Hepatic Bioactivation of Skin-Sensitizing Drugs to Immunogenic Reactive Metabolites.

The clinical use of some drugs, such as carbamazepine, phenytoin, and allopurinol, is often associated with adverse cutaneous reactions. The bioactivation of drugs into immunologically reactive metabolites by the liver is postulated to be the first step in initiating a downstream cascade of pathological immune responses. Current mechanistic understanding and the ability to predict such adverse drug cutaneous responses have been partly limited by the lack of appropriate cutaneous drug bioactivation experimental models. Although in vitro human liver models have been extensively investigated for predicting hepatotoxicity and drug-drug interactions, their ability to model the generation of antigenic reactive drug metabolites that are capable of eliciting immunological reactions is not well understood. Here, we employed a human progenitor cell (HepaRG)-derived hepatocyte model and established highly sensitive liquid chromatography-mass spectrometry analytical assays to generate and quantify different reactive metabolite species of three paradigm skin sensitizers, namely, carbamazepine, phenytoin, and allopurinol. We found that the generation of reactive drug metabolites by the HepaRG-hepatocytes was sensitive to the medium composition. In addition, a functional assay based on the activation of U937 myeloid cells into the antigen-presenting cell (APC) phenotype was established to evaluate the immunogenicity potential of the reactive drug metabolites produced by HepaRG-derived hepatocytes. We showed that the reactive drug metabolites of known skin sensitizers could significantly upregulate IL8, IL1β, and CD86 expressions in U937 cells compared to the metabolites from a nonskin sensitizer (i.e., acetaminophen). Thus, the extent of APC activation by HepaRG-hepatocytes conditioned medium containing reactive drug metabolites can potentially be used to predict their skin sensitization potential.

Introduction

Cutaneous adverse reactions to drugs administrated into systemic circulation remains a severe problem because it can result in significant morbidity and mortality such as Steven Johnson Syndrome (SJS) and toxic epidermal necrolysis (TEN).[1−3] Increasing evidence shows that this type of cutaneous drug reactions is frequently associated with the formation of reactive drug metabolites, which have been detected in circulating human blood plasma of patients.[4−7] It is postulated that these reactive metabolites can act as antigens when bound to plasma proteins (i.e., a hapten), which in turn activate antigen-presenting cells (APCs) in the blood circulatory system that will stimulate a cascade of downstream cytotoxic effector immune responses to attack epidermal keratinocytes.[8−11] However, there is a lack of reliable cellular experimental models to mimic these different cellular processes, which hampers efforts to understand and predict cutaneous drug reactions.

The in vitro formation of immunogenic reactive drug metabolites recapitulates an essential first step in modeling cutaneous drug reactions. Reactive drug metabolites, such as 2 hydroxyl carbamazepine (2-OH CBZ) and 3 hydroxyl carbamazepine (3-OH CBZ) from carbamazepine (CBZ),[12] 5-(4′-hydroxyphenyl)-5-phenylhydantoin (p-HPPH) from phenytoin (PHT),[13] and oxipurinol from allopurinol,[14−16] are often generated during Phase I metabolism of the parent compound by cytochrome P450 (CYP) enzymes and other enzymes in the liver as either the primary or secondary metabolites. To date, the bioactivation of paradigm skin-sensitizing drugs into their reactive metabolites has only been demonstrated in rat microsomes or rat hepatocytes[17] and human liver microsomes.[18] Since the CYP enzymatic profiles are significantly different between humans and animals, a human hepatocyte model would more reliably recapitulate the metabolism of a potential skin sensitizer into its constituent reactive metabolites.

HepaRG-hepatocytes (HepaRG-Heps) are derived from a bipotent human liver progenitor cell line and are increasingly being used as an alternative to primary human hepatocytes (PHHs) due to their unlimited expansion capacity and amenability to long-term in vitro culture.[19,20] Importantly, HepaRG-Heps consists of several essential CYP enzyme activities, which are responsible for generating the reactive metabolites during drug bioactivation that could probably induce skin sensitization.[21,22] To date, HepaRG-Heps have been extensively used to model drug metabolism in the context of predicting hepatotoxicity and drug–drug interactions. For example, paradigm skin sensitizers, such as carbamazepine,[22] phenytoin,[23] and allopurinol,[24] are being investigated using HepaRG-Heps to discover their CYP induction potential and consequent adverse drug–drug interactions.[25,26] However, the applicability of HepaRG-Heps in the generation of reactive drug metabolites with functional readouts related to skin sensitization potential has not been explored.

In this study, we established a HepaRG-hepatocyte model and highly sensitive accompanying liquid chromatography-mass spectrometry (LCMS) analytical assays to investigate the production of different reactive metabolite species from paradigm skin sensitizers under varying medium conditions. In addition, a functional assay, which was based on the activation of U937 monocytes into antigen-presenting cell (APC) phenotypes[27−29] was implemented to evaluate the immunogenicity potential of the reactive drug metabolites produced by HepaRG-Heps to distinguish between the three known skin-sensitizing drugs, namely, carbamazepine, phenytoin, and allopurinol, from a nonskin-sensitizing drug (acetaminophen, APAP).

