Global patterns of variation in allele and haplotype frequencies and linkage disequilibrium across the CYP2E1 gene

Abstract

Cytochrome P450 2E1, gene symbol CYP2E1, is one of a family of enzymes with a central role in activating and detoxifying xenobiotics and endogenous compounds. Genetic variation at this gene has been reported in different human populations, and some association studies have reported increased risk for cancers and other diseases. To the best of our knowledge, multi-SNP haplotypes and linkage disequilibrium (LD) have not been systematically studied for CYP2E1 in multiple populations. Haplotypes can greatly increase the power both to identify patterns of genetic variation relevant for gene expression as well as to detect disease-related susceptibility mutations. We present frequency and LD data and analyses for 11 polymorphisms and their haplotypes that we have studied on over 2,600 individuals from 50 human population samples representing the major geographical regions of the world. The diverse patterns of haplotype variation found in the different populations we have studied show that ethnicity may be an important variable helping to explain inconsistencies that have been reported by association studies. More studies clearly are needed of the variants we have studied, especially those in the 5′ region, such as the VNTR, as well as studies of additional polymorphisms known for this gene to establish evidence relating any systematic differences in gene expression that exist to the haplotypes at this gene.

Introduction

Haplotype diversity is a key to understanding population evolution as well as disease evolution. Heterogeneity in both Linkage Disequilibrium (LD) and haplotype frequencies across the genome have been observed among large numbers of diverse ethnic populations in several studies ^1–4. Earlier studies from our laboratory have shown that haplotype and LD patterns at different genes associated with diseases vary widely across different populations of the world ^{2, 5–7}. Studies on different genes associated with disease that included the CEPH diversity panel have also shown widely varying haplotype patterns ^8–9. These earlier studies demonstrate the importance of studying the variation patterns in multiple populations representing different regions of the world for genes that have been associated with disease.

Cytochrome P450 2E1 (CYP2E1) is a member of the cytochrome P450 multifamily of enzymes that play a central role in activating and detoxifying a wide variety of xenobiotics as well as endogenous compounds. Several drug effects have been identified. The anti-fungal drug miconazale has been found¹⁰ to inhibit CYP2E1 enzyme activity. Peterson et al.¹¹ have discussed the complex role CYP2E1 appears to play in the pharmacologic interaction of ciprofloxacin and pentoxifylline; genetic variation in CYP2E1 function may thus have complex secondary consequences. The review by Gonzalez and Yu¹² summarizes the evidence for the important role that genetic variation in the CYP2E1 enzyme plays in the susceptibility of patients to hepatitis induced by anti-tuberculosis drug therapy. The PharmGKB database has links to publications showing the relationship to alcohol-related liver diseases and also reports drug response studies involving CYP2E1 for acetaminophen, alcohol, ethanol, geldanamycin, and xenobiotics.

Considerable variation in allelic distributions at CYP2E1 and of CYP2E1 enzyme activity is found among different human populations ^13–16. Several polymorphic sites in the 5′-flanking and intronic region of CYP2E1 have been reported to be associated with increased risk factors for cancers and other diseases ^16–20. However, no consistent results were observed in studies of the effects of these SNPs on the expression of the gene and activity of the enzyme, and on the susceptibility to diseases ^21–24.

The promoter region and other regulatory variation in or near the gene will function in cis with any amino acid variation as one functional unit. Relevant variation can also include any variants that affect splicing or mRNA conformation. Thus, the haplotype encompassing all relevant variation is the relevant unit for association studies. LD may allow SNPs with no functional consequences to serve as surrogates for unknown and/or untyped variants with functional consequences. However, haplotype frequencies and LD patterns are expected to vary among populations.

Haplotypes and LD of the CYP2E1 gene region have been poorly studied. The aim of the present study has been to analyze polymorphisms across most of the CYP2E1 gene, document global ethnic variation in their allelic frequencies, and study the patterns that exist in haplotypes and LD. To those ends we present data on 11 polymorphisms, their frequencies and haplotypes in over 2,600 normal, healthy individuals from 50 population samples representing all major geographical regions of the world.

