxyl groups [114,118,119]. The planar and aromatic moieties of flavonoids also contribute to DNA intercalation. The electronic properties of the flavonoid rings induce hydrophobic and aromatic p–p stacking interactions with DNA bases [74,113], and rings A and C are fused to provide a planar molecule that can intercalate between the stacked nucleic acid bases [117]. The planar benzopyran-4-one portion is probably localized and intercalated in the double-stranded DNA, more or less parallel to the adjacent planes of nitrogenous bases because of its hydrophobic nature. This hydrophobic force plays the most important part in intercalation of flavonoids into DNA [120,121]. By contrast, the coplanar catecholic group (ring B) is more likely to be oriented toward the external medium where it can form hydrogen bonds with phosphate groups in the DNA backbone [122]. The benzopyran-4-one portion of quercetin does not interact electrostatically with DNA because it has a higher affinity for positively charged structures at physiological pH, whereas DNA is mainly negatively charged [122,123]. The non-planar and hydrophilic
REVIEWS
flavonol, dihydroquercetin, which is saturated across the C2 C3 double bond, has less affinity for internal hydrophobic interactions with DNA. Naringin, a flavonoid without a C2 C3 double bond, does not interact with DNA alone but is capable of interacting with the nitrogenous bases in the presence of Cu(II), with which it forms a complex [124]. Hydroxyl groups were found to enhance DNA binding. Hydroxylation at the 7 position, as in 7hydroxy flavone and 5,7-dihyroxy flavone, was identified as the most important for flavone–DNA interaction, whereas a methoxyl group at the 7 position reduced the binding constant [117]. There are several proposed mechanisms for protection against DNA damage conferred by flavonoids that are related to flavonoid– DNA interactions. First, intercalation of flavonoids into DNA double helices induces stabilization of DNA helical structure and condensation of DNA to a highly compact form that is less susceptible to attack by free radicals. Second, flavonoids have several hydroxyl groups, which easily donate hydrogen atoms to radicals formed in the DNA bases, as the flavonoid–DNA complex enables the flavonoid to repair the radical base very efficiently. Third, flavonoids can interact with the phosphate moiety of the DNA backbone through hydrogen bonding. The repair of sugar radicals is attributed to hydrogen donation from flavonoids through this bonding. Fourth, flavonoids act as reducing agents owing to their low oxidation potential and react easily with electron-accepting radicals before the ROS can attack DNA components. Complexes between flavonoids and metal ions intercalated into DNA with a higher affinity than flavonoid alone, resulting in more DNA cleavage. This type of DNA damage was observed with complexes formed between flavonoids and copper, manganese or zinc, as previously discussed [125–127].
Induction of hypoxia The induction of cellular hypoxia is related to the consumption of local oxygen. The production of ROS is decreased under hypoxia. This mechanism contributes to the protective effects of compounds against oxidative stress [128,129]. Several studies reported that flavonoids can induce expression of hypoxia-induced transcription factor-1 (HIF-1), a protein produced under hypoxia. HIF1 is a key mediator of the cellular response to tissue hypoxia and acts to promote tissue survival at low oxygen levels [130]. The induction of HIF-1 expression results in reduced ROS [131] and delays the progression of cells past the G1 phase of the cell cycle, causing cell cycle arrest and leading to the inhibition of cell growth [130,131]. There is also a relationship between HIF-1 and iron because iron has a major role in the regulation of HIF-1. Flavonoids chelate cellular iron, resulting in the depletion of intracellular iron, and might be responsible for the reduction in ROS in response to iron-mediated oxidation [130]. Although there is no clear mechanism for the induction of HIF-1 expression by flavonoids, in quercetin and myricetin the catechol moieties were largely responsible for the observed induction [131]. More experiments in this field are required to determine whether HIF-1 induction is related directly to hypoxia or to the iron-chelating effect of flavonoids. The induction of HIF-1 and the hypoxiaresponse pathway by flavonoids results in a reduction of ROS levels, by which flavonoids confer their protective effects against genotoxicity induced by ROS. www.drugdiscoverytoday.com
915
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Concluding remarks
Reviews KEYNOTE REVIEW
IR produces ROS that attack and damage several targets, most crucially DNA. DNA damage results in genomic instability and mutation and/or cell death if the damage is not properly repaired. Flavonoids, by virtue of their chemical structure, protect against genomic instability induced by ROS. These compounds have prooxidant activity and can cause DNA cleavage in vitro that is increased in the presence of copper. Although there is no research to suggest that this deleterious effect of flavonoids occurs in vivo in normal cells, co-exposure of cancer cells to IR and to flavonoids results in enhanced radiation damage, with flavonoids acting as radiosensitizers. This effect might be beneficial to patients undergoing radiotherapy. Several in vivo and in vitro studies have shown flavonoids to have protective effects against IR-induced genotoxicity. There are several mechanisms for the mitigation of genomic instability induced by IR, including free radical scavenging, increased expression of endogenous enzymes, stabilization of
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
the DNA double helix into a form less susceptible to damage and reduction of ROS levels by the induction of hypoxia. These effects of flavonoids are completely dependent on their structure, especially the number and position of hydroxyl groups and the C2 C3 double bond. Because flavonoids have a unique structure, these bioactive molecules have great potential as lead compounds for the design of new, more active compounds for ROS scavenging and DNA intercalation. These pharmacological properties might be useful in cancer treatment as novel anticancer or radio- and chemoprotective agents for patients during radiotherapy.
Acknowledgements This study was supported by a grant from Mazandaran University of Medical Sciences, Sari, Iran. The author thanks Dr Vladimir Tolmachev and his department members from Uppsala University, who have an excellent scientific atmosphere for the preparation of this review.
References 1 Schmidt-Ullrich, R.K. (2003) Molecular targets in radiation oncology. Oncogene 22, 5730–5733 2 Harborne, J.B. and Williams, C.A. (2000) Advances in flavonoid research since 1992. Phytochemistry 55, 481–504 3 Weiss, J.F. and Landauer, M.R. (2003) Protection against radiation by antioxidant nutrients and phytochemicals. Toxicology 189, 1–20 4 Weiss, J.F. and Landauer, M.R. (2009) History and development of radiationprotective agents. Int. J. Radiat. Biol. 85, 539–573 5 Hosseinimehr, S.J. (2007) Trends in the development of radioprotective agents. Drug Discov. Today 12, 794–805 6 Mikkelsen, R.B. and Wardman, P. (2003) Biological chemistry of reactive oxygen and nitrogen and radiation-induced signal transduction mechanisms. Oncogene 22, 5734–5754 7 Zhao, W. et al. (2007) Oxidative damage pathways in relation to normal tissue injury. Br. J. Radiol. 80, S23–S31 8 Gunther, M.R. et al. (2002) Nitric oxide trapping of the tyrosyl radical – chemistry and biochemistry. Toxicology 177, 1–9 9 Xu, G. et al. (2003) Radiolytic modification of basic amino acid residues in peptides: probes for examining protein–protein interactions. Anal. Chem. 75, 6995–7007 10 Stadtman, E.R. and Levine, R.L. (2003) Free radical-mediated oxidation of free amino acids and amino acid residues in proteins. Amino Acids 25, 207–218 11 Garrison, W.M. (1987) Reaction mechanisms in the radiolysis of peptides, polypeptides, and proteins. Chem. Rev. 87, 381–398 12 Dizdaroglu, M. and Gajewski, E. (1989) Structure and mechanism of hydroxyl radical-induced formation of a DNA–protein cross-link involving thymine and lysine in nucleohistone. Cancer Res. 49, 3463–3467 13 Edwards, J.C. et al. (1984) The effects of ionizing radiation on biomembrane structure and function. Prog. Biophys. Mol. Biol. 43, 71–93 14 Torreggiani, A. et al. (2006) Investigation of radical-based damage of RNase A in aqueous solution and lipid vesicles. Biopolymers 81, 39–50 15 Iyer, R. and Lehnert, B.E. (2000) Effects of ionizing radiation in targeted and nontargeted cells. Arch. Biochem. Biophys. 376, 14–25 16 Hutchinson, F. (1966) The molecular basis for radiation effects on cells. Cancer Res. 26, 2045–2052 17 Evans, M.D. et al. (2004) Oxidative DNA damage and disease: induction, repair and significance. Mutat. Res. 567, 1–61 18 Roginskaya, M. et al. (2006) Protection of DNA against direct radiation damage by complex formation with positively charged polypeptides. Radiat. Res. 166, 9–18 19 Gates, K.S. (2009) An overview of chemical processes that damage cellular DNA: spontaneous hydrolysis, alkylation, and reactions with radicals. Chem. Res. Toxicol. 22, 1747–1760 20 Zhang, Q. and Wang, Y. (2005) The reactivity of the 5-hydroxy-5,6dihydrothymidin-6-yl radical in oligodeoxyribonucleotides. Chem. Res. Toxicol. 18, 1897–1906 21 Hong, H. et al. (2007) Formation and genotoxicity of a guanine–cytosine intrastrand cross-link lesion in vivo. Nucleic Acids Res. 35, 7118–7127
916
www.drugdiscoverytoday.com
22 Jiang, Y. et al. (2007) In vivo formation and in vitro replication of a guanine– thymine intrastrand cross-link lesion. Biochemistry 46, 12757–12763 23 Wang, Y. (2008) Bulky DNA lesions induced by reactive oxygen species. Chem. Res. Toxicol. 21, 276–281 24 Obe, G. et al. (2002) Chromosomal aberrations: formation, identification and distribution. Mutat. Res. 504, 17–36 25 Dextraze, M.E. et al. (2010) DNA interstrand cross-links induced by ionizing radiation: an unsung lesion. Mutat Res. 26 Sun, J. et al. (1998) Role of antioxidant enzymes on ionizing radiation resistance. Free Radic. Biol. Med. 24, 586–593 27 Molla, M. et al. (2005) Protective effect of superoxide dismutase in radiationinduced intestinal inflammation. Int. J. Radiat. Oncol. Biol. Phys. 61, 1159–1166 28 Veldwijk, M.R. et al. (2009) Normal-tissue radioprotection by overexpression of the copper–zinc and manganese superoxide dismutase genes. Strahlenther. Onkol. 185, 517–523 29 Gauter-Fleckenstein, B. et al. (2008) Comparison of two Mn porphyrin-based mimics of superoxide dismutase in pulmonary radioprotection. Free Radic. Biol. Med. 44, 982–989 30 Okunieff, P. et al. (2008) Antioxidants reduce consequences of radiation exposure. Adv. Exp. Med. Biol. 614, 165–178 31 El-Missiry, M.A. et al. (2007) Ameliorative effect of melatonin against gammairradiation-induced oxidative stress and tissue injury. Ecotoxicol. Environ. Saf. 66, 278–286 32 Srinivasan, M. et al. (2009) Lycopene: an antioxidant and radioprotector against gamma-radiation-induced cellular damages in cultured human lymphocytes. Toxicology 262, 43–49 33 Misri, S. et al. (2008) Telomeres, histone code, and DNA damage response. Cytogenet. Genome Res. 122, 297–307 34 Sak, A. et al. (2000) Protection of DNA from radiation-induced double-strand breaks: influence of replication and nuclear proteins. Int. J. Radiat. Biol. 76, 749–756 35 Leadon, S.A. (1996) Repair of DNA damage produced by ionizing radiation: a minireview. Semin. Radiat. Oncol. 6, 295–305 36 Karagiannis, T.C. and El-Osta, A. (2004) Double-strand breaks: signaling pathways and repair mechanisms. Cell. Mol. Life Sci. 61, 2137–2147 37 van Veelen, L.R. et al. (2005) Analysis of ionizing radiation-induced foci of DNA damage repair proteins. Mutat. Res. 574, 22–33 38 Limoli, C.L. et al. (2001) Attenuation of radiation-induced genomic instability by free radical scavengers and cellular proliferation. Free Radic. Biol. Med. 31, 10–19 39 Limon-Pacheco, J. and Gonsebatt, M.E. (2009) The role of antioxidants and antioxidant-related enzymes in protective responses to environmentally induced oxidative stress. Mutat. Res. 674, 137–147 40 Dziegielewski, J. et al. (2008) WR-1065, the active metabolite of amifostine, mitigates radiation-induced delayed genomic instability. Free Radic. Biol. Med. 45, 1674–1681 41 Singh, V.K. et al. (2009) Tocopherol succinate: a promising radiation countermeasure. Int. Immunopharmacol. 9, 1423–1430
42 Shirazi, A. et al. (2007) A radiobiological review on melatonin: a novel radioprotector. J. Radiat. Res. (Tokyo) 48, 263–272 43 Hosseinimehr, S.J. et al. (2003) Radioprotective effects of citrus extract against gamma-irradiation in mouse bone marrow cells. J. Radiat. Res. (Tokyo) 44, 237–241 44 Hosseinimehr, S.J. et al. (2007) Radioprotective effects of hawthorn fruit extract against gamma irradiation in mouse bone marrow cells. J. Radiat. Res. (Tokyo) 48, 63–68 45 Hosseinimehr, S.J. (2009) Potential utility of radioprotective agents in the practice of nuclear medicine. Cancer Biother. Radiopharm. 24, 723–731 46 Dumont, F. et al. (2010) Radiation countermeasure agents: an update. Expert Opin. Ther. Pat. 20, 73–101 47 Lotito, S.B. and Frei, B. (2004) The increase in human plasma antioxidant capacity after apple consumption is due to the metabolic effect of fructose on urate, not apple-derived antioxidant flavonoids. Free Radic. Biol. Med. 37, 251–258 48 Pietta, P.G. (2000) Flavonoids as antioxidants. J. Nat. Prod. 63, 1035–1042 49 Jeong, K.W. et al. (2009) Screening of flavonoids as candidate antibiotics against Enterococcus faecalis. J. Nat. Prod. 72, 719–724 50 Guardia, T. et al. (2001) Anti-inflammatory properties of plant flavonoids. Effects of rutin, quercetin and hesperidin on adjuvant arthritis in rat. Farmaco 56, 683–687 51 Plochmann, K. et al. (2007) Structure–activity relationships of flavonoid-induced cytotoxicity on human leukemia cells. Arch. Biochem. Biophys. 460, 1–9 52 Ahmadi, A. et al. (2008) Chemoprotective effects of hesperidin against genotoxicity induced by cyclophosphamide in mice bone marrow cells. Arch. Pharm. Res. 31, 794–797 53 Nothlings, U. et al. (2007) Flavonols and pancreatic cancer risk: the multiethnic cohort study. Am. J. Epidemiol. 166, 924–931 54 Gates, M.A. et al. (2007) A prospective study of dietary flavonoid intake and incidence of epithelial ovarian cancer. Int. J. Cancer 121, 2225–2232 55 Wang, L. et al. (2009) Dietary intake of selected flavonols, flavones, and flavonoidrich foods and risk of cancer in middle-aged and older women. Am. J. Clin. Nutr. 89, 905–912 56 Bobe, G. et al. (2009) Flavonoid consumption and esophageal cancer among black and white men in the United States. Int. J. Cancer 125, 1147–1154 57 van der Hoeven, J.C. et al. (1984) Genotoxicity of quercetin in cultured mammalian cells. Mutat. Res. 136, 9–21 58 Rueff, J. et al. (1992) Oxygen species and the genotoxicity of quercetin. Mutat. Res. 265, 75–81 59 Caria, H. et al. (1995) Genotoxicity of quercetin in the micronucleus assay in mouse bone marrow erythrocytes, human lymphocytes, V79 cell line and identification of kinetochore-containing (CREST staining) micronuclei in human lymphocytes. Mutat. Res. 343, 85–94 60 Czeczot, H. and Kusztelak, J. (1993) A study of the genotoxic potential of flavonoids using short-term bacterial assays. Acta Biochim. Pol. 40, 549–554 61 Utesch, D. et al. (2008) Evaluation of the potential in vivo genotoxicity of quercetin. Mutat. Res. 654, 38–44 62 Harwood, M. et al. (2007) A critical review of the data related to the safety of quercetin and lack of evidence of in vivo toxicity, including lack of genotoxic/ carcinogenic properties. Food Chem. Toxicol. 45, 2179–2205 63 Okamoto, T. (2005) Safety of quercetin for clinical application. Int. J. Mol. Med. 16, 275–278 64 Michael McClain, R. et al. (2006) Genetic toxicity studies with genistein. Food Chem. Toxicol. 44, 42–55 65 McClendon, A.K. and Osheroff, N. (2007) DNA topoisomerase II, genotoxicity, and cancer. Mutat. Res. 623, 83–97 66 Di Virgilio, A.L. et al. (2004) Genotoxicity of the isoflavones genistein, daidzein and equol in V79 cells. Toxicol. Lett. 151, 151–162 67 Klein, C.B. and King, A.A. (2007) Genistein genotoxicity: critical considerations of in vitro exposure dose. Toxicol. Appl. Pharmacol. 224, 1–11 68 Azam, S. et al. (2004) Prooxidant property of green tea polyphenols epicatechin and epigallocatechin-3-gallate: implications for anticancer properties. Toxicol. In Vitro 18, 555–561 69 El Amrani, F.B. et al. (2006) Oxidative DNA cleavage induced by an iron(III) flavonoid complex: synthesis, crystal structure and characterization of chlorobis(flavonolato)(methanol) iron(III) complex. J. Inorg. Biochem. 100, 1208–1218 70 Cao, G. et al. (1997) Antioxidant and prooxidant behavior of flavonoids: structure– activity relationships. Free Radic. Biol. Med. 22, 749–760 71 Hadi, S.M. et al. (2007) Oxidative breakage of cellular DNA by plant polyphenols: a putative mechanism for anticancer properties. Semin. Cancer Biol. 17, 370–376 72 Yamashita, N. et al. (1999) Mechanism of oxidative DNA damage induced by quercetin in the presence of Cu(II). Mutat. Res. 425, 107–115
REVIEWS
73 Chen, J.W. et al. (2002) Structure–activity relationship of natural flavonoids in hydroxyl radical-scavenging effects. Acta Pharmacol. Sin. 23, 667–672 74 Zheng, X. et al. (2006) Electrochemical studies of ()-epigallocatechin gallate and its interaction with DNA. Anal. Bioanal. Chem. 386, 1913–1919 75 Rice-Evans, C.A. et al. (1996) Structure–antioxidant activity relationships of flavonoids and phenolic acids. Free Radic. Biol. Med. 20, 933–956 76 Dugas, A.J., Jr et al. (2000) Evaluation of the total peroxyl radical-scavenging capacity of flavonoids: structure–activity relationships. J. Nat. Prod. 63, 327–331 77 Heim, K.E. et al. (2002) Flavonoid antioxidants: chemistry, metabolism and structure–activity relationships. J. Nutr. Biochem. 13, 572–584 78 Kyle, E. et al. (1997) Genistein-induced apoptosis of prostate cancer cells is preceded by a specific decrease in focal adhesion kinase activity. Mol. Pharmacol. 51, 193–200 79 Lee, T.J. et al. (2006) Quercetin arrests G2/M phase and induces caspase-dependent cell death in U937 cells. Cancer Lett. 240, 234–242 80 Tan, J. et al. (2009) Regulation of survivin and Bcl-2 in HepG2 cell apoptosis induced by quercetin. Chem. Biodivers. 6, 1101–1110 81 Siegelin, M.D. et al. (2009) Quercetin promotes degradation of survivin and thereby enhances death-receptor-mediated apoptosis in glioma cells. Neuro-oncol. 11, 122–131 82 Song, L.H. et al. (2006) Protective effects of soybean isoflavone against gammairradiation induced damages in mice. J. Radiat. Res. (Tokyo) 47, 157–165 83 Davis, T.A. et al. (2007) Subcutaneous administration of genistein prior to lethal irradiation supports multilineage, hematopoietic progenitor cell recovery and survival. Int. J. Radiat. Biol. 83, 141–151 84 Katz, J. et al. (2008) Isoflavones and gamma irradiation inhibit cell growth in human salivary gland cells. Cancer Lett. 270, 87–94 85 Papazisis, K.T. et al. (2000) Protein tyrosine kinase inhibitor, genistein, enhances apoptosis and cell cycle arrest in K562 cells treated with gamma-irradiation. Cancer Lett. 160, 107–113 86 Krishnan, K. and Vijayalakshmi, N.R. (2005) Alterations in lipids & lipid peroxidation in rats fed with flavonoid rich fraction of banana (Musa paradisiaca) from high background radiation area. Indian J. Med. Res. 122, 540–546 87 McLaughlin, N. et al. (2006) The survivin-mediated radioresistant phenotype of glioblastomas is regulated by RhoA and inhibited by the green tea polyphenol ()epigallocatechin-3-gallate. Brain Res. 1071, 1–9 88 Yamamoto, T. et al. (2004) Protective effects of EGCG on salivary gland cells treated with gamma-radiation or cis-platinum(II)diammine dichloride. Anticancer Res. 24, 3065–3073 89 Turner, N.D. et al. (2002) Opportunities for nutritional amelioration of radiationinduced cellular damage. Nutrition 18, 904–912 90 Devipriya, N. et al. (2008) Quercetin ameliorates gamma radiation-induced DNA damage and biochemical changes in human peripheral blood lymphocytes. Mutat. Res. 654, 1–7 91 Nair, C.K. and Salvi, V.P. (2008) Protection of DNA from gamma-radiation induced strand breaks by Epicatechin. Mutat. Res. 650, 48–54 92 Benkovic, V. et al. (2008) Evaluation of the radioprotective effects of propolis and flavonoids in gamma-irradiated mice: the alkaline comet assay study. Biol. Pharm. Bull. 31, 167–172 93 Maurya, D.K. et al. (2005) Protection of cellular DNA from gamma-radiationinduced damages and enhancement in DNA repair by troxerutin. Mol. Cell. Biochem. 280, 57–68 94 Hosseinimehr, S.J. et al. (2009) Radioprotective effects of hesperidin against genotoxicity induced by gamma-irradiation in human lymphocytes. Mutagenesis 24, 233–235 95 Hosseinimehr, S.J. et al. (2009) Protective effects of hesperidin against genotoxicity induced by (99m)Tc-MIBI in human cultured lymphocyte cells. Nucl. Med. Biol. 36, 863–867 96 Jankovic, T. et al. (2008) Radioprotective effects of Gentianella austriaca fractions and polyphenolic constituents in human lymphocytes. Planta Med. 74, 736–740 97 Jagetia, G.C. and Reddy, T.K. (2002) The grapefruit flavanone naringin protects against the radiation-induced genomic instability in the mice bone marrow: a micronucleus study. Mutat. Res. 519, 37–48 98 Hosseinimehr, S.J. and Nemati, A. (2006) Radioprotective effects of hesperidin against gamma irradiation in mouse bone marrow cells. Br. J. Radiol. 79, 415–418 99 Parihar, V.K. et al. (2007) Anticlastogenic activity of morin against whole body gamma irradiation in Swiss albino mice. Eur. J. Pharmacol. 557, 58–65 100 Landauer, M.R. et al. (2003) Genistein treatment protects mice from ionizing radiation injury. J. Appl. Toxicol. 23, 379–385 101 Alcaraz, M. et al. (2009) Liposoluble antioxidants provide an effective radioprotective barrier. Br. J. Radiol. 82, 605–609
www.drugdiscoverytoday.com
917
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
102 Marfak, A. et al. (2003) Redox reactions obtained by gamma irradiation of quercetin methanol solution are similar to in vivo metabolism. Radiat. Res. 159, 218–227 103 Castillo, J. et al. (2000) Antioxidant activity and radioprotective effects against chromosomal damage induced in vivo by X-rays of flavan-3-ols (Procyanidins) from grape seeds (Vitis vinifera): comparative study versus other phenolic and organic compounds. J. Agric. Food Chem. 48, 1738–1745 104 Melidou, M. et al. (2005) Protection against nuclear DNA damage offered by flavonoids in cells exposed to hydrogen peroxide: the role of iron chelation. Free Radic. Biol. Med. 39, 1591–1600 105 Saada, H.N. et al. (2009) Grape seed extract Vitis vinifera protects against radiationinduced oxidative damage and metabolic disorders in rats. Phytother. Res. 23, 434–438 106 Kalpana, K.B. et al. (2009) Investigation of the radioprotective efficacy of hesperidin against gamma-radiation induced cellular damage in cultured human peripheral blood lymphocytes. Mutat. Res. 676, 54–61 107 Moridani, M.Y. et al. (2003) Dietary flavonoid iron complexes as cytoprotective superoxide radical scavengers. Free Radic. Biol. Med. 34, 243–253 108 Kostyuk, V.A. et al. (2004) Experimental evidence that flavonoid metal complexes may act as mimics of superoxide dismutase. Arch. Biochem. Biophys. 428, 204–208 109 Svoboda, P. and Harms-Ringdahl, M. (2005) Influence of chromatin structure and radical scavengers on yields of radiation-induced 8-oxo-dG and DNA strand breaks in cellular model systems. Radiat. Res. 164, 303–311 110 Mishra, K. et al. (2009) Netropsin, a minor groove binding ligand: a potential radioprotective agent. Radiat. Res. 172, 698–705 111 Tawar, U. et al. (2007) Nuclear condensation and free radical scavenging: a dual mechanism of bisbenzimidazoles to modulate radiation damage to DNA. Mol. Cell. Biochem. 305, 221–233 112 Martin, R.F. et al. (2004) In vitro studies with methylproamine: a potent new radioprotector. Cancer Res. 64, 1067–1070 113 Janjua, N.K. et al. (2009) Spectrophotometric analysis of flavonoid-DNA binding interactions at physiological conditions. Spectrochim. Acta A: Mol. Biomol. Spectrosc. 74, 1135–1137 114 Kanakis, C.D. et al. (2007) An overview of DNA and RNA bindings to antioxidant flavonoids. Cell Biochem. Biophys. 49, 29–36 115 Oliveira-Brett, A.M. and Diculescu, V.C. (2004) Electrochemical study of quercetin–DNA interactions: part II. In situ sensing with DNA biosensors. Bioelectrochemistry 64, 143–150
918
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
116 Wang, Z. et al. (2008) Evaluation of flavonoids binding to DNA duplexes by electrospray ionization mass spectrometry. J. Am. Soc. Mass Spectrom. 19, 914–922 117 Ragazzon, P.A. et al. (2009) Structure–activity studies of the binding of the flavonoid scaffold to DNA. Anticancer Res. 29, 2285–2293 118 Zhou, J. et al. (2001) Synthesis, characterization, antioxidative and antitumor activities of solid quercetin rare earth(III) complexes. J. Inorg. Biochem. 83, 41–48 119 Kanakis, C.D. et al. (2006) Interaction of antioxidant flavonoids with tRNA: intercalation or external binding and comparison with flavonoid–DNA adducts. DNA Cell Biol. 25, 116–123 120 Solimani, R. (1997) The flavonols quercetin, rutin and morin in DNA solution: UV– vis dichroic (and mid-infrared) analysis explain the possible association between the biopolymer and a nucleophilic vegetable-dye. Biochim. Biophys. Acta 1336, 281–294 121 Zhu, Z. et al. (2002) Electrochemical studies of quercetin interacting with DNA. Microchem. J. 71, 57–63 122 Solimani, R. (1996) Quercetin and DNA in solution: analysis of the dynamics of their interaction with a linear dichroism study. Int. J. Biol. Macromol. 18, 287–295 123 Marinic, M. et al. (2006) Interactions of quercetin and its lanthane complex with double stranded DNA/RNA and single stranded RNA: spectrophotometric sensing of poly G. J. Inorg. Biochem. 100, 288–298 124 Mello, L.D. et al. (2007) Electrochemical and spectroscopic characterization of the interaction between DNA and Cu(II)-naringin complex. J. Pharm. Biomed. Anal. 45, 706–713 125 Tan, J. et al. (2009) DNA binding and cleavage activity of quercetin nickel(II) complex. Dalton Trans. 24, 4722–4728 126 Oliveira-Brett, A.M. and Diculescu, V.C. (2004) Electrochemical study of quercetin–DNA interactions: part I. Analysis in incubated solutions. Bioelectrochemistry 64, 133–141 127 Li, Y.H. et al. (2007) Infrared and DNA-binding on ultraviolet and fluorescence spectra of new copper and zinc complexes with a naringenin Schiff-base ligand. Spectrochim. Acta A: Mol. Biomol. Spectrosc. 67, 395–401 128 Kouvaris, J.R. et al. (2007) Amifostine: the first selective-target and broad-spectrum radioprotector. Oncologist 12, 738–747 129 Allalunis-Turner, M.J. (1990) Reduced bone marrow pO2 following treatment with radioprotective drugs. Radiat. Res. 122, 262–267 130 Triantafyllou, A. et al. (2007) The flavonoid quercetin induces hypoxia-inducible factor-1alpha (HIF-1alpha) and inhibits cell proliferation by depleting intracellular iron. Free Radic. Res. 41, 342–356 131 Bach, A. et al. (2010) The antioxidant quercetin inhibits cellular proliferation via HIF-1-dependent induction of p21WAF. Antioxid. Redox Signal
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
This review focuses on the thermodynamic basis of the unfavorable changes observed in physicochemical properties in lead discovery and optimization programs and suggests that monitoring binding thermodynamics could contribute to an improvement in the quality of compounds identified.
Thermodynamics guided lead discovery and optimization Gyo¨rgy G. Ferenczy1 and Gyo¨rgy M. Keseru˝2,* 1 2
Sanofi-Aventis CHINOIN, 1-5. To´ u. Budapest, Hungary, H-1045, Hungary Discovery Chemistry, Gedeon Richter Plc., 19-21. Gyo¨mro˝i u´t, Budapest, H-1103, Hungary
The documented unfavorable changes of physicochemical properties during lead discovery and optimization prompted us to investigate the present practice of medicinal chemistry optimization from a thermodynamic perspective. Basic principles of binding thermodynamics suggest that discriminating between enthalpy-driven and entropy-driven optimizations could be beneficial. We hypothesize that entropy-driven optimizations might be responsible for the undesirable trend observed in physicochemical properties. Consequently, we suggest that enthalpydriven optimizations are preferred because they provide better quality compounds. Monitoring binding thermodynamics during optimization programs initiated from thermodynamically characterized hits or leads, therefore, could improve the success of discovery programs. Here, we summarize common industry practices for tackling optimization challenges and review how the assessment of binding thermodynamics could support medicinal chemistry efforts. Introduction More than a decade ago, Teague et al. [1] investigated the physicochemical profile of screening compounds and concluded that polar and low molecular weight (MW) starting points were more easily converted to leads than lipophilic and higher MW hits. They proposed that a suitable screening library should consist of compounds with a MW range between 100 and 350 and clogP = 1–3. It was suggested that hits from such lead-like libraries would provide a wider chemistry space during the optimization of potency, physicochemical and absorption, distribution, metabolism and excretion (ADME) properties. The effect of lead optimization on physicochemical properties has been analyzed in comparative studies between leads and corresponding drugs [1,2]. These studies demonstrated that leads are typically less complex (with lower MW, fewer rings and rotatable bonds) and less hydrophobic (lower clogP) than drugs and suggest that the lead optimization process results in more complex structures. Although there might be notable differences in the physicochemical profile of compounds optimized against different target families, Morphy showed [3] that property shifts associated with optimization vary only slightly across target families. Thus, we can conclude that lead optimization is a major contributor to the unfavorable change in properties of clinical candidates.
Gyo¨rgy G. Ferenczy Gyo¨rgy G. Ferenczy received a Ph.D. from the Eo¨tvo¨s University in Budapest, Hungary in 1988. His early career involved quantum chemical method development and applications to extended systems. He was a postdoctoral fellow at the University of Oxford, UK, and at the University of Nancy, France. He was in the Drug Design group of Gedeon Richter. In 1999 he joined Sanofi-Aventis CHINOIN as drug designer. From 2001 he spent three years as the Head of Drug Design at the Strasbourg site of the company. He has authored and co-authored over 50 scientific papers and patents. Gyo¨rgy M. Keseru˝ Gyo¨rgy M. Keseru˝ obtained his Ph.D. at Budapest, Hungary and joined Sanofi-Aventis CHINOIN heading a chemistry research lab. He moved to Gedeon Richter in 1999 as the Head of Computeraided Drug Discovery. Since 2003 has been responsible for lead discovery activity and more recently he was appointed as the Head of Discovery Chemistry at Gedeon Richter. He contributed to the discovery of the antipsychotic cariprazine and the analgesic radiprodil currently investigated in Phase III and Phase II clinical trials, respectively. His research interests include medicinal chemistry and drug design. He has published over 140 papers and more than 10 books and book chapters.
*
Corresponding author:. Keseru˝, G.M. ([email protected])
1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.08.013
www.drugdiscoverytoday.com
919
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Attempts to control property shifts
Reviews KEYNOTE REVIEW
MW and lipophilicity have a major impact on compound quality because – in addition to physicochemical properties – these parameters have a considerable impact on drug metabolism and pharmacokinetics (DMPK) and safety profile. MW increases in parallel with complexity, yielding large and complex molecules that are more likely to form suboptimal or repulsive interactions upon binding to proteins, as demonstrated by Hann et al. [4]. Moreover, highly lipophilic compounds have a greater chance of being promiscuous [5,6]; they typically have limited solubility and ADME/ DMPK problems [7]. Because these properties have a direct impact on clinical success rates [8], two strategic improvements have been introduced to the practice of lead optimization. First, compounds prepared in optimization programs are now screened extensively in physicochemical and in vitro ADME assays to evaluate their property profile. Second, following the original idea of Teague et al. [1], it was concluded that less complex, polar, low MW hits serve as better starting points for optimization. Fragment-based drug discovery (FBDD) straightforwardly realizes this concept. Restricted size and
complexity of fragments results in low binding affinity; therefore, identification of fragment hits requires high concentration screening and, consequently, high solubility for fragments. The fragment space defined by the ‘rule of three’ [9] typically fulfills these criteria. FBDD reviews frequently claim that physical properties could be more easily controlled in optimizing fragments than starting from higher affinity high throughput screening (HTS) hits (see, for example, Ref. [10]). The hope that unfavorable property shifts could be avoided by using FBDD strategies has generated considerable interest in the medicinal chemistry community. Investigating whether fragments could really help reduce property shifts, we collected fragment hit-lead pairs from the literature [11–14] and compared their physicochemical profiles to those of recent HTS hits and HTS leads. Our hit-to-lead database consists of 59 HTS [15] and 34 fragment hits and leads, respectively, which were all screened against the same set of targets, including proteases, kinases and GPCRs (Supplementary Data). HTS hits typically showed micromolar affinity with a mean pPotency of 6.01 (Fig. 1a), which was higher than the pPotency
[(Figure_1)TD$IG] (a)
(b)
Potency
8
0.8
6 5 4
FR
3
HTS
2
Tanimoto
p(Potency)
7
Lead to hit similarity
0.6 0.4 0.2
1 0.0
0
(d)
LE 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00
SFR SILE
3.00 2.50 2.00
SILE
LE
(c)
HT
Lead
Hit
FR
FR HTS
1.50 1.00
HTS
0.50 0.00
Lead
Hit (e)
(f)
LLE 5 4 4 3 3 2 2 1 1 0
FR HTS
Hit
Lead
Lead
Hit LELP 16 14 12 10 8 6 4 2 0
FR HTS
Hit
Lead Drug Discovery Today
FIGURE 1
Average properties of fragment (FR) and HTS hits and leads. 920
www.drugdiscoverytoday.com
obtained for less complex fragment hits (pPotency = 4.41), as expected. Hit-to-lead optimization from HTS hits resulted in HTS leads with a median pPotency of 7.60, representing an increase in affinity of approximately one order of magnitude. In the case of fragment hits, the mean pPotency of leads reached 7.20, revealing that fragment hits could be effectively optimized to leads having similar affinity to that of HTS leads. In addition to the significant increase in affinity, the average of Tanimoto similarities calculated between corresponding hit and lead pairs (Fig. 1b) suggests that hit-to-lead optimization of fragment hits led to structurally dissimilar leads. The larger chemistry space and the increased freedom of operation associated with fragment hits seem to be reflected in Tanimoto indices obtained for fragment- (0.53) and HTS-based (0.72) optimizations. Next, we investigated ligand efficiency measures, including original ligand efficiency (LE) [16] and size-independent ligand efficiency (SILE) [17]. Contrary to the frequently cited phenomena of the high LE of fragments [10], we found that HTS hits and leads have LE similar to that of fragment pairs (Fig. 1c). It seems that the high potency of HTS hits compensates for their more complex nature, resulting in high LE for those HTS hits that could be followed up and optimized to leads. It was interesting to see, however, that the initial LEs did not improve for either fragmentor HTS-based optimizations. Because the size dependency of LE became obvious recently, we also compared SILE for fragment and HTS hits and leads (Fig. 1d). SILE of HTS hits and leads was higher than that of the fragments, although their difference became marginal for leads. Comparing lipophilic efficiencies, we found that the better lipophylic ligand efficiency (LLE) [15] of fragment hits disappeared for leads having virtually identical LLE to that of HTS leads (Fig. 1e). Recently, we introduced a new metric – LELP, defined as the ratio of logP and LE [15] – to depict the price of LE paid in logP (i.e. a lower absolute value of LELP is better). The present analysis revealed that LELP does not improve during the optimization of HTS hits to leads. More importantly, increasing LELP values associated with hit-to-lead optimizations of fragments suggests that improved affinity of fragment leads was primarily achieved by adding lipophilicity (Fig. 1f). Although LELP indicates that lipophilic efficiency deteriorates during hit-to-lead optimizations, SILE values demonstrate that ligand efficiency improves for both fragment and HTS hits and, furthermore, that this improve-
REVIEWS
ment seems to be more significant for fragment hits. The significant increase in affinity combined with large structural changes during hit-to-lead optimization underlines the advantage of fragment-based approaches over conventional HTS-based optimization because the former provides diverse lead chemotypes with almost the same affinity for leads as HTS leads. Although these advantages made fragment-based approaches popular in earlyphase discovery, analyzing the basic properties of fragment hits and leads highlighted that, like HTS hits, the properties of fragments are also shifted unfavorably (Fig. 2). At the hit-identification phase, fragment hits are notably less complex and lipophilic than HTS hits, as described in most case studies reporting fragment-based hit discovery [11–14]. This is one of the conceptual advantages of fragment screening: we can pick up a less complex and more soluble starting point for hit-to-lead studies. Unfortunately, our most important finding is that these good-quality fragment hits are optimized to leads with high MW and logP. Contrary to previous hopes of fragment-based approaches, this observation suggests that maintaining or improving LE or SILE alone is not a guarantee for high-quality leads. Fragment leads have a MW almost identical to that of HTS leads, and both groups have a similarly high logP that is in line with present medicinal chemistry practice [5]. Although most of the HTS-based optimizations do start from hits of higher MW and lipophilicity, this seems to be under control, as suggested by average changes detected during their optimization (DMW 75.63, DlogP 0.49). Fragment optimization, however, increased MW and lipophilicity more significantly (DMW 173.33, DlogP 0.93) as compared to their HTS-based counterparts, indicating that efficient optimization of fragments to potent leads is challenging (i.e. physicochemical properties could not be controlled easily, despite the attractive initial properties). These data indicate that present medicinal chemistry practice optimizes both HTS and fragment hits to leads that end up in the center of drug-like space. Consequently, further optimization with the same practice would shift the resulting candidate to the edge of the Lipinski zone with suboptimal physicochemical properties. The impact of lead optimization on undesired property changes has been well documented. Analysis of a large number of hit and lead pairs identified by different strategies, including highthroughput screening, fragment screening, natural product
[(Figure_2)TD$IG] LogP 4.0 3.5 3.0
FR
LogP
MW
MW 500 450 400 350 300 250 200 150 100 50 0
2.5
FR
2.0
HL
1.5
HTS HL
MedChem
1.0
MedChem
HTS
0.5 0.0
Hit
Lead
Hit
Lead Drug Discovery Today
FIGURE 2
Median molecular weights and logP calculated for fragment (FR) and HTS hits and leads. Corresponding values of historical leads (HL) [15] and compounds representing the present medicinal chemistry practice (MedChem) [5] are depicted for comparison. www.drugdiscoverytoday.com
921
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews KEYNOTE REVIEW
screening and virtual screening, has demonstrated [15,18] that this unfavorable shift in physicochemical properties can be traced back to lead discovery. We concluded that the increase in logP and MW during hit-to-lead optimization is independent of the nature of the library screened, the detection technology applied and the lead discovery strategy used. Here, we show that fragment-based approaches cannot avoid unfavorable property shifts per se. These observations suggest that it is the optimization practice that is a major contributor to property shifts, which prompted us to investigate the thermodynamic basis of optimization.
