Worldwide variation in human drug-metabolism enzyme genes CYP2B6 and UGT2B7: implications for HIV/AIDS treatment

Abstract

Aim

Hepatic enzymes, CYP2B6 and UGT2B7 play a major role in the metabolism of the widely used antiretroviral drugs efavirenz, nevirapine and zidovudine. In the present study, we provide a view of UGT2B7 haplotype structure, and quantify the genetic diversity and differentiation at both CYP2B6 and UGT2B7 genes on a worldwide scale.

Materials & methods

We genotyped one intronic and three promoter SNPs, and together with three nonsynonymous SNPs, inferred UGT2B7 alleles in north American (n = 326), west African (n = 133) and Papua New Guinean (n = 142) populations. We also included genotype data for five CYP2B6 and six UGT2B7 SNPs from an additional 12 worldwide populations (n = 629) analyzed in the 1000 Genomes Project.

Results

We observed significant differences in certain SNP and allele frequencies of CYP2B6 and UGT2B7 among worldwide populations. Diversity values were higher for UGT2B7 than for CYP2B6, although there was more diversity between populations for CYP2B6. For both genes, most of the genetic variation was observed among individuals within populations, with the Papua New Guinean population showing the highest pairwise differentiation values for CYP2B6, and the Asian and European populations showing higher pairwise differentiation values for UGT2B7.

Conclusion

These new genetic distinctions provide additional insights for investigating differences in antiretroviral pharmacokinetics and therapy outcomes among ethnically and geographically diverse populations.

The introduction of highly active antiretroviral therapy (HAART) as a standard of care has markedly decreased AIDS incidence and improved HIV/AIDS prognosis [1]. However, a number of clinical factors, including regimen, adherence and stage of infection/disease at the initiation of therapy, are known to influence outcomes of HAART [2–4]. In addition, host-related genetic factors, including polymorphisms in antiretroviral drug-metabolism enzyme and transporter genes, may influence outcomes of HAART [5,6].

Among a variety of antiretroviral drugs that are included in HAART, the non-nucleoside reverse transcriptase inhibitors efavirenz (EFV) and nevirapine (NVP) and, the nucleoside reverse transcriptase inhibitor zidovudine (AZT) are widely used, particularly in resource-limited regions because of cost considerations. For example, NVP is widely used in Africa, especially to prevent mother-to-child transmission [101]. These drugs have narrow therapeutic windows [7–12], but show large interindividual variability in their pharmacokinetics [13,14], contributing to the variable outcomes seen among HIV-infected patients. Clinical studies have suggested that subtherapeutic plasma concentrations are associated with treatment failure, whereas supratherapeutic concentrations are associated with toxicities of these drugs [7–12]. Repeated exposure to subtherapeutic concentrations may also facilitate the emergence of antiretroviral drug resistance [9].

The metabolism of EFV, NVP and AZT is relatively well defined. The hepatic enzyme CYP2B6 plays a major role in the hydroxylation of EFV and NVP [15,16]. The CYP2B6 gene (chromosome 19) is highly polymorphic, containing ≥25 nonsynonymous, several synonymous and numerous promoter SNPs [17]. To date, a total of 29 alleles of CYP2B6 (*1A [wild-type] to *29) have been described [102], many of which are associated with increased, decreased or abolished enzyme activity [17–19]. In particular, CYP2B6*6 (516G>T [rs3745274, exon-4, Gln172His] and 785A>G [rs2279343, exon-5, Lys262Arg]) is associated with ‘slow metabolism’ of EFV [19,20]. Hofmann et al. demonstrated that 516G>T in CYP2B6*6 was alone responsible for aberrant splicing, resulting in a high-splice variant and low CYP2B6 expression phenotype [21].

A number of clinical studies conducted in HIV-infected patients of European and African ethnicities, have analyzed the association between CYP2B6 SNP genotypes, particularly 516G>T, a marker for *6, and/or 983T>C (rs28399499, exon-7, Ile328Thr), and the pharmacokinetics of EFV and NVP [22–29], as well as between the genotypes or pharmacokinetics and outcomes of treatment with these drugs [25,26,30,31]. Recently, similar studies have been conducted in other ethnically diverse patients [32–39]. Taken collectively, the results pertaining to the association between the genotypes and the pharmacokinetics of these drugs are mostly consistent across populations: 516TT and/or 983CC genotypes are associated with up to fivefold higher drug concentrations (~30-fold in one case [40]) compared with other genotypes. However, the results pertaining to the association between the genotypes or pharmacokinetics and outcomes of treatment with these drugs are less clear [25,26,31,37,39], possibly because the CYP2B6 genotype associations may be cohort-specific [31].

Another hepatic enzyme, UGT2B7, catalyzes direct glucuronidation of AZT, EFV and EFV metabolites [41,42]. The UGT2B7 gene (chromosome 4) is also polymorphic, containing three nonsynonymous, and many synonymous, intronic and promoter SNPs [43–45]. To date, a total of four alleles of UGT2B7 (*1a–*4) have been described [103], which may affect UGT2B7 expression and glucuronidation activity [44–46]. Among these alleles, UGT2B7*2 (802C>T [rs7439366, exon-2, His268Tyr]) affects the in vitro glucuronidation of AZT, but not of EFV [41,42]. To our knowledge, influence of UGT2B7 SNP genotypes on antiretroviral pharmacokinetics and clinical response was not reported until recently: 802C>T was found to be significantly associated with lower plasma EFV concentrations in HIV-infected African patients in univariate, but not in multivariate, analysis [47]. A synonymous SNP 735A>G (rs28365062, exon-2, Thr245Thr), linked to several intronic SNPs, was found to be associated with faster AZT clearance in African patients coinfected with HIV–tuberculosis, and with higher AZT glucuronidation in vitro [48]. However, another study, conducted in Belgian patients of European and African ethnicities, did not observe a significant effect of 802C>T or 735A>G on the pharmacokinetics of EFV [49]. Recently, in European–American (EA) and African–American (AFA) patients, we did not observe a significant association between 802C>T or the intronic SNP IVS1+985A>G (rs62298861, intron-1) genotypes and the HAART-related time to achieve virologic success [31]. These studies indicate that the effects of UGT2B7 SNP genotypes could also be cohort-specific [31,47–49].

