Approaches to assessing drug safety in the discovery phase: highlights of the society for medicines research symposium

Copyright 2010 Prous Science, S.A.U. or its licensors. All rights reserved.
APPROACHES TO ASSESSING DRUG SAFETY INTHE DISCOVERY PHASE HIGHLIGHTS OF THE SOCIETY FOR MEDICINESRESEARCH SYMPOSIUM HELD ON SEPTEMBER 24TH, 2009, AT THE NATIONAL HEART &LUNG INSTITUTE, KENSINGTON, LONDON, U.K.
J. Allen1, P. Jeffrey2, R. Williams3 and A.J. Ratcliffe4 1AstraZeneca, Alderley Park, Macclesfield, SK10 4TG, UK; 2GlaxoSmithKline, Immunoinflammation Centre of Excellence for Drug Discovery,Gunnels Wood Road, Stevenage, SG1 2NY, UK; 3Cancer Research UK, P.O. Box 123, WC2A 3PX, UK; 4Cellzome Ltd., Chesterford ResearchPark, Little Chesterford, Cambridge CB10 1XL, UK Professor Kevin Park (MRC Centre for Drug Safety Science,Department of Pharmacology, University of Liverpool, U.K.) deliv- The Society for Medicines Research symposium, sponsored by ered the opening lecture on molecular aspects of adverse drug reac- Apredica, Gwathmey Preclinical Services, Gentronix and Cyprotex, was tions from molecule to man. Professor Park suggested understand- held at the National Heart and Lung Institute, Kensington, London, ing adverse drug reactions at the molecular level, and linking them U.K. The meeting, organized by Jack Allen, Phil Jeffrey and Andrew to pharmacogenomics, offered a way of tailoring the medicine to the Ratcliffe, focused on approaches to assessing drug safety in the discov- individual patient to minimize unwanted adverse toxicity. To illus- ery phase. Topics included molecular aspects of adverse drug reactions trate the approach, warfarin therapy was given as an example.
from molecule to man, the relationship of physicochemical properties Warfarin produces anticoagulation by reducing the binding of coag- to toxicity, reactive metabolites, inhibition of the cardiac sodium chan- ulation factors to the vascular endothelium. At the mechanistic level, nel, drug-induced mitochondrial dysfunction, high-throughput screen- this is driven through inhibition of vitamin K epoxide reductase com- ing for genotoxicity and carcinogenicity, and the use of zebrafish as a plex (VKORC1), an enzyme that is a key operator in the post-transla- model for hepatotoxicity and developmental toxicity. tional γ-carboxylation of glutamic acid residues on the coagulation Attrition in drug development is still cripplingly high, with toxicity factors required for endothelial binding. Warfarin is administered as the leading cause at all stages in the drug development pipeline. It a racemate, with the more potent (S)-isomer undergoing metabo- has been estimated that a 10% improvement in predicting failure lism principally by cytochrome P450 CYP2C9. However, there are before the initiation of expensive and time-consuming clinical trials certain groups of patients which represent challenges to the dosing could save upwards of $100 million in the costs associated with drug and management of warfarin therapy (1-3). Patients with a common development. Furthermore, since approximately 70% of all toxicity- functionally defective CYP2C9, resulting in an inability to efficiently related failures that occur preclinically are comprised of toxicologi- eliminate (S)-warfarin from the systemic circulation through metab- cal outcomes for which the preclinical models are predictive of olism, run a higher risk of life-threatening bleeding. To compensate, human toxicity, then the benefits of identifying and predicting safe- a significantly lower maintenance dose regimen is required. There ty liabilities earlier in the drug discovery and development process are also patients with VKORC1 polymorphisms, which can result in could be of enormous benefit and value.
either warfarin sensitivity, as in the case of CYP2C9 polymorphism,or warfarin resistance, in which patients are stratified to higher doserequirements. As a consequence of understanding in detail themechanism of a drug’s action, and associated genetic and environ- Correspondence: Andrew J. Ratcliffe, Cellzome Ltd., Chesterford Research Park, Little
Chesterford, Cambridge CB10 1XL, UK. E-mail: andrew.ratcliffe@cellzome.com.
mental factors that avoid adverse drug reactions, new regulatory APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE labeling can be approved and subsequently used by physicians to bioactive thiol is further compromised, leading to nonresponsive- deliver improved healthcare. In the case of warfarin, new labeling ness and a higher rate of adverse clinical events. In an acute coro- was approved in 2007, in which genetic variations of CYP2C9 and nary setting this can take the form of death from cardiovascular VKORC1, along with other factors, should be considered when pre- causes, myocardial infarction or stroke. In comparison to clopidogrel, scribing the drug, although such tests should not delay initiating the onset of action of prasugrel is rapid and more sustained, and it is less prone to adverse events involving CYP2C19 polymorphism,given a lack of involvement of this CYP in its bioactive thiol genera- In another example, the importance of mechanistically understand- ing at the human level the conversion of prodrugs to their activeform, and placing in context the risk of adverse drug reactions, was The hepatocyte is well tuned to sensing chemical stress caused by demonstrated in comparing and contrasting the clinically approved exposure to reactive metabolites. With respect to the molecular receptor antagonists clopidogrel (1) and prasugrel (2) (4, 5).