Results

Functional Characterization of HepaRG-Derived Hepatocytes

Increasing evidence suggests that reactive drug metabolites are responsible for causing severe skin sensitization reactions, such as SJS and TEN, which necessitates a need to develop a cutaneous bioactivation model. Since HepaRG-derived hepatocytes (HepaRG-Heps) have shown comparable metabolic functions to primary human hepatocytes,[19] we hypothesize that they are capable of generating different skin-sensitizing drug metabolites, which can be detected through highly sensitive LCMS analytical assays.

After 14 days of differentiation, the bipotent HepaRG progenitor cells formed large cords of hepatocyte-like cells, which exhibited the characteristic polygonal morphology of primary human hepatocytes (PHHs), and were surrounded by biliary epithelial-like cells as shown in Figure A.[30] The identities of the 2 cell populations were confirmed by the hepatic marker (CYP3A4) and biliary marker (CK19) (Figure B). The derivation efficiency of HepaRG-Heps in the cultures was approximately 50% as reported by Cerec et al.[31] The albumin secretion of HepaRG-Heps could reach a comparable if not higher level than that of PHH after the first 7 days of differentiation (Figure C). The average albumin production rate of PHH was approximately 100 μg/day/106 cells, while that of HepaRG-Heps on days 7, 9, and 11 post differentiation were 99 ± 30, 200 ± 50, and 109 ± 40 μg/day/106 cells, respectively (Figure C). This indicated that functional HepaRG-Heps could be maintained in vitro for an extended period of time. The transcriptional expressions of various liver-specific genes (CYP1A2, CYP2B6, CYP3A4, Albumin, HNF-4α, and PXR) were also similar to those of PHH (Figure D).

Figure 1

Characterization of HepaRG-derived hepatocytes (HepaRG-Heps) as surrogates for primary human hepatocytes (PHHs) in drug bioactivation assessment in vitro. (A) Phase-contrast images demonstrated the morphology of HepaRG-Heps (up) and PHHs (down). Scale bars in HepaRG-Heps = 50 μm and PHHs = 100 μm. (B) Immunofluorescence images of HepaRG-Heps stained with the 4′,6-diamidino-2-phenylindole (DAPI) (blue), biliary marker CK19 (green), and hepatic marker CYP3A4 (red). Scale bar = 100 μm. (C) Albumin production in HepaRG-Heps (black bars) after D3, D5, D7, D9, and D11 of differentiation. Black dotted line represents the albumin secretion of PHHs from three different lots of culture for 24 h. (D) Quantitative polymerase chain reaction (qPCR) analysis depicted the hepatic marker gene expression in PHHs (gray bars) and HepaRG-Heps (black bars). (E) CYP1A2 activity and (F) CYP3A4 activity of PHHs (gray bars) and HepaRG-Heps (black bars). Data for HepaRG-Heps are average ± standard error of the mean (SEM) of three independent batches of differentiation. Data for PHHs were average ± SEM of three different lots of cryopreserved PHHs. Asterisks denote statistically significant differences (Student t-test, *p < 0.05).

Many of the known skin-sensitizing metabolites are products of phase 1 metabolism mediated by CYP enzymes. Therefore, we also examined the basal CYP1A2 and CYP3A4 activities of HepaRG-Heps by evaluating the conversion of enzyme-specific substrates (phenacetin for CYP1A2 and midazolam for CYP3A4) into their metabolic products.[32] The CYP1A2 activity of HepaRG-Heps was significantly higher than that of PHHs (HepaRG-Heps: 36 ± 2.8 ng/mL/min/106 cells vs PHH: 23 ± 3.7 ng/mL/min/106 cells) (Figure E). At the same time, HepaRG-Heps also exhibited higher midazolam hydroxylation activity catalyzed by CYP3A4 (105 ± 7.4 ng/mL/min/106 cells) than the PHHs (63 ± 11.9 ng/mL/min/106 cells) (Figure F). By the analysis of CYP enzyme-dependent substrate conversion, HepaRG-Hep preserved higher major human-relevant CYP activities (CYP1A2 and CYP3A4) compared to cultured PHH, which agreed with the gene expression results. Collectively, our data indicated that we are able to consistently generate functionally competent hepatocytes from a human progenitor cell line, which would provide a stable bioactivation cell model to study potential systemic skin sensitizers.

Generation and Analytical Measurement of Reactive Metabolites in Different Media

To examine the applicability of the HepaRG-Heps as a cutaneous drug bioactivation model, we treated HepaRG-Heps with three model drugs that are known to cause cutaneous reactions, including carbamazepine (CBZ), phenytoin (PHT), and allopurinol. We also established LCMS analytical assays to quantify the production of their antigenic reactive metabolites, including 2-OH CBZ, 3-OH CBZ, 5-(p-hydroxyphenyl)-5-phenylhydantoin (p-HPPH), and oxipurinol (Figure A). The drugs were administered to the HepaRG-Heps in various medium milieus to determine if their compositions may affect the detection of the drug metabolites.

