Evaluation of Human Hepatocyte Drug Metabolism Carrying High-Risk or Protection-Associated Liver Disease Genetic Variants

Metabolic-dysfunction-associated steatotic liver disease (MASLD), which affects 30 million people in the US and is anticipated to reach over 100 million by 2030, places a significant financial strain on the healthcare system. There is presently no FDA-approved treatment for MASLD despite its public health significance and financial burden. Understanding the connection between point mutations, liver enzymes, and MASLD is important for comprehending drug toxicity in healthy or diseased individuals. Multiple genetic variations have been linked to MASLD susceptibility through genome-wide association studies (GWAS), either increasing MASLD risk or protecting against it, such as PNPLA3 rs738409, MBOAT7 rs641738, GCKR rs780094, HSD17B13 rs72613567, and MTARC1 rs2642438. As the impact of genetic variants on the levels of drug-metabolizing cytochrome P450 (CYP) enzymes in human hepatocytes has not been thoroughly investigated, this study aims to describe the analysis of metabolic functions for selected phase I and phase II liver enzymes in human hepatocytes. For this purpose, fresh isolated primary hepatocytes were obtained from healthy liver donors (n = 126), and liquid chromatography–mass spectrometry (LC–MS) was performed. For the cohorts, participants were classified into minor homozygotes and nonminor homozygotes (major homozygotes + heterozygotes) for five gene polymorphisms. For phase I liver enzymes, we found a significant difference in the activity of CYP1A2 in human hepatocytes carrying MBOAT7 (p = 0.011) and of CYP2C8 in human hepatocytes carrying PNPLA3 (p = 0.004). It was also observed that the activity of CYP2C9 was significantly lower in human hepatocytes carrying HSD17B13 (p = 0.001) minor homozygous compared to nonminor homozygous. No significant difference in activity of CYP2E1, CYP2C8, CYP2D6, CYP2E1, CYP3A4, ECOD, FMO, MAO, AO, and CES2 and in any of the phase II liver enzymes between human hepatocytes carrying genetic variants for PNPLA3 rs738409, MBOAT7 rs641738, GCKR rs780094, HSD17B13 rs72613567, and MTARC1 rs2642438 were observed. These findings offer a preliminary assessment of the influence of genetic variations on drug-metabolizing cytochrome P450 (CYP) enzymes in healthy human hepatocytes, which may be useful for future drug discovery investigations.