mechanisms that come into play, a huge amount of fundamental Although both drugs require biotransformation to generate active information has been mapped out from a detailed understanding of thiols that subsequently bind irreversibly to cystine residues in the events caused by dosing of paracetamol (6). As part of the disposi- P2Y ADP receptor on platelets, the pathways to such species differ tion of paracetamol, a small proportion is converted to the reactive metabolite N-acetyl-p-benzoquinoneimine (NAPQI), which isquenched by hepatic glutathione (GSH) or cysteine residues within Through clever design, the ester group in prasugrel is hydrolyzed KEAP1, a cytosolic protein that binds to the redox-sensitive tran- using intestinal hydroxy esterases to the thiolactone (3), which is
scription factor Nrf2. Under normal physiological conditions, the role subsequently metabolized primarily by CYP3A4 to its active thiol (4).
of KEAP1 is to maintain Nrf2 transcription in check through a protea- In contrast, clopidogrel is transformed to its active thiol (5) via an
some-dependent degradation of the protein. This process is disrupt- intermediate thiolactone (6) solely through CYP1A2-, CYP3A4-,
ed by the covalent binding of NAPQI, with liberation of Nrf2 and sub- CYP2C9- and CYP2C19-dependent transformations. sequent translocation to the nucleus, where it orchestrates an The drug disposition of clopidogrel is further complicated by the fact antioxidant response through the activation of cell defense genes, that only approximately 15% of the drug is metabolized by the CYP such as glutathione transferases, NAD(P)H quinine oxidoreductase, pathway to its bioactive thiol form, with the majority suffering heme oxygenase and glucuronyltransferase. Unfortunately, at a high hydrolysis to an inactive acid derivative (7). As a result, patients often
(over) dose of paracetamol the extent of NAPQI formation seriously receive a relatively low exposure of the active form of the drug, which depletes hepatic GSH levels and overwhelms the antioxidant line of can lead by its very nature to a variable onset of action and response.
defense, resulting in the covalent modification of critical proteins, This is further heightened in patients carrying a CYP2C19 reduced such as γ-glutamylcysteine synthetase, glyceraldehyde-3-phos- functional allele, in which their ability to metabolically generate the phate dehydrogenase (GAPDH) and Ca2+/Mg2+-ATPase, which are Figure 1. Biotransformation of clopidogrel (1) and prasugrel (2) to active thiols.
THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE key in maintaining mitochondrial function and Ca2+ homeostasis.
To fill this void, GSK initiated an internal program aimed at defining The impact of such an insult on the cell can lead ultimately to cell certain guidelines. In addition, it was hoped that such an output death, either by apoptosis or necrosis. Of great interest is that both could help address the high proportion of compounds failing to forms of cell death release specific signature proteins that can be progress within the organization due to unacceptable hepatoxicity.
monitored in plasma, namely high-mobility group box protein 1 To this end, a database was established consisting of 200 hepato- (HMG-1) from necrotic cells and fragments of cytokeratin-18 (CK-18) toxic and 200 nonhepatotoxic known drugs. Criteria for entry into from apoptotic cells. From animal studies, increased doses of para- the hepatotoxic class of drug consisted of at least 50 reports of cetamol, up to 550 mg/kg, have been correlated with increased hepatotoxicity or at least 3 reports of life-threatening hepatotoxicity, plasma levels of these signature proteins, although each one has a warnings or precautions in label, including liver enzyme monitoring.
unique dose–response profile. Above doses of 550 mg/kg and up to In contrast, no reports of hepatotoxicity or only one report of “mild” 1000 mg/kg, the levels of CK-18 fragments captured dramatically hepatoxicity constituted the nonhepatotoxic class of drugs.
fall, while in contrast, levels of HMG-1 continue to rise. From the Initially, two methodologies, often used in lead optimization, were unraveling of the molecular mechanisms of paracetamol toxicity, used to assess the propensity for reactive metabolite formation. The HMG-1 and CK-18 fragments appear as potential biomarkers for cel- first centered on GSH trapping, where production of GSH conjugates lular necrosis and apoptosis, respectively. Such relationships have from microsomal incubations was monitored via loss of pyroglutam- found utility in the diagnosis and severity grading of a number of dis- ic acid (m/z 129). Using a signal:noise ratio approach, the extent of GSH conjugative production could be quantified (ratios: minor 1-10; Hypersensitivity is a common adverse event with many drugs that notable 10-100; and marked > 100). The second assay focused on can limit their therapeutic use; examples include carbamazepine, CYP time-dependent inhibition (TDI). The CYPs employed covered sulfamethoxazole, abacavir and nevirapine. There is a growing body 1A2, 2C9, 2C19, 2D6, 3A4DEF and 3A47BQ, and the extent of the of evidence that reactive drug metabolites may play a key role in elic- fold change with time allowed a ranking distribution of the iting the allergic reaction, which is believed to center on activation of drugs into subsets (fold shift: minor < 2; notable 2-5; and marked > the T-lymphocyte system (9). Although a lymphocyte transformation 5). Results from the GSH trapping screen revealed that 80% of GSH test (10) is available as an in vitro test to assess the propensity of a adducts classified as marked adhered to the hepatotoxic class of drug or its metabolite to cause hypersensitivity, further research has drugs. In the case of marked TDI, this figure dropped to 70%.
