Differential effects on human cytochromes P450 by CRISPR/Cas9-induced genetic knockout of cytochrome P450 reductase and cytochrome b5 in HepaRG cells

HepaRG cells are increasingly accepted as model for human drug metabolism and other hepatic functions. We used lentiviral transduction of undifferentiated HepaRG cells to deliver Cas9 and two alternative sgRNAs targeted at NADPH:cytochrome P450 oxidoreductase (POR), the obligate electron donor for microsomal cytochromes P450 (CYP). Cas9-expressing HepaRGVC (vector control) cells were phenotypically similar to wild type HepaRG cells and could be differentiated into hepatocyte-like cells by DMSO. Genetic POR-knockout resulted in phenotypic POR knockdown of up to 90% at mRNA, protein, and activity levels. LC–MS/MS measurement of seven CYP-activities showed differential effects of POR-knockdown with CYP2C8 being least and CYP2C9 being most affected. Further studies on cytochrome b5 (CYB5), an alternative NADH-dependent electron donor indicated particularly strong support of CYP2C8-dependent amodiaquine N-deethylation by CYB5 and this was confirmed by genetic CYB5 single- and POR/CYB5 double-knockout. POR-knockdown also affected CYP expression on mRNA and protein level, with CYP1A2 being induced severalfold, while CYP2C9 was strongly downregulated. In summary our results show that POR/NADPH- and CYB5/NADH-electron transport systems influence human drug metabolizing CYPs differentially and differently than mouse Cyps. Our Cas9-expressing HepaRGVC cells should be suitable to study the influence of diverse genes on drug metabolism and other hepatic functions.


Results
Genotyping of HepaRG cells. According to Gripon et al. 4 , HepaRG cells carry an additional chromosome 7, which however harbours a deletion that includes the location of the POR gene. Thus, HepaRG cells are diploid for POR, but to our knowledge, no information on POR genotype in HepaRG cells has been published. We therefore sequenced the entire POR gene including all exons and adjacent intron regions, resulting in genotype POR*1/*37. In contrast to the common allele *28 (A503V), which can influence CYP activities in diverse ways, *37 (A503V + V631I) is a rare allele (< 1%) that has not been functionally characterized (www.Pharm Var.org) 31,47,48 . As POR-knockout effects on CYP-activities may also depend on CYP genotype, we determined major alleles for the CYPs included in our study (Table 1). While previously described genotypes for CYPs 2C9 (*2/*2), 2C19 (*1/*1), 2D6 (*2/*9) and 3A5 (*3/*3) were verified 49 , we found that HepaRG cells are homozygous for CYP2C8*3 and heterozygous for CYP2B6*6 ( Table 1).  50 , we used lentiviral transduction of undifferentiated cells for delivery of Cas9 and sgR-NAs. The high transduction efficiency coupled with antibiotic selection lead to a high proportion of cells (> 75%) expressing Cas9 (Fig. 1a,b). We next examined whether transduced HepaRG VC cells are still able to differentiate with DMSO into hepatocyte-like cells. Indeed, using standard differentiation conditions, Cas9-expressing HepaRG VC cells were morphologically comparable to wild type HepaRG cells with respect to their ability to differentiate into hepatocyte-like cells and biliary cells (Fig. 1c-f). Moreover, enzyme activities simultaneously determined for seven CYPs correlated strongly (r S = 0.86) with those of wild type cells, although the absolute activities tended to be somewhat lower (Fig. 1g). Analysis of a broader set of genes showed also a tendency to lower expression in HepaRG VC versus HepaRG, but confirmed highly similar gene expression patterns (r S = 0.94; Fig. 1h). Taken together these findings suggested that HepaRG VC cells retained the most important characteristics of HepaRG and should thus be highly suitable for genome editing.
