A functional ABCA1 gene variant is associated with low HDL-cholesterol levels and shows evidence of positive selection in Native Americans

Abstract

It has been suggested that the higher susceptibility of Hispanics to metabolic disease is related to their Native American heritage. A frequent cholesterol transporter ABCA1 (ATP-binding cassette transporter A1) gene variant (R230C, rs9282541) apparently exclusive to Native American individuals was associated with low high-density lipoprotein cholesterol (HDL-C) levels, obesity and type 2 diabetes in Mexican Mestizos. We performed a more extensive analysis of this variant in 4405 Native Americans and 863 individuals from other ethnic groups to investigate genetic evidence of positive selection, to assess its functional effect in vitro and to explore associations with HDL-C levels and other metabolic traits. The C230 allele was found in 29 of 36 Native American groups, but not in European, Asian or African individuals. C230 was observed on a single haplotype, and C230-bearing chromosomes showed longer relative haplotype extension compared with other haplotypes in the Americas. Additionally, single-nucleotide polymorphism data from the Human Genome Diversity Panel Native American populations were enriched in significant integrated haplotype score values in the region upstream of the ABCA1 gene. Cells expressing the C230 allele showed a 27% cholesterol efflux reduction (P< 0.001), confirming this variant has a functional effect in vitro. Moreover, the C230 allele was associated with lower HDL-C levels (P = 1.77 × 10⁻¹¹) and with higher body mass index (P = 0.0001) in the combined analysis of Native American populations. This is the first report of a common functional variant exclusive to Native American and descent populations, which is a major determinant of HDL-C levels and may have contributed to the adaptive evolution of Native American populations.

INTRODUCTION

It has been suggested that genetic susceptibility of Hispanics to type 2 diabetes (T2D), obesity and dyslipidemia is related to their Native American heritage (1–3). We recently found a frequent non-synonymous variant (R230C, rs9282541) within the ATP-binding cassette transporter A1 gene (ABCA1) associated with low high-density lipoprotein cholesterol (HDL-C) levels (the most common dyslipidemia in populations with Native American ancestry), obesity and T2D in Mexican Mestizos (4,5). ABCA1 plays a key role in cholesterol efflux and transfer from peripheral cells to lipid-poor apolipoprotein A1 (ApoA1), the first step in HDL particle formation (6,7).

The R230C variant was initially described in the Oji-Cree population (8). To date, it has been found only in Native American and Mexican-Mestizo populations (4). We performed a large-scale analysis including individuals from 36 Native North and South American groups and assessed the effect of this variant on anthropometric and metabolic traits. Because it was previously suggested that R230C may have conferred selective advantage as a thrifty gene and/or resistance against certain infectious diseases (4), we performed a more thorough analysis seeking evidence of positive selection.

RESULTS

The C230 allele was present in the majority of the Native American populations at an average frequency of 12% (range 0–31%) (Fig. 1; Supplementary Material, Table S1), but was absent from 863 additional individuals belonging to different European (Spaniard and Dutch) and Asian groups (Han Chinese, Manchu, Mongolian, Siberians and Eskimos) (Fig. 1). The distribution of this allele was not structured according to language or geographic groups (North versus South America), as evidenced by analysis of molecular variance (AMOVA) (P = 0.978 and 0.895, respectively). However, the C230 allele frequency increased at tropical latitudes (between the tropics of Cancer and Capricorn) and gradually decreased at higher latitudes both to the North and South (Fig. 1), showing a significant correlation (r² = 0.328; P = 0.02).

Figure 1.

Frequency distribution of the C230 allele [(%) black-shaded area] in Native American, European, Asian and African populations. C230 frequency data from populations with an asterisk were obtained from previous reports (8–11 and HapMap and SNP500CANCER databases).

