A role for the pregnane X receptor in flucloxacillin-induced liver injury


  • Potential conflict of interest: Nothing to report.


Drug-induced liver injury (DILI) due to flucloxacillin is a rare but serious complication of treatment. There is some evidence that flucloxacillin is a human pregnane X receptor (PXR) agonist. This study was designed to investigate the relevance of PXR to flucloxacillin toxicity and to identify genes changing in expression in response to flucloxacillin. Changes in gene expression in human hepatocytes after treatment with 500 μM flucloxacillin for 72 hours were examined by expression microarray analysis. The ability of flucloxacillin to act as a PXR agonist was investigated with reporter gene experiments. Flucloxacillin DILI cases (n = 51), drug-exposed controls without toxicity (n = 64), and community controls (n = 90) were genotyped for three common PXR polymorphisms. Luciferase reporter assays were used to assess the significance of a promoter region PXR polymorphism. Seventy-two probe sets representing 50 different genes showed significant changes in expression of 1.2-fold or higher. Most genes showing changes greater than 3-fold were known to be rifampicin-responsive, and this suggested a PXR-dependent mode of regulation. Using a luciferase-everted repeat separated by 6 base pairs element construct, we confirmed that flucloxacillin was a PXR agonist. We found a difference in the distribution of a PXR polymorphism (rs3814055; C-25385T) between flucloxacillin DILI cases and controls with the CC genotype associated with an increased risk of disease (odds ratio = 3.37, 95% confidence interval = 1.55-7.30, P = 0.0023). Reporter gene experiments showed lower promoter activity for the C allele than the T allele. Conclusion: Flucloxacillin is a PXR agonist at pharmacologically relevant concentrations, and a functionally significant upstream PXR polymorphism is a risk factor for flucloxacillin-induced DILI. Hepatology 2010

Adverse hepatic reactions have accounted for 23% of postmarketing drug withdrawals in the United Kingdom since 1975 and have been the leading cause of such withdrawals in the last 3 decades.1 Despite progress in toxicological studies, the correlation between liver toxicity in animals and man is poor, and idiosyncratic hepatotoxicity remains one of the principal causes of termination in clinical trials of new chemical entities. A study based in France estimated a standardized annual incidence rate of symptomatic hepatic adverse reactions of 8 per 100,000 inhabitants.2

Flucloxacillin is an effective antimicrobial against staphylococcal infections and is used widely in Europe and Australia in the treatment of these infections. The use of flucloxacillin has been associated with characteristic cholestatic hepatitis, which appears to be more common in females than males3 and to increase in frequency with the age of the patient and the duration of treatment.4 The incidence of drug-induced liver injury (DILI) due to flucloxacillin has been estimated to be 8.5 per 100,000 new users in the first 1 to 45 days after treatment is started.5 This adverse event can be severe, with a mortality rate of around 1 in 20 and with transplantation required in some cases.3, 6 Reports from Europe7-9 and Australia10 have confirmed that flucloxacillin contributes significantly to the burden of DILI.

The mechanism underlying flucloxacillin DILI is unclear, although there is some evidence pointing to both immune and metabolic mechanisms. A recent study found a strong association with the class I human leukocyte antigen (HLA) allele B*5701, and this suggests an immune contribution to the disease involving inappropriate T cell responses.11 Evidence for specific T cell reactivity in a single case of the disease has been reported, although the level of stimulation was low in comparison with other drugs studied.12 In vitro studies using cultured human hepatocytes and biliary epithelial cells have shown that the 5′-hydroxymethyl metabolite of flucloxacillin, which is generated by cytochrome P450 3A4 (CYP3A4), is toxic to biliary epithelial cells but not to hepatocytes, whereas the parent compound is nontoxic, even at high concentrations, to either cell type.13

