Human hepatocyte incubation is an excellent in vitro system for drug biotransformation research due to its ability to reflect metabolism in the intact human liver. Human hepatocytes are isolated living cells containing the complete repertoire of phase I and phase II drug metabolizing enzymes, necessary cofactors, uptake and efflux drug transporters, and drug binding proteins (Diao and Huestis, 2017).

At the National Institutes of Health/National Institute on Drug Abuse, we established a strong collaboration with the United States Drug Enforcement Administration (DEA) to identify optimal marker metabolites of new SC. When DEA seizures of a particularly toxic new SC they would purify and provide us with the SC. We first determined the SC's half-life by quantifying the disappearance of the parent compound during incubation with HLM, in order to best design the SC human hepatocyte incubation experiment. By far, the most challenging aspect was the high-resolution mass spectrometry (HR-MS) analysis to identify the full spectrum of metabolites, and those metabolites that best differentiated the SC from its closest analogs. We also attempted to obtain authentic human urine samples following specific SC ingestion through international collaborations, enabling comparison of in vitro and in vivo metabolites and clarifying the targets for SCs drug testing.

This workflow was successful in predicting major urinary metabolites of many SCs, including AB-PINACA/5F-AB-PINACA (Wohlfarth et al., 2015), AB-FUBINACA (Castaneto et al., 2015), FDU-PB-22/FUB-PB-22 (Diao et al., 2016a), and NM-2201 (Diao et al., 2017b) etc. For AB-FUBINACA, the prominent metabolite following human hepatocyte incubation was amide hydrolysis product M11 (Figure 2A). Consistently, M11 was also the most abundant metabolite in human urine after β-glucuronidase hydrolysis. Besides M11, M6 (aliphatic hydroxylation) and M7 (amide hydrolysis product of M6) were the primary metabolites after β-glucuronidase hydrolysis in both human hepatocyte incubation and in human urine following AB-FUBINACA intake. For, AB-PINACA, metabolites A23 and A16 were the major metabolites in human hepatocyte incubation and in human urine after β-glucuronidase hydrolysis (Figure 2B).

HLM incubation is currently the most popular in vitro metabolism model, providing an affordable method to identify target metabolites following CYP and UGT metabolism. The popularity of this in vitro model is attributed to its simplicity and widespread availability, and the ability to determine specific metabolizing isozyme(s) by studying their activity in the presence of specific inhibitors (Bickett et al., 1993). HLM are hepatocyte endoplasmic reticulum vesicles prepared by differential centrifugation. HLM contain primarily CYP, UGT and esterase enzymes, accounting for about 95% of clearance mechanisms for the top 200 drugs prescribed in the United States in 2002 (Williams et al., 2004).

However, major metabolite discrepancies may occur between those noted in HLM incubations and those found in human urine following SC intake, i.e., in the metabolism of 5F-AKB-48 and AM-2201 (Figure 3A; Diao and Huestis, 2017). The probable reason is that the enzyme responsible for oxidative defluorination, the primary metabolic pathway for a fluoropentyl chain SC, is not located in HLM. In the case of AM-2201, the HLM metabolic profile did not match satisfactorily with metabolites identified in authentic urine specimens. AM-2201 HLM incubation produced N-desfluoropentyl, mono-hydroxyl, di-hydroxyl, and the most abundant dihydrodiol metabolite (Figure 3A; Sobolevsky et al., 2012). One researcher self-administered 5 mg AM-2201 to identify the major human urinary metabolites (Hutter et al., 2013). Four major metabolites were identified in the post administration urine samples, JWH-018 N-(5-OH-pentyl), JWH-018 N-pentanoic acid, AM-2201 6-OH-indole, and AM-2201 N-(4-OH-pentyl). The highest concentrations were for JWH-018 N-pentanoic acid and JWH-018 N-(5-OH-pentyl); however, AM-2201 shared major metabolites with JWH-018, i.e., JWH-018 N-(5-OH-pentyl) and JWH-018 N-pentanoic acid. Thus, it is challenging to differentiate AM-2201 from JWH-018 intake based on the detection of these two metabolites in urine specimens. To distinguish AM-2201 intake from JWH-018, detection of AM-2201 6-OH-indole and AM-2201 N-(4-OH-pentyl) is essential.