The C230 allele is located on a single genetic haplotype

To perform a phylogenetic reconstruction of the evolutionary relationships, 15 additional single-nucleotide polymorphisms (SNPs) within a 50 kb block were analyzed in 20 Native American and 25 Mexican-Mestizo trios to define haplotype blocks within the region. Together with data from HapMap populations, a total of 58 haplotypes were identified (Supplementary Material, Table S2). Seven haplotypes were found in Native Americans, and the C230 allele was clearly found in only one genetic block (haplotype 32) in all Mexican-Mestizo, North and South American native individuals analyzed. Maximum parsimony (MP)-based network analysis is shown in Figure 2. The phylogenetic reconstruction uncovered two major lineages (haplogroups A and B) defined by the non-synonymous polymorphism R219K within the ABCA1 gene. The C230 allele occurred in haplogroup B characterized by the ancestral R219 allele.

Figure 2.

MP-based network describing the evolutionary relationships of 11 distinct haplotypes. Native American and Mexican-Mestizo haplotypes were established in trios; the remainder were inferred from HapMap groups. Each haplotype is represented by a circle whose area reflects the overall number of copies observed and whose color-coding indicates the frequency of the haplotype in the HapMap groups and Native Americans and Mexican Mestizos. Line length is proportional to the number of differences between haplotypes. Non-filled circles represent non-sampled haplotypes reconstructed by the MP algorithm as evolutionary intermediaries between observed haplotypes. The phylogenetic reconstruction uncovered two major lineages (haplogroups A and B) defined by the non-synonymous polymorphism R219K. The R230C variant occurred on haplogroup B characterized by the ancestral R219 allele, which is frequent in Europe, Asia and America but infrequent in African populations. The C230 allele was found in only one genetic block in all Mexican-Mestizo and Native American individuals analyzed.

Positive selection testing

The long-range haplotype (LRH) test showed that the extension of linkage disequilibrium (LD) was much longer in C230 than in non-C230 chromosomes [relative extended haplotype homozygosity (REHH) = 9.8 and 5.9 at 365 and −434 kb from the core, respectively; Fig. 3]. The former value remained significant compared with REHH values from additional regions elsewhere in the genome, genotyped in a similar set of Native American populations (P = 0.036 at 300 kb from the core and P = 0.021 at a marker H of 0.04). REHH values were significant in both Kichwa (P = 0.018 at a distance of 300 kb and P = 0.007 at a marker H of 0.04) and Nahua populations (P = 0.043 at a distance of 300 kb and P = 0.021 at a marker H of 0.04). The only other two significant core haplotypes were located upstream the ABCA1 gene (Supplementary Material, Table S3). Furthermore, in Native Americans from the Human Genome Diversity Panel (HGDP) (R230C genotypes not available), the ABCA1 5′ region (∼75 kb upstream R230C) was clearly enriched for outliers of the integrated haplotype score (iHS) statistic genome-wide distribution (iHS > 2.5) (Fig. 4). In agreement with the REHH analysis, the highest iHS values are clustered upstream ABCA1.

Figure 3.

EHH and REHH of ABCA1/R230C and 23 additional SNPs × physical distance in Native American individuals. C230-bearing chromosomes (in gray) appear to have greatly extended LD compared with non-C230-bearing chromosomes (in black). Gray vertical lines in the upper track represent genotyped SNPs, and boxes indicate annotated genes.

Figure 4.

iHS values for individual SNPs flanking the ABCA1 region (2 Mb) × physical distance (top panel) and P-values for the composite likelihood ratio (CLR) test based on a 31-SNP sliding window analysis to detect local regions enriched for high iHS values (bottom panel) in the combined Native American sample. Filled circles indicate P-values <0.05 genome-wide significance level.

