A key event in human population history is the dispersal of early humans from Africa to other parts of the world with different climates, pathogens, sources of food and xenobiotic exposure, and the dynamic adaptation to the evolving environments (Jin and Su 2000). Current phenotypic differences between individuals/groups could be due partly to functional polymorphisms that facilitated survival in the ancestral populations (Di Rienzo and Hudson 2005). Resolving the underlying allelic architectures of environmental response and searching for their molecular adaptation to selective forces have been attractive indirect strategies for implementation and interpretation of genetic analysis of environmental response and complex disorders (Bamshad and Wooding 2003; Sabeti et al. 2007).

Members of the cytochrome P450 (CYP) 3A family are among the most important CYP enzymes in humans. They metabolize various endogenous and exogenous chemicals, such as clinically important drugs, environmental carcinogens, cholesterol, steroids, and other lipids (Wojnowski 2004). Members of the CYP3A family are expressed in organs primarily associated with xenobiotic and hormone disposition, such as the liver and gastrointestinal tract (Wojnowski 2004). Four functional CYP3A enzymes—CYP3A4, CYP3A5, CYP3A7, and CYP3A43—have been identified in humans. CYP3A4 is most abundant in adult liver and intestine and is the major enzyme involved in xenobiotic and drug metabolism (Fujita 2004). CYP3A5 is the predominant form in the kidney (Givens et al. 2003). CYP3A7, a fetoplacental enzyme found only in humans, is expressed mainly in fetal liver (Leeder et al. 2005) and extrahepatic tissues such as endometrium and placenta (Burk et al. 2002; Schuetz et al. 1993). The more recently cloned CYP3A43 is expressed predominantly in prostate and testis (Gellner et al. 2001). Interindividual variation in CYP3A activity up to 10-fold has been observed (Dorne et al. 2003). Because CYP3A activity is a major determinant of drug response and may be associated with risk for cancers such as breast and prostate cancer (Keshava et al. 2004), the identification of sequence variants at the CYP3A locus and functional characterization of their clinical relevance have been of long-standing interest in pharmacogenetics and toxicogenetics.

The human CYP3A gene cluster resides in a 231-kb region on chromosome 7q22 and consists of four genes and two pseudogenes, arranged in the order of CYP3A5, CYP3A5P1, CYP3A7, CYP3A5P2, CYP3A4, and CYP3A43, from centromere to telo mere (Finta and Zaphiropoulos 2000) (Figure 1). CYP3A43 is in the opposite orientation from all other CYP3A genes. Each intact gene encodes a protein consisting of 503 amino acids. Finta and Zaphiropoulos (2000) suggested that the locus arose through duplication of an ancestral CYP3A cassette of 40–50 kb. Many genetic variants have been identified in this locus and are available online ( http://www.imm.ki.se/CYPalleles/); some functional variants for CYP3A have been reported, such as the CYP3A5*3, which results in an incorrectly spliced mRNA and a truncated nonfunctional protein (Kuehl et al. 2001).

Parallel to exploring phenotypic effects and clinical relevance of CYP3A genetic polymorphisms, investigating molecular adaptations to the environment is an intriguing complementary strategy for pharmacogenetic and toxicogenetic studies. The CYP3A4 and CYP3A5 genes have a strong haplotype structure at varying frequencies across ethnic groups (Thompson et al. 2004). Using a comparative genomics approach and sequence-based neutrality test, Thompson et al. (2004) reported evidence of positive selection on the derived allele of the functional CYP3A5*3 in non-African populations and significant correlation of its allelic frequency with distance from the equator. In a genotype-based investigation on the CYP3A locus, Schirmer et al. (2006) proposed that negative natural selection acted primarily toward the elimination in non-African populations of the ancestral CYP3A4*1B allele rather than toward the reduction of CYP3A5 expression described by Thompson et al. (2004). It should be noted that these population genetics studies were based on a small fraction of sequence information or on only limited data of genotypes of single-nucleotide polymorphisms (SNPs), and both had limited coverage of the 230-kb genomic sequence of CYP3A locus. Therefore, the refined molecular targets of natural selection on the complete CYP3A locus and its evolutionary dynamics with respect to geographically and temporally fluctuating environments in our demographic history have not yet been explicitly pinpointed.

