Predictive value of in vitro safety studies
Willi Suter
The predictive value of in vitro safety studies is discussed for three important areas of pharmaceutical safety evaluations. In genetic toxicology, currently assays are sensitive for the prediction of cancer, but their overall predictive value is strongly diminished because of their low specificity. In the area of safety pharmacology blockage of hERG channel in vitro has recently been introduced to predict cardiac repolarization delay (QT interval prolongation) in patients. There is a plethora of in vitro methods to predict and characterize liver toxicity. However, little data is available that demonstrate a reliable prediction for hepatotoxicity in vivo over a wide range of chemical structures. In all three areas, further improvements are needed. ‘Omics’ technologies and new cell lines derived from stem cells are expected to strongly contribute to establish new and more predictive in vitro assays.
Introduction
In vitro testing is generally fast, relatively cheap and requires small amounts of test material. Therefore, in vitro safety tests are extensively used in the pharmaceutical and chemical industry in a high- to medium-throughput mode to get results, at the very beginning of product development, that are predictive for the outcome of the later animal safety studies and, ultimately, for prediction of eventual side effects in humans. Predictive in vitro safety experiments are also conducted in later phases of development to investigate mechanisms and pathways leading to toxicity.
This short review on predictive in vitro safety focuses on systems that are of most importance for drug development (i.e., genetic toxicology, cardiac in vitro electrophysiology and liver toxicity), and mainly discusses papers that have been published over the past two years. Until 2005, in vitro assessments were a requirement only for genetic toxicology. They were, however, used extensively in other areas of safety testing [1]. Recently, in vitro assessments also became mandatory for cardiac safety pharmacology [2] and to assess phototoxicity. Readers interested in the status and use of in vitro systems used to test chemicals in Europe are referred to some recent publications [3–5] discussing this area. Indans [6] summarized the use of in vitro data for cosmetics and Kirkland [7] discussed the problems to be expected for the application of the new genetic toxicology testing guideline for hair dyes.
Genetic toxicology
Genotoxicity assessment was the first area of toxicology for which in vitro studies became obligatory for marketing drugs, agrochemicals, cosmetics and other chemical products. Consequently, a large database exists to assess the predicitivity of genetox assays for cancer in animals, and much analysis of this data has been published. By far the most comprehensive recent evaluation is the publication by Kirkland et al. [8]. The performance of the most commonly used in vitro genotoxicity assays (i.e., the Ames test, also known as the Salmonella microsome test; mouse lymphoma assay; in vitro micronucleus test; and chromosome aberration test) has been evaluated for its ability to discriminate rodent carcinogens and non-carcinogens. Altogether, over 700 chemicals were evaluated. The sensitivity (percentage of rodent carcinogens found positive in the genetox assessment) of the test battery was very high (93% of the rodent carcinogens were positive in at least one of the above-mentioned genetox assays). For most of the carcinogens missed there is information suggesting that they are non-genotoxic. However, the specificity (percentage of non-carcinogens found negative in the genetox assessment) of the mammalian assays was very poor (below 45%). The Ames assay, on the other hand, had a reasonable specificity of 73.9%. In the discussion of their results, Kirkland et al. asked several critical questions. For example, whether arguments for reducing the number of animals are still valid and whether it would be better to reduce the in vitro genetox battery and use two in vivo endpoints in different organs together with the Ames test. These disappointing findings, after 30 years routine in vitro genotoxicity testing, call for a complete rethink of this field.
In the short term, improvements may come from changes in the currently used test conditions (cytotoxicity, solubility, etc.). Taking into account the satisfactory performance of the bacterial mutagenicity assessment, one has to focus on finding assays that complement the shortcomings of the Ames test. As pointed out by Brambilla and Martelli [9] the use of S9 rat liver homogenate is the main reason for the failure of the in vitro genetox assays to detect some genotoxic carcinogens. The use of cultured primary hepatocytes [10,11], cells with heterologous expression of cytochromes [12] or liver-derived cell lines [13] might enable new approaches to be developed. In addition, new endpoints (e.g., genotoxic effects related to genes relevant for tumorigenesis [14],DNA strand breaks [15] or cell transformation [16,17]) should be investigated as alternatives to the currently used endpoints (i.e., chromosomal aberration and mutation at the thymidine kinase locus). However, the results obtained so far with these methods do not give much reason to believe that these approaches can provide a solution for the lack of specificity of the regulatory mammalian genetox assays. Thus, a radically new approach is needed that tries to understand the mechanisms leading to chromosome aberrations in mammalian cells. Toxicogenomics might have the potential to provide data that will allow distinguishing between DNA-reactive and DNA-non-reactive mechanisms [18–20]. Chromosome aberrations caused by DNAnon-reactive mechanisms are thought to occur only at excessively high concentrations as currently used for regulatory in vitro genetox testing, and are probably not related to carcinogenesis.