Results

A total of 2,657 mostly unrelated individuals (by self-report) were typed and analyzed for each of these polymorphisms. Allele frequencies and sample sizes for the 11 polymorphisms in all 50 populations can be found in ALFRED (http://alfred.med.yale.edu/) using the UIDs in Tables 1 and 2. Allele frequency ranges for each polymorphism are given in Figure 1, and the ancestral allele frequencies for the 10 SNPs and the most common allele frequency for the VNTR are given in Supplemental Table S1. There were no significant deviations from Hardy-Weinberg ratios. The average heterozygosities across 50 population samples and F_st values for 11 markers are shown in Supplementary Figure S1. For most markers the average heterozygosities are low, ranging from 0.035 (marker 7) to 0.283 (marker 10). F_st values vary around the mean of 0.14 for a standard set of 369 SNPs ²⁵ but are high at markers 10 and 11, 0.254, 0.231, respectively, at the 3′ end of the gene. Only seven of the eleven markers segregate in all populations. The derived allele frequencies of SNPs at exon 4 (marker 6) and exon 6 (marker 7) are very low outside of Africa and these derived alleles are completely absent in the populations of East Asia and the Americas. The derived alleles of the upstream SNPs, except rs6413420 (marker 5), are observed in higher frequencies in Asia and the Americas than in Africa or Europe.

Table 1.

Description of 11 polymorphic markers studied in the CYP2E1 gene.

Marker	Function	Polymorphism	dbSNP rs#	Site Location	Positiond (bp)	Base pairs to next SNP	ALFRED UIDa	ADeles	Ancestral allele
1	none proven	VNTR		5′ upstream	135,188,828	885b	SI01 40900	4 alleles	NAc
2		C_2431875_10;Pstl	rs3813867	5′ upstream	135,189,595	240	SI000693S	G/G	G
3		Rsal	rs2031920	5′ upstream	135,189,835	703	S1000694T	C/T	C
4		C_15867697_1D	rs2070672	5′ upstream	135,190,538	281	SI001473P	G/A	A
5		C_25594209_10	rs641342Q	5′ upstream	135,190,819	4,846	SI001475R	T/G	G
6	Val179lle	C_30443971_10	rs6413419	exon 4	135,195,665	1,722	SI001468T	G/A	G
7	lle321lle	C_7468401_10	rs915909	exon 6	135,197,387	330	SI001476S	T/C	C
8		C_30173803_10; Mspl	rs4646976	intron 6	135,197,717	817	SI000692R	A/G	A
9		Dral	rs6413432	intron 6	135,198,534	2,593	SI014089W SI014088V	A/T	A
10		C_16026001_20	rs2070676	intron 7	135,201,127	225		GIC	G
11	Phe421Phe	C_16026002_10	rs2515641	exon 8	135,201,352		SI000174Q	TC	C

VNTR has two common (6-, 8-repeats described by Hu el al.1999) alleles; two rare alleles observed in African samples.

^a“UIDs” are unique identifiers in the ALFRED database for polymorphism descriptions and allele frequencies.

^bDistance to next SNP (bp) from the proximal end of the VNTR

^cNot Available; Failed to identify ancestral allele for VNTR

^dNCBI Map Build 36.3

Table 2.

The 50 Population samples studied with unique identifier links to full descriptions in the ALFRED database.

