The chemical and structural diversity of the proficiency chemicals used for (pre)validation needs to address both the chemical applicability domain of the chemical Universe for which the test method is intended to predict endpoint-specific toxicity, but also be structurally relevant for the biological role of the endpoint. In addition, in many cases, where known, natural and endogenous ligands should be included in the chemical selection (Waxman 1999; Jacobs and Lewis 2002) as these are the ligands that the anthropogenic chemicals of concern are mimicking.

A targeted reiterative but not systematic literature search was carried out to identify an initial selection pool of candidate proficiency chemicals for which human CYP modulation data was available for CYP1A1/1A2, CYP2B6, and CYP3A4, and that belonged to OECD TG Programme relevant chemical classes (including industrial chemicals, pesticides, and food additives). Data sources from both human cell line in vitro and human in vivo studies relating to specified CYP induction and directly related receptors were sourced and critically evaluated, but due to the general scarcity of in vivo human studies carried out with non-pharmaceutical chemicals, only human in vitro data was available for most chemicals. Support that this would be an acceptable approach to take was first established within the OECD WNT, given the scarcity of human in vivo data. In some cases, it was appropriate to use rodent data for weight of evidence support.

On this regulatory acceptable basis, between 2017 and 2021, Scopus and pubmed search engines were queried with respect to human relevant data, cytochrome P450, and chemicals, including pesticides/bacteriocides and excluding pharmaceuticals as these are already addressed in the draft TG. The results were filtered and prioritised for human relevant data including human relevant cell lines, e.g. HepaRG, HepG2, and the relevant CYPs in the draft validated TG, together with relevant references within the papers. These were critically reviewed and double checked by the authors and then external regulatory experts for the OECD TG Programme (WNT).

In the validated test method, overall CYP enzymatic activity is quantified in an in vitro human hepatocyte cell line (cryopreserved HPR116 differentiated HepaRG cells) before and after pretreatment with test chemicals, by measuring the rate of metabolic conversion of substrates that are selective for CYP1A1/1A2 (phenacetin to acetaminophen), CYP2B6 (bupropion to hydroxybupropion), and CYP3A4 (midazolam to 1′-hydroxymidazolam) using a LC-MS analytical technique (Bernasconi, Pelkonen et al., 2019). In such cell systems, the overall effect on enzyme activity is dependent on the extent to which a chemical up or downregulates not only CYP mRNA/protein levels, but also functional enzymatic activity. Data evaluated in the literature search therefore included in vitro mRNA/protein quantity and enzyme activity data generated using human cells, as well as enzyme activity data generated in human liver microsomes (HLMs) and recombinant enzyme preparations (REPS). CYP1A1/1A2 gene expression is well understood and extensively documented to be induced by activation of the AhR (e.g. Bock 2014; Vogel, Van Winkle et al., 2020), and activation of CAR and PXR has been shown to induce CYP2B6 and CYP3A4 (Tolson and Wang 2010; Wang, Ong et al., 2012). Data showing activation of these receptors was taken to indicate likely mRNA/protein upregulation of the respective CYPs. Preliminary assessment as to whether a candidate chemical is likely to induce, inhibit, or have no effect on a specific CYP in the test method was made on the basis of enzyme activity data from human cell systems with innate CYP expression, when this data was available. Where this data was not available, estimations were made based on available mRNA/protein quantity data and enzyme activity data from non-cell preparations, when possible. In the latter case, estimations were considered to be of lower reliability as compared to cases where cell-based enzymatic activity data was available, and insufficient data were considered to be available to estimate the effect of a chemical in the test method if overall the available good quality data was considered to be contradictory. In the design of validation experiments intended for applications beyond classification and prioritisation purposes, it is good practice to include proficiency chemicals that are expected to produce a potency range from negative to low, moderate, and strong effects, and to generate concentration-response information, as this is of greater utility for IATA approaches (Jacobs, Ezendam et al., 2022). Thus, for chemicals for which cell-based enzymatic activity data showed them to be an inducer, the magnitude of any observed CYP enzyme induction was categorized as low (≤3 fold), moderate (>3 to 4.5 fold), or strong (>4.5 fold), and this information was used in the chemical selection considerations. When cell-based enzymatic data was absent (not tested), the magnitude of an expected effect was categorized as uncertain.