Figure 2

Generation and analytical measurements of drug-reactive metabolites in different media. (A) Schematics of the bioactivation of paradigm skin-sensitizing drugs (i.e., CBZ, PHT, and allopurinol) by HepaRG-Heps to metabolites. The metabolites generated by HepaRG-Heps in either Krebs–Henseleit buffer (KHB), William E (WE), or RPMI after 6 h of incubation time were then quantified by an established liquid chromatography-mass spectrometry (LCMS) analytical assay. Generation of CBZ metabolites in 100 μM concentration of CBZ: (B) 10,11-epoxide carbamazepine (CBZ-E), (C) 2-hydroxy carbamazepine (2-OH CBZ), and (D) 3-hydroxy carbamazepine (3-OH CBZ). (E) Generation of PHT’s metabolites, hydroxyphenytoin, 5-(4′-hydroxyphenyl)-5-phenylhydantoin (p-HPPH) in 15 μM concentrations of PHT. (F) Generation of allopurinol’s metabolites, oxypurinol in 10 μM concentrations of allopurinol. Black bars with blank feature denote KHB, the bars with diagonal stripes represent William E (WE) medium and the bars with horizontal stripes demonstrated RPMI medium. Data are average ± SEM of 3 independent experiments. Asterisks denote statistically significant differences (Student t-test, *p < 0.05; **p < 0.01).

Krebs–Henseleit buffer (KHB) is a bicarbonate-based buffer, which is commonly employed for phase I and II drug biotransformation studies because they can maintain hepatocyte functions for short periods of time (30–120 min) while ensuring minimal interference in metabolite analysis and detection.[33] However, the immune responses triggered by skin-sensitizing drugs are mostly dependent on the concentration and exposure duration to antigenic reactive metabolites.[34] Therefore, it is essential for the cutaneous drug bioactivation models to generate reactive metabolites continuously for a sufficiently long duration. Being a minimal medium, KHB is not ideal for conducting longer-term drug bioactivation studies. Thus, we compared the generation of reactive drug metabolites by HepaRG-Heps in two complex media, namely, William E (WE) and RPMI basal media, compared to KHB over a period of 6 h.

We first validated the LCMS analytical method by examining its assay specificity, linearity, lower limits of quantitation (LLOQ), stability, precision, and accuracy. The assay specificity was evaluated by comparing the chromatograms of blank medium and quality control (QC) samples to the drug-treated samples in different media (i.e., WE, RPMI and KHB). Our results indicated that the composition of the different blank media did not interfere with the analysis of spiked metabolites and internal standards (IS) (Figures S1–S3). In addition, the linearity of the detection range was evaluated on three separate days with two sets of calibration curves per day. The calibration curves showed good linearity over a concentration range of 1–500 ng/mL for 3-OH CBZ, 2-OH CBZ, and CBZ-E; 1–250 ng/mL for p-HPPH; and 100–5000 ng/mL for oxipurinol, respectively, in all three media. Furthermore, Table S1 summarized all of the equations obtained from calibration curves using weighted (1/x2) least squared linear regression. The precision and accuracy as well as the recovery effect in different media for the inter-day and intra-day are shown in Table . The precision (relative standard deviation, RSD) and accuracy (relative error, RE) of the LCMS assay were within 20% at low concentrations and 1% at high concentrations of quality controls (QCs) in all metabolites. We also determined the matrix effect by normalizing LCMS readings of different metabolites spiked into different media to metabolites spiked into methanol (Table ). The ratios were within the range of 90–110%, which indicated the absence of media influence and all of the metabolites were stable in both simple and complex media. Meanwhile, we also examined the stability of metabolites in different media subjected to three different storage conditions: (i) after three freeze–thaw cycles, (ii) at room temperature for 24 h after preparation, and (iii) at −20 °C for at least 14 days and all of the results are summarized in Table S2. Similar to the RSD and RE, all of the results shown were all within 20% at low concentrations and 15% at medium and high concentrations of QCs, proving that the five reactive drug metabolites were stable in the different media under the three conditions tested.

Table 1

Precision, Accuracy, Recovery and Matrix Effects for Metabolites in the Different Matrices (media)

	spiked (ng/mL)	inter-day RSD (%)	intra-day RSD (%)	accuracy RE (%)	recovery (%)	matrix (media) effects (%)	media
2-OH CBZ	2	12.8	15.4	–12.1	83.4 ± 10.7
	50	9.6	8.9	–8.2			WE
	300	5.2	5.1	–2.4
	2	12.5	14.7	–14.9
	50	4.8	4.8	10.1			RPMI
	300	4.2	10.6	0.1
	2	5.0	4.5	5.9
	50	3.9	4.6	3.8			KHB
	300	5.1	4.8	4.2
3-OH CBZ	2	13.5	16.5	–11.2
	50	10.1	9.4	–8.9			WE
	300	5.8	5.2	5.1
	2	8.9	13.0	3.9
	50	3.4	2.6	8.1			RPMI
	300	4.1	4.4	7.4
	2	5.3	4.7	4.1
	50	3.9	4.6	1.5			KHB
	300	5.8	5.3	8.7
CBZ-E	2	13.7	17.0	–11.3
	50	8.2	8.7	–13.6			WE
	300	4.9	4.9	4.2
	2	9.1	6.5	–12.1
	50	6.2	1.4	–7.8			RPMI
	300	3.8	10.7	5.7
	2	4.0	3.9	0.1
	50	3.2	4.3	–1.2			KHB
	300	6.0	5.5	14.3
p-HPPH	2	7.3	6.2	–4.7
	50	6.7	7.2	3.6			WE
	200	6.6	7.1	–6.9
	2	19.2	17.8	4.2
	50	11.3	11.0	–6.5			RPMI
	200	4.3	4.5	4.1
	2	16.3	5.7	19.0
	50	11.4	12.1	–0.5			KHB
	200	3.6	3.8	–3.4
oxipurinol	200	5.4	6.7	2.9
	600	12.7	8.1	6.9			WE
	2500	11.0	7.5	7.4
	200	9.1	9.6	–9.7
	600	4.26	6.0	–0.9			RPMI
	2500	9.4	8.2	0.9
	200	11.3	7.9	–5.3
	600	8.9	6.3	–1.2			KHB
	2500	9.4	9.7	6.4