To further identify genetic variations in the human CYP3A locus and assess the effect of natural selection on the pattern of nucleotide diversity at this locus, we resequenced the human CYP3A locus in three populations and addressed evolutionary population genetics on the gene cluster.

Materials and Methods

Subjects and samples

We obtained human genomic DNA samples from two sources: a ) we extracted genomic DNA samples from venous blood for 29 unrelated healthy Chinese men (average age, 20 ± 2 years) who were chosen from the sample collection constructed for the Chinese Human Genome Diversity Project through a coordinated effort of several institutes (Chu et al. 1998); and b) we used DNA samples of 24 Caucasians (11 females, 13 males), 24 Africans (9 females, 15 males), and three apes [i.e., one common chimpanzee (Pan troglodytes), one lowland gorilla (Gorilla gorilla), and one orangutan (Pongo pygmaeus)] provided by Coriel Cell Repositories (Camden, NJ, USA). Written informed consent was obtained from all the Chinese subjects. This study was performed with the approval of the Ethical Committee of Chinese National Human Genome Center at Beijing.

Sequencing strategy

A schematic representation of genes and pseudogenes at the CYP3A locus and the resequenced regions is shown in Figure 1. We screened 54 genomic DNA fragments totaling 43,675 bases. The resequenced regions included nearly all exons (except exon 11 of CYP3A7 because of difficulty in sequencing the highly homologous sequences) and relevant exon–intron boundaries of four genes (CYP3A43, CYP3A4, CYP3A7, and CYP3A5) and two pseudogenes (CYP3AP1 and CYP3AP2) and approximately 2.5-kb promoter regions of the four intact genes. We also resequenced four randomly selected segments equally distributed across the CYP3A43–CYP3A4 gene interval and a fragment in CYP3AP2–CYP3A7 interval, with an average length of 1.1 kb.

Identification of polymorphisms and divergences

We identified polymorphisms by sequencing polymerase chain reaction (PCR) products from both ends. Long-range PCR (L-PCR) amplifications were performed first with primers specific for the target DNA regions to increase specificity [GenBank accession no. NG_000004.2 (National Center for Biotechnology Information 2009)]. DNA sequencing was performed using overlapping primers that covered the overall L-PCR fragments. We confirmed positions and individual genotypes of the variations by resequencing from the opposite strand or from overlapped sequencing. Fixed differences or divergences between humans and apes were inferred by aligning the resequenced sequences of humans with those of the three apes; we considered the alleles that occurred in the apes to be the ancestral alleles for polymorphic sites. Details regarding PCR and sequencing conditions, as well as PCR primers, are available on request. As a measure of quality control, sequence segments of individuals presenting singletons or ambiguous polymorphisms were reamplified and resequenced. We assessed SNP data validity by repeating 10% of the assays. The error rate was relatively low (1.2%).

Data analysis

We performed tests of fitness to the Hardy-Weinberg equilibrium for each polymorphic site using LDA software (Ding et al. 2003). We used the false discovery rate method to correct for multiple testing using Q-VALUE in R (Dabney and Storey 2002) as described by Storey and Tibshirani (2003). Three measures of nucleotide diversity were assessed: average nucleotide diversity (π), a statistic based on the average number of pairwise sequence differences and influenced mostly by intermediate-frequency variants (Tajima 1989); Watterson’s θ_W, which is based on the number of segregating sites and influenced mostly by low-frequency variations, theoretically equal to the neutral mutation parameter 4N_em (Watterson 1975); and θ_H (Fay and Wu 2000), a summary that gives more weight to high frequency–derived alleles. To test whether the frequency spectrum of polymorphisms conformed to the standard neutral model, we calculated the values of four test statistics: Tajima’s D, which considers the difference between π and θ_W (Tajima 1989); Fu and Li’s F* and D*, which compare the number of singletons with the number of nonsingletons (Fu and Li 1993); and Fay and Wu’s H, which compared the high-frequency with intermediate frequency–derived variants (Fay and Wu 2000). Significant values for these test statistics were estimated from 10⁴ coalescent simulations of a Wright-Fisher equilibrium model that conditioned on a constant sample size and the level of polymorphism as the observed data (Kaplan et al. 1991). We also performed the Hudson–Kreitman–Aguade (HKA) test to compare the diversity patterns in the sequences observed in our study and those observed in DMD intron 44 (Nachman and Crowell 2000). All neutrality tests and coalescent simulations were performed using DnaSP software, version 4.0 (Rozas et al. 2003).

