Dr Brockmöller Institute of Clinical Pharmacology, University Clinic Charité, Humboldt University, Schumannstrasse 20/21 D-10098 Berlin, Germany. Tel.: +49 30 28025318 Fax: +49 30 28025153 E-mail: firstname.lastname@example.org
Aims The genetically polymorphic cytochrome P450 enzyme CYP2C9 metabolizes many important drugs. We studied the frequency of the amino acid variants cysteine144 (CYP2C9*2 ) and leucine359 (CYP2C9*3 ) in a Turkish population and the correlation between genotype and phenotype using phenytoin as probe drug.
Methods CYP2C9 alleles *2 and *3 were measured in 499 unrelated Turkish subjects by PCR and restriction fragment length pattern analysis. Phenotyping was performed in a subgroup of 101 volunteers with a single oral dose of 300 mg phenytoin and concentration analysis in serum drawn 12 h after dosage.
Results CYP2C9 allele frequencies in 499 unrelated Turkish subjects were 0.794 for CYP2C9*1, 0.106 for CYP2C9*2 and 0.100 for CYP2C9*3. Mean phenytoin serum concentrations at 12 h after dosage were 4.16 mg l−1 (95% CI 3.86–4.46) in carriers of the genotype CYP2C9*1/1, 5.52 mg l−1 (4.66–6.39) in CYP2C9*1/2, and 5.65 mg l−1 (4.86–6.43) in CYP2C9*1/3. These differences were significant and accounted for 31% of total variability in phenytoin trough levels. Mean 12 h concentration ratios of 5-(para-hydroxyphenyl)-5-phenylhydantoin/phenytoin (p-HPPH/P) were 0.43 (0.39–0.47) for CYP2C9*1/1 compared with 0.26 (0.21–0.31) for CYP2C9*1/2, 0.14 (0.13–0.14) for CYP2C9*2/2, 0.21 (0.18–0.24) for CYP2C9*1/3, and 0.02 for CYP2C9*3/3; all mutant genotypes were significantly different compared with CYP2C9*1/1.
Conclusions Frequency of the two CYP2C9 variants in Turkish subjects was in a similar range as in other Caucasian populations. A significant proportion of the interindividual variability in phenytoin trough levels is explained by the genotypes. The 12 h serum concentrations after a single phenytoin dose may be used for routine phenotyping of CYP2C9 mediated metabolic clearance and the p-HPPH/P ratios may be even more sensitive indicators of CYP2C9 activity.
The cytochrome P450 2C9 (CYP2C9) is involved in hydroxylation of amitriptyline, fluoxetine, losartan, phenytoin, S-warfarin, tolbutamide, and many nonsteroidal antirheumatics including meloxicam [ 1, 2]. Recent studies have shown that there are at least two amino acid variants in CYP2C9 with functional relevance for metabolism and pharmacokinetics of phenytoin [ 3, 4], warfarin [ 5–8], and tolbutamide [ 9–11]. Apparently, these amino acid variants may not affect CYP2C9 mediated metabolism of some nonsteroidals like diclofenac to the same extent [ 1].
Many studies have also shown considerable interethnic differences in drug metabolism [ 12]. For instance, the proportion of carriers of gene duplications of CYP2D6 varies even within the European population from 1 to 7% from the north to the south and we have recently found that the proportion of these gene duplications in the Turkish population is about 4% (Aynacioglu et al. unpublished data). Therefore, we anticipated peculiarities also with respect to CYP2C9 and we studied a sufficiently large sample of the Turkish population on frequency of the CYP2C9 wild-type allele (termed CYP2C9*1 ), as well as of the allelic variants cysteine144 (CYP2C9*2 ) and leucine359 (CYP2C9*3 ).
Although phenotyping with CYP2C9 probes was included early in drug research [ 13], phenotyping for CYP2C9 is not as frequently performed compared with phenotyping for CYP2D6 or CYP2C19. A Medline search on phenotyping and CYP2C9 gave just 10 hits, compared with 238 hits for CYP2D6 and phenotyping and 66 hits for CYP2C19 and phenotyping. This indicates that the importance of CYP2C9 in drug metabolism and the relevance of genetic polymorphism in CYP2C9 is not yet widely recognized [ 1]. Probably CYP2C9 is the major metabolizing enzyme of more clinically relevant drugs than CYP2C19 [ 1] and contributes to about 20% of total hepatic cytochrome P450 protein [ 9]. Many CYP2C9 metabolized drugs may be used for phenotyping; tolbutamide has already been used for this purpose [ 14]. Choice of phenytoin as a CYP2C9 probe has the advantage that the concentration analysis methods are available in most clinical pharmacology laboratories. A major proportion of phenytoin (60–80%) is metabolized via CYP2C9 to 5-(p-hydroxyphenyl)-5-phenylhydantoin [ 15]. Therefore, we used phenytoin phenotyping in this population study to investigate the functional relevance of the CYP2C9 mutations in a Turkish population.
