|Home | About | Journals | Submit | Contact Us | Français|
A considerable number of reports identifying and characterizing genetic variants within the CYP2C9 coding region have appeared. Much less is known about polymorphic promoter sequences that also might contribute to interindividual differences in CYP2C9 expression. To address this problem, approximately 10,000 bp of CYP2C9 upstream information was re-sequenced using 24 DNA samples from the Coriell Polymorphism Discovery Resource. Thirty-one single nucleotide polymorphisms (SNPs) were identified; nine were novel while 22 were previously reported. Using both sequencing and multiplex single-base extension, individual SNP frequencies were determined in 193 DNA samples obtained from unrelated, self-reported Hispanic Americans of Mexican descent and compared to similar data obtained from a non-Latino White cohort. Significant inter-ethnic differences were observed in several SNP frequencies, some of which appeared unique to the Hispanic population. Analysis using PHASE 2.1 inferred nine common (>1%) variant haplotypes, two of which included the g.3608C>T (R144C) CYP2C9*2 and two the g.42614A>C (I359L) CYP2C9*3 SNPs. Haplotype variants were introduced into a CYP2C9/luciferase reporter plasmid using site-directed mutagenesis and the impact of the variants on promoter activity assessed by transient expression in HepG2 cells. Both constitutive and PXR-mediated inducible activities were measured. Haplotypes 1B, 3A, and 3B each exhibited a 65% decrease in constitutive promoter activity relative to the reference haplotype. Haplotypes 1D and 3B exhibited a 50% decrease and a 40% increase in induced promoter activity, respectively. These data suggest that genetic variation within CYP2C9 regulatory sequences likely contributes to differences in CYP2C9 phenotype both within and among different populations.
Many factors contribute to an individual’s response to medications including age, dietary intake, concomitant medication, and various innate pharmacokinetic and pharmacodynamic parameters (Gage et al., 2004; Kamali et al., 2004). Among these, genetic factors that influence drug metabolism play a major role and contribute substantially to observed interindividual variability in response. While numerous enzymes are involved drug metabolism, the cytochrome P450-dependent monooxygenase superfamily is widely recognized as having a prominent role. Consistent with these two facts, a recognized hallmark of cytochrome P450-dependent metabolism is large intersubject variation in the human population. Although environmental factors resulting in induction or suppression contribute to intersubject variation, our current understanding suggests that genetic variability is equally if not more important (Phillips et al., 2001).
CYP2C9 is a major human cytochrome P450 enzyme that accounts for approximately 20% of the total cytochrome P450 protein content in adult human liver (Shimada et al., 1994) and is responsible for the metabolism of approximately 16% of clinically used drugs cleared by oxidative pathways (Williams et al., 2004). Examples of small molecular weight therapeutics for which CYP2C9 is important for disposition include the anticoagulant, warfarin, the antidiabetic agents, tolbutamide and glipizide, the anticonvulsant, phenytoin, the antihypertensive, losartan, the antidepressant, fluoxetine and a number of nonsteroidal anti-inflammatory drugs such as ibuprofen, diclofenac and celecoxib (Miners and Birkett, 1998; Klose et al., 1998; Davies et al., 2000). Over the past several years, multiple CYP2C9 single nucleotide polymorphisms (SNPs) resulting in amino acid changes have been identified. Currently more than thirty allelic variants are listed on the cytochrome P450 allele website (http://www.cypalleles.ki.se), however, not all have been fully characterized. The most common CYP2C9 alleles include CYP2C9*1A (reference allele), CYP2C9*2 (three haplotype variants, all containing the g.3608C>T, R144C SNP), and CYP2C9*3 (two haplotype variants, all containing the g.42614A>C, I359L SNP). Less common are CYP2C9*4 (g.42615T>C, I359T) and CYP2C9*5 (g.42619C>G, D360E), and CYP2C9*6 (g.10601delA). The CYP2C9*2 and CYP2C9*3 variant alleles encode enzymes exhibiting significantly lower intrinsic clearance both in vivo and in vitro (reviewed in Lee et al., 2002).
As evidenced above, extensive research has been conducted to identify and characterize SNPs within the CYP2C9 coding region. However, much less is known about other variant sites, such as polymorphic promoter sequences, that also might contribute to observed interindividual differences in CYP2C9 expression. In a population study of Japanese epileptic patients, seven polymorphic sites were identified within the first 2000 bp upstream of the CYP2C9 transcription start site (Shintani et al., 2001). Several of the identified SNPs were in linkage disequilibrium, resulting in six unique haplotypes. One of the haplotypes (g.-1912T>C, g.-1885C>, g.-1538G>A, g.-1189C>T, and g.-982G>A) exhibited a 60% reduction in in vitro promoter activity. Further, an association between estimated phenytoin intrinsic clearance and promoter activity was observed with respect to the identified CYP2C9 upstream haplotypes (Shintani et al., 2001). However, because the promoter SNPs defining this pattern also were in linkage disequilibrium with the SNP defining CYP2C9*3, it is unclear what impact these variants might have on CYP2C9 phenotype in vivo. Similar findings were reported in a population study of Japanese and non-Latino White subjects using warfarin (Takahashi et al., 2004). However, when differences in S-warfarin CYP2C9 clearance were compared between Japanese and non-Latino White patients having reference alleles in both promoter (up to position -2100) and coding regions, the Japanese patients exhibited significantly greater intrinsic clearance than the non-Latino White patients. These data not only suggest the likelihood of significant interpopulation genetic differences, but also the possible presence of additional important CYP2C9 promoter elements and variants upstream of position -2100. Such a conclusion is not only supported by the study of Takahashi et al. (2004), but also has precedence based on our knowledge of other cytochrome P450 genes (Martinez-Jimenez et al., 2005).