revealed a genetic basis that may offer pharmacogenetic screening However, common to both screens was a cluster of drugs displaying as a means of de-risking patient drug hypersensitivity. In several marked GSH/TDI effects but not exhibiting hepatotoxicity. Further clinical studies of abacavir, a nucleoside reverse transcriptase investigation of this cluster revealed that many of the drugs were inhibitor designed to combat HIV, avoiding inclusion of patients with administered as a low dose (< 10 mg) or intermittently or topically the major histocompatibility complex (MHC) class I allele HLA- applied. Revisiting the analysis and setting the drug dose threshold B*5701 delivered a significantly reduced diagnosis of hypersensitivi- to > 100 mg/day delivered an improved discrimination. Of those ty (11). Such results represent a potentially important step in the clin- drugs that exhibited hepatotoxicity in man, 96% and 82%, respec- ical management of abacavir, given that patients presenting tively, gave marked GSH adducts and TDI effects.
hypersensitivity manifest conditions of fever, rash, gastrointestinal The availability of radiolabeled drugs allows a further assessment of and respiratory symptoms that can become more severe, rapid and the extent of drug protein adducts and the risk of toxicity, in partic- life-threatening if discontinuation of the drug is not immediately ular idiosyncratic toxicity (IDT), through bioactivation mechanisms.
instigated. The HLA-B*5701 genotype has also been advocated as a Covalent binding measurements can be performed in liver microso- high risk factor for the drug-induced liver injury caused by flu- mal or hepatocyte preparations from both animals and humans. In cloxacillin, and prescreening of patients, as in the case of abacavir, the case of animals, the level of adducts to both liver and plasma may provide enhanced patient compliance (12). proteins can be determined in the intact animal after oral dosing.
Mr. Andy Harrell (GSK, U.K.) presented a current industry perspective Seminal work by the Merck group (13) suggested 50 pmol drug on reactive metabolites. Since their identification in the 1930s, there eq/mg protein as a target upper limit, above which a compound has been great interest in the area. During the last 30 years of the would not be advanced into development unless other qualifying 20th century, the association of reactive metabolites with toxicity and considerations were taken into account, such as a daily dose < 10 mg withdrawal of certain classes of drugs, such as β-lactam antibiotics or if the disease under treatment was life-threatening.
(anaphylaxis), nonsteroidal anti-inflammatory drugs (NSAIDs; idio- As part of the GSK predictive toxicology initiative, 65 radiolabeled syncratic hepatoxicity) and arylamines (carcinogenicity and agranulo- drugs of known safety profiles were screened in a microsomal acti- cytosis), began to emerge. Several hypotheses (hapten and critical vation assay, in which residual radioactivity associated with unex- protein) were proposed to support the link at a mechanistic level, in tracted protein was measured. Analysis of the results suggested 200 which a central tenet focused on covalent binding of the reactive pmol drug eq/mg of protein as an alert. In a second GSK study, metabolite to proteins or DNA. As a consequence, during the last 10 analysis of ex vivo covalent binding data, easily generated from low- years a multitude of nonradioactive reactive metabolite screens have dose toxicology studies, suggested that > 30% unrecovered radioac- become engrained within drug discovery in an attempt to identify tivity at a concentration of > 1 pmol represented a further risk alert.
development candidates where risk from reactive metabolite genera-tion and toxicity is minimized. However, during the early period of this Reference was also made to recently published work by researchers paradigm shift to reactive metabolite screening, little scientific data at Daiichi Sankyo (14), who screened 42 radiolabeled drugs of known existed in the public domain to support decision-making.
safety profiles for covalent binding in human microsomes, human THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE hepatocytes and rat liver in vivo. They also concluded that 50 pmol cology of the compound were removed from the analysis. Defined drug eq/mg of protein was ineffective in grouping the safety cate- lists of significant toxicology markers, generated from histopatholo- gories, with drugs given a warming of IDT in the Physician Desk gy, clinical pathology or clinical signs, were used to label the com- Reference and Japanese drug labeling not exceeding this threshold, pounds. Free drug exposure could be substituted for total drug and 4 of the 12 so-called safe drugs exhibiting covalent binding in exposure, providing plasma protein binding data were available.
human microsomes that exceeded 100 pmol drug eq/mg of protein.
Unfortunately, due to a lack of historical data, only 72% of the origi- Indeed, the use of covalent binding alone in each of the three test nal dataset met this criteria. However, by following a similar line of systems failed to distinguish the safety categories. Only when clini- data deconvolution, the appropriate free drug C cal daily dose was taken into account did a significant distribution determined to be 1 µM. Regardless of whether free or total drug Cmax plot become clear, with the separation of drugs into safe, equivocal was examined, analysis of the drug toxicity classification using a wide range of physicochemical properties and descriptors revealed aconsistent link with both TPSA and CLogP. Calculation of toxicity In concluding the talk, Mr. Harrell suggested there was a move away odds ratios revealed an inter-relationship between these properties, from blanket screening for reactive metabolites to a more bespoke and by setting cut-offs for CLogP and TPSA of 3 and 75 A2, respec- application of various screening models that could be applied at dif- tively, identified compounds with CLogP > 3 and TPSA < 75 A2 as ferent stages of a project. Risk assessment guidelines were now being six times more likely to have an impactful toxic outcome than available from such models, with several incorporating a clinical compounds with CLogP < 3 and TPSA > 75 A2, whether based on dose input that led to a higher level of predictivity. As an underlying total or free drug exposure. Compounds exhibiting only a single risk theme, increased dose appeared to be a clear driver for increased factor gave a weak and inconsistent trend. The addition of a further 2 years of new IVT data reinforced the high CLogP (> 3)/low TPSA Professor Julien Blagg (Cancer Research U.K. Centre for Cancer (< 75 A2) trend, with an increase in the odds ratio to 10-fold using Therapeutics) delivered a talk on the role of medicinal chemistry as a reference point. Substitution to a free drug design in the avoidance of toxicity at clinically effective exposures. In analysis delivered a further rise to 27-fold.
