Applications of cytotoxicity assays and pre-lethal mechanistic assays for assessment of human hepatotoxicity potential

Abstract

While drug toxicity (especially hepatotoxicity) is the most frequent reason cited for withdrawal of an approved drug, no simple solution exists to adequately predict such adverse events. Simple cytotoxicity assays in HepG2 cells are relatively insensitive to human hepatotoxic drugs in a retrospective analysis of marketed pharmaceuticals. In comparison, a panel of pre-lethal mechanistic cellular assays hold the promise to deliver a more sensitive approach to detect endpoint-specific drug toxicities. The panel of assays covered by this review includes steatosis, cholestasis, phospholipidosis, reactive intermediates, mitochondria membrane function, oxidative stress, and drug interactions. In addition, the use of metabolically competent cells or the introduction of major human hepatocytes in these in vitro studies allow a more complete picture of potential drug side effect. Since inter-individual therapeutic index (TI) may differ from patient to patient, the rational use of one or more of these cellular assay and targeted in vivo exposure data may allow pharmaceutical scientists to select drug candidates with a higher TI potential in the drug discovery phase.

Introduction

It is estimated that for the drug research that was initiated in 2001, and resulting in a planned drug approved in 2013, the average cost is US$ 1.9 billion [1]. About half of this cost is the “time costs”, or “capitalizing out-of-pocket costs to the point of marketing approval”[1]. Therefore, tremendous amount of resources and time are wasted if a drug is withdrawn after launch due to toxicity reasons, not to mention the human suffering it may have caused to patients.

Drug-induced toxicity, or adverse drug reactions, have emerged as a major cause of post-market withdrawal of medications. During 1975–1999, 45 of previously approved drugs have acquired one or more black box warnings and 16 were withdrawn from the market for safety reasons [2]. Of the 45 that acquired black box warnings, 10 were due to hepatic toxicity (22%). Of the 16 that were withdrawn from the market, 5 were due to hepatic toxicity (31%). These statistics make hepatic toxicity the largest root cause issue in this cohort of black box warnings and withdrawals [2]. In addition to post-market attrition, toxicity is also a major reason of pre-launch attrition of drug candidates. It is thus highly desirable to devise a strategy to detect and therefore minimize such toxicity in the earlier phases of drug development.

A survey to examine the strengths and weaknesses of animal studies to predict human toxicity indicated that current safety studies in animal models do not adequately predict human hepatic toxicities [3]. In a recent study of predictivity of human drug toxicity by regulatory animal toxicity testing, 72% (167 of 230) of human toxicities were found to be associated with toxicity in animals [3]. Of these, 8% were in rodents only, 28% were in non-rodents only, and 37% were in both. Correlation between animal and human toxicities was high for cardiovascular, hematologic, and gastrointestinal toxicities. However, among human toxicities, hypersensitivity, cutaneous reactions, and hepatotoxicity had the poorest correlation with regulatory animal toxicity tests [3].

Drug-induced hepatic injury is the most frequent reason cited for withdrawal of an approved drug and accounts for more than 50% of cases of acute liver failure [4]. However, in only approximately half of new pharmaceuticals that produced hepatotoxicity in clinical drug development was there any concordance with animal toxicity studies [3]. Occurrence of these adverse drug reactions is largely attributable to the imperfect extrapolatability of animal study findings to humans. Several possible reasons have been proposed to explain such animal/human discrepancy, including:

• Animal species do not predict human metabolism.

• Animal species under well-controlled experimental settings do not predict the diversity of human patients living in heterogeneous conditions.

• “Normal”, healthy animal species do not predict diseased human patients.

• A sub-population of human patients, for reasons attributed to pharmacogenetic, environmental, and/or life-style susceptibility factors, are particularly sensitive to adverse drug reactions.

Given that there is a plethora of possible reasons for missing a toxic “signal” in current animal studies, it is unlikely that isolated and simple solutions exist to adequately predict human toxicity. Instead, a multitude of experimental and theoretical methods need to be explored and data combined to address this complex issue. This review will focus on the use of cell-based assay systems for the early screening of human drug toxicity in the discovery phase of drug development, especially of human hepatic toxicity. The rationale for the placement of these screens in the drug discovery phase is obvious: this is the phase when chemical choices and chemical modifications can still be made before significant development resources are committed to a particular new chemical entity (NCE).

The relevance of using cell-based system as surrogate models to organ toxicity in vivo is analogous to the well-accepted practice of screening for desired pharmacological effects. Therefore, all the caveats of the in vitro systems would apply. Most importantly, while the in vitro systems can identify a chemical's potential to cause an effect (in this case, side effect), the local exposure levels of an NCE in vivo would largely determine whether the chemical would ultimately be toxic to humans. The famous “Paracelsus doctrine” is just as relevant today as it was 500 years ago, the difference being today we should use “concentration” instead of “dose” to predict toxicity.

In addition to challenges in scalability between in vitro and in vivo exposure levels, the practice of human cellular toxicity also faces another challenge. That is, in the most extreme scenario, the mechanisms of drug toxicity to human organs could be as diverse as the mechanisms of human physiology. Even within a single cell type, e.g. hepatocytes, the mechanisms of chemical injury can be quite diverse. A recent review of hepatocyte toxicity explored at least 6 mechanisms (see Fig. 2 in [4]) and 10 manifestations (see Table 1 in [4]) of liver injury. In light of such complexity, how can one practically address so many potential possibilities of “chemical toxicity” within a finite amount of time in the drug discovery phase? A logical approach is to use a panel of better-characterized assays in parallel, in order to identify one or more chemical molecules with the best overall “therapeutic indices” in vitro. However, no in vitro systems today can mimic the potential complexities of the human body. Therefore, it is impractical to expect “perfect predictions” of in vivo outcomes by any in vitro systems. That said, some in vitro systems have better predictivity than others and they are the focus of this review.