Despite the fact that EFV, NVP and AZT are widely used antiretroviral drugs, the information regarding worldwide variation in CYP2B6 is limited [17] and, in UGT2B7, is not available. In recent years, the information regarding the prevalence of CYP2B6 alleles in ethnically diverse populations has become increasingly available [50–54], and two general conclusions can be drawn from these studies: CYP2B6*6 is the most prevalent mutant allele and there are significant interethnic differences in CYP2B6 allele frequencies. However, the information regarding worldwide patterns of genetic diversity and differentiation for this gene is still not available. Therefore, it is not known whether the observed differences in the allele frequencies between populations are due to demographic changes or, if they are indicative of the action of local (i.e., geographically restricted) natural selection. Furthermore, the data regarding haplotype structure of UGT2B7 in ethnically diverse populations are still scarce [44,55–57]. The present study takes the first step towards closing some of these gaps.

Previously, based on six CYP2B6 nonsynonymous SNPs, we reported the prevalence of alleles *1A–*7, *9 and *18 in six populations (four major north American ethnic groups, west African and Papua New Guinean [PNG]) [53,58]. The same populations were also analyzed for three UGT2B7 nonsynonymous SNPs [59]. In the present study, our aims were to provide a view of UGT2B7 haplotype structure and to quantify the genetic diversity and differentiation in both CYP2B6 and UGT2B7 genes on a worldwide scale. For these aims, we genotyped one intronic and three promoter SNPs, and together with the nonsynonymous SNPs, inferred UGT2B7 haplotypes (alleles) in north American, west African and PNG populations; we also included genotype data for five CYP2B6 and six UGT2B7 SNPs from an additional 12 worldwide populations analyzed in the 1000 Genomes Project [60]. Our results provide new insights regarding comparative variation in both CYP2B6 and UGT2B7 genes, which may help guide investigations into antiretroviral drug metabolism and therapy outcomes.

Materials & methods

North American, west African & PNG populations (study populations)

The study populations and sample collection procedures have been previously described[53,59,61,62]. Blood samples were collected from north American random blood donors, belonging to four major ethnic groups (EA, AFA, Asian–American [ASA] and Hispanic–American [HA]) [62], and from PNGs from the Wosera area, East Sepik Province [53,59]. DNA samples were collected from west Africans from five countries (Senegal, Guinea, Sierra Leone, Ivory Coast and Ghana) [61]. All samples were collected under protocols including the procedures for informed consent approved by the corresponding institutional review boards. The numbers of samples analyzed in this study (n = 607 [CYP2B6] and n = 601 [UGT2B7]) are provided in supplementary table s1 (see online at: www.futuremedicine.com/doi/suppl/10.2217/pgs.11.160). The ethnicity of the north American individuals was self-identified. All west African samples were collected from the West Atlantic subgroup of the Niger-Congo language family, and were analyzed as one population. The PNG samples were collected from individuals speaking Abelam, which belongs to the Sepik language family. Genomic DNA was extracted from the north American and PNG blood samples using the QIAamp^® 96 DNA blood kit (Qiagen) [53,59].

PCR & SNP genotyping

The aforementioned study populations have previously been genotyped for six CYP2B6 (64C>T [rs8192709, exon-1, Arg22Cys], 516G>T, 777C>A [rs45482602, exon-5, Ser259Arg], 785A>G, 983T>C, and 1459C>T [rs3211371, exon-9, Arg487Cys]) [53,58] and three UGT2B7 (211G>T [rs12233719, exon-1, Ala71Ser], 802C>T and 1192G>A [no rs number, exon-5, Asp398Asn]) [59] SNPs using the oligonucleotide ligation detection reaction-fluorescent microsphere assay on the Bio-Plex^TM suspension array system (Bio-Rad Laboratories). In the present study, these populations were genotyped for the UGT2B7 intronic SNP IVS1+985A>G, which is a haplotype-tagging SNP [44], and promoter SNPs −327G>A (rs7662029), −138G>A (rs73823859) and −125T>C (rs7668282) using the same ligation detection reaction-fluorescent microsphere assay. The method for genotyping IVS1+985A>G has been recently published [31]. The primers and conditions to selectively amplify the UGT2B7 promoter region (574 bp) are described in supplementary table s2. The ligase detection reaction primers and conditions to genotype the promoter SNPs, and the mean and 95% CI of log-transformed fluorescent values corresponding to each of the genotypes are also presented in supplementary table s2.

1000 Genomes Project populations (reference populations)

In addition, we included CYP2B6 and UGT2B7 data, pertaining to the SNPs that were genotyped in the study populations, from the 1000 Genomes Project [60]. The genotype calls were based on 629 individuals, belonging to 12 populations, from the 2010–08–04 sequence and alignment release of the project (northern and western European from UT, USA; Toscani from Italy; British from England and Scotland; Finnish from Finland [FIN]; African ancestry from southwest, USA; Yoruba from Ibadan, Nigeria; Luhya from Webuye, Kenya; Han Chinese from Beijing, China; Han Chinese from south China; Japanese from Tokyo, Japan; Mexican ancestry from Los Angeles, CA, USA; and Puerto Rican from Puerto Rico, USA [PUR]). Sample sizes of these globally distributed reference populations are provided in supplementary table s1.