Recently, Mogler et al. also reported discrepancies in the metabolism of SC CUMYL-PEGACLONE between HLM and human urine samples (Mogler et al., 2018). Following HLM incubation, M15, a mono-hydroxylation metabolite on the pentyl chain, was found in the highest concentration, and no parent, CUMYL-PEGACLONE was detected in any of the human urine samples (n = 30). Twenty-two different phase I CUMYL-PEGACLONE metabolites were detected in human urine. Metabolic pathways included mono-hydroxylation, di-hydroxylation, dehydrogenation, N-dealkylation, β-oxidation (pentyl side chain to a propionic acid metabolite), carbonyl formation at the pentyl side chain, and combinations of these biotransformations. The most abundant two metabolites were identified as M20 and M09 (Figure 3B); M20 was a metabolite with mono-hydroxylation on the γ-carbolinone core and M09 was a further pentyl chain carbonylated metabolite from M20. The authors proposed M20 and M09 as sensitive and specific urinary markers to prove intake of CUMYL-PEGACLONE. The metabolism of CUMYL-PEGACLONE was unexpectedly different from previous cumyl-derivatives (Kevin et al., 2017; Özt"urk et al., 2018; Staeheli et al., 2018), which were mainly hydroxylated on the pentyl side chain, whereas CUMYL-PEGACLONE was mainly hydroxylated on the γ-carbolinone core.

In pharmaceutical industry, early prediction of possible toxic metabolites is important to preclinical and clinical decision-making (Afzelius et al., 2007; T'Jollyn et al., 2011). Identification of possible toxic metabolites in the drug discovery stage enables earlier decisions on elimination of drug candidates from further development. Adams et al utilized metabolism prediction software to suggest that the urinary AMB-FUBINACA de-esterified acid metabolite could identify AMB-FUBINACA intake in a mass intoxication outbreak (Adams et al., 2017).

In 2011, T'jollyn et al. evaluated Meteor (Lhasa Ltd., Leeds, UK), MetaSite (Molecular Discovery Ltd., Middlesex, UK), and StarDrop (Optibrium Ltd., Cambridge, UK) software drug metabolism tools (T'Jollyn et al., 2011). Meteor is a rule-based (empirical) software tool (Langowski and Long, 2002). MetaSite is an automated docking model with a reactivity correction considering the reactivity components of an atom related to heme that is designed to predict phase I CYP450 metabolism (Cruciani et al., 2005). StarDrop uses a quantum mechanical approach for the prediction of the relative involvement of CYP3A4, 2D6, and 2C9 of the query compound (Earnshaw, 2010). Its mechanism is based on calculation of the energy barrier to electron removal, considered to be the rate-limiting step in product formation. The authors evaluating the state-of-the-art metabolite prediction software concluded that it has many advantageous features but needs refinement to obtain acceptable prediction profiles. Synergistic use of different software packages could prove useful. However, it is not practical for forensic laboratories to purchase expensive in silico software licenses.

In our previous investigation of SC THJ-2201 metabolism, we first utilized the MetaSite in silico prediction software prior to hepatocyte incubation and HR-MS analysis. The software predicted 8 first-generation and 7 second-generation metabolites (Diao et al., 2016c). The top predicted metabolites were N-depentyl-THJ-018, 1′-OH-THJ-2201, 1′-carbonyl-THJ-2201, and pent-1′-enyl-THJ-2201 (Figure 4). However, these top predicted metabolites were inconsistent with the scenario in human hepatocyte incubation and human urine. After incubating THJ-2201 with human hepatocytes, 27 metabolites were generated, with THJ-018 pentanoic acid (F25) and 5′-OH-THJ-018 (F26) as the most abundant metabolites. F26 was produced by oxidative defluorination of THJ-2201, with further oxidation to F25. In a SC screening method, THJ-018 N-pentanoic acid was the target marker metabolite for THJ-2201 intake (Gundersen et al., 2019).

MetaSite primarily focuses on CYP450 mediated metabolism and does not simulate reactions mediated by non-CYP450 oxidases, such as aldehyde oxidase. The performance of MetaSite for THJ-2201 metabolism most likely missed the oxidative defluorination reaction because it was not catalyzed by CYP450.

Clinical studies on SC effects are hampered by the lack of preclinical toxicology data. It also is impractical to conduct clinical trials on each new SC due to the constant introduction of novel SC into the illegal drug market. Thus, many researchers study SC pharmacodynamics and pharmacokinetics in vivo animal models. AM-2201 metabolism in rats has similarities and differences from that in humans (Jang et al., 2014). Predominant AM-2001 metabolites after hydrolysis of rat urine with β-glucuronidase were JWH-018 N-pentanoic acid and AM-2201 6-OH-indole (Figure 5A). Two additional metabolites, JWH-018 N-(5-OH-pentyl) and AM-2201 N-(4-OH-pentyl), were detected in lower abundance in rat urine. These four metabolites also were the top four metabolites in human urine after self-administration of 5 mg AM-2201 (Hutter et al., 2013). However, the relative abundance of these 4 metabolites in human urine was quite different from that in rat urine. In human urine, JWH-018 N-pentanoic acid and JWH-018 N-(5-OH-pentyl) were the most abundant metabolites (Figure 5A), while AM-2201 N-(4-OH-pentyl) and AM-2201 6-OH-indole were observed in lower concentrations.