A group of 499 (246 female, 253 male) unrelated subjects originating from Gaziantep in South-East Anatolia, Turkey, was studied after informed written consent. The study was approved by the local ethics committee at the Gaziantep University. The sample includes 280 outpatients with various trivial diagnoses and 218 healthy volunteers. 5–10 ml venous blood samples were drawn with EDTA as anticoagulant for molecular genetic characterization of the CYP2C9 genotypes.
Phenotyping with phenytoin
A subgroup of 101 healthy volunteers was phenotyped with phenytoin. These subjects were healthy according to their medical history. Phenotyping was done overnight in order to reduce discomfort due to sedation. After at least 4 h of fasting, each subject took a 300 mg phenytoin tablet with tap water at around 23.00 h and a blood sample was drawn 12 h later. Serum was kept frozen at −20° C until h.p.l.c. analysis.
Phenytoin and 5-(p-hydroxyphenyl)-5-phenylhydantoin (p-HPPH) reference substance was from Aldrich, Steinheim, Germany. Serum samples (0.1 ml) were treated for 30 min at 30° C with 0.1 ml β-glucuronidase (100 units) and extracted with 1.5 ml ethyl acetate. Hexobarbitone was used as internal standard. Phenytoin and p-HPPH were analysed by h.p.l.c. with u.v.-detection at 210 nm. The separation was performed on an octadecylsilane coated silica column (InertsilTM ODS-2, 5 μm particles, 250×4.6 mm) by isocratic elution with a mixture of 15% methanol, 18% acetonitrile, and 67% 20 mm sodium phosphate buffer pH 6.9. The flow rate was 1.1 ml min−1 at 50° C. The calibration curve was linear between 0.01 μg ml−1 and 40 μg ml−1. The limit of quantification was determined as 0.05 μg ml−1 for both substances. Other antiepileptic drugs like carbamazepine, the main metabolites of carbamazepine, phenobarbitone, primidone, ethosuximide, or valproic acid were chromatographically well separated from phenytoin and p-HPPH. Ratios of p-HPPH over phenytoin (p-HPPH/P ratios) were calculated on a molar basis. Assay precision and accuracy was tested with a blank serum sample spiked to 2.5 mg l−1 of phenytoin and 2.5 mg l−1 of p-HPPH. Within day coefficient of variation was 3.5% (mean: 2.51 mg l−1, true value: 2.50, n=12) for phenytoin and 3.6% (mean: 2.50 mg l−1; true value: 2.50; n=12) for p-HPPH as measured with the spiked sample. Between day coefficient of variation was 1.3% (mean: 2.61; true value: 2.50; n=10) for phenytoin and 2.5% (mean 2.75; true value: 2.50; n=10) for p-HPPH. A pooled serum sample from phenytoin treated patients served as an additional control. Between day coefficient of variation with this control was 6.9% (mean: 10.81 mg l−1; n=10) for phenytoin and 6.6% (mean: 3.22 mg l−1; n=10) for p-HPPH. This pooled serum sample was furthermore 1:5 diluted with blank serum giving a between day coefficient of variation of 9.6% (mean 2.19; n=10) for phenytoin and 6.0% (mean 0.67; n=10) for p-HPPH. All samples were measured with and without glucuronidase treatment, on average, only a fraction of 5% (range: 0–23%) of total p-HPPH in serum was unconjugated, while the phenytoin concentrations were not different when measured with vs without glucuronidase treatment.
