5′-Flanking region polymorphisms of CYP2C9 and their relationship to S-warfarin metabolism in white and Japanese patients

Population differences in the drug-metabolizing enzymes are important in bridging therapeutic doses and safety profiles of drugs from one population to another. Warfarin is a widely used oral anticoagulant and its effect is largely attributable to the pharmacologically more active S-enantiomer.¹  Recently, we observed that Japanese patients receiving warfarin therapy had a significantly greater body weight–normalized plasma unbound clearance (CLpo,u) of S-warfarin than white patients,²  which is predominantly reflective of CYP2C9-mediated hepatic metabolism.³  At present, 11 coding region variant alleles of the gene have been reported.⁴  Because whites have greater allelic frequencies than Southeast Asians⁵  of the 2 most common functionally defective variants (CYP2C9^*2 and CYP2C9^*3), it is possible that the population differences in S-warfarin metabolism may be attributed to this distribution difference. However, in our previous study a difference in S-warfarin metabolism was noted even when the 2 populations were matched with respect to the homozygous wild-type CYP2C9 genotype (CYP2C9^*1/^*1).²  This indicates that such a coding region polymorphism cannot fully account for the population differences in the CYP2C9 activity. Recently, Shintani et al⁶  reported 7 single nucleotide polymorphisms (SNPs) in the 5′-flanking region of CYP2C9 in Japanese subjects, some of which were associated with an altered level of gene transcription. Accordingly, we investigated SNPs within the -2.1-kb promoter region of CYP2C9 in white and Japanese patients receiving warfarin and assessed their contribution to the variability of in vivo CYP2C9 activity within and between the 2 populations.

The promoter region analysis was performed in the participants of an earlier investigation,²  22 white (9 men and 13 women) and 38 Japanese patients (25 men and 13 women), whose DNA samples were available. Details of the study have been reported.²  Briefly, the subjects were recruited at Vanderbilt University Hospital (Nashville, TN), and the International Medical Center of Japan. All patients received a constant oral dose of racemic warfarin once daily for at least 1 month before blood sampling. None had impaired hepatic function. Informed consent was obtained from each patient and the study protocol was approved by the institutional review boards at both institutions. Blood samples were obtained from patients at approximately 16 hours after oral administration of the last dose of warfarin. These samples were analyzed for the separate R- and S-enantiomers of warfarin, along with the extent of their plasma binding, resulting in estimates of the plasma unbound concentration (Cu) and unbound clearance (CLpo,u) of S-warfarin.^7-9 

Variants in the 5′-flanking region of CYP2C9 up to -2137 bp were analyzed according to the method of Shintani et al⁶  using DNA previously collected from the patients.²  Coding region SNPs (CYP2C9^*2, CYP2C9^*3, CYP2C9^*4, CYP2C9^*5, and CYP2C9^*6) were analyzed as previously described.² 

Population differences in the mean values were compared using an unpaired Student t test. Multiple comparisons for the mean CLpo,u for S-warfarin and other parameters obtained from different genotypes within each population were performed by analysis of variance followed by the Tukey-Kremer test. A P value of less than .05 was considered statistically significant.

There was no difference in the mean age of the white and Japanese populations (59 years vs 60 years). While there were significant (P < .01) differences in the mean body weight (84 kg vs 57 kg) and oral dosage of warfarin (16.0 μmol/d [5.3 mg/d as warfarin-Na] vs 9.2 μmol/d [3.2 mg/d as warfarin-K]), the mean body weight–normalized daily warfarin dosages (0.18 μmol/d/kg [0.060 mg/d/kg as warfarin-Na] vs 0.16 μmol/d/kg [0.056 mg/d/kg as warfarin-K]) were comparable between the populations. The mean body weight–normalized CLpo,u for S-warfarin in Japanese patients was approximately 2-fold greater (P < .01) than that in white patients (10.5 mL/min/kg vs 4.8 mL/min/kg). There were significant (P < .01) differences in the mean Cu of S-warfarin and international normalized ratio (INR) between white and Japanese patients: 15.6 nM (4.8 ng/mL) versus 6.5 nM (2.0 ng/mL) and 2.4 versus 1.6, respectively. These data are consistent with the recent studies demonstrating that Southeast Asians (Japanese and Chinese) may require lower INR values (eg, 1.5-2.5) for anticoagulation than whites.^10,11 