As the conversion of phenacetin to acetaminophen is catalyzed by both CYP1A1 and 1A2 (Kcat = 0.84 and 2.2 min⁻¹, respectively) (Huang, Deshmukh et al., 2012), production of acetaminophen is used as an overall marker of both CYP1A1 and 1A2 activity in the CYP enzyme induction assay; and CYP1A1/1A2 activity predictions were thus made based on the cumulative available data for both CYP1A1 and 1A2. Formation of 1′-hydroxymidazolam is widely used as a selective marker of CYP3A4 activity. The latter reaction is catalyzed to a significant extent by both CYP3A4 and 3A5 (Vmax = 35 and 72 nmol/min/nmol CYP, and Km = 5 and 14 μM, respectively) (Williams, Ring et al., 2002), but the HepaRG cell line contains two CYP3A5*3 alleles, which are known to be null due to expressed RNA instability (Jackson, Li et al., 2016). It is expected that CYP3A4/5 activity in HepaRG cells is predominantly attributable to CYP3A4 activity. Hydroxybupropion formation, on the other hand, has been shown to be catalyzed nearly exclusively by CYP2B6 (Faucette, Hawke et al., 2000; Hesse, Venkatakrishnan et al., 2000), with other CYP isoforms (including CYP2C19 and 3A4) being involved to only a negligible extent (Faucette, Hawke et al., 2001; Sager, Price et al., 2016).

As it is anticipated that on becoming an OECD TG, the test method will be widely utilised internationally, and will fall under the Mutual Acceptance of Data agreement for the OECD member country regulatory jurisdictions, it is important that proficiency chemicals should be associated with the minimum possible transport, supply and usage restrictions, and accommodate national and international limitations on use. Chemicals that are excessively expensive to procure or that have restricted availability in OECD member countries were therefore avoided when possible. Examples of chemicals with restricted availabilities include all chemicals listed under the Stockholm Convention on Persistent Organic Pollutants (POPs) (http://chm.pops.int/TheConvention/ThePOPs/TheNewPOPs/tabid/2511/Default.aspx, accessed 10 December 2021), and certain classes of controlled substances, such as drugs with abuse potential for example anabolic steroids and cannabinols. Chemical mixtures containing undefined or variable chemical constituents, including (stereo)isomers and racemic mixtures were also avoided when possible, due to the potential for between batch variability. Data and data sources that were considered to be of inadequate quality, due to lack of information regarding the successful establishment of the assay(s), or poor reproducibility (Taxvig 2020; Franzosa, Bonzo et al., 2021), were not utilised.

From the candidate pool of chemicals, a proposed set of proficiency validation chemicals was selected to enable adequate coverage of structural diversity but also a representative selection of chemicals from relevant sectors (including industrial chemicals, pesticides, and food additives) that was practically possible, on the basis of publicly available scientific literature. Importantly, to be able to fully evaluate the functioning and reliability of the test method, the proposed proficiency chemicals were also chosen to try to ensure the inclusion of a sufficient number of negative chemicals (a minimum of 25% of total tested), with the range of positive chemicals that would adequately probe the ability of the test method to detect individual induction of each of the four measured CYPs (CYP1A1/1A2, CYP2B6, and CYP3A4).

Using a targeted literature search and the selection criteria detailed in the methodology section, an initial pool of potential proficiency candidates consisting of a total of 23 chemicals were identified, and a tabular listing of these chemicals which includes summaries of available CYP activity data and chemical structure images are provided in Table 1, Parts A and B. From these candidates, on the basis of the review exercise, the following 13 chemicals were selected to augment the current proficiency chemical list, consisting of pharmaceuticals and to be proposed as additional proficiency chemicals for further validation experiments: tebuconazole, benfuracarb, atrazine, cypermethrin, chlorpyrifos, perfluorooctanoic acid (PFOA), bisphenol A (BPA), N,N-diethyl-m-toluamide (DEET), benzo-[a]-pyrene (B [a]P), fludioxonil, malathion, triclosan and caffeine (See Table 1 Part A). These 13 chemicals have a diverse range of structures and molecular weights and include representative examples of industrial chemicals, pesticides, and food and cosmetics additives. Excepting caffeine, all the chemicals display a degree of lipophilicity (experimental log P ranging from 1.97 to 7.75). Any potential solubility issues will be evaluated and addressed by solubility and cytotoxicity assessment that are part of the planned validation augmentation study design. These data will be reported following completion of the planned experiments.