Association of R230C with HDL-C levels and other metabolic traits

Overall, the prevalence of hypoalphalipoproteinemia (HA) was the most common dyslipidemia (65% in Mexican and South American natives; Supplementary Material, Table S4). Table 1 shows the effect of R230C on HDL-C and total cholesterol levels and body mass index (BMI). The R230C/C230C genotypes were significantly associated with low HDL-C levels in Pimas (P = 6.4 × 10⁻⁵) and in the combined analysis of eight Mexican native groups (P = 5.3 × 10⁻⁸). In South American native groups, biochemical data were available from two populations (Parkatejés and Kichwas), and although HDL-C levels were lower in C230 carriers, the differences did not reach statistical significance. Altogether, the combined results of all Native American groups showed a highly significant effect of the C230 allele (−4.2% per copy, P = 1.77 × 10⁻¹¹). Interestingly, differences in the effect of R230C on lipid profiles were observed in some Native American populations. R230C was strongly associated with low total cholesterol and triglyceride levels in Pimas (P = 1.8 × 10⁻⁷ and P = 7.0 × 10⁻⁴, respectively) and Mayans (P = 0.004 and 0.010, respectively). Although the combined analysis in Native American groups also showed a significant association with low total cholesterol levels (P = 7.15 × 10⁻⁵), it was clearly not as significant as the association with low HDL-C levels (P= 1.77 × 10⁻¹¹). The C230 allele was associated with higher BMI in the combined analysis of Mexican native groups (P = 0.012), in Native South American populations (P = 0.0003 and 0.050 for Parkatejé and Kichwas, respectively), but not in Pima Indians. Altogether, combined results showed that the C230 allele was associated with higher BMI (P = 0.0001), and this association was evidently more significant in male than in female individuals (P = 4.05 × 10⁻⁶ versus P = 0.02).

Table 1.

Association of R230C with lipid levels and BMI in Native American populations

Native American population (n)	HDL-C levels		Total cholesterol		BMI
	Effect (SE)	P-value	Effect (SE)	P-value	Effect (SE)	P-value
North America
USA
Pima (2563)a	−0.075 (0.019)	6.4 × 10−5	−0.071 (0.014)	1.8 × 10−7	0.008 (0.015)	0.586
Mexico
Yaquis (45)	—	—	—	—	0.044 (0.018)	0.012
Teenek (67)	−0.051 (0.026)	0.057	−0.010 (0.024)	0.671	0.001 (0.016)	0.978
Coras (123)	−0.033 (0.014)	0.021	−0.006 (0.013)	0.681	0.032 (0.011)	0.006
Purepechas (15)	−0.040 (0.074)	0.603	−0.048 (0.046)	0.333	0.034 (0.019)	0.097
Mazahuas (83)	−0.039 (0.036)	0.281	−0.031 (0.041)	0.444	0.006 (0.019)	0.758
Nahuas (267)	−0.040 (0.014)	0.014	−0.014 (0.020)	0.470	−0.004 (0.008)	0.617
Totonacas (113)	−0.028 (0.021)	0.180	−0.031 (0.015)	0.042	0.022 (0.013)	0.085
Zapotecs (106)	−0.047 (0.022)	0.038	0.007 (0.019)	0.723	0.007 (0.013)	0.605
Mayans (110)	−0.040 (0.017)	0.023	−0.043 (0.015)	0.004	−0.007 (0.013)	0.554
Mexican natives combined	−0.038 (0.007)	5.3 × 10−8	−0.019 (0.007)	0.027	0.010 (0.004)	0.012
South America
Kichwas (79)	−0.043 (0.030)	0.153	−0.005 (0.020)	0.791	0.024 (0.012)	0.050
Parkatejé (78)	−0.029 (0.026)	0.270	−0.002 (0.030)	0.945	0.046 (0.012)	0.0003
All Native Americans combined	−0.042 (0.006)	1.77 × 10−11	−0.021 (0.006)	7.15 × 10−5	0.011 (0.003)	0.0001

Effect values are presented as effect size per C230 allele copy, standard error (SE). Linear regression was performed on the basis of log-transformed values for HDL-C levels (mg/dl), total cholesterol levels (mg/dl) and BMI (kg/m²), adjusting for age, gender and diabetes status. HDL-C and total cholesterol levels were also adjusted for BMI.

^aP-value adjusted by age, gender, birth year, diabetes status and family membership.

