“Drugs on oxygen”: an update and perspective on the role of cytochrome P450 testing in pharmacology

Abstract

Low hit rates for lead compounds and high attrition remain a major problem for drug development. The reasons for compound failure range from poor pharmacokinetics to toxic metabolites and adverse drug interactions; all of which are frequently mediated by cytochrome P450-dependent monooxygenases (CYPs). However, despite some 30 years of assay development and refinement, CYP metabolism remains a critical issue during drug development. While current testing strategies succeed in characterizing single substance toxicity, they are challenged by practical issues such as assay standardization or complex scenarios such as multidrug usage. This editorial summarizes where we stand and highlights the major challenges we face with CYPs in drug development today. The article also tries to spell out the future direction of CYP testing. The latter will depend on the extended inclusion of polypharmacy into testing strategies, as well as on our capability to make use of upcoming complex in vitro test systems and their inclusion into tiered testing strategies.

Keywords::

• CYP screening
• cytochrome P450-dependent metabolism
• drug–drug interaction (DDI)
• drug testing
• in vitro testing

1. Drug attrition, cytochrome P450, and xenobiotic metabolism

1.1 Scale of the problem

Public perception is seeing drugs increasingly as a granted commodity. Treatment is available for many ailments and a high standard of safety testing makes the use of drugs safer than ever before. This reassuring drug safety, however, comes at a price. That is a worrying decrease in the number of new drugs reaching the market. The average hit rate of pharmaceutical library screens is 0.01% and a pass rate of 1 out of 10 during subsequent safety testing reduces this number further [1]. A major reason for this high rate of attrition is unfavorable Phase I metabolism by cytochrome P450-dependent monooxygenases (CYPs, EC 1.14.14.1). These enzymes catalyze the activation and introduction of molecular oxygen into saturated aromatic ring structures. The increased electrophilicity prepares the substrates for further breakdown or for Phase II conjugation and excretion. Nevertheless, with hindsight to xenobiotic metabolism the respective metabolites can as well be toxic (e.g., albumin binding metabolites of the PPARγ-agonist troglitazone), critically alter first pass metabolism and have an influence on the potential for drug–drug interactions (DDI) (e.g., modifiers of CYP3A4 activity such as the inhibitor clarithromycin, St. John's worth or dietary flavinoids) [2-7]. Altogether biotransformation-related toxicity is responsible for 23 – 27% of lead compound failures and CYPs make a major contribution as they are involved in 75% of drug metabolism [1,4,8].

1.2 CYP-mediated drug metabolism

Regarding their substrate specificity and transcriptional regulation, CYPs show a high variation across species and are thus not readily accessible by animal tests. Hence screens for metabolite identification and CYP induction rely primarily on humanized in vitro tests and focus on those CYPs that are responsible for ≥ 90% of CYP-mediated drug metabolism (CYPs 3A4, 2C9, 2E1, 1A2, 2A6, 2B6, and 2D6) [9]. Systems routinely used comprise recombinant enzymes and cellular test systems such as primary hepatocytes, metabolically adapted cell lines (e.g., CYP-expressing V79 and Hep2G cells), or hepatocyte-like cells derived from stem cells [10-13]. Usually, these assays are a good first reflection of hepatic metabolism and critical results can be subsequently confirmed in animal tests [14,15]. The situation is more difficult with regard to intestinal CYP metabolism, which is dominated by CYP3A4 and CYP2C9 [3]. In the vicinity of efflux transporters such as P-gp, MRP1 and MRP2, intestinal CYP activity can have a major impact on drug clearance and availability. A classical example is the lipid-lowering HMG-CoA reductase inhibitor atorvastatin, where intestinal metabolism via CYP3A4 and P-gp/OATP C contributes up to 95% to the overall clearance [16]. While the development of suitable in vitro systems is still at an early stage, improved modeling approaches have been used recently to assess intestinal first-pass metabolism [3,17].