The increase in intracellular bile acid concentrations seen during cholestasis initiates a number of negative feedback responses to limit hepatotoxicity, including decreased bile acid uptake and increased efflux into the bile canaliculi.14 The nuclear pregnane X receptor (PXR; nuclear receptor subfamily 1, group I, member 2) plays an important role in the regulation of the expression of a number of proteins contributing to bile acid homeostasis.15 A recent report suggests that flucloxacillin is a PXR ligand,16 and another isoxazolyl penicillin, dicloxacillin, has been shown to increase PXR-regulated gene expression at both the messenger RNA (mRNA) and protein levels.17 To study flucloxacillin interactions with PXR in more detail and to identify novel candidate genes that might contribute to flucloxacillin DILI, we have examined changes in gene expression patterns in human hepatocytes after treatment with flucloxacillin. We have also determined the relevance of the PXR genotype to flucloxacillin DILI because our microarray and additional functional studies point to an important role for PXR in cellular responses to flucloxacillin.

Materials and Methods


ABC, adenosine triphosphate-binding cassette; ADH1, alcohol dehydrogenase 1; AFM, alpha albumin; AGXT2L1, alanine-glyoxylate aminotransferase 2-like 1; AKR1B10, aldo-keto reductase family 1 member B10; ANOVA, analysis of variance; cDNA, complementary DNA; CI, confidence interval; CYP, cytochrome P450; DILI, drug-induced liver injury; GCRMA, GeneChip robust multi-array average; GNMT, glycine N-methyltransferase; HLA, human leukocyte antigen; MALAT1, metastasis-associated lung adenocarcinoma transcript 1; mRNA, messenger RNA; MRS2L, MRS2 magnesium homeostasis factor homologue; MTT, 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyltetrazolium bromide; OR, odds ratio; PCR, polymerase chain reaction; PGLYRP2, peptidoglycan recognition protein 2; PXR, pregnane X receptor; RT-PCR, reverse-transcriptase polymerase chain reaction; SNP, single nucleotide polymorphism; STAT, signal transducer and activator of transcription; TAT, trans-activator of transcription.

Cell Culture.

HepG2 cells (from human hepatoblastoma) and LS180 cells (from human colon adenocarcinoma) were cultured in Eagle's minimum essential medium (Bio Whittaker, Rockland, ME), which contained 10% heat-inactivated fetal calf serum, 1% nonessential amino acids, 2 mM glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin.

Primary human hepatocytes were obtained from liver resections through the UK Human Tissue Bank (De Montford University, Leicester, United Kingdom). Cells were seeded at 1.2 million cells per well on six-well plates and cultured in hepatocyte maintenance medium (Clonetics, MD) supplemented with ITS+ premix (BD Biosciences, Oxford, United Kingdom), penicillin/streptomycin, and dexamethasone. Cells were allowed to rest for 24 hours before being treated with either flucloxacillin (CP Pharmaceuticals, Ltd., Wrexham, United Kingdom; generally 500 μM) or a vehicle control for 72 hours. The maintenance medium was replaced every 24 hours.

MTT Test for Cytotoxicity.

Assays were performed in six-well tissue culture plates. After treatment with the drug or vehicle, the growth medium was replaced with fresh medium including 0.1 volumes of 5 mg/mL 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyltetrazolium bromide (MTT; Sigma, Poole, United Kingdom) in phosphate-buffered saline and incubated at 37°C for 2 hours. The MTT-containing medium was removed and replaced with 1 mL of acidified isopropanol. Plates were then incubated at 4°C for 45 minutes. After brief centrifugation, the supernatant was diluted 1:5 with phosphate-buffered saline, and the absorbance at 570 nm was measured.

Expression Microarray Analysis.

RNA was extracted from human hepatocytes (from three separate donors) with the RNeasy mini kit (Qiagen, United Kingdom). A GeneChip expression 3′-amplification 2-cycle complementary DNA (cDNA) synthesis kit was used to amplify 100 ng of total RNA (Affymetrix, High Wycombe, United Kingdom). The cDNA was amplified by in vitro transcription with biotin labeling using the GeneChip in vitro transcription labeling kit and then was fragmented and hybridized to Affymetrix U133+2 arrays according to the manufacturer's guidelines (http://www.affymetrix.com).