Sequencing and in vitro functional analysis

To rule out the presence of another possible causal variant in LD with C230, all 50 exons and the promoter region of ABCA1 were sequenced in a limited number of individuals (2 of each genotype); however, no promoter or coding variant in LD with R230C was found. Cholesterol efflux from Flp-In cell lines expressing the ABCA1 C230 allele was significantly lower (27%) than that of cells expressing the wild-type R230 allele (P < 0.001) (Fig. 5A), confirming that this variant has a functional effect in vitro. In contrast, the C230 and R230 cell lines showed no differences in phospholipid efflux. The C230 allele expressed ABCA1 protein at levels comparable with that of T1091 and wild-type alleles (Fig. 5B).

Figure 5.

Functional characterization of the ABCA1/R230C variant by lipid efflux assay. (A) Polyclonal stable cell lines expressing the ABCA1 wild-type (WT), variant R230C and mutant M1091T (known defective in lipid efflux) were generated, and efflux activity for cholesterol was performed as described in Materials and Methods. Data represent mean ± SD of two to five experiments as a percent of the ApoA1-dependent efflux induced by wild-type ABCA1. Each assay was performed in triplicate. *P < 0.001. (B) The expression of WT, variant R230C and mutant M1091T ABCA1 protein in Flp-In cells was assessed by western immunoblotting.

DISCUSSION

R230C, a private allele to the Americas

The R230C allele first identified in Oji-Crees and Mexican Mestizos was found in most Amerindian groups throughout the Americas, but not in any ethnic group from other continents (4,8). This is in agreement with previous studies that have not found this allele in 7717 Caucasian, Asian, African and South-Pacific Rim individuals (HapMap and SNP500CANCER databases) (8–11). This absence strongly suggests that C230 is a private allele (exclusive to Native American and Native American-derived populations), although it may be present in some non-Amerindian populations not included in this analysis. Its presence on the same haplotype in both North and South Americans suggests that it may have arisen among Native American founders in Beringia or North-East Asia. This is in agreement with recent studies suggesting that founder populations stayed in Beringia long enough to give rise to exclusive genetic variants (12–14). Although private alleles in the HLA system and other genes have been previously reported in some Native American populations (15,16), there is only one previous report of a common private autosomal allele (microsatellite D9S1120 9RA) ubiquitous in the Americas (17). It is noteworthy that D9S1120 and R230C (ABCA1) are both located on chromosome 9q, although separated by a 19 Mb distance. We genotyped DS9S1120 in 16 C230C Native American and Mestizo homozygotes and found no allele in LD, indicating that the two ancestry informative markers are independent.

The C230 allele distribution varied among different Amerindian populations (0–0.31). The complex demographic processes that these populations have gone through must have played a crucial role in this distribution. Initially, the moderate bottleneck in the out of Beringia process led to a relatively small effective population size, so genetic drift could have been one of the main causes of fluctuation (18,19). In groups that later expanded demographically to constitute societies formed by thousands of individuals such as Mesoamericans, genetic drift was less likely to cause differences in the distribution of C230 frequencies.

Evidence suggesting R230C underwent positive selection

Understanding the impact of natural selection acting on particular genes in human populations can provide insights into the genetic etiology of human disease. Interestingly, ABCA1 has been recognized as one of the genes most likely to have been subject to positive selection in humans since the divergence from the common ancestor of our lineage and that of chimpanzees (20,21). The results of the REHH and iHS analyses for the ABCA1 gene region in Native Americans are not compatible with a simple neutral evolutionary model, but are consistent with the hypothesis that the R230C variant resides on a haplotype which is the target of an ongoing directional selective sweep. It must be acknowledged, however, that with the currently available genotyping data, it is not possible to define whether the R230C haplotype is also responsible for the signal resulting from the iHS test.