Cardiac electrophysiology
Safety pharmacology is the second area in pre-clinical drug safety assessment where an in vitro assay has become mandatory for marketing approval. ICH guideline S7B [2] requires investigation of inhibition of the human ether-ago-go-related (hERG) channel measured by patch clamp in cells with heterologous expression of the hERG protein. The hERG channel is the rapid component of the delayed rectifying potassium current (IKr) that strongly determines the duration of cardiac repolarization in the human heart. Thus, it determines the duration of the QT interval (time interval measured in milliseconds from the beginning of the Q wave to the end of the T wave in the electrocardiogram) [21]. Several non-cardiovascular drugs have been withdrawn from the market because of their potential to induce QT interval prolongation in the surface electrocardiogram and potentially fatal arrhythmia, so-called torsade de pointes (TdP, malignant polymorphic tachyarrhythmia) [22]. QT interval prolongation was found to be strongly linked with blockage of the hERG channel. In a detailed analysis Redfern et al. [23] showed that most drugs associated with TdP in humans are also associated with hERG channel block at concentrations close to or the same as free plasma concentrations found in clinical use. Based on the analysis of 52 drugs covering a broad range of therapeutic classes, these authors found a 30-fold margin between the maximal therapeutic free plasma level (Cmax) and the hERG IC50 (concentration that inhibits the hERG current by 50%) to provide an acceptable degree of safety for arrhythmogenicity. However, interactions with multiple ion channels can either mitigate or exacerbate the QT interval prolongation, resulting from hERG blockage, and QT interval prolongation is not tightly correlated with the occurrence of TdP . Thus, hERG IC50 values are an important component of a proarrhythmia risk evaluation, but data from more complex models are necessary for accurate risk estimation.
The Langendorff-perfused female rabbit heart model (Screenit system) measures action potential duration (APD), conduction and the TRIaD parameters (triangulation, reverse use dependence, instability and dispersion) [25,26]. The Screenit system detects all changes in repolarization indices (TRIaD) and conduction velocity that might interfere with APD values. Changes in the TRIaD parameters have been found to be much more predictive for an arrhythmic potential than changes in APD. In several validation studies, Screenit was found to be highly predictive for the proarrhythmic potential of drugs [25–28]. The group of Antzelevitch [29–31] developed another proarrhythmia model based on the transmural heterogeneity of the expression of cardiac ion channels. Using a left ventricular wedge preparation of the canine heart, transmembrane action potentials from epicardial and M-regions were simultaneously measured. The data support the hypothesis that the risk for the development of TdP is related to the increase in transmural dispersion of repolarization, rather than to prolongation of the QT interval. Bottino et al. [32_] developed mathematical models that use hERG IC50 data and APD results measured from dog Purkinje fibers to predict drug interaction with other cardiac ion currents and dispersion of repolarization in transmural ECG. In addition, this in silico approach allows investigation of the influence of known clinical risk factors for arrhythmia (e.g., hypokalemia) on the proarrhythmic risk in patients.
Liver toxicity
Liver toxicity has been a major reason for ‘black box warnings’ and withdrawal from the drugs market over the past 25 years [33_]. Liver toxicity observed in preclinical and clinical studies is also a major reason for stopping the development of promising drug candidates. Safety experts in the pharmaceutical industry are therefore strongly interested in establishing predictive screening systems and mechanistic models to detect hepatotoxicity early in the drug development process. Farkas and Tannenbaum [34] gave a thorough overview about the currently available in vitro methods to study hepatotoxicity (i.e., the isolated perfused liver, liver slices, primary hepatocyte cultures, human liver-derived cell lines and flow-through bioreactors). This listing of systems gives an idea about the complexity of the situation, which is increased even more if one considers the plethora of endpoints that can be studied using these models. Xu et al. [33_] distinguished two sets of models: cytotoxicity assays (e.g., protein synthesis, membranes integrity and caspase-3 induction); and pre-lethal mechanistic assays (assays to assess steatosis, cholestasis, phospholipidosis, reactive metabolites, mitochondrial toxicity, oxidative stress and drug–drug interaction).