Today direct CYP-mediated drug toxicity is thus less of a problem than it was 20 years ago [8]. In 1991, unfavorable pharmacokinetics was the reason for 40% of drug failures. Some 10 years later this rate was down to 10%, not the least because of a better understanding of CYP-mediated drug metabolism and the appropriate redesign of drugs initially found to be toxic [1,8,18]. Moreover, data on CYP metabolism are now routinely used to design pro-drugs with improved pharmacokinetic properties (i.e., the CYP2A6- and CYP2C9-mediated activation of tegafur to 5-fluorouracil) [19].

1.3 Allelic variability

However, drugs that are preferentially metabolized by CYPs with a high genetic variability (i.e., CYP2D6 and CYP3A4) are still a considerable challenge [9,20]. The different genotypes for these enzymes express protein variants with high to poor enzymatic activity, thus rendering certain drugs less effective (i.e., codeine, high CYP2D6 activity) or harmful (i.e., perhexiline, low CYP2D6 activity). For CYP2D6 alone there are at least 74 different allelic variants and the enzyme is involved in the metabolism of about 25% of all marketed drugs. Reaction rates of major CYP2D6 substrates such as the antitussive dextromethorphan can vary up to 53-fold due to allelic variation. In this particular example, this will affect the conversion of the pro-drug to the pharmacologically active dextrorphan as well as clearance of the latter. The frequently used β-blockers and antidepressants are also typical substrates for CYP2D6 [9]. Consequently, caution has to be exercised in cases where metabolism studies indicate the involvement of genetically variable enzymes (i.e., CYP1A2, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP3A4, and CYP3A5). Subsequent tests should include mutant enzymes and data of any following Phase II and Phase III trials should be analyzed with hindsight.

2. Drug–drug interactions (DDI)

Things get even more challenging when it comes to evaluating potentially adverse drug metabolism in presence of altered CYP activity. The classical situation is the simultaneous presence of additional substrates, be it inducers or inhibitors. The range of substances in question can comprise natural CYP substrates like food ingredients (e.g., flavones) as well as drugs [21]. The latter is a particular problem as longevity is increasing worldwide and such is the likelihood for polypharmacy. The average patient above 65 years is likely to be medicated with 2 – 5 drugs, which in turn increases the risk for drug–drug interactions [22]. Pharmaceutical industry has reacted by implementing screening strategies for possible DDI early during drug development. In addition, regulatory authorities tried to further harmonize testing strategies by issuing respective guidelines such as the “Draft Guidance for Industry – Drug Interaction Studies” by the U.S. “Food and Drug Administration” (FDA) [23] or the “Guideline on the investigation of drug interactions” by the “European Medicines Agency” (EMA) [24]. Depending on the testing requirements systems use microsomal liver extracts, purified recombinant CYPs as well as defined CYP mixtures. Routine screens are limited to CYP3A4, CYP2D6, CYP2C9, CYP2C19, and CYP1A2 because these five enzymes cover up to 90% of human drug metabolism [9,15,25]. Enzyme kinetics are either measured with model substrates [26-28] or are based on quantitative mass spectroscopy [29,30]. The latter has the advantage of high accuracy and the option to use pharmaceutically relevant substrates, while the standardized setup of the first is ideal for high-throughput screening [25].

2.1 CYP inhibition

In case of irreversible enzyme inactivation, the major experimental challenge is a potential delay in its onset due to time-dependent inhibition [31]. Yet subsequent hazard assessment is relatively straightforward. The alleviation of enzyme inhibition requires new protein synthesis and this reduces the problem to the question if a reduced enzyme turnover is condonable (i.e., as consequence of the inactivation of CYP2B6 by N,N',N''-triethylenephosphoramide). However, in vivo only about a third of strong DDIs are caused by irreversible enzyme inhibition and the majority of inhibitory effects are reversible [32]. Typically one distinguishes four types of inhibition, namely competitive inhibition where the inhibitor competes directly with the substrate and which is opposed to uncompetitive inhibition where the inhibitor reacts with the enzyme–substrate complex, as well as mixed inhibition and non-competitive inhibition. The latter is a special case of allosteric inhibition where binding of the inhibitor occurs independently from the presence of the substrate. Nevertheless, the experimental detection of the mode of inhibition can be problematic due to data quality, the presence of multiple inhibitory effects, or enzyme source and quality. As a consequence discrepancies amongst studies determining the type of inhibition for the same CYP are as high as 35% [32]. The absence of a clear result will inherently prevent the determination of key kinetic parameters such as the inhibitory binding constant (K _i). It should be noted also that most strategies are designed to identify inhibition of specific CYPs. The correct quantification and prediction of promiscuous CYP inhibitors is still an unsolved challenge [33].