Given the data supporting the likely presence of important CYP2C9 regulatory polymorphisms, the evidence for interpopulation differences in both the presence and frequency of genetic variants, and the paucity of pharmacogenetic studies in the Hispanic population, the present study was designed to begin addressing this knowledge gap.
Custom oliogonucleotides were synthesized by MWG Biotech (High Point, NC). Herculase High Fidelity polymerase and Quick Change site directed mutagenesis kits were obtained from Stratagene (La Jolla, CA). The ExoSap-IT mix and shrimp alkaline phosphatase were purchased from United States Biochemical Corp. (Cleveland, OH). CEQ SNP-Primer Extension and Dye Terminator Cycle Sequencing kits were obtained from Beckman Coulter, Inc. (Fullerton, CA). Restriction endonucleases were purchased from New England Biolabs (Beverly, MA). For SNP discovery, the 24 sample subset of the Polymorphism Discovery Resource (24PDR) was obtained from the Coriell Institute (Camden, NJ). The luciferase reporter plasmid, pGL3Basic, and luciferase reporter assay kit were purchased from Promega (Madison, WI) whereas the luminescent β-galactosidase assay kit was obtained from BD Biosciences (Palo Alto, CA). High purity plasmid purification kits and the QIAmp DNA blood Midi kit were supplied by QIAGEN (Valencia, CA). Cell culture medium, fetal bovine serum, DMSO and rifampicin were purchased from Sigma-Aldrich (St. Louis, MO). Lipofectamine 2000, Opti-MEM reduced serum medium, Taq polymerase and the TA cloning kit were purchased from Invitrogen (Carlsbad, CA). The HepG2 human hepatoblastoma cell line was a gift from Dr. Barbara Knowles (Jackson Laboratories, Bar Harbor, ME). The HNF-4α expression plasmid, pCMVHNF4α (Stoffel and Duncan, 1997), was provided by Dr. Stephen A. Duncan (Medical College of Wisconsin, Milwaukee, WI). The HNF-1α expression plasmid, pBJ5HNF1α, (Kuo et al., 1990), and dimerization cofactor of HNF-1α (DCoH), pBJ5DCoH (Mendel et al., 1991), were generous gifts from Dr. Gerald R. Crabtree (Stanford University School of Medicine, Stanford, CA). The expression vectors for human PXR, pSG5hPXR (Lehmann et al., 1998), and human CAR, pCDM8hCAR (Baes et al., 1994), were kindly provided by Drs. Stephen A. Kliewer (Glaxo Wellcome Research and Development, Research Triangle Park, NC) and David D. Moore (Baylor College of Medicine, Houston, TX), respectively.
After consent, blood samples were collected from 193 women of self-reported Hispanic background who traced their Mexican ancestry back a minimum of two generations. Volunteers were recruited after admission to the labor and delivery units of Provena Saint Therese Medical Center or Victory Memorial Hospital, Waukegan, IL. DNA was extracted using the QIAmp DNA blood MIDI kit and stored at 4°C. This research protocol was approved by all involved Institutional Review Boards.
All PCR DNA amplification primers were designed using OLIGO V.6.45 (Molecular Biology Insights, Cascade, CO). Each primer pair was designed to have similar melting temperatures in order to facilitate high throughput processing using a 96-well format. Primer sequences are provided in supplemental data (see supplement Table S1).
Common CYP2C9 genetic variants were identified by sequencing the approximate first 10 kbp of CYP2C9 5′ flanking sequence in each sample from the Coriell 24PDR. Templates were prepared by PCR DNA amplification using 20 to 25 ng genomic DNA in a 20 μL reaction volume containing 0.2 mM each deoxyribonucleotide triphosphate, 0.5 μM each primer, and 1.25 units of Herculase High Fidelity polymerase (Stratagene). Cycles were as follows: denaturation at 92°C for 40 sec and annealing and extension at 56°C to 62°C for 1 min for a total of 30 cycles. All amplification reactions included an initial 10 sec hold at 94°C and a final 5 min hold at 75°C. After amplification, 4 μL of ExoSap-IT (United States Biochemical Corp.) was added and the reaction was incubated at 37°C for 30 min to remove unincorporated deoxy- and dideoxyribonucleotide triphosphates and primers.