Identification of CYP2C9 mutations
For CYP2C9 genotyping, 5–10 ml of blood were collected in EDTA tubes and DNA was extracted using a standard phenol/chloroform extraction. DNA samples were dissolved in 10 mm Tris/1 mm EDTA, pH 8.0 and stored at 4° C until PCR analyses. CYP2C9 exon 3 mutation C432T which is responsible for amino acid change Arg144Cys was detected by PCR-RFLP as described earlier [ 16] using primers CL1 (5′-CACTGG CTGAAAGAGCTAACAGAG) and CR1 (5′-GTGATA TGGAGTAGGGTCACCCAC) to amplify a 372-bp amplicon in a 50 μl PCR mix comprised of 10 mm Tris-HCl pH 8.3, 1.25 mm MgCl2, 50 mm KCl, 200 μm dNTPs, 0.2 μm of each of the primers, 2.5 U Taq polymerase (AmpliTaqTM, Perkin Elmer), and 1 μl of genomic DNA. PCR was performed with an initial denaturation for 2 min at 94° C followed by 35 cycles of 30 s at 94° C, 10 s at 60° C, 1 min at 72° C, and a terminal extension for 7 min at 72° C. For Arg144Cys detection, 20 μl of PCR product were digested overnight with restriction endonuclease Sau96I (New England Biolabs, Schwalbach, Germany), and analysed by 3% agarose gel electrophoresis. Wild type alleles (Arg) were cut into fragments of 179, 119 and 74 bp, whereas mutant alleles (Cys) showed fragments of 253 and 119 bp by loss of one restriction site. Exon 7 mutation A1077T which codes for the amino acid change Ile359Leu was detected by a PCR-RFLP assay using primers C5 (5′-AGGAAGAGATTGAACGTGTGA) and C6 (5′-GGCAGGCTGGTGGGGAGAAGGCCAA). PCR conditions were the same as described above. A 130-bp amplicon was digested with StyI (New England Biolabs); wild type alleles (Ile) remained uncut, but mutant alleles (Leu) were cleaved into two fragments of 104 and 26 bp. As a control, in each assay series a DNA sample known to be positive for the StyI restriction site (i.e. positive for the Leu variant) was included to confirm activity of the respective batch of StyI.
Population frequencies of alleles and allele combinations are given together with their 95% confidence limits. Significance of interethnic differences was calculated by the exact Fisher test. Significance of all functional differences attributable to the CYP2C9 polymorphism was calculated by analysis of variance after we confirmed that the frequency distributions of all subgroups did not differ significantly from normal distribution as calculated by the Kolmogorov-Smirnov test. In addition, nonparametric measures are given and significance of the trends related to the number of mutant genes was assessed by the nonparametric Jonckheere-Terpstra-test for trends calculated by the software SPSSTM for Windows (SPSS Inc, Chicago, USA). The program SYSTATTM (SPSS Inc, Chicago, USA) was used for the other statistical calculations.
Frequencies of the variant amino acids at positions 144 and 359 of CYP2C9 in the sample of 499 Turkish subjects is given in Table 1 and the CYP2C9 allele frequencies are presented in Table 2. These data correspond to an allele frequency of the mutant allele CYP2C9*2 of 0.106 (95% CI 0.085–0.130) and an allele frequency of the mutant allele CYP2C9*3 of 0.100 (95% CI 0.079–0.123). There were no differences in allele frequencies between hospital patients and healthy volunteers.
Table 1. Proportion of CYP2C9 amino acid variants in a Turkish population of 499 subjects.
Table 2. Frequency of CYP2C9 genotypes in a Turkish population of 499 subjects.
There was no evidence for a linkage of the Cys144 variant with the Leu359 variant on the same chromosome. In particular, in the large sample of 499 subjects there was no subject with the combination of Cys/Cys at amino acid 144 together with Ile/Leu at amino acid 359 or with Arg/Cys at position 144 together with Leu/Leu at position 359. Such combinations would have proven existence of both amino acid variants on the same chromosome, but were never found.
Trough levels taken 12 h after a 300 mg test dose of phenytoin were obtained in a subgroup of 101 healthy subjects. Mean 12 h serum concentration of phenytoin was 4.66 mg l−1 (range: 0.47–8.75 mg l−1 ) and mean 12 h serum concentration of p-HPPH was 1.65 mg l−1 (range: 0.09–3.52 mg l−1 ). Frequency distribution of the phenytoin 12 h concentrations is shown in Figure 1. The difference between the CYP2C9 genotype groups was significant (P<0.0001, F-test) and post hoc testing using Bonferroni adjustment showed significant differences between the wild-type CYP2C9*1/1 genotype and the following genotypes: CYP2C9*1/2 (difference, 95% CI of difference: 1.36, 0.46–2.26 mg l−1, P=0.009), CYP2C9*2/2 (2.42, 0.93–3.91, P=0.02), and CYP2C9*1/3 (1.49, 0.78–2.20, P=0.001). Activities within the genetically defined subgroups is described in Table 3. Significance of the differences due to genotype was also confirmed by the nonparametric Jonckheere-Terpsta-test which takes the trends with increasing number of active alleles into account. This test showed highly significant differences when comparing the groups with 0, 1, and 2 cysteine144 mutations (P<0.001) and when comparing the groups with 0, 1, and 2 leucine359 variants (P<0.001).