setting the scene, clinical attrition drivers over a 10-year period span- As a hypothesis, it was suggested that promiscuous binding to off- ning 1991-2000 were highlighted (15). In contrast to a dramatic target pharmacology was responsible for the increased incidence of decline in attrition through pharmacokinetic aspects, the most adverse outcomes associated with high CLogP/low TPSA space. To prominent cause in 1991 to a minor bit player in 2000, drug failure support the theory, CEREP data across the BioprintTM panel of 48 due to toxicology almost doubled to become a major factor entering assays of varied target class (G protein-coupled receptors [GPCRs], the 21st century. In addition a slight increase was also observed due enzymes, ion channels) on 108 compounds were analyzed, using to lack of clinical safety. However, Professor Blagg suggested pre- > 50% inhibition at 10 µM at three or more targets as a definition of clinical attrition rates due to safety and toxicity findings were likely to promiscuity. In line with the observations derived at a toxicity level, be significantly higher, given that such data often resided in confi- compounds of CLogP > 3 and TPSA < 75 A2 appeared 25 times more likely to have a significant off-target pharmacological profile than Although mechanistically adverse outcomes could be linked to pri- compounds of CLogP < 3 and TPSA > 75 A2. High lipophilicity has mary pharmacology, secondary pharmacology or the presence of a often been a design parameter used by the medicinal chemist in structural alert or toxicophore, meta-analysis of in vivo tolerance driving high potency against a target. Given its link to imparting an (IVT) studies conducted on a dataset comprised of 245 potential increased risk of toxicity and promiscuity, as well as high clearance, Pfizer drug candidates accumulated over a 5-year period support a Pfizer researchers have coined the term LipE, as defined by equation further origin in the form of physicochemical drug properties (16).
(1), as a lipophilicity efficiency measurement that can be used by the For each IVT study, the corresponding pharmacokinetic exposure medicinal chemist to ensure increases in potency and compounddesign are directed away from liphophilicity risk factors associated Cross-comparison of the chemical space property distribution map (MW, CLogP, TPSA) of the IVT compound dataset with that of a diverse subset of the Pfizer corporate file showed good overlap, In relation to idiosyncratic adverse events triggered by chemical thereby confirming that the set of compounds occupied sufficient structure, in particular through metabolic activation, Professor chemical diversity and the results of the meta-analysis hold general Blagg focused on reducing dose size as a means of preventing attri- application. Of the dataset, 50% were of a basic nature, 40% neutral tion through this mechanism. The closely related analogues cloza- pine (8) and olanzapine (9) served to illustrate the case (Fig. 2).
Clozapine, at a clinical dose of 300 mg/day, is known to form reac-
Key to the meta-analysis was the importance of data deconvolution, tive metabolites in vitro and cause a 1% incidence of argranulocyto- and at what specific exposure threshold should intrinsic toxicity be sis. In contrast to the restricted use of clozapine, olanzapine, dosed was used as the single parameter to reflect exposure, at 10 mg/day, shows no incidence of agranulocytosis, despite in vitro and 10 µM total drug was selected as a pragmatic threshold level studies confirming reactive metabolite formation. that delivered a distribution of compounds between toxic and clean,while at the same time minimizing the number of compounds clas- Furthermore, documented work has shown that drugs dosed at < 10 sified as uncertain. Data associated with compounds in which the mg/day were associated with a significantly lower incidence of idio- adverse in vivo outcome could correlate with the primary pharma- syncratic drug reactions (18). Notwithstanding imparting the desired THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE also profiled during candidate selection for cardiovascular effects invivo using telemetered dogs. Patch clamp assays are clearly more informative than binding assays, but as a manual assay throughputhas been limited. Newer, automated patch clamp robots (QPatch) are now available and have markedly increased compound through-put per FTE. A further emerging technique with potential utility to be deployed as part of a screening strategy is compound profiling in spontaneously beating cardiomyocytes derived from human embry- onic stem cells. Novartis investigates the effects of candidate com-pounds on cardiac action potentials in vivo using small animal mod- els. Rabbits and guinea pigs respond similarly to humans in response to sodium channel blockade, displaying prolonged P-R and QRS intervals. During preclinical development, Novartis con- ducts ECG profiling using telemetry in beagle dogs. Early sightingstudies are done using a noninvasive jacket telemetry system, with studies in chronically implanted animals being undertaken prior to Dr. Traebert briefly described a Novartis case study demonstratingthe utility of the aforementioned battery of assays to identify a com- Figure 2. Structures of clozapine (8) and olanzapine (9).