Introduction
Nonalcoholic fatty liver disease (NAFLD), or, as has recently been proposed, metabolicdysfunction-associated steatotic liver disease (MASLD), is a huge financial burden on the healthcare system.It affects 30 million adults in the US and is expected to increase to over 100 million by 2030 [1].MASLD encompasses a spectrum of liver damage ranging from simple steatosis (or NAFL) to metabolic-dysfunction-associated steatohepatitis (MASH), cirrhosis, end-stage liver disease (ESLD), and hepatocellular carcinoma (HCC).Genetic and environmental factors, as well as disease drivers such as inflammatory cytokines, including adipokines, bacterial products, and metabolites originating from the intestine and adipose tissue, contribute to the development and progression of MASLD [2].The pathologic hallmarks of MASLD include steatosis, inflammatory infiltrate, hepatocyte ballooning, and fibrosis, leading to decreased hepatocellular functions and eventually cirrhosis and HCC [3,4].Despite its public health importance and financial burden, there is currently no FDA-approved therapy for MASLD.The lack of therapeutic options reflects the complex interpatient pathogenesis and heterogeneity as well as the lack of experimental models that fully recapitulate disease phenotypes and genotypes involved in MASLD progression.
The most relevant and reproducible SNP identified across GWAS is in the patatin-like phospholipase-domain-containing 3 (PNPLA3) gene (rs738409 C>G p.IIe148Met) [6].This variant encodes an isoleucine-to-methionine substitution at position 148aa (I148M) in the protein, which causes PNPLA3's lipase activity to be lost, resulting in fat buildup in hepatocytes.A decrease in protein function is also caused by the MBOAT7 rs62641738 genetic variation, which causes it to lose its ability to acetylate lysophosphatidylinositol lipids and, therefore, its protective role in preventing hepatic steatosis [7,8].GWAS have revealed a significant link between GCKR genetic variation and various metabolic parameters.Specifically, the p.P446L GKRP substitution has been identified as a critical factor; this substitution leads to the destabilization of the GCK binding interface, explaining the observed inverse correlation between fasting glucose and triglycerides for individuals carrying this variant.Furthermore, the increased activity of hepatic GCK resulting from this genetic variation leads to reduced glucose levels and simultaneous increases in triacylglycerol (TAG) and glycogen synthesis in individuals with normal blood sugar levels (normoglycemia) [9].
In contrast to variants that increase MASLD risk, recent studies have identified two novel protective variants in the HSD17B13 (rs72613567 T>TA; 17-β hydroxysteroid dehydrogenase 13) and MTARC1 (rs2642438 G>A p.Ala165Thr; mitochondrial amidoxime reducing component 1) genes that are linked to lower risk of MASLD [10][11][12][13][14].However, little is currently known regarding the biological function of these MASLD high-risk or protective variants.HSD17B13 is a lipid-droplet-associated protein with retinol dehydrogenase activity.Hepatic expression of HSD17B13 has been shown to be significantly higher in MASLD patients, while the protective variant (rs72613567 T>TA) is associated with a decrease in fibrogenic pathways [10,13,14].MTARC1 is a molybdenum-containing enzyme that is localized to the outer mitochondrial membrane and has been reported to function in both drug detoxification and lipid metabolism [8,9].The protective variant (rs264238 G>A) has been associated with decreased inflammation and fibrosis in MASLD [11,12].Our understanding of how these genetic variants function in MASLD pathogenesis is lacking.Moreover, the impact of genetic variants on the levels of drug-metabolizing CYP enzymes in human hepatocytes has been largely unexplored.
This report describes the analysis of metabolic functions in human hepatocytes, including eighteen hepatic phase I/II activity studies, three CYP induction studies, and five studies on the inducibility and mRNA expression of CYP enzymes.Data from 126 human hepatocyte isolations are presented and classified by the presence of either high-risk or protectionassociated MASLD genetic variants.These data provide an initial overview of the impact of genetic variants on drug-metabolizing enzymes in normal human hepatocytes, which could be helpful for drug discovery studies in the future.

Study Design
To assess the role genetic variants might play in drug-metabolizing cytochrome P450 (CYP) enzymes, polymorphic sites associated with high-risk liver disease progression and protection were determined in 126 human hepatocytes.The cryopreserved primary hepatocytes were obtained from In Vitro ADMET Laboratories Inc. (IVAL, Columbia, MD, USA).The isolations were procured from human liver donors that were rejected for orthotopic liver transplantation.Liver tissue was equally distributed between males and females (M:F ratio, 1:1).The mean age was 44 years (range 0-77 years), while the ethnicity distribution of the processed specimens was 71.4% Caucasian, 16.7% Hispanic, 7.9% African American, and 3.9% other ethnic groups (e.g., Asian, Middle Eastern, or not provided).For all the cell isolation specimens, lower than 70% viability preparations were excluded from the study.We focused on five genetic variants in donors and recipients where robust GWAS have linked these variants to a spectrum of liver diseases ranging from steatosis to nonalcoholic steatohepatitis, hepatic fibrosis, hepatocarcinoma, ESLD, and increased risk of mortality in the general population [15].These include PNPLA3 rs738409, MBOAT7 rs641738, GCKR rs780094, and two previously reported protective genetic variants HSD17B13 rs72613567 and MTARC1 rs2642438 (Table S1) [10,[16][17][18][19][20][21][22].We found that these variants were similarly distributed among the cases analyzed.The frequency of the gene variants was as follows: PNPLA3 rs738409:G (minor homozygous), 6.35%; MBOAT7 rs641738:T (minor homozygous), 21.5%; GCKR rs780094:T, 11.9%.The protective HSD17B13 rs72613567:TA gene variant was present in 3.1% of the cases, while MTARC1 rs2642438:A was present in 7.1% of the cases (Table S1).Test for Hardy-Weinberg equilibrium (HWE) was performed.The p-value for the variants was as follows: PNPLA3 rs738409, p = 0.952; MBOAT7 rs641738, p = 0.549; GCKR rs780094, p = 0.886; HSD17B13 rs72613567, p = 0.659; and MTARC1 rs2642438, p = 0.775.As the p-values were >0.05, there is not enough evidence to reject the null hypothesis that each SNP is in HWE.