Expression data were preprocessed with the GeneChip robust multi-array average (GCRMA) algorithm on GeneSpring software. Principal component analysis was used to identify outlier chips, which were removed from further analysis because of batch effects. The remaining chips were then processed with microarray suite version 5.0 (Affymetrix) (MAS5) and GCRMA. Probe sets with absent calls in half the chips processed with MAS5 were discarded from further analysis. The remaining probe sets, normalized with GCRMA, were analyzed with RankProd (http://www.ncbi.nlm.nih.gov/pubmed/16982708) in BioConductor (http://www.ncbi.nlm.nih.gov/pubmed/15461798) to produce a list of probe sets significantly differentially expressed (P < 0.05) after 72 hours of treatment with 500 μM flucloxacillin versus the vehicle control. Subsequent visualization and investigation were performed with GeneSpring GX 10 (Agilent, Berkshire, United Kingdom).

Quantitative Reverse-Transcriptase Polymerase Chain Reaction (RT-PCR).

Expression microarray results were validated by quantitative polymerase chain reaction (PCR) with the same flucloxacillin and vehicle-treated RNA extracted for the microarray expression analysis. Studies were also performed with RNA extracted from LS180 cells treated with 20 μM rifampicin, 500 μM flucloxacillin, or a vehicle control for 72 hours. RNA (1 μg) was reverse-transcribed with a cDNA synthesis kit (Bioline, London, United Kingdom). Primer sequences for RT-PCR are listed in the supporting information.

cDNA PCR was run in a 25-μL final volume with SYBR Green JumpStart Taq ReadyMix for quantitative PCR (Sigma). We checked primer specificity by running melting curves. The Ct thresholds were set and checked with a standard curve of cDNA dilutions. All assays were run in triplicate. Expression levels were normalized to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase. Normalized values were used to calculate changes in expression between control and flucloxacillin-treated cells with the ΔΔCt method.18

Reporter Gene Assays.

To determine whether flucloxacillin was a PXR ligand, a plasmid construct containing a concatamer of three PXR response elements everted repeats separated by 6 base pairs (ER6) ligated to a luciferase reporter gene in the pGL3-promoter vector [(ER6)3-pGL3-prom-luc; Promega, Southampton, United Kingdom] was used.19 HepG2 cells were transfected with 0.5 μg of either the full construct or an empty construct lacking the ER6 element. The cells were cotransfected with 0.05 μg of a renilla-thymidine kinase construct (Promega) as an internal control. HepG2 cells express high levels of PXR,19 and therefore it was not necessary to additionally transfect cells with PXR. All transfections were performed in a serum-free medium with Genejuice (Merck, Nottingham, United Kingdom) for 2.5 hours before it was replaced with fresh medium. After transfection, cells were treated for 72 hours with vehicle only, flucloxacillin, or rifampicin.

To determine whether the PXR promoter region polymorphism at position -25385 affected promoter activity, a 945-bp section of the PXR promoter region was studied. This included the 212-bp maximal activity region directly upstream of the transcription start site20 together with sufficient upstream sequence to encompass the site of the -25385 polymorphism 831 bp upstream of the transcription start site.21 This region was cloned by PCR of genomic DNA from a -25385 CC homozygote and a -25385 TT homozygote. The PCR primers were GTCAGAGCTCATTTCCCCGGATATGAGACA (forward) and TGCTGGTACCTGAAGACAACTGTGGTCATT (reverse). Products were ligated into pGEM easy T (Promega) and then subcloned into pGL3 basic (Promega). After the cloning, the inserts were sequenced to check that no new mutations or additional polymorphisms were present. HepG2 cells plated at 1 × 104 cells per well in a 24-well plate were transfected with 5 μg of a pGL3 basic vector containing the insert. Cells were cotransfected with 0.05 μg of a renilla-thymidine kinase construct (Promega). All transfections were performed in a serum-free medium with Genejuice (Merck) for 5 to 8 hours before the addition of fresh complete medium. After 24 hours, the medium was replaced, and the cells were incubated at 37°C for 72 hours.

For both types of reporter assays, cells were lysed with a cell lysis buffer (Promega), and this was followed by an analysis with a dual luciferase reporter assay system (Promega). Levels of firefly luciferase and renilla luciferase were measured in a luminometer (EG&G, Berthold, Germany).