Conventional cytotoxicity assays rely on measuring one or more cytotoxic indicators, including loss of membrane integrity or cytolysis (e.g. LDH release, membrane-impermeable DNA stain), apoptosis (e.g. activation of caspases), loss of critical macromolecules (e.g. ATP depletion, GSH depletion), anti-metabolic effects (e.g. tetrazolium salt assays, Alamar blue assay), or anti-proliferative effects (e.g. inhibition of DNA or protein synthesis).

In vitro, known hepatotoxic chemicals can be frequently identified and ranked by their 50% inhibitory effect concentration (μM) in cellular models, such as HepG2 cell line (a human hepatoma cell line), primary hepatocytes, cryopreserved hepatocytes or Liverbeads, and WIF-B9 cell lines, by measures of any of the aforementioned cytotoxic indicators [5], [6]. It is noteworthy that in terms of drug metabolizing activity, the rank order among various cellular models is: primary hepatocytes > cryopreserved hepatocytes > HepG2 cell line ∼ WIF-B9 cell line. Liverbeads are cryopreserved hepatocytes entrapped in alginate beads; WIF-B9 cells are hybrids of rat hepatoma (FaO) and WI38 human fibroblasts. Cytotoxicity is not detected in these cell lines for some hepatotoxins, such as ethionine, valproate, galactosamine, thioacetamide, diclofenac and isoniazide. This may be in part attributable to incomplete metabolic competence of these cellular systems [5].

Assessment of the predictivity of drug-induced human hepatotoxicity by cytotoxicity assays can be made by comparison with that of regulatory animal toxicity testing. If the cytotoxicity assay has similar predictivity, it should be considered of high value for safety assessment in drug discovery because of its much lower cost and much faster turn around time. For such assessments, drugs that have been marketed were classified into three categories according to the severity of human hepatotoxicity they produce: severe or “bin 1” (producing >1% frequency of increased serum ALT plus 2 of jaundice, >3 reports of liver failure or black box warning; 42 drugs); moderate or “bin 2” (producing 0.1–1% frequency of increased serum ALT plus jaundice or label of occurrence of adverse effect; 283 compounds); none to minimal or “bin 3” (<0.1% frequency of increased ALT, no clinical symptoms; 286 compounds). Table 1 compares the predictivity of various cytotoxicity assays applied to these drugs [6] and demonstrates that these assays have up to half the sensitivity (proportion of positives identified) of animal tests [4]. These values may be substantially underestimated because they are based on retrospective assessment of concordance of test results with drugs that have already been marketed for usage in humans. Therefore, those discovery compounds that would have tested positive by in vitro cytotoxicity tests and animal toxicity tests and would have caused severe human hepatotoxicity (or in “bin 1”), were already excluded in this retrospective analysis. Although the in vitro assays were still relatively insensitive at detecting hepatotoxicity, when drugs were tested positive in these assays (e.g. greater than 50% effect at 30 μM in HepG2 cells), there was high probability of human toxicity (i.e. they had high specificity, see Table 1).

The predictivity of cell-based assays for in vivo toxicity depends in part on the point in the sequence of pathogenic events that they target (Table 2). Assays that target late events in the process of cell injury, when the cell is near death, are more likely to miss toxicities that require chronic exposure or exert adverse but non-lethal effects. Many of the above cytotoxicity assays (e.g. LDH release, mitochondria dye reduction, cell rupture, membrane blebbing, DNA degradation and nuclear condensation) are for late-stage toxicity and cellular events associated with a lethal apoptotic or necrotic effect [7], [8]. Such assays have low sensitivity (Table 1) for detection of adverse cellular effects and furthermore provide little mechanistic understanding of the toxicologic effects in humans. Conversely, assays that provide early assessment of specific toxicologic mechanisms in cells (Table 2) prior to the onset of the late stages of non-specific degeneration and apoptotic or necrotic death should theoretically have greater predictive power and extrapolatability across models and species.

While these retrospective analysis support the value of in vitro cytotoxicity screens to identify the “overtly” toxic compounds, it points to the need of further refinement in the method to predict subtle or sub-lethal adverse events that account for the majority of side effect profiles of human pharmaceuticals.

It is well recognized that even within the same organ, drug-induced toxicity can be caused by a variety of mechanisms. For example, at least six mechanisms that involve hepatocytes have been proposed to produce liver injury [4]. Therefore, a variety of pre-lethal mechanistic assays have been used by researchers to study potential hepatic side effects of drugs, each addressing a specific mechanism (steatosis, cholestasis, phospholipidosis, reactive intermediates, mitochondria membrane function, oxidative stress, intracellular calcium homeostasis, etc.). These assays have the advantage to identify the “subtle” and potentially more clinically-relevant lesions in a human or mammalian cell-based assay system. They also have the advantage to identify such lesions at much lower drug concentrations than the concentrations needed for gross or overt cytotoxicity (i.e. more sensitive). An ideal in vitro assay is such that both the assay endpoints and drug concentrations needed to elicit those effects are more relevant to the human clinical observations. For the remaining of this article, we will review those mechanism-specific pre-lethal cellular assays that have generally been considered useful by multiple practitioners in this field. This is not intended to be a comprehensive review of all of the available and applicable in vitro assays that exist in the literature. For readers who are interested in earlier reviews of hepatotoxic mechanisms, drugs and experimental tools, please refer to classical text books such as Zimmerman's Hepatotoxicity [9].