We calculated an unbiased small-sample estimator of pairwise populations F_ST (F-statistic/fixation index), a measure of allele frequency difference among populations and an estimate of interpopulation genetic differentiation, as described previously (Weir and Cockerham 1984). A bootstrapping method (1,000 bootstrap samples) was used to test the statistical significance of F_ST in each pairwise comparison between populations. Lewontin’s D′ and r² were applied to evaluate the pairwise linkage disequilibrium (LD) between biallelic polymorphisms. We used Fisher’s exact test to determine the statistical significance of pairwise LD, followed by false discovery rate correction for multiple testing. We used GOLD software (Center for Statistical Genetics 2009) as described by Abecasis and Cookson 2000) to draw the LD maps.

Haplotypes were constructed using the Phase 2.0 program (Stephens and Donnelly 2003). We then partitioned haplotype blocks with HaploBlockFinder (Zhang 2004) as described by Zhang and Jin (2003). The mutational relationships among haplotypes in CYP3A43 and CYP3A5 were shown by constructing minimum-spanning trees (MSTs), with the common chimpanzee as an out-group species. We calculated the networks by reduced median algorithm for CYP3A5 and median jointing algorithm for CYP3A43 implicated in the Network 4.1 package (Fluxus Technology 2008) as described by Bandelt et al. (1999).

Results

Data summary and sequence variations

We sampled 77 unrelated human individuals, including 24 Africans, 24 Europeans, and 29 Asians. The surveyed sequences spanned 43,675 kb and contained almost all of the coding sequences of CYP3A4, CYP3A5, CYP3A7, and CYP3A43. We observed a total of 167 segregating sites (including 165 biallelic SNPs and two biallelic indel polymorphisms) in the human DNA samples [see Supplemental Material, Table 1, available online (doi:10.1289/ehp.0800528.S1 via http://dx.doi.org/)], yielding an average density of one SNP per 262 bp. Of the variants identified, 11 were nonsynonymous, 2 resulted in a splicing defect, 1 was a frame shift mutation, and 8were synonymous.

Sequence divergence between human and chimpanzee was 0.95–0.97% for the total locus in subpopulations, which was comparable with the genome-wide average of 1.24% (Ebersberger et al. 2002). Nucleotide diversity was low in both Chinese and Caucasians: 1.6 × 10⁻⁴ and 1.4 × 10⁻⁴, respectively, for π; 2.8 × 10⁻⁴ and 3.1 × 10⁻⁴, respectively, for θ_W (Table 1). When the F_ST for three-way population comparisons were estimated with polymorphisms site-by-site, 11 F_ST values (ranging from 0.30 to 0.50) fell into the upper 0.05 tail of the empirical genomewide distributions estimated previously (Bowcock et al. 1991, Storz et al. 2004) [see Supplemental Material, Table 1 (doi:10.1289/ehp.0800528. S1)]. Comparison of F_ST for Africans and non-Africans using the empirical genome-wide F_ST distribution constructed with similar sets of pooled non-African and African samples (Fullerton et al. 2002) showed that estimated F_ST values for five variants (T147767C, G164751T, T165611C, A169228G, and A260167G) were extremely high (≥ 0.40; data not shown) and fell above the 95th percentile of the empirical distribution.

Selective neutrality tests

Frequency spectra of the derived alleles for the three populations are shown in Figure 2. Notably, we observed an excess of singletons and nearly fixed SNPs (derived allelic frequencies > 80%). Consistently, Tajima’s D, and Fu and Li’s F* and D* deviated significantly from expectations under neutrality in all three populations examined. When individual populations were considered, D and F* values in both Chinese and Caucasians were significant (p < 0.05), and D* was significant in Chinese (p < 0.05). We also observed a statistically significant H value in Caucasians (p = 0.046). Results of the HKA test indicated that the nucleotide diversity patterns in the entire surveyed sequences were significantly different from those in the neutral DMD intron 44 in all three subpopulations (p < 0.05).