Table 3. Functional effects of the CYP2C9 allele variants on phenytoin trough levels taken 12 h after dosage (n=101).
Since only about 80% of phenytoin clearance was explained by CYP2C9 metabolism to p-HPPH and trough levels are also dependent on bioavailability and volume of distribution, we wanted to test whether the p-HPPH/P ratios might be even better indicators in a CYP2C9 phenotyping procedure. Frequency distribution of these p-HPPH/P ratios is shown in Figure 2 and the statistical parameters are given in Table 4. The difference between the groups was significant (P<0.0001, F-test), and post hoc testing using Bonferroni adjustment showed significant differences between the wild-type CYP2C9*1/1 genotype and the other detected genotypes as follows: CYP2C9*1/2 (difference, 95% CI of difference: 0.17, 0.10–0.23, P=0.002), CYP2C9*2/2 (0.29, 0.25–0.33, P=0.006), and CYP2C9*1/3 (0.22, 0.17–0.26, P<0.0001). Significance of the differences due to genotype was also confirmed by the nonparametric Jonckheere-Terpsta-test which takes the trends into account (P<0.0001).
Table 4. Functional effects of the CYP2C9 allele variants on the ratio of p-HPPH to phenytoin (n=101).
CYP2C9 allele frequencies
Frequencies of both amino acid polymorphisms in the Turkish studied here were in a comparable range but tended to be lower compared with some Caucasian populations. For the cysteine144 variant an allele frequency of 0.103 was found. Furuya et al. [ 5] found in 94 British patients an allele frequency 0.191 which was significantly higher (P=0.002; two-sided Fisher’s exact test). In a sample of 127 Caucasians from Berlin [ 16], we found an allele frequency of 0.13 which was not significantly different from the Turkish population.
The frequency of the Cys144-mutant was almost identical in our Turkish sample compared with the study of London et al. [ 17] who reported a proportion of 19.3% carriers of one or two CYP2C9*2 alleles in Caucasians from Los Angeles (heterozygous and homozygous carriers of CYP2C9*2 together were 20.2% in our sample). In addition, the study of London et al. [ 17] leaves no doubt that the proportion of the CYP2C9*2 allele is significantly lower in Africans with a proportion of 7.1% carriers of one or two CYP2C9*2 alleles corresponding to an allele frequency of 0.04. The Cys144-allele frequency was zero in a small study of 39 Japanese subjects [ 9], which was significantly lower than in our sample but further studies in larger populations are required. Thus, from other studies there appear to be major differences in the population frequency of CYP2C9*2 between Caucasians, Africans and Asians. Larger samples will have to be studied to confirm differences within Caucasians.
There are fewer data published on the population frequency of the CYP2C9*3 which had an allele frequency of 0.10 in the Turkish population. Inuou et al. reported 3 heterozygous carriers among 39 Japanese (CYP2C9*3 allele frequency 0.038) and 3 heterozygous carriers among 45 Caucasians (allele frequency 0.033), which is lower compared with our sample but power is too small to make a general conclusion. In a sample of 266 unrelated and unselected Germans from Berlin, the CYP2C9*3 allele frequency was 0.081 [ 18] which was not significantly different compared with the Turkish group.
Comparisons of the DNA sequences from various sources revealed a number of other potentially polymorphic amino acids, namely cysteine/tyrosine175, phenylalanine/leucine239, tyrosine/cysteine358, and glycine/aspartic acid417 [ 11]. Earlier screening of more than 200 subjects could however, not identify any of these alleles in the German population. Therefore we concluded that these mutations are either extremely rare or may even be artefacts and we did not study these mutations here. Routine genotyping with the PCR-RFLP methods described here could be performed within 48 h with high accuracy. It is a particular advantage of the assay for the arginine/cysteine144 polymorphism that the PCR amplicon has an additional nonpolymorphic Sau98I cleavage site which serves as an internal control for enzyme activity [ 16]. The assay for the isoleucine/ leucine359 polymorphism might be further improved by artificial introduction of a similar nonpolymorphic restriction site.