Next, we proceeded to assess whether the different media influenced the production of reactive metabolites by the HepaRG-Heps. It was observed that HepaRG-Heps could generate the highest levels of all three CBZ metabolites in KHB buffer (Figure B–D), while there was no significant difference in the bioactivation of PHT into p-HPPH in the three media (Figure E). For oxipurinol, the metabolite production was the highest in WE medium and the lowest in the simple medium, KHB (Figure F). It is likely that different bioactivation pathways and enzymes were involved in the conversion of parent drugs into their metabolites, and different media could maintain the different metabolic enzymes to varying extents. Nonetheless, it was ascertained that different media compositions affected the bioactivation of different reactive drug metabolites by HepaRG-Heps, but not their detection by LCMS.

Assessment of Skin-Sensitizing Potential of Reactive Drug Metabolites Generated by HepaRG-Hep

Finally, we attempted to establish a functional assay to evaluate the immunogenic potential of the reactive drug metabolites generated by HepaRG-Hep. U937 is a myeloid cell line, which is known to undergo functional and phenotypic alteration from monocytes to antigen-presenting cells (APC), indicated by the upregulation of IL8, IL1β, and CD86 after treatment with dermal skin sensitizers.[27−29] Therefore, we postulated that by exposing U937 cells to HepaRG-Heps conditioned medium that contained the reactive drug metabolites, the U937 cells can be activated into an APC phenotype if the drug metabolites are potentially antigenic skin sensitizers (Figure A).

Figure 3

Optimization of conditioned medium toward functional activation of U937 cells into APCs by 2,4-dinitrochlorobenzene (DNCB) or CBZ’s reactive metabolites. (A) Schematics illustrating DNCB treatment in different composite media with varying ratios of KHB or WE to RPMI (1:5, 1:10, and 1:20). (B) Gene expression changes in inflammatory cytokines (IL8 and IL1β) and the antigen-presenting costimulatory receptor, CD86, in U937 cells maintained in RMPI medium after 12 h (blue-gray bars), 48 h (black bars), and 72 h (purple bars) of incubation with 1 μM DNCB. (C–E) Gene expression changes after 48 h of treatment with 1 μM DNCB in different composite media relative to untreated cells: (C) IL8, (D) IL1β, and (E) CD86. Black bars denote the composition of KHB/RPMI while light gray bars denote the medium composition of WE/RPMI. The dotted lines represent the gene expression of U937 after 48 h of 1 μM DNCB treatment in RPMI medium. (F–H) Growth rate and functional response of U937 to CBZ’s reactive metabolites in different ratios of KHB/RPMI conditioned medium after 48 h of incubation. (F) Percentages of U937’s growth rate after 48 h of treatment in different ratios of KHB/RPMI. Gene expression changes in U937 for (G) IL8 and (H) IL1β after being treated with reconstituted CBZ’s metabolites and HepaRG-Heps conditioned medium in different KHB/RPMI ratios at 48 h. Data are average ± SEM of 3 independent experiments. Asterisks denote statistically significant differences (Student t-test, *p < 0.05) while ns represents no significant differences.

U937 cells are routinely maintained in RPMI medium, which was not optimal for the generation of all drug metabolites (Figure B–F). Therefore, there is a need to determine a composite medium suitable for both metabolite productions by hepatocytes as well as U937 activation. Thus, the functional activation of U937 cells by a potent skin-sensitizing compound, 2,4-dinitrochlorobenzene (DNCB), into APCs was assessed in different media. U937 cells performed the best in its own RPMI medium. When U937 cells were stimulated with DNCB in RPMI medium, there was significant upregulation in the transcriptional expressions of IL8, IL1β, and CD86 compared to untreated U937 cells (Figure B), which was in accordance with prior studies.[27] However, this functional response of U937 cells was highly sensitive to the culture media composition. When the cells were stimulated with DNCB in RPMI adulterated with varying ratios of KHB or WE media (1:5, 1:10 and 1:20), the extent of IL8, IL1β, and CD86 upregulation was significantly attenuated compared to RPMI alone (Figure C–E). We observed that the KHB/RPMI composite media performed slightly better than the WE/RPMI composite media in producing an immunological response, especially at 1:20 ratio. Importantly, we found that in the growth rate of U937 in 1:20 KHB/RPMI medium was the highest (90.6 ± 2.8%) similar to that of U937 cells maintained in RPMI medium (Figure F).