We performed sliding-window analyses for the neutrality tests, with each window containing 20 variations and with a step size of one variation. The values of the neutrality statistics varied widely across the locus in all three populations and in subpopulations (Figure 3). We observed significantly negative D, D*, F*, and H values simultaneously in two regions in all the three populations. The first region covered the 40.5-kb sequence from the CYP3A43–CYP3A4 interval to CYP3A4 intron 6, or between SNPs C122872T and T163355G. The second region covered the 25.7-kb sequence from the CYP3A5P2–CYP3A7 interval to CYP3A7 exon 7, or between SNPs G191479A and A217213G. In addition, significantly negative values for all four test statistics were also observed in the region from CYP3A43 intron 10 to CYP3A43 5′-flank region (between SNPs A73018G and G106490T) in Chinese, and in the region from CYP3A5 exon 2 to CYP3A5 3′ untranslated region (between SNPs C256880T and C284792T) in Caucasians.

Linkage disequilibrium

When 24 SNPs shared by all three subpopulations were considered, mean pairwise |D′| values were significantly higher in Chinese (p = 0.001) and Caucasians (p < 0.001) compared with Africans: values were 0.93, 0.86, and 0.74, respectively. The Africans had far fewer SNP pairs with r² ≥ 0.5 compared with either Chinese or Caucasians (6.5%, 25.0%, and 22.8%, respectively). Five SNPs in CYP3A43 (G83269C, C83448T, A83566G, C83448T, and T88726C) were in complete LD with each other in all three populations. These five SNPs were also in significant LD with both the nonsynonymous substitution CYP3A43*3 and the frame shift mutation CYP3A43*2A in non-Africans. The patterns of LD structure in each population (Figure 4), delineated with polymorphisms with minor allelic frequency > 5% in individual populations, also showed remarkable differences in extent and strength of LD across the entire locus between Africans and non-Africans. When the threshold of |D′| value was set at 0.8, we observed five, four, and seven LD blocks in Chinese, Caucasians, and Africans, respectively. The longest blocks spanned 60.5 kb in Chinese (block D), 81.7 kb in Caucasians (block B), and 49.3 kb in Africans (block D) (Figure 4).

Haplotype distribution

Because all four neutrality tests indicated that nucleo tide diversity patterns in the regions between SNPs A73018G and G106490T in CYP3A43 and between SNPs C256880T and C284792T in CYP3A5 deviated significantly from expected under neutrality in non-Africans, we constructed the haplotypes in these regions of CYP3A43 and CYP3A5 and depicted their mutational relationships by constructing MSTs rooted with the chimpanzee [Figure 5; see also Supplemental Material, Figure 1 (doi:10.1289/ehp.0800528.S1)]. The haplotype MSTs showed two clusters of haplotypes for both regions, one main cluster dominated with only one common haplotype (Figure 5B, left) and one minor cluster scattered with haplotypes with low or intermediate frequencies (Figure 5B, right). The haplotype clusters at the CYP3A43 locus were separated by five nucleotides at positions 83269, 83448, 83566, 83643, and 88726, with high frequency–derived alleles at positions 83448 and 83643 (Figure 5B). The haplotype clusters at the CYP3A5 locus were separated by two high frequency–derived alleles at positions 260167 and 284792 [see Supplemental Material, Figure 1 (doi:10.1289/ehp.0800528.S1)]. The haplotype diversity test showed that haplotype diversity in the CYP3A43 region deviated obviously from that expected under neutrality in Chinese and Caucasians or non-Africans, and haplotype diversity in CYP3A5 deviated significantly from that expected under neutrality in Caucasians (Table 2).

Discussion and Conclusion

Several lines of evidence indicate signatures of natural selection on CYP3A4 and CYP3A7 in all three populations in our study. First of all, we observed extremely low nucleotide diversities at both gene loci in all three populations. The π values were obviously much lower than the genome-wide average for autosomes (7.7 × 10⁻⁴), across chromosome 7 (7.6 × 10⁻⁴) (Stephens et al. 2001), and in coding regions for autosomal genes (3.4 × 10⁻⁴) (Sachidanandam et al. 2001). Second, significantly negative values for all four statistics (Tajima’s D, Fu and Li’s D* and F*, and Fay and Wu’s H) were observed in CYP3A4 and CYP3A7. Third, the HKA test also indicated a local reduction in variability relative to divergence at the CYP3A locus compared with the neutral DMD intron 44 (Nachman and Crowell 2000).