The geographical distribution of the C230 allele clearly differs from the North-to-South gradient described for genome-wide neutral markers (22), suggesting the possibility of a climate-related adaptive process, as has been previously described for other genes involved in energy metabolism (23). In the context of Neel's hypothesis (24), R230C carriers could have had a selective advantage. Because the C230 ABCA1 protein shows decreased cholesterol efflux, the presence of this variant could favor intracellular cholesterol and energy storage. Specifically, adipose tissue benefits various biological functions including the ability to accommodate fluctuations in energy supply such as severe famine, the regulation of reproductive function and providing energy for the immune system now known to have a significant energy cost (25,26). However, under current westernized lifestyle changes, this allele may have become a major susceptibility allele for low HDL-C levels and other metabolic traits, which is consistent with the association of the R230C variant with higher BMI in Native American populations, and with obesity, T2D and metabolic syndrome in Mexican Mestizos (4,5). However, other environmental factors may also be involved in C230 allele frequency distribution. For instance, cholesterol plays an important role in various infectious processes such as the entry and replication of Dengue virus type 2 and flaviviral infection (27). The ABCA1 transporter is known to participate in infectious and/or thrombotic disorders involving vesiculation, since homozygous ABCA1 gene deletions confer complete resistance against cerebral malaria in mice (28,29). Interestingly, areas with higher C230 allele frequencies correspond to dengue, yellow fever and malaria distributions in the Americas (30). Altogether, these different lines of evidence suggest that the ABCA1 C230 allele may have been important for survival throughout the colonization of the Americas.

Association of R230C with HDL-C levels and other metabolic traits

Overall, the prevalence of low HDL-C levels was not only higher in Native Americans than in European, Asian and African individuals (3), but also the most common dyslipidemia (65% in Mexican and South American natives). The Pima population is known to have much lower HDL-C and total cholesterol levels than US Caucasians (31). R230C/C230C genotypes were strongly associated with low HDL-C levels in Native American rural populations, Pimas and urban Mexican Mestizos (4,9). In fact, the sole presence of the C230 allele explains ∼4% of the HDL-C level variation in these populations, which is higher than the variation explained by any other SNP associated with HDL-C levels identified through genome-wide scans in Europeans and Indian Asians (9). This is consistent with both in silico (PANTHER subPSEC score −4.27) and in vitro evidence confirming that the R230C variant is functional (27% decrease in cholesterol efflux) (22). The functional effect is significant, but mild compared with the T1091 allele previously identified in Tangier patients (32).

Environmental factors and further genetic variation (within ABCA1 or other genes) may play a relevant role in the association of the C230 allele with other metabolic traits. Lower total cholesterol and triglyceride levels were found in C230 carriers only in Pimas and Mayans. In addition, the C230 allele was associated with higher BMI in Mexican native groups, but not in the Pima Indians. Moreover, a gender effect was observed, as the association of R230C with higher BMI was more significant in males. Interestingly, in a previous study, a transcription factor 7-like 2 (TCF7L2) gene haplotype (HapA) with evidence of positive selection was also associated with higher BMI only in male individuals (33). Further studies are required to confirm the role of R230C in these metabolic and other fat storage-related traits such as non-alcoholic fatty liver disease, which is highly prevalent in Hispanic populations (34,35).

The C230 allele has also been associated with T2D in the Mexican-Mestizo population (5). The overall frequency of T2D in most Mexican native groups was also high (11.3%); however, the study design was not appropriate for a case–control association in these groups. Interestingly, the R230C was only marginally associated with T2D in Pimas (P = 0.06) despite the large sample size and the previous finding that HDL-C concentrations in non-diabetic Pima Indian women were negatively associated with the development of T2D (36). Impaired ABCA1 function causes cholesterol accumulation in beta cells in animal models, suggesting that beneficial reductions in plasma lipids may limit the extent of beta cell damage and could partially mask glucose homeostasis disturbances (37). The highly significant association of R230C with reduced total cholesterol and triglyceride serum levels observed in Pimas may be one of the factors explaining this marginal association. The role of R230C as a risk allele for T2D in Mexican native groups and its interaction with environmental factors requires further analysis.

In conclusion, to the best of our knowledge, this is the first report of a common functional variant exclusive to Native American and descent populations associated with low HDL-C levels and other metabolic traits. We present several lines of evidence in favor of positive selection for the R230C allele possibly contributing to the adaptive evolution of Native American populations and providing insight into the genetic etiology of currently prevalent metabolic disease.

