¹F Hoffmann-La Roche Ltd, Basel, Switzerland

²Roche Molecular Systems Inc., Alameda, CA, USA

³Laboratory of Statistical Genetics, The Rockefeller University, NY, USA

Correspondence to: Gonzalo Acuña, F Hoffmann-LaRoche Ltd, Ch-4070 Basel, Switzerland, E-mail: gonzalo.acuna@roche.com

^aThe first three authors contributed equally to this work.

A retrospective pharmacogenetic study was conducted to identify possible genetic susceptibility factors in patients in whom the administration of the anti-Parkinson drug, tolcapone (TASMAR^Ò), was associated with hepatic toxicity. We studied 135 cases of patients with elevated liver transaminase levels (ELT) of 1.5 times above the upper limit of normal, in comparison with matched controls that had also received the drug but had not experienced ELT. DNA samples were genotyped for 30 previously described or newly characterized bi-allelic single nucleotide polymorphisms (SNPs), representing 12 candidate genes selected based on the known metabolic pathways involved in the tolcapone elimination. SNPs located within the UDP-glucuronosyl transferase 1A gene complex, which codes for the enzymes involved in the main elimination pathway of the drug, were found to be significantly associated with the occurrence of tolcapone-associated ELTs.

The Pharmacogenomics Journal (2002) 2, 327-334. doi:10.1038/sj.tpj.6500123

pharmacogenetics; tolcapone; hepatotoxicity; UDP-glucuronosyl transferase 1; UGT1A

Introduction

Tolcapone is a selective and potent inhibitor of catechol-O-methyltransferase (COMT), and is used to improve the bioavailability of L-dopa in the treatment of Parkinson's disease. A number of significant tolcapone-related liver injuries were reported during the first year after introduction of the drug,^1,2,3 resulting in a marketing suspension in the European Union and stricter patient monitoring in North America and the rest of the world. Tolcapone-induced liver injuries are rare events: during the initial clinical trials none of the 3848 patients treated experienced clinical signs or symptoms indicative of serious drug-induced liver injury, but 1-3% of the patients did experience asymptomatic transient liver enzyme elevations. Animal toxicology studies failed to provide any insight into the pathomechanism of liver injury.^4,5 In an attempt to shed additional light on this adverse event, we embarked on a retrospective clinical study to examine the potential role of genetic variants as predisposing factors to tolcapone-induced liver injury.

The main metabolic routes for tolcapone metabolism⁶ are summarized in Figure 1, and genetic variations in these enzymes were analyzed with regard to their prevalence in ELT cases and controls, respectively. In addition, we investigated whether genetic variations in enzymes involved in oxidative stress response, mitochondrial uncoupling and glutathione-S-transfer (Table 1) might show associations with tolcapone-induced liver toxicity.

Results

The study was designed to include patients who had received tolcapone in previous clinical trials. The cases in this study (n=135) were Parkinson's patients who, while being treated with tolcapone, demonstrated ELT of 1.5´ the upper limit of normal (ULN) or more. The controls (n=274) were Parkinson's patients who showed no signs of ELT while treated with tolcapone. Patient data were categorized by case-control status and by genotype at any of the markers tested. First-pass analysis revealed statistically significant associations between prevalence of ELT and homozygosity for the G allele at marker UGT1A6-A528G (OR=2.76, 95% CI=1.5-5.06; P<0.001) as well as nine other markers in the UDP-glucuronosyltransferase gene complex, all of which are in significant linkage disequilibrium with UGT1A6-A528G (see Table 1). Stratification of the sample according to sex revealed that the observed association between ELT and UGT1A6-A528G genotype was limited exclusively to male patients (OR 5.18; 95% CI=2.37-11.34; P<0.0001) and was not observed in females (OR=1.08, 95% CI=0.38-3.00; NS).

