Genetic Variation at the CYP2C Locus and Its Association with Torsemide Biotransformation

Genetic variation at the CYP2C locus and its association with torsemide biotransformation

SV Vormfelde1, M Schirmer1, In 97 unselected volunteers and two additional homozygous carriers of 2 1 CYP2C9*3, we investigated the oral clearance of torsemide in relation to 37 MR Toliat , I Meineke , polymorphisms at the CYP2C gene locus. Torsemide total oral clearance was 3 2 J Kirchheiner ,PNu¨rnberg linearly associated with the number of CYP2C9*3 alleles (geometric mean: and J Brockmo¨ller1 59, 40 and 20 ml/min in carriers of no, one and two alleles) and so were the methyl- and ring-hydroxylation but not the carboxylation clearance. 1Department of Clinical Pharmacology, Haplotypes including the CYP2C9*3 allele were similarly associated with University Medical Centre, Georg-August- the clearances but no other variant and no haplotype not including the University, Go¨ttingen, Germany; 2Cologne Center for Genomics (CCG) and Institute for Genetics, CYP2C9*3 variant. The extended haplotype length (EHL) of the CYP2C9 University of Cologne, Cologne, Germany and haplotypes was positively associated with higher activity of the gene product. 3Department of Pharmacology of Natural Torsemide total oral clearance was predictable with r2 ¼ 82.1% using plasma Products & Clinical Pharmacology, University of concentrations at 0.5, 1, 2 and 24 h. In conclusion, torsemide’s biotransfor- Ulm, Ulm, Germany mation strongly depended on the CYP2C9*3 variant but no other. Higher Correspondence: clearance CYP2C9 haplotypes appear to be evolutionarily selected. Dr SV Vormfelde, Department of Clinical The Pharmacogenomics Journal (2007) 7, 200–211. doi:10.1038/sj.tpj.6500410; Pharmacology, University Medical Centre, published online 12 September 2006 Robert-Koch-Strae 40, 37075 Go¨ttingen, Germany. E-mail: [email protected] Keywords: CYP2C9; evolutionary selection; haplotype; hepatic clearance; polymorphism; torsemide

Introduction

The four drug metabolizing enzymes CYP2C8, CYP2C9, CYP2C18 and CYP2C19 contribute to the biotransformation of about 30% of all drugs. The coding genes are located at a common CYP2C gene locus. Considering the large number of clinically used substrates, including all vitamin K antagonistic oral anti- coagulants, almost all sulfonylurea oral antidiabetic drugs and most nonsteroidal antiinflammatory agents, CYP2C9 may be most important among the CYP2C enzymes.1–4 CYP2C8 plays a major role in the metabolism of cerivastatin, chloroquine, paclitaxel, repaglinide, rosiglitazone, verapamil and zopiclone.5,6 CYP2C19 predominantly metabolizes almost all proton pump inhibitors, several antidepressants, the protease inhibitor nelfinavir and the imidazole antimycotic voriconazol.3,7,8 The role of CYP2C18 is less clear; it is expressed poorly in the liver; CYP2C18 is expressed significantly in the skin and tissue-specific bioactivation of phenytoin by CYP2C18 has been associated with phenytoin’s dermal drug toxicity.9 Genetic variation in the four CYP2C genes and its clinical consequences have been widely studied.2,3,5,10–13 CYP2C9*3, an lle359Leu substitution affects the maximum reaction rate of almost all currently known CYP2C9 substrates, CYP2C9*2, Arg144Cys, differentially affects the metabolism of some drugs more Received 17 March 2006; revised 19 June 2 2006; accepted 27 June 2006; published than of others. Except for CYP2C8*2 and *3, CYP2C9*2 and *3 and CYP2C19*2 online 12 September 2006 functional amino acid substitutions in CYP2C genes are rare in Caucasian CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 201

populations (compare www.imm.ki.se/CYPalleles/). For for amino acid polymorphisms has been unsuccessful, we CYP2C9 substrates, more than 80% of the interindividual focused to detect polymorphisms affecting gene regulation. variation in pharmacokinetics remain unexplained by the We therefore calculated haplotypes and performed haplo- known polymorphisms. Thus, the search has focused on type–phenotype association. regulatory regions and indeed some evidence for functional polymorphisms in the CYP2C9 regulatory regions has been Results reported.14,15 The extended haplotype homozygosity (EHH) method is a Study population new screening method to detect recent positive selection in The 97 unselected healthy male Caucasian volunteers, who 16 the human genome from haplotype structure. It relies on participated in the clinical study had a mean (range) age of the breakdown of haplotypes including a new mutation over 28 years (19–50 years). The mean body weight and height time by recombination processes. Positive selection is were 78 kg (57–93 kg) and 182 cm (170–198 cm), and the assumed the longer a haplotype at a locus of interest mean body surface was 2.0 m2 (range 1.7–2.3 m2). extends in relation to its frequency and other haplotypes. The EHH method has the potential to differentiate between Genetic variation at the CYP2C locus variants, which were evolutionarily selected due to their In total, 37 potential polymorphisms were investigated functional benefits for the organism from those variants, (Figure 1). Four of them turned out to be monomorphic which have merely been selected due to linkage to a and are not discussed further. The allele frequencies of the primarily selected variant.17,18 minor alleles were above 0.05 except for rs1409654 in Torsemide is a loop diuretic drug quite specifically CYP2C18 and rs1058930 in CYP2C8 (Table 1). The poly- metabolized by CYP2C9.19,20 Methyl- and phenyl-hydro- morphisms formed three linkage disequilibrium (LD) blocks. xylation form metabolites termed M1 and M3. M1 is further One LD block included CYP2C18 and CYP2C19 and each oxidized to the carboxylated metabolite M5. Torsemide one LD block included CYP2C9 or CYP2C8 (Figure 2). There clearance was greatly reduced in carriers of the CYP2C9*3 are controversies concerning the LD block definition at the variant but not in carriers of the CYP2C9*2 variant in CYP2C locus.30,31 With our data set, all applied algorithms accordance with in vitro findings.2,19,21,22 Torsemide may for LD block definition came essentially to the same therefore be a valuable CYP2C9 probe drug with specific grouping with the only relevant difference concerning properties compared to the other commonly used substrates whether or not to include some of the 50 polymorphisms tolbutamide,23 warfarin 24–26 or losartan.27 of the CYP2C9 locus into the CYP2C18–CYP2C19 block. Phenotyping unselected samples has lead to the identifi- Inclusion of the most 50 CYP2C9 polymorphism subdivided cation of functionally relevant genetic polymorphisms.28,29 the most frequent haplotype at the CYP2C18–CYP2C19 Here, we applied this approach to the CYP2C locus using including LD block into two subhaplotypes. Inclusion of torsemide as the phenotyping drug. After extensive search further polymorphisms merely resulted in lower frequencies

Figure 1 Gene structure and location of the studied polymorphisms at the cytochrome P450 2C (CYP2C) locus. This figure depicts the location of the CYP2C locus on chromosome 10, the proportional size and location of the four genes including direction of transcription (arrows), gene borders and intron–exon structure and the chromosomal position and rs number of the investigated polymorphisms as given by the NCBI databases Map Viewer, Gene and SNP (NCBI build 35.1). In some cases, no rs number was available. Then the hCV-number referring to the Celera SNP database is given or the polymorphism is termed goe1 or goe1plus2. Putative polymorphisms that turned out monomorphic are given in grey letters.