Characterization of POR-knockout. For CRISPR/Cas9-mediated knockout of POR in HepaRG cells we designed one sgRNA (POR#1) near the 5′-end of exon 2 using the common CHOPCHOP tool (Fig. 2a). A previously reported sgRNA (POR#2), which binds near the 5′-FMN binding site in exon 4, was simultaneously analyzed for comparison 39 . Predicted CRISPR/Cas9 editing was validated for both sgRNAs using the T7E1 assay (Fig. 2b). Whereas both sgRNAs had comparable efficiency scores (51.6 and 52.6, respectively), sgRNA POR#2 was predicted to bind to three off-targets. However, gene editing in these regions could be excluded by T7E1 assay. Analysis of POR in differentiated HepaRG cells revealed that both sgRNAs were similarly effective and reduced POR mRNA and protein by 60 to 80% (Fig. 2c). Although there were minor differences in reduction of mRNA and protein between sgRNA POR#1 and POR#2, they were not consistent and we thus consider them  Impact of POR-knockdown on CYP enzyme activities. We used our previously established CYP cocktail LC-MS/MS assay to measure enzyme activities of seven CYPs simultaneously in microsomes isolated from both types of HepaRG -POR and HepaRG VC cells. As shown in Fig. 3a, all CYP activities except for amodiaquine N-deethylation were negatively affected by both sgRNAs. While the effects on individual CYP isoforms differed substantially, the pattern was similar for both sgRNAs. Across all CYPs measured, we observed 1.4-to 20-fold reductions in enzyme activity. The strongest effect of POR-knockdown was observed for CYP2C9 (85-95% reduction of activity), while no significant effect was seen for CYP2C8. Compared to other CYP activities, CYP2C8-catalyzed amodiaquine N-deethylation thus appeared to be rather insensitive against variable POR levels. Mean levels ± SD are shown relative to vector control (VC) set at 1 (dark grey: VC, light grey: POR#1, white: POR#2). Statistical significance was assessed by unpaired t-test (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant). www.nature.com/scientificreports/ In order to characterize these changes in more detail we performed kinetic experiments exemplarily for CYP2C8, CYP2C9 and CYP3A4 using the substrates amodiaquine, tolbutamide, and atorvastatin, respectively, and midazolam as a second CYP3A4 substrate ( Fig. 3b-e, Table 2). The effects of POR-knockdown on kinetic parameters of amodiaquine N-deethylation were again surprisingly small, reducing V max by only 26% (POR#1) and 13% (POR#2), while substantial reductions in V max were found for the other substrates (Table 2). Only for tolbutamide an increase in K M (~ twofold) was observed, effectively reducing intrinsic clearance for CYP2C9 to about 5% in HepaRG -POR compared to HepaRG VC cells. The kinetic measurements of the conversion of In depth analysis of CYP2C8-mediated amodiaquine N-deethylation. As the striking insensitivity of amodiaquine N-deethylation towards POR-knockdown was surprising, we made additional analyses. Using the potent and specific CYP2C8 inhibitor montelucast 51 we confirmed similarly strong inhibition of amodiaquine N-deethylation in HepaRG VC as well as in both types of HepaRG -POR microsomes (Fig. 4a), suggesting catalysis by CYP2C8.
To test POR sensitivity of amodiaquine N-deethylation in a different system we next used commercially available bactosomes containing recombinant CYP2C8, CYP2C9 and CYP3A4 coexpressed with high or low levels of POR (Fig. 4b-e). Indeed, in this system amodiaquine N-deethylation was clearly POR sensitive, suggesting that CYP2C8 may be supported by an alternative electron donor in HepaRG cells.
As a primary candidate we suspected the CYB5/CYB5 reductase system, which is lacking in the bacterial membrane vesicles but should be present in HepaRG cells 52 . Taking advantage of the fact that CYB5/CYB5 reductase depends on NADH rather than on NADPH, we performed comparative activity measurements in HepaRG VC and both types of HepaRG -POR microsomes. All seven CYP activities could be supported by NADH alone, albeit at a lower level (~ 15-50% compared to NADPH; Fig. 4f). All differences of NADH-versus NADPHsupported activities were statistically significant (p < 0.05; two-way ANOVA with Bonferroni correction). Interestingly, CYP2C8 activity was least affected by the cofactor change (50%) and in HepaRG -POR microsomes NADHdependent activity was significantly increased (POR#1, 65%; POR#2, 87%). Taken together these data suggested involvement of CYB5 in the various CYP activities, which seemed to be particularly strong for CYP2C8. To prove this assumption directly we created genetic CYB5A single-and POR/CYB5A double-knockout cells.