We determined pairwise strength of association between SNPs in the UGT1A gene. Figure 2 shows results for all pairs of the 13 SNPs in this gene for the case population. An analogous figure for the control population is not presented here because many SNPs were monomorphic. Clearly, except for SNPs located near the 5' or 3' ends of the gene (UGT1A8-C245A, UGT1A-G473T and 3'UTR-C908G), all were in complete disequilibrium with one another. Haplotypes for the 13 SNPs in the UGT1A gene were determined by Clark's algorithm,⁷ which does not require the assumption of Hardy-Weinberg equilibrium. Based on the genotypic results, a total of 44 haplotypes were inferred from 818 chromosomes and were assigned to each individual. Fifteen of the 44 haplotypes had frequency counts 5 in both case and control groups, accounting for approximately 93% of the total (Table 2). In these 15 haplotypes, two single sites (UGT1A8-C245A and UGT1A-G473T) are monomorphic. For one region, three consecutive SNPs (UGT1A7-G551T, UGT1A7-A555C and UGT1A7-A556G) are in complete linkage disequilibrium, that is, only two haplotypes (G-A-A and T-C-G) occur. Association between these 15 haplotypes and case-control status was statistically highly significant (P =0.0027). Haplotype 1, cAGAACgGGGCgC, and haplotype 4, cCTCGTgTAAAgG, render the highest statistical effect on differentiating cases from controls¾ removing either one of them decreases the statistical significance (the P-value increases about 10-fold), while removing other haplotypes had only slight effects (Table 2). For the two haplotypes, three sites were monomorphic (same lowercase nucleotides). At the remaining sites the two haplotypes show different alleles (uppercase nucleotides). Haplotype 1 is associated with disease (more frequent in cases than expected) while haplotype 4 is more common in control individuals than expected by chance.

To analyze the data further, principal components analysis was carried out among the 10 SNPs that showed P-values of 0.05 or less (Table 1), resulting in the identification of two principal components, P1 and P2, which accounted for 97% of the genotypic variation among the 10 markers (for case and control individuals combined). Multivariate regression analysis for P1, P2, and sex re-vealed that these three variables account for 21% of ELT variance (P<0.0001). The power of the marker geno-types was sufficient to allow accurate prediction in 61% of ELT status.

Discussion

Given that glucuronidation is the main pathway for tolcapone elimination (Figure 1), our findings are consistent with the hypothesis that impaired elimination of tolcapone may contribute causally to the observed liver toxicity. This interpretation is further supported by the finding that the UGT1A6 Ala181/Ser184 variant¾which arises from the UGT1A6 G754/C765 haplotype¾has previously been found to show reduced enzymatic activity in vitro compared with the wild type⁸ and might thus be associated with impaired clearance of the drug. This concurs with the findings from the current analysis, in which the presence of the Ala181 and Ser184 variants was associated with incrementally higher risks for ELT. For the polymorphism located in the 3'UTR of UGT1A the less frequent allele appears to have a protective effect, as it is more frequent in control patients than in patients with ELT. Since the UGT1A genes potentially share the 3' regulatory region, this polymorphism may affect the expression of several or all UGT1A genes⁹ involved in the metabolism of tolcapone (Figure 3). Alternatively, this polymorphism may be in linkage disequilibrium with another polymorphism that affects either the structure of the UGT1A proteins or the expression of the gene. The observation that the UGT1A polymorphisms are only significantly associated with ELTs in male but not in female patients could result from the lower number of females in the study, or it may suggest that the mechanism leading to tolcapone-induced ELT is different in males and females. Finally, no significant association was found between ELT and any of the other genetic variants tested (Table 1).

ELT is a well-established, yet indirect, indicator of liver damage. The relatively low cut-off value of 1.5 times above the ULN was selected to increase sensitivity of the study. As a consequence, the study population is perhaps more likely to include patients whose ELT was the product of factors other than tolcapone than had been the case with a higher cut-off value.

The relatively modest odds ratios found for association between ELT and individual marker genotypes are consistent with the concept that drug-induced liver toxicity represents a complex, multifactorial phenomenon. Thus, in addition to heritable predisposition (and the role of sex) presented here, additional factors¾such as additional genes not considered here, and/or extrinsic influences (such as co-medication, nutrition, etc) as well as complex interactions among genetic and environmental factors¾are all likely to contribute to the ELT phenotype. The limited positive predictive power of the risk genotype identified makes it an unlikely candidate for a clinically useful diagnostic tool for predicting patients at risk of tolcapone-induced liver toxicity. Nevertheless, the identification of a susceptibility gene/haplotype represents a first step towards the identification of a mechanism contributing or causing toxicity in humans. We believe that this study, therefore, represents an illustrative and pragmatic real-world example for what we may commonly expect to find in similar situations, with regard to the promise, but also the limitations of pharmaco- and toxicogenetic approaches.

Methods

Patient Selection

Initially, 645 patients who had received tolcapone in previous clinical trials were considered for inclusion in this retrospective genetic analysis. This included 215 patients (cases) who had developed ELT, and 430 patients matched by age (within ±10 years), and study protocol (which provided matching for disease severity). Of the 215 ELT patients, 135 agreed to participate in the study; another 31 patients did not participate because local Institutional Review Boards failed to approve the protocol (mainly due to local regulation or sensitivities concerning genetic research); and 49 patients were either lost to follow-up or deceased. Accordingly, 274 of the controls were retained to provide a 1 : 2 ratio. All cases and controls were of European origin and included a total of 261 males (74 cases, 187 controls) and 148 females (61 cases, 87 controls). Participation in this study was voluntary, and the patients' data were anonymized. All participating patients signed an appropriate informed consent before donating 9 ml of whole blood for DNA analysis.