SNP data

The 1000 Genomes Project provides low-coverage (2–4×) sequence data for each sample; at this depth, sequencing cannot provide the complete genotype of each sample, but may allow the detection of most variants with frequencies as low as 1%. Using BEAGLE genetic analysis software package [104], we included high-quality phased genotype data for the five CYP2B6 (64C>T, 516G>T, 785A>G, 983T>C and 1459C>T) and six UGT2B7 (−327A>G, −138G>A, −125T>C, 211G>T, IVS1+985A>G and 1192G>A) SNPs from the aforementioned 12 reference populations. The CYP2B6 777C>A and UGT2B7 802C>T SNP data were not available, possibly because these sites did not pass quality filters.

Statistical analysis

GenePop v4.0.10 [105] was used to calculate CYP2B6 and UGT2B7 SNP frequencies and to perform the Hardy–Weinberg exact tests (estimation of p-values were calculated using the Markov chain method). SHEsis online version [106] was used to compute Lewontin’s D’ and correlation coefficient (r²) parameters in order to measure linkage disequilibrium (LD) between SNPs. Since the six SNPs included in this study define well-known CYP2B6 alleles [102], CYP2B6 alleles in the study populations were identified manually. However, since the newly characterized SNP IVS1+985A>G is not yet part of the UGT2B7 alleles [103], UGT2B7 alleles in these populations were inferred using PHASE v2.1.1 [107]. An online 2 × 2 contingency table for Fisher’s exact test [108] was used to calculate differences in the SNP/allele frequencies between populations, and a two-tail value of <0.05 was considered to be significant. DnaSP v5.10 [109] was used to compute haplotype diversity (H), and Arlequin v3.11 [110] was used to compute analysis of molecular variance (AMOVA) and pairwise population differentiation (PA-F_st). Both H and F_st values range from 0 (no diversity, no differentiation) to 1 (every individual has a different haplotype, fixed difference between populations).

Results

SNP frequencies, LD patterns & allele profiles

CYP2B6 SNP frequencies

The CYP2B6 SNP frequencies are presented in table 1. 64C>T had a low prevalence, 1459C>T had a low-to-moderate prevalence, whereas 516G>T and 785A>G were highly prevalent in all populations. Among all of the 18 populations examined, the lowest frequency of 516G>T was observed in the FIN population. The 516G>T and 785A>G SNP frequencies in the PNG population were significantly higher than in any other population. 983T>C was low to moderately prevalent in the populations of African ancestry, and was also present in the PUR (0.05) and HA (0.01) populations. The PUR population arose as a result of admixing among Europeans, west Africans and Native Americans [63]. Overall, our CYP2B6 SNP frequency results are in agreement with those previously reported for comparable populations [52,54,64,65].

Table 1.

Frequencies of CYP2B6 SNPs in various populations.

Population†	n	64C>T	516G>T	777C>A	785A>G	983T>C	1459C>T
		f(T) ‡	f(T) ‡	f(A) ‡	f(G) ‡	f(C) ‡	f(T) ‡
European–American	59	0.03	0.34	0	0.38	0	0.07
Northern and western European from UT (USA)	90	0.04	0.25	NA	0.20	0	0.07
Toscani from Italy	92	0.07	0.29	NA	0.21	0	0.09
British from England and Scotland	43	0.03	0.26	NA	0.14	0	0.10
Finnish	36	0.04	0.08	NA	0.10	0	0.08
African–American	85	0.02	0.36	0	0.37	0.08	0.04
African ancestry from southwest USA	24	0.06	0.38	NA	0.35	0.04	0.02
Yoruba from Ibadan (Nigeria)	78	0.04	0.35	NA	0.22	0.12	0.01
West African	153	0.04	0.5	0	0.42	0.05	0.02
Luhya from Webuye (Kenya)	67	0.08	0.37	NA	0.33	0.06	0
Asian–American	61	0.07	0.23	0	0.27	0	0.02
Han Chinese from Beijing	68	0.03	0.13	NA	0.15	0	0
Han Chinese from south China	25	0.04	0.10	NA	0.10	0	0
Japanese from Tokyo (Japan)	84	0.01	0.21	NA	0.19	0	0.01
Hispanic–American	77	0.03	0.37	0	0.35	0.01	0.07
Mexican ancestry from Los Angeles (CA, USA)	17	0.06	0.29	NA	0.24	0	0.12
Puerto Rican from Puerto Rico (USA)	5	0	0.30	NA	0.30	0.05	0.05
Papua New Guinean	172	0	0.65*	0	0.64*	0	0.03

^†As described in the ‘Materials and methods’ section.

^‡Mutant allele.

^*p< 0.05 compared with any other population.

f: Frequency; NA: Not available.