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 202

Table 1 SNP-wise genotype–phenotype association

Identificationa Nucleotide exchangeb MAFc Cl torsemided Cl methyl hydroxylationd

Per 2 alleles (ml/min) Pe Per 2 alleles (ml/min) Pe

Referencef 61.9 (2.2) 33.9 (1.8)

CYP2C18 rs1010570 G4A 0.28 70.5 (6.4) 38.3 (5.2) rs12243416 C4T 0.26 55.5 (6.6) 34.7 (5.6) rs7900135 A4G 0.15 65.7 (8.0) 33.7 (6.6) rs7099637 T4G 0.26 56.1 (6.6) 33.5 (5.6) rs1409654 T4C 0.01 80.3 (28.4) 43.5 (23.2) rs2281891 G4A 0.15 65.3 (8.0) 33.3 (6.4) rs1042194 C4A 0.15 65.3 (8.0) 33.3 (6.4)

CYP2C19 rs4986894 T4C 0.15 65.3 (8.0) 33.3 (6.4) rs11597292 C4G 0.48 63.1 (6.6) 32.9 (7.4) rs12769205 T4C 0.13 62.3 (9.4) 31.3 (7.6) rs4244285 (*2) C4T 0.15 65.3 (8.0) 33.3 (6.4) rs12164742 C4G 0.23 51.5 (7.8) 28.3 (6.4) rs7475884 C4T 0.15 68.1 (8.0) 35.1 (6.6) rs4417205 C4G 0.15 66.1 (8.0) 34.1 (6.6) rs4917623 T4C 0.47 64.3 (6.8) 28.9 (5.6)

CYP2C9 rs4918758 T4C 0.41 67.9 (6.0) 39.1 (5.0) rs9332120 A4G 0.25 54.1 (6.8) 28.3 (5.8) rs1799853 (*2) C4T 0.13 59.9 (8.6) 31.1 (7.0) rs10509679 C4T 0.18 69.7 (7.4) 42.3 (6.0) rs12268897 G4A 0.33 62.5 (6.8) 32.7 (5.4) hCV27434889 C4T 0.19 68.9 (7.4) 42.1 (6.0) rs1934963 T4C 0.23 63.1 (6.6) 33.3 (5.6) rs1057910 (*3) T4G 0.09 20.7 (8.8) o0.0001 4.5 (7.4) 0.0001 rs9332238 C4T 0.23 60.7 (6.6) 28.3 (5.8) rs1934969 A4T 0.33 60.9 (7.0) 31.5 (5.6)

Between CYP2C9 and CYP2C8 rs2860975 C4A 0.35 63.7 (7.0) 34.9 (5.8)

CYP2C8 rs1058932 C4T 0.19 63.9 (7.4) 34.5 (6.0) rs2275620 A4T 0.32 63.7 (7.2) 37.1 (6.0) rs1934980 T4C 0.19 64.7 (7.4) 35.5 (6.0) rs1058930 (*4) C4G 0.05 78.1 (14.6) 31.3 (12.0) rs11572093 C4T 0.34 61.3 (6.4) 29.7 (5.2) rs11572080 (*3) C4T 0.13 56.5 (9.0) 37.7 (7.4) rs7912549 T4C 0.11 63.3 (10.0) 35.9 (8.2) aAccording to the NCBI database SNP. bMore frequent to less frequent allele. cMAF ¼ minor allele frequency. dMean (s.e.m.) of the total oral torsemide clearance (Cl torsemide) and the methyl hydroxylation clearance (Cl methyl hydroxylation) from linear regression analyses controlled for the effect of the CYP2C9*3 polymorphism (not controlled in the case of CYP2C9*3 itself). eP-value of respective linear regression analyses; only P-values o 0.05 are given. fThe mean total oral torsemide or methyl hydroxylation clearance in carriers of no CYP2C9*3 allele given by linear regression analysis. of the major haplotypes and the prediction of additional five tagging SNPs identified these frequent haplotypes at the infrequent haplotypes. respective LD block (Table 2). The frequent haplotypes Five, five and six frequent haplotypes (allele frequency amounted to 95.1, 98.6 and 98.1% of all haplotypes at 42%) were observed at the CYP2C18–CYP2C19, CYP2C9 the respective LD block. The multiallelic D0-values of and CYP2C8 LD block, respectively (Table 2). Four, four and the CYP2C9 haplotypes were 0.96 and 0.72 with the

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 203

Figure 2 The linkage disequilibrium (LD) among the studied polymorphisms. This figure depicts the linkage between the polymorphisms and the resulting LD blocks at the CYP2C locus. Grey shaded boxes denote the attribution of the polymorphisms to the genes encoding the cyptochrome P450 enzymes CYP2C18, CYP2C19, CYP2C9 and CYP2C8. LD block size is given as horizontal bars as calculated by the software package Haploview, according to the ‘4 gamete rule’, the ‘solid spine of LD’ and the ‘confidence interval’ method with the respective predefined settings. The LD blocks drawn below represent the LD blocks that were primarily used for the haplotype–phenotype association defined relying on the ‘n þ 1’ method of Zeggini et al.43 The polymorphisms are identified as given in Table 1 and Figure 1. Dark red boxes without a value denote a D0-value of 1.00; light blue boxes without a value denote a not determinable value. Numbers denote the 100-fold of the respective D0-values. The higher the value, the darker the fill colour of the boxes and the higher the LD.

CYP2C18–CYP2C19 and the CYP2C8 haplotypes, respec- in the unselected study population (52.5–56.4 ml/min tively. 7s.e.m., range 25.8–128.2 ml/min; Figure 4). The geometric Haplotype calculation including all polymorphisms re- mean value of the extrarenal torsemide clearance was vealed individual coupling among the haplotyes at the three 38.3 ml/min (36.8–39.9, range 16.1–97.5 ml/min). The geo- LD blocks described above (Figure 3). Concerning poly- metric mean methyl hydroxylation clearance was 26.4 ml/ morphisms known to be functional, H3, which includes the min (24.8–28.0, range 1.7–120.8 ml/min) and was an order CYP2C19*2 variant, is tightly coupled with the CYP2C9 of magnitude lower for the ring hydroxylation clearance haplotype H8 and the CYP2C8 haplotype H16 that includes with 1.4 ml/min (1.3–1.5, range 0.4–6.0 ml/min). Taken the CYP2C8*4 variant. The CYP2C9 haplotypes H9 and H10, together, renal, methyl and ring hydroxylation clearances defined by CYP2C9*2 and CYP2C9*3, are particularly tightly left about 15% of the total oral torsemide clearance coupled to neighbouring haplotypes at the CYP2C18– unexplained. The arithmetic mean of the normally distri- CYP2C19 locus. H9 is also coupled to the CYP2C8 haplotype buted carboxylation clearance was 156 ml/min (162–150, H14, which includes the CYP2C8*3 variant. Two subhaplo- range 31–315 ml/min). types of CYP2C9 haplotypes were observed. A subgroup of H8 was coupled to H5 and H12 (H5–H8–H12) and a Genoptype–phenotype association subgroup of H1 and H6 is tightly coupled to the CYP2C8 Most of the 17 heterozygous carriers of a CYP2C9*3 allele haplotype H15, a combination that has a frequency of 7.7% had total and hydroxylation clearances below the mean (H1–H15 and H6–H15). value; the carboxylation clearance was apparently independent from the CYP2C9*3 polymorphism (Figure 4). The Torsemide metabolic phenotype geometric mean values of the total oral torsemide clearance The total oral torsemide clearance, the extrarenal clearance was 58.6 ml/min (56.5–60.87s.e.m., range 28.7–128.2 ml/ and both hydroxylation clearances were log-normally dis- min) in the carriers of no CYP2C9*3 allele. In the hetero- tributed, the secondary carboxylation clearance was nor- zygous group, mean oral torsemide clearance was 40.3 ml/ mally distributed (Figure 4). The geometric (logarithmic) min (37.9–42.8, range 28.0–62.7 ml/min) and 19.9 ml/min mean of the total oral torsemide clearance was 54.4 ml/min (19.4 and 20.4 ml/min) in the two CYP2C9*3 homozygotes