Patient Recruitment.

Flucloxacillin DILI cases and controls were recruited from centers throughout the United Kingdom. The study was approved by the Leeds East research ethics committee, and all participants gave written, informed consent. Full details on cases (n = 51) and drug-exposed controls (n = 64), including clinical data and inclusion criteria, have been described elsewhere.11 A further control group consisted of 90 community controls from a population described previously.22 All cases and controls in this study were of self-reported white European origin.


DNA was extracted from blood samples as described previously.23 PCR reactions were performed in 25 μL with a final concentration of 0.025 units of Taq DNA polymerase per microliter (Promega), 0.1 mM deoxynucleotide triphosphates (Qbiogene, United Kingdom), and 0.25 μM forward and reverse primer in a 5× reaction buffer (Promega).

As the linkage disequilibrium patterns and possible functional significance of single nucleotide polymorphisms (SNPs) in PXR are quite well established,24, 25 we decided to genotype for the upstream C-25385T (rs3814055), A7635G (rs6785049) in intron 5, and A11156G (rs3814057) in the 3′-untranslated region. All PXR genotyping assays used PCR/restriction fragment length polymorphism; details for the primers, restriction enzymes, and digests are listed in the supporting information. After PCR, samples were diluted into the appropriate digestion buffer (New England Biolabs, Herts, United Kingdom) and digested for 16 hours with 2 U of enzyme. Digestion products were separated on 10% polyacrylamide gels in a trishydroxymethylaminomethane/borate/ethylene diamine tetraacetic acid buffer at 150 V for 4 hours. Assays were validated through the sequencing of selected PCR products of representative genotypes, and controls of a known genotype were included in all sets of analyses.

Samples were genotyped for B*5701 with Gold SSP B17 high-resolution kits (Invitrogen, Renfrew, United Kingdom) as described previously.11

Statistical Analysis.

Genotype distributions between cases and controls were compared with Fisher's exact test, and odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. Haplotypes were assigned with PHASE 2.1.26 For quantitative RT-PCR, statistical significance was assessed by two-way analysis of variance (ANOVA) followed by subgroup analysis with Bonferroni correction. For reporter gene experiments, statistical significance was assessed with t tests and one-way ANOVA.


Analysis of Changes in Gene Expression in Hepatocytes Following Flucloxacillin Treatment.

To identify genes subject to changes in expression when human hepatocytes were exposed to flucloxacillin, expression microarray analysis was performed. Before hepatocytes were treated with flucloxacillin, a suitable nontoxic dose needed to be established. This was done with an MTT cytotoxicity assay of hepatocytes from a single donor with flucloxacillin concentrations in the range of 100 μM to 1 mM; viability was checked after 72 hours. A decrease in viability greater than 20% was considered to be toxic. For 1 mM flucloxacillin, some toxicity was seen (72% of the viability of the vehicle-treated control, P = 0.004). It was therefore decided to use a 500 μM flucloxacillin treatment for 72 hours to examine changes in gene expression because of the absence of toxicity at this dose. Drug-treated cells were compared to vehicle-treated controls with hepatocytes from three donors in separate experiments. Initial principal component analysis was performed on the data to identify any outlier chips, which identified a batch effect, and four chips were excluded from further analysis. A total of 72 probe sets showed significant changes in expression with the rank products method27 with a P value cutoff of <0.05. Twenty-two of these probe sets, representing 10 different genes, showed increased expression, with 50 probe sets, which represented 40 different genes, showing decreased expression. All but two of the significant probe sets showed fold changes greater than 1.2. Probe sets and genes found to be increased or decreased in expression more than 3-fold are listed in Table 1. The genes listed in Table 1 have previously been reported to be subject to PXR regulation on the basis of microarray analysis studies with rifampicin (GSE 10410 from Gene Expression Omnibus data sets at http://www.ncbi.nlm.nih.gov), with the exception of aldo-keto reductase family 1 member B10 (AKR1B10), peptidoglycan recognition protein 2 (PGLYRP2), metastasis-associated lung adenocarcinoma transcript 1 (MALAT1), and alpha albumin (AFM). A complete list of all probe sets showing significant changes in expression is provided in the supporting information.