UGT2B7 SNP frequencies

The UGT2B7 SNP frequencies are presented in table 2. −138G>A and −125T>C had a low prevalence, whereas −327A>G and IVS1+985A>G were highly prevalent in all populations. Notably, the frequency of −327A>G in the European group (~0.5), and the frequencies of IVS1+985A>G in the Asian group (0.04–0.11) were lower than in any other group. 802C>T was also highly prevalent in the study populations, with the frequency of 802T (0.53) being significantly higher in the EA population than in any other population. The 802C>T SNP data for the reference populations were not available. However, the frequencies of 802T in northern and western Europeans from UT, USA; Yoruba from Ibadan, Nigeria; Han Chinese from Beijing, China plus Japanese from Tokyo, Japan populations (all are part of the low-coverage panel) have been reported as 0.49, 0.26 and 0.32, respectively [111]. These frequencies are comparable to those in EA, AFA and ASA study populations, respectively (table 2), suggesting worldwide interethnic differences. 211G>T was highly prevalent in the populations of Asian ancestry, and was also present in the FIN (0.04), Luhya from Webuye, Kenya (0.01) and HA (0.02) populations. The origin of the population of Finland has been debatable, and both European and Asian populations seem to have provided a contribution to the Finnish gene pool [66–68]. Luhya from Webuye, Kenya, is an admixed population in which individuals have varying continental ancestry proportions [69]. Overall, our UGT2B7 SNP frequency results are in agreement with those previously reported for comparable populations [44,56,57,70,71].

Table 2.

Frequencies of UGT2B7 SNPs in various populations.

Population†	n	−327A>G	−138G>A	−125T>C	211G>T	IVS1+985A>G	802C>T	1192G>A
		f(G) ‡	f(A) ‡	f(C) ‡	f(T) ‡	f(G) ‡	f(T) ‡	f(A) ‡
European–American	79	0.48	0.03	0.03	0	0.12	0.53*	0
Northern and western European from UT (USA)	90	0.51	0.02	0	0	0.16	NA	0
Toscani from Italy	92	0.48	0.08	0	0	0.13	NA	0
British from England and Scotland	43	0.47	0.01	0	0	0.13	NA	0
Finnish	36	0.51	0.04	0	0.04	0.24	NA	0
African–American	82	0.66	0.02	0.02	0	0.23	0.34	0
African ancestry from southwest USA	24	0.79	0.04	0.08	0	0.38	NA	0
Yoruba from Ibadan (Nigeria)	78	0.80	0.03	0.04	0	0.33	NA	0
West African	133	0.79	0.02	0.05	0	0.34	0.21	0
Luhya from Webuye (Kenya)	67	0.72	0.09	0.04	0.01	0.29	NA	0
Asian–American	81	0.61	0.01	0.09	0.09	0.09	0.4	0
Han Chinese from Beijing	68	0.68	0	0.07	0.12	0.04	NA	0
Han Chinese from south China	25	0.76	0	0.1	0.16	0.04	NA	0
Japanese from Tokyo (Japan)	84	0.67	0	0.07	0.13	0.11	NA	0
Hispanic–American	84	0.72	0.01	0.01	0.02	0.37	0.28	0
Mexican ancestry from Los Angeles (CA, USA)	17	0.66	0.09	0	0	0.31	NA	0
Puerto Rican from Puerto Rico (USA)	5	0.7	0.1	0	0	0.45	NA	0
Papua New Guinean	142	0.72	0	0	0	0.27	0.28	0

^†As described in the ‘Materials and methods’ section.

^‡Mutant allele.

^*p < 0.05 compared with African–American, west African, Asian–American, Hispanic–American or Papua New Guinean population.

f: Frequency; NA: Not available.

Populations in Hardy–Weinberg equilibrium

We calculated expected genotype numbers for each of the CYP2B6 and UGT2B7 SNPs in the study and reference populations. For all SNPs analyzed, the expected genotype numbers did not differ significantly from the observed genotype numbers in any of the populations (data not shown). However, a significant deficit of heterozygosity at CYP2B6 516G>T (p = 0.016) and 1459C>T (p = 0.003) was observed in the PNG population, which indicates either nonrandom mating or population substructure.

LD patterns

We quantified the extent of LD among the CYP2B6 and UGT2B7 SNP pairs in all populations. Strong LD, defined by high values for both D’ (≥0.8) and r² (≥0.5) parameters [72], was only observed between SNP pairs CYP2B6 516–785 (all populations) and UGT2B7 −327–802 (study populations; 802C>T SNP data not available for the reference populations) (table 3). The UGT2B7 SNP pair −327–IVS1+985 had high D’ (>0.8) and moderate r² (0.12–0.351) values in all populations except the Asian group. The UGT2B7 SNP pair −327–138, had high D’ (1) and moderate r² (0.106–0.259) values in the African group. All other SNP pairs had highly variable D’ (0–0.8) and low r² (<0.1) values in all populations (data not shown). CYP2B6 & UGT2B7 allele profiles Results regarding CYP2B6 six-SNP alleles in the study populations have been previously reported [53,58]. Briefly, a total of ten alleles were identified in the study populations, of which the wild-type *1A and the mutant *6 were the most prevalent alleles in all populations. The frequency of *6 in the PNG population (0.62) was significantly higher than in any north American (0.23–0.34) or west African (0.42) population [53]. CYP2B6*18 (983C) was prevalent in the populations of African ancestry (AFA [0.08] and west African [0.05]) [58]. Since we did not have the complete six-SNP frequency data for the reference populations (table 1), we could not perform the allele profile analysis on this set of populations. However, since we had the 516G>T, 785A>G, and 983T>C SNP data for all of the reference populations, we estimated frequencies of haplotypes 516T_785G (presumably *6 ), 516G_785A _983C (presumably *18) and 516G_785G_983C (presumably *16) (table 4). Overall, the results of this analysis were similar to those for the study populations [53,58]. Taken collectively, these results indicate that among all the populations examined, the frequency of *6 is lowest in the FIN population (0.07) and highest in the PNG population (0.62). In the African group and the PUR population, 983T>C was observed predominantly as *18. Previously, 983T>C has been found either as *18 in AFA, Ghanaian and other west African populations [52,58], or as *16 in Tanzanian and Turkish populations [73].