ELT was classified as at least one measurement above 1.5 times the ULN of the investigator's range for aspartate aminotransferase or alanine aminotransferase. The ULN range was used instead of the absolute values to avoid the variation resulting from the different measurement procedures applied by each investigator in the different centers.

SNP Discovery in the UGT1 Gene Complex

All exons of the UGT1 locus (Figure 3) were sequenced in DNA samples derived from 47 non-related individuals of different ethnic origin (obtained from Coriell Institute of Medical Research in Camden, NJ, USA). DNA sequencing and SNP identification were carried out as described.¹⁰ SNPs thus newly identified are annotated by an asterisk in Table 1.

Genotyping

DNA was extracted from 400 l whole blood using a silica gel-based extraction method (QiaAmp DNA Blood Kit, Valencia, CA, USA). Samples were genotyped using either automated fluorescent dideoxytermination-based DNA sequencing or the kinetic thermocycling method (KTC).¹¹ The primers used for PCR amplification and sequencing are listed in Table 3.

In the KTC format, the generation of double-stranded amplification product is monitored using a DNA intercalating dye and a thermal cycler which has a fluorescence-detecting CCD camera attached (PE-Biosystems GeneAmp 5700 Sequence Detection System). Fluorescence in each well of the PCR amplification plate is measured at each cycle of annealing and denaturation. The cycle at which the relative fluorescence reached a threshold of 0.5 using the SDS software from PE-Biosystems was defined as C_t. The amplification reactions were designed to be allele-specific, so that the amplification reaction was positive if polymorphism was present and negative if polymorphism was absent. For each bi-allelic polymorphism, one well of the amplification plate was set up to be specific for allele 1 and a second well was set up to be specific for allele 2. For each polymorphism to be detected, three primers were designed¾two allele-specific primers and one common primer (Table 4). The amplification conditions were as follows: 10 mM Tris pH 8.0, 40 mM KCl, 2 mM MgCl₂, 50 m each of dATP, dCTP and dGTP, 25 m of TTP and 75 m of dUTP, 4% DMSO, 0.2X SyBr Green (Molecular Probes, Eugene, OR, USA), 2% glycerol, uracil N-glycosylase (UNG, 2 units), Stoffel Gold DNA polymerase (15 units, obtained from D. Birch, Core Research Department, RMS)¹² and primers (0.4 M) in an 85 l volume for each well. Thirty nanograms of DNA in a 15 l volume was then added to each well. To reduce the possibility of contamination by pre-existing amplification product, the assay procedure included the incorporation of dUTP into the amplification product and an incubation step with uracyl N-glycosylase (UNG) for degradation of pre-existing uracyl-containing products.¹³ The thermal cycling conditions were as follows: 5 min at 50°C for UNG degradation of any previously contaminating PCR products, 12 min at 95°C for Stoffel Gold polymerase activation, 55 cycles of denaturation at 95°C and annealing at the annealing temperature indicated in Table 4, followed by a dissociation step of 1 min at 1° increments from 60°C to 95°C. The C_t of each amplification reaction was determined and the difference between the C_t for allele 1 and allele 2 (C_t) was used as the assay result. Samples with C_t's between -3.0 and 3.0 were considered heterozygous, and samples with C_t's below -3.0 or above 3.0 were considered homo-zygous for allele 1 or 2, respectively.

Statistical Analysis

Patient samples were stratified by case-control status and by genotype at any of the markers tested. Data were analyzed by the logistic regression analysis using recessive and dominant models. Odd ratios and 95% confidence intervals that excluded unity were regarded as statistically significant. This analysis was subsequently repeated after stratification by sex.

Multivariate Analysis and Interaction Modeling

A two-stage approach,¹⁴ with marker selection followed by a modeling stage, was employed, by first selecting all markers that were associated with case-control status at P<0.05 in the chi-square test for comparing geno-type frequencies between cases and controls (Table 1). Of these, the best linear combinations were obtained by principal components analysis and used, along with sex, and predictive variables for multivariate logistic regression analysis. The estimated logistic regression equation reads as follows: log(odds) = 0.46(P1)+1.36(P2)+0.40(sex)-0.29 (P1´S)-0.71(P2´S), where P1 and P2 are the main components, and S equals 1 for male and 2 for female (ie, the log(odds) is predicted differently for males and females). The corresponding sex-specific equations are

log(odds)=0.17(P1)+0.65(P2)+0.40 for males

log(odds)=-0.12(P1)-0.06(P2)+0.80 for females.