Effects of genetic CYB5A and combined POR/CYB5A knockout on CYP enzyme activities. Since lentivirally transduced HepaRG cells constitutively express Cas9, gene knockout can be easily accomplished by transient transfection of suitable gRNAs. Therefore, we transfected undifferentiated HepaRG VC and HepaRG -POR cells with two sgRNAs targeting CYB5A (Fig. 5a), resulting in CYB5A single-and POR/CYB5A double-knock-  www.nature.com/scientificreports/ down HepaRG cell lines. Characterization following differentiation revealed 50% decrease of CYB5 on mRNA level and ~ 60 to 90% decrease on protein level (Fig. 5b). To analyze the effect of the double-knockdown on CYP-activities we measured these directly in living cells (Fig. 5c). Although all seven CYP-activities appeared to be decreased by 20-40%, only the strongest difference seen for CYP2C8-dependent amodiaquine N-deethylase activity was statistically significant. Most activities were further diminished in the double-knockdown cells, again most profoundly for CYP2C8 activity. Taken together, these and the former NADPH/NADH experiments www.nature.com/scientificreports/ indicated that several of the human CYP enzyme activities we tested for were markedly influenced by the CYB5 electron donor system and that amodiaquine N-deethylation showed a particularly strong dependence on CYB5 with accordingly less dependence on POR.

Effects of POR-knockdown on gene expression. The effects of POR-knockdown on CYP expression
are summarized in Fig. 6. We observed a surprisingly strong increase in CYP1A2 protein level by 4.5-and 9-fold for sgRNAs POR#1 and POR#2, respectively, while CYP2C9 and CYP2D6 were decreased by 50-70% and 30-40%, respectively (Fig. 6a,b). Protein expression of CYPs 2B6, 2C8 and 3A4 was apparently not markedly changed by POR-knockdown. Our findings at the protein level were corroborated by measurements of mRNA expression levels, which also showed strongly CYP isoform-dependent effects (Fig. 6c). For CYPs 2B6 and 2C9 mRNA levels were decreased with generally stronger effects seen for sgRNA POR#2, in agreement with the CYP2C9 protein data. The strong induction of CYP1A2 protein was confirmed by an up to 3.13-fold induction of CYP1A2 mRNA. Induced mRNA levels of CYP2C8 as well as unchanged levels of CYP3A4 mRNA were also in good agreement with the protein data.

Discussion
Here we used CRISPR/Cas9 genome editing in HepaRG cells to study effects of POR and CYB5 on the activity and expression of seven human CYP enzymes. HepaRG cells are usually kept in a proliferative state and then differentiated to hepatocyte-like cells by DMSO 4 . Pilot cell cloning experiments indicated substantial phenotypic heterogeneity among individual cell clones, many of which had lost their differentiation capacity. As we considered it important to maintain cellular characteristics during gene editing, we dismissed the possibility of selecting single clones with heterozygously or homozygously edited target gene as previously described 14,16 and instead used lentiviral transduction. Delivery of Cas9 and a puromycin resistance gene resulted in Cas9expressing cells (HepaRG VC ) that closely resemble original HepaRG cells in their phenotypic appearance and the ability to differentiate in 2% DMSO to hepatocyte-like cells with a highly similar CYP expression profile, as shown by close correlation of gene expression and CYP activity profiles (Fig. 1). Using either one of two different POR-specific sgRNAs up to 90% reduction of POR enzyme activity was achieved. The low residual POR presumably derives from cells with heterozygous or no knockout that escaped puromycin selection. Previous POR-knockout studies in mammalian cells achieved between ~ 50% up to 100% depending on the technology used [33][34][35][36][37][38][39][40] . The effects of POR reduction on the various CYP-activities differed greatly with CYP2C8-catalyzed amodiaquine N-deethylation being least affected while CYP2C9-catalyzed tolbutamide being the most affected (Fig. 3). Kinetic measurements showed that the reduced CYP activities were generally attributable to reduced V max . The comparably low tolbutamide hydroxylase activity may be partially explained by the significantly reduced expression of CYP2C9 in HepaRG -POR microsomes (Fig. 6). Notably, HepaRG cells are homozygous for the decreased function allele CYP2C9*2 that has been described to show lower affinity towards POR 53 .