SNPs and Liver Enzymes
To investigate the extent to which the presence of these genetic variants might influence drug-metabolizing cytochrome P450 (CYP) enzyme activity, hepatic phase I activities were measured for several important cytochrome P450 family members, including CYPs 1A2, 2C8, 2C9, 2D6, 2E1, and 3A4 (Table 1).Interestingly, we found significant differences in CYP1A2 activity (p = 0.011) in human hepatocytes with MBOAT7 minor homozygous (TT) compared to nonminor homozygous (CC/CT) using the substrate phenacetin and the metabolite acetaminophen and in CYP2C8 activity (p = 0.004) in human hepatocytes with PNPLA3 minor homozygous (GG) compared to nonminor homozygous (CC/CG) using the substrate paclitaxel and the metabolite 6-6α-hydroxypaclitaxel (Figure 1 and Table S2).Moreover, it was also observed that the activity of CYP2C9 (p = 0.001) using the substrate diclofenac and the metabolite 4-hydroxydiclofenac was significantly lower in human hepatocytes with HSD17B13 minor homozygous (TTA/TATA) compared to nonminor homozygous (TT) (Figure 1 and Table S2).
Given that CYP450 enzymes exhibit substantial polymorphism, we verified whether these significant differences observed in the results of CYP1A2, CYP2C8, and CYP2C9 could be attributed to the specific genetic variants under investigation rather than generic CYP450 polymorphisms.For this purpose, we conducted a genotyping test for one polymorphism for each CYP450 enzyme: CYP1A2 rs762551, CYP2C8 rs11572080, and CYP2C9 rs1057910.When comparing individuals carrying minor homozygous for the MBOAT7 rs641738, PN-PLA3 rs738409, and HSD17B13 rs72613567 variants and minor homozygous of CYP450 polymorphism, we found no significant difference (p = 0.6052, p = not enough sample to evaluate, and p = 0.1929, respectively).We also compared individuals carrying nonminor homozygous for MBOAT7 rs641738, PNPLA3 rs738409, and HSD17B13 rs72613567 variants and carrying nonminor homozygous of CYP450 polymorphism, and we found no significant difference (p = 0.452, p = 0.4304, and p = 0.1331, respectively).These results confirm that the heightened activity observed was indeed attributed to the specific variant under examination and not because of CYP polymorphism (Figure S1).
Int. J. Mol.Sci.2023, 24, x FOR PEER REVIEW 5 of 17 polymorphism, and we found no significant difference (p = 0.452, p = 0.4304, and p = 0.1331, respectively).These results confirm that the heightened activity observed was indeed attributed to the specific variant under examination and not because of CYP polymorphism (Figure S1).On the other hand, there was no significant difference in activity of CYP2D6, CYP2E1, and CYP3A4 between human hepatocytes carrying genetic variants for PNPLA3 rs738409, MBOAT7 rs641738, GCKR rs780094, HSD17B13 rs72613567, and MTARC1 rs2642438 (Figure 1 and Table S2).   2 and Table S3).Among the metabolic pathways that can directly affect the integrity of not only the liver but also multiple other organs are phase II conjugation enzymes.The phase II detoxification enzymes comprise conjugation reactions that facilitate the elimination of harmful metabolites from the body, thus reducing their toxicity.The enzymes involved in these processes include sulfotransferases (SULT), UDP-glucuronosyltransferases (UGT), glutathione S-transferases (GST), and N-acetyltransferases (NAT) [23].

