FTO Genetic Variation and Association With Obesity in West Africans and African Americans

The study included two sets of participants: 1) 968 unrelated African Americans (estimated African ancestry 0.78 ± 0.11) enrolled from the Washington, D.C., metropolitan area in the U.S. as part of the Howard University Family Study, and 2) 517 unrelated West Africans enrolled as control subjects as part of the Africa-America Diabetes Mellitus (AADM) Study. After obtaining written informed consent, participants underwent an interview followed by a physical examination during which weight, height, and WC were measured using standard methods. BMI was computed as weight (measured in kilograms) divided by the square of the height (measured in meters). Body composition was estimated using bioelectric impedance analysis with validated population-specific equations as previously described (8). PFM was calculated as (fat mass/weight)*100. DNA was extracted from buffy coats using PureGene kits (Gentra).

Based on the International HapMap Project (HapMap) YRI data, tag single nucleotide polymorphisms (SNPs) at a pairwise r² >0.8 and with a minor allele frequency (MAF) ≥0.02 were selected for genotyping. The resulting 273 SNPs were genotyped as part of a custom Illumina panel using the Illumina GoldenGate Assay. Of the 273 SNPs, 264 were successfully genotyped, giving a locus success rate of 96.7%. The genotype call rate was 99.32%. The concordance rate for blind duplicates was 99.98%. Two SNPs deviated significantly (P < 0.001) from Hardy-Weinberg equilibrium and were excluded from further analysis, leaving 262 SNPs.

MAFs were computed and LD visualized using Haploview (9). Association with obesity traits was tested under an additive model with adjustment for age and sex using PLINK version 1.06. Potential population stratification was accounted for in each sample by adjusting for the first principal component derived from a set of 142 ancestry-informative SNPs genotyped in the African American subjects and by adjusting for ethnic group among the West African subjects. Each sample was analyzed separately. Then, combined analysis was done using a meta-analysis technique that computes weighted statistics for association and tests for heterogeneity, implemented in METAL (Meta Analysis Helper, available through the University of Michigan). Previous European studies have estimated that the FTO variant explains ∼1% of the phenotypic variance in BMI (1,–3). The present study has ∼88% power (for the African American sample) and ∼63% power (for West African sample) to explain ∼1% of the variance in BMI for a SNP with an MAF of at least 0.05 at a two-tailed α level of 0.05.

Given the strong prior information about the role of FTO variation in obesity, we considered our evaluation of the association between intron 1 SNPs and obesity a replication study; thus, nominal P values ≤0.05 were considered significant. For SNPs in the rest of the gene, tests of associations could be considered discovery rather than replication for which a Bonferroni-corrected P value threshold of 0.0002 (0.05/209 SNPs) would be significant.

The characteristics of the study subjects are shown in Table 1. The African Americans weighed significantly more and had higher BMI, WC, and PFM than the West Africans. MAFs were generally similar between the two samples (supplementary Table 1, available in an online appendix at http://care.diabetesjournals.org.cgi/content/full/db09-1252/DC1). However, seven (2.7%) SNPs showed a difference (δ) in allele frequencies ≥0.1, and 88 (33.6%) SNPs showed a δ ≥ 0.05 (supplementary Fig. 1).

The LD structure across the gene in both study samples is shown in Fig. 1. Using the CI method of Gabriel et al. (10) to construct haplotype blocks, the African American sample had 59 haplotype blocks with 138 SNPs outside blocks while the West African sample had 58 blocks with 142 SNPs outside blocks. The LD structure in the two study samples is more fragmented than European or Asian continental populations (represented by the HapMap CEU and JPT/CHB samples, respectively) (Fig. 1 and supplementary Fig. 2). This pattern of low LD was also observed for the intron 1 region surrounding rs9939609, the most consistently reported obesity-associated FTO SNP (supplementary Fig. 3). Using a common set of polymorphic SNPs, the proportion of SNPs exceeding specific r² thresholds of 0.2, 0.4, 0.6, and 0.8, respectively, in the HapMap CEU, CHB, and JPT samples were approximately threefold higher than those in the present study and the HapMap YRI sample (P < 0.0001, supplementary Fig. 2, and supplementary Table 2). In contrast, there were no significant differences in these proportions between the West African and African American samples in the present study or between these and the HapMap YRI sample.