pound with clear sodium channel liabilities that was dropped fromdevelopment. In vivo in dogs the compound affected P-R and QRSintervals, mean P wave duration, and also induced ventricular tachy- intrinsic potency at the target, such a low projected clinical dose cardia. Tachycardia was also observed in isolated rabbit heart. In demands in conjunction superb pharmacokinetic and physicochem- vitro the compound was active in cells stably transfected with the ical properties, all of which remain under the remit and influence of SCN5A gene. Dr. Traebert concluded by stressing the importance of screening for sodium channel activity during compound develop- The afternoon session of the symposium was opened by Dr. Martin ment, adding that consideration of this area is underestimated com- Traebert (Head of Safety Pharmacology EU, Novartis Institute for pared to the investigation of compound activity against hERG chan- Biomedical Research) with a presentation entitled “Addressing car- nels. A variety of in vitro, ex vivo and in vivo assays are available, and diac sodium channel liabilities during preclinical drug develop- significant inhibition that translates into ECG alterations should be a ment”. Dr. Traebert began by describing how, historically, pharma- ceutical industry interest in cardiac sodium channels had centered The second speaker in the afternoon session was Dr. James Dykens on trying to develop inhibitors as antiarrhythmic agents. Encainide (Director of Investigative Cellular Toxicity, Pfizer, U.K.). Dr. Dykens and flecainide were two drugs progressed into clinical development began his presentation by highlighting the problem caused by and investigated in the Cardiac Arrythmia Suppression Trial (CAST).
adverse drug reactions, referencing data from the U.S. Over 2.2 mil- Unfortunately, however, these drugs were shown to cause a 3.6-fold lion adverse drug reactions occur annually in hospitalized patients increase in arrhythmic death and a 2.5-fold increase in overall mor- alone in the U.S., leading to approximately 106,000 deaths/year.
tality (19). Molecular biology studies have revealed that the human Adverse drug reactions are the fourth leading cause of death in the cardiac sodium channel contains four transmembrane repeats and U.S. Dr. Dykens view was that serious drug toxicity liabilities were is encoded by the gene SCN5A. The channel displays strictly volt- clearly not being identified, and proposed that off-target drug age-dependent activation, and is responsible for the depolarization effects on mitochondria were important in this respect. Many drugs phase and “upstroke” of the action potential. Inherited loss-of-func- withdrawn from the market or receiving black box warnings have tion mutations of SCN5A are associated with a range of chan- been shown to impair mitochondrial function. The pharmaceutical nelopathies, including congenital long Q-T syndrome, idiopathic industry is now starting to research this area more closely using ventricular fibrillation (Brugada syndrome), isolated cardiac conduc- tion disease, atrial standstill, congenital sick sinus syndrome, sud- Mitochondria are the “gatekeepers” of cell death. If mitochondria den infant death syndrome and dilated cardiomyopathy.
die, then so does the cell. These organelles have evolved from Currently, although there is now increased awareness of sodium ancient bacteria, and mitochondrial DNA represents the only non- channel liabilities by regulatory authorities, there are no formal nuclear genome in all animals. Mitochondria generate > 90% of cel- guidelines on this topic. Drug companies are, however, deploying a lular energy, and the magnitude of this activity is highlighted by the range of preclinical assays to identify compounds possessing sodi- fact that human males turn over 193 lbs of ATP/day, while females um channel liabilities. Novartis routinely uses a binding assay, with turn over approximately 148 lbs/day. Studies have shown that drugs rat brain as a sodium channel source, as a screen during lead selec- can interfere with mitochondrial electron transfer and ATP genera- tion. Moving towards candidate selection, compounds are tested in tion at many points. A Pfizer study of 550 drugs revealed that 34% repolarization assays using the patch clamp technique with HEK- of drugs displaying organ toxicity impaired mitochondrial function.
293 or CHO cells stably transfected with SCN5A. Compounds are Drugs impairing mitochondrial function will have an adverse effect, THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE although clinical manifestation is dependent on the bioenergetic multiple measures of in vitro toxicity are used to aid compound opti- capacity of an individual animal or patient. Bioenergetic capacity is mization in order to reduce the attrition rate in the preclinical and determined by genetics and also age. Indeed, older animals have been shown to be more susceptible to troglitazone-induced toxicitythan younger healthy animals.
Dr. Dykens went on to discuss further his experience at Pfizer using a range of assays to detect drugs possessing mitochondrial toxicity.
The first assay described was a 96-well plate format mitochondrial respiration assay (20, 21). This assay revealed potent inhibitory activ- –Nongenotoxic carcinogenicity Human and rat AhR, rat PPARα ity for a number of thiazolidinediones, and subsequent work with apioglitazone photoaffinity probe pulled down MitoNEET, an atypical Glu, Calcein, Alamar Blue, Hoechst Luxcel 2Fe-2S protein, as a molecular target for this drug. A further mito- chondrial functional assay utilizes “Seahorse technology”. This sys- tem measures oxygen consumption rate and extracellular acidifica-tion rate in microchambers. Phenformin and butformin, drugs 1A1, 1A2, 2A6, 2C8, 2C9, 2C19, 2D6, 2E1, 3A4 withdrawn for causing lactic acidosis, were shown to be active in this Throughput: 40-80 compounds per assay batch assay, while metformin was inactive (22). Mitochondrial impairmentby thiazolidinediones and statins was also demonstrated using anassay format that measures the activity of individual oxidative phos- For fast genotoxicity screening, a panel of six screens is used, all of phorylation complexes following immunocapture (23).
which depend on sensitive luminescent luciferase-based assays. A Contemporary cell culture conditions typically contain glucose con- bacterial screen is performed in Salmonella, where the luciferase centrations five times the physiological levels. Elevated glucose expression is activated via a cascade of reactions known as the SOS inhibits mitochondrial respiration (Crabtree effect), and consequent- system, which is employed in the “Vitotox” assay to detect the geno- ly, drugs possessing mitochondrial toxicity will not be detected toxicity, cytotoxicity and mutagenic potency of the sample (Fig. 3).
under such conditions. However, cells grown in galactose do not The major benefit of the assay is that the entire DNA content of the generate ATP through glycolysis and become susceptible to drugs cell functions as a target for the genotoxin to display its effect, and inhibiting mitochondrial function (24). In terms of useful in vivo can be considered as a substitute for the full Ames mutagencity test.
models, heterozygous superoxide dismutase 2 (SOD+/–) mice display With respect to clastogenicity, a “RadarScreen” (reMYND, Leuven, clinically silent mitochondrial dysfunction, and in contrast to normal Belgium) assay is used to replace the sister chromatid exchange mice are susceptible to the hepatotoxic effects of troglitazone (25).