Sequence analysis was performed using 60 to 100 fmols of amplicon with 8 μL Quickstart Sequencing reagent (Beckman Coulter) in a MJ Research PTC-225 Peltier thermal cycler per manufacturer’s recommendations. Analysis was performed on both DNA strands. SNPs identified on a single allele were verified by a repeat analysis of an independently generated amplicon. In all instances, the coordinates of the identified SNPs follow the recommendations of the HUGO nomenclature working group, wherein the “A” of the ATG start codon is assigned +1 and uses contig NT_030059.12, build 36.1 as a reference. For purposes of clarity, the coordinates of all CYP2C9 DNA fragments also follow the same convention. Immediately following the sequencing reaction, unincorporated dideoxyribonucleotide triphosphates and primer were removed using CleanSeq magnetic beads and reagents (Agencourt) prior to analysis of samples by capillary electrophoresis in a CEQ8000 Genetic Analysis System (Beckman Coulter).
After reviewing all of the identified CYP2C9 SNPs, there were five unique clusters and eight remaining SNPs that fell outside of the clusters. Five PCR primer pairs were designed to amplify the DNA sequences containing the SNP clusters, resulting in amplicons ranging in size from 269 bp to 853 bp (see supplemental Table S1). Two CYP2C9 fragments containing the remaining eight sequence variants were amplified, resulting in amplicons of 1.6 kbp and 2.3 kbp. DNA amplifications were performed as described above using 20 to 25 ng genomic DNA as template. DNA amplification of sequences containing exons 3, 5, and 7 were generated for use as templates for genotyping of the previously identified SNP tags for the CYP2C9*2, CYP2C9*3, CYP2C9*4, CYP2C9*5, and CYP2C9*6 alleles (reviewed in Lee et al., 2002) (see supplemental Table S2).
The frequencies of previously identified CYP2C9 promoter variants (Veenstra et al., 2005), as well as any new variants identified within the discovery phase of this project, were determined within the Hispanic population using a combination of DNA sequencing (all SNPs that fell within the five unique cluster sets) and multiplexed single base extension (SBE) reactions (Lindblad-Toh et al., 2000) (eight remaining SNPs, as well as the SNP tags for the CYP2C9*2, CYP2C9*3, CYP2C9*4, CYP2C9*5, and CYP2C9*6 alleles) (reviewed in Lee et al., 2002). DNA sequencing was performed as described above. For genotyping by SBE, two muliplex primer sets were designed, consisting of four primers in each set, using OLIGO V. 6.45. The CYP2C9 -1537G>A variant was not identified in the initial SNP discovery. However, because of the previously reported linkage with the CYP2C9*3 haplotype (Veenstra et al., 2005), the frequency of this SNP subsequently was determined in the Hispanic study population using SBE. SBE reactions were performed in a 20 μL reaction volume consisting of a mixture of all four dye-labeled dideoxynucleotide terminators and the supplied proprietary polymerase, (Beckman Coulter), 20 fmols template, and approximately 1 to 10 pmols of each primer. Individual primer concentrations were optimized to ensure adequate signal intensities for each primer within the multiplex. SBE reactions were as follows: denaturation at 96°C for 10 sec, annealing at 50°C for 5 sec and extension at 72°C for 30 sec for a total of 25 cycles. Unincorporated dideoxynucleotides were eliminated by adding one unit of shrimp alkaline phosphatase (United States Biochemical Corp.) and incubating for 30 min at 37ºC. A 0.5 μL aliquot of the SBE reaction and 0.5 μL of SBE size standard 80 (Beckman Coulter) were mixed and analyzed by capillary electrophoresis on a CEQ8000 Genetic Analysis System (Beckman Coulter). As a quality control measure, 10% of the DNA samples were random selected and independently analyzed to confirm the original genotyping calls.
A CYP2C9 genomic fragment containing positions −1,515 to +184 was amplified from a single Hispanic DNA sample that, based on sequence analysis, was homozygous for the CYP2C9*1 allele. The amplification product was digested with the restriction enzymes HindIII (CYP2C9 position −1,454) and SacI (CYP2C9 position 66) and subsequently cloned into these same sites within the multiple cloning site of the pBluescript II KS+ vector (Stratagene). Site directed mutagenesis was performed to convert the adenosine residue at CYP2C9 position −1 to cytosine, thereby creating an NcoI site at the translation start codon. Following mutagenesis, the complete CYP2C9 insert was sequenced using M13 forward, M13 reverse and internal primers to verify the identity of the clone and validate its fidelity relative to the reference sequence. The mutagenized amplicon was excised from the pBluescript II KS+ vector using the restriction enzymes NcoI and HindIII, and cloned into these same sites within the pGL3Basic vector, generating pRNH905.
Due to persistent amplification difficulties, an alternate approach was used to isolate a CYP2C9 genomic fragment containing positions −5,910 to −1,325. A human BAC clone containing the region of interest (CTD2343I24, chromosome 10, position 96586284 to position 96697248) was purchased from Open Biosystem’s Clone Resource. The BAC clone DNA was isolated using a midi prep kit designed for large constructs (Qiagen), and then digested using EciI. The 16,589 bp fragment, representing CYP2C9 position −7,870 to +8,719, was gel purified and digested with HindIII and XmnI and subsequently cloned into the HindIII and SmaI sites of pRNH905, generating pRNH923.