Table 1. Probe Sets and Corresponding Genes Showing Alteration in Expression in Hepatocytes Greater than Three-Fold in Response to Treatment with 500 μM Flucloxacillin for 72 Hours
GeneProbe NameAverage Fold Change in mRNA Expression in Three Donors
  • All changes were statistically significant (P < 0.05).

  • Abbreviations: ADH1, alcohol dehydrogenase 1; GNMT, glycine N-methyltransferase; MRS2L, MRS2 magnesium homeostasis factor homologue; TAT, trans-activator of transcription.

  • *

    Genes not known to be regulated by PXR and not detected in a previous microarray expression analysis study using rifampicin (GSE 10410 in the Gene Expression Omnibus Database of the National Center for Biotechnology Information).

Increased expression  
Decreased expression  

Quantitative RT-PCR Analysis.

To confirm the expression microarray findings, quantitative RT-PCR for selected representative genes was performed on RNA from both hepatocytes and LS180 cells. The obtained data are summarized in Figs. 1 and 2. The gene for which expression increased most in the expression microarray experiment was CYP3A4. Quantitative PCR on mRNA from two separate hepatocyte preparations treated with 500 μM flucloxacillin found increases of 14-fold and 11.8-fold for CYP3A4 and vehicle-treated cells, respectively, (Fig. 1) with both changes significant (P < 0.001). This can be compared to the slightly lower 8.9-fold increase seen in the expression microarray study (Table 1). In general, the increased gene expression seen by the microarray analysis was confirmed, although the increase for one gene [alanine-glyoxylate aminotransferase 2-like 1 (AGXT2L1)] failed to show statistical significance in either donor. Quantitative PCR for CYP3A4 was also run on RNA from LS180 cells treated with 500 μM flucloxacillin for 72 hours. LS180 cells were used as they have been shown to be a good model cell line for PXR induction of both CYP3A4 and adenosine triphosphate-binding cassette B1 (ABCB1).28 We found a 6.8-fold increase in the level of the CYP3A4 transcript in comparison with the vehicle-treated control (P < 0.001; Fig. 2). As with hepatocytes, the level of ABCB1 induction was lower than that for CYP3A4. Levels of induction of both genes were higher for RNA from cells treated with 20 μM rifampicin, which was used as a positive control, although only the induction of CYP3A4 by rifampicin was statistically significant (P < 0.001).

Figure 1.

Quantitative RT-PCR on mRNA samples of hepatocytes from two different donors treated with flucloxacillin (500 μM) for 72 hours. Seven genes showing significantly increased expression by microarray analysis were selected for this further analysis. The fold change in expression with respect to vehicle-treated controls is shown. ***P < 0.001 and *P < 0.05 versus the corresponding vehicle control by subgroup analysis with Bonferroni correction following two-way ANOVA.

Figure 2.

Quantitative RT-PCR on mRNA samples of LS180 cells treated with flucloxacillin (500 μM) or rifampicin (20 μM) for 72 hours. The fold change in the expression of CYP3A4 and ABCB1 was determined. ***P < 0.001 versus the corresponding vehicle control by subgroup analysis with Bonferroni correction following two-way ANOVA.

Study of Flucloxacillin Induction Using Reporter Gene Constructs.

To determine more directly whether flucloxacillin is a PXR ligand, a reporter gene system with an ER6 response element specific to PXR was used. HepG2 cells were transfected with either ER6-positive or ER6-negative luciferase reporter gene constructs together with renilla-thymidine kinase constructs to normalize results. Cells were exposed to various flucloxacillin concentrations (500 μM to 2 mM) for 72 hours before reporter gene activity was measured. For the ER6-negative construct, no significant luciferase stimulation was detected by treatment with either rifampicin or flucloxacillin (P = 0.21). However, treatment of cells transfected with the ER6-positive construct with either rifampicin (20 μM) or flucloxacillin resulted in increased luciferase activity (P = 0.001). Using the method described by Plant et al.29 for calculation of the specific chemical effect, we saw a 5.1-fold stimulation of luciferase by the 20 μM rifampicin treatment versus 3.7-fold and 2.7-fold stimulation by flucloxacillin at 1 and 2 mM, respectively (Fig. 3). These data provided an approximate estimation of the 50% effective concentration for flucloxacillin of 460 μM; we assumed that the 20 μM rifampicin treatment was associated with maximal induction.