Thermodynamics of optimization The most important objective of hit or lead optimizations is improving ligand binding. The logarithm of binding affinity – usually quantified by Kd or Ki values – is proportional to the Gibbs binding free energy (DGbind, Eq. (1)). DGbind ¼ RT ln Kd RT ln Ki
(1)
where R is the gas constant, T is the absolute temperature, Kd is the apparent equilibrium dissociation constant and Ki is the inhibition constant defined by the Cheng–Prusoff equation [19]. DG is a function of the binding enthalpy (DH) and the binding entropy (DS). DGbind ¼ DH bind TDSbind
(2)
From a thermodynamic point of view, Eq. (2) suggests that the real challenge of medicinal chemistry optimization is to overcome enthalpy–entropy compensation [20]. There are two alternatives to achieve this goal: enthalpy-driven optimizations are characterized by decreasing DH that dominates over disfavored DS changes, and entropy-driven optimizations could be realized by increasing DS to compensate for DH penalties. Ligand binding is a multistep process that involves the conformational rearrangement and desolvation of both the ligand and the binding site and that is followed by the formation of the ligand– receptor complex. Assuming equilibrium thermodynamics, each of these elemental steps contributes to the binding thermodynamics of the ligand. Ligand binding is usually accompanied by conformational rearrangement of both the ligand and the receptor, and this typically represents an enthalpic penalty. Desolvation restructures organized water clusters around the ligand, which results in a significant entropic reward. Replacement of water from the binding site might be both enthalpic and entropic, depending on the binding interactions of the replaced waters. H-bonds broken upon desolvation, however, are responsible for an additional enthalpic penalty. Formation of the ligand–receptor complex is typically coupled to forming new interactions between the ligand and its binding site that are enthalpically beneficial. Molecular recognition of the ligand, however, limits its external rotational and translational freedom (as well as ligand and protein flexibility) and, therefore, represents an entropic penalty. Although the thermodynamic impact of long-range effects is usually neglected, they could also contribute to ligand binding. The net effect of these enthalpy and entropy components determines whether the binding is enthalpy or entropy dominated; thus, the optimization can be enthalpy or entropy driven, depending on which component contributes more significantly to the affinity improvement. Considering the enthalpic and entropic components of ligand binding, it was concluded that enthalpy-driven optimization is 922
www.drugdiscoverytoday.com
challenging [21]. Significant gain in binding enthalpy is associated with the formation of new contacts with optimal geometry that require new interaction partners such as charged groups, donors and/or acceptors at the ligand side. These new heteroatoms disfavor desolvation of the ligand and result in an enthalpic penalty. Because the new interactions formed upon binding reduce ligand and protein flexibility, they also contribute to the decrease of conformational entropy. Consequently, the gain in binding enthalpy could be easily compensated by enthalpic and entropic penalties caused by disfavored changes in desolvation and conformational entropy. In the case of entropy-driven optimizations, the gain in binding entropy could be realized by the increased lipophilicity of the ligand. More lipophilic compounds desolvate more easily, resulting in a significant reward in desolvation entropy. In addition, medicinal chemistry efforts reducing ligand flexibility – which usually increase MW and complexity – decrease the penalty arising from conformational entropy changes. Significant gain in desolvation entropy in conjunction with decreased penalty from conformational entropy is hardly compensated by enthalpic penalties. Because the optimization of specific interactions is far more difficult than increasing lipophilicity and complexity, entropy-driven optimization [21] seems to be a straightforward approach for medicinal chemistry teams working with strict timelines. In fact, entropy-driven optimization by adding lipophilic moieties and applying chain-ring strategies are successful tools routinely used in medicinal chemistry programs. Most of these optimizations, therefore, have significant entropy components, giving a thermodynamic rationale for undesirable property shifts. A recent article by Ladbury et al. [22] supports this hypothesis. The authors analyzed more than 400 isothermal calorimetry data obtained on more than 250 protein–ligand complexes and found a correlation between binding free energy and apolar surface burial upon complex formation. This finding is in accordance with the general medicinal chemistry observation that lipophilic interactions have a crucial role in binding affinity. Although the correlation between TDS and apolar surface burial was less remarkable, they identified a statistically significant trend indicating that increasing apolar surface burial is entropically favored. This result gives additional support to the notion that entropic optimization would be a major source of increasing lipophilicity and complexity documented in the medicinal chemistry literature.
Guidelines for thermodynamics-driven optimization Resolution of the binding free energy into entropy and enthalpy components goes beyond the usual characterization by affinity and it might be found useful at various stages of the hit to drug candidate process. A major challenge in optimization is that it is easier to achieve improved binding by increased hydrophobicity than by optimized polar interactions. Although it is a general assumption that improved polar interactions lead to more favorable binding enthalpy, we cannot fully control the enthalpy by simply engineering interactions. For this reason, it is advantageous to measure the binding free energy components at an early stage of drug discovery to guarantee advantageous polar interactions and to monitor them in the course of optimization. At decision points, such as hit or lead selection, the enthalpy content of binding is an
important piece of information to consider when one compares the potential of compound series (i.e. to assess whether an affinity gain together with a favorable physicochemical profile can be achieved upon optimization). In this respect, thermodynamic signature shows similarity to metrics that define LE and, in particular, to LLE [5] and LELP [15]. Compound characterization by LLE and LELP aims to support the optimization of affinity without increasing lipophilicity. Importantly, compound polarity does not necessarily correlate directly to the enthalpic component of binding. It is not the presence of the polar groups but their favorable interactions with the protein that contribute to the increase of binding enthalpy. This explains why no correlation was observed between the binding enthalpy and polarity-related Lipinski parameters for oral bioavailability [20]. The presence of polar groups is, however, a prerequisite for high binding enthalpy. A measure of the enthalpic content of the binding is enthalpic efficiency (EE) [23], which is defined as DH/Nhv, where Nhv is the number of non-hydrogen atoms. (An alternative called ‘specific EE’ and defined as DH/Npolar, with Npolar being the number of polar atoms, has also been proposed [23].) EE is similar to LE, but DH in EE replaces pKi in LE. It is well documented, however, that LE depends on the number of heavy atoms [17,24] and the SILE was defined as pKi/Nhv0.3 [17]. SILE enables an unbiased comparison of ligands of different sizes. Concerning EE, we have shown elsewhere [25] that it strongly depends on the number of atoms and that this dependence is different from that found for LE. The maximal observed DH increases (i.e. becomes less favorable) with increasing atom number and so does the maximal EE (EEmax). Nevertheless, the trend between Nhv and EEmax agrees with that found between Nhv and LEmax [17], although the parameters are different. The size dependency of EE would, therefore, potentially mislead chemists when compounds based on different scaffolds and sizes are compared. To avoid such size-biased prioritizations, we introduced [25] the size-independent enthalpic efficiency (SIHE), defined as SIHE = 0.018 DH N0.3 if DH is obtained at 300 K and is expressed in kcal/mol units. SIHE is a meaningful measure of the optimization potential of compounds helping series selection and optimization monitoring before late optimization. An optimal scenario of the optimization achieves affinity increase with only a modest growth in MW and lipophilicity. This can be realized by increasing the enthalpy content of the binding via the introduction of optimized polar interactions. Unfortunately, engineering of such interactions is a challenging task for several reasons. Polar interactions are highly sensitive to the relative positions of the interacting partners; in most cases, this sensitivity exceeds the precision our predictive tools can reach. Furthermore, the additivity of the free energy or its components is an approximation that also represents a hurdle in compound design [26]. Although it is reasonable to assume the additivity of the enthalpy of pairwise non-bonded interactions, the same cannot generally be assumed for the entropy and the free energy. This is due to the new or lost specific interactions that will change the number and population of available states for the system [27]. This issue is also related to the ubiquitous phenomenon of enthalpy–entropy compensation when, for example, the creation of a strong H-bond results in a favorable enthalpy gain that is largely compensated by an unfavorable entropy loss caused
REVIEWS
by the decrease of the available states for the system. These limitations of ligand design call for the experimental monitoring of the optimization by both structural studies and binding enthalpy measurements. Whereas X-ray or nuclear magnetic resonance (NMR) structures give a basically qualitative picture of the ligand–protein binding, thermodynamic data enable a quantification of the interactions. Thermodynamic data can come from the measurement of Kd at different temperatures followed by the application of the van’t Hoff equation to derive DH and TDS [28] or from isothermal titration calorimetry (ITC) experiments [29,30]. Both of the techniques provide the net thermodynamics of ligand binding that makes the structural interpretation of these data challenging. The former approach requires that enthalpy shows no temperature dependence (DCp 0), which is less typical for systems other than membrane proteins [31]. The ability to spot curvature in the plot caused by experimental error in Kd further complicates deriving thermodynamic parameters by the van’t Hoff equation. Steady-state measurements over a broad temperature range and rigorous curvature analysis, therefore, are suggested to obtain confident datasets. Finally, interdependency of DH and TDS impacts the interpretation of enthalpic and entropic components of ligand binding. ITC has the advantage of measuring DH directly, but its principal limitations are high protein requirement and low throughput. It should be noted that enthalpy values can change dramatically depending upon such conditions as temperature, pH, buffer, and so on. On one hand, enthalpy data should be corrected for superimposed protonation steps [26] and ion binding and release [32] if necessary. On the other hand, because enthalpy is an integral function of the heat capacity change, it might be important to measure DCp as well. Recent efforts with enthalpy arrays and automated ITC instruments promise to alleviate limitations in throughput [33,34]. Considering all of the limitations associated with the experimental evaluation of binding thermodynamics, carefully checked data generated for a series of compounds in unified conditions (a case typical in pharma optimizations) can be directly compared and analyzed. In most cases, quantitative structure–activity relationships (QSAR) use Kd or IC50, which are directly related to DG. A beneficial alternative is DH [35,36] because it better reflects the interactions between the ligand and the target. With accurate DH values made available by ITC, this avenue can readily be explored. Although DH is not the ultimate function we might want to optimize, a quantitative determination of DH enables us to have a better understanding of the interactions and to control the enthalpy content of binding. Although thermodynamic and structural studies are mutually corroborating and are best used together, in cases in which structural information is not available, thermodynamic experiments can provide us with information on the binding interactions and thus also with quantitative experimental feedback on the success of compound design. The thermodynamic analysis of ligand–receptor binding can, in some cases, also provide us with information on the agonist or antagonist nature of ligands. Agonists and antagonists can bind to the same receptor with different thermodynamic signatures. After the pioneering work of Weiland et al. [37], it turned out that the relative enthalpy and entropy components of agonist versus www.drugdiscoverytoday.com
923
Reviews KEYNOTE REVIEW
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews KEYNOTE REVIEW
antagonist binding are receptor dependent [38–42]. Furthermore, it was demonstrated that the discrimination depends on the experimental conditions applied [43,44]. This could suggest that the interactions of agonists and antagonists do differ, but this might not necessarily be manifested in the outcome of thermodynamic experiments because various contributions might cancel each other out. Appropriately chosen experimental conditions could perhaps affect the thermodynamic signatures of agonists and antagonists differently and, thus, a functional discrimination by thermodynamic experiments might become possible. Because DH affords a quantitative measure of the interactions, it can contribute to the localization of enthalpy hot spots of the active site and the identification of the crucial binding motifs of ligands. Structural changes in the ligand and its protein complex that are associated with significant favorable binding enthalpy are signs of new or optimized interactions. In this way, groups responsible for these advantageous interactions can be identified [44]. Thus, these interactions might be kept in the optimization of the compounds or they could serve as templates to introduce similar interactions in other compounds. Overcoming enthalpy–entropy compensation by designing specific new interactions is extremely difficult, making elimination of entropy-driven optimization unrealistic. Furthermore, the complexity of the binding event prevents delineating quantitative structure–thermodynamic relationships. We argue, however, that based on the evaluation of thermodynamic signatures, the practice of medicinal chemistry optimization could be thermodynamically more balanced. Basically, there are two strategies towards this goal. The first option is monitoring binding thermodynamics continuously during optimizations to support the design of thermodynamically balanced compounds in each round of the optimization cycle for follow-up [45]. The other option is the thermodynamic characterization of all available starting points and the selection of the enthalpically most favored ones for subsequent entropy-driven optimization. In addition, a combination of these strategies might also provide more viable leads. Balanced optimization can be achieved with favorable enthalpy and entropy contributions, which could give a limit on the desired entropy change. Independent of the approach used, the increasing enthalpic contributions to binding affinity would improve the quality of compounds optimized.
Case studies In this section, we discuss the practical utility of thermodynamic characterization used in early- and late-phase optimizations. Case studies of early optimization involve both HTS-based and fragment-based approaches (renin inhibitors and carbonic anhydrase, respectively). Late-stage optimizations are exemplified by HMGCoA and renin inhibitors.
Early-phase optimizations Renin inhibitors Renin is an aspartic protease of the renin–angiotensin system that cleaves its natural substrate, angiotensinogen, to angiotensin I. Angiotensin-converting enzyme processes angiotensin I further to the vasoconstrictor angiotensin II. The cleavage of angiotensinogen is the rate-determining step in the production of angiotensin II, suggesting that renin inhibitors are a promising therapy for 924
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
hypertension. The efficacy of Aliskiren, the first-in-class drug, provided clinical proof of concept for the development of nonpeptidic renin inhibitors. Although the first attempts to identify potent, orally active renin inhibitors were initiated almost 30 years ago, most of the peptidic or peptidomimetic compounds failed because of dissolution-limited absorption, high metabolic clearance and low oral bioavailability. Consequently, the identification and optimization of potent, non-peptidic, low MW, orally active renin inhibitors is desirable. Diaminopyrimidine-type renin inhibitors were discovered by an HTS campaign at Pfizer identifying 1 with double-digit micromolar affinity [46] (Fig. 3). Parallel synthesis of a 450-membered focused library around the diaminopyrimidine core resulted in 2, a low micromolar renin inhibitor. The X-ray structure of the renin-2 complex revealed that the diaminopyrimidine part of the molecule is stabilized by five hydrogen bonds; however, this analysis identified that the large S2 hydrophobic pocket and the smaller hydrophobic S3 subpocket were unoccupied. Because preliminary studies to fill the S2 pocket failed, 2 was first tethered by a tetrahydroisoquinoline (3) and a benzoxazinone (4) ring system, which were extended by a methoxypropyl side-chain toward the S3 subpocket [47,48]. X-ray analysis of the renin-3 complex showed that all hydrogen bonds stayed intact around the diaminopyrimidine core and that the methoxypropyl side-chain reached the S3 subpocket. Although this optimization increased the SILE significantly, there was only a small improvement in the SIHE and a marginal change in the LELP. The thermodynamic signature of 3 was recorded and compared to that of 2 (Fig. 4). New hydrophobic van der Waals contacts resulted in a moderate gain in DH that was less than 1 kcal/mol. Displacing ordered water molecules from the hydrophobic S3 subpocket, however, was entropically favored and compensated for the entropy loss associated with the decreased flexibility, as indicated by the almost 2 kcal/mol gain in TDS. These significant entropy effects identify the optimization of 2 to 3 as being basically entropy driven, as indicated by the marginal improvement detected in SIHE. Although the X-ray structure of the renin-4 complex is not publicly available, the complex of its 2,2-dimethyl-benzoxazinone analog was crystallized and showed that all the hydrogen bonds identified for 2 exist and that the S3 subpocket was filled by the methoxypropyl side-chain. This optimization increased both SILE and SIHE significantly and largely improved LELP simultaneously. Thermodynamic profiling of 4 revealed that in addition to the entropy gain caused by filling S3, significant enthalpic components are present (Fig. 4). The methyl group in position 2 formed favored van der Waals contacts and it is probable the benzoxazinone group was involved in new polar contacts within the active site. These new interactions yielded a significant gain in enthalpy (DH 4 kcal/mol) that was only partially compensated by the entropic penalty (TDS 2 kcal/mol) caused by the decreased flexibility and desolvation entropy. The significant enthalpy effects detected here mark this optimization as being enthalpy driven, as indicated by the significant improvement detected in SIHE. Comparing these two outcomes of the early-phase optimization, the entropically optimized compound (3) is somewhat more potent than that obtained by enthalpy-driven optimization (4). Compound 3, however, has a significantly higher MW and higher
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_3)TD$IG]
NH2 N NH2
NH2 N
N
N
O CH3
N
NH2 N
NH2
NH2
IC50=58 nM MW=417.5 clogP=4.2 SILE=2.53 SIHE=0.51 LELP=17.99
Reviews KEYNOTE REVIEW
N
3 HN
HN NH2 N
Cl
NH2
2
1 IC50=27,000 nM MW=388.3 clogP=4.5 SILE=1.72 LELP=25.61
N
F
F
Cl
IC50=6,600 nM MW=355.4 clogP=3.5 SILE=2.05 SIHE=0.46 LELP=17.57
N O
O
O CH3
IC50=132 nM MW=371.4 clogP=1.8 SILE=2.51 SIHE=0.65 LELP=7.06
CH3
4 NH2 N
N NH2
NH2 N
N
N
O CH3
N
NH2 N
NH2
IC50=58 nM MW=417.5 clogP=4.2 SILE=2.53 SIHE=0.40 LELP=17.99
NH2
3 HN
HN NH2 N
Cl
1 IC50=27,000 nM MW=388.3 clogP=4.5 SILE=1.72 LELP=25.61
N
F
F
Cl
NH2
2 IC50=6,600 nM MW=355.4 clogP=3.5 SILE=2.05 SIHE=0.35 LELP=17.57
N O
O
O CH3
IC=50132 nM MW=371.4 clogP=1.8 SILE=2.51 SIHE=0.50 LELP=7.06
CH3
4 Drug Discovery Today
FIGURE 3
Early-phase optimization of diaminopyrimidine-type renin inhibitors.
lipophilicity than compound 4. This finding is in line with the expectation that enthalpy-driven optimizations generate much less unfavorable shifts in physicochemical properties. SILEs are almost identical for both compounds, making the ranking difficult on this basis. The lipophilic efficiency defined by LELP [15] is, again, much better for compound 4. Comparison of the thermodynamic signatures shows that enthalpic components are much larger for compound 4 than for compound 3, as indicated by the corresponding SIHE values. Although the IC50 of compound 3 is half of that of compound 4, physicochemical and thermodynamic
data suggest the selection of compound 4 for further optimization. In fact, optimized compounds reported from these laboratories typically have benzoxazinone rather than tetrahydroisoquinoline rings [49].
Fragment-like carbonic anhydrase inhibitors Carbonic anhydrase (CA) is a Zn-containing metallo-enzyme that catalyzes the hydration of CO2 and the dehydration of bicarbonate. The human enzyme exists in several isoforms and is abundant in various tissues and cellular compartments. It plays a key www.drugdiscoverytoday.com
925
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Figure_4)TD$IG]
10
5
Kcal/mol
Reviews KEYNOTE REVIEW
0 2
3
4
15
16
17
-5
-10
dG dH -TdS
-15 Drug Discovery Today
FIGURE 4
Thermodynamic profile of renin inhibitors.
part in the regulation of pH and fluid balance in different parts of the body, and its inhibitors are used in various therapies. CA inhibitors are applied as diuretics, as antiglaucoma agents, in the management of mountain sickness and for the improvement of the arterial oxygenation in chronic obstructive pulmonary disease. CA is an ideal model system; its catalytic mechanism and structure are thoroughly studied and well characterized [35]. Furthermore, it binds benzene sulfonamides – small, fragmentlike compounds (MW < 250 Da) with low flexibility and high affinity to CA. The binding of benzene sulfonamides to CA occurs without gross conformational change of the enzyme. Recently, Scott and Jones reported the thermodynamic optimization of benzene sulfonamide (BSA)-type CA fragment inhibitors supported by ITC experiments and X-ray structure determinations [50,51] (Fig. 5). Their binding to CA is dominated by the interactions of the sulfonamide group. The binding free energies have favorable enthalpy and for the majority of the compounds a smaller favorable entropy component.
Substitutions on the benzene ring result in small changes in the binding affinity and thermodynamic signature of the ligands. To make meaningful comparisons of the binding free energies and their enthalpic and entropic components, the changes in DG, DH and DS with respect to the unsubstituted BSA (5) were investigated. The DD values enabled the tracking of subtle changes in the thermodynamics of binding (Fig. 6). m-Cl, m-CN and m-OMe (compounds 6–8) have similar thermodynamic signatures with large negative D(TDS) and smaller – but still large – positive DDH values that result in a small DDG that is either positive or negative. The o-Cl compound 9 shows a basically similar thermodynamic signature by having D(TDS) and DDH values with the same sign as those of the previous compounds but with lower absolute values. All four compounds have binding thermodynamics inferior to the reference BSA because the entropy content of their binding was increased at the expense of enthalpy loss. The m-F compound 10 is different in having a smaller favorable entropic and an almost absent enthalpic component. The o-F derivative 11 differs from all previously discussed compounds because it has a large favorable
[(Figure_5)TD$IG] O O
S
NH2
R1 R2
Compound 5 6 7 8 9 10 11 12 13 14
R1
R2 m-Cl m-OMe m-CN o-Cl m-F o-F
p-benzylamide p-benzylamide m-F p-benzylamide o-F
Kd/nM 839 591 1640 423 1125 118 314 27 20 5.7
LELP 0.49 2.30 1.04 0.38 1.29 1.11 0.68 4.2 4.36 4.84
SILE 3.05 3.03 2.75 3.02 2.90 3.37 3.17 3.08 3.09 3.31
SIHE 0.27 0.19 0.18 0.22 0.25 0.28 0.35 0.43 0.49 0.62
Drug Discovery Today
FIGURE 5
Fragment-like sulfonamide-type carbonic anhydrase inhibitors. 926
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_6)TD$IG]
4 3 2
Reviews KEYNOTE REVIEW
1
Kcal/mol
0 6
7
8
9
10
11
12
13
14
-1 -2 -3 -4 -5
ddG ddH
-6
-dTdS -7 Drug Discovery Today
FIGURE 6
Thermodynamic profile of carbonic anhydrase inhibitors of Figure 5. DDG, DDH and DTDS values relative to unsubstituted benzene sulfonamide 5.
enthalpic and a smaller, but still important, unfavorable entropic component. An analysis of the X-ray structures of the o-Cl (9), m-F (10) and o-F (11) complexes supports the interpretation of the different thermodynamic signatures. The fluorine atom of the o-F derivative points towards the main chain amide NH of Thr200 and this interaction presumably contributes to the enthalpy of binding. By contrast, the fluorine atom of the m-F derivative points in the opposite direction towards a dominantly hydrophobic surface of the enzyme. The aromatic ring of the o-Cl derivative is twisted with respect to the position of fluorine derivatives and places the Cl-atom in a hydrophobic pocket. These structural differences rationalize why m-F and o-Cl derivatives have decreased enthalpy and increased entropy content and explain the privileged binding thermodynamics of the o-F derivative. The SILE of the BSA derivatives 5–11 has a maximum at the m-F derivative 10, and the SIHE of these compounds has a maximum at the o-F BSA 11. The high SIHE of 11 accompanied with a reasonable SILE value suggests this compound as being a more appropriate starting point for further optimization owing to its more important enthalpy component. Compounds 12–14 include a p-benzylamide group (Fig. 5). Whereas 12 contains no further substituents, 13 and 14 are substituted by fluorine in the meta and ortho positions, to the sulfonamide group, respectively. Thus, within the pairs of 5 and 12, 10 and 13, and 11 and 14, either the absence of or the position of the F-substituent relative to the sulfonamide group is the same. The addition of the benzylamide group significantly improves binding in all three pairs, but the change in the binding free energy and its enthalpy and entropy components varies. This can be attributed to the varying intra- and intermolecular interactions in these compounds and in their CA complexes. Although
no experimental structure for these complexes is available, the binding features can be assumed from the X-ray structures of BSA derivatives and para-substituted benzylamide derivatives. The carboxamide group H-bonded to a water molecule that, in turn, is Hbonded to the enzyme. The F-substitution has a position-dependent effect on the H-bond acceptor ability of this carboxamide group. In addition, both the carboxamide and the sulfonamide groups can directly interact with an adjacent fluorine atom. Compound 14 shows the highest affinity and the highest enthalpy component, and its superiority is readily shown by its favorable SILE and SIHE values (Fig. 5). Although the most favorable enthalpy component among BSA derivatives 5–11 was already identified for o-F BSA 11, the addition of the p-benzylamide group not only increased the enthalpy content of binding but also resulted in the highest affinity compound. Thus, F-substitution ortho to the sulfonamide group has privileged properties in the BSA series, and this suggests that o-F BSA derivatives are particularly well suited for further investigations. In spite of this finding, a search in Prous Science Integrity1 [49] resulted in 15 m-F BSA derivatives associated with CA activity and no o-F BSA derivatives, as reported by Scott and Jones [50,51].
Late-phase optimizations Renin inhibitors Diaminopyrimidine-type renin inhibitors were next optimized toward the large hydrophobic S2 pocket (Fig. 7). NMR auxiliary screens identified an N-aryl-benzamide that binds to the S2 pocket. Inter-ligand nuclear Overhauser- effect (NOE) data suggested that the N-aryl-benzamide can be linked to the amino group located in position 4 of the diaminopyrimidine core. Compound 15 identified in this way showed somewhat less affinity (IC50 = 336 nM), probably www.drugdiscoverytoday.com
927
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Figure_7)TD$IG]
NH2
NH2 N
N N H N
Reviews KEYNOTE REVIEW
NH2 N
O
N
O CH3
H3C NH2
15 N O
O
NH2
HN O O CH3
N O
16
N
N F
S O O CH3
N NH2
O CH3
N
O CH3
O F
IC50=7 nM
CF3
O CH3
F
F
17
O
IC50=27nM
NH2
O
H N
NH2
N
IC50=132 nM
O CH3
H3C
NH2 N
N N H
IC50=336 nM
O CH3
CH3
4
N
O CH3
18
IC50=48 nM Drug Discovery Today
FIGURE 7
Late-phase optimization of diaminopyrimidine-type renin inhibitors.
because of the suboptimal contacts formed within the S2 pocket. Thermodynamic profiling of this compound (Fig. 4) indicated a significant loss in binding enthalpy and a large gain in entropy. Disfavored binding enthalpy suggested that polar groups in the aryl benzamide moiety could not form H-bonds in the S2 pocket and, thus, the penalty in desolvation could not be compensated by enthalpic factors. A huge gain in binding entropy could be rationalized by the displacement of ordered water molecules from the large hydrophobic S2 pocket. Based on these data, the team concluded that affinity could be improved by positioning substituents to interact with the negatively and positively polarized areas in the S2 pocket. Extensive optimization of S2 substituents led to the identification of 16. The thermodynamic signature of this compound nicely justified its design concept (Fig. 4) because significant gain in binding enthalpy was detected; however, new polar interactions decreased the flexibility of the inhibitor and the protein backbone, resulting in a less favored binding entropy. Consequently, the binding affinity of 16 was found to be more than 10 times higher than that of compound 15. An alternative optimization scheme focused on the central region and the S3 subpocket. Introducing the previously identified difluorophenyl moiety into position 2 of the benzoxazinone ring resulted in 17, with improved affinity. Although the S2 pocket remained empty in this case, the thermodynamic signature of this compound was similar to that of compound 16. Further optimization of the side-chain that fills the S3 subpocket finally led to compound 18, having somewhat less affinity but improved ADME 928
www.drugdiscoverytoday.com
properties. 18 showed good bioavailability both in rat (74%) and dog (19%), triggering its selection for preclinical development as identified from Prous Science Integrity1 [49].
HMG-CoA reductase inhibitors HMG-CoA reductase (HMGR) is an integral protein of endoplasmic reticulum membranes. HMGR catalyzes production of mevalonate from HMG-CoA, which is the rate-limiting step in cholesterol biosynthesis. HMGR inhibitors – such as statins – therefore effectively lower serum cholesterol levels. Statins prevent cardiovascular diseases and, although they are generally well tolerated, myalgia is a reported side-effect. This can be reduced by targeting hepatic tissues, and it has been shown that hydrophilic statins tend to be more hepatoselective [52]. Novel inhibitors of HMGR with improved preclinical efficacy and hepatoselectivity were sought [53]. Six series of analogs of earlier statins were investigated. They contain a central heteroaromatic ring substituted by 3,5-dihydroxyheptanoic acid and in most cases an adjacent aromatic group. Series 1 and 2 have imidazole cores and differ in the N-atom positions in the core. Series 3 contains pyrrole-based bicyclic compounds, and series 4–6 have pyrrole cores with different N positions. Fig. 8 shows the highest affinity compound from each series together with rosuvastatin (25), which is shown for reference. Biochemical assay, crystallography and ITC were used to characterize the compounds, the majority of which are in the nanomolar potency range. The most potent compounds in each series
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_8)TD$IG]
OH
O
OH
O O
N
OH
N H
N OH
O
N
F
22, Series 4 Kd = 27.4nM
19, Series 1 Kd = 13.5nM
OH
O
OH
O
OH
F
O
OH
N
OH
O
N
OH
N H
OH
OH
Reviews KEYNOTE REVIEW
N H
N H
N
OH
F F
F
F
20, Series 2 Kd = 26.9nM
23, Series 5 Kd = 3.2nM
O O N H
OH
OH
O
N
N H
OH
OH
O
O N
S
N
OH
O OH
F
24, Series 6 Kd = 21.3nM
21, Series 3 Kd = 12.7nM OH
OH
O
N N
OH N
S O O
F
25, Rosuvastatin Kd = 17.0nM Drug Discovery Today
FIGURE 8
HMG-CoA reductase inhibitors. Highest affinity compounds from each series and rosuvastatin.
have single-digit nanomolar activity. By contrast, enthalpy and entropy components vary considerably among the different series, although much less variation was observed within the series. Fig. 9 shows the thermodynamic signature of the highest affinity compound from each series. Series 1 and 2 have highly favorable binding free energy and the highest enthalpy component among all series. Unique structural features in series 1 and 2 include N-benzyl rather than N-phenyl
substitution of the amide group and the unsubstituted core Natom adjacent to the amide substituent. The authors proposed that these two factors are responsible for the favorable thermodynamic profile of these compounds. An analysis of the X-ray structure of 19 suggests that the phenyl to benzyl replacement has the advantage that the increased flexibility of the latter promotes the phenyl ring into an advantageous position and preserves the H-bond between the amide carbonyl and Ser565. These optimized www.drugdiscoverytoday.com
929
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Figure_9)TD$IG]
10
5
Reviews KEYNOTE REVIEW
0 Kcal/mol
19
20
21
22
23
24
25
-5
-10
-15
dG dH -TdS
-20 Drug Discovery Today
FIGURE 9
Thermodynamic profile of HMG-CoA reductase inhibitors.
interactions contribute to an increased binding enthalpy. The absence of the substituent ortho to the carboxamide group is also advantageous for the thermodynamic profile. In the complexes of other series, the partially solvent exposed substituents in this position result in an entropy gain at the expense of an enthalpy loss. Series 3 compounds have considerably lower binding enthalpy than compounds in series 1 and 2. The X-ray structure of 21 from series 3 reveals that the protein undergoes significant conformational movements that complicate the thermodynamic analysis of these complexes. The thermodynamic signatures of series 4, 5 and 6 are also inferior to those of series 1 and 2. This is rationalized by the replacement of the preferred N-benzyl by N-phenyl substituents and by the unfavorable substitution of the core N-atom (c.f. analysis of series 1 and 2). Series 4 and 5 are similar in structure and in terms of binding energy components. The change in the N position is not crucial for the interactions in accordance with the
observed similarity between series 1 and 2. Series 6 compounds contain a sulfonamide in position 4. Apparently, the replacement of carboxamide by sulfonamide has no significant effect on the binding, in line with the partial water exposure of these moieties. The conclusion of this analysis is that although the different series show similar affinities, they can be distinguished by their thermodynamic profiles. The high enthalpic content of binding of series 1 and 2 favors these series. Because series 1 is more hepatoselective than series 2 [53], the former is the most appropriate starting point for further development. It is also noteworthy that series 1 has a thermodynamic signature superior to rosuvastatin that was shown to bind with the highest enthalpy content among marketed statins [21,54]. Compounds from series 1 are shown in Fig. 10, and their thermodynamic signatures are in Fig. 11. With the exception of 30, they are characterized by high enthalpy that overcompensates for their unfavorable entropy. The best affinity was measured for 26, 27 and 29, although compound 19, with slightly lower
[(Figure_10)TD$IG] OH
O R2
N
N R3
O OH
N
R1
OH
Compound 19 26 27 28 29 30
R1 H F H H H H
R2 benzyl benzyl 4-Ph-benzyl (S)-CH(CH2OH)Ph (S)-CH(CH3)Ph 3-(2-F-Ph)-benzyl
R3 H H H H Me H
F Drug Discovery Today
FIGURE 10
HMG-CoA reductase inhibitors of series 1. 930
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_1)TD$IG]
10
5
19
26
27
28
29
30
Reviews KEYNOTE REVIEW
Kcal/mol
0
-5
-10
dG
-15
dH -TdS
-20
Drug Discovery Today
FIGURE 11
Thermodynamic profile of series 1 HMG-CoA reductase inhibitors.
affinity, has higher binding enthalpy. Owing to its advantageous balance of properties – including HMGR affinity, binding enthalpy, biological potency, and hepatoselectivity – 19 was selected for preclinical development [49,53].
Concluding remarks Although the undesirable shift in physicochemical properties of leads was first reported more than a decade ago, the current practice of drug discovery still tends to generate complex and apolar structures that are not ideally suited for clinical development. Recently, we showed that this phenomenon could be traced back to lead discovery (i.e. hit-to-lead optimization is responsible for an unfavorable shift of physicochemical properties). Comparing fragment and HTS hit and lead pairs screened against exactly the same set of targets, we demonstrated that fragment-based optimizations are also affected. Because unfavorable changes in properties seem to be independent of the lead discovery technology applied, we hypothesized that the optimization strategy should have a major impact on compound quality. Analyzing the thermodynamic basis of affinity optimizations, we concluded that entropy-driven optimization strategies contribute significantly to this undesired trend.
We argue that thermodynamically more balanced strategies might provide better quality leads and clinical candidates. Comparative thermodynamic analysis of lead candidates could help identify enthalpically favored starting points. Monitoring thermodynamic profiles along optimization pathways might enable a proper balance between entropic and enthalpic contributions. Promoting enthalpic optimizations, we introduce a size-independent measure of enthalpic efficiency (SIHE) that makes unbiased comparison of leads and the progression of compounds in discovery programs possible. Early case studies discussed here show that the concept of thermodynamics guided lead discovery and optimization could contribute to the success of both fragment-based and traditional medicinal chemistry programs.
Acknowledgements The authors are grateful to Ga´bor Maksay, Derek R. Buckle and the anonymous reviewers for their valuable comments.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.drudis.2010.08.013.
References 1 Teague, S.J. et al. (1999) The design of leadlike combinatorial libraries. Angew. Chem. Int. Ed. Engl. 38, 3743–3748 2 Oprea, T.I. et al. (2001) Is there a difference between leads and drugs? A historical perspective. J. Chem. Inf. Comput. Sci. 41, 1308–1315 3 Morphy, R. (2006) The influence of target family and functional activity on the physicochemical properties of pre-clinical compounds. J. Med. Chem. 49, 2969–2978 4 Hann, M.M. et al. (2001) Molecular complexity and its impact on the probability of finding leads for drug discovery. J. Chem. Inf. Comput. Sci. 41, 856–864 5 Leeson, P.D. and Springthorpe, B. (2007) The influence of drug-like concepts on decision-making in medicinal chemistry. Nat. Rev. Drug Discov. 6, 881–890 6 Hopkins, A.L. et al. (2006) Can we rationally design promiscuous drugs? Curr. Opin. Struct. Biol. 16, 127–136 7 Van de Waterbeemd, H. et al. (2001) Lipophilicity in PK design: methyl, ethyl, futile. J. Comput. Aided Mol. Des. 15, 273–286
8 Wenlock, M.C. et al. (2003) A comparison of physiochemical property profiles of development and marketed oral drugs. J. Med. Chem. 46, 1250–1256 9 Congreve, M. et al. (2003) A ‘rule of three’ for fragment-based lead discovery? Drug Discov. Today 8, 876–877 10 Murray, C.W. and Rees, D.C. (2009) The rise of fragment-based drug discovery. Nat. Chem. 1, 187–192 11 Alex, A.A. and Flocco, M.M. (2007) Fragment-based drug discovery: what has it achieved so far? Curr. Top. Med. Chem. 7, 1544–1567 12 Congreve, M. et al. (2008) Recent developments in fragment-based drug discovery. J. Med. Chem. 51, 3661–3680 13 Bembenek, S.D. et al. (2009) Ligand efficiency and fragment-based drug discovery. Drug Discov. Today 14, 278–283 14 Schulz, M.N. and Hubbard, R.E. (2009) Recent progress in fragment-based lead discovery. Curr. Opin. Pharmacol. 9, 1–7
www.drugdiscoverytoday.com
931
REVIEWS
Reviews KEYNOTE REVIEW
˝ , G.M. and Makara, G.M. (2009) The influence of lead discovery strategies on 15 Keseru the properties of drug candidates. Nat. Rev. Drug Discov. 8, 203–212 16 Hopkins, A.L. et al. (2004) Ligand efficiency: a useful metric for lead selection. Drug Discov. Today 9, 430–431 17 Nissink, J.W.M. (2009) Simple size-independent measure of ligand efficiency. J. Chem. Inf. Model. 49, 1617–1622 ˝ , G.M. and Makara, G.M. (2006) Hit discovery and hit-to-lead approaches. 18 Keseru Drug Discov. Today 11, 741–748 19 Cheng, Y. and Prusoff, W.H. (1973) Relationship between the inhibition constant (Ki) and the concentration of inhibitor, which causes 50 per cent inhibition (IC50) of an enzymatic reaction. Biochem. Pharmacol. 22, 3099–3108 20 Ruben, A.J. et al. (2006) Overcoming roadblocks in lead optimization: a thermodynamic perspective. Chem. Biol. Drug Des. 67, 2–4 21 Freire, E. (2008) Do enthalpy and entropy distinguish first in class from best in class? Drug Discov. Today 13, 869–874 22 Olsson, T.S. et al. (2008) The thermodynamics of protein–ligand interaction and solvation: insights for ligand design. J. Mol. Biol. 384, 1002–1017 23 Ladbury, J.E. et al. (2009) Adding calorimetric data to decision making in lead discovery: a hot tip. Nat. Rev. Drug Discov. 9, 23–27 24 Reynolds, C.H. et al. (2008) Ligand binding efficiency: trends, physical bias, and implications. J. Med. Chem. 51, 2432–2438 ˝ , G.M. Enthalpic efficiency of ligand binding. J. Chem. Inf. 25 Ferenczy, G.G. and Keseru Model. (in press), doi:10.1021/ci100125a 26 Baum, B. et al. (2010) Non-additivity of functional group contributions in protein– ligand binding: a comprehensive study by crystallography and isothermal titration calorimetry. J. Mol. Biol. 397, 1042–1054 27 Mark, A.E. and van Gunsteren, W.F. (1994) Decomposition of the free energy of a system in terms of specific interactions. J. Mol. Biol. 240, 167–176 28 Tellinghuisen, J. (2006) van’t Hoff analysis of K8 (T): how good . . . or bad? Biophys. Chem. 120, 114–120 29 Horn, J.R. et al. (2001) van’t Hoff and calorimetric enthalpies from isothermal titration calorimetry: are there significant discrepancies? Biochemistry 40, 1774– 1778 30 Horn, J.R. et al. (2002) van’t Hoff and calorimetric enthalpies II: effects of linked equilibria. Biochemistry 41, 7501–7507 31 Gilli, P. et al. (1994) Enthalpy-entropy compensation in drug receptor binding. J. Phys. Chem. 98, 1515–1518 32 Chaires, J.B. (2008) Calorimetry and thermodynamics in drug design. Annu. Rev. Biophys. 37, 137–151 33 Torres, F.E. et al. (2004) Enthalpy arrays. Proc. Natl. Acad. Sci. U. S. A. 101, 9517–9522 34 Recht, M.I. et al. (2009) Measurement of enzyme kinetics and inhibitor constants using enthalpy arrays. Anal. Biochem. 388, 204–212 35 Krishnamurthy, V.M. et al. (2008) Carbonic anhydrase as a model for biophysical and physical-organic studies of proteins and protein–ligand binding. Chem. Rev. 108, 946–1051
932
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
36 Luque, I. and Freire, E. (2002) Structural parameterization of the binding enthalpy of small ligands. Proteins 49, 181–190 37 Weiland, G.A. et al. (1979) Fundamental difference between the molecular interactions of agonists and antagonists with the b-adrenergic receptor. Nature 281, 114–117 38 Miklavc, A. et al. (1990) On the fundamental difference in the thermodynamics of agonist and antagonist interactions with b-adrenergic receptors and the mechanism of entropy driven binding. Biochem. Pharmacol. 40, 663–669 39 Borea, P.A. et al. (1996) Binding thermodynamics at A1 and A2A adenosine receptors. Life Sci. 59, 1373–1388 40 Maksay, G. (1994) Thermodynamics of g-aminobutyric acid type A receptor binding differentiate agonists from antagonists. Mol. Pharmacol. 46, 386–390 41 Dalpiaz, A. et al. (1996) Binding thermodynamics of 5-HT1A receptor ligands. Eur. J. Pharmacol. 312, 107–114 42 Wittmann, H-J. et al. (2009) Contribution of binding enthalpy and entropy to affinity of antagonist and agonist binding at human and guinea pig histamine H1receptor. Mol. Pharmacol. 76, 25–37 43 Harper, E.A. and Black, J.W. (2007) Histamine H3-receptor agonists and imidazolebased H3-receptor antagonists can be thermodynamically discriminated. Br. J. Pharmacol. 151, 504–517 44 Maksay, G. (2005) Activation of ionotropic receptors and thermodynamics of binding. Neurochem. Int. 46, 281–291 45 Ciulli, A. et al. (2006) Probing hot spots at protein–ligand binding sites: a fragmentbased approach using biophysical methods. J. Med. Chem. 49, 4992–5000 46 Freire, E. (2009) A thermodynamic approach to the affinity optimization of drug candidates. Chem. Biol. Drug Des. 74, 468–472 47 Holsworth, D.D. et al. (2007) Discovery of 6-ethyl-2,4-diaminopyrimidine-based small molecule renin inhibitors. Bioorg. Med. Chem. Lett. 17, 3575–3580 48 Sarver, R.W. et al. (2007) Binding thermodynamics of substituted diaminopyrimidine renin inhibitors. Anal. Biochem. 360, 30–40 49 Powell, N.A. et al. (2007) Rational design of 6-(2,4-diaminopyrimidinyl)-1,4benzoxazin-3-ones as small molecule renin inhibitors. Bioorg. Med. Chem. 15, 5912– 5949 50 Science Integrity1 (http://integrity.prous.com) (2010) ß Prous Science, a Thomson Reuters business; all rights reserved 51 Scott, A.D. et al. (2009) Thermodynamic optimization in drug discovery: a case study using carbonic anhydrase inhibitors. ChemMedChem 4, 1985–1989 52 Jones, L.H. (2009) Thermodynamic optimization of carbonic anhydrase fragment inhibitors MEDI 456. 238th National Meeting and Exposition, August 16–20, 2009, Washington, DC 53 Rosenson, R.S. (2004) Current overview of statin induced myopathy. ACC Curr. J. Rev. 13, 11–12 54 Sarver, R.W. et al. (2008) Thermodynamic and structure guided design of statin based inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase. J. Med. Chem. 51, 3804–3813
REVIEWS
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
PET tracers for the peripheral benzodiazepine receptor and uses thereof Pernilla J. Schweitzer1,2, Brian A. Fallon3,4, J. John Mann3,4,5,6 and J.S. Dileep Kumar3,4,5 1
College of Physicians and Surgeons, Columbia University, New York, NY 10032, USA Doris Duke Clinical Research Fellow, Columbia University, New York, NY 10032, USA 3 Department of Psychiatry, Columbia University, New York, NY 10032, USA 4 New York State Psychiatric Institute, New York, NY 10032, USA 5 Division of Molecular Imaging and Neuropathology, Columbia University, New York, NY 10032, USA 6 Department of Radiology, Columbia University, New York, NY 10032, USA 2
The peripheral benzodiazepine receptor (PBR) is expressed on the outer mitochondrial membrane of activated microglia and is implicated in the pathophysiology of a variety of central nervous system and peripheral diseases. The abundant receptor concentration makes PBR a potential biomarker and an attractive target for quantification in vivo using positron emission tomography. PBR can be an important target for monitoring disease progression, for evaluating the effect of therapy, and for investigating new treatment modalities. PBR is also emerging as a potential target in the treatment of neuroinflammatory and neuropsychiatric disorders. Here, we review the positron emission tomography radioligands employed for imaging PBR in living brain and their applications. Introduction The peripheral benzodiazepine receptor (PBR) is a hetero-oligomeric complex located in the outer mitochondrial membrane [1]. The PBR consists of at least three different subunits, including an 18 kDa protein, a 32 kDa voltage-dependent anion channel, and a 30 kDa adenine nucleotide carrier [2]. Evidence supports three main functions of the PBR: (i) cholesterol binding and transport for biosynthesis of steroids and bile salts, (ii) protein import for membrane biogenesis, and (iii) porphyrin binding and transport for heme biosynthesis [2,3]. The PBR receptor binding site is predominantly the 18 kDa protein, which has been named the ‘translocator protein’, reflecting its role in binding and transport of molecules across the mitochondrial membrane [2,3]. In the central nervous system (CNS), PBR ligands have been found to stimulate synthesis of neurosteroids involved in diverse functions, from regulation of apoptosis to reduction of anxiety via modulation of the GABAA receptor [2]. The PBR is found in many regions of the body, including the human iris, ciliary-body, heart, liver, adrenal and testis, blood cells Corresponding author. Kumar, J.S. Dileep ([email protected]) 1359-6446/06/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.drudis.2010.08.012
(including lymphocytes and erythrocytes), and brain [2]. In the CNS, the PBR is expressed primarily on microglia when they become activated in response to a wide variety of insults [1]. Upon activation, microglia undergo a change in morphology, migrate toward the site of neuronal damage, proliferate, synthesize numerous pro-inflammatory molecules and might release neurotoxic metabolites, resulting in progression of disease and, ultimately, loss of neurons through prolonged microglia-mediated damage [4]. Because the PBR is expressed mostly on activated microglia, it is present only in very low levels in normal brain parenchyma, except in certain areas, such as those constitutively without blood– brain barrier (BBB) (e.g. the choroid plexus and the ependymal cells lining the ventricles). That the PBR is expressed at low levels in normal brain parenchyma and is upregulated locally in response to damage makes it a potentially ideal and sensitive marker for the detection of small changes in the region of injury.