Name	Abbrev.	Location	N	Population ALFRED UID	Sample ALFRED UID
Africa
Biaka	BIA	S.W. Central African Rep.	70	PO000005F	SA000005F
Mbuti	MBU	eastern Dem. Rep. Congo	39	PO000006G	SA000006G
Yoruba	YOR	western Nigeria	78	PO000036J	SA000036J
Ibo	IBO	southern Nigeria	48	PO000096P	SA000099S
Hausa	HAS	northern Nigeria	39	PO000097Q	SA000100B
Chagga	CGA	Kilimanjaro area, Tanzania	45	PO000324J	SA000487T
Masai	MAS	Northern Tanzania	22	PO000456P	SA000854R
Sandawe	SND	North Central. Tanzania	40	PO000661N	SA001773S
African Americans	AAM	United States	90	PO000098R	SA000101C
Ethiopian Jews	ETJ	northwestern Ethiopia ‡	32	PO000015G	SA000015G
Somali	SOM	Somalia; refugees in Pakistan	40	PO000075M	SA0021380
S.W. Asia, Europe
Yemenite Jews	YMJ	Yemen ‡	43	PO000085N	SA000016H
Druze	DRU	Israel	106	PO000008I	SA000047L
Samaritans	SAM	Israel	41	PO0000950	SA000098R
Ashkenazi	ASH	eastern Europe ‡	83	PO000038L	SA000490N
Adygei	ADY	Krasnodar, Caurasus Mfns	54	PO000017I	SA000017I
Chuvash	CHV	easternmost Europe near Urals	42	PO000327M	SA0004910
Hungarians	HGR	Hungary	92	PO000453M	SA002023H
Russians	RUA	Kargopol, Archangelsk region	34	PO000019K	SA001530J
Russians	RUV	Vologda, northern Russia	48	PO000019K	SA000019K
Finns	FIN	Finland	36	PO000018J	SA000018J
Danes	DAN	Denmark	51	PO000007H	SA000007H
Irish	IRI	Ireland	118	PO000057M	SA000057M
Euro Americans	EAM	United States	92	PO000020C	SA000020C
N.W. Asia (Siberia)
Komi Zyriane	KMZ	N.W. Asia, near Urals	47	PO000326L	SA000489V
Khanty	KTY	N.W. Asia near Urals	50	PO000325K	SA000488U
S.C. Asia
Mohanna	MHN	Pakistan	112	PO000708P	SA002139P
Hazara	HZR	Pakistan	58	PO000575R	SA002140H
Negroid Makrani	NMK	Pakistan	56	PO000707O	SA002137N
Keralites	KER	Kerala, India ↕	30	PO000672P	SA001854S
N.E. Asia (Siberia)
Yakut	YAK	Sakha, N.E. Siberia	51	PO000011C	SA000011C
Pacific Islands
Nasioi	NAS	Bougainville, Solomon Islands	23	PO000012D	SA000012D
Micronesians	MCR	Micronesia, multiple islands	37	PO000063J	SA000063J
East Asia
Laotians	LAO	Laos	119	PO000671O	SA001853R
Cambodians	CBD	Cambodia	25	PO000022E	SA000022E
SF Chinese	CHS	southern Han, SF Bay Area	60	PO000009J	SA000009J
TW Chinese	CHT	Taiwan	49	PO000009J	SA000001B
Hakka	HKA	Taiwan	41	PO000003D	SA000003I
Koreans	KOR	Seoul, Korea	54	PO000030D	SA000936S
Japanese	JPN	Japan	51	PO000010B	SA000010B
Ami	AMI	eastern mtns, Taiwan	40	PO000002C	SA000002C
Atayal	ATL	eastern mtns, Taiwan	42	PO000021D	SA00002ID
Americas
Cheyenne	CHY	Oklahoma, U.S.A.	56	PO000023F	SA000023F
Pima, Arizona	PMA	Arizona, United States	51	PO000033G	SA000025H
Pitna, Mexico	PMM	northern Mexico	53	PO000034H	SA000026I
Maya	MAY	central Yucatan, Mexico	52	PO000013E	SA000013E
Quechua	QUE	Peru	22	PO000069P	SA000069P
Ticuna	TIC	Amazon, Brazil	65	PO000027J	SA000027J
R. Surui	SUR	Rondonia, Amazon, Brazil	47	PO000014F	SA000014F
Karitiana	KAR	Amazon, Brazil	57	PO000028K	SA000028K

^‡Samples collected in Israel

^↕Samples collected in U.S.A. from individuals born in Kerala

Figure 1.

Graphical representation of the average and range of ancestral allele frequencies in 50 population samples for each of 11 markers at CYP2E1 in 50 population samples.

We inferred 16 common haplotypes and estimated their frequencies (Figure 2). Most of the low frequency variation in the residual class of rare haplotypes is accounted for by a relatively small number of haplotypes in the 2 to 4% frequency range. The variation in haplotype frequencies among populations gives rise to a complex pattern of LD, both pairwise and as segments with high LD, that varies among populations (Supplemental Table S2 and Supplemental Figure S2).

Figure 2.

Frequencies for the haplotypes based on 11 markers at CYP2E1 for 50 populations. For each color-coded haplotype the alleles are shown for the sites in chromosome order as numbered in Table 1. Both the full allelic description and lower-case letter codes for the haplotypes are given. For each population the proportional length of each color bar represents the frequency of the respective haplotype. All haplotypes that have frequencies less than 5% in all the populations studied are grouped into the residual (gray bar) class. The ancestral haplotype, GCAGGCAAGC, for markers 2 through 11, is not found at common frequencies in any of the 50 populations studied.