Table 3.

Pairwise linkage disequilibrium in CYP2B6 and UGT2B7 SNPs in various populations.

Population†	SNP pair
	CYP2B6 516–785 D’ (r2)	UGT2B7 −327–802 D’ (r2)
European–American	0.873 (0.657)	1 (0.977)
Northern and western European from UT (USA)	0.778 (0.454)	–
Toscani from Italy	0.82 (0.447)	–
British from England and Scotland	1 (0.472)	–
Finnish	0.815 (0.561)	–
African–American	0.973 (0.901)	1 (1)
African ancestry from southwest USA	1 (0.914)	–
Yoruba from Ibadan (Nigeria)	0.955 (0.48)	–
West African	1 (0.719)	0.976 (0.911)
Luhya from Webuye (Kenya)	0.964 (0.763)	–
Asian–American	1 (0.803)	1 (1)
Han Chinese from Beijing	0.869 (0.63)	–
Han Chinese from south China	1 (1)	–
Japanese from Tokyo (Japan)	0.882 (0.695)	–
Hispanic–American	0.877 (0.727)	1 (1)
Mexican ancestry from Los Angeles (CA, USA)	1 (0.738)	–
Puerto Rican from Puerto Rico (USA)	0.762 (0.58)	–
Papua New Guinean	0.987 (0.926)	1 (1)

^†As described in the ‘Materials and methods’ section.

–: 802C>T SNP data not available.

Table 4.

Frequencies of certain CYP2B6 haplotypes in reference populations.

Population†	Haplotype
	516T_785G ‡	516G_785A_983C §	516G_785G_983C ¶
Northern and western European from UT (USA)	0.17	0	0
Toscani from Italy	0.18	0	0
British from England and Scotland	0.14	0	0
Finnish	0.07	0	0
African ancestry from southwest USA	0.35	0.04	0
Yoruba from Ibadan (Nigeria)	0.21	0.11	0.01
Luhya from Webuye (Kenya)	0.32	0.05	0
Han Chinese from Beijing	0.12	0	0
Han Chinese from south China	0.10	0	0
Japanese from Tokyo (Japan)	0.17	0	0
Mexican ancestry from Los Angeles (CA, USA)	0.23	0	0
Puerto Rican from Puerto Rico (USA)	0.25	0.05	0

^†As described in the ‘Materials and methods’ section.

^‡Presumably *6.

^§Presumably *18 [52,58].

^¶Presumably *16 [73].

We inferred UGT2B7 six-SNP allele profiles in the study populations using PHASE v2.1.1 (table 5). Assuming random associations among the six polymorphic UGT2B7 SNPs, one can predict 64 different alleles (2⁶). However, only seven alleles (6 + 1) would be possible if mutations are the only evolutionary forces acting to create new alleles, and other forces, such as recombination, recurrent mutation and gene conversion, do not occur [74]. We inferred a total of seven alleles in the study populations, which is the theoretical minimum. Of these, three alleles together accounted for the majority of total chromosomes examined: *1a, *1m, and *2a occurred at frequencies >0.1 in most of the populations. Notably, these were the only alleles present in the PNG population, whereas all other populations had five to six alleles. The new haplotype, tentatively designated as UGT2B7*1m, contains IVS1+985G, together with −327G and 802C, and was highly prevalent (table 5). The frequency of UGT2B7*1m was significantly higher in HA (0.37) than in any other north American population or in the PNG population. UGT2B7*2a (−327A_802T) was also highly prevalent; its frequency was significantly higher in EA (0.49) than in HA, AFA, west African or PNG population. Finally, UGT2B7*3 (211T) was prevalent (~0.1) in the ASA population, as previously reported for other populations of Asian ancestry [56,57].

Table 5.

Frequencies of UGT2B7 alleles in study populations.

Allele†	SNP							Population‡
	− 327	− 138	− 125	211	IVS1+985	802	1192	European– American (n = 79)	Hispanic– American (n = 82)	Asian– American (n = 81)	African– American (n = 84)	West African (n = 133)	Papua New Guinean (n = 142)
*1a	G	G	T	G	A	C	G	0.34	0.32	0.33	0.40	0.40	0.45
*1b	G	G	C	G	A	C	G	0.03	0.01	0.09	0.02	0.05	0
*1m	G	G	T	G	G	C	G	0.11	0.37*	0.09	0.23	0.34	0.27
*2a	A	G	T	G	A	T	G	0.49**	0.27	0.38	0.32	0.19	0.28
*2g	A	A	T	G	A	T	G	0.03	0.01	0.01	0.02	0.02	0
*2i	A	G	T	G	G	T	G	0.01	0	0	0	0	0
*3	G	G	T	T	A	C	G	0	0.02	0.09	0	0	0
*4	G	G	T	G	A	C	A	0	0	0	0	0	0

¹UGT2B7*1m and *2i represent new haplotypes (alleles), require validation.

^†Based on [103].

^‡As described in the ‘Materials and methods’ section.

^*p < 0.05 compared with any other north American or Papua New Guinean population;

^**p < 0.05 compared with any other north American (except Asian–American; p = 0.07), west African or Papua New Guinean population.