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 204

Table 2 Frequent haplotypes and their activity

No.a Nucleotidesb AFc Cl torsemided Cl methyl hydroxylationd

Per 2 alleles (ml/min) Pe Per 2 alleles (ml/min) Pe

Referencef 61.9 (2.2) 33.9 (1.8)

CYP2C18 and CYP2C19 tSNPg V ..V.V.....V... H1 GCATTGCTGTCCCCC 0.464 55.9 (6.6) 29.1 (5.6) H2 GTAGTGCTCTCGCCT 0.223 52.5 (7.8) 29.5 (6.4) H3 (*2) ACGTTAACCCTCTGT 0.127 65.9 (9.4) 32.7 (7.8) H4 ACATTGCTCTCCCCT 0.105 65.1 (8.0) 36.5 (6.8) H5 GTAGTGCTCTCCCCT 0.032 56.5 (7.4) 31.5 (4.4)

Sumh 0.951

CYP2C9 tSNP ..VVV.V... H6 TACCACTTCT 0.327 58.1 (5.8) 31.5 (5.6) H7 TGCCGCTTCA 0.250 56.9 (6.8) 28.3 (5.8) H8 CACTGTTTCA 0.182 69.7 (7.2) 42.3 (6.0) H9 (*2) CATTGT TTCA 0.132 60.5 (8.6) 31.1 (7.0) H10 (*3) CACCGCCGTA 0.095 20.7 (8.8) o0.0001 4.5 (7.4) 0.0001

Sum 0.986

CYP2C8 tSNP .VVV.VV H11 CATCTCT 0.345 63.1 (6.0) 29.3 (5.2) H12 CTTCCCT 0.212 60.9 (7.8) 36.3 (6.6) H13 TACCCCT 0.141 64.5 (7.2) 35.3 (6.0) H14 (*3) CATCCT T 0.132 66.5 (9.0) 37.7 (7.4) H15 CTTCCCC 0.106 63.9 (10.0) 36.1 (8.2) H16 (*4) TACGCCT 0.045 64.9 (11.2) 35.5 (10.0)

Sum 0.981 aHaplotype number assigned. bNucleotides of polymorphisms in the order given in Table 1. cAF ¼ allele frequency. dMean (s.e.m.) of the mean total oral torsemide clearance (Cl torsemide) and the mean methyl hydroxylation clearance (Cl methyl hydroxylation), respectively, calculated per two haplotype alleles in linear regression analyses controlled for the effect of haplotype 10 (not controlled in the case of haplotype 10 itself). eP-value of respective linear regression analyses; only P-values o0.05 are given. fThe mean total oral torsemide or methyl hydroxylation clearance in carriers of no CYP2C9*3 allele given by linear regression analysis. gTagging SNPs are marked ‘V’. hCumulative allele frequencies at the respective locus. Grey shading denotes the minor alleles of the respective polymorphisms.

(Figure 5). Thus, there was a linear relationship between the compared to 33.971.8 ml/min (r2 ¼ 14.6%, P ¼ 0.0001) and number of CYP2C9*3 alleles and the torsemide clearance. ring hydroxylation clearance was 0.370.4 ml/min com- The methyl and the ring hydroxylation clearances also pared to 1.770.1 ml/min (r2 ¼ 13.3%, P ¼ 0.0003). The depended linearly on the number of CYP2C9*3 alleles association with the carboxylation clearance and the (Figure 5). The major part of the secondary carboxylation residual clearance was not significant (P ¼ 0.34 and 0.96). clearance was independent of the CYP2C9*3 polymorphism Neither the CYP2C9*2 nor any other polymorphism except or any other of the investigated polymorphisms or haplo- the CYP2C9*3 allele was associated with any of the types (Figure 5). clearances even without correction for multiple testing Total oral torsemide clearance was 20.778.8 ml/min in (Table 1). carriers of two CYP2C9*3 alleles compared to 61.972.2 ml/ min in carriers of no CYP2C9*3 allele in linear regression Haplotype–phenotype association analysis (r2 ¼ 18.4%, Po0.0001; Table 1). Correspondingly, The group of carriers of the haplotype H10 was identical the methyl hydroxylation clearance was 4.577.4 ml/min with the group of CYP2C9*3 allele carriers resulting in

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 205

Figure 3 Coupling of haplotypes at the CYP2C locus. This figure Figure 5 Dependency of the total oral torsemide clearance and the depicts the coupling among the frequent haplotyes (those above 2% metabolite formation clearances on the CYP2C9 genotype. Data are total frequency) at the gene loci CYP2C18–CYP2C19 and CYP2C8 in given as geometric means (arithmetic mean for carboxylation clearance) relation to the five frequent haplotypes at the CYP2C9 locus. The box within the respective haplotypic groups7s.e.m. (in the groups with no height roughly resembles the haplotype frequency. CYP2C18–CYP2C19 and one H10 allele). The lines represent respective linear regression and CYP2C8 haplotypes are given and named when occurring with lines. above one percentage point frequency in the respective group defined by a CYP2C9 haplotype. Numbering of the haplotypes H1 through H16 is as in Table 2. *2, *3 and *4 in brackets denote the variant alleles identical association with the phenotype (Table 2). None of CYP2C19*2, CYP2C9*2 and *3 and CYP2C8*3 and *4. Coloured and dark the other haplotypes of CYP2C9 or of the CYP2C18– grey fields resemble single nameable haplotypes; light grey fields CYP2C19 or the CYP2C8 locus was associated with any of denote a number of haplotypes with less than one per cent point the phenotypes (Table 2). Haplotypes were also calculated frequency. H1–H15, H6–H15 and H5–H8–H12 denote subgroups of H1, using a number of further LD block definitions. These H6 and H8 identical to the main group of H1, H6 and H8 but tightly include those implied by Haploview (Figure 2, upper part), coupled to a specific haplotype (H15, H5 and H12) at another LD block. Haplotypes including named polymorphisms and H6–H15 are coloured an LD block over the whole CYP2C18–CYP2C19–CYP2C9 correspondingly to Figure 6. Numbers to the left denote the locus and an LD block including all frequent polymor- corresponding CYP2C9 haplotype numbering of Veenstra et al.13 phisms of the four CYP2C genes. In no case any frequent haplotype that did not include the CYP2C9*3 variant was associated with any phenotype even without correction for multiple testing. In silico analyses concerning haplotype breakdown at the CYP2C locus were performed with polymorphisms analysed in 120 Caucasian chromosomes in the HapMap project www.hapmap.org.32 The LD block including CYP2C9 identified by the software program Sweep spanned 16 polymorphisms and revealed eight haplotypes. Although only three of the polymorphisms were identical to those analysed in our sample, this data set revealed the haplotypes H7 to H10 having comparable frequencies (23, 17, 15 and 6%) as with our polymorphism set. The remaining four haplotypes amount to 40% and were subhaplotypes of H6. One of these four subhaplotypes was defined by the variant allele of rs2475377, had an allele frequency of 6.6% and due to its discrete linkage with variant alleles in the region of rs7912549 most likely resembled H6-H15 (Figure 3). The remaining three subhaplotypes were defined as a group by Figure 4 Distribution of the total oral torsemide clearance and the the variant allele of rs2253635, and subdivided into H6-1 metabolite formation clearances. This figure depicts the distribution with no further variant (15.8% haplotype frequency) H6-2 pattern of the total oral torsemide clearance and the three metabolite with the variant allele of rs2475376 (10.8%) and H6-3 with formation clearances in the unselected population of 97 healthy young the variant allele of rs9332197 (6.7%). The four subhaplo- men and the two additional CYP2C9*3 homozygotes (black coloured bars). Subjects heterozygous for the CYP2C9*3 allele are represented by types were tightly coupled with different variants outside grey coloured bars. The vertical lines indicate the respective mean values CYP2C9 suggesting that they are related but discrete as given in the text. haplotypes.