We also assessed whether the immunological response of U937 cells elicited by CBZ reactive metabolites was affected by different KHB/RPMI composite media. Both reconstituted reactive metabolites (i.e., CBZ-E, 2-OH CBZ, 3-OH CBZ, and combination of these metabolites) as well as HepaRG-hepatocyte conditioned medium were tested. In this experiment, the concentrations selected for reconstituted metabolites were based on the total amount of concentrations generated by HepaRG-Hep in different ratios of the conditioned medium as shown in Table S3. We observed that only HepaRG-Hep conditioned medium in 1:20 KHB/RPMI could produce a significant increase in IL8 (Figure G) and IL1β (Figure H) expressions compared to control cells treated with the parent drug, CBZ. It is also interesting to note that all of the concentrations of CBZ-E, 2-OH CBZ, and 3-OH CBZ in HepaRG-Hep conditioned medium at 1:20 dilution were much lower than that of the reconstituted metabolites (Table S9). Therefore, it is likely that the conditioned medium contained other reactive metabolites generated from CBZ itself or after the secondary metabolism of 2-OH CBZ and 3-OH CBZ. The results underscore the importance of having a highly functional hepatic cell model, such as HepaRG-Heps to generate a full repertoire of reactive metabolites as well as a suitable composite medium (1:20 KHB/RPMI) to maintain the immunological functions of U937 cells.

The optimized functional assay was employed to evaluate the skin-sensitizing potential of CBZ, PHT, and allopurinol. Briefly, test compounds were incubated with HepaRG-Heps in KHB medium for 6 h. The conditioned medium was then collected and reconstituted with RPMI at a ratio of 1:20 before being introduced to the U937 cells (Figure A). Acetaminophen (APAP) was included as a negative control drug because it gets metabolized into N-acetyl-p-benzoquinone imine, the reactive metabolites that will cause a toxic effect to the liver but not skin.[35] We observed that conditioned media incubated with all of the paradigm skin sensitizers stimulated upregulation of IL8, IL1β, and CD86 compared to parent drug alone (Figure B–D). In contrast, conditioned medium incubated with APAP did not trigger any upregulation of the APC markers (Figure B–D). These results indicated that the APC activation assay can differentiate the reactive metabolites of skin-sensitizing drugs from those of hepatotoxic drugs.

Figure 4

Assessment of skin-sensitizing potential on the metabolites generated by HepaRG-Heps. (A) Schematics of the experimental setup to study the effect of drug-reactive metabolites generated by HepaRG-Heps to activate the U937 skin-sensitizing model to antigen-presenting like cells (APCs) by upregulating the inflammatory markers IL8 and IL1β as well as costimulatory marker CD86. Gene expression of (B) IL8, (C) IL1β, and (D) CD86 changes in U937 after 48 h of parent drug (white bars) and HepaRG-Heps conditioned media (black bars) treatment. The conditioned media were in 1:20 KHB to the RPMI ratio. Data are average ± SEM of 3 independent experiments. Asterisks denote statistically significant differences (Student t-test, *p < 0.05; **p < 0.01).

Discussion

Severe cutaneous reactions associated with the therapeutic drugs used have been proposed to be caused by antigenic reactive drug metabolites, which can trigger a pathological adaptive immune response.[36,37] Various metabolic pathway studies have been performed using animal models to provide clues for the identification of possible reactive metabolites that might contribute to skin sensitization. For example, 33 metabolites from CBZ have been identified from rat and human urine[38] while PHT is found to be metabolized primarily into p-HPPH.[13] Wei Lu and colleagues further postulated that 2-OH CBZ, 3-OH CBZ, and p-HPPH were the reactive metabolites from CBZ and PHT, respectively, due to their ability to generate reactive oxygen species (ROS). These ROS can bind randomly to form protein adducts to initiate immunological response after incubating with liver microsomes and myeloperoxidase.[39] Despite recognizing the importance of liver-mediated drug biotransformation and immune activation in the etiology of severe cutaneous reactions, there has been limited effort to develop in vitro models that can predict the skin-sensitizing potential of a drug. In this study, we demonstrate a proof-of-concept in vitro testing scheme, involving the sequential application of paradigm skin sensitizers to a bioactivating-hepatic cell model followed by an immunologically reactive APC model to determine their skin sensitization potential. The hepatic model would convert the parent compound into its reactive metabolites, which would be subsequently cross-fed to an APC model to evaluate their immunogenicity potential.

We first showed that HepaRG-Heps were metabolically competent in producing various reactive drug metabolites of three paradigm skin sensitizers. The generation of reactive drug metabolites in the context of cutaneous reactions has only been evaluated in liver microsomes.[18,40,41] Compared to microsomes, which only contain a limited subset of liver metabolic enzymes,[42] human hepatocytes will be able to more accurately recapitulate the biotransformation of drugs into their corresponding reactive metabolite species. For instance, the generation of the antigenic reactive metabolite, oxipurinol, from allopurinol is mediated by xanthine oxidase (XO),[14,24] which is typically absent from liver microsome preparations.[42] In contrast, HepaRG-Heps have been discovered not only to have high CYP450 enzymes but also contain XO.[24] Indeed, we proved that incubation of HepaRG-Heps with allopurinol can generate oxipurinol in different media (Figure ). Therefore, our cutaneous bioactivation model using HepaRG-Hep serves to more comprehensively mimic the metabolism of a large range of drugs.