Genetic diversity & differentiation

CYP2B6 and UGT2B7 H values for the various populations are presented in Figure 1. Diversity values were higher for UGT2B7 (mean H = 0.708) than for CYP2B6 (mean H = 0.608); with particularly low H-values observed for CYP2B6 in the FIN (0.408, Han Chinese from Beijing, China (0.344); Han Chinese from south China (0.254); and Japanese from Tokyo, Japan (0.399) populations. Interpopulation diversity varied much more for CYP2B6 (H = 0.254–0.690) than for UGT2B7 (H = 0.603–0.737). Comparison of the genetic variation between CYP2B6 and UGT2B7 genes in all populations by AMOVA revealed that ≥90% of the variation was partitioned within populations, with some small but significant variation among groups (table 6), indicating significant between-population differences. These results are largely in agreement with previous findings, including those pertaining to absorption–distribution–metabolism–excretion genes, that the overwhelming majority of human genetic variation is found among individuals within populations [75,76].

Figure 1. CYP2B6 and UGT2B7 haplotype diversity in various populations.

Details regarding the study (EA, AFA, ASA, HA, west African and PNG) and the reference (CEU, TSI, GBR, FIN, ASW, YRI, LWK, CHB, CHS, JPT, MXL and PUR) populations are provided in supplementary table s1 and in the ‘Materials and methods’ section.

AFA: African–American; ASA: Asian–American; ASW: African ancestry from southwest USA; CEU: Northern and western European from UT, USA; CHB: Han Chinese from Beijing, China; CHS: Han Chinese from south China; EA: European–American; FIN: Finnish; HA: Hispanic–American; JPT: Japanese from Tokyo, Japan; LWK: Luhya from Webuye, Kenya; MXL: Mexican ancestry from Los Angeles, CA, USA; PNG: Papua New Guinean; PUR: Puerto Rican from Puerto Rico, USA; TSI: Toscani from Italy; YRI: Yoruba from Ibadan, Nigeria.

Table 6.

Distribution of CYP2B6 and UGT2B7 genetic variation among and within ethnic groups.

Source of variation	CYP2B6		UGT2B7
	Variation (%)	p-value	Variation (%)	p-value
Among groups	9.3	<0.001	6.25	<0.001
Among populations within groups	1.6	<0.001	0.4	0.039
Among individuals within populations	89.1	<0.001	93.35	<0.001

Ethnic groups: European (European–American; northern and western European from UT, USA; Toscani from Italy, British from England and Scotland; and Finnish), African (African–American; African ancestry from the southwest USA, Yoruba from Ibadan, Nigeria; west African; and Luhya from Webuye, Kenya), Asian (Asian–American; Han Chinese from Beijing; Han Chinese from south China; and Japanese from Tokyo, Japan), Hispanic (Hispanic–American; Mexican ancestry from Los Angeles, CA, USA; and Puerto Rican from Puerto Rico, USA), and Papua New Guinean. All these populations are described in the ‘Materials and methods’ section.

The PA-F_st values for CYP2B6 and UGT2B7 in all populations are presented in supplementary table s3a & s3b, respectively. The mean values for these genes were 0.062 and 0.061, respectively. For CYP2B6, the highest values were observed in comparisons involving a PNG population, in keeping with the very different allele-frequency distribution for this population [53]. For UGT2B7, the highest values tend to involve the populations with Asian ancestry (mean pairwise differentiation [PA-F_st] = 0.050–0.089), followed by the values for those with European ancestry (mean PA-F_st = 0.042–0.076). We also analyzed these results according to ethnic groups (European, African, Asian, Hispanic and PNG). In this analysis, the PNG population showed the highest differentiation at CYP2B6, and the Asian (PA-F_st = 0.043–0.068) and European (PA-F_st = 0.019–0.080) groups showed higher differentiation at UGT2B7 (table 7).

Table 7.

Pairwise genetic differentiation among various ethnic groups.

	European	African	Asian	Hispanic	Papua New Guinean
European	–	0.080†	0.044†	0.019†	0.059†
African	0.043‡	–	0.068†	0.021†	0.003†
Asian	0.012‡	0.076‡	–	0.043†	0.059†
Hispanic	0.012‡	0.011‡	0.048‡	–	0.008†
Papua New Guinean	0.254‡	0.112‡	0.338‡	0.163‡	–

^†Values represent pairwise differentiation at UGT2B7.

^‡Values represent pairwise differentiation at CYP2B6.

¹Ethnic groups: European (European–American; northern and western European from UT, USA; Toscani from Italy; British from England and Scotland; and Finnish), African (African–American; African Ancestry from southwest USA; Yoruba from Ibadan, Nigeria; west African; and Luhya from Webuye, Kenya), Asian (Asian–American; Han Chinese from Beijing, China; Han Chinese from south China; and Japanese from Tokyo, Japan), Hispanic (Hispanic–American; Mexican ancestry from Los Angeles, CA, USA; and Puerto Rican from Puerto Rico, USA) and Papua New Guinean. All these populations are described in the ‘Materials and methods’ section.

Discussion

This study attempted to provide comprehensive comparative information regarding variation in both CYP2B6 and UGT2B7 genes on a worldwide scale, by including north American, west African and PNG, as well as 1000 Genomes Project populations. Taken collectively, we observed that:

• ▪ There are significant differences in certain SNP and allele frequencies of CYP2B6 and UGT2B7 among worldwide populations, which in turn produce differences in H and PA-F_st;

• ▪ Diversity values were higher for UGT2B7 than for CYP2B6, although there was more diversity between populations for CYP2B6;

• ▪ For both genes, most of the genetic variation was observed among individuals within populations, with the PNG population showing the highest PA-F_st values for CYP2B6, and the Asian and European populations showing higher PA-F_st values for UGT2B7.

These observations are meant to help guide antiretroviral drug metabolism and therapy outcome studies that are designed to optimize drug effectiveness, while minimizing toxicity.