MATERIALS AND METHODS

Subjects

The study included a total of 4405 adult individuals from 36 different Native American groups and 863 Europeans and Asians. All Mexican and South American natives and their ancestors (two generations) were born in the same community and spoke their own native language. Field research was conducted by multidisciplinary teams.

Ethics statement

This study was conducted according to the principles expressed in the Declaration of Helsinki and was approved by the Ethics Committees of all participant institutions. Participants provided written informed consent. Local authorities gave their approval to participate in the study, and a translator was used as needed.

Anthropometric and biochemical analyses

Anthropometric and metabolic parameters were available for 2563 Pimas, 1016 Mexican and 157 South American native individuals (Supplementary Material, Table S4). The Pima subjects included are part of an ongoing longitudinal study of the etiology of T2D in the Gila River Indian community in Central Arizona (38). All biochemical measurements in 1050 Mexican natives and Kichwas (from Ecuador) were performed by the INCMNSZ with commercially available standardized methods as described by Villarreal-Molina et al. (4). Biochemical parameters of Parkatejé individuals have been previously described (39). T2D and HA were defined according to the American Diabetes Association and National Cholesterol Education Program (NCEP) criteria, respectively (40,41).

DNA sequencing of the ABCA1 gene

Genomic DNA was extracted from peripheral blood leukocytes. The 50 exons and proximal promoter region of the ABCA1 gene were amplified in samples from six individuals (two R230R, two R230C and two C230C) as described previously (8). Amplicons were sequenced using ABI PRISM BigDye Terminators version 3.1 on an ABI 3100 automated sequencer according to the manufacturer's protocol (Applied Biosystems, Foster City, CA, USA).

SNP genotyping

The R230C variant and 23 SNPs spanning an 800 kb region were genotyped using TaqMan assays (ABI Prism 7900HT Sequence Detection System; Applied Biosystems). The 23 SNPs were selected from the HGDP for being informative in Native American populations and were genotyped in 10 Kichwa, 7 Nahua and 3 Zapotec trios (mother–father–offspring). The names and chromosomal position of all SNPs analyzed are given in Supplementary Material, Table S5. Genotyping call rate exceeded 95% per SNP, and no discordant genotypes were observed in 40 duplicate samples. Deviation from Hardy–Weinberg equilibrium was not observed for any SNP.

Generation of R230C variant constructs and cell lines

Polyclonal stable cell lines expressing the ABCA1 R230C variant were generated using the Flp-In system (Invitrogen, Carlsbad, CA, USA) as described previously (42). The generation and detailed biochemical characterization of many of these cell lines are described elsewhere (43). Briefly, the R230C variant was generated by PCR-based site-directed mutagenesis using the primers 230F,5′-GAGCGAGTACTTTGTTCCAACATG and 230R,5′-CATGTTGGAACAAAGTACTCGCTC and cloned into pcDNA5/FRT (Invitrogen). The plasmid was completely sequenced prior to transfection into 293 Flp-In cells. The M1091T ABCA1 mutation previously identified in Tangier patients was used as control (32).

Cholesterol and phospholipid efflux assay

Efflux experiments were performed as described previously (42). Briefly, cells were loaded with 1 µCi of [³H] cholesterol or 2 µCi of [³H]-choline (Amersham Biosciences, Little Chalfont, Buckingham, UK) for 24 h. The following day, the medium was removed and replaced with serum-free medium containing 5 mg/ml delipidated bovine serum albumin (Sigma, St Louis, MO, USA). After 1 h of incubation, 20 μg/ml human ApoA1 (Athens Research and Technology, Athens, GA, USA) was added for 4 h. For cholesterol efflux, the medium was collected and cells were lysed in 0.1 N NaOH/0.1% SDS. For phospholipid efflux, the [³H]-choline/phospholipids/ApoA1 in the medium was collected by immunoprecipitation with ApoA1 antibody and cells were digested for protein assay. The radioactivity in the samples was quantified by scintillation counting. Cholesterol efflux is expressed as a percent of counts in medium over total (medium + cells). Phospholipid efflux is expressed as counts of the immunocollected [³H]-choline/phospholipid/ApoA1 normalized to cell protein. Data are expressed as percent of the ApoA1-dependent efflux induced by wild-type ABCA1. Significance was calculated using a one-way ANOVA test with a Newman–Keuls post-test using GraphPad Prism 4 software (San Diego, CA, USA). ABCA1 expression was determined by western blotting as described previously (44), using anti-ABCA1 or anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH) antibodies (Chemicon, Temecula, CA, USA).