Figure 3.

Effect of flucloxacillin on luciferase expression in HepG2 cells with the ER6 promoter luciferase construct. The specific chemical effect of flucloxacillin (0.5-2 mM) was determined with 20 μM rifampicin as the control. The effect of the vehicle-only treatment was taken to be 1.

Genotyping of Flucloxacillin DILI Cases and Controls.

To further explore PXR as a potential candidate gene for flucloxacillin-induced liver injury, three SNPs in the gene were genotyped in a cohort of flucloxacillin DILI cases, drug-treated controls, and community controls. In the case of one SNP, C-25385T, CC homozygotes were more frequent, with the frequency increasing from 28% in the control population to 57% among cases (Table 2). The OR for possession of the CC genotype in the case population compared with drug-exposed controls was 3.37 (95% CI = 1.55-7.30, P = 0.0023). The other two SNPs studied, at positions 7635 and 11,156, were not found to be significantly associated with flucloxacillin DILI (Table 2). All genotypes were in Hardy-Weinberg equilibrium in the community control groups, but genotypes for rs3814057 in the drug-exposed controls and for rs3814055 in the cases were not in equilibrium. The failure to see Hardy-Weinberg equilibrium for the at-risk SNP in the cases was not surprising, but there was no obvious explanation for this failure for rs3814057 in the drug-exposed controls. Genotyping errors were unlikely because validation of the assay by sequencing and genotyping for this polymorphism was repeated by another operator who was blinded.

Table 2. Genotype Frequencies of PXR Polymorphisms in Cases and Controls
  1. The OR for CC (rs3814055) in cases was 3.37 (1.55-7.30) versus drug-exposed controls (P = 0.0023) and 2.63 (1.30-5.34) versus community controls (P = 0.0079). The OR for AA (rs6785049) in cases was 1.32 (0.63-2.76) versus drug-exposed controls (P = 0.57) and 1.44 (0.73-2.89) versus community controls (P = 0.38). The OR for carriage for AA (rs3814057) in cases was 2.62 (95% CI = 1.08-6.32) versus drug-exposed controls (P = 0.04) and 1.33 (0.56-3.20) versus community controls (P = 0.66).

−25385 (rs3814055)CCCTTT
Cases (n = 51)29 (0.57)14 (0.27)8 (0.16)
Flucloxacillin controls (n = 64)18 (0.28)36 (0.56)10 (0.16)
Community controls (n = 90)30 (0.33)49 (0.54)11 (0.12)
7635 (rs6785049)AAAGGG
Cases (n = 51)25 (0.49)19 (0.37)7 (0.14)
Flucloxacillin controls (n = 64)27 (0.42)27 (0.42)10 (0.16)
Community controls (n = 90)36 (0.40)42 (0.47)12 (0.13)
11156 (rs3814057)AAACCC
Cases (n = 51)42 (0.82)9 (0.18)0 (0.0)
Flucloxacillin controls (n = 64)41 (0.64)22 (0.34)1 (0.02)
Community controls (n = 90)70 (0.78)19 (0.21)1 (0.01)

Haplotypes were also assigned to determine whether the PXR haplotype was more predictive for flucloxacillin DILI susceptibility than the -25385 genotype alone. However, there was no significant difference seen in the haplotype distribution between the cases and drug-exposed controls (data not shown).

In a previous study, possession of the HLA class I allele B*5701 was found to be associated with a significantly increased risk of flucloxacillin DILI with an OR of 80 (95% CI = 22.8-284.9) for cases compared with drug-exposed controls.11 To determine whether the effect of the PXR genotype differed in those positive for B*5701 and those negative for this allele, we compared PXR genotype distributions for cases positive for B*5701 and those negative for it separately against the distributions in all flucloxacillin-exposed controls. It was not feasible to do this against B*5701-positive controls because only 4 of the 64 controls were positive for this allele. As shown in Table 3, the B*5701-positive cases showed an increased OR of 3.91 (95% CI = 1.72-8.87, P = 0.0013) for possession of the CC genotype, whereas significance was lost for the small number of B*5701-negative cases.

