Gender-Dimorphic Impact of PXR Genotype and Haplotype on Hepatotoxicity During Antituberculosis Treatment

Supplemental Digital Content is available in the text


INTRODUCTION
T uberculosis (TB) remains a major infectious cause of deaths worldwide. To prevent transmission and future relapse, prompt and supervised anti-TB treatment for an extended period is very important. 1 However, the development of hepatotoxicity, which may be induced by anti-TB drugs or acute flare-up of concomitant viral hepatitis, is the most important adverse event leading to interruption or premature discontinuation of anti-TB treatment. 2 Among the first-line anti-TB drugs, isoniazid (INH), rifampin (RMP), and pyrazinamide (PZA) are all hepatotoxic and can increase the risk of hepatotoxicity further when used together.
For a long time, women have been reported to have a higher risk of hepatitis during antituberculosis treatment (HATT) than men. Although the definition of hepatotoxicity varies between studies, the hazard ratio of female sex ranges from 1.5 to 3.3. [3][4][5][6][7] A previous study in Taiwan revealed that female sex was a significant risk factor of drug-induced HATT and was independent of the N-acetyltransferase 2 (NAT2) status. 6 One possible reason is the activity of cytochrome P450 (CYP) 3A4, the most abundant enzyme in the hepatic CYP family that catalyzes the phase I reaction of many drugs, is higher in women, 8 given that many adverse drug reactions are caused by the CYP dependent activation of drugs into reactive metabolites. 9 The exact mechanism that leads to higher CYP3A4 activity in women than men is unknown. Single nucleotide polymorphisms (SNPs) in the coding region of the CYP3A4 gene occur only rarely and cannot explain the difference in CYP3A4 activity between men and women. 10 The pregnane X receptor (PXR), a member of the nuclear receptor superfamily, is a known regulator of the CYP3A4 gene. When bound by its ligand, PXR upregulates the expression of its target genes, including genes for phase 1 metabolizing enzymes such as CYP3A, and genes for all 3 phases of xenobiotic metabolism. 11,12 SNPs in the transcription factor binding sites of the PXR regulatory region (the promoter and intron1) have also been associated with altered PXR and CYP3A4 expressions, 13,14 as well as drug-induced liver injury. 15 Given the large spectrum of PXR activating ligands and target genes, and the association between SNPs in the PXR regulatory region and CYP3A4 expression, it is possible that gene variants in the PXR regulatory region may contribute to differences in risk of drug-induced HATT between male and female patients. We hypothesized that certain genotypes and haplotypes in PXR regulatory region SNPs may be risk factors for HATT, and the distribution of these genotypes and haplotypes may be different between male and female TB patients, leading to the increased risk of hepatotoxicity in females.

Study Population and Protocols
This prospective study was conducted at National Taiwan University Hospital, a tertiary-care center in Taiwan. The hospital's Institution Review Board approved the study (NTUH REB No.: 9561707008). All of the participants provided informed written consent.
From March 2007 to February 2010, adult patients (>16 years) with culture-confirmed pulmonary TB were enrolled as the derivation cohort. Mycobacterial culture and drug susceptibility testing were performed as previously described. 16 Subjects were excluded if they were pregnant, had a life expectancy <6 months, had abnormal baseline liver function test (LFT), or had Mycobacterium tuberculosis (MTB) isolates resistant to INH, RMP, or both. From March 2010 to February 2013, TB patients fulfilling these criteria were enrolled as the validation cohort.
Complete medical data and radiologic imaging were recorded. Alcohol abuse was defined as a daily consumption of !60 g of alcohol. 17 Malnutrition was defined as either serum albumin level <3.5 g/dL 6 or body-mass index <18.5 kg/m 2 . 18 For TB patients, LFT, including aspartate transaminase (AST) and alanine transaminase (ALT), direct and total bilirubin levels, were determined before the start of anti-TB treatment. Serologic tests for hepatitis B virus (HBV), hepatitis C virus (HCV), and human immunodeficiency virus, serum albumin, creatinine, and complete hemogram were also performed. The LFT was checked at 2, 4, 6, 8, 12, and 16 weeks after the start of anti-TB treatment or whenever symptoms of hepatitis developed during the initial 6 months of anti-TB treatment. 19 If there was elevated AST or ALT, the LFT was repeated and assessed weekly. For patients with concomitant HBV or HCV infection, serum HBV or HCV viral load was determined by quantitative PCR (Cobas Amplicor HBV and HCV monitor v2.0; Roche Diagnostics, Pleasanton, CA) simultaneously with LFT to document acute flare-up of viral hepatitis.
All TB patients received standard anti-TB treatment of daily INH, RMP, ethambutol (EMB), and PZA in the first 2 months, and daily INH and RMP for the next 4 months. The daily dosage of each drug was calculated by weight. 20 The regimen was modified by the primary care physician if necessary, for example, when there were adverse drug effects.