Discussion
In this study, we genetically profiled 126 healthy liver donors for five different genetic variants previously associated with predisposition to the development of liver disease (PNPLA3 rs738409, GCKR rs780094, and MBOAT7 rs641738) or protection of liver disease (HSD17B13 rs72613567 and MTARC1 rs2642438) [24][25][26][27][28].We analyzed minor and nonminor homozygous in the genotype frequencies to determine associations between these genetic variants and phase I and II liver enzyme activity and inducibility.
It is already known that CYP450 are highly polymorphic.For that reason and to guarantee that the significant results observed was due to the variants and not because of CYP450 polymorphism, we performed genotyping tests for the most cited polymorphism related to CYP1A2 using the substrate phenacetin, CYP2C8 using the substrate paclitaxel, and CYP2C9 using diclofenac as the substrate.No significant difference was found in any of the hepatocytes carrying the CYPs mutant allele (minor homozygous) and nonminor homozygous that also carries minor homozygous and nonminor homozygous MBOAT7 rs641738, PNPLA3 rs738409, and HSD17B13 rs72613567 variants [29][30][31].
We already know that elevated hepatic enzyme activity can potentially reduce pharmacotherapeutic effectiveness, leading to toxic liver disease.Consequently, the reduced metabolization capacity of phase II enzymes may lead to toxic consequences when using clinically prescribed medications.
As demonstrated previously for PNPLA3 rs738409, the presence of allele GG is linked to escalated liver enzyme levels and an increased susceptibility to liver toxicity compared to the CC allele.This risk is particularly notable when individuals are treated with medications such as cyclophosphamide, cytarabine, daunorubicin, doxorubicin, mercaptopurine, methotrexate, aspargase, thioguanine, and vincristine.Our research indicated that the PNPLA3 rs738409 variant increased CYP2C8 activity when assessed with paclitaxel and 7-ethoxycoumarin as substrates [32].Additionally, we observed that the MBOAT7 rs641738 variant was associated with an increase in CYP1A2 activity, while the presence of HSD17B13 rs72613567 led to a reduction in CYP2C9 activity [32].
This potentially translates to an increased risk of adverse effects due to reduced metabolization capacity of phase II enzymes when individuals are prescribed clinical medications.According to the literature, the GCKR allele T is linked to an increased response to fenofibrate medication in individuals with hypertriglyceridemia. Interestingly, our data did not indicate a significant association between the GCKR rs780094 genetic variant and any significant increase or decrease in liver enzyme levels [33].
Our studies highlight the potential association between metabolic and toxicological consequences that certain genetic variants may present in human hepatocytes in human livers.Most of these variants cause harmful effects of elevated levels of lipids (lipotoxicity) on cells and tissues, particularly when these lipids accumulate to excessive levels.
Lipotoxicity in the liver can cause inflammation, oxidative stress, and cell damage, leading to MASLD, which ranges from simple fat accumulation (steatosis) to more severe conditions such as nonalcoholic steatohepatitis (MASH) and cirrhosis [34].
It is essential to study the relationship between P450 levels and MASLD to better understand the underlying mechanisms that contribute to the disease.This study can potentially lead to the development of targeted therapies or interventions aimed at modulating P450 enzymes to manage or prevent MASLD.However, it is important to note that the exact mechanisms and specific roles of P450 enzymes in MASLD may vary depending on individual genetic factors, environmental influences, and overall metabolic health.
Most of the gene variant associations examined here have been studied, and lipid droplet biology, intracellular lipid synthesis and degradation, and secretion of very-lowdensity lipoproteins [15] have been found to play a role in the potential mechanisms behind the development of MASLD, although the mechanisms by which the variants lead to ESLD and cellular death are poorly understood.
About 90% of medications are processed by multiple human phase I enzymes that regularly participate in clinically relevant drug-drug interactions.By activating or inhibit-ing phase I enzymes, certain drugs can either speed up or slow down the metabolism of other drugs [35].On the other hand, enzymes from phase II are crucial for the metabolic inactivation of pharmacologically active substances as well as the biotransformation of endogenous substances and xenobiotics into more excretable forms [36].
Drug-induced liver injury (DILI) is a significant concern as it can result in mild and reversible liver dysfunction turning to severe and life-threatening conditions, such as acute liver failure.When investigating the role of DILI in MASLD, studies that employ RUCAM causality assessment provide valuable data to establish the association between specific drugs and liver injury, aiding in the safe prescribing and management of patients with metabolic-associated liver diseases [37][38][39].
In 2018, Rolf Teschke collected data from DILI cases based on the RUCAM system from international databases.A total of 3312 drugs were analyzed, with amoxicillinclavulanate (an antibiotic), flucloxacillin (an antibiotic), atorvastatin (a lipid-lowering agent), disulfiram (a substance abuse agent), and diclofenac (an NSAID) being the five most relevant medications related to DILI [40].
Patients with MASLD may already have underlying liver abnormalities due to fatty liver, inflammation, or fibrosis.They also often share risk factors, such as obesity, diabetes, and metabolic syndrome [41].When exposed to certain drugs, these patients may be more susceptible to DILI as their liver might have reduced resilience to handle additional stressors when certain drugs are introduced, thereby exacerbating liver injury [42].
We observed that patients who were minor homozygous for genetic variants previously related to predisposition to liver disease (PNPLA3 and MBOAT7) had increased phase I enzyme activity.As opposed to that, the HSD17B13 variant, which is related to protective effect in liver disease, led to a decrease in phase I enzyme activity.The relationship between phase I enzymes and pharmacotherapy is already known.Increased hepatic phase I activity has the potential to decrease the pharmacotherapeutic efficacy of recognized substrates by increasing their metabolism or increasing the production of reactive metabolites and oxidative stress [23,43].Thus, the lower capacity of metabolization of phase II enzymes can bring toxic consequences from clinically utilized medications [36].Conversely, decreased phase I enzyme activity also decreases the production of reactive metabolites and oxidative stress, as we observed in HSD17B13.
The current study on drug-metabolizing cytochrome P450 (CYP) enzymes identified the importance of additional potential mechanisms of cellular stress that, together with environmental factors and unexplored mechanisms of disease, may facilitate disease progression.This study is limited by the small number of human hepatocyte cases that were analyzed carrying relatively uncommon genetic variants.Thus, future studies will be dedicated to expanding the number of cases to improve the power of the current analysis.In summary, the studies described here highlight pharmacogenomic interactions and help to expand understanding of human liver diseases and preventive therapies, paving the way for precision medicine.