In agreement with our findings, Qiu et al. (2008) also detected particularly strong recent positive selections on protein-coding sequences of human CYP3A4 and CYP3A7 after the split of the chimpanzee and human lineages or early in hominoid evolution. In a recent phylogenetic investigation, Zawaira et al. (2008) also showed that the Gotoh’s substrate recognition sites of human CYP3A is associated with the sites under adaptive evolution. CYP3A is involved in the metabolism of foreign compounds, such as naturally occurring flavonoids, diterpenoids in germander, pyrrolizidine alkaloids (e.g., echimidine and jacobine), and herbal constituents (Zhou et al. 2007). CYP3A also normally metabolizes food-derived activators of pregnane X receptor and/or constitutive androstane receptor, thereby regulating the expression of a range of detoxifying systems in the liver (van Waterschoot et al. 2009). Therefore, molecular adaptation to the evolving environment is possible for the CYP3A locus or individual genes in this locus.

In the present study, we observed that the 40.5-kb region from the CYP3A4 5′-flanking region to CYP3A4 intron 6 was under a recent positive selection in human samples overall. Thompson et al. (2004) also observed an excess of rare variants, as well as a low number of polymorphisms, in the CYP3A4 gene in non-Africans but not Africans. Schirmer et al. (2006) observed that the region centromeric of the ancestral CYP3A4*1B allele (which is correlated with moderately increased expression of CYP3A4 mRNA and protein and increased activity) exhibits “high haplotype homozygosity in European Caucasians as oppose d to African Americans.” The slight difference between our study and that of Thompson et al. (2004) may be due to differences in data analysis, as all the neutrality tests were performed on the assumption of absence of genetic recombination across the locus. However, recombination at the CYP3A locus is possible, as indicated by LD structure in the present study. Neutrality tests that take recombination into account, such as the sliding-window analysis, may be more appropriate. Of course, it is still difficult to pinpoint the evolutionary driving force shaping the nucleotide diversity pattern of CYP3A4 in our study. In spite of its role in the metabolism of xenobiotics, CYP3A4 also metabolizes endogenous chemicals such as the eicosanoid metabolite endocannabinoid anandamide (Snider et al. 2007) and uroporphyrin (Franklin et al. 2000); factors that lead to fluctuation of these chemicals might be one of selective forces for human CYP3A4.

CYP3A7 shows distinctly high catalytic activities for the 16α-hydroxylation of steroids with a C17-keto group, such as estrone (Lee et al. 2003) and dehydroepiandrosterone (Ohmori et al. 1998), and plays a role in estriol synthesis (Siiteri and MacDonald 1963) and retinoic acid metabolism (Marill et al. 2002). Leeder et al. (2005) observed extremely low expression of CYP3A7 in livers of anencephalic human fetuses. This iso enzyme is supposed to be feto protective in humans. In the present study, we observed that a nonsynonymous substitution of CYP3A7, N192S, was driven to near fixation in all human samples and fixed in non-Africans. The N192S SNP occurs in a region close to the CYP3As substrate recognition site 2 (Xue et al. 2001). Alignments of amino acid sequences with all human CYP3A and the orthologs in rat (Cyp3a9) and mouse (Cyp3a13) by us indicated that the residue 192N is highly conserved (data not shown). It is possible that the N192S substitution affects the regioselectivity of CYP3A7 toward its substrates and thus is favored by natural selection. Interestingly, Rodríguez-Antona et al. (2005) recently reported that a common nonsynonymous variant CYP3A7*2 is in strong LD with CYP3A5*1. The CYP3A7*1/CYP3A5*3 haplotype is associated with CYP3A7.1 expression but without CYP3A5 expression, whereas the CYP3A7*2/CYP3A5*1 haplotype is associated with the expression of CYP3A5 and a more active form of CYP3A7 (Rodríguez-Antona et al. 2005). Unfortunately, however, we failed to identify the CYP3A7*2 variant in our study because of the highly homologous sequences of the CYP3A family, which prevented our addressing the possibility that CYP3A7*2 might be the target of positive selection.

One of the novel findings in our study was the unique natural selection pattern on CYP3A43 in non-Africans. Most CYP3A43 haplotypes in non-Africans contained the high frequency–derived alleles at G83448A and C83643T polymorphic sites. In addition, the CYP3A43*2 and CYP3A43*3 variants were also in significant LD with these SNPs in non-Africans. The skewed haplotype structure and significant result of haplotype diversity test suggest that the major haplotype in this region is favored by natural selection in non-Africans. The five variants that separated the two CYP3A43 haplotype clusters had extremely high F_ST values for the Chinese–African and Caucasian–African comparisons, which also indicates the impact of population-specific selective pressures on CYP3A43 in populations outside of Africa.