Because liver toxicity has frequently been described as species-specific and an idiosyncratic phenomenon, much work has been dedicated to establishing predictive models using human hepatocyte cultures and cell lines derived from the human liver [34,35]. These technologies have been strongly facilitated by the development of reliable methods for cryopreservation of human hepatocytes [36,37] and opened new possibilities to investigate human-specific processes that cannot be explored in the standard pre-clinical animal studies.
The ‘omics’ revolution provides new tools that allow the study of various mechanisms leading to liver toxicity [34,38]. Before using ‘omics’ tools to assess hepatotoxicity in vitro, the differences between the in vivo and the in vitro situation have to be carefully studied and the experimental conditions need to be standardized [39]. Under optimal conditions, predictive and mechanistic toxicogenomic data are concordant for compounds tested in vitro in primary hepatocytes and in vivo in the liver [40]. Because liver toxicity is a major clinical problem, the early detection of hepatotoxicity liabilities is of great importance to the pharmaceutical industry, and new screening approaches are needed. Miret et al. [41] compared different methods to assess cytotoxicity in HepG2 cells. They concluded that a battery of assays (ATP level, cell necrosis, caspase 3/7 activation) is needed to reliably detect cytotoxicity. Dambach et al. [42] proposed the use of a battery of five cell lines, four of which expressed one of the major drug metabolizing cytochromes P450 enzymes (3A4, 2C9, 2C19, and 2D6), to predict acute direct or metabolism-mediated hepatic necrosis. Using this system, correct predictions of 585/587 non-hepatotoxic drugs, 15/21 severely hepatotoxic drugs (i.e., with black box warnings or withdrawal from the market) and 51/71 variably hepatotoxic drugs were made. O’Brien and Siraki [43] developed the accelerated cytotoxicity mechanism screening using freshly isolated hepatocytes that were treated with test compounds in the presence of inhibitors or inducers of drug metabolism.
Conclusions
The three areas of in vitro safety evaluation discussed in this short review have developed quite differently. Genetic toxicology, the most successful field of in vitro toxicology over the past 30 years, has become strongly regulated, which has led to the creation of a huge database for the routinely used standard assays, but unfortunately also to a weakening of innovation and interest in the development of improved methods. The recent analysis of the predictivity of the current regulatory test systems for cancer in rodents showed that only very few genotoxic carcinogens are missed, indicating a high sensitivity. However, because of the low specificity of the mammalian in vitro assays, a positive result does not really predict that a compound is a carcinogen [8]. Improvement of the established set of assays and the development of fundamentally new approaches are therefore desirable to justify the value of in vitro genetox assays in regulatory toxicology.
In cardiac safety pharmacology, in vitro assays were mainly established over the past 10 years. Data from hERG assays became mandatory for marketing authorization last year [2], justified by the strong correlation between hERG channel blockage and the occurrence of TdP in patients [23]. Because all specialists agree that the hERG assay is not sufficient for a reliable prediction of a proarrhythmic potential, innovation will not be stifled by the new guideline.
New methods are in development, using the large biological knowledge of cardiac biology and physiology [31,24_]. The use of in vitro systems for the detection and characterization of liver toxicity is not requested by regulators. Because of this lack of a guideline and the many ways in which exposure to xenobiotics can lead to liver injury, a broad range of assays has been developed.These assays are used to characterize and investigate mechanisms and, thus, contribute to hazard identification and characterization. Some companies have established screening assays for liver toxicity early in development [41,43], but there is little information about the predictive power of these approaches [42]. Because the predictivity is limited in all areas of in vitro safety testing, improvements are necessary before in vitro data can be reliably used to assess risk to humans as proposed by Combes [44]. Stem cell technology, genetically modified cell lines [48] and ‘omics’ technologies [49,50] may provide new ways for in vitro safety assessments.
Update
In a recent article, Cimino [51] provided an overview of testing strategies used by regulatory agencies throughout the world. The author recommended strict adherence to internationally recognized protocols for regulatory testing, and discussion of the testing program with the regulatory body responsible for the regulatory oversight. Sanguinetti and Tristani-Firouzi [52] reviewed the current knowledge about the structure and function of the hERG channel as well as its role in cardiac repolarization. Although the crucial amino acid residues of the drug binding site seem to have been identified, further studies are needed to explain why the hERG channel is blocked by so many structurally different molecules.