Differences in SNP/allele frequencies CYP2B6

Our CYP2B6 SNP/allele frequency data suggest that among the four population groups, European, African, Asian and Hispanic, the slow metabolism phenotype of EFV, associated with CYP2B6*6, is likely to be more prevalent in the African (frequencies of *6; 0.21–0.41) and Hispanic (0.23–0.32) groups than in the European (0.07–0.31) and Asian (0.1–0.23) groups. Thus, it is conceivable that slower hepatic metabolism and the resulting increased systemic circulation level of EFV are more frequent in African and Hispanic groups than in European and Asian groups. Based on such allele frequency differences, as well as some recent proof-of-concept and pharmacokinetic model-based studies, it is being advocated that there is a need for prospective clinical studies to evaluate the utility of genotype-driven EFV dose adjustments in populations of diverse ancestry [77]. Another suggested approach is therapeutic drug monitoring to identify improperly dosed individuals within a population [13,78]. Genotype-based prediction and/or therapeutic drug monitoring could possibly lead to EFV dose reduction (an economic advantage) if many individuals in a given population are shown to be exposed to supratherapeutic drug concentrations. Results of a number of recent clinical studies conducted in the populations of Africa, where the frequency of *6 ranges from 0.3 to 0.5, support the use of such an approach [17,77,78]. Our worldwide *6 allele frequency data substantiates the need to consider such an approach in a PNG population.

Recent published estimates suggest that the PNG population has among the highest adult HIV prevalences in the Asia–Pacific region, estimated at 1.28% among people aged 15–49 years in 2007, although more recent estimates suggest national prevalence may be closer to 1.0% [112,113]. Regardless, HIV prevalence is projected to increase up to almost 3% by 2015 [113]. Current antiretroviral therapy coverage is ~75% [113]. EFV is used in first-line antiretroviral therapy as a substitute for NVP, and is preferred for cases of HIV–tuberculosis coinfection [114]. Despite the unique high frequency (0.62) of *6, data on EFV pharmacokinetics, clinical response and toxicity are not available in the PNG population. Therefore, in this population it would be most sensible to pursue genotype-driven EFV clinical studies. In all likelihood, this would permit a decreased overall use of EFV per patient, allowing the treatment of more patients. It would also contribute to a clearer understanding of the effects of high-frequency genetic mutations in the therapeutic setting.

On the other hand, we noted that among all the populations, the frequency of *6 was lowest in the FIN population (0.07). It is anticipated that the area under the 24-h plasma EFV concentration–time curve (AUC₀–24 h) may be lower among FIN than among other populations, although a formal pharmacokinetic analysis would be needed to confirm this prediction. Thus, the PNG and FIN populations, compared with the other populations in the present study, would present two different dosing approaches for EFV to be both effective and safe.

The populations of African ancestry, and with African admixture (PUR and HA in this study), may show further interindividual differences in EFV and NVP pharmacokinetics and/or treatment outcomes due to the presence of 983T>C, compared with the populations of non-African ancestry where this SNP is absent. 983T>C, together with 516G>T, was associated with fivefold higher mean plasma EFV concentrations in HIV-infected African patients [73]. Subsequently, a number of studies reported significant associations between 983T>C genotype and lower clearance/higher plasma concentrations of EFV [23] and NVP [28–30], and between 983T>C genotype and NVP-related adverse events [30,79] in HIV-infected African patients.

UGT2B7

Compared with CYP2B6 polymorphisms, fewer and inconclusive data are available in the literature regarding the effects of UGT2B7 polymorphisms on AZT/EFV metabolism [41,42], pharmacokinetics [47–49] and clinical response [31]. Here, it is important to consider that the pharmacokinetics and clinical response studies were conducted in patients of European and African ethnicities only, and it is possible that the outcome is cohort-specific. Therefore, further investigations that include more diverse study populations are needed. The UGT2B7 SNPs −327A>G, IVS1+985A>G, 802C>T and 211G>T, showing noticeable frequency differences in the present study, are functionally important. The −327A>G SNP may not be of direct functional importance, but is a part of the haplotype II that includes other promoter SNPs, including −161T>C (previously known as −102T>C). In an in vitro assay, the −161C variant resulted in a twofold increase in the transcriptional activity [80]. However, in patients, the effect of −161T>C genotypes on the glucuronidation of morphine is not clear and may depend on medical condition, daily dosage of morphine and administration of other drugs [80]. In human liver microsomes, IVS1+985A>G, a part of the haplotype 4, was significantly associated with increased mRNA expression and the glucuronidation of morphine [44]. A negative impact of 802C>T on the protein content [41] and specific activity of the enzyme has been indicated [70]. However, the SNP did not affect the in vitro glucuronidation of AZT [81] and EFV [42]. Finally, compared with the 211G allele, heterologous expression of the 211T variant in HeLa cells resulted in lower protein expression and a significant decrease in the glucuronidation of carvedilol, an α- and β-adrenergic blocking drug [82]. However, the effect of 211G>T on the pharmacokinetics of carvedilol may be cohort-specific [83,84]. Thus, it is important to determine whether the IVS1+985G and 211T enzyme variants result in altered catalytic activity towards AZT and EFV, and whether certain promoter haplotypes, in linkage with coding SNPs (−327A>G is in linkage with 802C>T; table 3), influence antiretroviral pharmacokinetics and clinical response.

A small number of studies that have analyzed the influence of a few UGT2B7 SNP genotypes on antiretroviral pharmacokinetics and clinical response have yielded varying results [31,47–49]. Using the common UGT2B7 haplotypes may be more meaningful to interpret HIV/AIDS treatment outcome measures than by analyzing those SNPs singly. In this direction, our UGT2B7 haplotype frequency data show substantial differences among ethnically diverse populations (table 5), thus providing highly valuable information for future pharmacogenetic studies of antiretroviral therapy.