PET imaging of PBR Positron emission tomography (PET) is an imaging technique in which tracer compounds labeled with positron-emitting radionuclides are injected into the subject of the study to track biochemical www.drugdiscoverytoday.com
933
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Fig._1)TD$IG]
11
O
H311C
O
O
N
N
N
Cl
11
CH3 O
O N CH3
F
N O
O
Reviews GENE TO SCREEN
Cl
N
N
Cl 11
[ C]Ro 5-4864
18
F
F
11
O
O
N N O
N O
CH3 N O N
N N N H3 C
O N
[18F]FEDAA1106
[11C]PBR28
[11C]DAA1106
[11C]PK11195
O 11 CH3
O
[11C]AC-5216
11
C
H2
N
O
O 11
N
CH3
O O
[11C]Vinpocetine
N
[11C]DPA-713 Drug Discovery Today
FIG. 1
PET tracers studied in human for PBR.
and physiological processes in vivo. Radiolabeling of PBR ligands has enabled the imaging of PBR expression using PET. Because it is based on a ligand–receptor interaction, PET imaging of the PBR benefits from well-validated concepts and tools from the neuroreceptor imaging field [5,6]. Several specific ligands for the PBR have been successfully radiolabeled and used for in vivo studies with human subjects [7–10] (Fig. 1). The development of radiolabeled ligands has enabled PET imaging of the PBR to be used in the study of neuroinflammatory and neurodegenerative conditions. PET imaging of the PBR strategy is already being used for quantitative assessment of disease progression and treatment response. PET findings involving the PBR in various human CNS diseases are summarized in Table 1. In this review, we focus on studies of radioligands for PET imaging of the PBR in human subjects, with a summary of findings published through March 2010.
[11C]Ro 5-4864 Ro 5-4864 is a 40 -chlorodiazepam (7-chloro-5-(4-chlorophenyl)-1methyl-1,3-dihydro-2H-1,4-benzodiazepin-2-one) compound. It is the only benzodiazepine with binding affinity 6 nM that has been radiolabeled with C-11 isotope in this context to date [11]. PET studies of [11C]Ro 5-4864 in human subjects with gliomas and meningiomas did not demonstrate increased uptake of this tracer in areas with tumor, known to have a high density of the PBR, as compared with uptake in normal brain tissue [12,13]. Moreover, it has been found that Ro 5-4864’s binding affinity is both temperature dependent and species dependent, providing markedly different 934
www.drugdiscoverytoday.com
results between rats and humans [14,15]. This ultimately limits its usefulness as a tool for studying the PBR.
[11C]-PK11195 (R)-[11C]-PK11195 Racemic PK11195 (1-(2-chlorophenyl)-N-methyl-N-(1-methylpropyl)-3-isoquinoline carboxamide) is the first non-benzodiazepine high-affinity and selective PBR ligand (KI = 9.3 nM) [5]. In vivo comparison of R and S enantiomers of [11C]PK11195 in rats with cortical focal lesion show a twofold higher affinity of (R)-[11C]PK11195, making it advantageous over racemic or S-enantiomer of PK11195 for imaging studies [16]. Although both Ro 54864 and PK11195 bind the PBR in a saturable and reversible manner with nanomolar affinity, they differ substantially in their kinetics and pharmacological profile in that PBR isoquinoline binding sites are more abundant in the normal human brain than PBR benzodiazepine sites by approximately threefold [17]. PK11195 has several kinetic properties that permit its use as an in vivo ligand: the extraction of PK11195 from blood to brain is rapid and high (>90%) and is unimpeded by the BBB (i.e. tracer delivery is similar in areas with and without a BBB) [18]. Although some have suggested that PK11195 is a substrate of the efflux transporter P-glycoprotein, other studies have not found this to be the case [19]. Quantification of [11C]-labeled PK11195 has been approached largely by either normalization of the uptake to a reference region such as the cerebellum or application of the simplified reference tissue model with a ‘reference’ region devoid of PBR derived from cluster analysis [20,21]. In addition to reference tissue modeling,
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
TABLE 1
In vivo PET imaging studies of the peripheral benzodiazepine receptor (PBR) in human diseases Disease
PET ligand 11
Effect
Refs
[ C]PK11195 (R)-[11C]PK11195 [11C]DAA1106 [(R)-[11C]PK11195
No increases in binding were identified relative to control subjects Increased binding in entorhinal, temporo-parietal and cingulate cortex Increased binding in various cortex regions, striatum and cerebellum Increased binding in various cortex regions
Groom et al., 1995 Cagnin et al., 2001 Yasuno et al., 2003 Edison et al., 2008
AD and MCI
(R)-[11C]PK11195 (R)-[11C]PK11195
No increased binding Increased binding in the frontal cortex of some subjects
Wiley et al., 2009 Okello, et al., 2009
ALS CV
(R)-[11C]PK11195 [(R)-[11C]PK11195
Increased binding in motor cortex, prefrontal cortex, pons and thalamus Increased binding in occipital and temporo-parietal cortex
Turner et al., 2004 Goerres et al., 2001
CBD
(R)-[11C]PK11195
Gerhard et al., 2004
(R)-[11C]PK11195
Increased binding in caudate, putamen, substantia nigra, pons, pre-postcentral gyrus and frontal cortex Increased binding in basal ganglia, temporal and parietal cortex
FASSc FTD Glioblastoma Glioblastoma
(R)-[11C]PK11195 (R)-[11C]PK11195 [11C]PK11195 [11C]Ro 5-4864
Reduction of binding in lung macrophages Increased binding in frontal temporal cortex Increased binding in area of tumor No increase in binding
Branley et al., 2008 Cagnin et al., 2004 Pappata et al., 1991 Junck et al., 1989
HE
(R)-[11C]PK11195
Cagnin et al., 2006
[11C]PK11195
Increased binding in the pallidum, right putamen and right dorsolateral prefrontal region No increase in binding
(R)-[11C]PK11195
Increased binding in primary and secondary projected neuron
Cagnin et al., 2001
Increased binding in thalamus, putamen, frontal, temporal and occipital cortex Increased binding in striatum and cortical regions including prefrontal cortex and anterior cingulate Increased binding in striatum and cortical regions
Hammoud et al., 2005
Increased binding in cerebral cortex Increased binding in the ipsilateral thalamus Increased binding in cerebral cortex Increased binding in primary lesion and remote pathological changes after Wallerian degeneration Increased binding in peri-infact zone
Ramsay et al., 1992 Pappata et al., 2000 Gerhard et al., 2000 Gerhard et al., 2005
Increased binding in MS plaques, cerebral central gray and in areas corresponding with clinical deficits Increased binding in Gadolinium-enhancing lesions and in normal-appearing white matter Increased binding in normal-appearing white matter Increased binding in MS plaques
Banati et al., 2000
Herpes encephalitis HIV encephalitis
11
(R)-[ C]PK11195 (R)-[11C]PK11195 (R)-[11C]PK11195
Ischemic stroke
11
[ C]PK11195 [11C]PK11195 [11C]PK11195 (R)-[11C]PK11195 (R)-[11C]PK11195
MS
11
(R)-[ C]PK11195 [11C]PK11195 [11C]PK11195 [11C]PK11195, [11C]vinpocetine
Reviews GENE TO SCREEN
AD
Henkel et al., 2004
Iverson et al., 2006
Pavese et al., 2006 Tai et al., 2007
Price et al., 2006
Debruyne et al., 2003 Versijpt et al., 2005 Vas et al., 2008
MSA
(R)-[11C]PK11195
Increased binding in prefrontal cortex, putamen, pallidum, pons and substantia nigra
Gerhard et al., 2003
PD
(R)-[11C]PK11195 (R)-[11C]PK11195 [11C]PK11195
Increased binding in midbrain Increased binding in pons, basal ganglia, frontal and temporal cortex Increased binding in contralateral putamen BP and midbrain BP, but not significantly different from normal controls
Ouchi et al., 2005 Gerhard et al., 2006 Bartels et al., 2010
PSH
[(R)-[11C]PK11195
Turner et al., 2005
PSP RA RE SZ SZ
(R)-[11C]PK11195 (R)-[11C]PK11195 (R)-[11C]PK11195 (R)-[11C]PK11195 (R)-[11C]PK11195
Increased binding in motor cortex and supplementary motor region contralateral to the affected limbs Increased binding in basal ganglia, midbrain, frontal cortex and cerebellum Higher binding in knee joints Increased binding in affected hemisphere Increased binding in total gray matter Increased binding in whole-brain gray matter, particularly in the hippocampus
Gerhard et al., 2006 Van der Laken et al., 2008 Banati et al., 1999 van Berckel et al., 2008 Doorduin et al., 2009
Abbreviations: AD, Alzheimer’s Disease; ALS, amyotrophic lateral sclerosis; CBD, corticobasal degeneration; CV, cerebral vasculitis; FASSc, fibrosing alveolitis associated with systemic sclerosis; FTD, frontal temporal dementia; HE, hepatic encephalopathy; HD, Huntington’s disease; MCI, mild cognitive impairment; MS, multiple sclerosis; MSA, multiple system atrophy; PD, Parkinson’s disease; PSH, progressive spastic hemiparesis; PSP, progressive supranuclear palsy; RA, rheumatoid arthritis; RE, Rasmussen’s encephalitis; SZ, schizophrenia.
full kinetic characterization of (R)-[11C]PK11195 with measurement of arterial input function has been reported with the application of a model with two tissue compartments and four rate constants [22]. Logan graphical analysis with arterial input function or with refer-
ence tissue input adds an accurate method for generating binding parameters [20,23]. In general, the use of a tissue input function has the advantage over an arterial input function in that the latter requires extensive arterial blood sampling in patients with brain www.drugdiscoverytoday.com
935
REVIEWS
Reviews GENE TO SCREEN
injury. Although these various evolutions in technique have led to some improvements in the quantification of [11C]PK11195, it is nevertheless difficult to assess their accuracy given that PET studies conducted in humans are not validated by postmortem receptor autoradiography [24,25]. Studies of whole-body distribution and metabolism of (R)[11C]PK11195 in humans have found a large individual variation in the amount of plasma radiometabolites [26]. The whole-body distribution of (R)-[11C]PK11195 showed the highest radioactivity levels in urinary bladder, adrenal gland, liver, salivary glands, heart, kidneys, and vertebral column. (R)-[11C] PK11195 seems to be eliminated through both the renal and the hepatobiliary systems. [11C]PK11195 has been used in a wide range of human CNS diseases, from Rasmussen’s encephalitis and multiple sclerosis (MS) to neurodegenerative diseases (Alzheimer’s, Parkinson’s, amyotrophic lateral sclerosis, and Huntington’s), infectious diseases (HIV and herpes encephalitis), and neuropsychiatric disorders such as schizophrenia. To date, the vast majority of PET imaging studies on PBR in human disease have been performed with [11C]PK11195 as the tracer, and – with the exception of the earliest studies – most have used the (R)-enantiomer. What follows is a summary of PET studies using racemic [11C]PK11195 or (R)[11C]PK11195 in human subjects. The first PET study to use [11C]PK11195 in humans was reported by Charbonneau et al. in 1986 [27]. The authors studied [11C]PK11195 binding in the heart of dogs and humans and found specific binding in both species. Junck et al. [12] later reported a comparative PET study of [11C]Ro 5-4864 and [11C]PK11195 in glioma patients and found increased binding of [11C]PK11195, whereas [11C]Ro 5-4864 failed to demonstrate specific binding. Banati et al. [28] observed a considerable increased binding of (R)-[11C]-PK11195 in the affected cerebral hemisphere of two Rasmussen’s encephalitis patients, as compared with the unaffected contralateral hemisphere. The histologic distributions of microglial staining and distribution pattern from MRI were also correlated with areas of increased (R)-[11C]PK11195 binding. By contrast, patients with hippocampal sclerosis showed no increase in (R)-[11C]PK11195 binding, which the authors interpret as evidence that the in vivo (R)[11C]PK11195 signal is preferentially caused by the presence of activated microglia, as opposed to glial scar tissue. In a study of patients with MS, Banati et al. [18] found increased PBR expression with (R)-[11C]PK11195 in areas of focal pathology identified by MRI, particularly in gadolinium-enhancing lesions. Binding of (R)-[11C]PK11195 was also increased in brain areas corresponding to ongoing or recent clinical deficit; for example, in patients with visual dysfunction, signals were observed in anatomical locations along the pathway of the neuronal network controlling visual processing or eye movement. No significant correlations were seen between the global hemispheric (R)[11C]PK11195 lesion load and disability (total expanded disability status scale and individual sub scores), disease duration, or the interval since the last relapse. The authors suggest that (R)[11C]PK11195 binding might relate better to clinical change than to cumulative measures of long-standing and recent disability as measured by the expanded disability status scale. PET studies by Debruyne et al. [29] in a small population of MS patients also revealed increased brain uptake of [11C]PK11195 in MS patients as compared with normal controls, particularly in patients imaged 936
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
during an acute MS relapse. Binding was increased in gadoliniumenhanced lesions, as well as in MRI normal-appearing white matter (NAWM). A later study by Versijpt et al. [30] elaborated on this finding of increased [11C]PK11195 in areas of MRI-NAWM by also demonstrating a significant correlation between total NAWM [11C]PK11195 uptake and disease duration, as well as a correlation between NAWM [11C]PK11195 uptake and brain atrophy as measured by MRI. Vas et al. [31] compared [11C]PK11195 and [11C]vinpocetine binding in MS patients and observed greater binding of both tracers in affected brain regions than in unaffected regions; however, BP of [11C]vinpocetine was higher than [11C]PK11195. Taken together, these studies suggest that microglial activation is of central importance in the pathophysiology of MS and can be visualized with PET imaging using radioligands for the PBR. Several studies have found increased binding of [11C]PK11195 and (R)-[11C]PK11195 in patients with ischemic stroke [32–35]. Binding of (R)-[11C]PK11195 has been found to correspond with areas where T1-weighted MRI shows intensity changes [32]. In patients with chronic middle cerebral artery infarcts, [11C]PK11195 has been used to visualize increased microglial activation in the ipsilateral thalamus [33]. PET imaging with [11C]PK11195 thus seems to be a promising tool in the study of cerebral infarction and could also be useful in the evaluation of neuroprotective strategies, especially with respect to the consequences of microglial activation. Molecular imaging of inflammation in Alzheimer’s disease (AD) and dementia has been reviewed by Versijpt et al. [36]. In a study of eight patients with AD, Cagnin et al. [37] showed in vivo microglial activation in the brain of patients with mild to moderate AD in several brain regions using PET with the radioligand (R)-[11C]PK11195. The spatial distribution of (R)[11C]PK11195 binding matched with the regional distribution of cerebral hypometabolism, as detected with [18F]fluorodeoxyglucose, and correlated with brain atrophy assessed by longitudinal MRI scans. Cagnin et al. also reported high levels of (R)[11C]PK11195 binding in regions not traditionally thought to be involved in AD, such as the thalamus and the brainstem. The authors interpret these findings as a result of microglial activation in regions connected to areas of primary pathology, which might be amplified in the thalamus by the high density of corticothalamic connections. As one review article notes, however, it is not possible to confirm the histological presence of activated microglia in these regions for studies of human subjects [38]. These increases might reflect regional variations in the constitutive PBR expression that are independent of the disease pathology, or the increases might be a non-uniform element of non-specific (R)-[11C]PK11195 binding. One early study in AD patients failed to show any significant increase in [11C]PK11195 uptake compared with controls [39], but this study can be distinguished from later studies by several important methodological differences, the sum of which are thought to have resulted in a substantially lower sensitivity. The early study used a racemic [11C]PK11195 as opposed to the higher affinity R-enantiomer and did not acquire PET data in 3D model. Later studies also used a different application of tracer kinetic modeling for the generation of quantitative parametric maps. Several groups have evaluated various modeling methods for quantification of (R)[11C]PK11195 in AD and baseline scans. Comparative study of
(R)-[11C]PK11195 and [11C]PIB, a specific tracer for imaging bamyloid in AD and mild cognitive impairment, revealed an increase in microglial activation in cortical areas; however, the amyloid deposit load was twofold greater in cortical areas in AD patients [40–42]. More recently, Wiley et al. [43] observed no differences in brain (R)-[11C]PK11195 retention when subjects were grouped by clinical diagnosis or by the presence or absence of b-amyloid pathological findings as indicated by analyses of [11C]PIB retention. These findings suggest that either microglial activation is limited to later stages of severe AD or (R)[11C]PK11195 is too insensitive to detect the level of microglial activation associated with mild to moderate AD. (R)-[11C]PK11195 has been studied in many other forms of dementia, including those of infectious etiology. In a PET study of patients with AIDS, Hammoud et al. [44] observed greater binding of (R)-[11C]PK11195 than in normal controls. However, patients with HIV-associated dementia did not show significant differences in binding when compared to HIV+ nondemented patients. Increased microglial activation has been observed in the PET scanning of (R)-[11C]PK11195 in patients with Huntington’s disease [45]. The formation of microglia has been observed in both symptomatic and presymptomatic Huntington’s disease gene carriers, and the degree of microglial activation in the striatum has been found to correlate with D2 receptor dysfunction as measured by [11C]raclopride PET, as well as with clinical measures of disease severity [46]. In multiple system atrophy (MSA) patients, (R)-[11C]PK11195 binding has been shown to match the known distribution of neuropathologic changes, with increased binding in the dorsolateral prefrontal cortex, putamen, pallidum, pons, and substantia nigra [47]. These findings suggest that microglial activation is an indicator of disease activity and that (R)-[11C]PK11195 PET can thus be used to characterize the in vivo neuropathology of MSA. Dodel et al. [48] recently reported the effect of minocycline in MSA patients with (R)-[11C]PK11195 and PET. Two out of three MSA patients treated with minocycline in this clinical trial showed decreased PBR binding with (R)-[11C]PK11195, whereas most patients receiving placebo experienced a mean increase in binding. Further studies with larger populations are required to address the clinical relevance of (R)-[11C]PK11195 with PET. In patients with Parkinson’s disease (PD), Gerhard et al. [49] reported increased (R)-[11C]PK11195 binding in the pons, basal ganglia, frontal, and temporal cortex as compared with normal controls. These authors also performed longitudinal PET studies using (R)-[11C]PK11195 and [18F]DOPA and found that binding of (R)-[11C]PK11195 in patients with PD remained stable over the two year period of follow-up. Microglial activation was not found to correlate with clinical index or [18F]DOPA binding in this longitudinal study. These findings suggest that microglial activation might be important early in the disease process but remains stable as the disease progresses. Bartels et al. [50] have used [11C]PK11195 in PD patients to evaluate the medication effect of celecoxib, a COX-2 inhibitor. PET analyses showed that PD patients possessed higher contralateral putamen binding potential (BP) and midbrain BP than controls, although considerable overlap was seen and differences were not statistically significant. Unexpectedly, the BP and distribution volume (DV) after celecoxib were slightly higher. These findings suggest that [11C]PK11195 might not be a suitable tracer for the reliable assessment of PD.
REVIEWS
The utility of (R)-[11C] K11195 might be greater in patients with more severe movement disorders. For example, in patients with corticobasal degeneration, increased binding of (R)[11C]PK11195 has been observed in cortical regions and basal ganglia as compared with normal controls [51]. In a study of patients with progressive spastic hemiparesis, Turner et al. [52] reported that two of three patients showed increased binding of (R)-[11C]PK11195 in the motor cortex and supplementary motor region contralateral to the affected limbs, as visualized on PET imaging. No focal areas of increased binding were seen in the cerebral cortex of the third patient, who had a high cervical cord lesion and was presumed to have extra-cerebral inflammatory disease. In amyotrophic lateral sclerosis, patients were found to have significantly increased binding in the motor cortex, pons, dorsolateral prefrontal cortex, and thalamus relative to healthy control subjects. Furthermore, there was a significant correlation between binding in the motor cortex and the burden of upper motor neuron signs clinically [53]. Taken together, these findings suggest that cerebral microglial activation can be detected in vivo during the evolution of several degenerative movement disorders. In a PET study of patients with hepatic encephalopathy (HE), Cagnin et al. [54] found significant increases in glial (R)[11C]PK11195 binding bilaterally in the pallidum, right putamen, and right dorsolateral prefrontal region. The patient with the most severe cognitive impairment had the highest increases in regional (R)-[11C]PK11195 binding. These findings support earlier experimental evidence from rodent models of liver failure and suggest that an altered glial cell state, as evidenced by the increase in cerebral PBR, might be causally related to impaired brain functioning in HE [55]. Iversen et al. [56], however, reported no significant differences in the volume of distribution (VT) of [11C]PK11195 between regions studied or between the HE and control group. These latter results might be interpreted as meaning that microglial activation with expression of PBR is not an important mechanism of degeneration in patients with HE, or the results might simply reflect the inadequacy of the tracer, particularly in its racemic form. One of the latest uses of PK11195 has been in the study of neuropsychiatric disorders such as schizophrenia. In a study of ten patients with recent-onset schizophrenia, BP of (R)-[11C]PK11195 in total gray matter was increased relative to healthy age-matched controls, suggesting that activated microglia might play a part in the loss of gray matter associated with this disease [57]. In a study of seven patients within the schizophrenia spectrum who had recently experienced an episode of psychosis, a significantly higher BP of (R)-[11C]PK11195 was found in the hippocampus relative to healthy age-matched volunteers. It was also found that patients with recent psychosis had 30% higher (R)-[11C]PK11195 BP in the whole-brain gray matter, despite the fact that MR images revealed no visual abnormalities [58]. The results of these two studies suggest that neuroinflammation might have an important role in schizophrenia and that PK11195 combined with PET might provide insight into this process beyond the realm of traditional neuroimaging techniques. [11C]PK11195 has even been used to guide therapy. In a recent report involving a child with refractory seizures secondary to encephalitis, Kumar et al. [59] used [11C]PK11195 to detect areas www.drugdiscoverytoday.com
937
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews GENE TO SCREEN
of neuroinflammation, which were then removed surgically. This led to significant recovery. [11C]PK11195 has also been used in the study of inflammatory diseases outside the CNS. In a study of patients with rheumatoid arthritis, Van der Laken et al. [60] reported higher binding of (R)[11C]PK11195 in PET imaging of knee joints. PET tracer uptake in joints correlated significantly with PBR staining in the sublining of synovial tissue. In a study of patients with the chronic lung disease scleroderma fibrosing alveolitis (FASSc), Branley et al. [61] reported a trend of reduced uptake of (R)-[11C]PK11195 in FASSc patients versus controls and also found that uptake correlated inversely with lung density, which was significantly elevated in FASSc. Reduced uptake of the radiotracer was thought to represent a change in the morphology of lung macrophages with disease progression and accumulation of permanent scar tissue. These results of studies outside the CNS demonstrate that inflammatory cell traffic can be reliably imaged in regions as diverse as knee joints and lung tissue and that PK11195 can be used to assess disease progression.
Limitations of PK11195 Although it is well established that (R)-[11C]PK11195 and [11C] PK11195 show increased CNS retention in a wide array of neurological disorders, there are several methodological and kinetic issues that limit the interpretation and potential of this radioligand. Lower than desirable BP has been attributed to both low affinity for its receptor and relatively low total brain uptake, which results from substantial binding of the tracer to organs in the periphery. In addition to this low level of total binding in the normal brain, [11C]PK11195 has a high level of non-specific binding owing to its lipophilic nature; together, these factors lead to a poor signal-to-noise ratio. The low level of binding in normal brain also renders the modeling of this tracer particularly difficult because effects of no interest such as tissue heterogeneity and vascular signal become predominant [24]. Yet another complicating factor is that [11C]PK11195 demonstrates highly variable kinetic behavior, which further complicates quantitative analysis of PBR receptor density. This variability is thought to result from its extensive binding to plasma proteins, some of which are acutephase reactants that vary in inflammatory conditions, both systemically and locally at the site of acute injury [24]. Refinements in the modeling methods for analysis of PK11195 imaging data have led to incremental improvements in the quantification of PBR [62], but the lack of sensitivity and specificity of [11C]PK11195 and (R)-[11C]PK11195 have so far precluded the development of a standard method of analysis easily applicable to all subjects. Until now, therefore, the interpretations of results have been limited to an emphasis on the general agreement between brain areas of increased PK11195 uptake and the known distribution of a given pathology [5,10]. In addition to these drawbacks, the short half-life of 11C (20.4 min) is a limitation to the dissemination of [11C]-labeled PK11195 for clinical purposes. The longer half-life of 18F (109.8 min) enables remote cyclotron creation of the tracer and is preferred to facilitate both distribution and multiple use of the radiotracer production batches. Another important consideration is in regards to the extent of microglial activation that needs to be present in the CNS before a 938
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
signal can be detected using PK11195. For example, in patients with mild cognitive impairment (a condition that often progresses to Alzheimer’s disease), some studies have observed increased (R)[11C]PK11195 binding in a portion of patients [42], whereas others have not found any significant differences between patients and normal controls [43]. Similarly, (R)-[11C]PK11195 PET imaging was not able to distinguish between HIV-infected patients with dementia and HIV-infected patients without dementia [44]. These studies suggest that low levels of microglial activation might not be detected using PK11195 and underscore the importance of developing ligands that bind to microglia with greater sensitivity and specificity. There are also certain discrepancies in this body of work that remain to be addressed. For example, (R)-[11C]PK11195 has shown increased binding in regions traditionally not associated with neuroinflammation such as the thalamus in several diseases such as AD, PD, and Huntington’s disease [37]. One potential explanation for these findings is that the PBR is being induced in cells other than microglia, such as astrocytes, which has been suggested in animal models of neuronal injury [63,64], as well as in humans [65]. These findings might also reflect regional variations in the constitutive PBR or some degree of non-uniform non-specific binding in the CNS because it is not possible to histopathologically ascertain microglial activation in these regions in human PET studies. These concerns highlight the importance of developing newer ligands that have greater specificity and sensitivity to activated microglia for PET imaging. In light of the limitations of PK11195, many groups worldwide are actively engaged in a search for new ligands with improved capacities to quantify PBR expression. During the past few years, more than 50 new PBR radioligands have been reported in the literature, labeled with the short-lived positron emitters carbon-11 and fluorine-18 (half-life: 109.8 min) or with the single-photon longer-lived emitter iodine-123 (halflife: 13.2 h) [7–10]. Most of these potential tracers are still in the early stages of investigation.
[11C]Vinpocetine Vinpocetine (eburanamenine-14-carboxyic acid ethyl ester), a vinca alkaloid, is an agent that is currently used in the treatment of acute and chronic stroke patients [66] because it is thought to interfere with various stages of the ischemic cascade. The uptake of [11C]vinpocetine in human brain has been shown to distribute rapidly and heterogeneously among brain regions [67]. When compared with the cerebellum, the highest regional uptake has been found in the thalamus, upper brain stem, striatum, and cortex [68]. In a study with oral administration, [11C]vinpocetine accumulated in stomach, liver, brain, and kidney with distribution of brain identical to that of intravenous administration [69]. Vinpocetine binds the PBR with low affinity in vitro (IC50 = 0.2 mM), but its role as a PBR ligand is supported by the finding that pretreatment with vinpocetine substantially decreases later uptake of [11C]PK11195. The uptake of [11C]vinpocetine is increased in brain after PK11195 pretreatment, presumably owing to blockade of PBRs in the periphery [67]. One drawback of vinpocetine is that it binds to other receptors with an affinity similar to its affinity for the PBR (including adrenergic a2b receptors, IC50 = 0.9 mM), a finding that raises questions about its in vivo
specificity for the PBR [68]. [11C]vinpocetine has been used with success in a small PET imaging study of MS patients, in which it showed greater global brain uptake than [11C]PK11195 and increased BP in plaque regions for all four MS patients, whereas [11C]PK11195 showed increased binding in plaque regions of only one of the four patients [70]. However, the affinity of vinpocetine for other receptors and the presence of [11C]ethanol as radiometabolite make quantification of PBR difficult and clinical results less reliable [10].
[11C]DAA1106 and [18F]FEDAA1106 DAA1106, N-(2,5-dimethoxybenzyl)-N-(5-fluoro-2-phenoxyphenyl)acetamide, is a 2-phenoxy-5-fluoroanilide derivative with high affinity and selectivity for the PBR [71]. DAA1106 has a five-fold to six-fold higher affinity for PBR than PK11195. In vivo imaging in monkey brains has demonstrated fourfold higher levels of [11C]DAA1106 binding compared with [11C]PK11195, as well as higher levels of specific binding in experimentally lesioned areas [72]. The [11C]DAA1106 binding was markedly inhibited by unlabeled DAA1106 and PK11195 in the monkey brain, suggesting that most of the [11C]DAA1106 binding represents specific binding. Because of its higher affinity, DAA1106 might serve as a better ligand for labeling of the PBR, as well as for addressing some of the issues related to non-specific binding seen in studies with [11C]PK11195. Ex vivo autoradiography and PET imaging in vivo showed greater retention of [11C]DAA1106 compared with [11C]PK11195 in animal models of neuroinflammation induced with either lipopolysaccharide or 6-hydroxydopamine [73]. DAA1106 binds with higher affinity to microglia in rat models of neuroinflammation when compared with PK11195. In a study comparing the pharmacological binding properties of [3H]PK11195 and [3H]DAA1106, Venneti et al. [74] looked at binding patterns in postmortem tissues from patients with cerebral infarcts, amyotrophic lateral sclerosis, AD, frontotemporal dementia, and MS (n = 10 each). In all diseases, [3H]DAA1106 showed a higher binding affinity, as reflected by lower dissociation constant values than those of [3H]PK11195. Moreover, specific binding of both ligands correlated with the presence of activated microglia identified by immunohistochemistry in situ. These studies suggest that DAA1106 might possess binding characteristics superior to those of PK11195, which might be beneficial for in vivo PET imaging. Use of [11C]DAA1106 in a small study of patients with AD yielded promising results [75]. Mean BP was increased significantly in the brain of AD patients compared with control subjects in areas of known AD pathology, including the dorsal and medial prefrontal cortex, lateral temporal cortex, parietal cortex, occipital cortex, anterior cingulate cortex, striatum, and cerebellum. [11C]DAA1106 binding was also observed in more widespread regions in the AD patients than in earlier studies using [11C]PK11195. The authors suggest that this might be due to the higher affinity and lipophilicity of DAA1106 as compared with PK11195 in the quantification of PBR in vivo. Interpretations of these results should be made cautiously, however, given that DAA1106 and PK11195 were not directly compared in the same subjects. More recently, Fujimura et al. [76] studied various analytic methods for quantification of [18F]FEDAA1106 (fluoroethyl derivative of DAA1106) in healthy humans. The DV was estimated by
REVIEWS
nonlinear least-squares (NLS), Logan plot and multilinear analysis (MA), and these methods were found to be significantly correlated. There was also significant correlation between BP with NLS and DV with NLS, Logan plot or MA; however, the interindividual differences in the DV of the free and non-specific binding compartment (K1/k2) were large. In a simulation study, variation of the DV estimated by Logan plot was small, but it was underestimated as the noise increased. By MA, the bias of DV was smaller, but the variation of DV was larger than by Logan plot. Within a 3% noise level, there was almost no difference between Logan plot and MA in both bias and variation. DVs estimated by both Logan plot and MA were underestimated by 10–20%. Although the variation of DV was larger by NLS than by Logan plot, it was small enough in the noise level of volume of interest analysis, and the bias of DV was 0–2%. These results suggest that NLS is a suitable method for the estimation of [18F]FEDAA1106 binding to PBRs.
[11C]PBR28 PBR28, N-(2-methoxybenzyl)-N-(4-phenoxypyridin-3-yl)acetamide, is a PBR ligand with a lower lipophilicity than PK11195 and DAA1106 [77]. [11C]PBR28 has shown high brain uptake in monkey brain in areas consistent with known PBR distribution in monkey [78]. Similar results were observed in a rat model of cerebral ischemia and stroke [79]. Biodistribution of [11C]PBR28 in healthy humans has been found to match known patterns of PBR distribution, with highest uptake in the PBR-rich organs: lungs, kidneys, and spleen [80]. Kinetic analysis of [11C]PBR28 in healthy human subjects revealed that DVs were only approximately 5% of what had been observed in monkeys [80]. The time– activity curves in two of the twelve subjects seemed as if they had no PBR binding (i.e. rapid peak of uptake and fast washout from brain). The cause(s) of these unusual findings are unknown, but both subjects were also found to lack binding to PBRs in peripheral organs such as lung and kidney. Similarly, one in seven subjects had less binding of [11C]PBR28 (60–90%) in kidneys, spleen, and lungs. The activity in the baseline monkey scans was greater than that in humans for organs with high PBR densities. For this reason, the human effective dose was overestimated by 60% with monkey biodistribution data.
[11C]AC-5216 AC-5216 ([11C]-AC-N-benzyl-N-ethyl-2-(7-methyl-8-oxo-2pheyl-7,8-dihydro-9H-purin-9-yl)acetamide) is a dihydropurin with high affinity for the PBR [81]. [11C]AC5216 has been validated successfully in several TPSO models, including kainic acid rat models [81,82]. PET studies in monkey brain demonstrated high uptake of [11C]AC-5216 in the occipital cortex, a rich PBRdense area in the primate brain [83]. PET studies in human showed that the highest BP, compared with nondisplaceable uptake (BPND), was in the thalamus (4.6 1.0) and binding was lowest in the striatum (3.5 0.7) [84]. The total VT obtained by an NLS method of graphical analysis showed regional distribution similar to BPND. There was no correlation between BPND and VT, however, because of the interindividual variation of K1/k2. BPND obtained with data from a scan time of 60 min was in good agreement with that from a scan time of 90 min (r = 0.87). Regional distribution of [11C]AC-5216 was in good agreement with previous PET studies of PBRs in the human brain. BPND is more www.drugdiscoverytoday.com
939
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Fig._2)TD$IG]
18
O O
O
F
N
N
O
Cl
O O
N
O
O
O CH3
H311C
11
Reviews GENE TO SCREEN
[18F]PBR06
[11C]PBR01
I
N
N
[11C]CLINME 18
F
11
N
N
O N
N
O
N N
CH3 N O N O
N N
O
O 11
N
N
H3 C [11C]DAC
[11C]AC-5216
N [18F]DPA-714 Drug Discovery Today
FIG. 2
Examples of promising radiotracers studied in animals for PBR.
appropriate for estimating [11C]AC-5216 binding than VT is because of the interindividual variation of K1/k2, and with this method, [11C]AC-5216 is a promising PET ligand for quantifying PBR in the human brain.