Definition and Etiology of Hepatitis During Antituberculosis Treatment (HATT)
HATT was defined as increased serum AST and/or ALT >3 times the upper limit of normal (ULN) in symptomatic patients, or >5 times the ULN in asymptomatic patients. 2 Once HATT occurred, potentially hepatotoxic drugs (INH, RMP, and PZA) were stopped. Anti-TB drugs were reintroduced after serum levels of AST and/or ALT returned to <3 times the ULN and the clinical symptoms of hepatitis resolved. As in a previous study, 6 the diagnosis of INH-or RMP-induced hepatitis required a positive rechallenge test (at least doubling of serum AST or ALT levels and recurrence of clinical symptoms of hepatitis after rechallenge), whereas PZA-induced hepatitis was diagnosed either by a positive rechallenge test or by exclusion. Virusinduced HATT was diagnosed if the rise in serum AST and ALT was associated with a concomitant rise in viral load.

Genotyping for PXR SNPs
Genotyping for the PXR and NAT2 genes for TB patients was performed on genomic DNA extracted from peripheral white blood cells. Laboratory technicians were blinded to the status of the participants during the entire process of PXR and NAT2 genotyping.
The PXR gene, also known as the nuclear receptor subfamily 1, group I, member 2 (NR1I2) gene, is located in chromosome 3. We first obtained a list of PXR SNPs based on predicted regulatory function or known association with diseases. 13,21-23 All SNPs were used as input files for the Haploview v4.1 (http:// www.broadinstitute.org/mpg/haploview) to search for tag SNPs in the genomic region of PXR that encompasses 36 kb and contains 33 polymorphic sites, using squared correlation (r 2 ) cutoff !0.95 and minor allele frequency (MAF) !0.1. We selected 6 tag SNPs in the regulatory region of PXR gene to be investigated: rs3814055 (located in the 5 0 untranslated region), rs12488820, rs2461823, rs7643645 (all located in intron 1), rs6785049 (located in intron 5), and rs3814057 (located in the 3 0 untranslated region) (Supplementary Figure S1, http:// links.lww.com/MD/A303).

Data Analysis and Statistical Analysis
All SNPs were tested for Hardy-Weinberg equilibrium. 25 Double data entry were performed to ensure data quality. Differences between groups were analyzed by independent sampled t test for continuous variables and by chi-square test or Fisher exact test for categorical variables. Linkage disequilibrium analysis was performed using Haploview. 26 Haplotype frequencies were calculated from PXR genotype data and analyzed using the EM algorithm by TagSNPs. 27 We collapsed rare haplotypes (frequency <0.05) into a category in final haplotype analyses.
The association between drug-induced HATT and clinical factors, NAT2 and PXR genotype frequencies, PXR allele frequencies, and PXR haplotype frequencies were analyzed using chi-square method, univariate, and multivariate logistic regression model. In the multivariate logistic regression analysis, an interaction variable between sex and PXR genotypes and haplotypes was also included. The identified predictors of drug-induced HATT were then validated using a validation cohort. A 2-sided p < 0.05 was considered significant. All analyses were performed using the SAS (Version 9.2, SAS Institute Inc., Cary, NC).