The selected set contains the following number of expected enzyme inducers: 7 x CYP1A1/1A2 (four low, two strong, one uncertain), 6 x CYP2B6 (two low, two moderate, two uncertain), and 8 x CYP3A4 (four low, two moderate, one strong, one uncertain). These will allow for a sufficient evaluation of the ability of the assay to accurately measure the activity of all four CYP isoforms that are covered by this assay. Additionally, the selected set also contains two expected inhibitors for CYP1A1, CYP2B6, and CYP3A4, which will allow also for an evaluation of the performance of the test method in detecting CYP enzyme inhibitors. For validation experiments, whilst it is considered good practice to include a proficiency chemical set that contains at least 25% of negative chemicals, unfortunately it was not possible to fully meet this criterion for this augmented chemical set, as sufficient data regarding no CYP activity was only available for two of the 23 candidate pool chemicals. Therefore, the augmentation chemical set contains two chemicals that are expected to have no effect on CYP1A1/1A2 activity, but no additional non pharmaceutical chemicals that are expected to have no effect on CYP2B6 or CYP3A4 activity. Following the additional chemical augmentation validation confirmatory testing, the relative potencies of the chemicals in Table 2 will be consolidated.

Validation data for the test method has previously been generated for the 12 pharmaceutical proficiency chemicals shown in Table 2, and in silico chemical space evaluations are reported to demonstrate that the structures of the latter pharmaceuticals are representative of EU REACH (https://echa.europa.eu/regulations/reach/legislation), Drugbank, and Tox21 listed chemicals (Bernasconi, Pelkonen et al., 2019). Overall, the total set of 25 proficiency chemicals (including the 12 pharmaceuticals and the proposed 13 additional industrial, pesticidal, and food additive chemicals) is considered to be sufficiently diverse and representative of OECD TG Programme relevant chemical classes to facilitate draft TG approval at the OECD.

CYP induction data can indicate whether and to what extent a chemical is likely to undergo CYP-mediated metabolism, and results showing significant CYP induction could be used as an indicator that (pre-)incubation steps to account for in vivo metabolism should be included in any toxicity assays lacking metabolic competence. In relevant situations, CYP induction and inhibition data could facilitate the selection of optimal in vivo test chemical doses for human and other animal studies, and also indicate the possible involvement of CYP metabolism/metabolites in adverse or PK effects. Perturbations in the levels of endogenous chemicals that are metabolised by CYPs are associated with several adverse effects, and CYP induction data could also be used to support the contribution of CYP-mediated mechanisms in adverse outcome pathways and IATAs as shown in Figure 1B. Furthermore, the performance of a number of QIVIVE/PBPK models and PK databases such as MetaPath (Kolanczyk, Schmieder et al., 2012) would be substantially expanded by the incorporation of available CYP induction data. It would be particularly beneficial to generate CYP data for food chemical classes for example, as in Europe, little mammalian in vivo data is available for these.

Moreover, CYP induction and inhibition data are currently commonly used to predict potential CYP mediated clinical PK drug interactions for pharmaceuticals (EMA 2012; US FDA, 2020) and the availability and use of a validated and adopted OECD CYP induction test method, that will fall under the Mutual Acceptance of Data agreement, will therefore also be valuable for drug discovery and regulatory application for pharmaceutical CYP drug interaction evaluations.

Relevant in vitro metabolism data would greatly benefit several specific IATAs, including the developmental neurotoxicity (Bal-Price, Crofton et al., 2015) and non-genotoxic carcinogenicity (Jacobs, Colacci et al., 2020) IATAs currently under development, and the OECD Conceptual Framework for Endocrine Disruptors (updated OECD 2018). With respect to the latter, as proposed in 2013 (Jacobs, Laws et al., 2013), in vitro metabolism test method can be added at level 2 of the Endocrine Disruptor Conceptual Framework as shown in Figure 1A,B below, but will also have great utility in informing all levels of the Conceptual Framework. This will begin to accommodate the additional in vitro assay flexibility needs for the regulatory identification of endocrine disruptors (Solecki, Kortenkamp et al., 2017), by filling the metabolism translational gap between in vitro level 2 assays and the WHO definition of endocrine disruptor, as “an exogenous substance or mixture that alters the function(s) of the endocrine system and consequently causes adverse effects in an intact organism, or its progeny, or (sub) populations”. Quoting from the consensus paper ‘a) Alterations of the function of the endocrine system may arise from interaction with hormone receptors, changes in circulating levels of the hormone, and from the impact of chemical(s) on hormone synthesis, transport, metabolism and other factors’. c) The term “intact organism” is understood to mean that the effect would occur in vivo, either observable in a test animal system, epidemiologically or clinically. However, it does not necessarily mean that the adverse effect has to be demonstrated in an intact test animal, but may be shown in adequately validated alternative test systems predictive of adverse effects in humans and/or wildlife‘ (Solecki, Kortenkamp et al., 2017).