[11C]DPA-713 DPA-713 (N-diethyl-2-[2-(4-methoxyphenyl)-5,7-dimethyl-pyrazolo[1,5-a]pyrimidin-3-yl]-acetamide) is a high-affinity PBR ligand in the pyrazolopyrimidine group [85] (Fig. 2). [11C]DP-713 was successfully validated in a variety of rat inflammation models and in monkeys [86,87]. In a PET study comparing [11C]DPA-713 to [11C]PK11195, Endres et al. [88] found that in the healthy brain, the average plasma-to-tissue clearance and the total VT of [11C]DPA-713 were an order of magnitude larger than those measured for [11C]PK11195. Studies in patient populations are needed to determine whether [11C]DPA-713 is sensitive enough to evaluate localized elevations in PBR expression. Other promising candidates for the labeling of PBR include [11C]PBR01, [18F]PBR06, [11C]CLINME, [11C]DAC, [11C]AC-5216, and [18F]DPA-714 (Fig. 2). [11C]PBR01 and [18F]PBR06 have shown a high degree of displaceable specific binding in the brain in PET imaging studies of rhesus monkeys [87,89–92]. Of these, [18F]PBR06 might have better kinetics for quantitative analysis, with relatively low non-specific uptake [90]. [11C]CLINME has compared favorably with (R)-[11C]PK11195 in PET imaging of rodents with induced local neuroinflammation, where uptake of [11C]CLINME was identical to that of (R)-[11C]PK11195 in the brain lesion but significantly lower in the intact contralateral hemisphere [91]. [18F]DPA-714 has also been studied in PET imaging of rodents with induced local neuroinflammation, where it performed better than [11C]DPA-713 and (R)-[11C]PK11195, with the highest uptake ratio and BP [91,92].
Concluding remarks Extensive research during the past decade has led to the development of several new PET radioligands for the visualization of the 940
www.drugdiscoverytoday.com
PBR, which have surpassed 11C-labeled PK11195 in efficacy. Among these, [11C]DAA1106 and [11C]PBR28 have demonstrated significantly increased binding in human brain. [18F]DPA-714 is also a promising PET tracer and has recently demonstrated high affinity for the PBR with better uptake and BP than [11C]-(R)PK11195. Comparative studies of these new radioligands relative to PK11195 are needed to further determine which will lead to the best results in human clinical research. Although (R)[11C]PK11195, [11C]DAA1106 and [11F]PBR28 are currently the most extensively studied PET ligands for the quantification of PBR in human, the development of a new PET tracer that is easier to use, more reliable, and more quantifiable is still required for clinical studies to become more widespread and productive. Most research efforts in this field are currently at the level of animal studies and are aimed at the following key issues: improving ligand metabolic stability, decreasing non-specific binding, developing reliable tracer-kinetic modeling, and finding more reliable methods for PBR quantification. Currently available PBR ligands combined with PET have enabled the study of neuroinflammatory conditions in ways beyond the scope of conventional imaging techniques. With additional improvements, PET imaging of the PBR could offer a non-invasive modality for early diagnosis of CNS diseases, as well as a strategy for quantitative assessment of disease progression and treatment response. Although most work to date has focused on adult neurological conditions, detection of PBR expression also has potential applications in pediatric disorders, such as in the detection of perinatal brain injury in newborns exposed to intrauterine insults and as a prognostic indicator for the development of white matter injury and cerebral palsy [93]. There is also evidence to suggest that PET imaging of the PBR could be useful in guiding therapy, such as in outlining seizure foci for surgical removal. Finally, early work in animal models suggests that PBR ligands may have several potentially useful effects, from decreasing inflammation [94,95] to promoting neuronal survival
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
and regeneration in various models of injury [96–98]. PBR ligands have even shown promise as anxiolytic agents, acting via the production of neurosteroids that target the GABAA receptor [99,100]. These initial findings suggest that PBR ligands
REVIEWS
have potential as therapeutic agents for the treatment of neurological and psychiatric disorders and that PET imaging with PBR ligands could help in the further development of drugs for human use.
1 Banati, R.B. (2002) Visualising microglial activation in vivo. Glia 40, 206–217 2 Papadopoulos, V. et al. (2006) Translocator protein (18kDa): new nomenclature for the peripheral-type benzodiazepine receptor based on its structure and molecular function. Trends Pharmacol. Sci. 27, 402–409 3 Casellas, P. et al. (2002) Peripheral benzodiazepine receptors and mitochondrial function. Neurochem. Int. 40, 475–486 4 Kreutzberg, G.W. (1996) Microglia: a sensor for pathological events in the CNS. Trends Neurosci. 19, 312–318 5 Cagnin, A. et al. (2002) In vivo imaging of neuroinflammation. Eur. Neuropsychopharmacol. 12, 581–586 6 Frankle, W.G. et al. (2005) Neuroreceptor imaging in psychiatry: theory and applications. Int. Rev. Neurobiol. 67, 385–440 7 James, M.L. et al. (2006) Development of ligands for the peripheral benzodiazepine receptor. Curr. Med. Chem. 13, 1991–2001 8 Dolle, F. et al. (2009) Radiolabelled molecules for imaging the translocator protein (18kDa) using positron emission tomography. Curr. Med. Chem. 16, 2899–2923 9 Doorduin, J. et al. (2008) PET imaging of the peripheral benzodiazepine receptor: monitoring disease progression and therapy response in neurodegenerative disorders. Curr. Pharm. Des. 14, 3297–3315 10 Chauveau, F. et al. (2008) Nuclear imaging of neuroinflammation: a comprehensive review of [11C]PK11195 challengers. Eur. J. Nucl. Med. Mol. Imaging 35, 2304–2319 11 Watkins, G. et al. (1988) A captive solvent method for rapid N-[11C]methylation of secondary amides: application to the benzodiazepine, 4’-chlorodiazepam (RO54864). Int. J. Radiat. Appl. Istrum. A 39, 441–444 12 Junck, L. et al. (1989) PET imaging of human gliomas with ligands for the peripheral benzodiazepine binding site. Ann. Neurol. 26, 752–758 13 Bergstrom, M. et al. (1986) Peripheral benzodiazepine binding sites in human gliomas evaluated with positron emission tomography. Acta Radiol. Suppl. 369, 409–411 14 Farges, R. et al. (1994) Site-directed mutagenesis of the peripheral benzodiazepine receptor: identification of amino acids implicated in the binding site of Ro 5-4864. Mol. Pharmacol. 46, 1160–1167 15 Wang, J.K.T. et al. (1980) Properties of [3H]diazepam binding sites on rat blood platelets. Life Sci. 27, 1881–1888 16 Shah, F. et al. (1994) Synthesis of the enantiomers of [N-methyl-11C]PK11195 and comparison of their behaviors as radioligands for PK binding sites in rats. Nucl. Med. Biol. 21, 573–581 17 Rao, V.L. and Butterworth, R.F. (1997) Characterization of binding sites for the v3 receptor ligands PK11195 and Ro 5-4864 in human brain. Eur. J. Pharmacol. 340, 89–99 18 Banati, R.B. et al. (2000) The peripheral benzodiazepine binding site in the brain in multiple sclerosis: quantitative in vivo-imaging of microglia as a measure of disease activity. Brain 123, 2321–2337 19 Ishiwata, K. et al. (2007) In vivo evaluation of P-glycoprotein modulation of 8 PET radioligands used clinically. J. Nucl. Med. 48, 81–87 20 Anderson, A.N. et al. (2007) A systematic comparison of kinetic modelling methods generating parametric maps for [(11)C]-(R)-PK11195. Neuroimage 36, 28–37 21 Turkheimer, F.E. et al. (2007) Reference and target region modeling of [11C]-(R)PK11195 brain studies. J. Nucl. Med. 48, 158–167 22 Kropholler, M.A. et al. (2005) Development of a tracer kinetic plasma input model for (R)-[11C]PK11195 brain studies. J. Cereb. Blood Flow Metab. 25, 842–851 23 Schuitemaker, A. et al. (2007) Evaluation of methods for generating parametric (R)[11C] PK11195 binding images. J. Cereb. Blood Flow Metab. 27, 1603–1615 24 Chen, M.K. and Guilarte, T.R. (2008) Translocator protein 18kDa (TSPO): molecular sensor of brain injury and repair. Pharmacol. Ther. 118, 1–17 25 Venneti, S. et al. (2006) The peripheral benzodiazepine receptor (translocator protein 18kDa) in microglia: from pathology to imaging. Prog. Neurobiol. 80, 308–322 26 Hirvonen, J. et al. (2010) Human biodistribution and radiation dosimetry of 11C(R)-PK11195, the prototypic PET ligand to image inflammation. Eur. J. Nucl. Med. Mol. Imaging 37, 606–612
27 Charbonneau, P. et al. (1986) Peripheral-type benzodiazepine receptors in the living heart characterized by positron emission tomography. Circulation 73, 476–483 28 Banati, R.B. et al. (1999) [11C](R)-PK11195 positron emission tomography imaging of activated microglia in vivo in Rasmussen’s encephalitis. Neurology 53, 2199–2203 29 Debruyne, J.C. et al. (2002) Semiquantification of the peripheral-type benzodiazepine ligand [11C]PK11195 in normal human brain and application in multiple sclerosis patients. Acta Neurol. Belg. 102, 127–135 30 Versijpt, J. et al. (2005) Microglial imaging with positron emission tomography and atrophy measurements with magnetic resonance imaging in multiple sclerosis: a correlative study. Mult. Scler. 11, 127–134 31 Vas, A. et al. (2008) Functional neuroimaging in multiple sclerosis with radiolabelled glia markers: preliminary comparative PET studies with [11C]vinpocetine and [11C] PK11195 in patients. J. Neurol. Sci. 264, 9–17 32 Gerhard, A. et al. (2005) Evolution of microglial activation in patients after ischemic stroke: a [11C](R)-PK11195 PET study. Neuroimage 24, 591–595 33 Gerhard, A. et al. (2000) In vivo imaging of activated microglia using [11C] PK11195 and positron emission tomography in patients after ischemic stroke. Neuroreport 11, 2957–2960 34 Pappata, S. et al. (2000) Thalamic microglial activation in ischemic stroke detected in vivo by PET and [11C]PK11195. Neurology 55, 1052–1054 35 Price, C.J. et al. (2006) Intrinsic activated microglia map to the peri-infarct zone in the subacute phase of ischemic stroke. Stroke 37, 1749–1753 36 Versijpt, J. et al. (2005) Functional imaging and psychopathological consequences of inflammation in Alzheimer’s dementia. In Bioimaging in Neurodegeneration (Broderick, P.A., Rahni, D.N., Kolodny, E.H., eds), pp. 75–83, Humana Press 37 Cagnin, A. et al. (2001) In-vivo measurement of activated microglia in dementia. Lancet 358, 461–467 38 Venneti, S. et al. (2009) Imaging microglial activation during neuroinflammation and Alzheimer’s disease. J. Neuroimmune Pharmacol. 4, 227–243 39 Groom, G.N. et al. (1995) PET of peripheral benzodiazepine binding sites in the microgliosis of Alzheimer’s disease. J. Nucl. Med. 36, 2207–2210 40 Edison, P. et al. (2008) Microglia, amyloid, and cognition in Alzheimer’s disease: an [11C](R) PK11195-PET and [11C]PIB-PET study. Neurobiol. Dis. 32, 412–419 41 Kropholler, M.A. et al. (2007) Evaluation of reference regions for (R)-[11C] PK11195 studies in Alzheimer’s disease and mild cognitive impairment. J. Cereb. Blood Flow Metab. 27, 1965–1974 42 Okello, A. et al. (2009) Microglial activation and amyloid deposition in mild cognitive impairment. Neurology 72, 56–62 43 Wiley, C.A. et al. (2009) Carbon 11-labeled Pittsburgh Compound B and carbon 11labeled (R)-PK11195 positron emission tomographic imaging in Alzheimer Disease. Arch. Neurol. 66, 60–67 44 Hammoud, D.A. et al. (2005) Imaging glial cell activation with [11C]-R-PK11195 in patients with AIDS. J. Neurovirol. 11, 346–355 45 Tai, Y.F. et al. (2007) Imaging microglial activation in Huntington’ s disease. Brain Res. Bull. 72, 148–151 46 Tai, Y.F. et al. (2007) Microglial activation in presymptomatic Huntington’s disease gene carriers. Brain 130, 1759–1766 47 Gerhard, A. et al. (2003) [11C](R)-PK11195 PET imaging of microglial activation in multiple system atrophy. Neurology 61, 686–689 48 Dodel, R. et al. (2010) Minocycline 1-year therapy in multiple-system-atrophy: effect on clinical symptoms and [(11)C] (R)-PK11195 PET (MEMSA-trial). Mov. Disord. 25, 97–107 49 Gerhard, A. et al. (2006) In vivo imaging of microglial activation with [11C](R)PK11195 PET in idiopathic Parkinson’s disease. Neurobiol. Dis. 21, 404–412 50 Bartels, A.L. et al. (2010) [11C]-PK11195 PET: quantification of neuroinflammation and a monitor of anti-inflammatory treatment in Parkinson’s disease? Parkinsonism Relat. Disord. 16, 57–59 51 Gerhard, A. et al. (2004) In vivo imaging of microglial activation with [1C](R)PK11195 PET in corticobasal degeneration. Mov. Disord. 19, 1221–1226 52 Turner, M.R. et al. (2005) Mills’ and other isolated upper motor neurone syndromes: in vivo study with 11C-(R)-PK11195 PET. J. Neurol. Neurosurg. Psychiatry 76, 871–874
www.drugdiscoverytoday.com
941
Reviews GENE TO SCREEN
References
REVIEWS
Reviews GENE TO SCREEN
53 Turner, M.R. et al. (2004) Evidence of widespread cerebral microglial activation in amyotrophic lateral sclerosis: an [11C](R)-PK11195 positron emission tomography study. Neurobiol. Dis. 15, 601–609 54 Cagnin, A. et al. (2006) In vivo imaging of cerebral ‘‘peripheral benzodiazepine binding sites’’ in patients with hepatic encephalopathy. Gut 55, 547–553 55 Itzhak, Y. and Norenberg, M.D. (1994) Ammonia-induced upregulation of peripheral-type benzodiazepine receptors in cultured astrocytes labeled with [3H]PK11195. Neurosci. Lett. 177, 35–38 56 Iversen, P. et al. (2006) Peripheral benzodiazepine receptors in the brain of cirrhosis patients with manifest hepatic encephalopathy. Eur. J. Nucl. Med. Mol. Imaging 33, 810–816 57 van Berckel, B.N. et al. (2008) Microglia activation in recent-onset schizophrenia: a quantitative (R)-[11C]PK11195 positron emission tomography study. Biol. Psychiatry 64, 820–822 58 Doorduin, J. et al. (2009) Neuroinflammation in schizophrenia-related psychosis: a PET study. J. Nucl. Med. 50, 1801–1807 59 Kumar, A. et al. (2008) Epilepsy surgery in a case of encephalitis: use of 11C-PK11195 positron emission tomography. Pediatr. Neurol. 38, 439–442 60 van der Laken, C.J. et al. (2008) Noninvasive imaging of macrophages in rheumatoid synovitis using 11C-(R)-PK11195 and positron emission tomography. Arthritis Rheum. 58, 3350–3355 61 Branley, H.M. et al. (2008) PET scanning of macrophages in patients with scleroderma fibrosing alveolitis. Nucl. Med. Biol. 35, 901–909 62 Tomasi, G. et al. (2008) Novel reference region model reveals increased microglial and reduced vascular binding of 11C-(R)-PK11195 in patients with Alzheimer’s disease. J. Nucl. Med. 49, 1249–1256 63 Ji, B. et al. (2008) Imaging of peripheral benzodiazepine receptor expression as biomarkers of detrimental versus beneficial glial responses in mouse models of Alzheimer’s and other CNS pathologies. J. Neurosci. 28, 12255–12267 64 Chen, M.K. et al. (2004) Peripheral benzodiazepine receptor imaging in CNS demyelination: functional implications of anatomical and cellular localization. Brain 127, 1379–1392 65 Cosenza-Nashat, M. et al. (2009) Expression of the translocator protein of 18kDa by microglia, macrophages and astrocytes based on immunohistochemical localization in abnormal human brain. Neuropathol. Appl. Neurobiol. 35, 306–328 66 Bereczki, D. and Fekete, I. (2008) Vinpocetine for acute ischaemic stroke. Cochrane Database Syst. Rev. 1, CD000480 67 Gulya´s, B. et al. (2005) [11C]vinpocetine: a prospective peripheral benzodiazepine receptor ligand for primate PET studies. J. Neurol. Sci. 230, 219–223 68 Gulyas, B. et al. (2002) PET studies on the brain uptake and regional distribution of [11C] vinpocetine in human subjects. Acta Neurol. Scand. 106, 325–332 69 Gulyas, B. et al. (2002) Drug distribution in man: a positron emission tomography study after oral administration of the labelled neuroprotective drug vinpocetine. Eur. J. Nucl. Med. Mol. Imaging 29, 1031–1038 70 Vas, A. et al. (2008) Functional neuroimaging in multiple sclerosis with radiolabelled glia markers: preliminary comparative PET studies with [11C]vinpocetine and [11C]PK11195 in patients. J. Neurol. Sci. 264, 9–17 71 Probst, K.C. et al. (2007) Strategy for improved [(11)C]DAA1106 radiosynthesis and in vivo peripheral benzodiazepine receptor imaging using microPET, evaluation of [(11)C]DAA1106. Nucl. Med. Biol. 34, 439–446 72 Maeda, J. et al. (2004) Novel peripheral benzodiazepine receptor ligand [11C]DAA1106 for PET: an imaging tool for glial cells in the brain. Synapse 52, 283–289 73 Venneti, S. et al. (2007) A comparison of the high-affinity peripheral benzodiazepine receptor ligands DAA1106 and (R)-PK11195 in rat models of neuroinflammation: implications for PET imaging of microglial activation. J. Neurochem. 102, 2118–2131 74 Venneti, S. et al. (2008) The positron emission tomography ligand DAA1106 binds with high affinity to activated microglia in human neurological disorders. J. Neuropathol. Exp. Neurol. 67, 1001–1010 75 Yasuno, F. et al. (2008) Increased binding of peripheral benzodiazepine receptor in Alzheimer’s disease measured by positron emission tomography with [11C]DAA1106. Biol. Psychiatry 64, 835–841 76 Fujimura, Y. et al. (2006) Quantitative analyses of 18F-FEDAA1106 binding to peripheral benzodiazepine receptors in living human brain. J. Nucl. Med. 47, 43–50
942
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
77 Briard, E. et al. (2008) Synthesis and evaluation in monkey of two sensitive 11C-labeled aryloxyanilide ligands for imaging brain peripheral benzodiazepine receptors in vivo. J. Med. Chem. 51, 17–30 78 Imaizumi, M. et al. (2008) Brain and whole-body imaging in nonhuman primates of [11C]PBR28, a promising PET radioligand for peripheral benzodiazepine receptors. Neuroimage 39, 1289–1298 79 Imaizumi, M. et al. (2007) PET imaging with [11C]PBR28 can localize and quantify upregulated peripheral benzodiazepine receptors associated with cerebral ischemia in rat. Neurosci. Lett. 411, 200–205 80 Brown, A.K. et al. (2007) Radiation dosimetry and biodistribution in monkey and man of 11C-PBR28: a PET radioligand to image inflammation. J. Nucl. Med. 48, 2072–2079 81 Zhang, M.R. et al. (2007) 11C-AC-5216: a novel PET ligand for peripheral benzodiazepine receptors in the primate brain. J. Nucl. Med. 48, 1853–1861 82 Yanamoto, K. et al. (2007) In vitro and ex vivo autoradiography studies on peripheral-type benzodiazepine receptor binding using [11C]AC-5216 in normal and kainic acid-lesioned rats. Neurosci. Lett. 428, 59–63 83 Zhang, M.R. et al. (2007) 11C-AC-5216:a novel PET ligand for peripheral benzodiazepine receptors in the primate brain. J. Nucl. Med. 48, 1853–1861 84 Miyoshi, M. et al. (2009) Quantitative analysis of peripheral benzodiazepine receptor in the human brain using PET with (11)C-AC-5216. J. Nucl. Med. 50, 1095–1101 85 James, M.L. et al. (2005) Synthesis and in vivo evaluation of a novel peripheral benzodiazepine receptor PET radioligand. Bioorg. Med. Chem. 13, 6188–6194 86 Boutin, H. et al. (2007) 11C-DPA-713: a novel peripheral benzodiazepine receptor PET ligand for in vivo imaging of neuroinflammation. J. Nucl. Med. 48, 573–581 87 Chauveau, F. et al. (2009) Comparative evaluation of the translocator protein radioligands 11C-DPA-713, 18F-DPA-714, and 11C-PK11195 in a rat model of acute neuroinflammation. J. Nucl. Med. 50, 468–476 88 Endres, C.J. et al. (2009) Initial evaluation of 11C-DPA-713, a novel TSPO PET ligand, in humans. J. Nucl. Med. 50, 1276–1282 89 Yanamoto, K. et al. (2009) Evaluation of N-benzyl-N-[11C]methyl-2-(7-methyl-8oxo-2-phenyl-7,8-dihydro-9H-purin-9-yl)acetamide ([11C]DAC) as a novel translocator protein (18kDa) radioligand in kainic acid-lesioned rat. Synapse 63, 961–971 90 Imaizumi, M. et al. (2007) Kinetic evaluation in nonhuman primates of two new PET ligands for peripheral benzodiazepine receptors in brain. Synapse 61, 595–605 91 Boutin, H. et al. (2007) In vivo imaging of brain lesions with [11C]CLINME, a new PET radioligand of peripheral benzodiazepine receptors. Glia 55, 1459–1468 92 Doorduin, J. et al. (2009) [11C]-DPA-713 and [18F]-DPA-714 as new PET tracers for TSPO: a comparison with [11C]-(R)-PK11195 in a rat model of herpes encephalitis. Mol. Imaging Biol. 11, 386–398 93 Kannan, S. et al. (2009) Positron emission tomography imaging of neuroinflammation. J. Child Neurol. 24, 1190–1199 94 da Silva, M.B. et al. (2004) Involvement of steroids in anti-inflammatory effects of PK11195 in a murine model of pleurisy. Mediators Inflamm. 13, 93–103 95 Torres, S.R. et al. (2000) Anti-inflammatory effects of peripheral benzodiazepine receptor ligands in two mouse models of inflammation. Eur. J. Pharmacol. 408, 199–211 96 Veiga, S. et al. (2005) Ro5-4864, a peripheral benzodiazepine receptor ligand, reduces reactive gliosis and protects hippocampal hilar neurons from kainic acid excitotoxicity. J. Neurosci. Res. 80, 129–137 97 Soustiel, J.F. et al. (2008) The effect of oxygenation level on cerebral post-traumatic apoptosis is modulated by the 18-kDa translocator protein (also known as peripheral-type benzodiazepine receptor) in a rat model of cortical contusion. Neuropathol. Appl. Neurobiol. 34, 412–423 98 Mills, C. et al. (2008) Ro5-4864 promotes neonatal motor neuron survival and nerve regeneration in adult rats. Eur. J. Neurosci. 27, 937–946 99 Taliani, S. et al. (2009) Translocator protein ligands as promising therapeutic tools for anxiety disorders. Curr. Med. Chem. 16, 3359–3380 100 Rupprecht, R. et al. (2009) Translocator protein (18kD) as target for anxiolytics without benzodiazepine-like side effects. Science 325, 490–493
REVIEWS
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
The role and impact of quantitative discovery pathology Steven J. Potts, G. David Young and Frank A. Voelker Flagship Biosciences, Flagstaff, AZ, United States
The decision to advance an early-stage compound into formal preclinical testing depends on confidence in mechanism, efficacy and toxicity profiles. A substantial percentage of this confidence comes from histopathology interpretation, as the local tissue environment contains strong signals of both efficacy and toxicity. Accessing this tissue information is made difficult by biological variability across organs and tissues, an insufficient pool of pathology experts working in discovery, and the high subjectivity and individual isolation of microscope-based observations. This article describes how whole-slide imaging and quantitative analysis by trained pathologists are improving early-stage decision-making.
Introduction Histopathological evaluations of animal tissues are an integral part of business decisions in drug development, especially early in the process, when key decisions are made at small biotech companies or pharmaceutical discovery groups for prioritization of good laboratory practice (GLP) preclinical testing. In science boardroom discussions, the advocate for the compound presents the pathologist’s conclusions and the histopathological findings, often delivered by a consulting or internal pathologist who evaluates many studies with other candidate compounds. The advocate presents small representative images of pathologic changes or, if used, immunohistochemistry (IHC)-stained sections demonstrating the presence or absence of a biomarker protein to support the hypothesis. Unfortunately, the representative images are not typically any more characteristic of the pathologic change than a few hash marks would be of an entire football field.1 The pathologist who performed the study is often not present and busy reading and preparing the results of the next study. This common decision-making scenario in drug development illustrates three problems: a shortage of trained pathologists dedicated to discovery efforts, the biased individual isolation and selection of microscope-based observations, and the wide natural biological variability across single tissue cross-sections. New digital pathology applications that provide whole-slide imaging and automated quantitative analysis are helping to Corresponding author:. Potts, S.J. ([email protected]), ([email protected]) 1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.09.001
address these problems and will become mainstream as the technology becomes more widespread at pharmaceutical discovery groups. With the use of digital pathology systems, the entire tissue section on a glass slide is scanned at high resolution, producing a large whole-slide image that can be stored in a secure Internetaccessible database. The two driving reasons for producing a whole-slide image are to eliminate geography barriers2 and to make the entire tissue section available for quantitative image analysis.
Shortage of pathologists assigned to discovery efforts Table 1 shows the distribution of employment of North American veterinary and medical anatomic pathologists in large pharmaceutical companies and contract research organizations. Within pharmaceutical companies, the industry average seems to be approximately one veterinary pathologist for every one billion dollars in annual sales, with one pathologist supporting on average ten research and development compounds. There is a strong consolidation of veterinary pathologists at several leading preclinical contract research organizations, where they are primarily supporting GLP studies. Several articles have described the impact that pathologists can have in discovery3–6 if they have time available from their GLP preclinical work. Within some large pharmaceutical companies, there are discovery pathology groups (e.g. Discovery Toxicology, Investigative Pathology and Predictive Toxicology), enabling their preclinical pathologists to support earlystage discovery projects with 5–10% of their work time. www.drugdiscoverytoday.com
943
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
TABLE 1
Employment of veterinary and human anatomic pathologists in the pharmaceutical industry.
Reviews GENE TO SCREEN
Pharmaceutical company
Ethical drug salesa ($ millions)
Number of DVM pathologists b
Number of MD anatomic pathologistsc
Number of R&D drugsd
Number of own drugs
Number of drugs under license
Pfizer
43,363
54
–
163
117
46
GlaxoSmithKline
36,506
36
–
249
157
92
Bristol-Myers Squibb
17,715
26
–
84
60
24
Merck & Co
26,191
19
–
173
115
58
Amgen
15,794
16
–
56
40
16
AstraZeneca
32,516
22
–
169
123
46
Eli Lilly
19,140
14
1
131
98
33
Wyeth
15,682
13
–
100
75
25
Novartis
36,172
17
3
165
99
66
Hoffmann-La Roche
30,336
21
4
191
123
68
Abbott
19,466
16
–
82
65
17
Sanofi-Aventis
35,642
20
–
160
103
57
Schering-Plough
14,253
8
–
86
40
46
Boehringer Ingelheim
11,595
11
–
54
38
16
Allergan Johnson & Johnson Biogen-Idec
3502
6
–
34
21
13
29,425
9
1
141
83
58
4097
5
–
41
20
21
e
Number of DVM pathologists
Number of MD pathologists f
Contract research organization
Sales ($ thousands)
Charles River
1343
55
–
Covance
1827
50
3(?)
EPL
29
–
MPI Research
11
–
Huntingdon
10
–
WIL Research
8
–
–
35
Quest Clin Trials
177
Genzyme
–
a
2008 Ethical drug sales, from 2008 company annual reports. b A DVM pathologist is defined in this survey as an employee of a pharmaceutical company having either a veterinary medicine degree (DVM) and board certification (DACVP or ECVP). Compiled from the 2009 Society of Toxicology Pathology membership list, cross-referenced with the ACVP/ECVP membership lists. Only pathologists in North America and Europe were considered in this study. c Data from an informal survey of pharmaceutical executives and managers, October 2009. Only anatomic pathologists in North America supporting pathology work clinical trials were considered in this study. d Number of active R&D projects (July 2009) in early preclinical study through launch, compiled by Pharma Documentation Ring. Data from PharmaProjects (http:// www.pharmaprojects.com/). e 2008 sales from company annual reports. Data not listed for private companies. f From company websites or annual reports.
The individual isolation of microscope-based observations One of the primary roles of the pathologist is to effectively communicate histopathological study results to non-experts on the project team. The solitary nature of the microscope makes this difficult, especially when the discovery client teams are not closely co-located or the pathologist is part of a consulting group or a different organization. Solutions that have been implemented in pathology laboratories include multi-headed microscopes to enable fellow scientists and pathologists to view the same slide at the same time and having the pathologist take representative film or digital photos or images of areas of interest within a tissue section. Historically, these technologies have served well in providing information to others about what the study pathologist interpreted but were not without limitations. Viewing slides using multi-headed micro944
www.drugdiscoverytoday.com
scopes required all individuals to gather at the scope at the same time and spend considerable amounts of time to view the images. Conversely, looking at representative digital or film images allowed individuals to view them at their selected times and in varying location; although this allowed some flexibility in viewing the images, they were forced to rely on the pathologist to take accurate representative images and were further hindered by not discussing the results in a real-time environment.
Wide natural biological variability across tissue sections Although the local tissue environment can provide remarkable insight into both efficacy and toxicity of chemical agents, tissue presents an extremely complex architecture with variable heterogeneity within most tissues. Even in similar tissue types across a
single section, the variability can be high. Fig. 1 illustrates this variability across several tissue samples in different internal studies by the authors, in which cellular signals have been measured at various locations in a single section and then analyzed for their intra-tissue variability. For example, coefficient of variation in a given xenograft for randomly sampled regions of manual microvessel counting averaged 46% and automated area measurements averaged 37%. Neutrophil and lymphocyte counts in rodent livers had lower intra-tissue variability but still higher variability than one would expect in more homogenous samples such as urine or serum. In individual human breast tissue sections, average variability ranged from 10% to 63%. The variability of single regions in a section is clearly high across a tissue section. A much larger study used for FDA clinical submission looked at composite variability (Fig. 1) of regions across individual tissue sections.7 In this case, 15–20 regions were chosen by a trained pathologist and analyzed with image analysis to give an overall score for that tissue. The process was then repeated with two other pathologists, who worked with the same blinded sample, chose their own 15–20 regions and ran image analysis to give an overall score for each tissue. The variation with a composite score is generally much reduced but still ranged from a low of 11% for estrogen receptor measurements to a high of 33% for progesterone receptor measurements. This variability is not a reflection of the pathologist’s capabilities but illustrates the challenging and heterogeneous nature of tissue samples. Because all of the regions analyzed represented similar tissue within the section (e.g. only cancer cells were sampled), it leads to the logical conclusion that it is important to sample as large a region as possible in a tissue section, provided the regions all contain the appropriate target tissue. This variability is leading the industry into the increased use of stereological principals to ensure representative sampling.
Whole-slide imaging Paralleling the shift from processed sheet film to digital images in radiology departments, pathology departments are also – albeit slowly – making the change from glass slides to whole-slide digital images viewed on a computer monitor. The technology for scanning an entire tissue section on a glass slide has been reviewed previously for its utility in drug development.8,9 For the first time, investigators and other pathologists are provided full transparency of a complete tissue section, either viewed at their leisure or with the capability of communicating real-time with the pathologist, via web-based conferencing or internal customized image sharing. Providing complete tissue images for viewing and analysis saves time and empowers investigators to take a more active role in identifying or communicating the areas and possible lesions of interest to the pathologist, saving countless hours of pathology time trying to find what the investigator is looking for. An extremely valuable practice is providing digital images to researchers and, after reviewing the basic anatomical features of target tissues, letting investigators sort through the digital images and come back to the pathologist with questions and/or potential conclusions. This not only frees up the pathologist’s time from sitting with a researcher and reviewing tissues using a multiheaded scope but also encourages the researchers to study closely
REVIEWS
the images in support of their hypothesis. An added benefit is that the researcher spends the time to examine specifics within the tissue without having to repeatedly return to the pathologist for clarification. Although final conclusions are always the responsibility of the pathologist, the interactions with digital slides greatly increase researcher–pathologist interaction, engage researchers and help them to understand the nuances of tissue variability.
Image analysis in histopathology When given a choice, investigators almost always prefer quantitative efficacy data from their pathologists to the usual qualitative pathologist grades (e.g. minimal, mild, moderate and marked). Hence, image analysis of digitized images provides a practical quantifiable means of measuring cellular change and, consequently, replaces subjective with objective evaluation. One can divide histopathology image analysis into three basic overlapping approaches: area-based measurements, cell-based measurements and object-based measurements (Fig. 2). Area-based measurements, in which stains are isolated from each other, can be useful in measuring areas and intensities of various color hues in target tissues. Every stain, such as hematoxylin, eosin or diaminobenzene (DAB), comprises different proportions of the primary color (red, blue or green) elements. The user defines the stain of interest, either as hue on a color wheel ranging from 0 to 1 (e.g. 0.1 = DAB, 0.33 = green, 0.66 = blue, 1.0 = red), or by entering RGB decimal values. Area-based approaches are useful for measuring a single, visually apparent substance in tissue regardless of color and have the advantage that hue width and threshold can be precisely adjusted to ‘digitally see’ what the pathologist sees on the slide. The color is binned into three or more grades or levels of intensity (e.g. 0, 1+, 2+ or 3+ or low, medium, high) and a count is made of the area occupied by each in terms of either pixels or mm2. Color deconvolution is a widely used image analysis process that digitally separates each histology stain.10 The great advantage is that a stain of interest is accurately measured and is not partially obscured by other stains on the slide. In other words, the algorithm can see what the pathologist does not see on the slide. An example of its use would be a case in which DAB is not obscured by hematoxylin in an IHC stain. Another example would be a case in which red coloration of a periodic acid Schiff reaction for glycogen would not be obscured by hematoxylin in a liver section. Cell-based analysis measures RGB values inherent in stains such as DAB or hematoxylin and typically focus on either nuclear or membrane stain affinity. In nuclear stains, each nucleus is identified by the intensity of a counterstain (e.g. hematoxylin) and each cell is assumed to contain one nucleus. Negative and positive nuclei can be determined and counted. Nuclei can also be simultaneously categorized and segregated by size, shape and/or stain intensity, as reflected in cell type, cell-cycle stage or cancerous nuclear effects. This capability can be useful in characterizing changes such as cellular hyperplasia or hypertrophy, cellular infiltration, nuclear pleomorphism and tabulation of specific cell types within a tissue. An example of its use might be to determine percent and degree of DAB positivity of neoplastic cell nuclei. Stromal cell nuclei, which results in false negative nuclear counts, www.drugdiscoverytoday.com
945
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[()TD$FIG]
Tissue
Count
Number of sections sampled in a tissue
Number of tissues
Biological variability across a tissue section (average coefficient of variation)
Variation of individual regions across a single tissue section Microvessel count (manual) 10 9 46% Microvessel area 10 9 37% (automated) Neutrophil counts Mouse livers 8 2 16% T lymphocyte counts Rat livers 21 2 32% HER2 H Score Human breast 7 to 11 5 10% ER H Score Human breast 6 to 10 5 41% PR H Score Human breast 6 to 11 5 63% Variation of different composite regions averaged across a single tissue section 15-20 averaged as HER2 (+3,+2,+1,0 scoring) 17% Human breast 180 one composite region 15-20 averaged as Human breast ER percent positive cells 11% 180 one composite region 15-20 averaged as PR percent positive cells 33% Human breast 180 one composite region
Reviews GENE TO SCREEN
Mouse xenografts Mouse xenografts
Drug Discovery Today
FIG. 1
Tissue cross-sections exhibit high levels of biological variability. The table shows typical coefficients of variation by sampling multiple small related sections across a large tissue section. Typical sampled regions are shown for a human ER breast cancer tissue at middle left, a HER2 breast cancer tissue at middle right, and a rat liver sampled at bottom for T lymphocytes.
946
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[()TD$FIG]
Cell-based IHC nuclear, membrane and some cytoplasm markers apoptosis (TUNL)
Area-based colocalization
Reviews GENE TO SCREEN
proliferation (ki67)
Object-based kidney glomeruli
cytoplasm markers
pancreatic islets rare event detection
area of lesion necrotic tissue size
microvessel analysis
Drug Discovery Today
FIG. 2
Image analysis measurements can be classed as cell-based, area-based, or object-based.
can be excluded if shape and size are substantially different from adjacent neoplastic cell nuclei. More advanced image analysis tools, such as proximity measurements, are required to exclude stromal nuclei with shape and size more similar to neoplastic cell nuclei. Other examples might include measuring hepatocellular hypertrophy (nuclear number/unit area of hepatic parenchyma) and measuring numbers of infiltrating inflammatory cells per unit area in a given tissue. With membrane analysis, the nuclei are also identified with a nuclear counterstain, to assist in finding cells, and individual cell membrane staining is then quantified for intensity and completeness of the membrane stain. Although most commercially available software programs for membrane analysis have been optimized for the clinically widely used human epidermal growth factor receptor 2 (HER2)-stained breast tissue measurements, they can also be used for other tissue types and membrane stains. Cells are classified as 0, 1+, 2+ and 3+ based on their membrane stain intensity, with completeness and percentages of each integrated into a standard HER2 scoring scheme. The longstanding solution to estimating marker expression in both nuclear and membrane analyses is a subjective scoring system using a convention called the ‘H-score’ method.11 With this method, the evaluation scores staining features of cells (e.g. cytoplasmic, nuclear or membranous staining) by intensity of stain according to grades of 0, 1+, 2+ or 3+
and assigns relative percentages of the tumor cells having each grade. Accordingly: H score ¼ ð1Þ ð%1þÞ ð%AreaÞ þ ð2Þ ð%2þÞ ð%AreaÞ þ ð3Þ ð%3þÞ ð%AreaÞ
Example : ð1Þðþ5%1þÞ þ ð2Þ ð65%2þÞ þ ð3Þ ð15%3þÞ ¼ 180 With image analysis, the thresholds that define the intensity levels are programmed into the analysis algorithm (usually with pathologist review and input) and an H-score calculation can be consistently applied across various samples. Object-based measurements are typically more complex than area- or cell-based measurements. In this approach, individual cells are grouped into a single object for measurement and computation of object statistics. For example, a cross-section of a microvessel might contain multiple endothelial cells arranged around a lumen. In this case, what is of interest is the microvessel object, with the identification of positively stained endothelial cells only an intermediate step to finding the complete object. In microvessel analysis, a stain for endothelial cells (e.g. CD31 or Factor VIII) is coupled with color deconvolution methods to identify endothelial cells, and then the algorithm is programmed to identify vessel-like objects.12 The user can automatically filter www.drugdiscoverytoday.com
947
REVIEWS
Reviews GENE TO SCREEN
out existing larger vessels and identify and compute morphology statistics on only smaller angiogenic vessels. Output statistics include individual and mean vessel area, vessel perimeter, lumen area, vascular area and vessel wall thickness computed on each vessel. A common application of microvessel analysis is in classifying vessel types and quantitating degree of angiogenesis in a mouse xenograph neoplasm. There are many other potential object-based measurements, including micrometastasis analysis for locating rare tissue events, such as tumor metastases, or identifying amyloid plaque objects in Alzheimer rodent models.
Whole-section analysis with histology pattern recognition Whether the approach is to measure area, cells or objects, one of the biggest challenges of obtaining accurate image analysis results is the segregation of target tissue (region-of-interest) from other tissues on a slide. Before whole-slide imaging, one commonly captured a dozen representative images of small square areas that contained only the target tissue with a microscope-mounted camera (film and then digital), then performed evaluations and/ or image analysis on these areas. The use of whole-slide imaging enables one or two orders of magnitude more tissue to be analyzed because the image analysis can be conducted across the entire tissue. However, whereas choosing a few representative regions puts the burden on the pathologist to choose appropriate locations, with whole-slide image analysis the burden is again on the pathol-
[()TD$FIG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
ogist to analyze the appropriate tissue on the section. To do this manually it requires tedious manual separation of regions of interest, which can be ineffectively time consuming. To solve the challenge of target tissue specificity, various software vendors (Visiopharm, CRI VectraTM, Definiens Tissue StudioTM and Aperio GenieTM) have introduced automated pattern recognition software as a preprocessing utility to segregate target and nontarget tissue during analysis (Fig. 3). The basic concept is for the computer to recognize microscopic tissue patterns after training by an operator. An individual tissue type can then be selected for subsequent image analysis, without the pathologist making tedious region-of-interest drawings on the image. It thus serves as a preprocessing utility for subsequent analysis by any of the cell-, areaor object-based approaches or can be used independently to calculate various areas of the different tissue types across a slide. In cancer applications, pattern recognition enables measurements such as total neoplastic cell area, total numbers of neoplastic cell nuclei, mean nuclear size and mean neoplastic cell size. There is also great potential for the use of pattern recognition software in toxicologic pathology. Examples in hematoxylin & eosin (H&E)-stained sections include extramedullary hematopoiesis or periarteriola lymphoid sheaths of mouse spleen, hepatocellular necrosis or bile duct hyperplasia of rat liver, follicular cell hypertrophy of rat thyroid gland and myocardial fibrosis or inflammatory cell foci of mouse heart. In diabetes research, the ability to successfully analyze tissue components of the pancreas is a valuable asset to investigators.13,14 Pancreatic islets and acinar
FIG. 3
With histology pattern recognition a pathologist teaches the computer to run analyses across all tissue of a similar type, greatly expanding the sample size. In the image below, histology pattern recognition has identified all the neoplastic regions in a tumor tissue sample, which will then be used to measure protein expression only on these areas. 948
www.drugdiscoverytoday.com
tissue are easily recognized by pattern recognition in H&E-stained sections of mouse pancreas. It is also possible to measure alpha and beta islet cell masses through the use of double IHC staining for alpha and beta cells, in which image analysis could easily recognize both cell types, in addition to acinar cells.