Haplotype diversity is much higher in Africa (with 6–10 common haplotypes) than outside of Africa (with about 1–6 common haplotypes). The most common 11-MARKER haplotype, 6GCAGGCATCC (dark green in Figure 2), is very frequent in all populations outside of Africa and in Ethiopia, but not in other populations of Africa. Two haplotypes, 6CTAGGCAACC (light yellow) and 8GCGGGCGTGT (light blue), are not seen in African populations and rarely seen (<5%) in European populations, but are more frequent in most East Asian (0.0 – 0.273 and 0.073 – 0.262) and Native American (0.065 – 0.443 and 0.023 – 0.184) populations.

In order to understand the evolution of the haplotypes we estimated haplotypes with fewer SNPs across shorter segments of the gene. We identified three core regions that have evolved common haplotypes solely by accumulation of mutations from the ancestral core haplotype. These cores involve markers 1 through 5 (core A), markers 6 through 9 (core B), and markers 10 and 11 (core C) (Figure 3). No recurrent mutations are required to explain all of these core haplotypes. The full 11-marker haplotypes can be explained by combinations of the haplotypes of the three cores (Figure 4 and Supplemental Table S3). These combinations have arisen by accumulation of mutations (as depicted in Figure 3) and historical crossovers. It is difficult to be certain of orders of all events, mutations and crossovers, when the three cores are considered together, in part because other combinations that could have been intermediate now are either absent or exist among the rare haplotypes.

Figure 3.

The evolutionary relationships among haplotypes of three core segments of CYP2E1. In all cases the schema starts with the ancestral human haplotype and gives the pattern of mutational accumulation for that core. A: Core A comprised of markers 1 through 5. B: Core B comprised of markers 6 through 9. C: Core C comprised of markers 10 and 11.

Figure 4.

The composition of the 11 full haplotypes in terms of combinations of individual core haplotypes. The core haplotypes are numbered as in Figure 3 with core A on the left, core B in the center, and core C on the right. The lower case letters for the full haplotypes correspond to those in Figure 2. (See also Supplemental Table S3.)

In contrast to the global frequency patterns of the whole 11-marker haplotypes, the individual core haplotypes show different global patterns (Supplemental Figures 3, 4, & 5). Core A haplotypes show greater frequency similarity between African and European populations than between European and both East Asian and Native American populations. One core A haplotype, 6GCAG (#1 in Figure 3A), exists at frequencies of 56% to 95% in the Africans and Europeans. Another core A haplotype, 6CTAG (#5 in Figure 3A), is not seen in Africans, is rare in Europeans, but is frequent in East Asian and Native American populations. To the degree these 5′ markers encompass the major regulatory regions, it is possible that East Asians and Native Americans may have a common derived variant in regulation that is very uncommon to absent in the rest of the world.

Discussion

We are unaware of any publications of the CYP2E1 gene that have included all of the polymorphisms that we present here. Certainly, none of these markers has been studied previously on such a large and ethnically diverse set of individuals. This study is an explicit example of the type of global perspective on pharmacogenetic variation within and among populations discussed in an editorial by Marsh²⁶. Even this dataset does not probe the full extent of the genetic diversity of this small segment of DNA. Public databases report multiple additional polymorphisms across the gene (including 5′ and 3′ UTRs).

We cannot precisely relate the 16 common haplotypes (Figure 2) we have observed to the standard CYP2E1 allelic designations in the “cypalleles” web site (http://www.cypalleles.ki.se/cyp2e1.htm) because, from a genetic transmission perspective, each of the haplotypes we report is an allele and the “cypalleles” web site does not give full haplotype specifications for the allele designations they summarize, precluding a strict comparison. Moreover, we have not included SNPs with rare or uncommon variants that have not been studied widely. To distinguish the haplotypes we have identified from those in the “cypalleles” nomenclature, we have used letter designations rather than numbers in Figures 2 and 4 and Supplemental Table S3. As an example of the difficulty of establishing precise correspondences, the mutation G→A at marker 6 (corresponding to 179 Val→Ile) defines core B haplotype 4 (Figure 3B) and appears to represent one mutational event. That core B haplotype exists in two combinations with core A haplotypes and two combinations with core C haplotypes for a total of three 11-marker haplotypes: g, k, and n (Figures 2 & 4). All three of these haplotypes correspond to allele CYP2E1*4 in the “cypalleles” nomenclature. We expect the haplotypes encompassing the gene to become more complex as more SNPs and rare variants are included in an even more comprehensive study of the gene.