The cross-feeding of conditioned medium from drug-treated hepatocytes to immune cells have mostly been performed in the context of idiosyncratic drug-induced liver injury. For example, Kato et al. tested the supernatant generated by the FLC-4 hepatocyte cell line with THP-1 derived macrophages;[43] Kegel et al. tested the supernatant from human primary hepatocytes treated with hepatotoxic drugs on isolated human Kupffer cells;[44] and Oda et al. used the supernatant from drug-treated HepG2 and HepaRG to treat with human promyelocytic leukemia HL-60 cell.[45] In these studies, the immune cell model exhibited macrophage-like phenotype and functions because they were used as proxies for Kupffer cells. In the case of adverse cutaneous reactions, the pathological immunological reaction is initiated by antigenic reactive metabolites activating APCs, which will in turn activate cytotoxic effector cells.[46] Therefore, we have selected the use of U937 cells as the immune cell model because they are used in well-established standardized OECD tests to detect dermal skin sensitizers. Indeed, the activation of APC by U937 cells was not only affected by dermal skin sensitizers, such as DNCB (Figure B–E), but also by antigenic reactive metabolites from skin-sensitizing drugs, including CBZ, PHT, and allopurinol (Figures G–H and 4B–D). Importantly, the reactive metabolite of acetaminophen, which is known to cause hepatotoxicity but not skin sensitization, did not trigger a positive response in the U937 cells (Figure B–D). Therefore, our immune functional assay is specific to cutaneous reactions.

In this study, the presentation of reactive metabolites was realized via the cross-feeding of conditioned medium from drug-treated HepaRG-Heps to U937 cells. Since HepaRG-Heps and U937 cells were maintained in different culture media, there was a need to evaluate the functions of the 2 cell types in different media. Our results indicated that both simple (KHB) and complex (WE and RPMI) media did not interfere with their detection by LCMS (Table ). However, the generation of different drug metabolites by HepaRG-Heps differed significantly between different media (Figure ). For example, the metabolism of CBZ is mainly mediated by CYP450 enzymes, which are well-maintained in KHB buffer.[33] Thus, the production of CBZ metabolites was the highest in KHB (Figure B–D). In contrast, the highest production of oxipurinol was observed in WE medium (Figure F). This is likely due to the presence of dexamethasone in WE medium, which induces XO,[47] the main enzyme that bioactivates allopurinol to oxipurinol. Similarly, we found that the activation of U937 cells was highly sensitive to the medium composition. The stimulation of U937 cells into APC-like cells was most robust in its own maintenance medium (i.e., RPMI). Addition of WE medium consistently attenuated the APC activation response to a greater extent than KHB buffer (Figure C–E). This may be due to the presence of inhibitory factors present in a more complex medium. The ratio of KHB/RPMI must also be carefully titrated as a higher proportion of hepatocyte medium (1:5 and 1:10), adversely affecting both the growth rate as well as the APC activation response (Figure ). This resulted in the dilution of reactive metabolites present in the conditioned medium collected from drug-treated HepaRG-Heps, which may explain why the APC activation response was relatively weak (Figure ). A compartmentalized coculture of liver and immune that can maintain different types of cells in their own medium and allow the diffusion of metabolites would circumvent this limitation to activate a robust immune response.

Conclusions

We have shown that HepaRG-Heps could generate skin-sensitizing metabolites resulting from both CYP450 and XO metabolism. Thus, HepaRG-Heps can potentially serve as the hepatic bioactivation model in the development of in vitro adverse cutaneous drug reaction testing. A functional assay based on the activation of APC by U937 cells was also established to prospectively evaluate the immunogenic potential of drug-reactive metabolites. This assay could distinguish between the reactive metabolites of 3 known skin sensitizers (i.e., CBZ, PHT, and allopurinol), from a nonskin sensitizer (APAP). However, the cross-feeding of conditioned medium from drug-treated hepatocytes to immune cells was nonideal due to the sensitivity of the U937 cells to medium compositions. Nonetheless, this study demonstrates that the extent of APC activation by reactive drug metabolites generated by HepaRG-Heps can potentially be used to predict potential skin-sensitizing drugs.

Materials and Methods

Materials

All drugs, chemicals, and reagents were purchased from Sigma Aldrich, Singapore unless otherwise stated.

HepaRG-Derived Hepatocyte (HepaRG-Hep) Cultures

HepaRG progenitor cells (Biopredic International, France) were cultured and differentiated according to the manufacturer’s protocols.[48] Briefly, the HepaRG progenitor cells were proliferated and matured in growth medium for 14 days. The cells were then adapted in a medium with a 1:1 ratio of HepaRG growth and differentiation medium for the next 3 days before undergoing the differentiation process for another 14 days. The HepaRG growth medium and differentiation medium were prepared using William’s E medium (WE) supplemented with HepaRG growth additives (Biopredic, France) and differentiation additives (Biopredic, France), respectively. After a total period of 14 days of growth and 17 days of differentiation, HepaRG-derived hepatocytes (HepaRG-Hep) were used for gene expression analysis, CYP basal activities, and drug testing experiment.

Primary Human Hepatocyte (PHH) Culture

Cryopreserved primary human hepatocytes (PHHs) were obtained from Life Technologies (Carlsbad, CA) and BD Biosciences (Franklin Lakes, NJ). Three different lots of cryopreserved PHHs were used for the experiments. The cells were cultured in William’s E (WE) medium supplemented with 1 mg/mL bovine serum albumin (BSA), insulin transferrin, and selenium, 50 ng/mL linoleic acid, 50 nM dexamethasone, and 100 U/mL of penicillin/streptomycin. Culture medium was replenished daily. Freshly thawed PHHs were used for gene expression analysis. PHHs cultured for 24 h were used for the examination of albumin production, and CYP basal activities.