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 206

The extended haplotype length (EHL) of the CYP2C9 including those shown in Figure 7a. Of total oral torsemide haplotypes was remarkably long. The median EHL of the clearance, 82.1% prediction was achieved calculating torse- CYP2C9 haplotypes was 450 kb compared to 105 and 47 kb mide AUC from only four plasma samples at 0.5, 1, 2 and at the two other LD blocks near the CYP2C locus, which also 24 h after torsemide intake (r2 ¼ 82.1%, median prediction spanned 16 polymorphisms. H6–H15 and H6-1 turned out error of 7.8%, linear model forced through origin, Figure the haplotypes with the greatest EHL, H6-2, H6-3, H8 and 7b). Reduced sampling clearance underestimated the clea- H9 intermediate and H7 and H10 had the shortest EHL rance compared to nonparametric analysis by 10.7%. (Figure 6). Also the variants themselves that define the CYP2C9 haplotypes were associated with relatively long range EHH. The variant defining H10 (CYP2C9*3) had an EHL of approximately 300 kb and those defining H8 and H9 (CYP2C9*2) both had an EHL of approximately 400 kb. The H6-2 and H7 defining variants were associated with an EHL of 500 kb and the variants defining H6-3 and H6-15 with 700 and 900 kb EHL (H6-1 was not definable by a single variant in that data set). These EHLs compare to a median EHL of about 160 kb in those of the 717 variants with 0–0.25 allele frequency. Unfortunately, EHL could not be calculated in our Go¨ttingen population because the extension of the CYP2C9 haplotypes exceeded the range of the analysed polymorphisms. This at least did not contradict the findings in the HapMap data set.

Nongenetic predictors of the torsemide metabolic phenotype Age, height, weight, body mass index (BMI), body surface and serum albumin concentration were not correlated with total torsemide clearance or any of the partial metabolite formation clearances even without adjustment for multiple testing.

CYP2C9 phenotype prediction with reduced sampling Best phenotype prediction was achieved calculating total oral torsemide clearance with area under the curve (AUC) estimates using a reduced number of plasma samples

Figure 7 Phenotyping with reduced sampling. (a) For 13 AUC Figure 6 Breakdown of CYP2C9 haplotypes with distance. This figure calculation methods the employed time points used for the calculation depicts the decline of the EHH with distance in four of the eight CYP2C9 are depicted. The association strength (r2-value in percent) of the haplotypes in the 120 Caucasian HapMap chromosomes (www.hapmap. resulting clearance (calculated as dose/AUC) with the clearance values org). The depicted haplotypes are H6–1, H6–H15, H9 (defined by the from the nonparametric analysis is given on the left side. Blue rectangles CYP2C9*2 allele) and H10 (defined by the CYP2C9*3 allele, numbering denote employed time points. Linear extrapolation was used among according to Table 2). In the gene track at the top, the open boxes time points that are shown adjacent and those connected by a straight proportionally resemble the size of genes at the locus, the genes line. A bended line denotes log-linear extrapolation (calculation details CYP2C18, CYP2C19, CYP2C9 and CYP2C8 are named above the gene in the text). (b) For individual clearances resulting from the AUC track. The vertical bars below the boxes resemble each one genotyped calculation method 4 (0, 0.5, 1, 2 and 24 h) the association with the polymorphism, the bold ones resemble those 16 in the CYP2C9 including clearance values from nonparametric analysis is depicted. The given r2- linkage disequilibrium (LD) block. The horizontal bars below denote the value and the formula refer to the shown regression line, which was size and the position of the LD blocks automatically defined by Sweep. forced through origin.

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 207

Reduced sampling clearance was associated with the number The CYP2C9*2 polymorphism is thought to alter the of CYP2C9*3/H10 alleles with P ¼ 0.0002. interaction of the CYP2C9 enzyme with the NADPH:cyto- Restricting to three plasma samples clearance calculation chrome P450 oxidoreductase.2,33 In vitro, torsemide’s meta- from 1, 2 and 24 h samples reached acceptable r2 ¼ 79.6%. bolism was not affected by the CYP2C9*2 polymorphism Omitting the 24 h sample resulted in r2-values of maximally supposedly as torsemide induces an anodic shift of CYP2C9 55.3% (model 10 in Figure 7). Restricting to two plasma in favour of catalytic activity.2,19,34 Clinical effects of the concentrations, the only prediction value above r2 ¼ 50% CYP2C9*2 variant, however, may also be due to linkage with was observed using the 2 and the 24 h concentration (model other polymorphism alike CYP2C8*335 (Figure 3) or differ- 13 in Figure 7). Single torsemide plasma concentrations and ential contributions of other CYP2C enzymes. In our study, derivatives thereof were even less associated with the total none of the haplotypes at the CYP2C locus including the oral torsemide clearance, irrespective of the association CYP2C9*2 variant was associated with torsemide’s metabolic model employed. As a maximum, the plasma concentration phenotype (Table 2). at t ¼ 24 h p.a. predicted r2 ¼ 24.4% of the total oral An extensive resequencing project covering 60 kb at the torsemide clearance (linear model). CYP2C9 locus revealed at least five distinct haplotype To circumvent blood sampling, phenotype prediction groups in 192 European American warfarin patients.13 The from urine may be preferred. The relation AeM3/Ae torsemide five haplotypes in our population corresponded well to and Ae torsemide itself in the 10–24 h interval turned out the these five haplotype groups despite using different poly- best predictors of the total oral torsemide clearance morphisms. The haplotypes H9 and H10, including (r2 ¼ 22.8 and 19.8% in linear models and below 35 % in CYP2C9*2 and CYP2C9*3, have been termed groups 2 and all other models). 3 by Veenstra et al. (Figure 3). H6, H7 and H8 compare to Veenstra’s groups 1A, 1B and 1C. Extending over CYP2C9, we observed strong linkage of the CYP2C9 haplotypes with Discussion the haplotypes at the CYP2C18–CYP2C19 and the CYP2C8 locus (Figure 3). Knowledge of such LD may be crucial to the In an earlier study in 36 selected subjects selected for interpretation of genotype–phenotype associations in that CYP2C9*2 and *3 alleles, we demonstrated that torsemide’s apparent association may in fact be due to association with a in vivo biotransformation depended on the CYP2C9*3 haplotype at a neighbouring gene. polymorphism but not on the CYP2C9*2 polymorphism.21 The great length of the CYP2C9 haplotypes as a group As over 80% of interindividual variation in the metabolic observed in the 120 Caucasian HapMap chromosomes (and clearance of torsemide remained unexplained, we investi- not contradicted by the findings in our subjects) implies gate a larger population with an extended set of analysed recent positive selection of the now present CYP2C9 polymorphisms. The CYP2C9*2 and the CYP2C9*3 allele haplotypes. Among them, the EHL is positively associated each defined exactly one CYP2C9 haplotype. The dominant with the clearance function of the resulting enzyme. This parts of the total oral torsemide clearance and the hydro- association might indicate positive selection for high xylation clearances are linearly associated with the number clearing CYP2C9 alleles. As CYP2C9 metabolizes organic of alleles of the haplotype defined by CYP2C9*3 but not anions, it is striking that in a search for signals of with CYP2C9*2 or any other of the 33 studied polymor- evolutionary selection in 168 genes related to immune phism. As the LD block definition at the CYP2C locus has function the haplotype with the most outstanding haplo- recently been a matter of discussion,30,31 we evaluated type length in Caucasians was the most frequent allele of the haplotype–phenotype associations with haplotypes from a multispecific organic anion transporter gene ABCC1.17 number of LD block definitions. This repetitively revealed Well-established in vivo phenotyping probes for CYP2C9 the association with haplotypes including the CYP2C9*3 activity are warfarin and tolbutamide.4,26,36–38 Tolbutamide allele but not others. The secondary carboxylation clearance will probably not be available as a registered drug in the next was independent from any variation in CYP2C9, including years in many countries and warfarin requires more CYP2C9*3. None of the tested potential non-genetic sophisticated enantiospecific analysis to be optimal.39 Thus, predictors including body weight and body surface had a one may think about alternative CYP2C9 phenotyping significant impact on the phenotype. methods. Torsemide as a probe would be specific for CYP2C9 The linear form of the relationship between the number of and pharmacogenetically only dependent on the CYP2C9*3 CYP2C9*3 alleles and the torsemide clearance correspond to variant. Such selectivity may be helpful to study effects of a codominant mode of inheritance (linear gene-dose effect) CYP2C9 induction or inhibition including or not the effect where the expression of the two alleles is (1) independent of the CYP2C9*3 variant but without interference by the from each other and (2) not regulated via feedback substrate CYP2C9*2 variant. For the ease of study performance or a or product inhibition. In the present study, 1.1 ml/min or potential clinical application one may want to circumvent 13% of the methyl hydroxylation clearance and 0.3 ml/min 24 h serum profiles or 24 h urine sampling as has recently, or 20% of the ring hydroxylation clearance were not for example, been suggested for CYP3A4 phenotyping.40 attributed to the CYP2C9*3 polymorphism (compare Figure Prediction values of 82.2 and 79.8% using four or three 5) marking the upper limit of a contribution of enzymes plasma concentrations, respectively, are comparable to other than CYP2C9 to the formation of these metabolites. single point values well above r2 ¼ 80% with CYP3A4