(SCE), chromosomal aberration (CA) assays and micronuclei tests, Dr. Dykens concluded his talk by reiterating his view that a key con- which are time-consuming and have low compound throughput.
tributor to idiosyncratic drug responses is mitochondrial dysfunction.
The assay is based on activation of the RAD54 promoter linked to a Individual susceptibility is determined by an individual’s “bioener- β-galactosidase reporter gene in yeast, as shown below (Fig. 4). getic threshold”. Pfizer has moved assessment of mitochondrial tox-icity to the lead selection phase of discovery, where there are poten- Good sensitivity, selectivity and predictability for the clastogenicity tially diverse hit series and the greatest opportunity to “de-risk” for test are observed, although prediction for mutagenicity is relatively hepatotoxicity, nephrotoxicity and neurotoxicity. low, which is the opposite for the Vitotox assay. More mechanisticassays for genotoxicity in Hep G2 cells were described, in which four The next speaker was Dr. Willem Schoonen (Department of Toxicol-ogy and Drug Disposition, Schering-Plough, the Netherlands), who promoters of the luciferase assay were used. An advantage of these presented his company’s approach to “High-Throughput Screening cells is their ability to metabolize/activate certain drugs such as for Toxicity Testing” in early drug discovery. Despite the efforts by the benzo[a]pyrene, aflaxtoxin B1 and etoposide, without the need for pharmaceutical industry over the last 15 years, some 50% of new the S9 metabolic activating system (28-30). Although the cells do chemical entities (NCEs) still fail due to toxicity, which represents a not have CYP2C9 activity, it can be induced. major cost to industry (15), particularly in the areas of hepatotoxici- All of these assays were validated against a large number of geno- ty, cardiotoxicity, skin toxicity, CNS side effects, genotoxicity and car- toxic (and nongenotoxic) compounds (190) with diverse mecha- cinogenicity. The preclinical cost is even more significant, as safety nisms, including direct-acting genotoxins, topoisomerase inhibitors, screening is often the final hurdle in drug discovery before the NCE nucleotide/DNA synthesis inhibitors, reactive oxygen species gener- enters the clinic, and few effective strategies for avoiding toxicity ators and aneugenes (change in number of chromosomes). In gen- exist to guide medicinal chemistry programs before then (26, 27).
eral, the sensitivity, specificity and predictivity of the assays are Dr. Schoonen described the assays Schering-Plough uses to assess acceptable, with a total of 110 compounds having a positive Ames or DNA and membrane damage, cell-, organ- or organelle-specific tox- clastogenicity score (31). It was concluded that Vitotox prediction is icity, nuclear receptor activation and CYP enzymes and induction, relatively high, RadarScreen is very good and prediction of all four which are used to select/deselect and/or rank compounds in lead different Hep G2 assays is relatively low in comparison with optimization. What assay in which phase depends on sample Salmonella and yeast genotoxicity assays. Nevertheless, in vitro throughput, the amount of compound available and the effort high-throughput screening might be more valuable for the predic- required in conducting the assay. Multiple leads are screened and tion of human genotoxicity (32, 33).
THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE RecA cleavage of LexA repressor Figure 3. Vitotox assay principle.
tool for the last 40 years, more recently they have been used as an in vivo vertebrate model to support early drug discovery for screening NCEs. The small size of the larvae makes them amenable to high- throughput screening using very small quantities of compound.
Their metabolism, physiology and development are apparently com- parable to humans, and since they are transparent, multiple end- points can be visualized without dissection. Dr. Hill went on to describe several assays that have been developed in conjunction with a number of pharmaceutical companies, as detailed below Uptake into zebrafish can vary considerably from compound to com- pound, and no one physicochemical property can adequately predict compound uptake through the skin. A generic protocol is therefore employed to measure the uptake of compounds using conventional Figure 4. RadarScreen assay principle.
bioanalytical methods, which helps to identify false negatives and to correlate the extent of exposure with toxicity.
The last speaker of the day was Dr. Adrian Hill (Evotec, U.K.), who Data were presented on a cardiac functional assay, where it was con- gave a presentation on the use of zebrafish screening in early drug cluded that the cardiophysiological response of zebrafish was discovery, with reference in particular to predicting hepatotoxicity, predictive of the human cardiovascular response. Some 64 com- cardiotoxicity and developmental toxicity (embryotoxicity and ter- pounds were used in the validation, a summary of which is present- atogenicity). Although zebrafish have been used as an experimental Table I. Multifunctional uses of zebrafish in drug discovery.
Developmental toxicity (embryotoxicity & teratogenicity) THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE Table II. Zebrafish endpoints associated with teratogenicicity.