U937 Cell Cultures

U937 cells were obtained from ATCC. The cells were cultured in RPMI 1640 medium (Life Technologies, Singapore) supplemented with 2 mM glutamine, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 1 mM sodium pyruvate, 100 U/mL penicillin–100 μg/mL streptomycin, and 10% fetal bovine serum (FBS) (HyClone). Cells were passaged every 3 days. The cell number and viability were measured using the Trypan blue dye and the growth rate of U937 was defined as the percentage of viable cells in conditioned medium to the viable cells in RPMI.

Immunostaining

HepaRG-Heps were fixed with 4% paraformaldehyde for 30 min, followed by permeabilization with 0.1% Triton-X in PBS for 20 min and blocking in a solution of 2% bovine serum albumin (BSA) in PBS for 1 h. The sample was then stained with primary antibodies: rabbit anti-CYP3A4 (Abcam, Cambridge, U.K.) and mouse anti-CK19 (Abcam, Cambridge, U.K.) overnight. Subsequently, after rinsing three times with 1× PBS, the sample was stained with secondary antibodies (Alexa Fluor 488-conjugated anti-mouse and 555-conjugated anti-rabbit, Life Technologies) for 1 h in room temperature. Nuclei were counterstained with DAPI (Thermo Fisher). Imaging was performed using a fluorescence (Nikon Ti-E, Japan) microscope.

Quantitative PCR (qPCR)

Total RNA was purified from samples using an RNeasy mini kit (Qiagen, Singapore). tRNA was then transcribed into cDNA using a Tetro cDNA synthesis kit (Bioline, U.K.). Real-time quantitative PCR was performed with FastStart Universal SYBR Green Master reagents (Rox) (Roche, Germany) in a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific).

Fold changes of transcripts in U937 cells relative to untreated control samples were determined by ΔΔCt while log2(GAPDH-gene) was used for hepatocyte gene expression. The primers used for human hepatocytes and U937 cells are listed in Tables S4 and S5, respectively.

Cytochrome P450 Activity

Basal activities of two important cytochrome P450 (CYP) enzymes (i.e., CYP1A2 and CYP3A4), were determined. Following differentiation, HepaRG-Hep were incubated in 100 μL of Krebs–Henseleit buffer (KHB) with a cocktail of CYPs substrates (Table S6) in 96 well plates for 2 h at 37 °C. Dimethyl sulfoxide (DMSO) was selected as a solvent to dissolve the CYP substrates as stock solutions. The stock solutions of CYP substrates were diluted in KHB to a final concentration such that DMSO did not exceed 0.1%. The metabolites generated by HepaRG-Hep were then analyzed using liquid chromatography-mass spectrometry (LCMS). In general, 100 μL of supernatants generated by HepaRG-Hep were spiked with 100 μL of APAP D4 as internal standards (Table S6). The mixtures were dried using a vacuum concentrator (Eppendorf, Germany) in room temperature for at least 6 h to eliminate the solvent. 100 μL of methanol containing 0.1% formic acid was added into each of the dried residues and followed by a vortex mixing. After centrifuging at 10 000g for 10 min at 4 °C, 60 μL of the sample supernatant was collected for LCMS analysis (LC: 1100 series, Agilent, Singapore; MS:LCQ Deca XP Max, Finnigan, Singapore). The drug metabolite products are acetaminophen (APAP) and hydroxy midazolam for CYP1A2 and CYP3A4, respectively.

Albumin Secretion

The assessment of albumin function in hepatocytes was performed using the human albumin ELISA quantitation kit (Bethyl Laboratory Inc.) based on the manufacturer’s recommendation. Culture media were collected from HepaRG-Heps on days 3, 5, 7, 9, and 11 for albumin analysis. The measurements of albumin production rates were normalized to the number of cells in the culture, which was determined by measuring the DNA content of the samples using a Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies, Singapore) and was subsequently extrapolated from a standard curve established with a known cell number.

High-Performance Liquid Chromatography-Mass Spectrometry (HPLC-MS/MS) analysis

The quantitative analysis of the skin-sensitizing drugs and their metabolites was performed using an Agilent 1290 Infinity Rapid Resolution Liquid Chromatography System (Agilent, Lexington, MA), connected to an Agilent 6495 triple quadrupole mass spectrometer. The Agilent Poroshell 120 EC-C18 (50 mm × 3.0 mm, i.d., 2.7 μm) was used to carry out the chromatographic separation. The detail of the mobile phase is shown in Table S7. The details of multiple reaction modes are shown in Table S8 for ion detection. The conditions for mass spectrometry were also optimized with 3000 V of capillary voltage, 16 L/min of gas flow rate, 150 °C of dry gas temperature, 350 °C of sheath gas temperature, and 40 psi of nebulizer.

Calibration Standard and Quality Control Samples

A primary stock solution of metabolites and internal standard (IS) were prepared by dissolving the weighed metabolites and IS in methanol, respectively. A series of working standards were prepared by serially diluting the stock solution of metabolites with methanol. All of the solutions were stored below 4 °C. The calibration standards were prepared by adding 10 μL of working standard solution of metabolites to 90 μL of blank media. Three levels of quality control (QC) samples in different media (WE, RPMI, and KHB) were prepared separately using the same method.