Acknowledgements

We would like to thank all the patients that participated in this study and acknowledge the contributions of the study investigators who contacted the patients and collected the blood samples. These data were submitted to the European drug regulatory authorities (CPMP) on 15 July 2000. We would also like to thank the technical staff (T Burchill, J Mangaccat, R Olmos and Y Kidane) who ran the genotyping assays. Partial support through grant HG00008 (to JO) from the National Human Genome Research Institute is gratefully acknowledged.

DUALITY OF INTEREST

1 Olanow CW. Tolcapone and hepatotoxic effects. Arch Neurol 2000; 57: 263-267.

2 Spahr L, Rubbia-Brandt L, Burkhard PR, Assal F, Hadengue A. Tolcapone-related fulminant hepatitis. Dig Dis Sci 2000; 45: 1881-1884.

3 Assal F, Spahr L, Hadengue A, Rubbici-Brandt L, Burkhard PR. Tolcapone and fulminant hepatitis. Lancet 1998; 352: 958.

4 Schlappi B, Morimoto H, Kobayashi K, Horii I. Six-month oral toxicity study with tolcapone on beagle dogs. Jpn Pharmacol Ther 1996; 24: ((Suppl)) 77-102.

5 Schlappi B, Jovanovic D, Okada M, Kobayashi K, Horii I. Six-month oral toxicity study of tolcapone in rat (feed admix). Jpn Pharmacol Ther 1996; 24: ((Suppl)) 49-75.

6 Jorga K, Fotteler B, Heizmann P, Gasser R. Metabolism and excretion of tolcapone, a novel inhibitor of catechol-O-methyltransferase. Br J Clin Pharmacol 1999; 4: 513-520.

7 Clark AG. Inference of haplotypes from PCR-amplified samples of diploid populations. Mol Biol Evol 1990; 111: 22.

8 Ciotti M, Marrone A, Potter C, Owens IS. Genetic polymorphisms in the human UGT1A6 (planar phenol) UDP-glucuronosyl-transferase: pharmacological implications. Pharmacogenetics 1997; 7: 485-495.

9 Di Paola R, Frittitta L, Miscio G, Bozzali M, Baratta R, Centra M et al. A variation in 3'UTR of hPTP1B increases specific gene expression and associates with insulin resistance. Am J Hum Genet 2002; 70: 806-812. Article MEDLINE

10 Lehnert V, Holzwarth J, Ott M, Thompson A, Demmak S, Foernzler D. A semi-automated system for analysis and storage of SNPs. Hum Mutat 2001; 17: 243-254.

11 Germer S, Higuchi R. Single-tube genotyping without oligonucleotide probes. Genome Res 1999; 9: 72-78. MEDLINE

12 Birch DE. Simplified hot start PCR. Nature 1996; 6581: 445-446.

13 Longo MC, Berninger MS, Hartley JL. Use of uracil DNA glycosylase to control carry-over contamination in polymerase chain reactions. Gene 1990; 93: 125-128. MEDLINE

14 Hoh J, Wille A, Zee R, Cheng S, Reynolds R, Lindpaintner K et al. Selecting SNPs in two-stage analysis of disease association data: a model-free approach. Ann Hum Genet 2000; 64: 413-417. Article MEDLINE

Figure 1 Primary metabolic routes of tolcapone metabolism (ST= sulfotransferase; NAT=N-acetyltransferase; COMT=catechol-O-methyl transferase; UGT=UDP-glucuronosyltransferase; CYP3A4=cytochrome P450 3A4).

Figure 2 Representation of the pairwise association strength between the 13 SNPs in the UGT1A gene of the case population. The color gradient from blue to red represents strong (blue) to weak (red) association.

Figure 3 The UGT1A gene consists of at least 9 promoters and first exons (indicated as black boxes; first exons 3, 11 and 12 are pseudogenes) which can be spliced with 4 common exons (gray boxes, indicated by roman numerals) to result in 9 different UGT1A enzymes. The relative position of the SNPs tested in this gene is indicated. The nucleotide position of the SNPs corresponds to their position within the indicated exon sequence (see Table 1 for the sequence annotation number).

Table 1 Summary of the genetic markers analyzed

Table 2 Haplotype frequencies

Table 3 PCR and sequencing primers for genotyping the UGT1A gene

Table 4 Primers used to genotype markers by kinetic thermocycling