Tissue-staining approaches Image analysis can be used either on standard stains or on special stains. Standard stains, such as hematoxylin and eosin, or other histochemical stains delineate the material but are not identity specific. Image analysis approaches can be used in evaluating cellular hypertrophy or atrophy, cell numbers, tissue infiltrates (e.g. fibrosis) and other structural alterations. A good example would be quantitating cross-sectional area of fibrosis (blue stain) in tissue using a Masson’s trichrome stain. Alternatively, specific stains from protein histochemistry, IHC or in situ hybridization can be quantified using image analysis. However, the complexities of immune staining can lead to false negatives and false positives. False negatives can result when antibody is inappropriate, denatured or used at the wrong concentration; through a loss of antigen through poor fixation and/or diffusion; or because of the presence of antigen at a density below level of detection. Similarly, false positives can occur from several factors, including cross-reactivity of antibody with unintended antigens, nonspecific binding of the antibody to the tissue, presence of endogenous peroxidase or avidity for the avidin–biotin complex, entrapment of normal tissues by tumor cells and leakage of protein from adjacent cells with subsequent absorption by target tumor cells.15 Despite the potential for error in staining, the biggest obstruction to effective analysis occurs when biomarkers are accurately stained but occur naturally in both target and nontarget tissues.
Image analysis results depend on proper histological processing There are several histology guidelines that need to be followed to ensure optimal accuracy in image analysis.16–18 First and foremost is the fixation. The longer a tissue sits unfixed at the prosector’s bench, the greater the chance for degradation of the target biomarker, especially if it is protein and subject to autolytic processes. Protein-preserving fixatives such as Hepes-glutamic acid buffermediated organic solvent protein effect provide preservation of protein antigens for IHC. Samples should be prepared by thinly slicing the fresh tissue into small pieces, placing them individually in perforated cassettes and immersing the filled cassettes quickly into cold buffered formalin. This is because 10% buffered formalin penetrates tissue at a relatively slow rate of approximately 1 mm/ hour, and thick tissue sections might not fix adequately. Insufficient fixation is frequently manifested by uneven IHC or in situ hybridization staining in which the edges of the sections have more intense staining than the center of the section (‘edge effect’). Uneven staining of tissue sections will cause problems in image analysis. Occasionally, immediate exposure to the fixative will actually inhibit staining or inactivate the biomarker, and this will be evidenced by lack of staining at the margins of the section. If tissue samples are being fixed from a large study, it is important that all samples from all animals be handled in a uniform manner to avoid any possible bias in staining during later processing. Any possible inadvertent
REVIEWS
dose-related alterations as a result of uneven fixation should be minimized by randomization of animals for submission to necropsy. It is also important that standard operating procedures be strictly followed by all necropsy prosectors to assure that tissues are harvested in a uniform manner at necropsy. This will assure that given tissues will be removed by similar handling and be subjected to similar short periods of time before fixation. IHC tissue samples should not be fixed for periods longer than 24 hours to avoid excessive cross-linkage of protein. Trimming of the tissues for embedding should be performed in a uniform manner such that the same plane of section is allowed for all samples in a study. All similar tissues for a study should be processed in the same batch, especially if automated processors are used. If this is not possible, it is imperative that identical conditions are documented and that there is overlap of the same tissue in different batches to enable quality control comparison of staining for the same sample. After staining has been completed and the slides have been scanned, the digital images of the samples must be subjected to a preliminary evaluation by the image analysis tools to be used. This is to enable analysis values to be adjusted optimally for all slides in the study and will assure that varying stain intensities between slides are fully recognized by the analysis tool. After this, the final image analysis can be performed in a batch processing action across all images in a study for data collection purposes. Anatomic consistency in sampling is mandatory in obtaining accuracy in image analysis for pathologists. Tissues have distinct regional differences in anatomical substructure and metabolism that must be taken into consideration for any type of analysis. For example, pancreatic islets vary in number depending on the region of the pancreas examined; intestinal and respiratory epithelia also vary in a regional manner, and hepatocytes differ from centrilobular to periportal regions. Despite these inconsistencies, sampling specificity will depend on the universality of the marker being measured. If a marker is known to be widespread in the liver, for example, then the sample area can be larger and could be so nonspecific as to include numbers of lobules rather than just one portion of a lobule. Such a case could lend itself more readily to random sampling procedures for statistical advantages. With regard to representative (or random) sampling, it might be easier to analyze the entire cross-section of an organ than to institute a complex and difficult random sampling procedure.
Concluding remarks The ability to digitize entire tissue specimens on slides and subsequently perform morphometric analysis on the images is valuable in the rapid and consistent measurement of tissue features and biomarkers for pharmaceutical research and development. Image analysis of specific target tissues can be particularly challenging, such as in cases with large and morphologically intricate areas of tissue or when tissue staining is nonspecific. Histology pattern recognition is a useful preprocessing utility capable of identifying and categorizing specific histologic tissue types, thus enabling subsequent analysis of target regions by standard image analysis tools. New technologies in scanning and image analysis help to overcome the lack of access to pathology experts in discovery research and will give investigators more confidence in quantitative data concerning efficacy and toxicity. www.drugdiscoverytoday.com
949
Reviews GENE TO SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
References
Reviews GENE TO SCREEN
1 Cardiff, R.D. et al. (2008) ’One medicine – one pathology’: are veterinary and human pathology prepared? Lab. Invest. 88, 18–26 2 Potts, S.J. (2009) Digital pathology in drug discovery and development: multisite integration. Drug Discov. Today 14, 935–941 3 Germann, P.G. et al. (2002) Importance and impact of discovery pathology. Exp. Toxicol. Pathol. 54, 165–167 4 Burkhardt, J.E. et al. (2002) Practical aspects of discovery pathology. Toxicol. Pathol. 30, 8–10 5 Ryan, A.M. (1999) Commentary: role of the pathologist in the identification and characterization of therapeutic molecules. Toxicol. Pathol. 27, 474–476 6 Cockerell, G.L. et al. (1999) Focus on: discovery pathology. Toxicol. Pathol. 27, 471 7 Nassar, A. et al. (2008) Trainable IHC HER2 image analysis system for Dako’s HercepTest and Ventana’s PATHWAY HER2. United States and Canadian Academy of Pathology (USCAP) Annual Meeting 8 Ying, X. and Monticello, T. (2006) Modern imaging technologies in toxicologic pathology. Toxicol. Pathol. 34, 815–826 9 Ying, X. et al. (2004) Digital microscopy imaging and new approaches in toxicologic pathology. Toxicol. Pathol. 32 (Suppl. 2), 49–58 10 Ruifrok, A.C. and Johnston, D.A. (2001) Quantification of histochemical staining by color deconvolution. Anal. Quant. Cytol. Histol. 23, 291–299
950
www.drugdiscoverytoday.com
11 McCarty, K.S., Jr et al. (1985) Estrogen receptor analyses. Arch. Pathol. Lab. Med. 109, 716–721 12 Potts, S.J. (2008) Angiogenesis measurements using digital pathology. Am. Soc. Clin. Pathol. Lab. Med. 39, 265–271 13 Duttaroy, A. et al. (2005) The DPP-4 inhibitor vildgliptin increases pancreatic beta cell neogenesis and decreses apoptosis. Poster presented at American Diabetes Association Annual Scientific Sessions, June 10–14, San Diego, CA, Poster 572-P 14 Duttaroy, A. et al. (2005) Head-to-head comparison of the DPP-4 inhibitor vildagliptin with exendin-4 in a model of pancreatic beta cell injury. Presented at American Diabetes Association Annual Scientific Sessions, June 10–14, San Diego, CA, Oral Presentation 267-OR 15 Rojo, J. (2004) Immunohistochemistry. Surgical Pathology (9th edn), p. 46, Mosby 16 Goldstein, N.S. et al. (2007) Recommendations for improved standardization of immunohistochemistry. Appl. Immunohistochem. Mol. Morphol. 15, 124–133 17 Walker, R.A. (2006) Quantification of immunohistochemistry—issues concerning methods, utility and semiquantitative assessment I. Histopathology 49, 406–410 18 Taylor, C.R. and Levenson, R.M. (2006) Quantification of immunohistochemistry— issues concerning methods, utility and semiquantitative assessment II. Histopathology 49, 411–424
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews INFORMATICS
Dynamics and flexibility of G-protein-coupled receptor conformations and their relevance to drug design Nagarajan Vaidehi Division of Immunology, Beckman Research Institute of the City of Hope, 1500 E. Duarte Road, Duarte, CA 91010, USA
G-protein coupled receptors (GPCRs) are seven helical transmembrane proteins with functional diversity and form the largest superfamily of drug targets. The functional diversity of these receptors stems from the conformational flexibility of the receptor, the nature of the ligand activating the receptor, and the intracellular protein that the receptor couples to. A molecular level understanding of the influence of each of these factors will greatly aid the design of functional selective drugs. In this review, the current state of our understanding of the conformational flexibility and dynamics of class A GPCRs derived from a confluence of biophysical and computational techniques is elucidated.
Introduction G-protein-coupled receptors (GPCRs), also known as the seventransmembrane (TM) receptors, form a superfamily of membranebound receptors that mediate cell signaling and are important drug targets. GPCRs share the topology of seven TM helices connected by intracellular and extracellular loops. Despite having a common structural topology, the amino acid sequences of GPCRs are diverse, and so are their functions. This functional diversity could be associated with subtle three-dimensional structural differences such as helical kinks and tilts and the rotation orientation and dynamics of these structural features [1,2]. The GPCR functional diversity is also dependent on the proteins the receptor couples to, such as the G-protein subtypes, the proteins in the Gprotein-independent pathways (e.g. the b-arrestin pathway) [3] or other complex regulatory pathways. An understanding of the three-dimensional structural information and its related dynamics for GPCRs is vital to aid drug design. A seminal breakthrough came with the recent surge of five highresolution crystal structures of human b2-adrenergic receptor (b2AR), avian b1-adrenergic receptor, adenosine receptor A2A, squid rhodopsin and ligand-free opsin, with and without the Gprotein peptide bound [4–10]. This surge facilitates crystallization of other class A GPCRs and opens new doors for understanding the dynamics of GPCR conformations and drug discovery research on class A GPCRs. The remarkable similarity of the structures of E-mail address: [email protected]. 1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.08.018
b-adrenergic receptors [4,6] and rhodopsin [11,12] with less than 20% sequence similarity reveals high structural similarity in the TM regions in class A GPCRs, in accordance with the results of earlier biophysical studies on biogenic amine receptors. There are, however, subtle but important differences in the structures, especially in the extracellular and intracellular loop regions that are crucial to ligand access to the binding site, ligand binding and Gprotein coupling.
Dynamic behavior of GPCR structures When activated by a ligand, GPCRs in turn activate the G-protein trimer, which leads to secondary messenger production after a cascade of signaling events. It should be noted that certain sevenTM receptors mediate signaling through b-arrestin and not require G-protein activation [13]. Many GPCRs or their constitutive active mutants show a basal activity producing the secondary messenger, even in the absence of an agonist. There is ample experimental evidence showing that GPCR active- and inactivestate conformations are in dynamic equilibrium [14], and the relative concentration of the states is determined by the receptor basal activity. An understanding of the structural features of the various active and inactive conformations of GPCRs and the molecular mechanism of the conformational changes from the inactive to the active states are important in designing functionally specific drugs [15,16]. In fact, this dynamic flexibility is a major bottleneck in obtaining pure protein for biophysical structural studies of GPCRs [17,18]. www.drugdiscoverytoday.com
951
REVIEWS
[(Figure_1)TD$IG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Ligands with different efficacies stabilize varied receptor conformations
Reviews INFORMATICS
In this review, the word ‘efficacy’ means the measured level of activity of the ligand at a GPCR. Ligands to GPCRs are of different sizes and shapes, ranging from small molecules to large proteins. Because GPCRs are drug targets, there are several known synthetic ligands with different efficacies. An agonist is a ligand, the binding of which leads to full activation of the receptor, resulting in a measurable increase or decrease in the secondary messenger. Partial agonists elicit less than maximal response, even at saturating concentrations of the ligand. Inverse agonists reduce the basal activity of the receptor, and antagonists prevent the agonist binding and activation. Several optical techniques, as well as NMR, have been used to study the dynamics of GPCR conformations in the presence of ligands of varied efficacies, as detailed later in this review. A detailed account of the similarities of and differences between the crystal structures of rhodopsin, b1- and b2-adrenergic receptors, adenosine receptor and opsin structures has been given in other reviews [19,20]. In brief, comparison of the crystal structures of the four GPCRs solved to date (namely bovine rhodopsin with covalently bound inverse agonist 11-cis retinal [12], the human b2AR bound to the inverse agonists carazolol [4] and timolol [21], avian b1AR bound to the antagonist cyanopindolol [6], and the human A2A adenosine receptor bound to the antagonist ZM241385 [7]) shows that the overall structures are similar, with most of the differences stemming from the extracellular surfaces. The root mean square deviation in coordinates for the different ˚ [20]. There are several differences in the receptor is within 3 A location of the inverse agonist or antagonist binding sites in these receptors. Whereas the antagonist ZM241385 binds vertically between TM helices 3 and 6 and extends upwards toward extracellular loop 2 (ECL2) in A2A adenosine receptor, the inverse agonists carazolol and timolol in b2AR and antagonist cyanopindolol in b1-adrenergic receptor bind horizontally in similar locations between TM3, TM5, TM6 and TM7. The ‘ionic lock’, which is the salt bridge between Arg1353.50 and Glu2476.30 present in rhodopsin inactive (or dark) state, is not seen in the crystal structures of the adrenergic or the adenosine receptors. This could mean that there are multiple inactive states or simply that the conformations that have been crystallized are not the inactive states of the receptor. Here, we have used the Ballesteros and Weinstein numbering for class A GPCRs.a
Conformational flexibility and activation mechanism of class A GPCRs The substituted-cysteine accessibility method (SCAM) is used to measure the solvent accessibility of residues by mutating them to cysteine residues and measuring the reactivity to sulfhydryl reagents. SCAM was used to map the residues involved in ligand binding, as well as residues inside the TM bundle accessible to water, in the dopamine D2 receptor. At a time when very little structural information was available on the conformational flexibility of GPCRs, SCAM experiments provided valuable informaa The first number in the superscript represents the TM helix in which the residue is located, and the second number refers to the residue number with respect to the most conserved residue in each helix among class A GPCRs [22].
952
www.drugdiscoverytoday.com
FIGURE 1
Cyan is the rhodopsin (1GZM) crystal structure; yellow is the opsin crystal structure (3DQB). (a) Y2235.58 in the opsin structure (yellow) breaks the ionic lock between E2476.30 and R1353.50. (b) Elongation of the helical region in the intracellular part of TM5 in the opsin structure. (c) TM5 in opsin is more tilted toward TM6, and both TM5 and TM6 are tilted outward from the TM bundle of the receptor.
tion on the flexibility of the receptor conformations, especially the residues on TM5, in the D2 dopamine receptor [23]. Ballesteros et al. [24,25] proposed that highly conserved motifs in class A GPCRs, characterized as ‘functional micro-domains’, are involved in the receptor conformational changes and, hence, in the GPCR activation process. They proposed that movements of these functional micro-domains could be conserved in the activation mechanism of class A GPCRs and in constitutively active mutants of GPCRs [24,26–28]. For example, in rhodopsin and adrenergic receptors, movement about the helical kink formed by the highly conserved WXP motif in TM6, the movement in TM7 around the NPXXY motif and the movement in the region around the ionic lock between Arg3.50 and Glu6.30 are all possibly conserved activation switches triggered in these functional micro-domains. Comparison of the rhodopsin and opsin crystal structures also shows considerable conformational changes that occur upon activation. The ionic lock between Arg1353.50 and Glu2476.30 present in rhodopsin inactive state is broken in the opsin structure. The residue Tyr2275.58 is pointing toward the lipid in the inactive rhodopsin structure, whereas this residue points toward the core of the TM region, between TM3 and TM6, in the opsin structure. This, in turn, leads to the disruption of the ionic lock in the partial active opsin structure (Fig. 1a). This is a possible conformational switch for the activation of class A GPCRs. Another activation conformational switch, known as the ‘rotamer toggle switch’, was originally observed in rhodopsin [29,30]. Upon activation, the side-chain rotamer of the residue Trp2656.48 flips, breaking the water-mediated hydrogen bond with Asn3027.49. Turning on the rotamer toggle switch upon activation leads to more flexibility in the active state than in the inactive state of rhodopsin, showing again that GPCRs have conformational flexibility. The differences between the partially active state opsin crystal structure and the crystal structure of rhodopsin in the inactive state are substantial and comparable to the differences between different GPCRs with less than 30% sequence
identity. There are substantial rearrangements in all three intracellular loops in the opsin structure, and the intracellular end of TM5 shows elongation of the helical regions (Fig. 1b). As shown in Fig. 1c, TM5 is also more tilted toward TM6, and both TM5 and TM6 are tilted outward from the TM bundle of the receptor [10]. Thus, it is evident that activation is associated with considerable conformational changes in the receptor. Spin labeling and EPR studies on rhodopsin before the publication of the opsin structure have shown that the ionic lock breaks upon activation and the rotamer toggle switch is also used for the activation of rhodopsin [29,31]. Solid-state NMR studies showed that the interhelical contact between the side chain of Glu1223.37 and the backbone oxygen of His2115.46 breaks upon activation of rhodopsin [32,33]. More recent NMR experiments combined with NMR-derived distance-restrained molecular dynamics (MD) simulations elucidated a chain of molecular events leading to activation in rhodopsin [34]. The light-activated isomerization of 11-cisretinal to all-trans retinal leads to changes in the conformation of ECL2, which triggers the movement of TM5, TM6 and TM7 [35]. Movement of these helices leads to breaking of the ionic lock between Arg1353.50 on TM3 and Glu2476.30 on TM6. Side-chain conformations of Trp2656.48, Tyr2235.58 and Tyr3067.53 change to assist the breakage of the ionic lock [36]. Vogel and coworkers studied the effect of membrane on the activation of rhodopsin and concluded that the membrane environment is enthalpically less favorable for the receptor activation steps such as the coupling of the breaking of the ionic lock and the uptake of protons by the cytoplasmic Glu1343.49 [37]. Experiments engineering metal ion binding sites and tethered ligands show that the extracellular regions of TM6 and TM7 move inward toward TM3 upon activation as the intracellular regions of TM6 and TM7 move outward from the protein core [38]. Biophysical experiments on purified fluorescently labeled b2AR to study the time evolution of the receptor conformational change upon exposure to various ligands of varied efficacies have shown that partial and full agonists stabilize conformationally distinct active states of the receptor [39]. Full agonists show a biphasic behavior in fluorescence intensity change with time, with a fast step preceding the slow step, whereas a strong partial agonist showed only a slow step and a weak partial agonist showed only a fast step. Yao et al. [40] showed that the conformational switches in the functional micro-domains used for activation by agonists of varied efficacies could also be different. Several other fluorescence experiments, such as fluorescence resonance energy transfer and bioluminescence resonance energy transfer in intact cells, have shown that peptide or protein agonist binding shows biphasic behavior. Kinetic fluorescence resonance energy transfer studies have shown that the efficacy of an agonist is also dependent on the rate of conformational change in the GPCRs [41]. More recently, Kobilka and coworkers have used NMR experiments to show that the extracellular loop conformations are distinct and vary when an agonist is bound as compared to an inverse agonist [42]. Thus, there is ample evidence that ligands of different efficacies could stabilize or select distinct receptor conformational states that have varied affinities to other downstream proteins such as the G proteins. The conformational flexibility, therefore, poses considerable complexity to drug design because it requires the knowledge of various
REVIEWS
conformational states to design a drug to shut down a particular active state of the receptor [43].
Relevance of GPCR conformational flexibility to drug design The GPCRs constitute the largest single group of targets for approved drugs, but their importance for future drug design has increased with the awareness that new approaches, such as functional selectivity [44], can permit novel approaches to drug design, even for ‘old’ GPCR targets. The receptor conformations selected or stabilized by ligands of varied efficacies could be different. There is also a growing body of literature that GPCRs signal through barrestin [45], a G-protein-independent pathway. The receptor conformations adapted for b-arrestin binding could be different from that conformation when coupled to the G protein [46]; hence, knowledge of the dynamics of various receptor conformations would immensely aid the design of biased ligands or functionally selective ligands.
Computational methods in studying GPCR conformational flexibility Obtaining a static image of any of the GPCR conformations through X-ray crystallography is by itself a tedious task, largely because of the conformational flexibility of these receptors. It is considerably more difficult to fully describe the dynamics of the receptor conformational states, using any single experimental method. Computational methods have a vital role in providing an atomistic-level model of the GPCR conformational dynamics and thereby aid design of conformation-specific drugs [47–55]. Predicted GPCR structural models can also be used to generate hypotheses as to which residues to label for activation studies. Here, I detail the state of the computational methods and demonstrate how they could be used to gain insights into conformational dynamics of GPCRs.
Molecular dynamics simulations for understanding early events in the activation of class A GPCRs Starting from crystal structures of the inactive states, MD simulations provide an atomistic-level mapping of the conformational changes representing the early events leading to activation of GPCRs. There have been several MD simulation studies, ranging from several hundreds of nanoseconds to a few microseconds, to study the early events in the activation of rhodopsin [56–60]. Crozier et al. [57] found substantial conformational changes in the TM helices TM5 and TM6, apart from shifting of counterions of the Schiff base in retinal from Glu1133.28 to Glu181 on the ECL2. But the counterion switch mechanism was not supported by the microsecond simulations by Grossfield et al. [59], stating that in the metarhodopsin-I state, it is more probable that the Schiff base forms a complex counterion with both Glu1133.28 and Glu181. The activation mechanism of rhodopsin from MD trajectories has also been analyzed in light of structural changes that occur in functional micro-domains. Saam et al. [56] observed that the ionic lock between Arg1353.50 on TM3 and Glu2476.30 on TM6 is weakened upon activation during the MD simulations. Although it is known from experiments that receptor lipid interactions are crucial to the function of GPCRs [4,61,62], atomic-level insights on the role of lipid–receptor interactions were provided by long-time-scale MD simulations. Khelashvili et al. [60] have shown www.drugdiscoverytoday.com
953
Reviews INFORMATICS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews INFORMATICS
recently that cholesterol interaction with rhodopsin modulates the functional micro-domain formed by TM1–TM2–TM7–helix8. Although many of these conformational changes are known to occur from experiments, MD simulations provide detailed atomiclevel understanding of the experimental data on the mechanism of activation. Unlike in the rhodopsin crystal structure, the ionic lock is not present in any of the other GPCR crystal structures solved so far. Dror et al. [63] performed microsecond-level MD simulations on the b2AR crystal structure with and without T4L lysozyme bound. They showed that both the inverse agonist carazolol-bound b2AR and the apo protein b2AR show a statistically significant population of the ionic lock formation in the inactive state of b2AR, although it was not found in the crystal structure. The formation of the ionic lock, however, was less likely to occur during the MD simulations of b2AR with carazolol and T4L lysozyme attached, as observed in the crystal structures. Thus, studying the dynamics complements the information obtained from the crystal structures. There are several caveats in the simulation methods that should be noted when interpreting the results. MD simulations on GPCRs use an explicit description of the lipid bilayer and water. Only one type of lipid is used in these simulations, however, and this in no way reflects the variety of lipids and their composition in the cell membrane. The time scale of simulations is a serious bottleneck for MD simulations. Very few research groups can afford microsecond-level simulations in explicit membrane and water. The longtime-scale MD simulations (low microseconds) starting from an inactive state of the receptor have not shown any features of the active-state receptor conformation. If the energy barriers involved in conformational changes are high enough for MD simulations at room temperature, the sampling could be limited to only those conformations that are accessible at room temperature. This limitation in MD simulations can be overcome using steered MD simulation techniques. Steered MD simulations enable transition from one conformational state to another using constraints. Such simulations, however, require prior knowledge of the active-state conformation or the activation process. Simulated annealing MD techniques combined with the available experimental data on the active-state conformation as constraints to optimize the activestate model provide atomic level insight but are limited by the constraints imposed on the system [64]. The methods described in the section below do not have the limitation of crossing energy barriers.
Other computational methods to understand the activation of GPCRs Elastic network models (ENMs) have been used to study the possible alternate conformations sampled by a protein in its native state, around the global minimum. This method involves computing the principal components of molecular motion by inverting the Hessian matrix calculated from applying a uniform harmonic potential applied to inter-residue contacts from the native state or from experimental data. Application of the ENM to study rhodopsin active state has shown good agreement in predicting the residues in the binding site of all-trans retinal and the motion opening the TM bundle in the intracellular regions [58]. 954
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Starting with the intermediate structure of lumirhodopsin, Costanzi and coworkers simulated the activation pathway of going from lumi to meta-II rhodopsin by combining MD simulations and ENM coarse-grain modeling, in two steps [65]. The first step was to map the local changes in the binding site of all-trans-retinal using restraints obtained from experimental measurements combined with MD simulations. Subsequently, the global changes in conformations were mapped by deriving an ENM model from the MD simulation results of the first step. The simulation of the pathway to meta-II structure showed a strong correlation between the forward and backward transition rates and the phenotype of several rhodopsin mutants. Some of the constitutively active mutants showed faster transitions from the inactive to the meta-II state compared to the mutations that reduce the activity of the receptor [65].
LITiCon method and its applications to study activation mechanisms GPCRs have a seven-TM helical topology that enables the receptor conformations to be sampled on a coarse-grain level, by performing systematic rigid body motion of the seven helices allowing a certain level of flexibility in the helices. Such methods could overcome energy barriers and enable sampling of various conformational states that are not accessible by MD simulations. The ligand-induced transmembrane conformational changes (LITiCon) computational method is one such method [66–68]. It optimizes the ligand-selective receptor conformation for a given ligand, by simultaneous optimization of the rotational orientations of the seven TM helices (including the side-chain rearrangements), while allowing for the ligand to reorient and redock into the reshaped binding site. The change in the ligand-binding energy for each sampled receptor conformation is calculated. This procedure generates a multidimensional binding energy surface from which all the local minima are identified. The local minima are then sorted, and the best minimum energy conformation – ranked by the binding energy and receptor stability – is chosen as the ligand-stabilized GPCR conformation. The LITiCon procedure enables the backbone and the side chains of the receptor to optimize to the ligand conformation and hence can be used as a receptor-flexible ligand-docking program for GPCRs. Because it is more difficult to obtain structural information for the active state of the receptor, the LITiCon method can be used to map the gross conformational changes in the receptor upon ligand binding. We have demonstrated the use of the LITiCon method for providing a molecular level insight integrating all experimental data available on the activation of rhodopsin. For example, although it was known from experiments that the TM6 in rhodopsin undergoes a dynamic change in the kink angle, that the rotamer of Trp2656.48 toggles upon activation [29] and that the ionic lock between TM3 and TM6 breaks upon activation, the mechanistic link behind these processes and their energetics was not known. Using the LITiCon method, Bhattacharya et al. [66] showed that the isomerization of 11-cis-retinal to all-trans-retinal frees up the side chain of Trp2656.48, leading to a change in its sidechain conformation, which in turn modulates the bend angle of TM6 around the conserved Pro2676.50. They showed a direct correlation between the change in the helical kink angle and the rotamer side-chain angle of Trp2656.48; the energy required
for the rotamer toggle comes from water-mediated hydrogen bonds. As a result, the intracellular ends of TM5 and TM6 move outward from the protein core, causing large conformational changes at the cytoplasmic interface. The predicted outward movements of TM5 and TM6 together as a unit were found to be in good agreement with the crystal structure of opsin that was published subsequent to this computational study [66] (Fig. 1c). The predicted active-state conformation of rhodopsin also leads to the formation of new stabilizing interhelical hydrogen-bond contacts, such as those between Trp2656.48 and His2115.46 and Glu1223.37 and Cys1674.56. These hydrogen-bond contacts serve as potential conformational switches, offering new opportunities for future experimental investigations.
Monte Carlo method for calculating the activation pathways The LITiCon method has been used to predict the active-state model of b2AR with ligands of varied efficacies. Using a biased Monte Carlo method on the multidimensional coarse-grain binding energy surface generated by LITiCon, Bhattacharya and Vaidehi calculated the possible activation pathways taken by the receptor in the presence of ligand of varied efficacies [68]. Analysis of the ligand-binding energy landscapes showed that the inverse agonist carazolol stabi-
[(Figure_2)TD$IG]
REVIEWS
lizes the receptor in a deep energy well with high barriers to access the agonist-stabilized states (shown in Fig. 2a). Agonist-bound binding energy landscape, however, such as that of norepinephrine and epinephrine, are highly flexible with an ensemble of degenerate states, showing a favorable energy channel connecting the carazolol-bound conformational state to the norepinephrine-bound state (Fig. 2b). This implies that the receptor is flexible and able to sample the inverse agonist states while bound to the agonist, norepinephrine. On the contrary, the inverse agonist carazolol trapped the receptor in the inactive conformation, making the agonist-bound states inaccessible, thus reducing the basal activity of the receptor. In the salbutamol, a partial agonist stabilized binding energy surface (Fig. 2c), the carazolol and the salbutamol states were parts of two distinct energy wells separated by a barrier and the salbutamolbound receptor potential well was deeper and broader than the carazolol-bound well. Binding of partial agonists such as salbutamol stabilizes different receptor conformations. Subsequent virtual ligand screening calculations on the epinephrine-bound predicted model showed enrichment of agonists, and the crystal structure showed poor enrichment for agonists and substantial enrichment for the antagonists. This is a clear demonstration of the usefulness of these predicted receptor state models for drug design. Mapping of
FIGURE 2
Binding energy surfaces for b2AR with (a) inverse agonist carazolol; (b) full agonist norepinephrine; (c) partial agonist salbutamol. The various calculated ligandstabilized states are marked. Norepinephrine is abbreviated as norepi. (d) Structures of b2AR ligands. The inverse-agonist-bound surface shows distinct separations between conformational states, whereas the agonist-bound surface shows more degeneracy among the conformation states. The above figure has been taken from reference [68] with permission. www.drugdiscoverytoday.com
955
Reviews INFORMATICS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
the flexibility of the energy surface is also important in designing mutants that stabilize various receptor conformations [18,69].
Concluding remarks
Reviews INFORMATICS
GPCRs exhibit dynamic conformational flexibility, leading to a dynamic equilibrium between several inactive- and active-state conformations. On the basis of our calculation of the potential energy surfaces of GPCR crystal structures, I postulate that there is a continuum of GPCR conformational states that could be stabilized by ligands of different efficacies for a given signaling pathway and the proteins that couple to the receptor. For example, the Gprotein-coupled signaling pathways could select a receptor conformation that is different from the states selected by b-arresting coupled pathways. An understanding of the continuum of states and the energetics of the dynamics between them is important in
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
designing ligands with functional specificity to a signaling pathway, the so-called ‘biased ligand’ design. I have enumerated various computational methods available in literature to model the dynamics of the GPCR conformations. Of course, there are several approximations involved in all these calculations that I have enumerated in each section, and these approximations should be improved upon in the future. However, there is a strong need for computational methods: first, to generate an ensemble of structural models that can be used for drug design and structure–activity studies, and second, to identify residues to be labeled for activation mechanism studies. Computational methods to calculate the potential energy surface of these receptors will also aid in understanding the various discrepancies observed in the crystal structures that stem from the artifacts of the protein construct used for the crystallization of GPCRs.
References 1 Yohannan, S. et al. (2004) The evolution of transmembrane helix kinks and the structural diversity of G protein-coupled receptors. Proc. Natl. Acad. Sci. U. S. A. 101, 959–963 2 Hall, S.E. et al. (2009) Position of helical kinks in membrane protein crystal structures and the accuracy of computational prediction. J. Mol. Graph. Model. 27, 944–950 3 Kovacs, J.J. et al. (2009) Arrestin development: emerging roles for beta-arrestins in developmental signaling pathways. Dev. Cell 17, 443–458 4 Cherezov, V. et al. (2007) High-resolution crystal structure of an engineered human beta(2)-adrenergic G protein-coupled receptor. Science 318, 1258–1265 5 Rosenbaum, D.M. et al. (2007) GPCR engineering yields high-resolution structural insights into beta(2)-adrenergic receptor function. Science 318, 1266–1273 6 Warne, T. et al. (2008) Structure of a beta(1)-adrenergic G-protein-coupled receptor. Nature 454, 486–491 7 Jaakola, V.P. et al. (2008) The 2.6 angstrom crystal structure of a human A2A adenosine receptor bound to an antagonist. Science 322, 1211–1217 8 Murakami, M. and Kouyama, T. (2008) Crystal structure of squid rhodopsin. Nature 453, 363–367 9 Park, J.H. et al. (2008) Crystal structure of the ligand-free G-protein-coupled receptor opsin. Nature 454, 183–187 10 Scheerer, P. et al. (2008) Crystal structure of opsin in its G-protein interacting conformation. Nature 455, 497–502 11 Palczewski, K. et al. (2000) Crystal structure of rhodopsin: a G protein-coupled receptor. Science 289, 739–745 12 Li, J. et al. (2004) Structure of bovine rhodopsin in a trigonal crystal form. J. Mol. Biol. 343, 1409–1438 13 Violin, J.D. and Lefkowitz, R.J. (2007) Beta-arrestin-biased ligands at seventransmembrane receptors. Trends Pharmacol. Sci. 28, 416–422 14 Deupi, X. and Kobilka, B. (2007) Activation of G-protein coupled receptors. Adv. Protein Chem. 74, 137–166 15 Galandrin, S. et al. (2007) The evasive nature of drug efficacy: implications for drug discovery. Trends Pharmacol. Sci. 28, 423–430 16 Mailman, R.B. and Murthy, V. (2010) Ligand functional selectivity advances our understanding of drug mechanisms and drug discovery. Neuropsychopharmacology 35, 345–346 17 Weis, W.I. and Kobilka, B.K. (2008) Structural insights into G-protein coupled receptor activation. Curr. Opin. Struct. Biol. 18, 734–740 18 Tate, C.G. and Schertler, G.F. (2009) Engineering G protein-coupled receptors to facilitate their structure determination. Curr. Opin. Struct. Biol. 19, 386–395 19 Kobilka, B. and Schertler, G.F. (2008) New G-protein-coupled receptor crystal structures: insights and limitations. Trends Pharmacol. Sci. 29, 79–83 20 Rosenbaum, D.M. et al. (2009) The structure and function of G-protein coupled receptors. Nature 459, 356–363 21 Hanson, M.A. et al. (2008) A specific cholesterol binding site is established ˚ structure of the human beta2-adrenergic receptor. Structure 16, by the 2.8 A 897–905 22 Ballesteros, J. and Weinstein, J. (1995) Integrated methods for modeling G-protein coupled receptors. Methods Neurosci. 25, 366–428
956
www.drugdiscoverytoday.com
23 Javitch, J.A. et al. (2002) Use of the substituted cysteine accessibility method to study the structure and function of G-protein coupled receptors. Methods Enzymol. 343, 137–156 24 Ballesteros, J. (1998) Functional microdomains in G-protein-coupled receptors: the conserved arginine-cage motif in the gonadotropin-releasing hormone receptor. J. Biol. Chem. 273, 10445–10453 25 Han, D.S. et al. (2008) Active state-like conformational elements in the beta2-AR and a photoactivated intermediate of rhodopsin identified by dynamic properties of GPCRs. Biochemistry 47, 7317–7321 26 Visiers, I. et al. (2002) Three-dimensional representations of G-protein-coupled receptor structures and mechanisms. Methods Enzymol. 343, 329–371 27 Filizola, M. et al. (2006) Dynamic models of G-protein coupled receptor dimers: indications of asymmetry in the rhodopsin dimer from molecular dynamics simulations in a POPC bilayer. J. Comput. Aid. Mol. Des. 20, 405–416 28 Weinstein, H. (2006) Hallucinogen actions on 5-HT receptors reveal distinct mechanisms of activation and signaling by G protein-coupled receptors. AAPS J. 7, E871–E884 29 Farrens, D.L. et al. (1996) Requirement of rigid-body motion of transmembrane helices for light activation of rhodopsin. Science 274, 768–770 30 Shi, L. et al. (2002) Beta2 adrenergic receptor activation: modulation of the proline kink in transmembrane 6 by a rotamer toggle switch. J. Biol. Chem. 277, 40989–40996 31 Hubbell, W.L. et al. (2003) Rhodopsin structure, dynamics and activation: a perspective from crystallography, site directed spin labeling, sulfhydryl reactivity and disulfide cross-linking. Adv. Protein Chem. 63, 243–290 32 Patel, A.B. et al. (2004) Coupling of retinal isomerization to the activation of rhodopsin. Proc. Natl. Acad. Sci. U. S. A. 101, 10048–10053 33 Patel, A.B. et al. (2005) Changes in interhelical hydrogen bonding upon rhodopsin activation. J. Mol. Biol. 347, 803–812 34 Hornak, V. et al. (2010) Light activation of rhodopsin: insights from molecular dynamics simulations guided by solid-state NMR distance restraints. J. Mol. Biol. 396, 510–527 35 Ahuja, S. et al. (2009) Helix movement is coupled to displacement of the second extracellular loop in rhodopsin activation. Nat. Struct. Mol. Biol. 16, 168–175 36 Ahuja, S. and Smith, S.O. (2009) Multiple switches in G-protein coupled receptor activation. Trends Pharmacol. Sci. 30, 494–502 37 Ludeke, S. et al. (2009) Rhodopsin activation switches in a native membrane environment. Photochem. Photobiol. 85, 437–441 38 Schwartz, T.W. et al. (2006) Molecular mechanism of 7TM receptor activation – a global toggle switch model. Annu. Rev. Pharmacol. Toxicol. 46, 481–519 39 Swaminath, G. et al. (2005) Probing the beta2 adrenoceptor binding site with catechol reveals differences in binding and activation by agonists and partial agonists. J. Biol. Chem. 280, 22165–22171 40 Yao, X. et al. (2006) Coupling ligand structure to specific conformational switches in the B2-adrenoceptor. Nat. Chem. Biol. 2, 417–422 41 Lohse, M.J. et al. (2008) Optical techniques to analyze real-time activation and signaling of G-protein-coupled receptors. Trends Pharmacol. Sci. 29, 159–165 42 Bokoch, M.P. et al. (2010) Ligand-specific regulation of the extracellular surface of a G-protein-coupled receptor. Nature 463, 108–112
43 Kenakin, T. (2003) Ligand-selective receptor conformations revisited: the promise and the problem. Trends Pharmacol. Sci. 24, 346–354 44 Urban, J.D. et al. (2007) Functional selectivity and classical concepts of quantitative pharmacology. J. Pharmacol. Exp. Ther. 320, 1–13 45 DeWire, S.M. et al. (2007) Beta-arrestins and cell signaling. Annu. Rev. Physiol. 69, 483–510 46 Violin, J.D. and Lefkowitz, R. (2007) Beta arresting biased ligands at seven transmembrane receptors. Trends Pharmacol. Sci. 28, 416–422 47 Filizola, M. et al. (1998) BUNDLE: a program for building the transmembrane domains of G-protein-coupled receptors. J. Comput. Aid. Mol. Des. 12, 111–118 48 Lomize, A.L. et al. (1999) Structural organization of G-protein coupled receptors. J. Comput. Aid. Mol. Des. 13, 325–353 49 Shacham, S. et al. (2001) Modeling the 3D structure of GPCRs from sequence. Med. Res. Rev. 21, 472–483 50 Vaidehi, N. et al. (2002) Prediction of structure and function of G protein-coupled receptors. Proc. Natl. Acad. Sci. U. S. A. 99, 12622–12627 51 Fanelli, F. and De Benedetti, P.G. (2006) Inactive and active states and supramolecular organization of GPCRs: insights from computational modeling. J. Comput. Aid. Mol. Des. 20, 449–461 52 Schlyer, S. and Horuk, R. (2006) I want a new drug: G-protein coupled receptors in drug development. Drug Discov. Today 11, 481–493 53 Reggio, P.H. (2006) Computational methods in drug design: modeling G protein-coupled receptor monomers, dimers, and oligomers. AAPS J. 8, E322–E336 54 Houston, D. et al. (2008) Development of selective high affinity antagonists, agonists, and radioligands for the P2Y1 receptor. Comb. Chem. High Throughput Screen. 11, 410–419 55 Vaidehi, N. et al. (2009) Modeling small molecule-compound binding to G-proteincoupled receptors. Methods Enzymol. 460, 263–288 56 Saam, J. et al. (2002) Molecular dynamics investigation of primary photoinduced events in the activation of rhodopsin. Biophys. J. 83, 3097–3112
REVIEWS
57 Crozier, P.S. et al. (2007) How a small change in retinal leads to G-protein activation: initial events suggested by molecular dynamics calculations. Proteins 66, 559–574 58 Isin, B. et al. (2008) Mechanism of signal propagation upon retinal isomerization: insights from molecular dynamics simulations of rhodopsin restrained by normal modes. Biophys. J. 95, 789–803 59 Grossfield, A. et al. (2008) Internal hydration increases during activation of the Gprotein-coupled receptor rhodopsin. J. Mol. Biol. 381, 478–486 60 Khelashvili, G. et al. (2009) Structural and dynamic effects of cholesterol at preferred sites of interaction with rhodopsin identified from microsecond length molecular dynamics simulations. Proteins 76, 403–417 61 Ruprecht, J.J. et al. (2004) Electron crystallography reveals the structure of metarhodopsin I. EMBO J. 23, 3609–3620 62 Schertler, G.F.X. (2005) Structure of rhodopsin and the metarhodopsin I photointermediate. Curr. Opin. Struct. Biol. 15, 408–415 63 Dror, R.O. et al. (2009) Identification of two distinct inactive conformations of the beta2-adrenergic receptor reconciles structural and biochemical observations. Proc. Natl. Acad. Sci. U. S. A. 106, 4689–4694 64 Gouldson, P.R. et al. (2004) Toward the active conformations of rhodopsin and the b2-adrenergic receptor. Proteins Struct. Funct. Bioinform. 56, 67–84 65 Tikhonova, I.G. et al. (2008) Atomistic insights into rhodopsin activation from a dynamic model. J. Am. Chem. Soc. 130, 10141–10149 66 Bhattacharya, S. et al. (2008) Agonist-induced conformational changes in bovine rhodopsin: insight into activation of G-protein-coupled receptors. J. Mol. Biol. 382, 539–555 67 Bhattacharya, S. et al. (2008) Ligand-stabilized conformational states of human beta(2) adrenergic receptor: insight into G-protein-coupled receptor activation. Biophys. J. 94, 2027–2042 68 Bhattacharya, S. and Vaidehi, N. (2010) Ligand-stabilized conformational states of human beta(2) adrenergic receptor: insight into G-protein-coupled receptor activation. J. Am. Chem. Soc. 132, 5205–5214 69 Balaraman, G. and Vaidehi, N. Structural insights into conformational stability of wild type and mutant b1-adrenergic receptor. Biophys. J. (in press)
www.drugdiscoverytoday.com
957
Reviews INFORMATICS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews POST SCREEN
Self-emulsifying drug delivery systems: an approach to enhance oral bioavailability Kanchan Kohli1, Sunny Chopra1, Deepika Dhar2, Saurabh Arora1 and Roop K. Khar1 1 2
Department of Pharmaceutics, Faculty of Pharmacy, Jamia Hamdard, New Delhi 62, India Research and Development Center, Ranbaxy Research Labs, Gurgaon, India
Self-emulsifying drug delivery systems are a vital tool in solving low bioavailability issues of poorly soluble drugs. Hydrophobic drugs can be dissolved in these systems, enabling them to be administered as a unit dosage form for per-oral administration. When such a system is released in the lumen of the gastrointestinal tract, it disperses to form a fine emulsion (micro/nano) with the aid of GI fluid. This leads to in situ solubilization of drug that can subsequently be absorbed by lymphatic pathways, bypassing the hepatic first-pass effect. This article presents an exhaustive account of various literature reports on diverse types of self-emulsifying formulations with emphasis on their formulation, characterization and in vitro analysis, with examples of currently marketed preparations.