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 208

probes40 and were sufficient to detect the association the supplier and proprietary pyrosequencing software with the CYP2C9*3 genotype. However, no single point (Biotage AB, Uppsala, Sweden). Controls were included to measure or measure from urine sampling reached any exclude mix-ups and other errors during genotyping: DNA- acceptable prediction value. Such weak predictability of free reaction mix and dedicated DNA. All polymorphisms the clearance from single point measures might indicate were in Hardy–Weinberg equilibrium. high variation in the volume of distribution or in the absorption time course. LD block definition and EHH test In conclusion, torsemide is a CYP2C9 substrate with a LD blocks were identified using the software Haploview, hepatic clearance dependent on the CYP2C9*3 but not the version 3.2.41 The linkage plot relied on the D0 value of CYP2C9*2 variant. We could not elucidate the basis for Lewontin.42 LD block definition is a critical matter of the about 80% variation in CYP2C9 activity unexplained by current evaluation and ongoing discussion.43–45 Recently, the CYP2C9*3 variant. Further studies like formal twin the LD block definition at the CYP2C locus has been studies are required to reveal how much of the 80% discussed controversially.30,31 The three nearly gene-wise unexplained inter-individual variability is really genetic. If LD blocks mainly used in this study were defined relying on truly genetic, this variation may be caused by trans-acting the ‘n þ 1’ method of Zeggini et al.43 The individual factors localized far away from the CYP2C locus or on other haplotypes were calculated using the software program chromosomes. Sufficient prediction of total oral torsemide Phase, version 2.1.1.46,47 Ten runs were performed starting clearance may be achieved with only three to four plasma from the seed values 2, 1536, 2936, 3122, 4957, 5283, 6757, samples. Finally, we suggest an evolutionary preference for 7992, 8633 and 9045. The haplotype pair most often higher functional CYP2C9 alleles. assigned to the individuals was used for the genotype– phenotype association. Materials and methods The EHH and the EHL were explored using the software program Sweep, version 1.0, freely available from http:// Clinical study www.broad.mit.edu/mpg/sweep/. Standard settings were An unselected population of 108 healthy healthy, nonsmok- used throughout. The EHH of single polymorphisms was ing, male volunteers aged between 19 and 50 years was explored defining LD blocks consisting solely of the very screened. Ninety-seven of them plus two homozygous polymorphism. The EHL is the range demarked by 50% EHH carriers of the CYP2C9*3 allele (total of 99 subjects) breakdown to both sides of a given haplotype.48 With our completed an open-label single-dose pharmacokinetic study. population we obtained the hint on relatively long EHL. All subjects were white. All volunteers were healthy as However, the apparent extension of the haplotypes beyond confirmed by physical examination and routine laboratory the CYP2C locus prevented meaningful description of the analyses. The study was approved by the Ethics Committee haplotype breakdown, namely calculation of EHL, in our of the Georg August University, Go¨ttingen, Germany, and population. In addition, maximal density of genotyped was carried out in accordance with the principles of the polymorphism is mandatory for complete haplotype sub- Declaration of Helsinki. All subjects provided written division as incomplete haplotype subdivision results in informed consent before inclusion in the study and after falsely fast apparent breakdown of the incompletely sub- full explanation of all study procedures. A subpopulation of divided haplotype.48 We therefore used the publicly avai- the study population presented here formed the population lable phased data of 717 polymorphisms spanning a 2.4 Mb published earlier.21 After an overnight fast, 10 mg torsemide region from 95.2 to 97.6 Mb on chromosome 10 for the 120 (Unat; Roche Diagnostics GmbH, Mannheim, Germany) was Caucasian (CEU) chromosomes of the HapMap project from administered orally after the subjects had emptied their www.hapmap.org (release no. 19). bladder. Volunteers fasted until 4 h after drug administra- tion. Blood samples were taken immediately before and at Drug concentration and pharmacokinetic analysis 0.5, 1, 2, 3, 4, 6, 8, 10 and 24 h after dosage. Urine was The methods to measure torsemide, its methyl hydroxylated collected in the respective intervals, weighed and a sample metabolite M1, M5, which is the corresponding 3-carboxy- stored together with the plasma samples at À201C. lated metabolite, and the phenolic metabolite M3 in plasma and urine samples have been published earlier.21,49 The limit Genotyping of quantification was 10 ng/ml for all analytes from plasma Genotyping was performed in the 108 screened subjects plus samples and 1 ng/ml for all analytes from urine samples. the two CYP2C9*3 homozygotes (total of 110 subjects). The Pharmacokinetic analysis was performed with the nonpara- location of the studied polymorphisms in the CYP2C gene metric analysis module of the WinNonlin software, version locus is illustrated in Figure 1. PCR primers and condition 2.1. Torsemide total oral clearance was calculated as dose/ details are given in Table 3. PCRs were carried out in a AUC with extrapolation to infinity (AUCtorsemide 0–N). The reaction volume of 15 ml with 6 ng genomic DNA, 10 mM renal torsemide clearance was calculated as Ae torsemide 0–N/ dNTP, 10 mM of each forward and reverse oligonucleotide AUCtorsemide 0–N, where Ae torsemide 0–N is the amount of PCR primers, Taq polymerase (5 U/ml), 1.5 mM MgCl2 torsemide excreted unchanged in the urine extrapolated to (25 mM). Genotyping was performed by Pyroseqencing on infinity; the extrarenal torsemide clearance was then total the PSQ HS96A System using standard protocols defined by oral minus renal torsemide clearance. Three metabolite