We found 13 SNPs, 1 insertion, and 2 deletions within the -2.1-kb 5′-flanking region of CYP2C9 (Figure 1). Because all white and Japanese participants possessed the sequence -2080T, -2079A, -2078G, Ins T-839, Del T-828, and Del A-820, as compared with the reference sequence (GenBank accession no. L16877), we considered the above sequence to be the “wild-type” in our populations (ie, pattern 1). Because some of the SNPs occurred together, they were classified into 9 different patterns (ie, patterns 1, 2, 3, 4, 5-1, 5-2, 6-1, 6-2, and 7; Figure 1). Patterns 6-1 and 6-2 were in complete linkage disequilibrium either with CYP2C9^*1/^*3 or CYP2C9^*3/^*3 in both Japanese and white patients. Similarly, patterns 5-1 and 5-2 were linked either with CYP2C9^*1/^*2 or CYP2C9^*2/^*2 in white patients. A white patient possessing CYP2C9^*2/^*3 had all 8 SNPs for the patterns 5-1 and 6-1, and another such patient with pattern 2 was found to have CYP2C9^*11. Patterns 5-1, 5-2, and 7 were found only in white patients. Our white patients appeared to have CYP2C9 variants in a greater frequency than those previously reported.⁵  However, further studies with a large number of subjects will be required for comparing population differences in the frequencies of haplotype (combinations of the 5′-flanking and coding region SNPs) of CYP2C9.

Previous studies have shown putative binding sites for transcription enhancers in 5′-flanking region of CYP2C9 (eg, TATA box, CAAT box, HepG2-specific factor-1 [HPF-1], C/EBP, AP-1, glucocorticoid response elements [GREs], Barbie box, hepatic nuclear factor-1 [HNF-1], and DR4 motif).^12-15  Some of the promoter region SNPs are located at or around these transcription enhancer responsive elements (Figure 1), indicating that they may contribute to within and between population variability in in vivo CYP2C9 activity.

Figure 2 shows doses of warfarin, INR values, and body weight–normalized CLpo,u for S-warfarin in the white and Japanese patients possessing the various patterns of SNPs in the CYP2C9 promoter region. Japanese patients with pattern 6-1 had significantly (P < .01) reduced CLpo,u for S-warfarin as compared with those with pattern 1; the warfarin doses in these patients were also lower. This finding is in agreement with that of Shintani et al⁶  based on studies in Japanese patients with another CYP2C9 substrate, phenytoin. However, because pattern 6-1 in the promoter region is in complete linkage disequilibrium with the functionally defective allele CYP2C9^*3, it cannot be excluded that this observation might instead be attributable to this coding region variant rather than the promoter pattern 6-1. Because the promoter SNP patterns of 5-1 and 5-2 were completely linked with CYP2C9^*2 in the white patients, these promoter SNPs would have only a limited effect on in vivo CYP2C9 activity as shown for CYP2C9^*2 previously.⁵ 

When the population differences in the CYP2C9 activity were compared between those having the wild-type alleles in both promoter and coding regions of CYP2C9 gene (Figure 2, pattern 1), the Japanese patients had significantly (P < .01) greater CLpo,u for S-warfarin (11.8 mL/min/kg ± 3.0 mL/min/kg) than the white patients (5.0 mL/min/kg ± 1.9 mL/min/kg). These data strongly indicate that the population differences in the in vivo CYP2C9 activity cannot be accounted for either the 13 SNPs in the promoter region or the 6 SNPs in the coding region. However, we cannot exclude the possibility that unidentified SNPs located more upstream in the 5′-flanking region than we currently investigated (eg, constitutive androstane receptor–responsive elements around -2.9 kb)¹⁶  might account for the in vivo findings. Further studies are, therefore, required to identify other possible factors (eg, diet and environmental factors) contributing to the population difference in the CYP2C9 activity.

The authors thank Dr Joyce A Goldstein, National Institutes of Environmental Health Sciences, Research Triangle Park, NC, for the generous gift of the genomic DNA for CYP2C9^*6 and Miss Mamiko Shimoyama, Miss Keiko Miyazaki, Miss Masumi Matsumoto, and Miss Eri Yamamoto for their excellent technical assistance.