Table 3. PXR C-25385T Genotypes with Respect to HLA-B*5701 Genotypes in Cases and Drug-Exposed Controls
  1. The OR for CC in B*5701 positives versus all drug-exposed controls was 3.91 (95% CI = 1.72-8.87, P = 0.0013). The OR for CC in B*5701 negatives versus all drug-exposed controls was 1.53 (95% CI = 0.33-7.10, P = 0.68).

B*5701-positive (n = 43)26107
B*5701-negative (n = 8)341

A comparison of gender, causality score, average time on the drug, and age at onset was performed: both CC genotypes for PXR were compared against the other genotypes, and those positive for both PXR CC and HLA-B*5701 were compared against all other genotypes, but no significant differences between groups were detected (data not shown).

Cases and controls were also genotyped for CYP3A5*3 because of a possible role for CYP3A5 in flucloxacillin metabolism, the clear functional significance of the polymorphism defining this allele, and the regulation of CYP3A5 by PXR. There was no significant difference in genotype distribution between cases and controls (data not shown).

Effect of Genotype on PXR Promoter Activity.

To determine whether the C-25385T polymorphism has a functionally significant effect on PXR expression, a luciferase reporter gene system was used. Luciferase activity in HepG2 cells transfected with the -25385C-containing promoter region was found to be approximately 4-fold lower (P = 1 × 10−4) than the promoter activity in the -25385T-transfected cells (Fig. 4).

Figure 4.

Comparison of luciferase reporter gene activity between -25383C-containing and -25385T-containing PXR promoter regions. Luciferase activity is shown for the C-containing and T-containing constructs together with the cloning vector lacking a PXR promoter region.


DILI, like other idiosyncratic adverse drug reactions, is believed to arise as a result of both genetic and environmental factors. To identify novel candidate genes for flucloxacillin DILI, we performed an expression microarray study on human hepatocytes treated with flucloxacillin. The concentration used was comparable to the peak levels reported to be achieved in plasma after a 1-g dose.30 The results show that flucloxacillin alters the expression of a number of genes, including CYP3A4 and ABCB1, in human hepatocytes, and they confirm previous, more preliminary observations by others.16 Most genes whose expression changed were PXR targets, and few novel candidates emerged. However, PXR appeared to be an interesting candidate gene in its own right. We extended the previous study16 by performing reporter gene studies with the ER6 construct; this showed directly that flucloxacillin was a PXR agonist, though weaker than rifampicin. Although CYP3A4 could also be viewed as a potential candidate gene for flucloxacillin DILI, previous studies have shown that polymorphisms in this gene are generally rare and fail to fully explain the observed interindividual variation in CYP3A4 activity (for a review, see Ingelman-Sundberg et al.31). The homologous CYP3A5 gene, which is expressed only in approximately 10% of Europeans, was an additional candidate gene. The common polymorphism in CYP3A5 giving rise to an absence of protein (CYP3A5*3 allele) has been clearly shown to be a good predictor of levels of CYP3A-mediated metabolism for some drugs (for a review, see Daly32), but no differences in genotype frequency between our cases and controls for the diagnostic SNP were detected. Whether flucloxacillin is a CYP3A5 substrate is still unclear.

To determine whether certain SNPs in the PXR gene were relevant to flucloxacillin DILI, we undertook genotyping in a small group of flucloxacillin DILI cases and drug-exposed controls. The large number of genes and biological processes under PXR control means that a polymorphism altering PXR function could have a large number of downstream consequences. Genotyping the cases and controls for the -25385 PXR promoter region polymorphism revealed an increased frequency of C homozygotes in cases in comparison with drug-treated controls and healthy community controls. A borderline significant difference in the frequency of A homozygotes for rs3814057 in the cases in comparison with the drug-exposed controls was seen, but there was no significant difference when cases were compared with the community controls. The community control genotype frequencies are comparable to those reported previously for this SNP in an Irish population,24 but the drug-exposed control frequencies show a lower frequency for the homozygous A genotype. The reason for this discrepancy remains unclear.