The Pharmacogenomics Journal Table 3 Primer sequences and PCR conditions of the Pyrosequencing assays

Identificationa Forward primer (5030) Reverse primer (5030) Internal primer (5030) PCR thermoprofile

CYP2C18 rs1010570 b CAGGCATGATTTGCTTTT CTGCTGCTTTCTTTGCTT CCTGAAGGAAAATGGTA 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs12243416 b AGGCAAACCAATCTATCTTCTT CAGCAGAGGGGGAGACTAAA AAATTCAGGTTGATGGA 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs7900135 b ATTTTGCGGGCATTATTCTG AACTGGATTTGGGGGTTTG GGGAAAAACCACCAG 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs7099637 b TAGGTGGTTAATGTCCAAACTC GCCAGCCATTATTATATTCCAG AATGAATAAGAATATAGAAA 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs1409654 b GCAGATTCTGGCTCTGCAT TGCCAAATTGGTTATCATCC AAACAAGTGTGTGGCTAA 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs2281891 AGCACAGAAGTCAGGGATGTT b GTCTGTCTGGGTGGGAATGT TCAGGGATGTTATTATGG 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs1042194 b CAATGAACATCCAACCTCCA AGCAGAGAATAAATGGCTCTGAA AATCATTTTTCTTTGTTTT 95 50, 45(9515s,5830s,7215s)7250, 4 for ever

CYP2C19 rs4986894 GGGACGAAGGAGAACAAGAC b TAAGACAACCGTGAGCTTGC GGTGATTGGCCACTT 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs11597292 b CTTGGAGGGGTTTGTTCCAT ACAGTGCAAAAACCCAAGGTG GGCAGAAGGAAACCAT 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs12769205 b GTCAGCTTCCTCTTTCTTGC GGTCCTCAATGCTCCTCTT CAGGAAGAGAAACGAAA 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs4244285 b CAGAGCTTGGCATATTGTATC GTAGTAAACACAAAACTAGTCAATG TTAAGTAATTTGTTATGGGT 95 50, 45(9545s,5545s,7245s)7250, 4 for ever rs12164742 TTTTAGAATAAGCACGATGTG b TTAGGCTCCCACACAATAA GGTCCACTTGGTCCAG 95 50, 45(9530s,5430s,7230s)7250, 4 for ever rs7475884 AACCAGCGTTGGCTGTCTT b GCGTCTCCCAAAAGAGACATA GGCTTTGATTTGAATTT 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs4417205 CCGCTCCTATTCAATATTTTT b GACTGTTGGTTTTGCTTTTC AATTTTGAATTTACTGTCAT 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs7474800 TTCCTGGCTGCTTTGTTTAC b TGCACCACGAGATTATATCCC TGTGCTAGCAATTAGTGAG 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs4917623 TCTGTCTGGAAATGGTACTGCT b ATGGTTGTGCCCTAGAAACA TGCTCTTCTTTGGAATG 95 50, 45(9515s,5830s,7215s)7250, 4 for ever

CYP2C9 b 0 0

rs4918758 ACTGTGGGTCCATTTAGTGAT AGACATGGCTGCTTTCTATT CCCTACCTCCCATCTT 95 5 , 45(9515s,6030s,7215s)725, 4 for ever Vormfelde SV CYP2C goe1 and goe1plus2c TCCATGTCTGGAGGTGACTGG b AAACAGCACCAGAGATGCTCA CGTTTCACTTCTGCAGT 95 50, 45(9515s,6030s,7215s)7250, 4 for ever b 0 0

rs9332120 CATTTTCCCACTGGCTGAAA ATCCCAGGCAAGAAAGAGGA GCTCTGCCCGAGG 95 5 , 45(9515s,6030s,7215s)725, 4 for ever metabolism torsemide and variation genetic rs12414460 b CCTGGGATCTCCCTCCTAGT ACCCACCCTTGGTTTTTCTC ATGAGGGAGAAACGC 95 50, 45(9515s,6030s,7215s)7250, 4 for ever b 0 0 rs1799853 GTATTTTGGCCTGAAACCCATA CACCCTTGGTTTTTCTCAACTC GGGAAGAGGAGCATTGAGGAC 95 5 , 45(9545s,5545s,7245s)725, 4 for ever al et rs10509679 b ACTGAATAGGGCAGGGGTGT CCAGCTCAGCACAGGGTATAAG AACCAATAATTTTGTCATTC 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs12268897 TTCAGAAGAGTACCCAGCCTTG b CAAAGCTTGAGACCTGAGAATG AGTTAGGCTACAAGGGG 95 50, 45(9515s,6030s,7215s)7250, 4 for ever hCV27434889 b CTCAAACTCCTGACCCTGT ACAGGCCAATATCCCTGAT CATCAAAAAGCTTATCCAC 95 50, 45(9515s,6130s,7215s)7250, 4 for ever rs1934963 ATCTCCCTGCCTACGTTTCCT b CTCATGCCTTTTCTGGCACT CATCAGGTAACTGATTTCC 95 50, 45(9515s,5830s,7215s)7250, 4 for ever rs1057910 b TGCACGAGGTCCAGAGAT GATACTATGAATTTGGGACTTC TGGTGGGGAGAAGGTC 95 50, 45(9545s,5545s,7245s)7250, 4 for ever rs9332238 and rs1934969cbGCACCCATCCACCCATCTA AGGGCTTCTCCCACACAAAT ATGCAATAACTGTTCAGTG 95 50, 45(9515s,5830s,7215s)7250, 4 for ever

Between CYP2C9 and CYP2C8 rs2860975 b GTTTCCCTTCCTTTCAGGAT ACTCCTGCCTAAGACATTGG AGCAAGAAGAGCCAGA 95 50, 45(9515s,5830s,7215s)7250, 4 for ever