Sample Preparation and Extraction

HepaRG-Heps were incubated with various concentrations of carbamazepine (100 μM), phenytoin (15 μM), and allopurinol (10 μM) in either William E (WE), RPMI, or KHB for 2 or 6 h. The final concentrations of each drug were selected according to their solubility and previously reported IC50 values.[49] A volume of 100 μL medium sample in different media (WE, RPMI, and KHB), QC, and calibration standard were spiked with 10 μL of IS and vortexed for 10 seconds, respectively. The mixture extraction was first carried out using 1 mL of ethyl acetate with 2 min of vortex mixing and followed by centrifugation to separate the aqueous and organic layers at 5000 rpm for 3 min. The organic layer was collected and dried using a vacuum concentrator. 100 μL of methanol was added into the dried residues and vortexed for 30 seconds before centrifuging at 12 000 rpm for 5 min. A sample with a volume of 2 μL was collected for HPLC-QqQ-MS analysis.

LCMS Method Validation

Specificity

Analysis specificity was calculated by comparing the ion chromatograms from (1) blank WE, RPMI, and KHB media, (2) the blank media spiked with reconstituted metabolites as QC, and (3) the supernatants collected from HepaRG-Heps after being treated with skin sensitization drugs (CBZ, PHT, and allopurinol) in different media.

Linearity and Lower Limits of Quantitation (LLOQ)

A volume of 10 μL of mixed working solutions and 10 μL of IS solution were added into 90 μL of blank media to prepare the calibration samples. The samples were then pretreated as described in Section . The resulting media contained 1–500 ng/mL of 2-OH CBZ, 3-OH CBZ, and CBZ-E; 1–250 ng/mL of p-HPPH and 100–5000 ng/mL of oxipurinol. The calibration curve was first constructed by calculating the peak area ratio of metabolites/IS plotted against the corresponding concentrations. A 1/x2 weighted linear least-squares regression model was then calculated using Mass Hunter Quantitative Analysis Software (version B.07.00, Agilent) to establish the calibration curve. All of the concentrations of unknown samples were calculated by interpolation from the established calibration curve, and the LLOQ represented the lowest concentration.

Precision and Accuracy

The evaluation of precision and accuracy was performed using QC samples at different concentrations (i.e., low, medium, and high), where precision of analysis was indicated by the relative standard deviation (RSD in %); while the accuracy of analysis was indicated by relative error (RE in %). The assessment of intra-day, inter-day precision and accuracy were performed for three consecutive days with five replicates. The precision and accuracy of sample analyses were calculated using calibration curves obtained daily.

Extraction Recovery and Matrix Effects

The extraction recovery efficiency was calculated by determining the ratio of the amounts of QC samples after sample preparation to the original spiked solution. The effect of different media composition on the separation and detection of drug species (henceforth referred to as the matrix effect) was evaluated by normalizing the LCMS readings of samples where drug metabolites were spiked into blank media (WE, RPMI, and KHB) to the readings where metabolites were spiked into methanol at three different QC concentrations. A ratio of 100% indicated that the components of the culture media did not affect LCMS detection of the drug species. All analyses were conducted using five replicates.

Stability

The long-term stability and post-preparation stability of metabolites in different matrices were evaluated using the quality control (QS) samples there were stored at either −20 °C for 14 days or at room temperature for 24 h with three freeze–thaw cycles.

Systemic Drug-Induced Skin Sensitization Assay

2 × 105 U937 cells were seeded in each well of a 12-well plate containing 2 mL of RPMI or HepaRG-Hep conditioned medium. For HepaRG-Hep conditioned medium, HepaRG-Heps were first grown for 14 days and differentiated for another 14 days in 96 well plates before incubating with parent drugs (i.e., CBZ, PHT, allopurinol, and APAP) for 6 h containing 100 μL of KHB. The number of HepaRG-Heps was around 3 × 104 cells for each well in the 96 well plate culture, which was determined by measuring the DNA content of the lysed HepaRG-Heps’ samples using the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies, Singapore). The KHB containing drug metabolites generated by HepaRG-Heps was then collected and reconstituted with fresh RPMI medium in the ratio of 1:20, 1:10, and 1:5 as the conditioned medium before adding to the U937 cells. The concentration of supplements used in the conditioned medium (e.g., FBS and Pen/strep) were adjusted to a final concentration of 10% FBS and 1% Pen/strep. Details on the composition of the conditioned medium are shown in Table S9. Negative and positive controls were included in every experiment. The negative control corresponded to U937 cells in RPMI or conditioned medium (in ratio 1:20, 1:10 and 1:5) without any drug treatment; while the positive control was U937 cells treated with 1 μM of 2,4-dinitrochlorobenzene (DNCB) for 48 h since it is a strong contact skin sensitizer.[27] After incubating U937 cells with parent drugs, reconstituted metabolites and supernatant collected from drug-treated HepaRG-Heps in either RPMI or different ratios of conditioned medium at 37 °C with 5% CO2 for 48 h, U937 cells were harvested for real-time qRT-PCR analyses to examine the changes of gene expression of IL8, IL1β, and CD86.

Statistical Analysis

All data were presented as the mean value ± standard error of the mean (SEM) for each sample.