A similar PCR approach as that described above for the proximal promoter fragment was used to amplify a CYP2C9 genomic fragment from position −10108 to −5517. The resulting amplicon was treated with Taq polymerase (Invitrogen) in order to add a single deoxyadenosine (A) to the 3′ end of the PCR product and subsequently cloned into pCR2.1 (Invitrogen). The relative orientation of the insert was checked by digestion with Bst1107I and SacI. A clone yielding product sizes of 4509 bp and 3990 bp was selected and the 4509 bp fragment was subsequently cloned into these same sites in pRNH923, generating pRNH924. Thus, this final construct contained CYP2C9 sequences from position −10,008 to −1 (chromosomal coordinates 96678320 to 96688429) directing the expression of the luciferase gene.
Three CYP2C9 fragments were excised from pRNH924 and cloned into the pBluescript II KS+ vector to perform site directed mutagenesis and introduce the various haplotype sequences into the CYP2C9/luciferease reporter vector. pRNH971 was constructed by digesting pRNH924 with HindIII and SalI and the resulting 3,376 bp fragment containing the luciferease cassette and CYP2C9 position −1 to −1453 was cloned into the HindIII and SalI sites of the pBluescript II KS+ vector. pRNH972 was constructed by digesting pRNH924 with HindIII and XbaI and the resulting 4,379 bp fragment cloned into these same sites of pBluescript II KS+. The third mutagenesis plasmid, containing CYP2C9 sequences from position −5,833 to −10,008, was constructed by digesting pRNH924 with XbaI and SacI and cloning the resulting 4,321 bp fragment into these same sites in pBluescript II KS+, generating pRNH973. After DNA sequence analysis of each plasmid, multiple rounds of site-directed mutagenesis were performed to introduce all combinations of SNPs present in each of the nine inferred CYP2C9 variant haplotypes, as well as the reference. The individual fragments were then re-assembled within the pGL3Basic backbone, thus generating pRNH954 (haplotype 1), pRNH955 (haplotype 1A), pRNH974 (haplotype 1B), pRNH956 (haplotype 1C), pRNH957 (haplotype 1D), pRNH958 (haplotype 1E), pRNH959 (haplotype 2A), pRNH960 (haplotype 2B), pRNH962 (haplotype 3A), and pRNH961 (haplotype 3B).
HepG2 human hepatoma cells were cultured in Eagle’s minimal essential media supplemented with 10% fetal bovine serum, penicillin (50 U/mL) and streptomycin (50 μg/mL), and maintained in a humidified incubator at 37°C in an atmosphere of 5% CO2. Cultures were never allowed to grow beyond approximately 80% confluence. All experiments were performed with cells between passage 4 and 15. For transfection studies, 1.5 × 105 cells were seeded into 24-well culture dishes and 24 hours later, were transfected with 1.4 μg of Lipofectamine 2000, 0.5 μg of test luciferase reporter plasmid, and 50 ng each of pCMVβgal, pCMVHNF-4α, pBJ5HNF-1α and pBJ5DCoH in Opti-MEM reduced serum medium. After incubation for 24 hrs at 37ºC, transfection medium was replaced with normal growth medium, and cells were incubated for an additional 24 hrs. For induction experiments, transfection medium was replaced with Opti-MEM reduced serum medium supplemented with 10 μM rifampicin or vehicle (0.1% DMSO). The rifampicin concentration and the time of harvest post-treatment were based on results from initial optimization experiments in which rifampicin concentrations of 5 μM, 10 μM, and 25 μM were added for 10, 16, and 24 hrs (data not shown). Cell lysates were prepared and luciferase assays performed according to the manufacture’s instructions. Data were normalized with respect to β-galactosidase activity to correct for transfection efficiency and expressed as relative luciferase activity. Data are reported as the mean ± standard deviation (SD) of at least three determinations performed with two independently prepared luciferase reporter plasmids.
DNA sequence data was analyzed using DNAStar software (LaserGene, Madison, WI); minimal acceptable quality scores values were set at ≥ 12. Sequences were scanned for potential transcription factor recognition sequences using the Match Program and the TRANSFAC Professional V11.3 database (BIOBASE Corporation, Wolfenbuettel, Germany). Both the Liver-Specific and Vertebrate Non-Redundant Matrix Profiles were used with search criteria that minimized the identification of false positives. All SNPs identified were tested for deviations from Hardy-Weinberg equilibrium with the use of a chi-square test. The frequencies of individual SNPs were compared using Fisher’s exact test (GraphPad InStat V 3.05, San Diego, CA). Based upon the observed frequencies of each SNP, haplotype analysis was inferred using PHASE V 2.1, with all parameters set at default values except that 5000 iterations were performed with a thinning value of one and a burn in of 1000 (Stephens and Donnelly, 2003). Functional differences among the different haplotypes were assessed using transient expression assays and compared by one-way ANOVA with a Holm-Sidak post hoc test for multiple comparisons (SigmaStat V3.11, Systat Software, Chicago, IL). Inferred haplotype frequencies between non-Latino Whites and Hispanics of Mexican descent were compared using a Student’s t-test (SigmaStat V3.11, Systat Software, Chicago, IL). In all instances, an α value of 0.05 was accepted as significant.