Differences in CYP2B6 & UGT2B7 diversity & differentiation

In addition to CYP2B6 and UGT2B7 SNP/allele frequency data, our genetic diversity and differentiation data may also be helpful in interpreting differences in outcomes of antiretroviral therapy among worldwide populations. Worldwide, interpopulation diversity varied much more for CYP2B6 (H = 0.254–0.690) than for UGT2B7 (H = 0.603–0.737), which reflects more pronounced frequency differences in the CYP2B6 SNPs than in the UGT2B7 SNPs. This suggests that interpopulation differences in outcomes of antiretroviral therapy are likely to be more correlated with the differences in CYP2B6 diversity than with the differences in UGT2B7 diversity. This is also supported by larger between-population differences for CYP2B6 than for UGT2B7 in the AMOVA (10.9 and 6.65%, respectively) and pairwise population differentiation (mean PA-F_st = 0.017–0.226 and 0.030–0.089, respectively) results. Based on the pairwise population differentiation results (supplementary tables s3a & s3b), the PNG (mean PA-F_st [CYP2B6] = 0.226), FIN (mean PA-F_st [CYP2B6] = 0.101), Asian (mean PA-F_st [UGT2B7] = 0.050–0.089) and European (mean PA-F_st [UGT2B7] = 0.042–0.076) populations are likely to show the most noticeable differences in outcomes of antiretroviral therapy.

Genetic differentiation among populations reflects the action of both random processes (e.g., genetic drift) and natural selection. An interesting question is: does natural selection appear to have played a role in the patterns of genetic differentiation observed at CYP2B6 and UGT2B7, particularly at CYP2B6 in the PNG population? The present data are limited, and further analyses, including in-depth investigation of the patterns of genetic variation in and around these loci, would be required to investigate this question. However, regardless of whether demography or natural selection is responsible for the SNP frequency differences and corresponding patterns of population differentiation, our results can still be considered as a starting point in identifying the contribution of genetic variation in CYP2B6 and UGT2B7 to (variable) HIV/AIDS treatment outcomes among diverse populations.

In summary, despite the complexity of HIV/AIDS treatment-related genetic investigations, many associations have been described between CYP2B6 variants, antiretroviral pharmacokinetics and/or therapy outcomes; however, some of those associations must still be validated. On the other hand, only a few studies have taken UGT2B7 variation into consideration; these studies have yielded inconsistent results. It is worth noting that for each reported association, or where there was no apparent association, the implicated CYP2B6/UGT2B7 variant differs considerably in frequency among ethnic groups. Although CYP2B6 and UGT2B7 variation has important clinical implications for HIV/AIDS treatment, recent studies indicate that the contributions of other drug-metabolism enzyme genes [47,85], drug-transporter genes [86], transcription factor genes [64,87] and protein–protein interactions [88] should not be discounted. Furthermore, in addition to SNPs, other key genomic mechanisms are likely to underly phenotypic variability; these include recently discovered alternative splicing of UGT2B7, which gives rise to multiple mRNA splice variants with novel functions [89]. Finally, the presence of key functional SNPs, such as CYP2B6 983T>C and UGT2B7 211G>T in admixed populations emphasizes the need to consider population history while assessing the genotype associations. In the future, antiretroviral treatment-related clinical studies need to consider more comprehensive genetic information, which would enable a systematic understanding of the interplay between ethnicity and HIV/AIDS treatment outcomes. Using this knowledge for optimizing antiretroviral effectiveness, while minimizing toxicity, could have substantial public health benefits, particularly in resource-limited countries.

Executive summary.

Background

• ▪ The key to a long life with HIV is highly active antiretroviral therapy, the HIV medications that changed the lives of all those affected and infected with HIV. Highly active antiretroviral therapy is defined as treatment with at least three antiretroviral medications, typically two nucleoside or nucleotide reverse transcriptase inhibitors plus a non-nucleoside reverse transcriptase inhibitor or a protease inhibitor.

• ▪ The non-nucleoside reverse transcriptase inhibitors efavirenz or nevirapine and nucleoside reverse transcriptase inhibitor zidovudine are included in many highly active antiretroviral therapy regimens used worldwide. Pharmacokinetics of, and outcomes of treatment with these drugs vary interindividually as well as interethnically. Genetic changes in the enzyme metabolism of these medications may contribute to clinical variability.

• ▪ The hepatic enzymes CYP2B6 and UGT2B7 play a major role in the metabolism of efavirenz, nevirapine and zidovudine. The information regarding comparative variation in both CYP2B6 and UGT2B7 genes on a worldwide scale is limited.

Materials & methods

• ▪ In the present study, we provide a view of UGT2B7 haplotype structure and quantify the genetic diversity and differentiation at both CYP2B6 and UGT2B7 genes on a worldwide scale, by analyzing key SNPs in north American, west African, Papua New Guinean and 1000 Genomes Project populations.

Results

• ▪ We found significant differences in certain SNP and allele frequencies of CYP2B6 and UGT2B7 among worldwide populations, which in turn produce differences in diversity and pairwise differentiation. Diversity values were higher for UGT2B7 than for CYP2B6, although there was more diversity between populations for CYP2B6. For both genes, most of the genetic variation was observed among individuals within populations, with the Papua New Guinean population showing the highest pairwise differentiation values for CYP2B6, and the Asian and European populations showing higher pairwise differentiation values for UGT2B7.

Conclusion

• ▪ These findings are meant to help guide investigations into antiretroviral drug metabolism and therapy outcomes, aimed at optimizing effectiveness and minimizing toxicity of these drugs.