The small effect of POR-knockdown on amodiaquine N-deethylase V max was surprising. Since the activity was strongly inhibited by the specific and potent CYP2C8 inhibitor montelucast 51 , an analytical artefact seemed unlikely. Since commercial bactosomes coexpressing CYP2C8 and different amounts of POR demonstrated a strong influence of POR on the same activity in this different system, a possible explanation was that HepaRG cells support this activity with an alternative redox partner that can compensate for lacking POR. An obvious candidate was CYB5, which is being regenerated by CYB5 reductase and NADH as cofactor. Since POR exhibits only marginal NADH-dependent activity 54 , our observation that all seven CYP-activities could be supported by NADH alone (~15-50% for HepaRG VC , Fig. 4) suggests a broad role of CYB5 as electron donor in HepaRG cells. Previous studies in other systems had indicated that CYB5 can act as sole electron donor for several mouse Cyps and at least for human CYP1A1 55 and that it markedly influences the activity of several drug metabolism activities catalyzed by human CYPs 27,42,45,46 . Our genetic CYB5A single-and POR/CYB5A-double knockout experiments directly confirmed the effect of CYB5 on some CYP-activities and on amodiaquine N-deethylation in particular (Fig. 5). With the possible exception of CYPs 1A2 and 2D6, the effect of CYB5A-knockdown was even greater in the POR-knockdown cells, suggesting that the CYB5/CYB5 reductase system compensates in part for lacking POR, probably due to unique but overlapping interaction sites for CYB5 and POR 43,56 . The strongest impact of CYB5A-knockdown on amodiaquine N-deethylation, especially in the double-knockdown cells, further demonstrated a particularly strong role of CYB5 for CYP2C8. Although there is evidence for CYB5 showing preferences for certain cytochrome P450 isozymes as well as particular reactions, we are not aware of any studies that included CYP2C8 57 . Interestingly, we found that HepaRG cells are homozygous for CYP2C8*3, a relatively common variant reported to show altered interaction properties with POR and CYB5 in vitro 58,59 .
One surprising result of the liver-specific deletion of Por in mice were the generally increased Cyp expression profiles, which suggested some compensatory mechanisms that have not yet been clarified nor have they been confirmed in a human cell culture system to our knowledge 24,25 . In HepaRG -POR cells we observed differential effects with CYP1A2 being severalfold induced while CYP2C9 was severalfold decreased, with data on mRNA and protein level being well in concordance (Fig. 6). The reasons for these profound differential effects on CYP expression as well as for the apparent species differences remain speculative. Due to the wide differences in CYP genes between mice and humans, which concern number, sequence and structure, it is difficult to directly compare gene regulatory effects. As a cautionary note it should also be kept in mind that HepaRG is an immortalized cell line that appears to reflect many aspects of human drug metabolism correctly but may also differ in certain aspects of dynamic gene regulation in comparison to other cell models or the liver as an organ.