Nearly half of the new drug candidates that reach formulation scientists have poor water solubility, and oral delivery of such drugs is frequently associated with low bioavailability [1,2]. To overcome these problems, various formulation strategies have been exploited, such as the use of surfactants, lipids, permeation enhancers, micronization, salt formulation, cyclodextrins, nanoparticles and solid dispersions. The availability of the drug for absorption can be enhanced by presentation of the drug as a solubilizate within a colloidal dispersion [3]. Much attention has focused on lipid solutions, emulsions and emulsion preconcentrates, which can be prepared as physically stable formulations suitable for encapsulation of such poorly soluble drugs. Emulsion systems are associated with their own set of complexities, including stability and manufacturing problems associated with their commercial production. Self-emulsification systems are one formulation technique that can be a fitting answer to such problems [4]. Self-emulsifying drug delivery systems (SEDDS) are isotropic mixtures of drug, lipids and surfactants, usually with one or more hydrophilic cosolvents or coemulsifiers [5]. Upon mild agitation followed by dilution with aqueous media, these systems can form fine (oil in water) emulsion instantaneously. ‘SEDDS’ is a broad term, typically producing emulsions with a droplet size ranging Corresponding author:. Chopra, S. ([email protected])
958
www.drugdiscoverytoday.com
from a few nanometers to several microns. ‘Self-microemulsifying drug delivery systems’ (SMEDDS) indicates the formulations forming transparent microemulsions with oil droplets ranging between 100 and 250 nm. ‘Self-nano-emulsifying drug delivery systems’ is a recent term construing the globule size range less than 100 nm [6].
Suitable drug candidate identification for SEDDS One of the primary challenges to any oral formulation design program is maintaining drug solubility within the gastrointestinal tract and, in particular, maximizing drug solubility within the prime absorptive site of the gut [7]. For lipophilic drug compounds that exhibit dissolution-rate-limited absorption, SEDDS can offer an improvement in rate and extent of absorption, resulting in reproducible blood time profiles. Logically speaking, however, use of SEDDS can be extended to all four categories of biopharmaceutical classification system (BCS) class drugs [6]. These systems can help in solving the under-mentioned problems of all the categories of BCS class drugs, as depicted in Table 1. Lipinski’s rule of five has been widely proposed as a qualitative predictive model for oral absorption trends. In the discovery setting, the ‘rule of five’ predicts that poor absorption or poor permeation is more likely when there are more than five H-bond donors, there are more than ten H-bond acceptors, the molecular weight >500 and the calculated log P > 5.
1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.08.007
TABLE 1
Application of SEDDS in various BCS category drugs BCS class
Problems
Class I
Enzymatic degradation, gut wall efflux
Class II
Solubilization and bioavailability
Class III
Enzymatic degradation, gut wall efflux and bioavailability
Class IV
Solubilization, enzymatic degradation, gut wall efflux and bioavailability
The question arising is whether solubility and log P are sufficient to identify potential drug candidates for such formulations. Although classification systems such as the BCS and Lipinski’s rule of five are useful, particularly at the initial screening stage, they have limitations. It is considered that the rule of five only holds for compounds that are not substrates for active transporters, and with increasing evidence suggesting that most drugs are substrates for some efflux or uptake transporters, this limitation might be notable. Aqueous solubility and/or log P alone are unlikely to be sufficient for identifying the suitability of a lipid-based formulation approach because they do not adequately predict potential in vivo (i.e. physiological) effects. It has been found that individually, these poorly water-soluble compounds, which are generally classified as ‘lipophilic’, behave differently in similar vehicles, thus highlighting the need to assess candidate compounds on an individual basis.
Lipid Formulation Classification System The Lipid Formulation Classification System was introduced as a working model in 2000 [3], and an extra ‘type’ of formulation was added in 2006. The main purpose of the Lipid Formulation Classification System is to enable in vivo studies to be interpreted more readily and, subsequently, to facilitate the identification of the most appropriate formulations for specific drugs (i.e. with reference to their physicochemical properties) [8]. Table 2 indicates the fundamental differences between types I, II, III and IV formulations [9].
Regulatory aspects of lipid excipients Initially excipients were considered inert substances that would be used mainly as diluents, fillers, binders, lubricants, coatings, solvents and dyes in the manufacture of drug products. Over the
REVIEWS
years, however, advances in pharmaceutical science and technology have facilitated the availability of a wide range of novel excipients. It is now recognized that not all excipients are inert substances and some might be potential toxicants [10]. In the United States, the Food and Drug Administration (FDA) has published listings in the Code of Federal Regulations for Generally Recommended as Safe (GRAS) substances that are generally recognized as safe (http://www.fda.gov/Food/FoodIngredients Packaging/GenerallyRecognizedasSafeGRAS/default.htm). Over the years, the Agency has also maintained a list entitled ‘Inactive Ingredient Guide’ for excipients that have been approved and incorporated in marketed products (http://www.accessdata.fda. gov/scripts/cder/iig/index.cfm). This guide is helpful in that it provides the database of allowed excipients with the maximum dosage level by route of administration or dosage form for each excipient. For new drug development purposes, once an inactive ingredient has appeared in an approved drug product for a particular route of administration, the inactive ingredient is not considered new and might require a less extensive review the next time it is included in a new drug product. In this context, the FDA has published a guidance document for industry on the conduct of nonclinical studies for the safety evaluation of new pharmaceutical excipients [11]. Existing human data for some excipients can substitute for certain nonclinical safety data. In addition, an excipient with documented prior human exposure under circumstances relevant to the proposed use might not require evaluation with a full battery of toxicology studies. There is no process or mechanism currently in place within the FDA to independently evaluate the safety of an excipient. For a drug or biological product subject to premarketing approval, their excipients are reviewed and approved as ‘components’ of the drug or biological product in the application. This is particularly true for lipid excipients, in view of their distinct physicochemical properties and the potential complex interactions with other ingredients or the physiological environment that might occur in vivo. In addition, oils can turn into severe cytotoxic agents when reduced to nano range in situ (as in the case of self-nano-emulsifying drug delivery systems), so scientists must be very careful while using oils in such systems. Surfactants can also be a source of severe gastrointestinal tract (GIT) irritation if used in higher concentrations.
TABLE 2
The Lipid Formulation Classification System: characteristic features, pros and cons of the four essential types of ‘lipid’ formulations Formulation Excipients
Properties
Pros
Cons
Type I
Oils without surfactants (e.g. tri-, di- and monoglycerides)
Nondispersing, requires digestion GRAS status; simple; excellent capsule compatibility
Type II
Oils and water-insoluble surfactants SEDDS formed without watersoluble components
Unlikely to lose solvent capacity on dispersion
Type III
Oils, surfactants and cosolvents (both water-insoluble and water-soluble excipients)
SEDDS/SMEDDS formed with water-soluble components
Clear or almost clear dispersion; Possible loss of solvent capacity drug absorption without digestion on dispersion; less easily digested
Type IV
Water-soluble surfactants and cosolvents (no oils)
Formulation disperses typically to form a micellar solution
Formulation has good solvent capacity for many drugs
Formulation has poor solvent capacity unless drug is highly lipophilic Turbid o/w dispersion (particle size 0.25–2 mm)
Likely loss of solvent capacity on dispersion; might not be digestible
www.drugdiscoverytoday.com
959
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Formulation of SEDDS
Reviews POST SCREEN
The SEDDS formulation should instantaneously form a clear dispersion, which should remain stable on dilution. The hydrophobic agent remains solubilized until the time that is relevant for its absorption [12]. Silva et al. [13] found that two main factors, small particle size and polarity of resulting oil droplets, determine the efficient release of the drug compounds from SEDDS. In o/w microemulsions, however, the impact of polarity of oil droplets is not considerable because the drug compound incorporated within the oil droplets reaches the capillaries. Isotropic liquids are preferable to waxy pastes because if one or more excipient(s) crystallize(s) on cooling to form a waxy mixture, it is very difficult to determine the morphology of the materials and, most importantly, the polymorphism properties of the drug within the wax. As a general rule, it is sensible to use the simplest effective formulation, restricting the number of excipients used to a minimum.
Screening of excipients for SEDDS With a large variety of liquid or waxy excipients available, ranging from oils through biological lipids and hydrophobic and hydrophilic surfactants to water-soluble cosolvents, there are many different combinations that could be formulated for encapsulation in hard or soft gelatin or mixtures that disperse to give fine colloidal emulsions. The following points should be considered in the formulation of a SEDDS: (i) the solubility of the drug in different oil, surfactants and cosolvents and (ii) the selection of oil, surfactant and cosolvent based on the solubility of the drug and the preparation of the phase diagram. The backbone of SEDDS formulation comprises lipids, surfactants and cosolvents. The right concentration of the above three decides the self-emulsification and particle size of the oil phase in the emulsion formed. These ingredients are discussed below. Lipids. Lipid is a vital ingredient of the SEDDS formulation. It can not only solubilize large amount of lipophilic drugs or facilitate self-emulsification but also enhance the fraction of lipophilic drug transported via intestinal lymphatic system, thereby increasing its absorption from the GIT [14,15]. Natural edible oils, comprising medium-chain triglycerides, are not frequently preferred in this regard owing to their poor ability to dissolve large amounts of lipophilic drugs [16]. Modified longand medium-chain triglyceride oils, with varying degrees of saturation or hydrolysis, have been used widely for the design of SEDDS [17]. These semisynthetic derivatives form good emulsification systems when used with a large number of solubilityenhancing surfactants approved for oral administration [18,19]. Surfactants. A surfactant is obligatory to provide the essential emulsifying characteristics to SEDDS. Surfactants, being ampiphilic in nature, invariably dissolve (or solubilize) high amounts of hydrophobic drug compounds. The two issues that govern the selection of a surfactant encompass its hydrophilic–lipophilic balance (HLB) and safety. The HLB of a surfactant provides vital information on its potential utility in formulation of SEDDS. For attaining high emulsifying performance, the emulsifier involved in formulation of SEDDS should have high HLB and high hydrophilicity for immediate formation of o/w droplets and rapid spreading of formulation in aqueous media in this context. It would keep drug solubilized for a prolonged period of time at 960
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
the site of absorption for effective absorption, so precipitation of drug compound within GI lumen is prevented [20,21]. A range of industrial nonionic surfactants were screened for their ability to form SEDDS with medium-chain and long-chain triglycerides by Pouton and Porter [8] and Werkley et al. [22], using subjective visual assessment. Cosolvents. Usually, the formulation of an effective SEDDS requires high concentrations of surfactant. Accordingly, cosolvents such as ethanol, propylene glycol and polyethylene glycol are required to enable the dissolution of large quantities of hydrophilic surfactant. The lipid mixture with higher surfactant and cosurfactant:oil ratios leads to the formation of SMEDDS [23,24]. Alcohol and other volatile cosolvents have the disadvantage of evaporating into the shell of soft or hard gelatin capsules, leading to precipitation of drug.
Mechanism of self-emulsification According to Reiss [25], self-emulsification occurs when the entropy change that favors dispersion is greater than the energy required to increase the surface area of the dispersion. The free energy of the conventional emulsion is a direct function of the energy required to create a new surface between the oil and water phases and can be described by the equation: DG ¼ SN i pr i 2S
(1)
where DG is the free energy associated with the process (ignoring the free energy of mixing), N is the number of droplets of radius r and S represents the interfacial energy. The two phases of emulsion tend to separate with time to reduce the interfacial area and, subsequently, the emulsion is stabilized by emulsifying agents, which form a monolayer of emulsion droplets, and hence reduces the interfacial energy, as well as providing a barrier to prevent coalescence [17,26].
In vitro dissolution problems for poorly water-soluble drugs Traditionally, dissolution testing has fulfilled two principal functions. As a mechanism to control quality, dissolution is a sensitive, reproducible and straightforward test that can be used to effectively monitor batch-to-batch variability and ensure bioequivalence once bioavailability has been established [27]. In some circumstances, in vitro dissolution can be used as a surrogate indicator of the likely in vivo dissolution profile and, therefore, as a tool to predict the extent of absorption where dissolution is limiting [28,29]. The principal determinants of the dissolution rate of a poorly water-soluble compound, however, are the degree of wetting and extent of drug solubility in the intestinal contents. The use of simple aqueous media to assess the dissolution profile of poorly water-soluble drugs is often limited by the low intrinsic aqueous solubility of the drug (and, therefore, difficulty in maintaining sink conditions), which – when coupled with analytical sensitivity and logistical issues such as drug binding to filters – can make reproducible dissolution assessment difficult. These problems can be overcome, especially for dissolution tests principally designed to perform a quality control function, by using nonaqueous dissolution media or simple surfactant solutions. Similarities between these media and the likely gastrointestinal environment,
REVIEWS
however, are limited. To improve the accuracy of in vivo dissolution prediction using in vitro dissolution, several studies have developed and defined modified dissolution media that more accurately reflect the solubilization power of the in vivo GI tract. The components of these various dissolution media have been modeled primarily on the likely levels of endogenous bile salts and phospholipids in the fasted and postprandial intestine [30,31]. Although for many poorly water-soluble compounds, solubility in the stomach is not sufficient for appreciable dissolution to occur before gastric emptying, for some compounds (and particularly for weak bases), the stomach is the principal site of dissolution. In these circumstances, simulated gastric fluid, as described in the United States Pharmacopoeia (USP) (2 g NaCl, 3.2 g pepsin, 7 ml HCl, H2O to 1000 ml, pH 1.2), can be used to simulate fasted gastric conditions, and homogenized long-life milk (3.5% fat, pH 6.5) has been suggested to simulate fed gastric conditions [32]. Dressman and colleagues have compared the dissolution profiles of several poorly water-soluble compounds using different dissolution media [33,34]. For nonionizable drugs, good correlations were found between the difference in dissolution profiles in [(Figure_1)TD$IG] simulated fasted-state intestinal media and simulated fed-state
media and the difference in plasma profiles obtained after fasted and postprandial administration. For compounds with appreciable ionization over the likely physiological pH range, the situation is complicated by the impact of both ionization and solubilization on solubility. The increased solubilizing power of the postprandial intestine is, at least in part, sufficient to overcome the low solubility of the unionized species, and the dissolution rate of such drugs under simulated fed-state intestinal conditions is only slightly lower than under fasted-state gastric (i.e. acidic) conditions [33].
Assessment of lipid-based formulations using in vitro lipolysis The design of self-emulsifying lipid-based formulations has focused on optimizing the solubility of the drug in the formulation and on the in vitro emulsification efficiency and particle size of the dispersion obtained on dilution in aqueous media [35]. In recent years, however, in vitro dispersion tests and in vitro lipid digestion models that are more reflective of the gastrointestinal environment have been developed to better predict the in vivo dissolution profile of poorly water-soluble drugs. Pictorial representation of this can be seen in Fig. 1.
pH-meter controller
pH measurement
Experiment initiation/ Add formulation & Pancreatic lipase/ Co-lipase NaOH titration to control pH Computer
Extent of digestion calculated from NaOH addition
Stiner Temperature controlled
Autoburette Drug Discovery Today
FIGURE 1
Lipid digestion models for in vitro assessment of lipidic formulations. (a) The vessel contains digestion buffer, bile salts and phospholipids (to represent a model intestinal fluid), into which lipid-based formulations are introduced and digestion initiated by the addition of pancreatic lipase or colipase. The onset of lipid digestion results in the liberation of fatty acids (FA), which, in turn, causes a transient drop in pH. (b) The drop in pH is quantified by a pH electrode that is coupled to a pH-stat controller and autoburette, which, together, automatically titrate the liberated FA via the addition of an equimolar quantity of sodium hydroxide. This maintains the pH at a set point (thereby enabling the pH-sensitive process of digestion to continue) and facilitates indirect quantification of the extent of digestion (via quantification of the rate of sodium hydroxide addition and assumption of stoichiometric reaction between FA and sodium hydroxide). (c) Throughout the digestion process, samples can be taken and ultra-centrifuged to separate the digest into a poorly dispersed oil phase, a highly dispersed aqueous phase and a precipitated pellet phase. Quantification of the mass of drug that is subsequently trafficked through to the highly dispersed aqueous phase and which does not precipitate provides an indication of the relative proclivity of the formulation with respect to in vivo precipitation and, therefore, a mechanism to (at least) rank order the likely in vivo performance of a series of lipidic formulations [54]. www.drugdiscoverytoday.com
961
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews POST SCREEN
In contrast to traditional dissolution testing, in which the dissolution of drug from the solid state into both simple and biorelevant dissolution media is measured, assessment of the utility of lipid-based formulations is more appropriately based on evaluation of the rate and extent of drug precipitation with respect to time (rather than solubilization) because drug is usually already dissolved in the formulation. In this case, in vitro ‘dispersion’ testing seems to be a more accurate description of the process of monitoring the ability of a formulation to maintain the drug in a solubilized state during dispersion in the stomach and subsequent processing of the formulation in the presence of pancreatic and biliary fluids.
Turbidity measurement
Evaluation of self-emulsifying formulations using in vitro lipolysis
Droplet size
In vitro lipid digestion models have been used to examine the potential in vivo performance of self-emulsifying and self-microemulsifying formulations [36,37]. A reasonable rank-order correlation was described between the patterns of solubilization obtained on in vitro digestion and the plasma profile of Danazol after oral administration of SMEDDS formulations containing long-chain lipid SMEDDS and medium-chain lipid SMEDDS to fasted beagle dogs. Sek et al. [38] examined the impact of a range of surfactants on the in vitro solubilization behavior and oral bioavailability of lipid-based formulations of Atovaquone. They observed no difference in the solubilization behavior of two SEDDS formulations comprising long-chain glycerides, ethanol and either Cremophor EL or Pluronic L121 on in vitro digestion, despite both formulations displaying different self-emulsifying properties. In agreement with the lipolysis data, drug bioavailability after oral administration of the Cremophor EL- and Pluronic L121-containing formulations to beagle dogs was indistinguishable, confirming the importance of consideration of the impact of formulation digestion on solubilization behavior when assessing the likely in vivo performance of lipid-based formulations. It seems, therefore, that reasonable correlation between differences in drug solubilization profiles during in vitro lipolysis and differences in in vivo exposure are evident for typical lipid-based self-emulsifying formulations.
Characterization of SEDDS The primary means of self-emulsification assessment is visual evaluation [39,40]. The various ways to characterize SEDDS are compiled below.
Equilibrium phase diagram Although self-emulsification is a dynamic nonequilibrium process involving interfacial phenomena, information can be obtained about self-emulsification using equilibrium phase behavior. There seems to be a correlation between emulsification efficiency and region of enhanced water solubilization and phase inversion region, formation of lamellar liquid crystalline dispersion phase on further incorporation of water. An equilibrium phase diagram enables comparison of different of different surfactants and their synergy with cosolvent or cosurfactant [41]. The boundaries of one phase region can easily be assessed visually. The phase behavior of a three-component system can be represented by a ternary phase diagram. 962
www.drugdiscoverytoday.com
This identifies efficient self-emulsification by establishing whether the dispersion reaches equilibrium rapidly and in a reproducible time [42]. These measurements are carried out on turbidity meters, most commonly the Hach turbidity meter and the Orbeco-Helle turbidity meter [43,44]. This apparatus is connected to a dissolution apparatus and optical clarity of formulation is taken every 15 s to determine clarity of nano or micro emulsion formed and emulsification time. Turbidity can also be observed in terms of spectroscopic characterization of optical clarity (i.e. absorbance of suitably diluted aqueous dispersion at 400 nm) [45].
This is a crucial factor in self-emulsification performance because it determines the rate and extent of drug release, as well as the stability of the emulsion. Photon correlation spectroscopy, microscopic techniques or a Coulter Nanosizer are mainly used for the determination of the emulsion droplet size [46,47].
Electron microscopic studies Freeze-fracture electron microscopy has been used to study surface characteristics of such dispersed systems [48]. Because of the high lability of the samples and the possibility of artifacts, electron microscopy is considered a somewhat misleading technique. Particle size analysis and low-frequency dielectric spectroscopy have been used to examine the self-emulsifying properties of Imwitor 742 (a mixture of mono- and diglycerides of capric and caprylic aciss) and Tween 80 systems [49].
Zeta potential measurement This is used to identify the charge of the droplets. In conventional SEDDS, the charge on an oil droplet is negative because of the presence of free fatty acids [50].
Determination of emulsification time Pouton et al. [51] quantified the efficiency of emulsification of various compositions of the Tween 85 and medium-chain triglyceride systems using a rotating paddle to promote emulsification in a crude nephelometer. This enabled an estimation of the time taken for emulsification. Once emulsification was complete, samples were taken for particle sizing by photon correlation spectroscopy, and self-emulsified systems were compared with homogenized systems. The process of self-emulsification was observed using light microscopy. It was clear that the mechanism of emulsification involved erosion of a fine cloud of small particles from the surface of large droplets, rather than a progressive reduction in droplet size.
Cryo-TEM studies For Cryo-Transmission Electron Microscopy (TEM), samples were prepared in a controlled environment verification system. A small amount of sample is put on carbon film supported by a copper grid and blotted by filter paper to obtain thin liquid film on the grid. The grid is quenched in liquid ethane at 1808C and transferred to liquid nitrogen at 1968C. The samples were characterized with a TEM microscope.
Liquefaction time This test is designed to estimate the time required by solid SEDDS to melt in vivo in the absence of agitation to simulated GI conditions. One dosage form is covered in a transparent polyethylene film and tied to the bulb of a thermometer by means of a thread. The thermometer with attached tablets is placed in a roundbottom flask containing 250 ml of simulated gastric fluid without pepsin maintained at 37 18C [52]. The time taken for liquefaction is subsequently noted.
Small-angle neutron scattering Small-angle neutron scattering can be used to obtain information on the size and shape of the droplets. The term ‘droplet’ is used to describe micelles, mixed micelles and oil-swollen micelles throughout the present work. Small-angle neutron scattering experiments use the interference effect of wavelets scattered from different materials in a sample (different scattering-length densities).
Small-angle X-ray scattering This a small-angle scattering technique in which the elastic scattering of X-rays by a sample that has inhomogeneities in the nm range is recorded at very low angles (typically 0.1–108). This angular range contains information about the shape and size of macromolecules, characteristic distances of partially ordered materials, pore sizes and other data. Small-angle X-ray scattering is capable of delivering structural information of macromolecules between 5 and 25 nm, of repeat distances in partially ordered systems of up to 150 nm. Small-angle X-ray scattering is used for the determination of the microscale or nanoscale structure of particle systems in terms of such parameters as averaged particle sizes, shapes, distribution and surface-to-volume ratio. The materials can be solid or liquid and they can contain solid, liquid or gaseous domains (so-called ‘particles’) of the same or another material in any combination. In addition to these tools, others – such as nuclear magnetic resonance and differential scanning colorimetry – have also been exploited to characterize these self-emulsifying systems for a better insight.
Biological aspects in selection of SEDDS Very few biopharmaceutical studies have been performed with SEDDS [2], and there is a need for more comparative studies,
REVIEWS
particularly against simple oils and solid dosage forms. At this stage, however, it is worth speculating on the issues that will influence the absorption from SEDDS [53]. The rate of gastric emptying of SEDDS is similar to solutions, so they are particularly useful where rapid onset of action is desirable. Conversely, if the therapeutic index of the drug is low, the rapid onset and accompanying high Thalf might lead to undesirable side-effects. With regard to bioavailability, there are differences between formulations that contain water-soluble surfactants or cosolvents and those that do not. The former systems can produce emulsions or micellar solutions with a lower capacity for solubilization of drugs, which might result in precipitation of drugs in the gut. SEDDS formed with relatively hydrophobic surfactants (HLB < 12), such as Tween 85 or Tagat TO, which do not migrate into the aqueous phase, tend to have lower solvent capacities for drugs unless log P (drug) > 4. These SEDDS should be preferable, however, if the drug can be dissolved to an adequate extent. Highly potent but poorly water-soluble drug candidates are a common outcome of contemporary drug discovery programs and present several challenges to drug development – most notably, the issue of reduced systemic exposure after oral administration [54].
Application SEDDSs present drugs in a small droplet size and well-proportioned distribution and increase the dissolution and permeability. Furthermore, because drugs can be loaded in the inner phase and delivered by lymphatic bypass share, SEDDSs protect drugs against hydrolysis by enzymes in the GI tract and reduce the presystemic clearance in the GI mucosa and hepatic first-pass metabolism.
Solid self-emulsifying drug systems This approach enables the development of tablets using a liquid SEDDS for a poorly water-soluble drug. A high content of liquid SEDDS can be loaded (up to 70%) onto a carrier, which not only maintains good flowability but also enables the production of tablets with good cohesive properties and good content uniformity in both capsules and tablets. This clearly expands the options available to the formulator. In addition to providing the obvious in vivo benefits of a SEDDS system in tablet dosage form (improved drug absorption, and so on), the benefits of developing a solid SEDDS system are that a high content of liquid SEDDS can be loaded onto a carrier and the
TABLE 3
Marketed formulations of SEDDS Active moiety
Trade name
Dosage forms
Tretinoin
Vesanoid (Roche)
Soft gelatin capsule, 10 mg
Isotretinoin
Accutane (Roche)
Soft gelatin capsule, 10, 20 and 40 mg
Cyclosporine
Panimum bioral (Panacea Biotec)
Capsule, 50 and 100 mg
Cyclosporin A
Gengraf (Abbott)
Hard gelatin capsule, 25 and 100 mg
Cyclosporin A
Sandimmune (Novartis)
Soft gelatin capsule, 25, 50 and 100 mg
Lopinavir and Ritonavir
Kaletra (Abbott)
Soft gelatin capsule, Lopinavir 133.33 mg and Ritonavir 33.3 mg
Sanquinavir
Fortovase (Roche)
Soft gelatin capsule, 200 mg
Tipranavir
Aptivus (Boehringer Ingelheim)
Soft gelatin capsule, 250 mg
Amprenavir
Agenerase (GSK)
Soft gelatin capsule
www.drugdiscoverytoday.com
963
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Reviews POST SCREEN
process gives good content (granule) uniformity. In terms of functionality and performance, the solubilizing properties of the final solid dosage form should remain unaffected by both the adsorption of the liquid SEDDS onto a carrier and the state of the drug in the lipid formulation (solubilized versus suspended). The formulation and process is remarkably straightforward and few challenges can be envisaged at the industrial scale. This technique offers formulators an additional option in the quest to achieve product performance, product design and manufacturability. Some examples of bioavailability enhancement achieved with various drugs using SEDDS are Indomethacin [55], Coenzyme Q 10 [56], Ontazolast [19], Simvastatin [57], Celecoxib [45], Carvedilol [40], Paclitaxel [58], Ramipril [59], Ibuprofen, Ketoprofen [60] and PNU-91325 [61]. Table 3 shows some of the marketed formulations of SEDDSs available for oral delivery of various drugs.
Although the potential utility of SEDDS has been known for some time, it is only in recent years that a mechanistic understanding of their impact on drug disposition has emerged. To this end, the use of a combination of in vitro dispersion and digestion methodologies has enabled a much improved understanding of the role of intestinal lipid processing on the solubilization behavior of lipidbased formulations. This in situ emulsion-forming system can be taken as an emulsion premix with high stability as such in the formulation. With future developments in this novel technology, SEDDS will remove deficiencies associated with delivery of poorly soluble drugs. Thus, this field requires further exploration and research to bring out a wide range of commercially available selfemulsifying formulations. To conclude, we can say that this system is not only about lipids and surfactants but also about their selection and the ratio in which they are used.
Concluding remarks
Acknowledgement
For poorly soluble drug candidates, SEDDS provide an effective and practical solution to the problem of formulating drugs where low solubility in the fluids of the GIT limits drug exposure.
The authors would like to acknowledge the assistance provided by the Library of Jamia Hamdard, New Delhi, India, for collection of literature.
References 1 Tang, J.L. et al. (2007) Self emulsifying drug delivery systems: strategy for improving oral delivery of poorly soluble drugs. Curr. Drug Ther. 2, 85–93 2 Humberstone, A.J. and Charman, W.N. (1997) Lipid based vehicles for oral delivery of poorly water soluble drugs. Adv. Drug Deliv. Rev. 25, 103–128 3 Pouton, C.W. (2000) Lipid formulations for oral administration of drugs: nonemulsifying, self-emulsifying and ‘self-microemulsifying’ drug delivery systems. Eur. J. Pharm. Sci. 11 (Suppl. 2), S93–S98 4 Venkatesh, G. et al. (2010) In vitro and in vivo evaluation of self-microemulsifying drug delivery system of buparvaquone. Drug Dev. Ind. Pharm. 36, 735–745 5 Gursoy, R.N. and Benita, S. (2004) Self-emulsifying drug delivery systems (SEDDS) for improved oral delivery of lipophilic drugs. Biomed. Pharmacother. 58, 173–182 6 Singh, B. et al. (2009) Self-emulsifying drug delivery systems (SEDDS): formulation development, characterization, and applications. Crit. Rev. Ther. Drug Carrier Syst. 26, 427–521 7 O’Driscoll, C.M. and Griffin, B.T. (2008) Biopharmaceutical challenges associated with drugs with low aqueous solubility – the potential impact of lipid-based formulations. Adv. Drug Deliv. Rev. 60, 617–624 8 Pouton, C.W. and Porter, C.J. (2008) Formulation of lipid-based delivery systems for oral administration: materials, methods and strategies. Adv. Drug Deliv. Rev. 60, 625–637 9 Pouton, C.W. (2006) Formulation of poorly water-soluble drugs for oral administration: physicochemical and physiological issues and the lipid formulation classification system. Eur. J. Pharm. Sci. 29, 278–287 10 Chen, M.L. (2008) Lipid excipients and delivery systems for pharmaceutical development: a regulatory perspective. Adv. Drug Deliv. Rev. 60, 768–777 11 Swenson, E.S. et al. (1994) Intestinal permeability enhancement: efficacy, acute local toxicity and reversibility. Pharm. Res. 11, 1132–1142 12 Patel, D. and Sawant, K.K. (2009) Self micro-emulsifying drug delivery system: formulation development and biopharmaceutical evaluation of lipophilic drugs. Curr. Drug Deliv. 6, 419–424 13 Silva, B.F. et al. (2009) Size, shape, and charge of salt-free catanionic microemulsion droplets: a small-angle neutron scattering and modeling study. J. Phys. Chem. B 113, 10230–10239 14 Gershanik, T. and Benita, S. (2000) Self-dispersing lipid formulations for improving oral absorption of lipophillic drugs. Eur. J. Pharm. Biopharm. 50, 179–188 15 Devani, M. et al. (2004) The emulsification and solubilisation properties of polyglycolysed oils in self-emulsifying formulations. J. Pharm. Pharmacol. 56, 307–316 16 Odeberg, J.M. et al. (2003) Lipid drug delivery and rational formulation design for lipophilic drugs with low oral bioavailability, applied to cyclosporine. Eur. J. Pharm. Sci. 20, 375–382 17 Constantinides, P.P. (1995) Lipid microemulsions for improving drug dissolution and oral absorption: physical and biopharmaceutical aspects. Pharm. Res. 12, 1561–1572
964
www.drugdiscoverytoday.com
18 Kimura, M. et al. (1994) Relationship between the molecular structures and emulsification properties of edible oils. Biosci. Biotechnol. Biochem. 58, 1258–1261 19 Hauss, D.J. et al. (1998) Lipid-based delivery systems for improving the bioavailability and lymphatic transport of a poorly water-soluble LTB4 inhibitor. J. Pharm. Sci. 87, 164–169 20 Shah, N.H. et al. (1994) Self emulsifying drug delivery systems (SEDDS) with polyglycolyzed glycerides for improving in vitro dissolution and oral absorption of lipophilic drugs. Int. J. Pharm. 106, 15–23 21 Attwood, D. and Florence, A.T. (1983) Surfactant Systems: Their Chemistry, Pharmacy and Biology. Chapman and Hall 22 Werkley, M.G. et al. (1986) Self emulsification of vegetable oil-non-ionic surfactant mixtures. ACS Symp. Ser. 311, 242–255 23 Meinzer, A. et al. (1995) Microemulsion: a suitable galenical approach for the absorption enhancement of low soluble compounds? BT Gattefosse 88, 21–26 24 Vonderscher, J. and Meinzer, A. (1994) Rationale for the development of Sandimmune Neoral. Transplantation Proceedings (Vol. 26) pp. 2925–2927 25 Reiss, H. (1975) Entropy induced dispersion of bulk liquids. J. Colloid Interface Sci. 53, 61–70 26 Craig, D.Q. et al. (1995) An investigation into the mechanism of self emulsification using particle size analysis and low frequency dielectric spectroscopy. Int. J. Pharm. 114, 103–110 27 Porter, C.J. and Charman, W.N. (2001) In vitro assessment of oral lipid based formulations. Adv. Drug Deliv. Rev. 50 (Suppl. 1), S127–S147 28 Dressman, J.B. et al. (1998) Dissolution testing as a prognostic tool for oral drug absorption: immediate release dosage forms. Pharm. Res. 15, 11–22 29 Horter, D. and Dressman, J.B. (2001) Influence of physicochemical properties on dissolution of drugs in the gastrointestinal tract. Adv. Drug Deliv. Rev. 46, 75–87 30 Hernell, O. et al. (1990) Physical–chemical behavior of dietary and biliary lipids during intestinal digestion and absorption, 2. Phase analysis and aggregation states of luminal lipids during duodenal fat digestion in healthy adult human beings. Biochemistry 29, 2041–2056 31 Fausa, O. (1974) Duodenal bile acids after a test meal. Scand. J. Gastroenterol. 9, 567– 570 32 Galia, E. et al. (1998) Evaluation of various dissolution media for predicting in vivo performance of class I and II drugs. Pharm. Res. 15, 698–705 33 Dressman, J.B. and Reppas, C. (2000) In vitro–in vivo correlations for lipophilic, poorly water-soluble drugs. Eur. J. Pharm. Sci. 11 (Suppl. 2), S73–S80 34 Nicolaides, E. et al. (2001) Biorelevant dissolution testing to predict the plasma profile of lipophilic drugs after oral administration. Pharm. Res. 18, 380–388 35 Porter, C.J. et al. (2008) Enhancing intestinal drug solubilisation using lipid-based delivery systems. Adv. Drug Deliv. Rev. 60, 673–691
36 Fatouros, D.G. et al. (2007) Structural development of self nano emulsifying drug delivery systems (SNEDDS) during in vitro lipid digestion monitored by small-angle X-ray scattering. Pharm. Res. 24, 1844–1853 37 Fatouros, D.G. et al. (2007) Morphological observations on a lipid-based drug delivery system during in vitro digestion. Eur. J. Pharm. Sci. 31, 85–94 38 Sek, L. et al. (2010) Examination of the impact of a range of Pluronic surfactants on the in-vitro solubilisation behaviour and oral bioavailability of lipidic formulations of atovaquone. J. Pharm. Pharmacol. 58, 809–820 39 Cui, S.X. et al. (2009) Preparation and evaluation of self-microemulsifying drug delivery system containing vinpocetine. Drug Dev. Ind. Pharm. 35, 603–611 40 Wei, L. et al. (2005) Preparation and evaluation of SEDDS and SMEDDS containing carvedilol. Drug Dev. Ind. Pharm. 31, 785–794 41 Pouton, C.W. et al. (1987) Self-emulsifying systems for oral delivery of drugs. International Symposium on Control Release Bioactive Materials pp. 113–114 42 Gursoy, N. et al. (2003) Excipient effects on in vitro cytotoxicity of a novel paclitaxel self-emulsifying drug delivery system. J. Pharm. Sci. 92, 2411–2418 43 Nazzal, S. et al. (2002) Preparation and in vitro characterization of a eutectic based semisolid self-nanoemulsified drug delivery system (SNEDDS) of ubiquinone: mechanism and progress of emulsion formation. Int. J. Pharm. 235, 247–265 44 Palamakula, A. and Khan, M.A. (2004) Evaluation of cytotoxicity of oils used in coenzyme Q10 self-emulsifying drug delivery systems (SEDDS). Int. J. Pharm. 273, 63–73 45 Subramanian, N. et al. (2004) Formulation design of self-microemulsifying drug delivery systems for improved oral bioavailability of celecoxib. Biol. Pharm. Bull. 27, 1993–1999 46 Goddeeris, C. et al. (2006) Light scattering measurements on microemulsions: estimation of droplet sizes. Int. J. Pharm. 312, 187–195 47 Yang, S. et al. (2004) Enhanced oral absorption of paclitaxel in a novel self microemulsifying drug delivery system with or without concomitant use of P-glycoprotein inhibitors. Pharm. Res. 21, 261–270 48 Vyas, S.P. and Khar, R.K. (2002) Submicron emulsion. In Targeted and Controlled Drug Delivery Novel Carriers Systems. CBS Publishers and Distributors pp. 291–294
REVIEWS
49 Craig, D.Q. et al. (1993) An investigation into the physicochemical properties of self emulsifying systems using low frequency dielectric spectroscopy, surface tension measurements and particle size analysis. Int. J. Pharm. 96, 147–155 50 Gershanik, T. and Benita, S. (1996) Positively charged self-emulsifying oil formulation for improving the oral bioavailability of progestrone. Pharm. Dev. Technol. 1, 147–157 51 Pouton, C.W. (1997) Formulation of self emulsifying drug delivery systems. Adv. Drug Deliv. Rev. 25, 47–58 52 Attama, A.A. et al. (2003) The use of solid self-emulsifying systems in the delivery of diclofenac. Int. J. Pharm. 262, 23–28 53 Patel, D. and Sawant, K.K. (2007) Oral bioavailability enhancement of acyclovir by self-microemulsifying drug delivery systems (SMEDDS). Drug Dev. Ind. Pharm. 33, 1318–1326 54 Porter, C.J. et al. (2007) Lipids and lipid-based formulations: optimizing the oral delivery of lipophilic drugs. Nat. Rev. Drug Discov. 6, 231–248 55 Kim, J.Y. and Ku, Y.S. (2000) Enhanced absorption of indomethacin after oral or rectal administration of a self-emulsifying system containing indomethacin to rats. Int. J. Pharm. 194, 81–89 56 Kommuru, T.R. et al. (2001) Self-emulsifying drug delivery systems (SEDDS) of coenzyme Q10: formulation development and bioavailability assessment. Int. J. Pharm. 212, 233–246 57 Kang, B.K. et al. (2004) Development of self-microemulsifying drug delivery systems (SMEDDS) for oral bioavailability enhancement of simvastatin in beagle dogs. Int. J. Pharm. 274, 65–73 58 Gao, P. et al. (2003) Development of a supersaturable SEDDS (S-SEDDS) formulation of paclitaxel with improved oral bioavailability. J. Pharm. Sci. 92, 2386–2398 59 Shafiq, S. et al. (2007) Development and bioavailability assessment of ramipril nanoemulsion formulation. Eur. J. Pharm. Biopharm. 66, 227–243 60 Araya, H. et al. (2005) The novel formulation design of self-emulsifying drug delivery systems (SEDDS) type O/W microemulsion II: stable gastrointestinal absorption of a poorly water soluble new compound, ER-1258 in bile-fistula rats. Drug Metab. Pharmacokinet. 20, 257–267 61 Gao, P. et al. (2004) Enhanced oral bioavailability of a poorly water soluble drug PNU-91325 by supersaturatable formulations. Drug Dev. Ind. Pharm. 30, 221–229
www.drugdiscoverytoday.com
965
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews POST SCREEN
Unearthing a source of medicinal molecules Edwin L. Cooper1 and M. Balamurugan2 1 Laboratory of Comparative Neuroimmunology, Department of Neurobiology, David Geffen School of Medicine at UCLA, University of California, Los Angeles, CA 90095-1763, USA 2 Division of Vermibiotechnology, Department of Zoology, Annamalai University, Annamalai Nagar 608002, India
Bioprospecting introduced enormous promise of additional sources of beneficial products that now confirm the ancient practices associated with Ayurveda (India), TCM (China) and Kampo (Japan). With the surge of immunology in the early 1960s, there was a need for a more encompassing view revealing the pervasiveness of the immune system demonstrated as rejection of transplants in earthworms. Analogous with the importance of natural products derived from plants, it was clear that similar useful natural substances might be found in worms. Fortunately, marine animals emerged and so did the earthworm, the humoral immune systems of which provided components useful in current approaches to complementary and alternative therapies. This introduced the enormous promise of additional sources of beneficial products now confirming ancient practices: Ayurveda (India), traditional Chinese medicine (China) and Kampo (Japan). This review provides basic knowledge on drug discovery from the earthworm. My aim as a biologist was to challenge the anthropocentric view that immunity was purely mammalian there for human [1–6]. What emerged and is widely accepted is the universality of innate immunity of the cellular and humoral type—essential throughout the living world and necessary for the full initiation and completion of the adaptive immune response that vertebrates, including humans, inherited. The importance of natural products derived from plants and animals when analyzed in appropriate assays could be shown to be relevant in our quest for defining useful treatments applicable to human maladies. Sources such as marine species emerged – and so did the earthworm – as animals whose humoral immune systems provided the fodder for components in earlier and even current approaches to complementary and alternative therapies. Thus, bioprospecting came on the scene.