–Atrioventricular decoupling: 31 of 36 (86%) Two causing no cardiac effect (erythromycin, sotalol) – poor uptake Three causing bradycardia (mibefradil, nortriptyline, ranolazine) One causing atrioventricular decoupling (diltiazem) –Overall predictivity = 91% (excluding bioanalysis data) Information was also presented from a Novartis (blinded) study using 20 compounds, in which bradycardia was correctly predictedfor 5 of 5 compounds. In terms of Q-T-prolonging drugs: 91% werecorrectly identified by atrioventricular decoupling (reducing to 71%due to false negatives with poor uptake). Excluding unclassified DISCLOSURES
compounds, the overall predictivity amounted to 94%.
J. Allen, P. Jeffrey and A. Ratcliffe are in the paid employ of their In terms of developmental toxicity (embryotoxicity and teratogenici- respective companies. All authors are SMR Committee members for ty), a number of institutions have used a zebrafish assay employing different protocols and strains/ages of larvae. Endpoints associatedwith teratogenicity are detailed in Table II. Overall predictivity REFERENCES
1. Rieder, M.J., Reiner, A.P., Gage, B.F. et al. Effect of VKORC1 haplotypes on Finally, in terms of hepatotoxicity, a zebrafish assay was described transcriptional regulation and warfarin dose. N Engl J Med 2005, 352(22): using a number of endpoints, including liver abnormalities, changes in size and shape of the liver (hepatomegaly), effects on yolk absorp- 2. Jorgensen, A.L., Al-Zubiedi, S., Zhang, J.E. et al. Genetic and environmen- tion (yolk retention), lethality, gastrointestinal toxicity and loss of tal factors determining clinical outcomes and cost of warfarin therapy: A bile. The assay was compared to a novel cell-based model using a prospective study. Pharmacogenet Genomics 2009, 19(10): 800-12.
high-content screening (HCS) model (34). In addition to 14 well- 3. Limdi, N.A., Veenstra, D.L. Warfarin pharmacogentics. Pharmacotherapy known toxic compounds, a subset of 36 compounds from the HCS assay was selected for screening including: 4. Norgard, N.B., Abu-Fadel, M. Comparison of prasugrel and clopidogrel in patients with acute coronary syndrome undergoing percutaneous coronaryintervention. Vasc Health Risk Manag 2009, 5: 873-82.
5. Motovska, Z., Widimsky, P. Improving outcomes in patients undergoing –Human hepatotoxic compounds that gave a false negative in the HCS percutaneous coronary intervention: Role of prasugrel. Vasc Health Risk (stavudine, novobiocin, bupropion, diethylcarbamazine, clofibrate, 6. Park, K.B., Dalton-Brown, E., Hirst, C., Williams, D.P. Selection of new –Toxic compounds to animals that gave a false negative in the HCS (rido- chemical entities with decreased potential for adverse drug reactions. Eur J –Toxic compounds to humans but NOT liver toxic that gave a false posi- 7. Kocsis, Á.K., Szabolcs, A., Hofner, P., Takács, T., Farkas, G., Boda, K., tive in the HCS (metformin [pancreas], pamidronate [kidney], astemi- Mándi, Y. Plasma concentrations of high-mobility group box protein 1, sol- zole [acute], temozolomide [bone marrow], gentamicin) uble receptor for advanced glycation end products and circulating DNA inpatients with acute pancreatitis. Pancreatology 2009, 9(4): 383-91.
–Nontoxic compounds: false positive in the HCS (picotamide) 8. Luft, T., Conzelmann, M., Benner, A. et al. Serum cytokeratin-18 fragments as quantitative markers of epithelial apoptosis in liver and intestinal graft-versus-host disease. Blood 2007, 110(13): 4535-42.
The HCS resulted in 6 of 9 false positives (gentamicin, metformin, 9. Park, K., Williams, D.P., Naisbitt, D.J., Kitteringham, N.R., Pirmohamed, astemizole, temozolomide, picotamide, pamidronate) and 9 of 27 M. Investigation of toxic metabolites during drug development. Toxicol false negatives (clofibrate, stavudine, novobiocin, bupropion, Appl Pharmacol 2005, 207(2, Suppl.): 425-34.
diethylcarbamazine, pravastatin, ridogrel, acaftadine, valproate).
10. Merk, H.F. Diagnosis of drug hypersensitivity: Lymphocyte transformation Overall, the HCS assay sensitivity was 67%, specificity 33% and pre- test and cytokines. Toxicology 2005, 209(2): 217-20.
dictivity 58%. In comparison, the zebrafish assay gave 3 of 50 false 11. Mallal, S., Phillips, E., Carosi, G. et al. HLA-B*5701 screening for hypersen- positives (gentamicin, praziquantel, astemizole) and 5 of 50 false sitivity to abacavir. N Engl J Med 2008, 358(6): 568-79.
negatives (oxyphenisatin, valproate, ketoconazole, novobiocin, 12. Daly, A.K., Donaldson, P.T., Bhatnagar, P. et al. HLA-B*5701 genotype is a stavudine), leading to an increased sensitivity of 86%, specificity of major determinant of drug induced liver injury due to flucloxacillin. Nat 77% and overall predictivity of 84%.