CYP2C8 b 0 0 h hraoeoisJournal Pharmacogenomics The rs1058932 TTTCTCTGCCACCCTCATACC GGATGCTTATGGGATATTGAGTG TGAAGAATGCTAGCCC 95 5 , 45(9515s,6030s,7215s)725, 4 for ever rs2275620 ATCCCCAAGGTAAGCTTGTT b TTGCACTTCTCTCATCTTGTGTT ATTTCTGTACTTCTGAAAT 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs1934980 TGTCTTTGCTTGGGGAATAA b AGAAAGTGGGCAAACAACA GGGAATAATCAATGAAGC 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs1058930 GGGAGAAAGTAAAAGAACACC b CATTCTAGCCATTGGACAATT CAATCCTCGGGACTTT 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs11572093 TAAAGATTGGAGGCTGAAA b GAAGAAAATAATTACCACCTTTG GGAGGCTGAAAATTTTG 95 50, 45(9515s,5430s,7215s)7250, 4 for ever rs11572080 AGTCACCCACCCTTGGTTTTTC b CACAACCTTGCGGAATTTTGG CCTCTTGAACACGGTC 95 50, 45(9515s,6030s,7215s)7250, 4 for ever rs7912549 b ATTTATGACCTTGAGGGAAATC GGGCTAAGTCTCCTATTTTTTGAT ATTACAATGTACATTTTTTA 95 50, 45(9515s,6030s,7215s)7250, 4 for ever

a rs number according to the NCBI database SNP. b Biotinylated primer. c

These polymorphisms were determined together in our run. 209 CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 210

formation clearances were calculated after adjustment for Acknowledgments the different molecular weights by multiplying Ae,M1 and Ae,M3 by 0.956 and Ae,M5 by 0.9184. Torsemide methyl The study was in part supported by the national genome research network Grants 01 GS 0107 and 01 GR 0416. The skilful hydroxylation clearance was calculated as (AeM1þ AeM5)/ AUC . Carboxylation clearance to form M5 from M1 contributions of Franziska Tuchen, Michaela Torn and Jan Wester- torsemide mann to the clinical study part and that of Sabine Engelhardt to the was calculated A /AUC . Torsemide ring hydroxylation eM5 M1 drug concentration analyses are gratefully acknowledged. clearance was calculated by AeM3/AUCtorsemide.

Duality of Interest CYP2C9 phenotype prediction with reduced sampling For phenotyping purposes of large samples, reduced sam- The authors are unaware of any further duality of interest. pling may greatly increase feasibility of clinical studies. We explored a number of potential predictors of total oral References torsemide clearance calculated by nonparametric analysis as gold standard. These included single and reduced numbers 1 Miners JO, Birkett DJ. Cytochrome P4502C9: an enzyme of major of serum concentrations of torsemide, M1 and M5 and the importance in human drug metabolism. Br J Clin Pharmacol 1998; 45: 525–538. summed serum concentration of M1 and M5. Reduced 2 Kirchheiner J, Tsahuridu MWJ, Roots I, Brockmo¨ller J. The CYP2C9 sampling clearance was calculated as dose/AUC. Then, AUC polymorphism: from enzyme kinetics to clinical dose recommendations. was calculated as the sum of interval AUCs. Most interval Personalized Med 2004; 1: 63–84. AUCs were calculated by the linear trapezoidal rule as 3 Wilkinson GR. Drug metabolism and variability among patients in drug response. N Engl J Med 2005; 352: 2211–2221. individual duration of the interval times the mean of the 4 Kirchheiner J, Brockmo¨ller J. Clinical consequences of cytochrome P450 respective plasma concentrations at the beginning and the 2C9 polymorphisms. Clin Pharmacol Ther 2005; 77: 1–16. end of the interval. In most instances, in the interval 5 Totah RA, Rettie AE. Cytochrome P450 2C8: substrates, inhibitors, including the concentration at 24 h, the AUC was calculated pharmacogenetics, and clinical relevance. Clin Pharmacol Ther 2005; 77: 341–352. by the log-linear trapezoidal rule as ten to the power of the 6 Rendic S. Summary of information on human CYP enzymes: human mean of the logarithms to the basis 10 of the concentrations P450 metabolism data. Drug Metab Rev 2002; 34: 83–448. at the beginning and at the end of the interval. In the model 7 Klotz U, Schwab M, Treiber G. CYP2C19 polymorphism and proton specified in the results section, a value of 10 mg/l was added pump inhibitors. Basic Clin Pharmacol Toxicol 2004; 95: 2–8. 8 Kirchheiner J, Nickchen K, Bauer M, Wong ML, Licinio J, Roots I et al. to the 24 h concentration resulting in a calculation of the Pharmacogenetics of antidepressants and antipsychotics: the contribu- clearance as tion of allelic variations to the phenotype of drug response. Mol Psychiatry 2004; 9: 442–473. 10mg P 9 Kinobe RT, Parkinson OT, Mitchell DJ, Gillam EM. P450 2C18 catalyzes 0:5Ãðlog ðconcentration at tÞþlog ðconcentration at 24hþ10 mg=lÞÞ interval AUCs up to time tþ10 10 10 the metabolic bioactivation of phenytoin. Chem Res Toxicol 2005; 18: 1868–1875. Further tested potential predictors included inverse serum 10 de Morais SM, Wilkinson GR, Blaisdell J, Nakamura K, Meyer UA, concentrations and the quotients of M1, M5 or M1 þ M5 per Goldstein JA. The major genetic defect responsible for the polymorph- A ism of S-mephenytoin metabolism in humans. J Biol Chem 1994; 269: torsemide concentrations. We also tested the e of torse- 15419–15422. mide, the three metabolites and the summed metabolites 11 DeLozier TC, Lee SC, Coulter SJ, Goh BC, Goldstein JA. Functional M1 þ M5 and M1 þ M3 þ M5 as well as the inverses of these characterization of novel allelic variants of CYP2C9 recently discovered each in the respective interval and summed up from the in southeast Asians. J Pharmacol Exp Ther 2005; 315: 1085–1090. 12 DeLozier TC, Tsao CC, Coulter SJ, Foley J, Bradbury JA, Zeldin DC et al. time after intake to the end of the very interval. Finally, we CYP2C44, a new murine CYP2C that metabolizes arachidonic acid to explored the quotients of the Ae of the metabolites and their unique stereospecific products. J Pharmacol Exp Ther 2004; 310: 845–854. sums per Ae torsemide. We explored a number of models 13 Veenstra DL, Blough DK, Higashi MK, Farin FM, Srinouanprachan S, including linear, potential and exponential relationship. Rieder MJ et al. CYP2C9 haplotype structure in European American warfarin patients and association with clinical outcomes. Clin Pharmacol Besides the determination coefficient we calculated the Ther 2005; 77: 353–364. mean prediction error (bias) that has been discussed to 14 Takahashi H, Ieiri I, Wilkinson GR, Mayo G, Kashima T, Kimura S et al. 50- provide additional information.50 Flanking region polymorphisms of CYP2C9 and their relationship to S- warfarin metabolism in white and Japanese patients. Blood 2004; 103: 3055–3057. 15 Shintani M, Ieiri I, Inoue K, Mamiya K, Ninomiya H, Tashiro N et al. Data analysis Genetic polymorphisms and functional characterization of the 50- Association with the torsemide clearances was performed by flanking region of the human CYP2C9 gene: in vitro and in vivo studies. two-sided linear regression analysis using the software Clin Pharmacol Ther 2001; 70: 175–182. package SPSS, version 12.0.0, P 0.05 was considered 16 Sabeti PC, Reich DE, Higgins JM, Levine HZ, Richter DJ, Schaffner SF o et al. Detecting recent positive selection in the human genome from significant. Linear regression analysis was performed with haplotype structure. Nature 2002; 419: 832–837. the number of CYP2C9*3 as independent predictor. Any 17 Walsh EC, Sabeti P, Hutcheson HB, Fry B, Schaffner SF, de Bakker PI further linear regression analysis was performed controlled et al. Searching for signals of evolutionary selection in 168 genes related for the effect of the CYP2C9*3 allele, if not stated otherwise. to immune function. Hum Genet 2006; 119: 92–102. 18 Schirmer M, Toliat MR, Haberl M, Suk A, Kamdem LK, Klein K et al. The Shapiro–Wilk test was used to test for normal distribu- Genetic signature consistent with selection against the CYP3A4*1B allele tion of the data. in non-African populations. Pharmacogenet Genomics 2006; 16:59–71.