Statistical analyses

Population genetics

Allele and genotype frequencies, Hardy–Weinberg equilibrium and AMOVA were calculated using Arlequin 3.11 software (45). The linguistic classification of Native American languages was adopted from Campbell (46). Network analyses of haplotypes from HapMap, Mexican-Mestizo and Native American population data were performed using a median-joining and maximum-parsimony method (Network 4.510 software) (47).

Positive selection tests

The LRH test was applied to examine the decay of LD (Sweep software) (48) within an 800 kb region flanking R230C using data obtained from the 20 Native American trios described earlier. LD decay was then compared with LRH data generated elsewhere covering a total of 24 Mb of the genome in Native American populations (49). To further explore the presence of positive selection signatures in the ABCA1 region, the iHS was estimated as described previously (50,51), using publicly available genotype data for ∼650 000 SNPs genome-wide distributed in five Native American groups from the HGDP (52). With an approach similar to that described by Nielsen et al. (53) to detect regions with aberrant allele frequency spectra (test 1), we applied a composite likelihood test to detect regions with aberrant ‘iHS spectra’. We first categorized |iHS| in bins of size 0.1 and then estimated the probability of observing an SNP in each bin, both in the whole-genome data set (background distribution) and in each 31-SNP sliding window over the whole genome. Two composite likelihoods were estimated for each window, multiplying the probability of observing each SNP in the window, by either the probability of observing the SNP in the genome-wide background or that estimated from the window. A log-likelihood ratio was then estimated comparing both likelihoods, where extreme values indicate unusual iHS patterns compared with the rest of the genome.

Associations with HDL-C and metabolic traits

Associations of R230C genotypes with HDL-C and other metabolic traits were tested using linear regression models (assuming an additive model) adjusting for covariates including age, sex and BMI (SPSS, version 15.0, statistical package; Chicago, IL, USA). All variables tested were log-transformed for the analysis. Combined association tests were conducted using a Mantel–Haenszel-like model (54). The combined estimated effect was computed as a weighted average of the individual estimated effects using weights proportional to the inverse of the standard errors squared (33).

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG online.

FUNDING

This research was supported by grant 69856 from the Consejo Nacional de Ciencia y Tecnología (CONACYT) México, and partly supported by grant 660 from the Fundación Mexicana para la Salud-Silanes; by the Intramural Research Program of the National Institute of Digestive and Kidney Diseases (NIDDK), NIH; and by a grant of the Canadian Institutes of Health Research (CIHR). V.A.-A., T.F.-D., M.V.-C. and L.J.-A. were supported by fellowships from CONACYT, México. J.K.K. was supported by fellowships of the Michael Smith Foundation for Health Research (MSFHR) and the CIHR. M.R.H. holds a Canada Research Chair in Human Genetics.

WEB RESOURCES

The URLs for data presented herein are as follows:

Online Mendelian Inheritance in Man (OMIM), http://www.ncbi.nlm.nih.gov/Omim/

HapMap database, http://www.hapmap.org/

SNP500CANCER, http://snp500cancer.nci.nih.gov

HGDP-CEPH database, http://hgdp.uchicago.edu/cgi-bin/gbrowse/HGDP

Arlequin 3.11 software, http://cmpg.unibe.ch/software/arlequin3/

Network 4.510 software, http://www.fluxus-engineering.com

Sweep software, http://www.broad.mit.edu/mpg/sweep/index.html