In summary, using lentiviral transduction we have created a HepaRG cell line that constitutively expresses Cas9 and that retains the ability to differentiate into hepatocyte-like cells with cytochrome P450 expression and Scientific Reports | (2021) 11:1000 | https://doi.org/10.1038/s41598-020-79952-1 www.nature.com/scientificreports/ www.nature.com/scientificreports/ activity profiles highly similar to those of the parent cell line. Genetic knockout of POR and CYB5A resulted in differential, CYP isoform-dependent effects on CYP expression and activities. Our data support a general role of CYB5 for drug metabolism in HepaRG cells as previously observed in other human drug metabolism systems. The Cas9-expressing HepaRG VC cells should be a versatile tool to study the influence of diverse genes on drug metabolism and other hepatic functions in a metabolically competent human hepatic cell line. Cytochrome C reduction assay. Cytochrome P450 reductase activity was determined in 15-30 µg of microsomal protein using the cytochrome C reduction assay as described elsewhere 31,65 .

Methods
Microsomes and cytochrome P450 activity measurements. For microsome preparation HepaRG cells were cultivated and differentiated in T175 flasks, harvested and mechanically disrupted. Bactosomes containing human CYP2C8 (CYP/EZ017, CYP/EZ047), CYP2C9 (CYP/EZ019, CYP/EZ006) and CYP3A4 (CYP/ EZ002, CYP/EZ010) and human POR in high and low concentrations coexpressed in Escherichia coli were purchased from Cypex Ltd. (Dundee, UK). For determination of CYP enzyme activities in microsomal preparations a liquid chromatography with tandem mass spectrometry (LC-MS/MS) substrate cocktail assay was used essentially as described elsewhere 12,66 . For determination of CYP enzyme activities in cell culture supernatant, the cells were incubated with the substrate cocktail for 3 h at 37 °C in the incubator. Metabolites were quantified in supernatants as described 66 . For microsomal kinetic experiments, 2.5-20 µg of microsomes in 0.1 N potassium phosphate buffer (pH 7.4) were incubated with substrates and/or inhibitors at the indicated concentrations, and NADPH-regenerating system for 20-30 min depending on the substrate. For direct comparison of NADPH-and NADH-driven metabolism, the cofactors were used as pure substances at 1 mM concentration with substrate cocktail as described above. www.nature.com/scientificreports/ anti-human CYP3A4 antibody (#458234 Gentest Corp.). Secondary antibodies were fluorescence labeled goat anti-mouse IRDye 680 and 800 and goat anti-rabbit IRDye 680 and 800 (LI-COR Biosciences GmbH, Bad Homburg, DE), and visualized using the IR imaging system Odyssey (LI-COR Biosciences GmbH).
Quantitative real-time PCR for gene expression analysis. 100-200 ng of total RNA (isolated using RNeasy Plus Kit (Qiagen)) was reverse transcribed with TaqMan Reverse Transcription Reagents (Applied Biosystems, Foster City, USA). cDNA was preamplified using TaqMan PreAmp Mastermix (Applied Biosystems) and expression of 82 genes was quantified using the Biomark HD system (Fluidigm, San Francisco, USA) with a 48:48 Dynamic Array Chip (Fluidigm). Expression levels were determined in triplicates and normalized to the geometric mean of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), RPLP0 and β-actin expression 68 and relatively quantified using the ΔΔct method.
Statistical analyses. Statistical analyses were performed using Graphpad Prism V5 software (GraphPad, San Diego, USA). Spearman's correlation coefficients and corresponding tests were used to assess the associations between wildtype HepaRG and HepaRG VC on CYP activity and gene expression level. Repeated measurements ANOVA with Bonferroni correction or appropriate paired/unpaired t-test statistics were applied on log-transformed data, significance level was set to p < 0.05. Results are shown as means ± standard deviation (SD) or with 95% confidence intervals (95% CI), as appropriate. Kinetic parameters K M and V max were determined by Michaelis-Menten model or by substrate inhibition model, inhibition parameters IC 50 and K i were determined by one site competition model using Graphpad Prism V5 software (GraphPad). Internal clearance (Cl int ) was calculated using the following equation: Cl int = V max K M Received: 23 July 2020; Accepted: 14 December 2020