2.2 CYP induction

With regard to the possible consequences of CYP induction, the overall situation is even less satisfactory. Screening for inhibitors of inducible CYPs will help to identify obvious risks due to cross inhibition and induction of CYPs is not necessarily a contraindication for drug safety [4]. That said the induction of CYP3A4 by the antibiotic rifampicin for example is likely to increase the risk of hepatic injury by predominant CYP3A4 substrates such as acetaminophen [10,34]. Although several hepatic CYPs are known to be inducible, amongst them some of the main drug metabolizers (e.g., CYP3A4, CYP1A2, CYP2C9, and CYP2B6), assays for the identification of potential enzyme inducers are not common practice and there is little consensus on a standardized experimental setup [10]. The major problem is the complex regulatory network that governs CYP expression. While some CYPs are regulated via mRNA stabilization (i.e., CYP2E1), transcription factor-mediated induction is more common. Three transcription factors are involved in the regulation of most CYPs. These are the arylhydrocarbon receptor (AHR) (CYPs 1A1, 1A3, 1B1 & 2S1), the constitutive androstane receptor (CAR) (CYPs 2A6, 2B6, 2C9, 2C19 & 3A4) and the pregnane X receptor (PXR) (CYPs 1A2, 2B6, 2C8, 2C9, 2C19, 3A4, 3A5 & 3A7) [35]. Translocation of the respective factor into the nucleus and subsequent transcriptional activation is usually not only dependent on ligand binding but on the activity of cofactors such as the nuclear translocator ARNT for AHR or the retinoid X receptor RXR for CAR or PXR. Things are further complicated by cross-regulation, overlapping target genes and modulation of receptor activity by other signaling pathways such as the β-catenin pathway (AHR) [35,36]. For example pairing of PXR with the vitamin D receptor VDR will activate transcription of CYP2C9, while activation of CYP2C6 requires pairing of PXR with the glucocorticoid receptor (GR) [37]. Similarly, transcription of CYP3A4 can be triggered by PXR in combination with CAR as well as with other receptors. Meanwhile, activity of PXR can be subject to cyclin-kinase-dependent modulation [38].

Screening assays often focus on receptor binding or on receptor activation [35]. Nevertheless, both strategies struggle to integrate effects of possible cross-regulation and are thus frequently replaced by quantifying target gene expression in cellular assays.

2.3 Systemic challenges and regulatory issues

The next challenge is the transfer of any in vitro data to the in vivo situation. In vivo overall CYP inhibition depends on pharmacokinetics, subsequent cellular transport and most importantly, the in situ concentration of the perpetuator drug (including any inhibitory metabolites, [25],). Testing for the latter is a major challenge and not necessarily part of routine testing strategies, as it is not required by the FDA or EMA. A recent literature survey revealed that 106 out of analyzed 129 CYP inhibitors had circulating metabolites, several of them being inhibitors themselves [39]. In the case of sulfinpyrazone for example the two major metabolites inhibit CYP2C9 more effectively than the parent compound. In some cases such multiple inhibitors can be assessed on the basis of their estimated inhibitor affinity (IC₅₀) [32].