The wild-type C variant at position -25385 in PXR has previously been associated with lower induced CYP3A expression in comparison with homozygotes for the T variant by in vivo studies using the erythromycin breath test as a measure of CYP3A activity.25 This lower expression may result in a higher accumulation of unmetabolized drug within the hepatocyte, especially if the drug is also a PXR inducer. In addition, the -25385 SNP (rs3810455) and other SNPs in complete linkage disequilibrium with it, including rs1523130, rs3842689, and rs11421631, lie within binding sites for a number of transcription factors, including signal transducer and activator of transcription 1 (STAT1), STAT3, STAT6, nuclear factor of activated T cells, and hepatocyte nuclear factor-1, and within a stress response element.21 Some of these sites are either created or removed when the variant allele is present. Our reporter gene data suggest that C homozygotes would have significantly less basal PXR expression, and this would potentially affect CYP3A induction and flucloxacillin disposition. In line with the findings from Zhang et al.,25 it is possible that, after exposure to flucloxacillin, any difference in the expression of CYP3A4 and other PXR-regulated genes between C homozygotes and T carriers would be accentuated. The large number of genes under the transcriptional regulation of PXR means that a small change in functional capacity seen in vitro might be magnified more in vivo and affect downstream processes, including those limiting bile acid accumulation within the hepatocyte. Susceptibility to flucloxacillin DILI can therefore be determined in part by PXR genotype, with homozygotes for C-25385 being more likely to develop this toxicity. Published pharmacokinetic data on flucloxacillin, especially with respect to disposition after administration for several days, are limited, but it is possible that interindividual variation in the induction of both CYP3A4 and ABC transporters may result in higher drug levels in some individuals than others, especially after prolonged use. Approximately 10% of flucloxacillin is converted to the hydroxymethyl metabolite at standard doses, with penicilloic acid metabolites also formed.33 The main route of excretion of the parent drug and the metabolites is renal, but a small percentage is believed to undergo biliary excretion.34 Excretion of either the parent drug or the metabolite via this route could be particularly relevant because of the injury to bile ducts often seen in this form of DILI.6 If the metabolite is reactive with protein, those with high PXR expression might form more adducts because of greater CYP3A4 induction resulting in a higher metabolite concentration. The opposite seems to occur here, and this suggests that the parent drug may also be important in the toxicity mechanism.

We have reported previously that possession of the HLA-B*5701 allele is associated with a highly significant risk for the development of flucloxacillin DILI.11 A similar association is well established for a generalized hypersensitivity reaction to the antiretroviral agent abacavir, although this reaction has a higher incidence in individuals carrying the at-risk genotype than flucloxacillin DILI, which occurs in approximately 1 in 500 to 1 in 1000 B*5701 carriers exposed to the drug. Individuals positive for the rs3814055 CC genotype in addition to HLA-B*5701 may be at further increased risk of DILI development, with a possible disease incidence of 1 of every 150 individuals with this combined genotype according to the present findings. Flucloxacillin DILI may require an additional event promoting drug accumulation within the liver, such as the apparently lower PXR expression associated with the rs3814055 CC genotype. Our analysis, stratified for the HLA-B*5701 genotype, indicates that the PXR genotype may indeed have an added effect on flucloxacillin-induced liver injury. However, this finding needs to be confirmed in a separate population.

Because the symptoms of flucloxacillin DILI typically occur 20 to 30 days after the initial drug intake, which is usually well after completion of the prescribed course, it seems unlikely that the major contributor to toxicity involves effects on bile acid metabolism or transport mediated by PXR. Instead, the lower PXR expression seen with the C form may result in higher levels of the drug within the liver favoring adduct formation, which would lead to an immune response. The formation of flucloxacillin adducts has been demonstrated to occur in rat livers,35 and adducts have also been detected recently in human serum.36


The authors thank Julia Graham and Julian Leathart for technical support.