The Pharmacogenomics Journal CYP2C genetic variation and torsemide metabolism SV Vormfelde et al 211

19 Miners JO, Coulter S, Birkett DJ, Goldstein JA. Torsemide metabolism by 34 Johnson DL, Lewis BC, Elliot DJ, Miners JO, Martin LL. Electrochemical CYP2C9 variants and other human CYP2C subfamily enzymes. characterisation of the human cytochrome P450 CYP2C9. Biochem Pharmacogenetics 2000; 10: 267–270. Pharmacol 2005; 69: 1533–1541. 20 Miners JO, Rees DL, Valente L, Veronese ME, Birkett DJ. Human hepatic 35 Yasar U, Lundgren S, Eliasson E, Bennet A, Wiman B, de Faire U et al. cytochrome P450 2C9 catalyzes the rate-limiting pathway of torsemide Linkage between the CYP2C8 and CYP2C9 genetic polymorphisms. metabolism. J Pharmacol Exp Ther 1995; 272: 1076–1081. Biochem Biophys Res Commun 2002; 299: 25–28. 21 Vormfelde SV, Engelhardt S, Zirk A, Meineke I, Tuchen F, Kirchheiner J 36 Depre M, Van Hecken A, Oeyen M, De Lepeleire I, Laethem T, et al. CYP2C9 polymorphisms and the interindividual variability in Rothenberg P et al. Effect of aprepitant on the pharmacokinetics pharmacokinetics and pharmacodynamics of the loop diuretic drug and pharmacodynamics of warfarin. Eur J Clin Pharmacol 2005; 61: torsemide. Clin Pharmacol Ther 2004; 76: 557–566. 341–346. 22 Kerdpin O, Elliot DJ, Boye SL, Birkett DJ, Yoovathaworn K, Miners JO. 37 Miners JO, Birkett DJ. Use of tolbutamide as a substrate probe for Differential contribution of active site residues in substrate recognition human hepatic cytochrome P450 2C9. Methods Enzymol 1996; 272: sites 1 and 5 to cytochrome P450 2C8 substrate selectivity and 139–145. regioselectivity. Biochemistry 2004; 43: 7834–7842. 38 Tomalik-Scharte D, Jetter A, Kinzig-Schippers M, Skott A, Sorgel F, 23 Kirchheiner J, Bauer S, Meineke I, Rohde W, Prang V, Meisel C et al. Klaassen T et al. Effect of propiverine on cytochrome P450 enzymes: Impact of CYP2C9 and CYP2C19 polymorphisms on tolbutamide a cocktail interaction study in healthy volunteers. Drug Metab Dispos kinetics and the insulin and glucose response in healthy volunteers. 2005; 33: 1859–1866. Pharmacogenetics 2002; 12: 101–109. 39 Scordo MG, Pengo V, Spina E, Dahl ML, Gusella M, Padrini R. Influence 24 Rieder MJ, Reiner AP, Gage BF, Nickerson DA, Eby CS, McLeod HL et al. of CYP2C9 and CYP2C19 genetic polymorphisms on warfarin main- Effect of VKORC1 haplotypes on transcriptional regulation and warfarin tenance dose and metabolic clearance. Clin Pharmacol Ther 2002; 72: dose. N Engl J Med 2005; 352: 2285–2293. 702–710. 25 Kaminsky LS, de Morais SM, Faletto MB, Dunbar DA, Goldstein JA. 40 Chaobal HN, Kharasch ED. Single-point sampling for assessment of Correlation of human cytochrome P4502C substrate specificities with constitutive, induced, and inhibited cytochrome P450 3A activity with primary structure: warfarin as a probe. Mol Pharmacol 1993; 43:234–239. alfentanil or midazolam. Clin Pharmacol Ther 2005; 78: 529–539. 26 Kim JS, Nafziger AN, Gaedigk A, Dickmann LJ, Rettie AE, Bertino Jr JS. 41 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization Effects of oral vitamin K on S- and R-warfarin pharmacokinetics and of LD and haplotype maps. Bioinformatics 2005; 21: 263–265. pharmacodynamics: enhanced safety of warfarin as a CYP2C9 probe. 42 Lewontin RC. The interaction of selection and linkage. Ii. Optimum J Clin Pharmacol 2001; 41: 715–722. models. Genetics 1964; 50: 757–782. 27 Yasar U, Dahl ML, Christensen M, Eliasson E. Intra-individual variability 43 Zeggini E, Barton A, Eyre S, Ward D, Ollier W, Worthington J et al. in urinary losartan oxidation ratio, an in vivo marker of CYP2C9 activity. Characterisation of the genomic architecture of human chromosome Br J Clin Pharmacol 2002; 54: 183–185. 17q and evaluation of different methods for haplotype block definition. 28 Wedlund PJ, Aslanian WS, McAllister CB, Wilkinson GR, Branch RA. BMC Genet 2005; 6: 21. Mephenytoin hydroxylation deficiency in Caucasians: frequency of a 44 Ding K, Zhou K, Zhang J, Knight J, Zhang X, Shen Y. The effect of new oxidative drug metabolism polymorphism. Clin Pharmacol Ther haplotype-block definitions on inference of haplotype-block structure 1984; 36: 773–780. and htSNPs selection. Mol Biol Evol 2005; 22: 148–159. 29 Cascorbi I, Drakoulis N, Brockmo¨ller J, Maurer A, Sperling K, Roots I. 45 Zhu X, Zhang S, Kan D, Cooper R. Haplotype block definition and its Arylamine N-acetyltransferase (NAT2) mutations and their allelic linkage application. Pac Symp Biocomput 2004; 9: 152–163. in unrelated Caucasian individuals: correlation with phenotypic activity. 46 Stephens M, Donnelly P. A comparison of bayesian methods for Am J Hum Genet 1995; 57: 581–592. haplotype reconstruction from population genotype data. Am J Hum 30 Ahmadi KR, Weale ME, Xue ZY, Soranzo N, Yarnall DP, Briley JD et al. A Genet 2003; 73: 1162–1169. single-nucleotide polymorphism tagging set for human drug metabo- 47 Stephens M, Smith NJ, Donnelly P. A new statistical method for lism and transport. Nat Genet 2005; 37: 84–89. haplotype reconstruction from population data. Am J Hum Genet 2001; 31 Walton R, Kimber M, Rockett K, Trafford C, Kwiatkowski D, Sirugo G. 68: 978–989. Haplotype block structure of the cytochrome P450 CYP2C gene cluster 48 Sabeti PC, Walsh E, Schaffner SF, Varilly P, Fry B, Hutcheson HB et al. on chromosome 10. Nat Genet 2005; 37: 915–916; author reply 916. The case for selection at CCR5-Delta32. PLoS Biol 2005; 3: e378. 32 Anonymous. The International HapMap Project. Nature 2003; 426: 49 Engelhardt S, Meineke I, Brockmo¨ller J. Improved solid-phase extraction 789–796. and HPLC measurement of torasemide and its important metabolites. 33 Crespi CL, Miller VP. The R144C change in the CYP2C9*2 allele alters J Chromatogr B Analyt Technol Biomed Life Sci 2006; 831: 31–35. interaction of the cytochrome P450 with NADPH:cytochrome P450 50 Sheiner LB, Beal SL. Some suggestions for measuring predictive oxidoreductase. Pharmacogenetics 1997; 7: 203–210. performance. J Pharmacokinet Biopharm 1981; 9: 503–512.

The Pharmacogenomics Journal