TCM: chemical ingredients for Di Long, the common name for the earthworm Lumbricus The earthworm, an ancient Ayurvedic source
Corresponding author. Cooper, E.L. ([email protected])
From another ancient culture, in India, within the Ayurvedic tradition, there are similar practices to those in China (Dilong is
966
www.drugdiscoverytoday.com
In Chinese medical journals, Di Long is a medicinal preparation based on extracts of the earthworm species Lumbricus rubellus used in traditional Chinese medicine (TCM) for a wide variety of disorders, from convulsions and fevers to rheumatoid arthritis and blood stasis syndromes [7–15]. Earthworms are also used in treating blepharoptosis, or drooping of the upper eye lid, along with other phlegm herbs (e.g. Dan Nan Xing, Jiang Can, Ban Xia, Tian Ma and Bai Fu Zi) (Figure 1). Di Long comes in two variants, Guang Di Long (native to Guangdong, Guangxi and Fujian and collected from spring to autumn) and Tu Di Long (collected during the summer in many regions of China). For initial assays, the abdomen of an earthworm is cut open immediately after capture; then, viscera and other contents are removed. The abdomen is washed clean and dried in the sun or indoors at low temperatures. According to TCM, Di Long is associated with the bladder, liver, lung and spleen meridians and has salty and cold properties. Di Long is thought to work by draining liver heat, by clearing lung heat and by clearing heat in the collateral channels. Earthworm’s ‘channelopening’ properties are thought to derive from its habits of burrowing through the earth, constantly searching out new spaces in which to slither. In nutritive components, Di Long can contain lumbrofebrine, lumbritin, terrestro-lumbrolysin, hypoxanthine, xanthine, adenine, guanine, choline, guanidine, ornithine, lysine, serine, proline, glycine, cystine, valine, phenylalanine, tryptophan, neutral lipids, cholesterol, free fatty acids, triglycerides, complex lipids, phosphatidylcholine, phosphatidyl ethanolamine, phosphatidylserine, dehydrogenase isoenzyme and esterase isoenzyme. Earthworms clearly have offered clues to their value as sources of a healing extract.
1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.09.004
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[()TD$FIG]
Reduce or Prevent Oxidative stress
Reactive oxygen groups ↑
NAPQ I
Induce genes SOD, CAT and GPx ↑
Reduce or Prevent Oxidative stress GSH depletion ↑
Inflammation
Damaged liver cells Inhibition Point
ALP, AST and ALP ↑
Bilirubin (antioxidant) ↑
Drug Discovery Today
FIGURE 1
How does earthworm extract ameliorate liver damage?
a transliteration of the Chinese term for the animal itself). According to Cooper et al. [16], indigenous people throughout the world, more particularly in Asia (including India, Myanmar, China, Korea and Vietnam) have traditionally practiced extracting and using biologically active compounds from earthworms [17]. Table 1 shows the activities of some of these compounds with fibrinolytic activity. Table 2 shows a more varied functional diversity, including – for example – mitogenicity, hemolytic hemagglutination, perforin-like activity, hemolysis, cytotoxicity, vasodepressor and antibacterial activity. Earthworms have a dense nutritional content because of their soil-based origin. Previous studies, using earthworms in the context of vermibiotechnology, have shown their antipyretic, antispasmodic, detoxic, diuretic, antihypertensive, antiallergic, antiasthmatic, spermatocidal, antioxidative, antimicrobial, anticancer, antiulceral and anti-inflammatory activities [4,5,18–21]. This, of course, reminds us of the earthworm’s innate immune properties. In the first of several experiments, in rats whose livers have been
damaged by paracetamol, earthworm extracts (EEs) were involved in a hepatoprotective role, whether by preventing damage owing to oxidative stress, by enhancing antioxidant activity, by reducing lipid peroxidation or by all these mechanisms.
Antiulcer and antioxidative therapeutic properties Lampito mauritii (Kinberg), an indigenous earthworm species widely used in Siddha and Ayurveda, possesses anti-inflammatory, antiulceral and antioxidative properties. To test and confirm this assertion, Prakash and Ranganathan [20] have analyzed the antiulceral and antioxidant properties of ‘earthworm paste’ (EPA) derived from Lampito mauritii. EPA effects were compared with a standard antiulceral drug, ranitidine, on the Wistar strain albino. Controls, administration of 200 mg/kg aspirin, increased the volume of gastric juice secretion, total acidity, free acidity and ulcer index and reduced the pH. Experimental animals also had decreased antioxidant levels (i.e. reduced glutathione, glutathione peroxidase, catalase and superoxide dismutase) but increased levels of thiobarbituric-acid-reactive substances. Pre-treatment with the standard drug, ranitidine (50 mg/kg) and different doses of EPA (20, 40, 80, 160 and 320 mg/kg) in rats with induced ulcers enhanced the pH, decreased the volume of gastric juice, free acidity and total acidity and reduced the ulcer index. Moreover, activities of reduced glutathione, glutathione peroxidase, catalase and superoxide dismutase were increased, whereas the thiobarbituric-acid-reactive substance decreased. Results were more considerable in rats administered with 160 mg/kg EPA than by the application of ranitidine and other doses of EPA. These results suggest that EPA possesses antiulcer and antioxidative therapeutic properties. Anti-inflammatory activity and antioxidant property associated with high polyphenolic content. After these encouraging results, a newer challenge began an analysis aiming to understand more clearly those therapeutic properties, such as anti-inflammatory, antioxidative, hematological and serum biochemical markers, associated with EPA [20]. Initially, investigators compared EPA activity with the standard anti-inflammatory drug, aspirin, on rats. Administration of EPA (80 mg/kg) to albino rats that had been previously induced to present an inflammatory response showed reduced inflammation and restored levels of antioxidants but reduced
TABLE 1
Therapeutic properties of earthworms (fibrinolytic) Earthworm
Active fraction
Activity
Refs
Eisenia fetida
G-90
[44]
Pheretima Lumbricus rubellus E. fetida E. fetida
Fibrinolytic enzyme Freeze-dried powder Glycosylated fibrinolytic enzyme Fibrinolytic enzymes extracted from earthworms reared in different substrates Fibrinolytic enzymes Fibrinolytic protease (F-III-2) Serine protease Fibrinolytic enzyme (eFE-D) SDS-activated fibrinolytic enzyme Earthworm protease-II and III
Fibrinolytic activity against malignant tumor patients’ blood Fibrinolytic Antithrombotic and fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic Fibrinolytic
[48] [49] [50] [51] [52] [53]
L. rubellus L. rubellus L. rubellus E. fetida E. fetida E. fetida
www.drugdiscoverytoday.com
[45] [31] [46] [47]
967
Reviews POST SCREEN
Paracetamol (overdose)
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
TABLE 2
Therapeutic properties of earthworms (non-fibrinolytic)
Reviews POST SCREEN
Earthworm
Active fraction
Activity
Refs
Eisenia fetida E. fetida E. fetida and Lumbricus terrestris E. fetida E. fetida E. fetida E. fetida
Coelomic Cytolytic Factor-1 (CCF-1) G-90 Coelomic fluid Small coelomocytes Lysenin Eiseniapore Fetidins
[54] [55] [56] [57] [58] [59]
E. fetida, L. terrestris and Aporrectodea caliginosa Eisenia fetida andrei E. fetida andrei
Hemolytic and hemagglutinating proteins in the coelomic fluid Eisenia fetida andrei factor Hemolysins
Cytolytic Mitogenicity Hemolytic and hemagglutinating Perforin-like activity Hemolysis, cytotoxicity and vasodepressor Cytolytic Antibacterial Hemolytic and hemagglutinating Hemolytic and hemagglutinating Hemolytic activity Hemolytic
[62] [63]
glutathione, glutathione peroxidase, superoxide dismutase, catalase and thiobarbituric-acid-reactive substances. Moreover, at the cellular level, treated rats showed restoration to normal values of erythrocytes and leukocytes. More specifically, differential levels of neutrophils, lymphocytes and eosinophils were restored, as were hemoglobin and serum biochemical components. These included protein, albumin, glucose, cholesterol, and acid and alkaline phosphatase and electrolytes (e.g. sodium, potassium and chloride). According to one interpretation, the anti-inflammatory activity together with antioxidant property of EPA seems to be due to the high polyphenolic content of earthworm tissues; polyphenolic compounds have been extensively studied and have shown great potential as antioxidants. Because they are present in the EE, the antioxidant properties of EEs might be a result of the component polyphenols. Restoration of liver histoarchitecture. Developing on previous studies, Balamurugan [21] demonstrated that EE was capable of concomitantly restoring histoarchitecture in paracetamolinduced liver damage in rats. Whole-tissue extract provided dose-dependent (over a range of 100–300 mg/kg body weight) liver protection to rats given paracetamol at 2 g/kg to induce liver damage. The treatment produced a statistically significant reduction (P < 0.05) for the hepatic marker enzymes aspartate transaminase, alanine transaminase and alkaline phosphatase that was similar to silymarin administered at 150 mg/kg. Histopathological observations of liver tissues confirmed such biochemical evidence.
Anti-inflammatory and antipyretic activities of EE EE stimulated properties similar to glycolipoprotein complex (G90). Administration of indomethacin (10 mg/kg), paracetamol (150 mg/kg) and/or various doses of EE (50, 100 and 200 mg/kg) were all capable of reversing histamine-induced and turpentineinduced inflammation (in both acute and chronic phase) and Brewer’s yeast-induced pyrexia in rats, effectively back to normal [20].
Modes of action Hepatoprotective and antioxidant properties Balamurugan et al. [22] have suggested that the hepatoprotective properties observed in paracetamol-induced liver injury in rats might have a similar mode of action to silymarin, a standard 968
www.drugdiscoverytoday.com
[60] [61]
hepatoprotective drug. Consequences of liver injury include eduction in liver antioxidants (such as glutathione, superoxide dismutase, glutathione peroxidase and catalase), as well as reduced serum total protein, and increased serum components such as alkaline phosphatase, aspartate aminotranferase (AST), alanine aminotranferase (ALT), bilirubin and liver thiobarbituric-acidreactive substances. Administration of EE (100, 200 and 300 mg/ kg) increased activities of liver glutathione, superoxide dismutase, glutathione peroxidase and catalase and elevated serum total protein levels but resulted in decreased levels of serum alkaline phosphatase, AST, ALT, bilirubin and liver thiobarbituric-acidreactive substances. The changes in these liver markers were similar to those produced by the hepatoprotective drug silymarin at a dose of 150 mg/kg. The mode of action of EE suggests that it might prevent the formation of reactive oxygen species, or that EE might assist in reactive oxygen species scavenging. As a result, EE prevents damage to hepatic cells, modulates the genes responsible for synthesis of antioxidant enzymes (glutathione peroxidase, catalase and superoxide dismutase in liver tissue), and decreases serum enzymatic activities such as alkaline phosphatase, AST and ALT.
Inter-species properties By way of confirmation, extension and validation of these observations, Prakesh used another earthworm species (Perionyx excavatus) to examine certain pharmaceutical effects [23]. Analyses focused on the hepatoprotective and antioxidant properties of earthworm powder (EPO), not EE or EPA, in a model of alcoholinduced hepatotoxicity. Rats with alcohol-induced hepatotoxicity showed elevation in the lipid-peroxidative marker, thiobarbituricacid-reactive substance. There was a decrease in the activities of the antioxidant enzymes superoxide dismutase, catalase and glutathione peroxidase and in the non-enzymatic antioxidants vitamin C, vitamin E and reduced glutathione. Oral administration of dried EPO (500 mg/kg) for 42 days reversed these parameters towards normal levels. Results similar to those of others suggest that this indigenous earthworm could afford a considerable hepatoprotective and antioxidant effect in alcohol-induced hepatotoxicity in rats. What is even more crucial is that EPO seemed to be as effective as EE in altering the pathology induced by alcohol, thus potentially confirming a universality of effects caused by earthworm products.
BOX 1
Functional properties of lumbrokinasea Leading researchers and doctors report on the power of lumbrokinase to: Dissolve clots and protect against ischemic heart disease and stroke Lower fribrinogen levels in cancer patients, which is strongly associated in scientific studies with better outcomes, less metastasis, and slower growth of tumors Dissolve bacterial biofilms present in chronic infections in conditions like autism and Lyme disease, allowing antimicrobials to work effectively Offer antiplatelet, anti-thrombotic and anti-apoptotic activity, remarkably regulating hypercoagulation a
Adapted from a table from Ref. [7].
Lumbrokinase from Lumbricus rubellus Basic characteristics related to clinical trials The earthworm has been used as a drug for various diseases in China and the Far East for thousands of years, although without rigorous scientific evaluation [6]. According to records in the most famous Chinese ancient medical publication, Ben Cao Gang Mu (Compendium of Materia Medica), the traditional medico material ‘Di Long’ (earthworm) was regarded as being effective in treating limb numbness and hemiplegia. Moreover, Di Long was also effective as an antipyretic, capable of sedation and able to improve blood circulation and ameliorate clot formation. According to Cooper [7], lumbrokinase (LK) is a member of a group of proteolytic enzymes, which include plasminogen activator and plasmin that has been isolated from a particular earthworm (Box 1). The plasminogen activator (e-PA) in LK is similar to the plasminogen activator (t-PA) from other species. This enzyme can only demonstrate thrombolytic activity in the presence of fibrin; therefore, LK has the advantage of not causing hemorrhage owing to hyperfibrinolysis during treatment, in contrast with either streptokinase or urokinase. According to acute and subacute toxicological experiments, there were no reported negative effects of LK on the nervous, cardiovascular, respiratory or blood systems of rats, rabbits or dogs. In addition, long-term damage to hepatic or renal systems could not be demonstrated in animal toxicological experiments. Embryonic development, similarly, was not affected by such treatment; there have been no reports of either teratogenesis or mutagenesis in embryonic rats.
Treatment of embolism In experimental acute pulmonary artery embolism in rabbits, the embolus was labeled with 125I and the radioactivity in blood tested at 0.5, 1.0, 2.0, 3.0 and 5.0 h after duodenal administration of LK. There was a marked dose-dependent increase in radioactivity in the blood at 3.0 and 5.0 h after administration. Inferior vena cava thrombosis tests in rats showed a reduction in thrombosis after rectal administration of LK. LK decreases fibrinogen, thus lowering blood viscosity and reducing platelet aggregation. LK has also been reported to be effective in treating and preventing ischemic cerebrovascular diseases and other embolic and thrombotic diseases: coronary, myocardial infarction, deafness, arterial sudden thrombosis of the central retinal vein, embolism in peripheral veins and pulmonary infarction [24–32].
REVIEWS
Boluoke: fibrin-dissolving enzyme from earthworms that might be applicable to disseminated intravenous coagulation Boluoke (LK) is the only fully researched oral enzyme that supports healthier blood. It shows great promise in supporting a healthy balance of coagulation and fibrinolysis in the body [7,33,34] (http://www.naturodoc.com/library/heart/boluoke.htm). The coagulation system is a complicated and highly regulated system. Hypercoagulation means the body is producing fibrin strands faster than it can break them down. When this happens, the fibrin can become deposited on capillary walls, impeding the delivery of oxygen and nutrients from the blood into tissues and waste chemicals from them out to the bloodstream. If this situation is sustained for a prolonged period, the body’s tissues will gradually become hypoxic (lacking in oxygen) and malnourished. The end result is often an acidic tissue environment, pain, lack of energy and the decline or loss of organ functions. Hypercoagulation also predisposes a person to clot formation (as in strokes or heart attacks) in blood vessels if atherosclerosis is already present. Research has clearly shown that hypercoagulation is often present in patients with chronic illnesses. If hypercoagulation is not diagnosed, patients often do not improve, or they relapse easily. In meningitis, there is often associated purpura fulminans. LK could be a candidate for amelioration of this condition.
Applications of earthworm lysins to membranes Relevance to human diseases Bioprospecting upon careful analyses reveals useful products synthesized and secreted even by sponges—the first multicellular members of the second animal phylum, the Porifera [4,5]. If we turn now to more complex multicellular species for which there is substantial centuries-old information, witness the literature pertaining to the earthworm’s healing properties [6]. Earthworm lytic molecules are antimicrobial and fibrinolytic and might prove useful as antibacterial agents and prophylactic molecules [24–32]. Two molecules, lysenin and eiseniapore, can target intracellular lipid-trafficking mechanisms. Trafficking dysfunction can directly produce pathology – for example, in Tangier disease and Niemann-Pick disease type C – or might be able to contribute to pathology in diseases such as Alzheimer’s and atherosclerosis. Lysenin reacts specifically with fibroblast membranes from patients with Niemann-Pick disease, a rather curious finding, but one that might have some clinical relevance [35,36] (Table 3). Niemann-Pick diseases are genetic diseases grouped under sphingolipidoses, or lipid storage disorders. For them, excessive lipids accumulate in the spleen, liver, lungs, bone marrow and brain. There is a classic infantile type A variant that causes complete deficiency of sphingomyelinase. Sphingomyelin is a component of cell membrane, including the organellar membranes; thus, the enzyme deficiency blocks degradation of lipid. This results in the accumulation of sphingomyelin within lysosomes in the macrophage-monocyte phagocyte lineage (the body’s phagocytic cells). When cells are affected, they often become enlarged, approximately 90 microns in diameter. Microscopic analysis reveals lipid-laden macrophages in the bone marrow, whereas pathologic examination shows ‘sea-blue histiocytes’. Numerous small vacuoles of uniform size are created. www.drugdiscoverytoday.com
969
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
TABLE 3
Peptides derived from earthworm that mediate cytolysisa,b
Reviews POST SCREEN
Nomenclature: molecular weight and investigators
Accession number
Homology
Site of expression
Proposed function
Lysenin EfL1 (L1) EfL2 (L2) EfL3 (L3) 42-kDa Sekizawa et al., 1997 Gene Fetidin [64]
D85846 D85847 D85848
U02710 (Fetidin 1)
Coelomic fluid; chlorogocytes; coelomocytes
Contracts rat smooth muscle
Fetidin 1 40-kDa Four isoforms Fetidin 2 45-kDa Monomorphic Lassegues et al., 1997 [65] Eu J Biochem Coelomic cytolytic factor 1
U02710
D85848 of (Lysenin)
Coelomic fluid; chlorogocytes; inducible; increases after injecting pathogenic bacteria
Hemolysis; bacteriolysis; agglutination; clotting; opsonization; heme-binding enzymes; horseradish peroxidase; protein U1 32; herpes simplex virus
CCF-1 42-kDa Beschin et al., 1998 [66] J Biol Chem Lumbricin
AF030028
AF395805 of (CCF precursor)
Coelomic fluid b-1, 3-glucan LPS
Cytolysis; opsonization; hemolysis
Lumbricin I 7.2-kDa Cho et al., 1998 [67] Biochem Biophys Acta
AF060552
Whole worm; not inducible; constitutive
Antimicrobial; not hemolytic; not induced by infection
a b
All derived from Eisenia foetida, except antimicrobial Lumbricin I from Lumbricus rubellus. Modified from Bioessay 24.4 Cooper et al. 2002 [52].
This gives a foamy appearance to the cytoplasm. The relevance and curiosity derives from the finding that fibroblast membranes can be recognized specifically by lysenin derived from earthworms (Table 3).
Specific binding of lysenin and sphingomyelin Specific binding of lysenin to sphingomyelin on cellular membranes might prove to be a useful tool to probe the molecular motion and function of sphingomyelin in biological membranes, especially in an effort to explain the mechanism of lysis caused by earthworm products. These results emphasize simultaneously the need for expanded analyses of various lytic pathways that might be mediated by and within the context of the earthworm’s immunodefense system. The products of sponges and those of other animals have potential importance that is at least considerable as that of the products being exploited from plants. One future goal will be to successfully introduce some of these compounds as treatments for human diseases. Such an approach would raise public awareness concerning the richness and diversity of natural products that could be carefully harvested for human benefit without damaging ecosystems. Serendipity has won again: how scientific approaches reveal solid evidence that certain molecules might now be ready for testing.
fractions: the ethanolic precipitate fraction, the alkaline fraction and the acidic fraction. Of the three fractions, the acidic fraction showed the most potent spasmolytic effects on histamine-induced contractions in isolated guinea pig tracheal rings. In addition, it inhibits increase of short circuit current induced by carbachol in isolated rat tracheal epitheliums with the IC50 values of 0.15 and 0.08 mg/ml, respectively. A short circuit is an abnormal low-resistance connection between two nodes of an electrical circuit that are meant to be at different voltages, as applied. The half-maximal inhibitory concentration (IC50) is a measure of the effectiveness of a compound in inhibiting biological or biochemical function. This quantitative measure indicates how much of a particular drug or other substance (inhibitor) is needed to inhibit a given biological process or component of a process (i.e. an enzyme, cell, cell receptor or microorganism) by half. Further in vivo studies concluded that the acidic fraction could protect an experimental asthma model induced by the combination of histamine and acetylcholine chloride in guinea pigs by prolonging the latent periods of asthma (P < 0.05) and significantly decreasing the cough frequency caused by ammonia water in mice (P < 0.001). This approach offers new and different ways to analyze the potential of material derived from earthworms because they might affect aspects of immunologic function such as immediate hypersensitivity [37].
In vitro approach confirms in vivo outcome According to Chu et al. [37], Pheretima (family Megascolecidae) have been documented as a potent agent for the treatment of cough and breathing difficulty in traditional Chinese medicine for nearly 2000 years. The water extract of Pheretima was separated into three 970
www.drugdiscoverytoday.com
Concluding remarks: perspectives on drug discovery from the earth This article has highlighted a substantial amount of information outlining the uses and potential modes of action of several
BOX 2
Hypercoagulation is often present in the following conditionsa Cancer Diabetes Fibromyalgia Crohn’s Disease Lyme Disease Multiple sclerosis Meniere’s Disease Chronic Fatigue Syndrome Chronic infections Lupus Gulf War Syndrome Excessive heavy metal burden Elevated serum fibrinogen Elevated serum CRP Elevated serum Lp[a] Elevated homocysteine Angina Heart attack history Transient ischemic attacks Ischemic stroke history On birth control pills On hormone replacement therapy Thrombocythemia Deep venous thrombosis Being on long air flights Hip fracture Ulcerative colitis Polycythemia Vascular dementia Autism ADD/ADHD Habitual miscarriages Infertility a
http://www.naturodoc.com/library/heart/boluoke.htm.
potentially interesting components of EEs and EPO. Primarily, model systems have been directed demonstrating the ability of
REVIEWS
such components to ameliorate aspects of the inflammatory process. According to Wallace [38] and Haefner [39], despite recent developments in combinatorial chemistry that can rapidly generate thousands of new chemicals, the pharmaceutical industry still relies heavily on a staggering array of as yet undiscovered possibilities from the natural environment. The terrestrial environment has been successfully mined for compounds and, currently, promising ones such as those from the earthworm – a vital animal from the earth – are being exploited, notably in the Far East. These chemicals are structurally complex, thus challenging organic chemists with approaches that might mimic synthetic versatility. Later, these products could well yield entirely new classes of drugs that would be valuable in treating human disease. Experimental trials have been directed to aspects of inflammation and lytic responses associated with hypercoagulation of blood (Box 2). Applications are pertinent, especially where coagulation and the need for dissolution are imperative. We need breakthroughs, overtures, discoveries and innovations in the face of rising medical costs and the fact that a considerable fraction of the general public (at least in the USA) is underinsured and often disgruntled with modernday Western medical practice. Clearly, the most prevalent of products from the earthworm consist of their role in providing lytic activity against a variety of pathological conditions [40– 43]. The earthworm has been crucial to the essence of this review concerned with aspects of inflammation [68]. This review now extends and clarifies information presented earlier [69].
Acknowledgements I express great appreciation for Lok-Hin Law for assistance in preparing this manuscript. I also acknowledge the collaboration of M.B. and his colleagues.
References 1 Cooper, E.L. (2008) From Darwin and Metchnikoff to Burnet and beyond. In Trends in Innate Immunity (Egesten, L.A., Herwald, L.H., Schmidt, W.A., eds), pp. 1–11, Karger 2 Cooper, E.L. (2004) Drug discovery, CAM and natural products. Evid. Based Complement. Alternat. Med. 1, 215–217 3 Cooper, E.L. (2005) CAM, eCAM, bioprospecting: the 21st century pyramid. Evid. Based Complement. Alternat. Med. 2, 125–127 4 Cooper, E.L. (2005) Bioprospecting: a CAM frontier. Evid. Based Complement. Alternat. Med. 2, 1–3 5 Engelmann, P. et al. (2002) Comparative analyses of earthworm immune system using cell surface and intracellular markets. In A New Model for Analyzing Antimicrobial Peptides with Biomedical Applications (Cooper, E.L. et al. eds), pp. 53–57, IOS Press 6 Cooper, E.L. (2002) The earthworm: a new model with biomedical applications. In A New Model for Analyzing Antimicrobial Peptides with Biomedical Applications (Cooper, E.L. et al. eds), pp. 3–28, IOS Press 7 Cooper, E.L. (March 2009) New enzyme complex isolated from earthworms is potent fibrinolytic: lumbrokinase has anti-platelet, anti-thrombotic activity. In Focus: Allergy Research Group Newsletter. 2–5 8 Wei, W. (1991) Forecast of research on anti-cancer effects of Di Long capsules. Chin. J. Clin. Res. Tumor 18, 131–132 9 Cheng, N.N. et al. (1992) Di Long phospholipid structure and platelet activating factor synthesis. J. Pharm. 27, 886 10 Li, W. et al. (1996) Using HPLC to determine quantity of hypoxanthine, xanthine, uracil, and uridyl in Di Long. J. Tradit. Chin. Med. Mater. 19, 625–627
11 Chen, B.Y et al. (1996) Analgesic and antipyretic effects of Di Long powder in mice, rats and rabbits. J. Shanghai Med. Univ. 23, 225–226 12 Zhang, F.X. et al. (1996) Experimental research on products from earthworm’s body that kills sperms. Shaanxi J. TCM 17, 234–236 13 Chen, J.B. et al. (1997) Chemical composition analysis of Di Long. Journal of Chinese Patented Medicine 19, 35–36 14 Zhang, F.C. et al. (1998) Di Long’s enhancing effect on macrophages’ immune activity. Chin. J. Pharm. 33, 532–535 15 Wang, G.Z. et al. (1998) Chemical composition analysis of Di Long related herbal material. J. Tradit. Chin. Med. Mater. 21, 133–135 16 Cooper, E.L. et al. (2004) Earthworms: sources of antimicrobial and anticancer molecules. In Complementary and Alternative Approaches to Biomedicine (Cooper, E.L. and Yamaguchi, N., eds), pp. 359–390, Kluwer Academic/Plenum Publishers 17 Balamurugan, M. et al. (2008) Hypothetical mode of action of earthworm extract with hepatoprotective and antioxidant properties. J. Zhejiang Univ. Sci. B 9, 141– 147 18 Ranganathan, L.S. (2006) Vermibiotechnology—from Soil Health to Human Health, Agrobios 19 Ismail, S.A. et al. (1992) Anti-inflammatory activity of earthworm extracts. Soil Biol. Biochem. 24, 1253–1254 20 Prakash, M. and Ranganathan, L.S. (2007) Anti-ulceral and anti-oxidative properties of ‘‘earthworm paste’’ of Lampito mauritii (Kinberg) on Rattus norvegicus. Eur. Rev. Med. Pharmacol. Sci. 11, 9–15
www.drugdiscoverytoday.com
971
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews POST SCREEN
21 Balamurugan, M. (2007) Earthworm paste (Lampito mauritii, Kinberg) alters inflammatory, oxidative, haematological and serum biochemical indices of inflamed rat. Eur. Rev. Med. Pharmacol. Sci. 11, 77–90 22 Balamurugan, M. (2009) Anti-inflammatory and antipyretic activities of earthworm extract- Lampito mauritii. J. Ethnopharmacol. 121, 330–332 23 Prakash, M. (2008) Effect of earthworm powder on antioxidant enzymes in alcohol induced hepatotoxic rats. Eur. Rev. Med. Pharmacol. Sci. 12, 237–243 24 Cho, J.H. et al. (1998) Lumbricin I, a novel proline-rich antimicrobial peptide from the earthworm: purification, cDNA cloning and molecular characterization. Biochim. Biophys. Acta 1408, 67–76 25 Dong, G.Q. et al. (2004) Molecular cloning and characterization of cDNA encoding fibrinolytic enzyme-3 from earthworm Eisenia foetida. Acta Biochim. Biophys. Sin. (Shanghai) 36, 303–308 26 Zhao, J. et al. (2005) Earthworm fibrinolytic enzyme. Stud. Nat. Prod. Chem. 30, 825– 847 27 Ge, T. et al. (2005) Cloning of thrombolytic enzyme (lumbrokinase) from earthworm and its expression in the yeast Pichia pastoris. Protein Expr. Purif. 42, 20– 28 28 Hu, Y. et al. (2005) Cloning and expression of earthworm fibrinolytic enzyme PM246 in Pichia pastoris. Protein Expr. Purif. 43, 18–25 29 Cheng, M.B. et al. (2008) Characterization of water-in-oil microemulsion for oral delivery of earthworm fibrinolytic enzyme. J. Control. Release 129, 41–48 30 Hahn, B.S. et al. (1997) Evaluation of the in vivo antithrombotic, antocoagulant, and fibrinolytic activities of Lumbricus rubellus earthworm powder. Arch. Pharm. Res. 20, 17–23 31 Kim, Y.S. et al. (1998) Dose dependency of earthworm powder on antithrombotic and fibrinolytic effects. Arch. Pharm. Res. 21, 374–377 32 Mihara, H. et al. (1991) A novel fribrinolytic enzyme extracted from the earthworm, Lumbricus rubellus. Jpn. J. Physiol. 41, 461–472 33 Jin, L. et al. (2000) Changes in coagulation and tissue plasminogen activator after the treatment of cerebral infarction with lumbrokinase. Clin. Hemorheol. Microcirc. 23, 213–218 34 Ge, T. et al. (2007) High density fermentation and activity of a recombinant lumbrokinase (PI239) from Pichia pastoris. Protein Expr. Purif. 52, 1–7 35 Yamaji, A. et al. (1998) Lysenin, a novel sphingomyelin-specific binding protein. J. Biol. Chem. 273, 5300–5306 36 Kobayashi, H. et al. (2004) Biology of lysenin, a protein in the coelomic fluid of the earthworm Eisenia foetida. Int. Rev. Cytol. 236, 45–99 37 Chu, X. et al. (2007) In vitro and in vivo evaluation of the anti-asthmatic activities of fractions from Pheretima. J. Ethnopharmacol. 111, 490–495 38 Wallace, R.W. (1997) Drugs from the sea: harvesting the results of aeons of chemical evolution. Mol. Med. Today 3, 291–295 39 Haefner, B. (2003) Drugs from the deep: marine natural products as drug candidates. Drug Discov. Today 8, 536–544 40 Cooper, E.L. (2008) eCAM: an emerging linkage with ethnopharmacology? Evid. Based Complement. Alternat. Med. 5, 365–366 41 Wang, F. et al. (2003) Purification, characterization and crystallization of a group of earthworm fibrinolytic enzymes from Eisenia fetida. Biotechnol. Lett. 25, 1105–1109 42 Cho, I.H. et al. (2004) Purification and characterization of six fibrinolytic serineproteases from earthworm Lumbricus rubellus. J. Biochem. Mol. Biol. 37, 199–205 43 Tang, Y. et al. (2002) Crystal structure of earthworm fibrinolytic enzyme component A: revealing the structural determinants of its dual fibrinolytic activity. J. Mol. Biol. 321, 57–68 44 Hrzenjak, T.M. et al. (1998) Fibrinolytic activity of earthworms extract (G-90) on lysis of fibrin clots originated from the venous blood of patients with malignant tumors. Pathol. Oncol. Res. 4, 201–211 45 Hu, W. and Fu, T. (1997) Isolation and properties of a novel fibrinolytic enzyme from an earthworm. Zhong Yao Cai 20, 78–81 46 Zhao, J. et al. (2003) Hydrolysis of fibrinogen and plasminogen by immobilized earthworm fibrinolytic enzyme II from Eisenia fetida. Int. J. Biol. Macromol. 32, 165–171
972
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
47 Liu, X.H. and Ge, F. (2002) Factors influencing the activity of fibrinolytic enzymes from earthworm, Eisenia foetida. Zhongguo Zhong Yao Za Zhi 27, 423–426 48 Mihara, H. et al. (1991) A novel fibrinolytic enzyme extracted from the earthworm, Lumbricus rubellus. Jpn. J. Physiol. 41, 461–472 49 Nakajima, A. et al. (1996) Specific clonal T cell accumulation in intestinal lesions of Crohn’s disease. J. Immunol. 157, 5683–5688 50 Sugimoto, M. and Nakajima, N. (2001) Molecular cloning, sequencing, and expression of cDNA encoding serine protease with fibrinolytic activity from earthworm. Biosci. Biotechnol. Biochem. 65, 1575–1580 51 Xing, B.D. et al. (1997) Purification and characterization of the fibrinolytic enzyme (eFE-D) from earthworm Eisenia fetida. Acta Biochim. Biophys. Sin. (Shanghai) 29, 609– 612 52 Yang, J.S. and Ru, B.G. (1997) Purification and characterization of an SDS-activated fibrinolytic enzyme from Eisenia fetida. Comp. Biochem. Physiol. B Biochem. Mol. Biol. 118, 623–631 53 Zhao, X.Y. et al. (2006) A component of earthworm fibrinolytic enzyme having higher thrombolytic activity than total components in vivo. Yao Xue Xue Bao 41, 1068–1073 54 Olivares Fontt, E. et al. (2002) Trypanosoma cruzi is lysed by coelomic cytolytic factor-1, an invertebrate analogue of TNF, and induces phenoloxidase activity in the coelomic fluid of Eisenia foetida. Dev. Comp. Immunol. 26, 27–34 55 Hrzenjak, T. et al. (1992) New source of active compounds- earthworm tissue (Eisenia fetida, Lumbricus rubellus). Comp. Biochem. Physiol. 102, 441–447 56 Kauschke, E. and Mohrig, W. (1987) Comparative analysis of hemolytic and hemagglutinating activities in the coelomic fluid of Eisenia foetida and Lumbricus terrestris (Annelida, Lumbricidae). Dev. Comp. Immunol. 11, 331–341 57 Kauschke, E. et al. (2001) Evidence for perforin-like activity associated with earthworm leukocytes. Zoology (Jena) 104, 13–24 58 Kobayashi, H. et al. (2004) Biology of lysenin, a protein in the coelomic fluid of the earthworm Eisenia fetida. Int. Rev. Cytol. 236, 45–99 59 Lange, S. et al. (1999) Biochemical characteristics of Eiseniapore, a pore-forming protein in the coelomic fluid of earthworms. Eur. J. Biochem. 262, 547–556 60 Milochau, A. et al. (1997) Purification, characterization and activities of two hemolytic and antibacterial proteins from coelomic fluid of the annelid Eisenia fetida andrei. Biochim. Biophys. Acta 1337, 123–132 61 Mohrig, W. et al. (1996) Crossreactivity of hemolytic and hemagglutinating proteins in the coelomic fluid of lumbricidae (Annelida). Comp. Biochem. Physiol. Physiol. 116A, 19–30 62 Roch, P. et al. (1981) Protein analysis of earthworm coelomic fluid. II. Isolation and biochemical characterization of the Eisenia fetida andrei factor (EFAF). Comp. Biochem. Physiol. 69B, 829–836 63 Roch, Ph. et al. (1989) Interactions between earthworm hemolysins and sheep red blood cell membranes. Biochim. Biophys. Acta 983, 193–198 64 Sekizawa, Y. et al. (1997) Molecular cloning of cDNA for lysenin, a novel protein in the earthworm, Eisenia foetida that causes contraction of rat vascular smooth muscle. Gene 191, 97–102 65 Lassegues, M. et al. (1997) Sequence and expression of an Eisenia foetida derived cDNA clone that encodes the 40 kD fetidin antibacterial protein. Eur. J. Biochem. 246, 756–762 66 Beschin, A. et al. (1998) Identification and cloning of a glucan- and lipopolysaccharide-binding protein from Eisenia foetida earthworm involved in the activation of prophenoloxidase cascade. J. Biol. Chem. 273, 24948–24954 67 Cho, J.H. et al. (1998) Lumbricin I a novel proline-rich antimicrobial peptide from the earthworm: purification, cDNA cloning and molecular characterization. Biochim. Biophys. Acta 1408, 67–76 68 Balamurugan, M. (2007) Restoration of histoarchitecture in the paracetamolinduced liver damaged rat by earthworm extract, Lampito mauritii. Eur. Rev. Med. Pharmacol. Sci. 11, 407–411 69 Cooper, E.L. et al. (2002) Digging for innate immunity since Darwin and Metchnikoff. Bioessays 24, 319–333
REVIEWS
Reviews POST SCREEN
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
What is the value of human FMRI in CNS drug development? Richard G. Wise1 and Cliff Preston2 1 2
Cardiff University Brain Research Imaging Centre, School of Psychology, Cardiff University, Park Place, Cardiff CF10 3AT, UK Portfolio & Decision Analysis Group, Pfizer Ltd, Ramsgate Road, Sandwich CT13 9NJ, UK
Functional neuroimaging has the potential to improve the decision-making process in the development of new drugs. With the high cost of failure of compounds in later stages of development, there is a need to establish, early in man, reliable measures of drug activity and efficacy in the brain. Functional magnetic resonance imaging (FMRI) is a tool for serially examining normal and pathological brain function at the systems level. FMRI is helping us to understand therapeutic mechanisms and can provide clinically relevant markers of disease responses to drugs. An analysis of the value of FMRI to aid decision-making requires an appreciation of the techniques and their validation, a task that has begun and which necessitates an investment of its own.