Material and Methods
This study analyzed metabolic activity and enzyme induction of liver enzymes in healthy liver donors (n = 126).Supplementary Table S1 presents the detailed characteristics of the cohorts.Fresh isolated primary hepatocytes were obtained from healthy liver donors, and liquid chromatography-mass spectrometry (LC-MS) was performed.Metabolic activities were obtained for selected phase I and phase II liver enzymes by LC-MS.Table 1 presents the metabolic pathways, substrates with their concentration, the marker metabolites analyzed, iron mode applications, and mass transition monitoring.Genotyping was also performed using genomic DNA extracted from hepatocyte samples.For the cohorts, participants were classified into minor homozygotes and nonminor homozygotes (major homozygotes + heterozygotes) for the five investigated gene polymorphisms (PNPLA3 rs738409, MBOAT7 rs641738, GCKR rs780094, HSD17B13 rs72613567, and MTARC1 rs2642438) (Table S1).

Human Primary Hepatocytes
Cryopreserved primary hepatocytes (n = 126) from healthy donors were obtained from In Vitro ADMET Laboratories Inc. (IVAL, Columbia, MD, USA).These hepatocytes were isolated from liver specimens of donors who were negative for hepatitis C virus (HCV), hepatitis B virus (HBV), and human immunodeficiency virus (HIV).