SNP discovery was accomplished by sequencing overlapping amplicons spanning approximate 10,000 bp upstream of the CYP2C9 transcription start site using 24 DNA samples obtained from the Coriell Polymorphism Discovery Resource (Camden, NJ). Thirty-one SNPs were identified, nine novel and 22 previously reported in the literature and/or on the human CYP2C9 allele nomenclature website (http://www.cypalleles.ki.se/) (Table 1). To determine if any of the novel or previously identified SNPs were located at or near putative transcription factor binding sites, a search within the 10,000 bp of CYP2C9 upstream sequence was performed using the Match Program along with the TRANSFAC Professional V11.3 database (BioBase Biological Databases) using both the Liver-Specific and Vertebrate Non-Redundant Matrix Profiles. Comparisons also were made against previous reports on functional CYP2C9 regulatory elements (Ibeanu and Goldstein, 1995; Chen et al., 2005; Kawashima et al., 2006) (Table 1). Only sites whose core or matrix match score was impacted, or which were eliminated or created by variant sequences were considered. None of the identified SNPs were located within any previously identified regulatory elements and only three SNPs were located in putative transcription factor binding sites. The g.-8416T>G SNP falls within a key residue of a putative glucocorticoid receptor IR3 element and would be predicted to eliminate or reduce binding at this site while the g.-3360T>C SNP falls immediately 5’ to a putative core NF1/CTF site, reducing the matrix match score. The g.-3597A>G SNP creates a putative new CREB binding site.
The PDR24 is an anonymous and blinded DNA panel from individuals representative of the United States population, but provides no insight into the presence or frequency of specific genetic variants within defined ethnic or racial groups. Thus, further characterization of the identified SNPs was needed in the Hispanic population. To define CYP2C9 promoter region haplotypes, allelic frequencies of the 31 SNPs identified during the discovery phase, the g.-1537G>A SNP previously shown to be linked to the CYP2C9*3 allele in non-Latino Whites (Veenstra et al., 2005), as well as the SNPs defining the CYP2C9*2, CYP2C9*3, CYP2C9*4, CYP2C9*5 and CYP2C9*6 alleles were determined in 193 Hispanic Americans of Mexican descent by either sequence analysis (for SNP clusters) or multiplex SBE. All SNPs identified were in Hardy-Weinberg equilibrium except g.-1188T>C (χ2 = 5.37) and g.-4302C>T (χ2 = 6.19), both genotyped using multiplex SBE. To eliminate possible assay error for these two positions, all DNA samples with variant alleles at g.-1188T>C or g-4302C>T were resequenced and in all instances, the original genotype call was confirmed. Five of the nine novel SNPs identified in the 24PDR were observed in the Hispanic population (g.-8422A>G, g.-5146G>C, g.-5143A>C, g.5140A>T and g.-4302C>T) (Table 2). The g.-4302C>T SNP was observed at a frequency of 10.9% whereas all other novel SNPs were observed at frequencies <0.5%. The most common promoter variant in the Hispanic population was g.-1188T>C at an observed frequency of 20.7%, whereas the g.-8422A>G, g.-8416T>G and g.-7336G>A SNPs were the least commonly observed variants at frequencies <1%. Two previously reported CYP2C9 structural variants, g.3608C>T (CYP2C9*2) and g.42614A>C (CYP2C9*3) (Rettie et al., 1994; Haining et al., 1996) also were observed in the Hispanic study population at 7.0% and 4.4%, respectively (Table 3). Other previously reported CYP2C9 exon variants, g.42615T>C (CYP2C9*4), g.42619C>G (CYP2C9*5) and g.10601delA (CYP2C9*6) (Imai et al., 2000; Dickmann et al., 2001; Kidd et al., 2001) were not observed.
The CYP2C9 promoter SNP frequencies observed in the Hispanic population were compared to the previously reported frequencies in a non-Latino White cohort (Veenstra et al., 2005) (Table 2). Five upstream genetic variants (g.-8422A>G, g.-5146G>C, g-5143A>C, g.-5140A>T and g.-4302C>T) observed in the Hispanic population were absent in the non-Latino White population, one of which, g.-4302C>T, was not found due to lack of ascertainment of that region in the later population (see GenBank accession AY702706, - "region not scanned" due to repetitive sequences). A total of 21 SNPs were observed in both ethnic groups, five of which occurred at different frequencies. For instance, the g.-620G>T, g.-1096A>G, g.-1188T>C, g.-2663delTG and g.-3089G>A SNPs were observed in the Hispanic population at frequencies of 4.9%, 7.0%, 20.7%, 7.3% and 10.9%, respectively, compared to 11.0%, 12.0%, 35.0%, 17.0% and 17.0%, respectively, in the non-Latino White population (p<0.05).