Introduction There is a discrepancy between the high spending levels in drug discovery and development and the comparatively small number of effective compounds for CNS disorders reaching the market [1,2]. There is a well-recognized need to cut down attrition rates in the clinical stages of drug development. The earlier in the process this is achieved, the greater the potential resource savings and overall greater therapeutic success across a portfolio. This is especially important given the trend towards personalized medicines and the threat therefore posed to the blockbustera [3,4]. Although stratification of patients is likely to bring improved treatment efficacy, the increasingly smaller target groups bring smaller financial returns. Drug development costs, however, remain essentially similar. Research tools are needed to improve our understanding of CNS disease mechanisms and their modification in treatment and to improve the efficiency of decisionmaking in early clinical phases, reducing the likelihood of costly failure in phase III. Although pharmacological functional magnetic resonance imaging (FMRI) is not the only available technique, it is emerging as a candidate for improving the efficiency of the drug discovery and Corresponding author:. Wise, R.G. ([email protected]) a A blockbuster medicine is defined as being one that achieves annual revenues of more than US$ 1 billion at a global level.
1359-6446/06/$ - see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.drudis.2010.08.016
development process, particularly in human studies. The cost of FMRI examination is not trivial; therefore, the widespread adoption of FMRI in drug development will require it to demonstrate penetration of a compound into the brain (central penetration) or clinically relevant markers of disease or treatment-based recovery or safety with a substantially greater sensitivity than currently available readouts, which might be cheaper. Where these improvements are demonstrated, a reduced number of patients can be studied. Careful assessment of the performance of FMRI in this process is still required to establish the real benefits that might accrue across the lifecycle of development of a novel or newly indicated compound. The assessment of FMRI as a tool has begun with some substantial investments from the pharmaceutical industry (http://cic.gsk.co.uk/), including public–private academic collaborations, in-house imaging facilities and the establishment of imaging consortia [5].
What is FMRI? FMRI is rarely out of the news, being applied to fields of research as diverse as neuromarketing (http://news.bbc.co.uk/1/hi/sci/tech/ 8569087.stm) and persistent vegetative state (http://news. bbc.co.uk/1/hi/health/8497148.stm). The technique has spread to the point at which many universities have now invested in their own imaging facilities as brain research tools. FMRI can be performed in humans on most newer MRI systems, although www.drugdiscoverytoday.com
973
REVIEWS
benefits are seen at higher field strengths (e.g. 3 T), with generally improved sensitivity to detect changes in brain activity. FMRI encompasses a collection of evolving non-invasive MRI techniques sensitized to the haemodynamic state of the brain that are used to infer changes in neural activity through changes in blood oxygenation, blood flow and cerebral blood volume (Fig. 1). FMRI provides a systems-level view of brain function in humans and animals. Reviews POST SCREEN
BOLD FMRI Blood-oxygenation-level-dependent (BOLD) FMRI (Fig. 1) is by far the most commonly implemented method [6,7] for pharmacological and non-pharmacological studies because it provides the best functional image contrast-to-noise ratio. A rapid imaging technique known as echo-planar imaging provides whole-brain measurements with a resolution of typically 3 mm 3 mm 3 mm every 3 s. Localized BOLD image contrast arises from the exquisite capability of the brain to control blood flow on a small spatial scale. Increased neural activity results in a local vasodilatation, in which the fractional increase in cerebral blood flow (CBF) is a factor of two, or more, larger than the fractional increase in metabolic oxygen consumption [8–10]. The quantity of deoxyhaemoglobin on the venous side of the local vasculature decreases. Deoxyhaemoglobin, being paramagnetic, distorts the magnetic field from the MRI scanner in and around the blood vessels (capillaries, venules and veins), reducing the coherence and hence the net intensity of the NMR signal from hydrogen nuclei in that area. A reduction of these distortions with increased blood flow increases the signal intensity: the basis of the BOLD effect. The degree of BOLD signal change, or BOLD response, is small, typically 1% of the image intensity in response to a change in neural activity. This varies widely depending on the magnetic field strength, the brain region and underlying physiology or pathology (local blood volume and vascular responsiveness), and the type of stimulation task given to the subject (long blocks or short events).
Perfusion FMRI Regional CBF measurement with dynamic imaging of bolus contrast agents has been performed for some time [11]. With recent improvements in scanner software and hardware, however, noninvasive (without exogenous contrast) measurements of CBF can be made in a few minutes (Fig. 1), using techniques known generically as arterial spin labelling (ASL) [12]. Blood on its way to the brain is ‘tagged’ using a radiofrequency excitation pulse. It flows into the brain and water is exchanged with the tissue; the tracer is delivered. The magnetization state of the tissue water is then interrogated with an imaging read-out. The procedure is repeated without tagging the inflowing blood to form a control image, and the difference in regional signal intensity between tag and control image is proportional to CBF [13]. Assumptions are normally made about the transit time for the blood, providing a convenient measure of CBF in ml/100 g tissue/min. These assumptions, however, must be examined where pathology might intervene (e.g. the increase of arterial transit time in stroke or where a pharmacological agent substantially increases the CBF). The functional contrast-to-noise ratio is normally lower for ASL CBF measures than for BOLD FMRI, but CBF offers the advantage of being a physical quantity and less susceptible to physiological alterations, 974
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
which can modulate BOLD contrast [14,15] and confound its interpretation. ASL CBF measures provide similar information to gold standard of oxygen-15 PET [16] but without the need for a radioactive tracer. ASL MRI, therefore, is more appropriate for human studies, especially where repeat scanning is beneficial, such as in cross-over studies. To date, ASL CBF measures have been applied in preclinical pharmacological studies [17,18] but are only just making their way into human pharmacological investigations [19–22] (Fig. 1). More widely, the potential of ASL perfusion measures in studying longer term changes in brain activity is now being recognized [23] and will undoubtedly become more important in investigating regional pharmacological activity in the brain. The sensitivity of BOLD FMRI is poor when the changes in brain activity of interest occur over a timescale longer than a couple of minutes because BOLD signal is subject to slow signal drifts. ASL perfusion is less sensitive to such noise and so is more appropriate for measuring haemodynamic changes over the timescale of minutes to days to months [24,25]. ASL perfusion measures, therefore, have the potential to measure long-term changes in local CBF such as those arising from chronic oral dosing.
Brain activity BOLD contrast probably reflects most closely the input and intracortical synaptic processing of a brain area, rather than its spiking output [26]. This is important to consider when interpreting the pharmacological interventions in FMRI studies. An observed site of altered BOLD or blood flow response might be distant from the binding site. Logothetis [27] has reviewed the limitations of the interpretation of FMRI signal. The dynamic nature of the BOLD signal has led to two principal manipulations of interest in pharmacological studies. The first is the classical FMRI approach of modulating the response to brief stimuli using a pharmacological agent. There are many examples of this; for example, the use of pain signals to demonstrate the time course of pharmacodynamics of an analgesic in the brain [28]. The second is the newer field of resting-state FMRI [29]. In BOLD and ASL signals, the brain exhibits long-range temporal correlations in the absence of an explicit task being performed by the volunteer (e.g. left and right motor cortices) [30]. On the basis of what we know about regional specialization of function, the areas in which BOLD signals co-vary in time seem to be robustly organized into plausible functional networks (e.g. visual, sensory motor [31] and one termed the ‘default mode’, which seems to be more active in the absence of a cognitively engaging task [32]). The strength of the temporal covariation is commonly interpreted, perhaps somewhat hopefully, as indicating the strength of functional connectivity. This has opened the way for pharmacological modulation of this property and the interpretation of drug effects on the communication between brain regions [33]. There are many pitfalls in the evaluation of FMRI data [15]. The application of what are normally arbitrary statistical thresholds – although important for rigour – arguably places too much importance on the apparent presence or absence of ‘activity’, which could lead to over interpretation when using results for decision-making in the development pipeline. In addition, in a pharmacological study, the molecular interactions of the compound might intervene in the signal transduction process, disrupting our ability to interpret
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
Reviews POST SCREEN
[(Figure_1)TD$IG]
FIGURE 1
Steps in the generation of the BOLD FMRI signal. Alterations in neural activity demand more blood flow, resulting in a net decrease in venous deoxyhaemoglobin and an increase in MRI signal: blood-oxygenation-level dependent (BOLD). The cerebral blood flow (CBF) increase can also be detected directly and non-invasively using arterial spin labelling FMRI. A drug might influence the neural activity, signalling to the blood vessels or the vascular responsiveness of the brain region under examination.
BOLD signal changes as being a faithful representation of neural activity. This can occur through drug-induced changes in neurovascular coupling, vascular reactivity and/or dynamic alterations in the basal physiological (oxygenation) state at the global or local level. We and other groups are developing strategies to improve the specificity of FMRI in measuring drug-induced changes in neural activity. Such strategies need to be tested for each class of compounds under investigation. They include quantification of druginduced changes in regional vascular reactivity [34], cerebral perfusion [35], altered components of physiological noise [36], metabolic oxygen consumption [20] and concurrent measurements of electrophysiological (electroencephalographic, or EEG) activity to compare with the haemodynamic (FMRI) response [37]. Simultaneous EEG-FMRI is particularly appropriate for examining coupling relationships between fast (millisecond-timescale) synaptic currents, measured at the scalp, and the highly spatially resolved (millimetrescale) vascular BOLD FMRI response. The marriage of EEG and FMRI might provide more sensitive and specific markers of drug effects on receptor groups by supporting changes seen in FMRI as originating from alterations in synaptic currents.
Uses of FMRI To make efficient and realistic use of FMRI in drug development, it is important to understand what it can and cannot offer. It would be unfortunate for the opportunity to exploit the true value of the technique to be missed through disillusionment arising from the dashing of false hopes. FMRI is likely to have a well-focussed and
important role at specific stages of drug discovery and development (Fig. 2).
Early stages Before the introduction of a drug, FMRI can improve our understanding of the normal function in animals and man of the network of brain regions (or circuits) believed to be related to the disease impairment. In pain, for example, FMRI has shed much light on the cortical and subcortical networks engaged in pain processing [38], providing the foundations for new neuroanatomical and hence neuropsychological hypotheses for treatment interventions. In a similar manner, the characteristics of disease models can inform us about potential disease processes and suggest relevant imaging markers for them (e.g. sensitization in neuropathic pain [39]). In the preclinical stages, FMRI can be combined with more invasive measures to aid in target selection, providing some early indication of a relevant pharmacodynamic effect and, by extension, lead optimization. By virtue of the haemodynamic origin of the FMRI signal and the brain-network-based information that it reveals, FMRI is probably a closer reflection of behaviour and the associated interacting streams of information processing than it is the receptor distribution associated with a specific target. FMRI is likely to be sensitive to changes in brain activity downstream of the direct site of action, engaging multiple neurotransmitter systems, as well as potentially at that site, where it might be associated with a metabolic demand of binding. Although this property of FMRI signals results in www.drugdiscoverytoday.com
975
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
[(Figure_2)TD$IG]
Drug discovery
Drug development Preclinical
Reviews POST SCREEN
Markers in disease models
Identify brain circuits
Biomarker fingerprint
Mechanistic understanding
Target selection and testing
Selection of relevant model
Animal replacement
Lead optimization
Clinical Phase 1 Healthy volunteers
Early signal of efficacy (ESOE)
Systems neuroscience
Patient stratification
Approved & safe failed drugs: novel uses
Optimizing FMRI for decision making FMRI marker specific to mechanism
Normal function Human and animal disease models to understand disease processes
Phase 3 Biomarker of efficacy and disease Safety signals
Go/NoGo decision making, faster and cheaper
Unexpected effects
Basic neuroscience
Phase 2a 2b Proof of Dose concept ranging
Maximum sensitivity to ESOE: optimized design and analysis
Building confidence: validation of specific FMRI markers of efficacy in multiple trials (marketed compounds, negative control etc.) Drug Discovery Today
FIGURE 2
The roles for FMRI in the drug development process. The highlighted area indicates the role of FMRI early in the process in man to produce signals of efficacy and central penetration. One key aim of the use of FMRI is to assist in go–no-go decision-making.
reduced specificity to different target receptors, it does provide a translational systems-based tool to compare the effects of a drug on both animal and human. Translation from animals to man can provide reassurance of drug activity on the desired system and might promote replacement of animals with early studies in small cohorts of human volunteers. Conversely and more innovatively, translation from man back to animals could offer imaging measurements with which to refine the relevance of the animal model to the human condition and its treatment [40]. In preclinical models, FMRI also has the potential to indicate unexpected effects on brain systems, in drugs targeted centrally or indeed at the periphery, with a potential future treatment role or safety implication. One of the key challenges in the early stages of drug development is to demonstrate some action of the compound in the CNS or central penetration and central activity. Although PET offers a direct approach to examining receptor binding, PET ligands might not be available for the compound in question. The haemodynamic nature of FMRI techniques means they cannot unequivocally distinguish between central action and the central consequences of a peripheral action of a compound; however, they might provide some additional confidence in making this distinction when combined with information from other sources (e.g. behavioural studies). ASL perfusion measurements, we believe, are likely to become more important in early studies in man in establishing central penetration (crossing the blood–brain barrier). Where a drug alters neural activity as a consequence of receptor binding or other activity, vascular tone, or metabolic activity, this is likely to result in a modulation of CBF. Blood flow is coupled to neural and metabolic activity. Therefore, the demonstration of a local change in CBF in 976
www.drugdiscoverytoday.com
response to a single dose of a compound in a functionally plausible network of brain areas at the most basic level would indicate a probable central drug effect, whether that be vascular, neuronal or metabolic, and provide some guidance on the choice of dose.
Clinical biomarkers FMRI would be most useful as a sensitive indicator of pharmacological responses to a therapeutic intervention [1,41,42]. To have the greatest value in decision-making in the drug development process, FMRI needs to index changes in clinically relevant patterns of brain activity that can be mechanistically related to the disease process or treatment effect. At that stage, FMRI would be able to validate novel drug targets and predict drug responses, with a concomitant important contribution to gaining regulatory approval. Despite certain applications making good progress in the pharmacological FMRI field, such as psychiatry [43] and pain [44], FMRI has not yet yielded extensively validated biomarkers with a predictive value demonstrated in novel compounds. Most research effort has focussed on investigations with well-characterized licensed compounds. The lack of work in evaluating novel compounds is partly a question of time and the recency of the technique. We draw an analogy between the role of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in cancer trials and the role of FMRI in drug discovery. DCE-MRI has become a well-used marker in the cancer field for examining angiogenesis. It is used in early clinical trials of anti-angiogenic compounds, in which it provides evidence of efficacy and dose-dependent responses to treatment [45]. FMRI, with further validation of its role, might prove equally useful in clinical trials in neurology and psychiatry.
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_3)TD$IG]
beneficial effects. Such an approach could conceivably continue into clinical practice.
The value of FMRI in decision-making It is clear from the above that FMRI has the potential to be applied at various stages of drug development, but is the information it provides valuable and is the value sufficient to justify the additional costs of performing an FMRI study? One way to address this question is to develop a decision-analysis valuation framework [50], such as in the following example based on the development of a novel drug for treating neuropathic pain. Because the purpose of the analysis is to quantify the benefit of the incremental cost of the FMRI study, the first step is to set out the development path that would be followed in the absence of an FMRI option. This will form the base case against which a decision to invest in FMRI will be assessed. Typically, the development path consists of a conventional phase I program, followed by a phase IIa ‘proof-of concept’ study in a highly selected population of pain patients. If this is successful, it will be followed by phase IIb studies to test a wider range of doses and a more broadly defined patient population. If these studies are successful, they will be followed by phase III confirmatory studies, registration and marketing. The evaluation of the base case is performed through a decision tree describing the cost, timing and probability of success of each phase, resulting in a range of commercial value that can be achieved if the drug successfully reaches the market (Fig. 3). Metrics can then be derived from the analysis representing the
Conventional program P=0.5
$20m
Launch / Commercialisation
$200m
P=0.2
Phase IIb, III, registration
Phase IIa proof of concept
Phase I
Fail Fail Risk-adjusted value $127m
Risk-adjusted return on investment 4.4
FMRI program P=0.5 $200m
P=0.5 $20m
P=0.4 $2m Phase I
Phase IIb, III, registration
Phase IIa proof of concept
FMRI
Launch / Commercialisation
Fail Fail
Fail Risk-adjusted value $132m
Risk-adjusted return on investment 5.6 Drug Discovery Today
FIGURE 3
FMRI as a simple filter. An example comparison between the ‘base case’, a conventional drug development program and one including an FMRI filter stage, assuming that phase I has already been successful. We suggest that the use of FMRI might increase the probability of success at phase IIa. Values are after tax. www.drugdiscoverytoday.com
977
Reviews POST SCREEN
An FMRI biomarker might predict clinical benefit or harm and could be in the form of a fingerprint of altered brain activity in response to the therapy [40,46], taking advantage of newer multivariate classification analysis techniques [47] rather than the more traditional univariate image voxel-wise or region-of-interest based analysis [48]. FMRI biomarkers and surrogate endpoints are in the process of being validated by academic and industrial centres. They will need to be validated for each indication and class of compounds to build up experience in interpreting each fingerprint, probably in a multivariate analysis of image data incorporating many brain regions. FMRI is likely to be particularly important in objectively quantifying subjective reports provided by patients with the potential for reducing variability in the data [49]. This is likely to be particularly important in psychopharmacology for psychiatric and certain neurological conditions because of the qualitative nature of self-reports and the potentially poor animal models for these human diseases. FMRI has the advantage of being non-invasive, permitting safe repeat scanning and facilitating cross-over studies with an appropriate washout period, within-subject dose–response investigations and comparisons of different agents. It also offers the possibility of long-term followup for treatment markers and safety signals in the late stages of development and post-marketing, particularly when combined with structural MRI markers such as regional atrophy. With a trend towards smaller target patient groups for any given compound, FMRI could also play an important part in stratifying patients to enter developmental trials to maximize the observed
REVIEWS
Reviews POST SCREEN
value of the project (risk-adjusted net present value, or rNPV) and the efficiency (risk-adjusted return on investment). The rNPV metric (also frequently referred to as ‘expected net present value’) takes into account both the timing and the uncertainty associated with future cash flows. It is calculated by first taking all future costs or revenues and applying discount factors to convert them into their equivalent in today’s money. These discounted figures are then risk-adjusted by multiplying them by their probability of occurrence and summed to give a single figure, which represents the overall value of the project [51]. Calculating risk-adjusted return on investment (also referred to as ‘expected productivity index’) requires the additional step of dividing the rNPV by the development costs (which are also discounted and risk-adjusted to reflect the probability they will be paid). This gives a ratio that reflects the efficiency of the investment, in which zero is the break-even point and positive figures represent attractive investments. The next step in the analysis is to determine how the FMRI study will be incorporated into the development program and how the results of the FMRI study will change the pattern of cost, risk and timing relative to this base case. One key observation is that from a decision-analytic perspective, information itself is not intrinsically valuable. Information only has value to the extent that it can influence a decision. In the current example, the best place to insert an FMRI study is before the phase IIa proof-of-concept study, to resolve as much of the efficacy risk as possible for minimal cost. The value of the FMRI study will consequently depend on the decisions that are made on its outcome; if the decision is to proceed to the proof-of-concept study anyway (albeit perhaps with increased confidence in the probable result), then the FMRI study is worthless. An example of this approach is illustrated in
[(Figure_4)TD$IG]
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
Fig. 3, which contrasts the base-case plan in the top panel with an FMRI-based plan below. In the example, for simplicity’s sake, it is assumed that the FMRI study has no appreciable false negative rate and, therefore, that the overall chance of reaching the market is unchanged; it is also assumed that the study can be inserted into the development program without being on the crucial path. In an analysis of a real example, either or both of these simplifying assumptions could be relaxed. Also for simplicity, in the example the development steps after proof-of concept are depicted as one block whereas, in reality, each phase has its own cost and associated probability of success. The final step in such an approach should typically involve sensitivity analysis around the assumptions involved. This can frequently be the most important step of all because it reveals which assumptions are crucial for making the investment decision and which do not matter. The procedure involves systematically varying one of the assumptions (such as the cost or duration of the FMRI study) until the economics no longer favour the FMRI approach over conventional development. From the current example, it should be obvious that the cost of the FMRI study is important, and if it approaches the cost of the proof-of-concept study then FMRI offers no benefit. Likewise, if the FMRI study forms part of the critical path and produces a substantial increase in the length of the drug development program then its benefits will be eroded. Perhaps less intuitively, the value of the FMRI study will also depend on the prior probability that the drug will successfully achieve proof-of concept because the main benefit of the FMRI study is in providing a relatively cheap means of terminating projects destined to fail. For example, to take an extreme case, if a project is certain to succeed then inclusion of the FMRI study simply adds $2m to the development cost, which results in a net
FIGURE 4
The FMRI filter: analysis of costs. In this example, using FMRI can save US$ 100 million per successful project, or an average of US$10 million for each project that completes phase I. These savings could, in turn, be used to fund more projects in the portfolio. 978
www.drugdiscoverytoday.com
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
REVIEWS
[(Figure_5)TD$IG]
FMRI program with 3-way decision P=0.5 Launch / Commercialisation
$220m
P=0.05
Enhanced phase IIb, III, registration
Fail
P=0.43 P=0.35 $20m $2m Phase I
Launch / Commercialisation
Phase IIb, III, registration
Phase IIa proof of concept
FMRI
Reviews POST SCREEN
P=0.5 $200m
Fail Fail
Fail
Drug Discovery Today
FIGURE 5
FMRI could also support a decision for fast development. Simple success in the FMRI study leads to the full proof-of-concept study, as before (Fig. 3). A superior result in the FMRI study could provide sufficient confidence to proceed directly to phase IIb, theoretically enabling an earlier launch and higher commercial value, which further adds value to the FMRI-based decision. For clarity, it is again assumed here that the FMRI study does not change the overall chance of launch, but in this example it can influence timing.
decrease in the value of the project. The ‘sweet spot’ for using FMRI will be for projects that have a moderate chance of success. Sensitivity analysis can also help determine acceptable false positive and false negative rates from the FMRI study. Too high a false negative rate will result in missed opportunities through the termination of some projects that could potentially become successful drugs. Too high a false positive rate will simply result in extra cost because more projects destined to be unsuccessful progress to (and fail in) the phase IIa proof-of-concept study. The valuation framework can be used to help decide whether the FMRI study should be incorporated in any particular project and, if so, what success criterion should be used to best trade-off these possible outcomes. In the current example (Fig. 3), because the FMRI study does not affect time to market or the overall probability of success, a shortcut can be used that simply compares the average or risk-adjusted cost of the two programs. This is easy to calculate, particularly if one applies the thinking to a collection of projects and then tracks their fate through the developmental cascade. Figure 4 shows an example of this, in which use of FMRI saves an average of $10m per project. In Fig. 5, a more ambitious decision framework is illustrated, in which the FMRI study could support a third branch based on superior results. This could result in the same overall success rate but a faster path to market built upon increased confidence in the magnitude of the efficacy signal. Again, sensitivity analysis can be used to help decide whether such a decision framework truly adds additional value or whether the simpler filter approach is sufficient. Of course, these examples assume that appropriate FMRI protocols and analysis toolkits are already in place with sufficient validation to support decision-making. Before the application of FMRI to the decision-making process, a decision must have been made to invest in developing the appropriate FMRI protocol itself. That decision would typically take into account the cost and staging of the validation studies, as well as the probability that the validation will be successful. This will be compared with the
average value FMRI can add to each compound it is applied to and the probable number of projects that will benefit. The number of projects might be limited if there is a danger that a new technology will emerge that has advantages over the validated model. The organization making this initial investment (or organizations, in the case of a consortium approach) will, therefore, benefit most if they have a broad range of projects to which the technology can be applied, occurring with a fair degree of certainty within a reasonably short time.
Concluding remarks When a compound is going to fail in the development pipeline, it is desirable to have it fail as fast and as cheaply as possible. FMRI is likely to contribute to this decision-making process, but we are currently still in the phase of validating FMRI for this purpose. More experience of the technique in the context of drug development is needed to provide data to model well the cost-effectiveness of the widespread application of FMRI. This experience is being gained largely through partnerships between industry and academia and is perhaps best done in an industry-wide precompetitive phase [5]. This task is enlarged by the need to constantly challenge the assumptions underlying the measurements at every stage of their development, particularly in applying them to patient populations such as the elderly where the underlying cerebral physiology might be altered. In FMRI, however, we have a tool that has proved very valuable for neuroscience in the past 15 years and when applied strategically in the drug development pipeline could result in substantial cost savings across a portfolio of compounds.
Acknowledgements RW wishes to acknowledge the generous support of the UK Medical Research Council for his Career Development Award. CP thanks colleagues Bill Vennart, Keith Tan, John Huggins, Trevor Smart and Duncan McHale. www.drugdiscoverytoday.com
979
REVIEWS
Drug Discovery Today Volume 15, Numbers 21/22 November 2010
References
Reviews POST SCREEN
1 Frank, R. and Hargreaves, R. (2003) Clinical biomarkers in drug discovery and development. Nat. Rev. Drug Discov. 2, 566–580 2 Kola, I. and Landis, J. (2004) Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug Discov. 3, 711–715 3 Cutler, D.M. (2007) The demise of the blockbuster? N. Engl. J. Med. 356, 1292–1293 4 Frantz, S. (2005) 2004 approvals: the demise of the blockbuster? Nat. Rev. Drug Discov. 4, 93–94 5 Borsook, D. et al. (2008) A ‘BOLD’ experiment in defining the utility of fMRI in drug development. Neuroimage 42, 461–466 6 Ogawa, S. et al. (1990) Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. U. S. A. 87, 9868–9872 7 Kwong, K.K. et al. (1992) Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc. Natl. Acad. Sci. U. S. A. 89, 5675–5679 8 Hoge, R.D. et al. (1999) Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex. Proc. Natl. Acad. Sci. U. S. A. 96, 9403–9408 9 Davis, T.L. et al. (1998) Calibrated functional MRI: mapping the dynamics of oxidative metabolism. Proc. Natl. Acad. Sci. U. S. A. 95, 1834–1839 10 Buxton, R.B. et al. (2004) Modeling the hemodynamic response to brain activation. Neuroimage 23 (Suppl. 1), S220–S233 11 Sourbron, S. et al. (2009) Quantification of cerebral blood flow, cerebral blood volume, and blood–brain-barrier leakage with DCE-MRI. Magn. Reson. Med. 62, 205–217 12 Petersen, E.T. et al. (2009) The QUASAR reproducibility study, part II: results from a multi-center arterial spin labeling test-retest study. Neuroimage 49, 104–113 13 Wong, E.C. et al. (1998) Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II). Magn. Reson. Med. 39, 702–708 14 Tjandra, T. et al. (2005) Quantitative assessment of the reproducibility of functional activation measured with BOLD and MR perfusion imaging: implications for clinical trial design. Neuroimage 27, 393–401 15 Iannetti, G.D. and Wise, R.G. (2007) BOLD functional MRI in disease and pharmacological studies: room for improvement? Magn. Reson. Imaging 25, 978–988 16 Bokkers, R.P. et al. (2009) Arterial spin labeling perfusion MRI at multiple delay times: a correlative study with H(2)(15)O positron emission tomography in patients with symptomatic carotid artery occlusion. J. Cereb. Blood Flow Metab. 30, 222–229 17 Bruns, A. et al. (2009) Validation of cerebral blood perfusion imaging as a modality for quantitative pharmacological MRI in rats. Magn. Reson. Med. 61, 1451–1458 18 Luo, F. et al. (2009) Differential responses in CBF and CBV to cocaine as measured by fMRI: implications for pharmacological MRI signals derived oxygen metabolism assessment. J. Psychiatr. Res. 43, 1018–1024 19 Chen, Y. and Parrish, T.B. (2009) Caffeine dose effect on activation-induced BOLD and CBF responses. Neuroimage 46, 577–583 20 Qiu, M. et al. (2008) Anesthetic effects on regional CBF, BOLD, and the coupling between task-induced changes in CBF and BOLD: an fMRI study in normal human subjects. Magn. Reson. Med. 60, 987–996 21 Qiu, M. et al. (2008) Spatial nonuniformity of the resting CBF and BOLD responses to sevoflurane: in vivo study of normal human subjects with magnetic resonance imaging. Hum. Brain Mapp. 29, 1390–1399 22 MacIntosh, B.J. et al. (2008) Measuring the effects of remifentanil on cerebral blood flow and arterial arrival time using 3D GRASE MRI with pulsed arterial spin labelling. J. Cereb. Blood Flow Metab. 28, 1514–1522 23 Tracey, I. and Johns, E. (2010) The pain matrix: reloaded or reborn as we image tonic pain using arterial spin labelling. Pain 148, 359–360 24 Wang, J. et al. (2003) Arterial spin labeling perfusion fMRI with very low task frequency. Magn. Reson. Med. 49, 796–802 25 Borogovac, A. et al. Mapping brain function using a 30-day interval between baseline and activation: a novel arterial spin labeling fMRI approach. J. Cereb. Blood Flow Metab. (in press), doi:10.1038/jcbfm.2010.89
980
www.drugdiscoverytoday.com
26 Logothetis, N.K. et al. (2001) Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150–157 27 Logothetis, N.K. (2008) What we can do and what we cannot do with fMRI. Nature 453, 869–878 28 Wise, R.G. et al. (2004) Using fMRI to quantify the time dependence of remifentanil analgesia in the human brain. Neuropsychopharmacology 29, 626–635 29 Fox, M.D. and Raichle, M.E. (2007) Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat. Rev. Neurosci. 8, 700–711 30 Biswal, B. et al. (1995) Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn. Reson. Med. 34, 537–541 31 Beckmann, C.F. et al. (2005) Investigations into resting-state connectivity using independent component analysis. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 360, 1001–1013 32 Raichle, M.E. et al. (2001) A default mode of brain function. Proc. Natl. Acad. Sci. U. S. A. 98, 676–682 33 Rack-Gomer, A.L. et al. (2009) Caffeine reduces resting-state BOLD functional connectivity in the motor cortex. Neuroimage 46, 56–63 34 Pattinson, K.T. et al. (2007) Pharmacological FMRI: measuring opioid effects on the BOLD response to hypercapnia. J. Cereb. Blood Flow Metab. 27, 414–423 35 Pattinson, K.T. et al. (2009) Opioids depress cortical centers responsible for the volitional control of respiration. J. Neurosci. 29, 8177–8186 36 Harvey, A.K. et al. (2008) Brainstem functional magnetic resonance imaging: disentangling signal from physiological noise. J. Magn. Reson. Imaging 28, 1337–1344 37 Rosenkranz, K. and Lemieux, L. Present and future of simultaneous EEG-fMRI. Magma (in press), doi:10.1007/s10334-009-0196-9 38 Tracey, I. and Mantyh, P.W. (2007) The cerebral signature for pain perception and its modulation. Neuron 55, 377–391 39 Zambreanu, L. et al. (2005) A role for the brainstem in central sensitisation in humans. Evidence from functional magnetic resonance imaging. Pain 114, 397–407 40 Borsook, D. et al. (2006) A role for fMRI in optimizing CNS drug development. Nat. Rev. Drug Discov. 5, 411–424 41 (2001) Biomarkers Definitions Working Group. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin. Pharmacol. Ther. 69, 89–95 42 Lesko, L.J. and Atkinson, A.J., Jr (2001) Use of biomarkers and surrogate endpoints in drug development and regulatory decision making: criteria, validation, strategies. Annu. Rev. Pharmacol. Toxicol. 41, 347–366 43 Minzenberg, M.J. and Carter, C.S. (2007) The quest for developing new treatments from imaging techniques: promises, problems and future potential. Expert Opin. Drug Discov. 2, 1029–1033 44 Schweinhardt, P. et al. (2006) Pharmacological FMRI in the development of new analgesic compounds. NMR Biomed. 19, 702–711 45 O’Connor, J.P. et al. (2007) DCE-MRI biomarkers in the clinical evaluation of antiangiogenic and vascular disrupting agents. Br. J. Cancer 96, 189–195 46 Borsook, D. et al. (2002) Utilizing brain imaging for analgesic drug development. Curr. Opin. Investig. Drugs 3, 1342–1347 47 Woolgar, A. et al. Multi-voxel coding of stimuli, rules, and responses in human frontoparietal cortex. Neuroimage (in press), doi:10.1016/j.neuroimage.201004.035 48 Mitsis, G.D. et al. (2008) Regions of interest analysis in pharmacological fMRI: how do the definition criteria influence the inferred result? Neuroimage 40, 121–132 49 de Visser, S.J. et al. (2003) Biomarkers for the effects of benzodiazepines in healthy volunteers. Br. J. Clin. Pharmacol. 55, 39–50 50 Clemen, R.T. (1996) Making Hard Decisions: An Introduction to Decision Analysis. Duxbury Press 51 Bogdan, B. and Villiger, R. (2008) Valuation in Life Sciences. Springer-Verlag
DIARY/CLASSIFIED
Drug Discovery Today • Volume 15, Numbers 21/22 • November 2010
diary Each month we publish brief details of forthcoming meetings. The diary is updated regularly. If you are organizing a future meeting, please help us to maintain an efficient service by sending the appropriate details to the Editor well in advance. Entries new to this issue are marked with an asterisk *. 1–2 December 2010 Pharmaceutical Cold Chain Distribution Marriott Regents Park, London, UK (URL: http://www.smi-online.co.uk/ 2010coldchain21.asp/) 6–8 December 2010 BioPharma India Convention 2010 Grand Hyatt, Mumbai, India (URL: http://www.terrapinn.com/2010/ biopharmaindia/) 6–8 December 2010
6–8 December 2010
23–24 February 2011
28–31 March 2011
Pharma Trials World India 2010 Grand Hyatt, Mumbai, India (URL: http://www.terrapinn.com/2010/ pharmatrialsind/)
Oncology Biomarkers Congress Hilton Deansgate, Manchester, UK (URL: http://www.biomarkers-congress. com/)
4th Annual Pharma Partnering World Asia 2011 Marina Bay Sands, Singapore (URL: http://www.terrapinn.com/2011/ partneringasia/)
7–9 December 2010
28 February–1 March 2011*
1st Annual SBS China Conference – Innovation in Drug Discovery: Science & Technology Pudong Shangri-La Hotel, Shanghai, China (URL: http://www.sbsonline.org/i4a/pages/ index.cfm?pageid=3925/)
Bio/Pharmaceutical Cold Chain China Venue to be confirmed, China (URL: http://www.pharmacoldchainchina. com/)
10–12 January 2011 Biotech Showcase 2011 Parc 55 Wyndham-Union Square, San Francisco, CA, USA (URL: http://www.ebdgroup.com/bts/ index.php/) 10–14 January 2011* PepTalk Hotel del Coronado, San Diego, CA, USA. (URL: http://www.chi-peptalk.com/) 19–20 January 2011 BioBusiness 2011 Radisson Blu Portman Hotel, London, UK (URL: http://www.biobusiness-conference. com/) 7–8 February 2011 Pharma R&D World Asia Congress Singapore (URL: http://www.pharmaworldasiacongress.com/)
Biologic Manufacturing World India 2010 Grand Hyatt, Mumbai, India (URL: http://www.terrapinn.com/2010/ biologicindia/)
7–10 February 2011
6–8 December 2010
23–24 February 2011
Drug Discovery World India 2010 Grand Hyatt, Mumbai, India (URL: http://www.terrapinn.com/2010/ drugindia/)
6th Annual Biomarkers Congress Hilton Deansgate, Manchester, UK (URL: http://www.biomarkers-congress. com/)
3rd International Conference on Drug Discovery and Therapy Dubai, UAE (URL: http://www.icddt.com/)
7–10 March 2011* The Traumatic Brain Injury & Neurorehabilitation Conference Washington, DC, USA (URL: http://www.tbiconference.com/ home/) 16–17 March 2011* Innovations in Healthcare Management and Informatics 2011 Kuala Lumpur, Malaysia (URL: http://www.healthcareinformaticsasia. com/) 28–31 March 2011 4th Annual Biologic Manufacturing World Asia 2011 Marina Bay Sands, Singapore (URL: http://www.terrapinn.com/2011/ biologicasia/) 28–31 March 2011 4th Annual Drug Discovery World Asia 2011 Marina Bay Sands, Singapore (URL: http://www.terrapinn.com/2011/ drugdiscovery/) 28–31 March 2011
4–5 April 2011 4th Annual Proteins Congress Copthorne Tara Hotel, London, UK (URL: http://www.proteins-congress.com/) 4–5 April 2011 Vaccines Congress Copthorne Tara Hotel, London, UK (URL: http://www.vaccines-congress.com/) 11–13 April 2011* Personalized Medicine Partnerships Conference Washington, DC, USA (URL: http://personalizedmedicinepartnerships.c om/) 18–20 April 2011 Pharma-Nutrition 2011 Amsterdam, The Netherlands (URL: http://www.pharma-nutrition.com/) 8–9 June 2011 12th Annual Drug Discovery Leaders Summit Montreux Palace, Switzerland (URL: http://www.drugdiscovery-summit. com/) 8–9 June 2011 2nd Annual Drug Development Congress Montreux Palace, Switzerland (URL: http://www.drugdevelopmentsummit.com/)
4th Annual BioPharma Asia Convention 2011 Marina Bay Sands, Singapore (URL: http://www.terrapinn.com/2011/ biopharmaasia/)
www.drugdiscoverytoday.com
981
Elsevier Supporting and informing our community of
OVER 500,000 REVIEWERS • 30-day access to Scopus and ScienceDirect • Reference Linking - enabling reviewers to have direct access to referenced articles from over 20,000 journals • Advice and practical information from Elsevier journal editors • Online Reviewers' Information Pack • Online support 24/7 via the Elsevier Editorial System (EES) • Live chat service from a dedicated Customer Support Team
www.elsevier.com/reviewers
Watch your language Elsevier Language Editing Services offer high-quality English language editing for researchers preparing articles for publication in international scientific and medical journals. Elsevier’s editorial expertise and long history of publishing peer-reviewed journals will ensure that the English language in your manuscript is clear and free of distracting errors before you submit.
visit us at languageediting.elsevier.com Native English speakers from the top universities Expert input from Ph.D.s or Ph.D. candidates matched to your ¿eld of study
Customer comments
Journal Editors’ comments
“...I am very grateful to you for helping me. The manuscript which you revised for me is excellent. Thanks a lot!” PhD. Qingyu Guan
“...my current view is that the author should take as many steps as possible to make their article right before it is submitted...” “Poor English is a worrying problem, especially when the scienti¿c work is novel and deserves publication.” “... authors have to be responsible for ensuring at least reasonable language.”
CONFERENCE
“Trends in Medicinal Chemistry” 9th December 2010 National Heart & Lung Institute, Kensington, London
PROGRAMME OF SPEAKERS: ITC data in compound development John E. Ladbury, Anderson Cancer Center Biophysical Techniques in Drug Design - A Medicinal Chemistry Perspective David Millan, Pfizer MetAP-2 review & rationale Nigel Parr, GSK Gyrase AZ Ann Eakin, AstraZeneca Thermofluor Matthew Todd, Janssen Pharmaceuticals Selective GlyT1 Inhibitors: Discovery of RG1678, a Promising Novel Medicine to Treat Schizophrenia Emmanuel Pinard, Roche Alpha 7 positive modulators Veronique Birault, GSK Student bursaries are available Please see our web site or contact the Secretariat. To view meeting abstracts and to register on-line please visit:
www.smr.org.uk Or alternatively, contact the SMR Secretariat at: 840 Melton Road, Thurmaston, Leicester, LE4 8BN Tel: 0116 2691048 Fax: 0116 2640141 Email: [email protected]