In addition to multiple SNPs across this gene, copy number variation (CNV) encompassing CYP2E1 has been reported^27–27. Our typing methods are not designed to detect CNVs but we can exclude any common occurrence in our samples because there is no significant deviation from HW ratios in any of the populations.

We have studied the allele, haplotype, and LD variation patterns for 11 polymorphisms in 50 populations from different geographical regions of the world across the CYP2E1 gene and have shown that there are large differences in these patterns worldwide. The haplotypes were useful in inferring recombination events in the recent evolution of the gene. The current study focuses attention on the core haplotype lineages that appear to have involved no recombination and on the combinations that have arisen because of historical recombinations. These cores and their combinations provide the framework for future expression studies. Depending upon when in one of the evolutionary lineages a functional variant arose, we would expect it to either define a new sublineage or be inherited into the descendant haplotypes in Figure 3. Thus, the evolutionary lineages may explain multiple haplotypes (alleles) having similar functional properties, even if the causative variant has not yet been identified. Other SNPs within the molecular extent of the region spanned will likely fall within this framework; some might refine the locations of the inferred historical crossovers.

Supplementary information is available at the The Pharmacogenomics Journal’s website.

Materials and Methods

Samples studied

DNA was purified from lymphoblastoid cell lines from 2657 healthy adults from 50 populations from around the world (Table 2). Population membership was designated by the subjects and all blood samples were obtained with individual informed consent following protocols approved by the Institutional Review Boards at Yale University School of Medicine, at the University of Karachi, and at multiple other relevant institutions in countries where samples were collected. The average population sample size is 53 individuals.

Markers studied

We studied ten SNPs and one VNTR across 13.9 kb that encompasses the 5′ region of the CYP2E1 gene and almost the entire coding region (Figure 5). We typed the VNTR and four SNPs in the upstream region, three SNPs in the coding regions, and three SNPs in the intronic regions of the gene (Table 1). The markers are referred to by their numeric position (1–11) in Table 1. The SNPs are all diallelic and the VNTR is essentially diallelic, as initially described²⁹. Two other very rare VNTR alleles have been seen in some African populations in the course of this study (data not shown); they were excluded from the haplotype analyses.

Figure 5.

Map of CYP2E1 on chromosome 10 and the markers typed. The filled boxes represent the exons of the gene; the number below each box is the exon number. The vertical lines represent positions of the markers studied.

Typing methods

The samples were typed by TaqMan assays (markers 2, 4–8, 10, and 11), by fragment length analysis on agarose gels (marker 1) after PCR, and by restriction fragment length after enzyme digestion of the PCR products for markers 3 (RsaI) and 9 (DraI).

Determining Ancestral Alleles

For each allele, the ancestral state in humans was determined by inference from the allele present in several other primate species. The ancestral allele of the VNTR (marker 1) could not be determined but by inference is 6.

Statistical methods

Allele frequencies of the VNTR and SNPs were calculated by gene counting assuming co-dominant inheritance. All the sites were also tested for Hardy-Weinberg ratios by chi-square test and/or exact test. Expected heterozygosities were estimated as 1-Σp_i². Haplotype frequencies were estimated by the EM algorithm using HAPLO ³⁰. Haplotypes with estimated frequencies of less than 5% in each of the population samples go into the residual class. The 5% threshold is a reasonble boundary for determining what are the common and rare haplotypes given the sample sizes in this study. While some estimated haplotypes below the 5% threshold have very clear evidence of occurrence, the standard errors on estimated frequencies increase along with some erroneous inferences due to the small number of observations available in the rare zone and the fact that the LD levels between sites vary. Pair-wise LD estimates were done as r² [refs. ^{31, 32}] with significance levels determined by a permutation test³³. Comparative plots of LD for all the populations were done using HAPLOT ³⁴.

Abbreviations

CYP2E1

cytochrome P450, family 2, subfamily E, polypeptide 1

LD

linkage disequilibrium

PCR

polymerase chain reaction

SNP

single nucleotide polymorphism

UID

unique identifier

VNTR

variable number of tandem repeats