Using the determined minor allelic frequencies of upstream and structural variants, a total of 34 CYP2C9 haplotypes were inferred using PHASE V 2.1, ten of which occurred at a frequency greater than 1%. This latter group was compared to the previously reported haplotypes in a non-Latino White study group (Table 4) (Veenstra et al., 2005). Importantly, for the purposes of comparison, the SNPs previously reported in the non-Latino White cohort were re-analyzed, but only including the promoter and *2 and *3 SNPs. Haplotype 1 was designated as the reference sequence and was inferred to occur at a frequency not significantly different from that reported in the non-Latino White population (Table 4). Four CYP2C9 haplotypes were deduced in the Hispanic population, but not in the non-Latino White population. Haplotype 1A (g.-8553C>A) was present at 3.2%, 1B (g.-4302C>T) at 10.0%, 1E (g.-1188T>C) at 1.3% and 2A (g.-3597A>G, g-3360T>C, g.-1188T>C, g.-1096A>G, g.-485T>A, g.-484C>A, g.3608C>T) at 2.1%. Five variant CYP2C9 haplotypes inferred in both the Hispanic and non-Latino White populations were present at different frequencies. For example, haplotype 1C (g.-3089G>A, g.-1188T>C) was deduced in the Hispanic population at a higher frequency than that observed in the non-Latino White population, i.e., 3.1% compared to 0.27%, respectively. In contrast, haplotype 1D was inferred to occur more frequently in the non-Latino White population, 17.2%, compared to the Hispanic population, 6.6%.
In addition to the five haplotypes (1A, 1B, 1C, 1D, and 1E) comprised solely of SNPs within the CYP2C9 regulatory region, four inferred haplotypes (2A, 2B, 3A, and 3B) consisted of upstream SNPs in linkage disequilibrium with the previously described g.3608C>T (CYP2C9*2) and g.42614A>C (CYP2C9*3) structural allelic variants (Table 4). Of these four haplotype variants, haplotype 2A (g.-3579A>G, g.-3360T>C, g.-1188T>C, g.-1096A>G, g.-485T>A, g.-484C>A, and g.3608C>T) appeared to be unique to the Hispanic population. Haplotypes 2B, 3A, and 3B were present at different frequencies in the two population groups (Table 4).
To explore the possible functional significance of the identified CYP2C9 variant haplotypes on constitutive promoter activity, we used site-directed mutagenesis to introduce nine common CYP2C9 haplotype variants into a reporter construct containing 10,008 bp of CYP2C9 upstream information driving the expression of the luciferase reporter gene. Negligible luciferase activity was observed when the CYP2C9/luciferase construct was transiently expressed alone (data not shown). Given the important roles HNF1α, and HNF4α (Chen et al., 2005; Kawashima et al., 2006) have in regulating CYP2C9 promoter activity, the initial experiments were repeated co-expressing both of these factors along with the CYP2C9/luciferase reporter construct. A substantial increase in promoter activity was observed in the presence of both HNF1αand HNF4α, but either factor alone had a minimal effect. As such, all subsequent transient expression experiments included expression vectors for both of these factors in the protocol. Of the nine variant haplotypes, 1B, 3A, and 3B exhibited 2.5-, 3.2-, and 2.6-fold decreased CYP2C9 promoter activity compared to the reference construct (Figure 1). Comparing the SNPs constituting each of these hypomorphic promoter variants against the variants located within putative transcription factor binding sites (Table 1), only the g.-3360T>C SNP is located immediately 5′ to a putative NF1/CTF core element (position -1). The T>C transition reduces the matrix match score from 0.747 to 0.722 and of the eight binding sites selected for defining this matrix, five had a T at position -1 and none had a C.
In addition to the contribution of genetic polymorphisms to interindividual variability in CYP2C9 activity (Schwarz, 2003; Lee et al., 2002), induction by exogenous agents also can contribute to observed differences (Williamson et al., 1998; Niemi et al., 2001). To determine whether any of the identified CYP2C9 variant haplotypes alter CYP2C9 induction by rifampicin, HepG2 cells were co-transfected with pSG5hPXR, a human PXR expression plasmid, and the various CYP2C9 reporter constructs and treated with 0.1% DMSO (vehicle control) or 10 μM rifampicin for 24 hrs. No effect was observed when HepG2 cells were co-transfected with pSG5hPXR and pRNH954 (reference haplotype) and treated with 0.1% DMSO (Figure 2A). HepG2 cells co-transfected with pSG5hPXR and pRNH954 (reference haplotype) and then treated with 10 μM rifampicin for 24 hrs, exhibited a 3.2-fold induction of luciferase activity (Figure 2A). However, transfection with pRNH957 (haplotype 1D) resulted in only a 1.6-fold induction (49% decrease relative to the reference haplotype), whereas pRNH961 (haplotype 3B) resulted in a 4.6-fold induction (44% increase relative to the reference haplotype) (Figure 2B). No differences in induction were observed with any of the other plasmids representing the other haplotypes.