In addition to mediating detoxification, CAR, PXR and AhR have been implicated in the regulation of a broader range of physiological functions (Kretschmer and Baldwin 2005; Wang and Tompkins 2008; Yi, Fashe et al., 2020), where dysregulation can lead to adverse effects (Hakkola, Bernasconi et al., 2018) and receptor and CYP induction have well documented roles for instance in inflammation (Christmas, 2015; Rubin, Janefeldt et al., 2015), cholestasis, steatosis (Gomez-Lechon, Jover et al., 2009), hepatotoxicity (Woolbright and Jaeschke 2015), carcinogenesis (De Mattia, Cecchin et al., 2016; Pondugula, Pavek et al., 2016; Fucic, Guszak et al., 2017), and thyroid disruption (OECD 2006; OECD 2014). Thus, in these cases, induction of specified CYP enzymes may serve as a biomarker for key events associated with adverse health effects. In addition, the use of CYP induction data, in combination with assays that can address foetal and early life chemical exposure mediated via the placenta, such as the hPlacentox human placental JEG-3 cell line model (Rat, Olivier et al., 2017; Olivier, Wakx et al., 2021) will assist regulatory in vitro tool development, reducing metabolism uncertainties and improving the evidence base to address ‘disruption of the programming role of hormones during prenatal and postnatal development [that] can cause adverse effects that do not become evident until later in life’ (Solecki, Kortenkamp et al., 2017). CYP3A induction increases by approximately 2-fold during pregnancy, and has been shown to be mediated by cortisol plasma concentrations during pregnancy (Sachar, Kelly et al., 2019).

CYP induction testing is a critical initial testing aspect in IATAs such as that for non-genotoxic carcinogenicity (Figure 1B).

The HepaRG CYP enzyme induction test method as validated thus far can address CYP1A, 3A4 and 2B6, as primary targets. Looking forward, assays for additional CYP isoforms will be needed for specific IATAs, as identified in Figure 1B, and for metabolism evaluations for chemicals that are metabolised by other inducible CYPs. For instance, an assay for CYP2E1 is being assessed within the OECD non-genotoxic carcinogenicity IATA expert group. Furthermore, among the xenobiotic-metabolizing CYPs, CYP2D6 results in a large contribution of genetic variation to the interindividual variation in CYP enzyme activity (Ingelman-Sundberg, Sim et al., 2007), and an FDA guideline on in vitro drug interaction testing of pharmaceuticals recommends also evaluating induction of CYP2C8, 2C9, 2C19 (US FDA, 2020). Thus, while inclusion of additional CYP enzymes (such as CYP2D6) and contribution of nuclear receptors such as the glucocorticoid receptor (GR) which induces CYP2D6 (Farooq, Kelly et al., 2016), and PPARs (which induce CYP4A, Kliewer, Lehmann et al., 1999, Waxman, 1999) would be useful to explore in the near future, the scope of this study was to augment the chemical applicability domain of the validated draft TG without de novo validation of additional targets.

It is noted that GR is functional in the HepaRG cell line (Hart, Li et al., 2010; Farooq, Kelly et al., 2016; Sachar, Kelly et al., 2019) and thus subsequent additional testing with glucocorticoids (such as prednisone, prednisolone, cortisone, corticosterone, dexamethasone, betamethasone, triamcinolone, 6-methylprednisolone, 21-deoxy cortisol, deflazacort and hydrocortisol) could be run in the assay to develop further model applications, to ascertain glucocorticoid activity via GR, especially when running the CYP HepaRG assay in relation to in vitro GR transactivation and placental models, or for inclusion of such information in a metabolic disruption IATA.