Derivation Cohort Case Enrollment
From March 2007 to February 2010 a total of 964 cases of culture-confirmed pulmonary TB were identified. Of the 964 cases, 222 cases were excluded due to the following reasons: 21 cases with multidrug-resistant MTB isolates which were resistant to both INH and RMP, 104 cases with INH-resistant MTB isolates, 3 cases with RMP-resistant MTB isolates, 51 cases with a life expectancy <6 months, 10 cases with abnormal baseline LFT (7 due to congestive heart failure and 3 due to excessive alcohol consumption), and 33 cases withdrew their consent later. Of the remaining 742 TB patients, 355 (36.8% of 964) agreed to participate, completed the study and formed the derivation cohort. Their mean age was 57.6 AE 19.4 years and 233 (65.6%) were male. The overall follow-up duration was 2004.5 person-months (5.6 months/patient).

Genotype of PXR SNPs and Association With
Overall Drug-Induced HATT All of the participants had PXR SNPs successfully genotyped, except for 1 man and 1 woman in rs2461823, 1 man in rs7643645, and 2 men in rs12488820 and rs3814057. All PXR polymorphisms met the Hardy-Weinberg equilibrium (p > 0.05) except for rs12488820 (p < 0.001) and rs3814057 (p ¼ 0.002). Table 2 shows the risk of drug-induced hepatitis among TB patients with different SNP genotypes in PXR gene (see also Supplementary Figure S2, http://links.lww.com/MD/A303). In male TB patients, none of the 6 SNPs in the PXR gene was associated with the development of overall drug-induced HATT, with similar genotype frequencies between male patients with hepatitis and those without. In female TB patients, 2 PXR variants were significantly associated with the risk of overall druginduced hepatitis. The frequency of drug-induced HATT in female patients with AA genotype at rs2461823 site (50%) was significantly higher compared to female patients carrying other genotypes (AG, 22%; GG, 15%, p ¼ 0.007). The higher risk of AA genotype at rs2461823 in females was also shown in the recessive model ( Table 1, NAT2 variants, PXR genotypes, allele numbers, haplotypes, and sex interaction, was performed to identify independent predictors of overall and individual druginduced HATT. Table 4 shows that the only independent predictors of overall drug-induced HATT were PXR variants genotypes, allele numbers, and haplotypes. Genotype analysis revealed that the female genotype AA at rs2461823 site (OR 4.64 [1.96-11.0]), a risk genotype, and the female genotype AA at rs7643645 site (OR: 0.14 [0.02-1.02]), a protective genotype, were both independently associated with overall druginduced HATT. Allele number analysis showed that the number  Bold value signifies the statistical significance. CI ¼ confidence interval, OR ¼ odds ratio, PXR ¼ pregnane X receptor.   Table 5 shows the influence of genotypes at each PXR SNP site on the risk of drug-induced HATT. Although none of the SNP genotypes at rs7643645 or rs2461823 were significantly associated with HATT by chi-square test, risk of HATT was higher in female patients with AA genotype at rs2461823 (56%) and GG genotype at rs7643645 (39%). Table 6 compares risk of drug-induced HATT in patients with or without risk predictors. It is evident that the risk of drug-induced HATT was significantly different between the 3 subgroups with different combination of genotypes at rs7643645 and rs2461823 (p ¼ 0.018), borderline different in the 3 subgroups with different number of G allele at rs7643645 (p ¼ 0.057), and significantly different in the 4 subgroups with different haplotypes (p ¼ 0.008). Although the P values in the female patients alone did not reach statistical significance, probably owing to the small sample size of the validation group, these findings suggested that the risk of drug-induced HATT was influenced by above-mentioned PXR variants and haplotypes only in females.