To determine the effect of promoter haplotypes on phenobarbital-dependent CYP2C9 induction, HepG2 cells were co-transfected with pCMVhCAR, a human CAR expression plasmid, and the variant CYP2C9 reporter constructs. No effect was observed with the CAR expression plasmid alone and no induction was observed with the reference or any of the variant haplotypes following treatment with 0.5 mM phenobarbital (data not shown). In contrast, a 2-fold induction was observed in HepG2 cells co-transfected with pGS5hPXR following treatment with 0.5 mM Phenobarbital. Similar differences in induction as those observed with human PXR and rifampicin were seen with the pRNH957 (haplotype 1D) and pRNH961 (haplotype 3B) constructs (data not shown).
Relative to other large population groups, the contribution of genetic polymorphisms to interindividual differences in CYP2C9 expression in Hispanics is poorly understood. In the current study, five of the nine novel SNPs identified in the 24PDR (g.-8422A>G, g.-5146G>C, g.-5143A>C, g.-5140A>T, and g.-4302C>T) were present in the Hispanics but were absent in the non-Latino Whites. A total of 22 SNPs were present in both populations, five of which (g.-620G>T, g.-1096A>G, g.-1188T>C, g.-2663delTG and g.-3089G>A) occurred at significantly different frequencies (Table 2). Thus, similar to what has been observed for CYP2C9 structural variants, ethnic differences exist in the presence and frequency of CYP2C9 regulatory polymorphisms, which may contribute to interpopulation differences in CYP2C9-dependent metabolism.
The observed frequency of two previously reported CYP2C9 structural variants, g.3608C>T (CYP2C9*2) and g.42614A>C (CYP2C9*3) were not significantly different between the Hispanics and non-Latino Whites, nor were the determined frequencies different than those reported by Llerena et al. (2004) for Hispanic Americans of Mexican descent. In contrast, there was a significant difference when compared to the frequencies reported by Xie et al. (2002). However, the ancestral background of the Hispanic population was not defined in the latter study. Thus, the discrepancy in these data is likely explained by a population of mixed ancestry.
Based upon the observed minor allele frequencies of 32 upstream and five structural variants, 10 common (>1%) CYP2C9 haplotypes were inferred. The observed haplotype 1 (reference) frequency in Hispanics was 60.53%, compared to 63.64% in non-Latino Whites, consistent with a relatively simple haplotype structure. Four CYP2C9 haplotypes (1A, 1B, 1E, and 2A) were inferred in Hispanics, but not in non-Latino Whites (Table 2).
The possible contribution of upstream CYP2C9 genetic variability to interindividual pharmacokinetic differences is more controversial. Variant effects on promoter activity in vitro have been observed, but failed to correlate with differences in in vivo activity (Shintani et al., 2001). Other studies have failed to show an independent effect of CYP2C9 upstream haplotype variants on mean warfarin clearance or dose (King et al., 2004; Takahashi et al., 2004; Veenstra et al., 2005). However, the design of these studies precluded the ability to test the impact of the CYP2C9 upstream polymorphisms independently of the variants impacting CYP2C9 catalytic activity. In the current report, haplotype 1B, consisting of only the g.-4302C>T variant, exhibited a significant decrease in promoter activity in vitro. Combined with its relatively high frequency (i.e., 10.0%), these data would be consistent with reduced constitutive CYP2C9 expression in this population. It is interesting that the frequency of the g.-4302C>T SNP was not in Hardy-Weinberg equilibrium, suggesting a possible selective pressure for the presence of the variant allele, although this deviation may also be due to recent population admixture. Significant decreases in in vitro constitutive promoter activity also were observed with haplotypes 3A and 3B. Thus, the results of our study suggest that, in addition to the defective enzymatic function of CYP2C9.3, a decrease in basal CYP2C9 transcription also may contribute to the overall observed CYP2C9*3 phenotype.
In vitro studies using primary hepatocytes have shown that CYP2C9 mRNA, protein, and catalytic activity are all increased by drugs such as rifampicin, hyperforin, and phenobarbital through a PXR-dependent mechanism (Chen et al., 2004). Rifampicin treatment also has been reported to enhance the clearance of CYP2C9 substrates, indicative of CYP2C9 induction in vivo (Williamson et al., 1998; Niemi et al., 2001). Co-transfection studies in HepG2 cells with a human PXR expression plasmid and CYP2C9 reporter constructs were used to investigate the possible functional effects of CYP2C9 haplotype variants on rifampicin-dependent enhancement of CYP2C9 promoter activity. With haplotype 1 (reference construct), addition of human PXR alone had no effect on basal CYP2C9 promoter activity, consistent with results reported by Ferguson et al. (2002). These results, however, conflict with those of Chen et al. (2004), who reported a 3.8-fold increase in CYP2C9 promoter activity upon addition of human PXR alone. When the co-transfected cells were treated with 10 μM rifampicin, a 3.2-fold increase in promoter activity was observed, similar to the approximate 3-fold increase reported by Chen et al. (2004). Yet Ferguson et al. (2002) failed to see an effect with rifampicin. The discrepancies in these data may result from differences in the amount of CYP2C9 upstream sequences present in the reporter construct or differences in transfection conditions. When co-transfection studies were performed with the CYP2C9 promoter variants, two of the inferred CYP2C9 haplotypes exhibited an altered induction profile; haplotype 1D resulted in a 1.6-fold reduction in inducibility, whereas haplotype 3B resulted in a 1.4-fold increase in inducibility. Given the frequency of these two haplotypes (6.6% and 2.4%, respectively) and the magnitude of this observed effect, it is possible these variants contribute to observed interindividual differences in CYP2C9 phenotype.