Moving beyond molecular initiating events, to subsequent key events, there will need to be specific considerations for the direct and indirect CYP activity in relation to inflammation for example, for carcinogenicity.

Following successful adoption and application of the HepaRG CYP enzyme induction test method, it would be useful to develop complementary components of in vitro human metabolism systems into OECD TGs, including induction/inhibition assays for additional CYP isoforms, and assays for Phase II metabolism, metabolite profiling, and metabolic transformation pre-incubation steps (as discussed above). The use of human microsomes and/or S9 mixes, or cryopreserved primary human hepatocytes (PHH) have utility in research and drug discovery, but for chemical hazard assessment purposes, cryopreserved PHH with very wide ranging variability in responses, had reproducibility issues in a validation exercise (Bernasconi, Pelkonen et al., 2019), and following OECD peer review, were considered too variable for TG development. There are also (unknown) viral transmission concerns with the use of primary human tissues in routine chemical testing.

The HepaRG CYP enzyme induction test method was agreed by the OECD member countries to be the best (longer term) option to take forward for regulatory use in 2008 (Jacobs, Janssens et al., 2008; OECD 2008; Jacobs, Laws et al., 2013), and following successful validation with pharmaceuticals (Bernasconi, Pelkonen et al., 2019) it is the most ready and reliable system available, and superior in validation performance to cryopreserved hepatocytes. A variety of additional non validated in vitro human liver on a chip and 3D liver organoid systems are currently in development, but the complexity, between batch variability, and lack of validation data for these models (Telles-Silva, Pacheco et al., 2022) mean that they are not ready for test guideline development at present. However available literature evidence indicates that HepaRG cells are well suited to all of these applications, as mRNA for many key proteins involved in xenobiotic metabolism are expressed in differentiated HepaRG cells, including xenobiotic sensing nuclear receptors (AhR, PXR, CAR, and PPAR-α), CYPs (CYP1A1, 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, 3A4, 3A5, 3A7), other Phase I metabolic enzymes [including various isoforms of alcohol dehydrogenase, aldehyde dehydrogenase, and flavin-containing monooxygenase], and Phase II metabolic enzymes [including various isoforms of glutathione S-transferase, UDP-glucuronosyltransferase, N-acetyltransferase, and sulfotransferase] (Aninat, Piton et al., 2006; Guillouzo, Corlu et al., 2007; Josse, Aninat et al., 2008; Antherieu, Chesne et al., 2010; Hart, Li et al., 2010). Enzymatic activity for CYP1A1/1A2, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, and 3A4 has also been confirmed to be present (Aninat, Piton et al., 2006; Josse, Aninat et al., 2008; Antherieu, Chesne et al., 2010; Lubberstedt, Muller-Vieira et al., 2011). Furthermore, chemical-induced induction and inhibition of CYP1A1/1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 3A4 has been demonstrated in HepaRG cells (Aninat, Piton et al., 2006; Josse, Aninat et al., 2008; Kanebratt and Andersson 2008; Turpeinen, Tolonen et al., 2009; Antherieu, Chesne et al., 2010; Yajima, Uno et al., 2014). The metabolic profile data generated for several chemicals in HepaRG cells have also been shown to be equivalent to that produced using PHH, including aflatoxin B1 and acetaminophen (Aninat, Piton et al., 2006) but without the inherent biological variability observed with different batches of PHH. Intrinsic clearance values generated for a large number of reference drugs in HepaRG cells have also been shown to be equivalent to the values generated in PHH (Lubberstedt, Muller-Vieira et al., 2011; Zanelli, Caradonna et al., 2012). In the future a potentially promising approach to explore would therefore be to develop the HepaRG cell system into an all-in-one system/assay providing all the desired abovementioned in vitro metabolic functionalities.

For now however, it is really important to first address the outstanding steps required to enable the successful adoption of the HepaRG CYP enzyme induction test method as a TG, and the work described herein provides the essential concrete chemical selection and applications step required to enable international TG progress within an immediate timeframe. Following the planned additional validation experiments with this candidate chemical selection list there will be an evidence basis upon which to refine the chemical list for use as additional proficiency chemicals for the HepaRG CYP enzyme induction test method, and to do any further analyses that may be warranted.