DNA Isolation and Genotyping
Genomic DNA was isolated from the isolated hepatocytes using the DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany) following the manufacturer's instructions.DNA quantity and quality were measured using a NanoDrop Lite spectrometer (ThermoFisher Scientific, Waltham, MA, USA).Genotyping reactions containing 1× TaqMan™ Genotyping Master Mix (Applied Biosystems, Foster City, CA, USA), 1× TaqMan ® Genotyping Assays (Applied Biosystems, Foster City, CA, USA), and 4.5 pg of genomic DNA were prepared in MicroAmp Fast Optical 96-well plates (Applied Biosystems, Foster City, CA, USA).Real-time PCR was performed using the StepOnePlus system (Applied Biosystems, Foster City, CA, USA).
In order to confirm that the significant differences observed in the results of the enzyme inductions and the variability in the mutations of CYP1A2, CYP2C8, and CYP2C9 were due to the tested genetic variants and not due to a CYP450 polymorphism when comparing individuals with minor and non-minor homozygous for MBOAT7 rs641738, PNPLA3 rs738409, and HSD17B13 rs72613567, we performed the genotyping for CYP1A2 rs762551, CYP2C9 rs1057910and CYP2C8 rs11572080.Details of the TaqMan ® Genotyping Assays are listed in Table S7.
Incubation with DME substrates: All incubations were performed in 96-well cell culture plates (Falcon, obtained from VWR Inc., West Chester, PA, USA).The cryopreserved human hepatocytes were thawed in a 37 • C water bath, with 1 mL of the thawed suspension added to 50 mL of universal cryopreservation recovery medium (UCRM, In Vitro ADMET Laboratories Inc. (IVAL), Columbia, MD, USA) in a 50 mL conical cell culture tube (Falcon, obtained from VWR Inc.) and centrifuged at 100× g for 10 min.The cell pellet from each conical tube was resuspended in 4 mL of hepatocyte incubation medium (HQM, IVAL) for viability determination (Trypan blue exclusion) and cell concentration determination.The cell suspension was then adjusted with HQM to 2× of the final cell density.All DME substrates were prepared in HQM at 2× of the final concentrations and added at a volume of 50 µL per well in a 96-well cell culture plate.The hepatocyte and substrate plates were prewarmed to 37 • C for 15 min in a cell culture incubator before the initiation of incubation by pipetting 50 µL of hepatocytes into each well of the 96-well plates containing the substrates and returned to the cell culture incubator without shaking for the incubation duration of 30 min.At the end of each incubation, 100 µL of acetonitrile was added into each well to terminate metabolism.The plates after termination were stored in a −80 • C freezer for later LC/MS-MS quantification of metabolite formation.
LC/MS/MS Analysis: Upon thawing, an aliquot of 200 µL of each sample was transferred from each well into a labeled 96-well plate, followed by the addition of 100 µL of acetonitrile solution containing the internal standard tolbutamide (250 nM) and mixing.All samples were centrifuged at 3500 rpm for 5 min.An aliquot of 100 µL of supernatant from each sample was transferred to a 96-well plate and diluted with 200 µL of deionized water by mixing before LC/MS/MS analysis.CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4 (midazolam 1 -hydroxylation), CYP3A4 (testosterone 6b-hydroxylation), ECOD, UGT, SULT, GST, FMO, MAO, AO, NAT1, and NAT2 metabolites were quantified using API 4000 QTRAP mass spectrometer with an electrospray ionization source (AB SCIEX, Framingham, MA, USA) connected to an Agilent 1200 series HPLC (Agilent Technologies, Santa Clara, CA, USA) using LC/MS/MS MRM mode, and the mass transitions (parent to daughter ion) were monitored (Table 1).An Agilent Zorbax Eclipse Plus C18 column (4.6 × 75 mm i.d., 3.5 µm; Agilent Technologies, Santa Clara, CA, USA) at a flow rate of 1 mL/min was used for the chromatography separation.The mobile phase consisted of 0.1% formic acid in acetonitrile (A) and 0.1% formic acid in water (B).The gradient for the positive ion mode operation was programed as follows: 0 to 2.5 min, increase B from 5 to 95%; 2.5 to 3.5 min, 95% B; 3.5 to 3.6 min, decrease B to 5%; run time, 5 min.The gradient program for the negative ion mode was follows: 0 to 3 min, increase B from 5 to 95%; 3 to 4 min, 95% B; 4 to 4.2 min, decrease B to 5%; run time, 6 min.Data acquisition and data processing were performed with the software Analyst 1.6.2(AB SCIEX, Framingham, MA, USA).
CYP450 inducibility: The hepatocytes were treated with omeprazole (100 µM), phenobarbital (1000 µM), and rifampicin (20 µM) 24 h after plating for a duration of 48 h, with the treatment medium refreshed after the first 24 h of treatment.
Data Analysis: Data are presented as mean and standard deviation of triplicate incubations derived using the Microsoft Excel 6.0 software.Statistical analysis was performed using the Student's t-test with the Microsoft Excel 6.0 software, with the probability of p < 0.05 considered as statistically significant.Specific activity (pmol/min/million hepatocytes) of each drug-metabolizing enzyme pathway was determined by dividing the total metabolite formed by the incubation time and normalized to cell concentration.