Due to the complexity of the issue there are no harmonized testing guidelines for the evaluation of CYP inhibition. For reversible inhibition there is basic consensus to define a DDI risk index based on the ratio of the inhibitor plasma concentration to the inhibition constant K _i whenever possible but the respective experimental strategies to obtain these values remain disputed [31,40,41]. Literature values of the K _i for classical CYP3A4 inhibitors such as fluconazole, ketoconazole and itraconazole differ up to 10-fold, probably due to different experimental conditions [41]. The situation is further complicated by the fact that different regulatory authorities recommend different strategies. The FDA advises the use of total concentration-based values, while EMA asks to refer to unbound plasma concentrations [41]. There is good reason for both approaches. Using unbound plasma concentrations lowers inter-laboratory assay variability, while the FDA's use of total concentrations tends to give more conservative results, at least for the prediction of hepatic metabolism. Correspondingly the EMA guideline seems more suitable for the assessment of orally administered CYP3A4 inhibitors as it suggests to include potential intestinal DDIs, an issue the FDA does not specifically address. While efforts to identify a common strategy are ongoing [40], it has to be acknowledged that the overall number of variables opposes a “one size fits all” solution.

2.4 The role of QSAR

Apart from improved screening active site modeling could be a solution (e.g., warfarin binding in CYP2C9) [42]. However, this requires detailed information about the oxygenases involved as well as extensive expertise to assess all possible interactions. Unsurprisingly, molecular modeling is thus predominantly used in academic case studies [43]. Similarly, the successful application of QSAR is currently restricted to ligand-based approaches that complement already existent experimental data sets [43]. Efforts to extend QSAR to predict DDI during cancer therapy or polypharmacy have shown some promise but found the method to be over-predictive [44,45].

3. Conclusion

Despite all remaining limitations the introduction of CYP screens has undoubtedly been a success, initially with regard to the assessment of toxic metabolites and optimization of pharmacokinetic properties but recently for the assessment of potential DDI as well. Back in 1998, the calcium channel blocker mibefradil had to be withdrawn from the market due to severe adverse drug reactions [46]. Today this drug would not pass early safety testing, because inhibitors of multiple CYPs (i.e., CYPs 2D6, 1A2, and 3A4) are rated critical. However, with increasing polypharmacy there is greater need for an improved characterization of multidrug metabolism. Clinically, relevant adverse reactions usually occur at a rate of 1:10,000. Therefore, clinical trials are no adequate safeguard to detect less obvious DDIs once they were missed during early drug development [46].

4. Expert opinion

Current strategies suffer from three major limitations that is the insufficient inclusion of transport phenomena and Phase II reactions, organ specificity and the inherent limitations as to the number of drugs that can be screened simultaneously. The latter could be tackled by careful assessment of likely drug combinations and a subsequent inclusion of the most frequent ones into appropriate high-throughput screens. The bigger challenge is the evaluation of coupled Phase II and Phase III reactions. Hepatotoxic injury of the anti-inflammatory diclofenac for example is partly attributed to impairment in the glucuronidation of its CYP2C9-metabolite 4-hydroxydiclofenac [47]. Ultimately, a complete evaluation of CYP-mediated toxicity thus requires an integrated approach for which animal experiments still are the accepted gold standard. However, with regard to CYPs and DDI animal experiments often suffer from the aforementioned species barrier and their elaborate design and time requirements put them at odds with the required need for increased testing efficiency and high-throughput. A solution might be the implementation of tiered testing strategies where CYP assays are subsequently coupled to cellular assays for Phase II metabolism and metabolite transport. Most organ-mimicking systems are still at their early stages of development and especially kidney-mimicking assays are still in their infancy. However, progress is rapid and there are promising systems available for liver, lung, and heart [13,48]. With their suitability for high-throughput screening, such cellular assays are ideal experimental filters to precede further in vivo assessments.

Last but not least current testing strategies focus on liver as the primary target organ for any xenobiobic metabolism. However, examples such as the metabolism of the anesthetic propofol by CYP2B in the brain and the existence of other extrahepatic CYPs (e.g., CYP2S1) demonstrate the need for a wider focus. With hindsight to the numerous problems with hepatic testing strategies, in vitro organ specific testing of CYPs is a substantial challenge. In addition to already used primary cells and the aforementioned organ-mimicking systems, stem cells show particular promise here. Current differentiation protocols include, amongst others, the generation of cardiomyocytes and neurons [13]. Nevertheless, it remains to be shown how these developments can be used to complement current testing strategies in the years to come.

Declaration of interest

The authors state no conflict of interest and have received no payment in preparation of this manuscript.