DISCUSSION
By prospective observation and genotyping in 355 cultureconfirmed pulmonary TB patients and validation in another 182 cases, this is the first study to show that the SNPs of the PXR gene are independent risk factors of overall drug-induced hepatitis in TB patients, and that the association occurs only in women. Risk factors of hepatitis for individual first-line anti-TB drugs vary, but PXR SNP genotypes and haplotypes remain significant risk factors and show gender difference.
Both rs7643645 and rs2461823 were located at intron 1b of the regulatory region close to the PXR promoter. Rs7643645 is located in the binding site of the transcription factor, hepatic nuclear factor (HNF)4a, and a change from the wild A allele to the mutant G allele leads to a loss of the HNF4a binding site. 13 Holloway et al 28 found that in a mouse model, HNF4a exerted both a positive and a negative regulatory effect on many hepatic genes. With loss of the HNF4a, 82% of the 4994 HNF4adependent genes were suppressed in males. In contrast, only 56% of HNF4a-dependent genes were suppressed, while some of the HNF4a-dependent genes were upregulated upon loss of HNF4a in females. 28 Such findings suggest that PXR gene expression (and its regulation on target xenobiotic metabolism genes) in response to HNF4a loss is different between the 2 sexes.
As for rs2461823, there has been no transcription factor binding sequence identified in this site to date. Nonetheless, it is in strong linkage disequilibrium with several other SNP sites in the PXR promoter region, including rs2472677, which is located within a hepatocyte nuclear factor 3-b (HNF3b) binding site. Its CC genotype is associated with high level PXR induction by RMP. 29 In mice, the HNF3b binding site has been associated with female predominant expression of the CYP2b9 gene. 30 Thus, the rs2461823 AA genotype may occur in strong linkage disequilibrium with certain genotypes of other SNPs (such as rs2472677) that bear a transcription factor binding sequence, and this may become associated with different risk of HATT between males and females.
A previous study found that the rs7643645 GG and rs2461823 AA genotypes were associated with increased risk and/or severity of nonalcoholic fatty liver disease, but without any sex difference. 21 Another study found that the rs2461823 GG genotype was associated with susceptibility to intrahepatic cholestasis of pregnancy. In that study all of the subjects were females. 29 These reports, together with the current findings, suggest that these 2 PXR SNPs affect detoxification and elimination pathways of drugs like INH and PZA, either in females alone or in both sexes. The underlying mechanisms of these observations may be a loss of PXR transcription factor binding site. The findings that SNP in rs2461823 of the PXR gene is associated with both INH-and PZA-induced hepatitis have never been reported. The hepatotoxin that leads to INH-induced hepatitis has been proposed to be hydrazine and other hydrazine metabolites that can generate free radicals. The conversion from INH to hydrazine involves either a classical P450 oxidase that can be induced by PXR ligands like RMP, phenobarbital or dexamethasone, 31 or an amidase which can also catalyze the reaction from monoacetylhydrazine to hydrazine (Supplementary Figure S3, http://links.lww.com/MD/A303). 32 Monoacetylhydrazine can be converted to free radical hepatotoxins by a P450-dependent CYP2E1. Therefore, PXR could increase the production of hydrazine, free radical hepatotoxins, and thus the risk of INH-induced hepatitis, through P450 oxidase and CYP2E1. Whether PXR could influence amidase is unknown. In addition, a recent study has demonstrated that cotreatment with INH and RMP causes liver injury through PXR-mediated alteration of the heme biosynthesis pathway. 33 As for PZA-induced hepatitis, recently PZA metabolites pyrazinoic acid and 5-hydroxypyrazinoic acid have been reported to correlate with PZA-induced hepatitis. 34 The conversion of PZA to pyrazinoic acid, and that of 5-hydroxy-PZA to 5-hydroxypyrazinoic acid both involve an amidase (Supplementary Figure S4, http://links.lww.com/MD/A303). 35 The other source of 5-hydroxypyrazinoic acid production is from pyrazinoic acid via xanthine oxidase. There has been no report regarding to the effect of PXR on the activity of amidase or xanthine oxidase. Since both INHinduced and PZA-induced hepatitis were associated with rs2461823 variant, it was also possible that PXR's effect was through its regulation on amidase, which metabolizes both INH and PZA. Further studies are needed to clarify these hypotheses.