Functional effects of CYP2C9 variants
Several biochemical studies showed that the two CYP2C9 amino acid variants are functionally relevant [ 4, 19, 20], but clinically the correspondence between CYP2C9 genotype and phenotype has not been as extensively studied as in other enzymes like cytochrome P450 2D6. Phenotyping should optimally provide data on metabolic clearance mediated by the studied enzyme. Such clearances can precisely only be obtained by AUC measurements, but 12 h trough levels as used here correlate with AUC and use of phenytoin 12 h trough levels as indicators of exposure to this drug is well established in the context of therapeutic drug monitoring. Serum ratios of phenytoin over its 4′ hydroxylated metabolite have also earlier been used to confirm CYP2C9 deficiency in volunteers [ 21]. Theoretically, phenytoin might even be used as a probe for two enzyme systems, namely CYP2C9 and CYP2C19. CYP2C9 mainly catalyses formation of the S-enantiomer of p-HPPH whereas CYP2C19 generates (R) p-HPPH [ 22]. However, the serum concentrations of CYP2C9 formed (R) p-HPPH were only about 2% of the amount of CYP2C19 formed (S) p-HPPH in the study of Ieiri et al. [ 23]. Therefore, it is unlikely that the CYP2C19 polymorphism confounds phenotyping of CYP2C9 with a non-enantioselective methods as applied in our study.
Distribution of the phenytoin trough levels was not different from normal distribution (P=0.65 using the Kolmogorov-Smirnov test; Figure 1 and 2). The two genetic polymorphisms explained 31.4% of the variability in phenytoin trough levels according to the anova r2 value. Frequency distribution of the p-HPPH/P ratios was also not significantly different from normal distribution (P=0.66) but Figure 2 might suggest that there is a bimodal distribution with all heterozygously and homozygously mutated people on one side and the carriers of the CYP2C9*1 allele on the other side. 36.5% of the variability in the p-HPPH/P ratios was explained by genetic polymorphism. The persons with homozygous wild-type genotype had only about 30% lower levels of phenytoin compared with both frequent heterozygous genotypes ( Table 3). This may be relevant in treatment with substrates like phenytoin itself and warfarin. It corresponds with earlier detected differences in the mean warfarin dose [ 5]. The functional effects of the homozygously mutant genotypes were more pronounced, but sample size was too small to get means with narrow confidence limits in this sample ( Table 3 and 4). In accordance with earlier data, the Ile/Leu359 variant was the variant with the more pronounced reduction in activity.
Phenytoin as an in vivo probe drug for CYP2C9 activity
The test dose and time schedule applied in this study was derived from feasibility considerations, namely dosage at night to avoid any discomfort due to phenytoin induced sedation, and blood sampling at the usual working time 12 h after dosage. With this dosage and time schedule all concentrations were well above the limit of quantification of our analytical method. Later sampling times may be preferable if phenytoin concentations are used as indicators of CYP2C9 metabolic clearance but should not be used if the p-HPPH/phenytoin ratios are used since p-HPPH concentrations may be below the limit of quantification. A test dose of 300 mg was chosen to allow easy and accurate quantification by h.p.l.c. with u.v.-detection. We were aware that phenytoin metabolism might become saturated in a few subjects. A clinical study on dose-dependent metabolism of phenytoin in subjects stratified for the five CYP2C9 genotypes should be performed in the future. The method for analysis of phenytoin and p-HPPH was robust and precise as documented above. It should be noted that β-glucuronidase treatment is essential when the metabolic ratio of p-HPPH over phenytoin is formed, since on average only 5% of p-HPPH was found unconjugated in the samples of this study. Both, the phenytoin levels or the p-HPPH/phenytoin ratios may be used as in vivo parameters of CYP2C9 activity. The metabolic ratios appeared to be slightly more sensitive indicators of CYP2C9 metabolism, but it might be speculated that the metabolic ratios are also modified by the rate of p-HPPH glucuronidation, a problem which does not exist for the phenytoin 12 h trough levels.
The CYP2C9 amino acid variants Cys144 and Leu359 existed in Turkish people with a frequency comparable with other Caucasian populations. About 30% of the interindividual variability in phenytoin trough levels is explained by these genotypes. The serum 12 h concentrations after a single dose of 300 mg phenytoin appear to be useful as indicators of CYP2C9 metabolic clearance in routine phenotyping and the p-HPPH/P ratios appeared to be even more sensitive in vivo indicators of CYP2C9 activity. The long-known dose dependency of phenytoin metabolism may be studied in the future in volunteer groups selected for the specific CYP2C9 genotypes.
The study was supported by grants 01EC9408 and 01ZZ9511 of the German Federal Ministry of Education, Science, Research, and Technology. A.S. Aynacioglu was supported by a research fellowship grant of the Association of Clinical Pharmacology Berlin/Brandenburg.