THOMSON REUTERS – Drugs of the Future 2010, 35(1) APPROACHES TO ASSESSING DRUG SAFETY IN THE DISCOVERY PHASE 13. Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A. Drug-protein 24. Marroquin, L.D., Hynes, J., Dykens, J.A., Jamieson, J.D., Will, Y.
adducts: An industry perspective on minimizing the potential for drug Circumventing the Crabtree effect: Replacing media glucose with galactose bioactivation in drug discovery and development. Chem Res Toxicol 2004, increases susceptibility of HepG2 cells to mitochondrial toxicants. Toxicol 14. Nakayama, S., Atsumi, R., Takakusa, H. et al. A zone classification system 25. Ong, M.M.K., Calivarathan, L., Boelsterli, U.A. Troglitazone induced for risk assessment of idiosyncratic drug toxicity using daily dose and cova- hepatic necrosis in an animal model of silent genetic mitochondrial lent binding. Drug Metab Dispos 2009, 37(9): 1970-7.
abnormalities. Toxicol Sci 2007, 97(1): 205-13.
15. Kola, I., Landis, J. Can the pharmaceutical industry reduce attrition rates? 26. Kramer, J.A., Sagartz, J.E., Morris, D.L. The application of discovery toxi- Nat Rev Drug Discov 2004, 3(8): 711-6.
cology and pathology towards the design of safer pharmaceutical lead candidates. Nat Rev Drug Discov 2007, 6(8): 636-49.
16. Hughes, J.D., Blagg, J., Price, D.A. et al. Physiochemical drug properties associated with in vivo toxicological outcomes. Bioorg Med Chem Lett 27. Stevens, J.L. Future of toxicology - Mechanisms of toxicity and drug safe- ty:where do we go from here. Chem Res Toxicol 2006, 19(11): 1393-401.
28. Westerink, W.M.A., Schoonen, W.G.E.J. Cytochrome P450 enzyme levels 17. Ryckmans, T., Edwards, M.P., Horne, V.A. et al. Rapid assessment of a in HepG2 cells and cryopreserved primary human hepatocytes and their novel series of selective CB2 agonists using parallel synthesis protocols: A induction in HepG2 cells. Toxicol In Vitro 2007, 21(8): 1581-91.
lipophilic efficiency (LipE) analysis. Bioorg Med Chem Lett 2009, 19(15): 29. Westerink, W.M.A., Schoonen, W.G.E.J. Phase II enzyme levels in HepG2 cells and cryopreserved primary human hepatocytes and their induction in 18. Uetrecht, J. Prediction of a new drug’s potential to cause idiosyncratic reac- HepG2 cells. Toxicol In Vitro 2007, 21(8): 1592-602.
tions. Curr Opin Drug Discov Devel 2001, 4(1): 55-9.
30. Westerink, W.M.A., Stevenson, J.C.R., Schoonen, W.G.E.J. Pharmacologic 19. The Cardiac Arrhythmia Suppression Trial (CAST) Investigators.
profiling of human and rat cytochrome P450 1A1 and 1A2 induction and Preliminary report: Effect of encainide and flecainide on mortality in a ran- competition. Arch Toxicol 2008, 82(12): 909-21. dom trial of arrhythmia suppression after myocardial infarction. N Engl J 31. Westerink, W.M.A., Stevenson, J.C.R., Lauwers, A., Griffoen, G., Horbach, G.J., Schoonen, W.G.E.J. Evaluation of the Vitotox and RadarScreen assays 20. Dykens, J.A., Marroquin, L.D., Will, Y. Strategies to reduce late stage attri- for the rapid assessment of genotoxicity in the early research phase of drug tion due to mitochondrial toxicity. Expert Rev Mol Diagn 2007, 7(2): 161- development. Mutat Res 2009, 676(1-2): 113-30.
32. Knasmüller, S., Mersch-Sundermann, V., Kevekordes, S. et al. Use of 21. Will, Y., Hynes, J., Ogurtsov, V.I., Papkovsky, D.M. Analysis of mitochondr- human-derived liver cell lines for the detection of environmental and dietary ial function using phosphorescent oxygen sensitive probes. Nat Protoc genotoxicants; current state of knowledge. Toxicology 2004, 198(1-3): 315- 22. Dykens, J.A., Jamieson, J., Marroquin, L., Nadanaciva, S., Billis, P.A., Will, 33. Mersch-Sundermann, V., Knasmüller, S., Wu, X.-J., Darroudi, F., Kassie, F.
Y. Biguanide-induced mitochondrial dysfunction yields increased lactate Use of a human-derived liver cell line for the detection of cytoprotective production and cytotoxicity of aerobically poised HepG2 cells and human antigenotoxic and cogenotoxic agents. Toxicology 2004, 198(1-3): 329- hepatocytes in vitro. Toxicol Appl Pharmacol 2008, 233(2): 203-10.
23. Nadanaciva, S., Dykens, J.A., Bernal, A., Capaldi, R.A., Will, Y.
34. O'Brien, P.J., Irwin, W., Diaz, D. et al. High concordance of drug-induced Mitochondrial impairment by PPAR agonists and statins identified via human hepatotoxicity with in vitro cytotoxicity measured in a novel cell- immunocaptured OXPHOS complex activities and respiration. Toxicol based model using high content screening. Arch Toxicol 2006, 80(9): Appl Pharmacol 2007, 223(3): 277-87.
THOMSON REUTERS – Drugs of the Future 2010, 35(1)

Source: http://smr.org.uk/smr/archive/pastmeetings/Downloads/SMR_Sept_09_Report.pdf