Statistical Methods
The data were analyzed when at least 3 or more data cases in each group were available.Correlation, multivariate, and postestimation analyses were performed using Stata/SE v18.0.After univariate analysis, pairwise correlation between independent variables was assessed using the pwcorr command with Bonferroni adjustment at a significance level of 0.05.All pairs of independent variables were not significantly correlated.Linear regression was then performed using the regress command with the activity/inducibility/mRNA level of each drug-metabolizing enzyme as the dependent variable and the genotypes for the five SNPs, age, and sex as the independent variables.After linear regression, heteroskedasticity was assessed using the ivhettest, nr2 command, a Lagrange multiplier version of the Breusch-Pagan test identified by Wooldredge that returns the standard White/Koenker nR-sq test statistic for heteroskedasticity.Heteroskedasticity was not observed in the dataset.Multicollinearity was also assessed by determining variance inflation factor (VIF) for each of the independent variables using the estat vif function.All VIFs were less than 2.5, suggesting the absence of multicollinearity.To account for multiple testing, significant p-values were determined following the Benjamini and Hochberg false discovery rate (FDR) method.The p-value for each enzyme is described in Tables S2-S6, and the ones where we observed significant differences are highlighted in blue and noted as significant in the notes of the table [44][45][46].
Test for Hardy-Weinberg equilibrium (HWE) was performed using Stata/SE v18.0.HWE for each SNP was assessed using the hwsnp command at a significance level of 0.05.

17 Figure 3 . 5 NAT2Figure 3 .
Figure 3. Functional assay showing the association of variants in phase II enzymes among healthy donors.The box plots represent minor homozygous and nonminor homozygous genotypes of MBOAT7 rs641738 (orange), GCKR rs780094 (green), and MTARC1 rs2642438 (gray) for phase I liver enzymes: SULT (n = 44), SULT (n = 36), UGT (n = 44), GST (n = 36), NAT1 (n = 45), and NAT2 (n = 41).Each black dot represents a donor, and an unpaired two-sided Mann-Whitney U test with 95% CI was used to assess the difference in metabolite levels between minor homozygous and nonminor homozygous groups.The p-values are shown at the top of each box plot.

Figure 4 .
Figure 4. Phase I CYP450 enzymes induction among healthy donors.The box plots represent minor homozygous and nonminor homozygous genotypes of PNPLA3 rs738409 (blue), MBOAT7 rs641738 (orange), and GCKR rs780094 (green) for phase I CYP450 liver enzymes: CYP1A2 (n = 34), CYP2B6 (n = 39) and CYP3A4 (n = 39).Each black dot represents a donor, and an unpaired two-sided Mann-Whitney U test with 95% CI was used to assess the difference in enzyme induction levels between minor homozygous and nonminor homozygous groups.The p-values are shown at the top of each box plot.* p < 0.05.

Figure 4 .
Figure 4. Phase I CYP450 enzymes among healthy donors.The box plots represent minor homozygous and nonminor homozygous genotypes of PNPLA3 rs738409 (blue), MBOAT7 rs641738 (orange), and GCKR rs780094 (green) for phase I CYP450 liver enzymes: CYP1A2 (n = 34), CYP2B6 (n = 39) and CYP3A4 (n = 39).Each black dot represents a donor, and an unpaired two-sided Mann-Whitney U test with 95% CI was used to assess the difference in enzyme induction levels between minor homozygous and nonminor homozygous groups.The p-values are shown at the top of each box plot.* p < 0.05.

Table 1 .
LCMS conditions for the quantification of metabolites from drug-metabolizing enzymeselective pathways.