Recently, Kevin et al. investigated MN-18 metabolism (Kevin et al., 2018) in rat after intraperitoneal administration of 3 mg/mL MN-18. Only two metabolites were identified in rat urine, the hydroxylated metabolite M3 and its glucuronide M10 (Figure 5B). However, following human hepatocyte incubation with MN-18, 13 metabolites were observed (Figure 5B), with the top 3 metabolites 1-pentyl-1H-indazole-3-carboxylic acid (M5), naphthalene hydroxylated MN-18 (M7), and pentyl-carbonylated MN-18 (M12) (Diao et al., 2017a). To reduce confusion, metabolite names were directly taken from the original literature; if the same nomenclature occurred in both literatures, the font of the metabolites from the latter one is in italicized.

The zebrafish (Danio rerio) was initially introduced by Streisiger as an animal model in genetic studies in the early 1980s; the zebrafish is a small teleost (3–4 cm) typically from sweet waters (Streisinger et al., 1981). Some important advantages are associated with zebrafish for research, such as small size, easy maintenance, low cost of breeding, and high reproductive rate. Zebrafish is emerging as a predictive vertebrate animal model for in vivo assessment of drug efficacy, toxicity, and safety. Interestingly, some studies evaluated the ability of zebrafish larvae at different stages of ripening to generate xenobiotic metabolites (Alderton et al., 2010; Chng et al., 2012).

The similarities between adult zebrafish and human metabolism were evaluated on the generation of phase I metabolites of the sports doping agent—sibutramine (de Souza Anselmo et al., 2017). Adult zebrafish produce several sibutramine metabolites (Figure 6A), including demethylsibutramine (nor-sib), bi-demethylsibutramine (bis-nor-sib), hydroxylated nor-sib (OH-nor-sib1 and OH-nor-sib2), and hydroxylated bis-nor-sib (OH-bis-nor-sib1 and OH-bis-nor-sib2). These metabolites were identified in zebrafish culture solution samples after hydrolysis with β-glucuronidase. The authors claimed that the study demonstrated that adult zebrafish could absorb, oxidize, and excrete several metabolites in a manner similar to humans. However, although these metabolites were observed in human urine, the major metabolites were quite different. In zebrafish incubation samples, the top two metabolites were demethylsibutramine (nor-sib) and bi-demethylsibutramine (bis-nor-sib), but, these two metabolites were only minor metabolites in human urine (Figure 6B). Eight phase II carbamoyl glucuronides of nor-sib and bis-nor-sib were the main metabolites in human urine (Link et al., 2006). Of note, the author did not hydrolyze the human urine sample with β-glucuronidase solution, so we cannot conclude whether nor-sib/bis-nor-sib or OH-nor-sib/OH-bis-nor-sib were the major human urine marker metabolites.

The use of microorganisms, and particularly fungus C. elegans as models to study human metabolism is well established and provides another approach to producing SC metabolites (Asha and Vidyavathi, 2009; Murphy, 2015). A review on C. elegans metabolism reported that the fungus has some similarities with human metabolism for various drugs (Asha and Vidyavathi, 2009). This cost-efficient system is capable of producing large quantities of metabolites. In addition, the fungus culture is easy to grow and can be transferred to new agar plates with ease (Choudhary et al., 2007).

Watanabe et al. made significant contributions on investigating the metabolism of SCs with the C. elegans model, and evaluated the similarity of metabolites generated in this incubation system with those observed in human metabolism (Watanabe et al., 2016, 2017, 2018a,b). SC metabolites for JWH-018, AM2201, JWH-073, PB-22, 5F-PB-22, XLR-11, and UR-144 were evaluated in the C. elegans incubation model. Some of the major phase I human metabolites of previously investigated SCs were documented, although this model was less effective in producing phase II human metabolites. The authors proposed that the fungus C. elegans is a complementary model to study human metabolism of novel SC and can generate sufficient SC metabolites for definitive structure elucidation.

When comparing C. elegans SC metabolites to human SC metabolites, there are similarities for some SCs but there were inconsistencies as well. For JWH-018 (Figure 7A), JWH-018 N-(4-OH-pentyl) and JWH-018 N-pentanoic acid were the top two metabolites in human urine collected from individuals with SC intoxication (Diao and Huestis, 2017). In the C. elegans incubation system, JWH-018 N-(4-OH-pentyl) was also the most abundant metabolite, matching quite well with the human metabolites (Watanabe et al., 2016). But the other abundant human metabolite, JWH-018 N-pentanoic acid, was a minor metabolite in C. elegans incubation model. For AM-2201 (Figure 7B), the primary human urinary metabolites were JWH-018 N-(5-OH-pentyl), JWH-018 N-pentanoic acid, AM-2201 6-OH-indole, AM-2201 N-(4-OH-pentyl). Although these metabolites were detected in the C. elegans incubation model, they had low abundance (Watanabe et al., 2016). In the C. elegans incubation model, the top two metabolites were Mc25 (AM-2201 dihydrodiol) and Mc47 (JWH-073). This is most likely attributed to the lack or low activity of the enzyme responsible for oxidative defluorination, which phenomenon was also observed in HLM incubation system.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.