Previous studies have identified two constitutive androstane receptor (CAR)-responsive elements within the CYP2C9 promoter at positions -2,898 and -1,839 that bind human CAR in vitro and transactivate reporter constructs (Ferguson et al., 2002). Further, Gerbal-Chaloin et al. (2001) described a 4-fold increase in CYP2C9 mRNA following treatment with phenobarbital, a known human CAR ligand. CAR-mediated phenobarbital induction of CYP2C9 promoter activity was not observed in the current study, although a two-fold, PXR-mediated phenobarbital enhancement of CYP2C9 promoter activity was observed. This latter observation supports the recent findings of Chen et al. (2004) who determined that human PXR is responsible for the induction of human CYP2C9 by both rifampicin and phenobarbital.
To gain some insight into which SNP, or combination of SNPs, might be responsible for the observed altered activities, a careful comparison between all observed haplotypes was made. Reduced constitutive activity was observed with haplotypes 1B, 3A and 3B. As indicated earlier, the g.-4302C>T SNP is unique to haplotype 1B and as such, is assumed to be causative, yet does not alter any known or putative transcription factor binding sites. Several SNPs are found in common between haplotypes 3A and 3B, but not in other haplotypes (g.-8897C>A, g.-7419A>G, g.-5813A>G, g.-5661C>A, g.-4877G>A, -1911T>C, -1885C>G, g.-1537G>A, and g.-981G>A). However, similar to the haplotype 1B g.-4302C>T SNP, none of these variants alter known or putative regulatory elements. Altered PXR-mediated inducibility was observed with both haplotypes 1D and 3B. In the case of haplotype 1D (g.-3089G>A, g.-2663delTG, g.-1188T>C), two other inferred haplotypes (1C and 1E) also contain g.-3089G>A and/or g.-1188T>C but exhibit promoter activities no different from the reference construct. The remaining SNP in haplotype 1D, g.-2663delTG, is not present in any other inferred haplotype and as such, may be causative for the reduced inducibility of this promoter variant. Haplotype 3B resulted in an increase in rifampicin-induced CYP2C9 promoter activity, whereas haplotype 3A exhibited no difference compared to the reference control. Yet, the only difference between these two haplotypes is the presence of the g.-1188T>C SNP in haplotype 3B. This observation suggests that the g.-1188T>C transition might be responsible for the observed difference in induction profiles. However, two other haplotypes (1C and 1E) also contain g.-1188T>C, yet exhibit no difference relative to the control construct. Further, a recent study by Sandberg et al. (2004) found that the g.-1188T>C variant did not affect gene expression in vitro. Finally, haplotype 1D also contains the g.-1188T>C SNP, and, in contrast to haplotype 3B, was associated with decreased induction of CYP2C9 promoter activity. These findings suggest that g.-1188T>C cannot solely be responsible for the observed increase in activity, but rather contributes to a combinatorial effect. The simplest explanation as to how individual or clusters of SNPs might function to alter gene regulation is by modifying or eliminating transcription factor binding. However, this does not appear to be the case for the variants identified herein and is clearly not the case for many regulatory polymorphisms. A survey of 247 known promoters and 647 haplotype variants by Buckland et al. (2005) revealed that only 35% of the functional regulatory variants identified were localized within predicted transcription factor binding sites. Thus, many regulatory polymorphisms may impact gene expression through other yet to be identified sequence-specific mechanisms.
In summary, this is the first extensive study of CYP2C9 haplotype and in vitro functional analysis in a Hispanic population of Mexican descent. Interindividual and interethnic differences were observed in the incidence and frequency of CYP2C9 regulatory polymorphisms. Further, based on in vitro assays, several of the inferred haplotypes are predicted to significantly decrease and/or alter PXR-mediated rifampicin-dependent CYP2C9 induction. These observations suggest that genetic variation within CYP2C9 regulatory sequences likely contributes to differences in CYP2C9 phenotype both within and among different populations. However, in vivo studies will be required to determine the relevance of these haplotype variants, if any, to clinical outcomes.
The authors would like to acknowledge the expert assistance of Min Le, Ph.D. and Sevasti B. Koukouritaki, Ph.D. in completing these studies.
*This work was supported in part by National Institutes of Health Grant GM068797 and funds from the Children Research Institute, Children's Hospital and Health Systems.