All neutrality tests showed significant departure from neutrality for the sequences from the CYP3A5 exon 2 to the CYP3A5 3′ untranslated region in Caucasians. The remarkable interpopulation differences in both frequency spectrum and haplotype structure in CYP3A5 also indicate a recent selective sweep in Caucasians, which is in accordance with the findings by Thompson et al. (2004). CYP3A5 is important in the metabolism of endogenous substrates such as cortisol and in environmental chemicals such as aflatoxin B₁. CYP3A5*3 reportedly undergoes strong selective pressure (e.g., salt sensitivity) (Thompson et al. 2004). Other selective pressures, especially from dietary chemical exposure, may also play a role in shaping the unusual nucleotide diversity patterns in CYP3A5 in Caucasians.

It has become clear that populations and individuals have their own individual finger-print of unique allelic architecture coding the xenobiotic response system, and these genetic variants have functional relevance with respect to drug response and environmentally related diseases (Nebert and Dieter 2000). Drug-metabolizing enzyme (DME) genes are among the most favorable targets of natural selection for their role in metabolism of environmental compounds. Recent genome-wide analyses indicate that the human xeno biotic metabolism and disposition system may have undergone unique natural selection (Sabeti et al. 2007). It has also been shown in candidate gene–based studies that positive selection has acted on DME genes such as CYP1A2 (Wooding et al. 2002), CYP3A4 and CYP3A5 (Schirmer et al. 2006; Thompson et al. 2004), and FMO3 (flavin-containing mono-oxygenase 3) (Allerston et al. 2007), all phase I DME genes; NAT2 (N-acetyltransferase 2) (Patin et al. 2006) and UGT2B17 (UDP-glucuronosyltransferase 2 family, polypeptide B17) (Xue et al. 2008), both phase II DME genes; ABCB1 [ATP-binding cassette, subfamily B, member 1; also named multidrug resistance 1 (MDR1)] (Tang et al. 2004; Wang et al. 2007), and ABCC1 (Wang et al. 2005), both phase III DMEs. Among these genes, CYP3A and ABCB1 are very interesting outliers because of their unique genetic and biochemical properties. The CYP3A gene cluster is located at chromosome 7, just 119 kb away from the ABCB1 locus, which codes MDR1 (P-glycoprotein), the best-characterized phase III drug transporter and one of the major determinants of the absorption, distribution, metabolism, and excretion/toxicologic profiles for a large range of hydro phobic exogenous substrates, including nearly every category of clinically important drugs, with a substrate spectrum similar to that of CYP3A enzymes. In the liver, lung, kidney, and intestine, there is a close correlation between the expressions of CYP3A4 and ABCB1 and their transcriptional factor, the pregnane X receptor (Miki et al. 2005), which coordinately regulates xenobiotic/drug metabolism and efflux through trans-activating the expression of the two genes (Synold et al. 2001). We and other groups have previously reported positive selection on functional derived variants of ABCB1 in its coding and regulatory regions in human populations, which are correlated with increased expression and enhanced transporter activity (Tang et al. 2004; Wang et al. 2007). The selection hotspot on the two closely linked major xenobiotic response genes, CYP3A and ABCB1, superimposed with their closely linked genomic map, their finely coordinated trans-activation for xenobiotic disposition and clearance, and their overlapping tissue expression profile and substrate spectrum, may provide a good model in evolutionary environmental medicine to decipher the integral and dynamic profile of the organization, function, and evolution of the xenobiotic disposition system.

In summary, this evolutionary population genetics study on the human CYP3A locus confirmed the previously reported positive selection on CYP3A5 in non-Africans, revealed that CYP3A4 and CYP3A7 are under a recent or ongoing positive natural selection in the general human population, and also pointed to positive natural selection on CYP3A43 in non-Africans. The complex pattern of natural selection on the allelic structure of the CYP3A locus may underlie the phenotypic variation in CYP3A activity in populations from different geographic regions and having inter-ethnic differences in response to drugs and herbal medicine as well as ethnic differences in predisposition to complex disorders resulting from CYP3A-substrate exposure.