The mechanism of hepatotoxicity due to rifampicin and its derivatives remains unknown. It is hydrolyzed in liver by acrylacetamide deacetylase (AADAC; for rodents, Aadac), a member of the carboxylesterase-5 family, which catalyze the hydrolysis of many ester-and amide-containing chemicals. 36 A previous study in mice revealed that Aadac mRNA was highly expressed in mouse livers and was suppressed by 2 PXR ligands-pregnenolone-16a-carbonitrile and dexamethasone. 37 However, whether the Aadac mRNA expression was altered by the SNP polymorphisms of the PXR gene was not studied. The finding that the 1 genotype and 2 haplotypes of the PXR gene were significantly associated with RMP-induced hepatitis in the study suggests that PXR may regulate the metabolism of RMP through AADAC and alter the risk of hepatitis. Further studies are needed to confirm these finding and investigate the underlying mechanisms.
From the results of genotype distribution we calculated the risk of anti-TB drug-induced HATT associated with haplotypes, and found that 2 haplotypes were also associated with the risk of HATT by multivariate analysis. We observed that all the 3 PXR SNPs (rs2461823, rs7643645, and rs6785049) contributed to the association between haplotypes and HATT.
We also observed that NAT2 slow acetylator and malnutrition were independent risk factors for INH-induced hepatitis both in males and females, consistent with previous reports. 6,38 Malnutrition has been reported to reduce the activity of hepatic glutathione S-transferase and increase vulnerability to oxidative injury, 39 leading to increased risk of INH-induced hepatotoxicity in TB patients.
The finding that end-stage renal disease is associated with RMP-induced hepatotoxicity is unexpected since 60% to 65% of RMP dose appears in feces, and it does not accumulate in patients with impaired renal function (Rifadin package insert, Marion Merrell Dow, Ohio, US). Because uremic toxins can alter the hepatic clearance of many drugs, either by downregulation of specific isoforms of CYP via affecting promoter 40 or by impaired hepatic uptake mediated by uptake transporters, 41 it is possible that the increased risk of RMP-induced hepatotoxicity in patients with end-stage renal disease was associated with the accumulation of uremic toxin. The present study has limitations regarding the interpretation of its findings. First, the mechanisms of INH-, RMP-, or PZA-induced HATT are likely to be different, and mixing together all patients with drug-induced HATT may mask unknown risk factors. Yet we analyzed risk factors for individual drugs and found that risk factors for individual drugs also included PXR variants. Second, drug-metabolizing enzymes other than PXR and NAT2 were not genotyped. Third, serum levels of drugs and toxic metabolites were not measured. Nonetheless, even if serum levels of metabolites have been shown to be highly predictive for some of HATT, routine measurement is not practical. Fourth, the validation cohort is small in sample size and did not directly replicate the results of the derivation cohort. Yet when we further analyzed patients with the presence or absence of the risk factors that were identified in the derivation cohort, the influence of PXR variants and haplotypes was still evident in the female patients of the validation cohort. Lastly, the study was conducted in a medical center and nearly half of the TB patients had underlying comorbidities that might influence laboratory results and radiographic findings.

CONCLUSIONS
This study is the first to show that 2 PXR SNP genotypes and 2 haplotypes influenced the risk of HATT only in females. The PXR gene variants have sex-dimorphic impact that contributes to the increased risk of drug-induced HATT in females.