Table 2 shows the SNPs displaying P ≤ 0.01 in either study population or in the meta-analysis. The top scoring SNPs for BMI were in intron 8 (Fig 2). For WC and PFM, the most significantly associated SNPs were in intron 8 or intron 1 (Table 2, Fig 2). The effect sizes for BMI were 0.9–1.7 kg/m² in West Africans, ∼1–1.6 kg/m² in African Americans, and ∼0.8 kg/m² in the meta-analysis for the two significant SNPs showing consistent direction of effect. The effect sizes for WC and PFM are shown in Table 2. Some SNPs (one for BMI, one for WC, and three for PFM) showed significant heterogeneity of association between the two study samples (Table 2). Haplotype analysis around these top-scoring SNPs showed that none of the haplotypes had a stronger association with any of the traits than the single SNPs they contained (data not shown). The only nonsynonymous coding SNP, rs16952624 (A405V), in this study (MAF 0.041 in West Africans, 0.024 in African Americans) showed no association with any of the obesity phenotypes in the West African (P 0.52–0.70) or African American sample (P 0.18–0.67). We note that none of the discovery SNPs in intron 8 reached statistical significance at a P value threshold of 0.0002 after correcting for multiple comparisons using the conservative Bonferroni-correction method.

We tested SNPs in the FTO gene previously reported to be associated with obesity in populations of European ancestry, including rs9939609, rs1121980, rs17817449, and rs7204609. The minor allele frequencies in our samples and the P values for association are shown in Table 2. We replicated the association with obesity for both rs1121980 and rs7204609 at P ≤ 0.05 in the West African sample but not in the African American sample. Similarly, we looked for association in the set of 16 SNPs tested in the only published African sample to date (11) and found that rs7204609, rs17817288 and rs12447107 showed significant association with PFM in our West African sample (supplementary Table 4). As in the Gambian study (11), these SNPs did not show association with BMI in this study.

One of the potential advantages of studying an African population is the opportunity to fine-map loci that have been identified in populations with stronger LD. We therefore looked for evidence of association of SNPs with an r² > 0.2 with the strongest reported FTO variant (rs9939609) in the HapMap CEU. Only one of 13 such SNPs (rs8055197, r² = 0.58 in HapMap CEU) showed a small P value in the present study (P = 0.012 for BMI and P = 0.015 for WC, both in African Americans). Conversely, none of the SNPs in Table 1 showed strong LD with rs9939609 in the present study (maximum r² = 0.337 among West Africans, 0.223 among African Americans, (supplementary Table 5).

The FTO gene is one of the few genes to show consistent association with human obesity, particularly in populations of European ancestry (1,12,,,–16), with few exceptions (17). In contrast, evidence from studies of other continental populations has been less consistent. Some studies of multiple Asian populations did not confirm this association (18,–20) while other studies did (21,22). There are few studies from populations of African origin, and the data from such studies have been largely negative. In the first study that included African Americans, Scuteri et al. (1) did not replicate the FTO association with obesity in a sample of 1,100 African Americans, neither did two other studies (11,23). The first study that found association between FTO variants and obesity in an African-ancestry population was a study of childhood obesity (24). More recently, Wing et al. (25) showed that rs8050136 and rs9939609 were associated with BMI and WC among African Americans in the Insulin Resistance Atherosclerosis Study (IRAS) Family Study. The only study so far of a sub-Saharan African population (11) was negative. Potential explanations for these differences between populations include differences in sample size, allele frequency differences, different LD patterns, and heterogeneity (as demonstrated for a few variants in this study).

In the present study, we tagged the entire span of the FTO gene (not just previously associated intron 1 SNPs) in two populations of African ancestry (African Americans and West Africans) to investigate association with obesity in these populations. To our knowledge, the set of 262 tag SNPs we studied is the largest set of SNPs to be directly genotyped in the FTO gene in any single study to date. Multiple SNPs in intron 8 (in addition to intron 1 SNPs) showed significant nominal association with the obesity phenotypes in the present study. This finding is significant because, to date, the focus on FTO variants associated with obesity has been, for the most part, on the specific SNPs or other intron 1 SNPs reported in the initial genome-wide association study. When significant associations are found within a gene, replication studies in other populations (especially those of a different continental ancestry) that scan the entire gene (rather than just one or a few SNPs) may lead to the discovery of other important genetic variants. Consistent with expectations, we found that both the West African and African American study samples showed weaker LD patterns across the gene when compared with other continental (e.g., European and Asian) populations. Studying populations of different ancestries (especially those with smaller LD) could help fine-map disease or trait loci. However, since the rs9939609 association in intron 1 of FTO reported in Europeans did not replicate in this study (as well as several other studies of African ancestry populations), African populations may not be the optimum choice to fine-map this locus.

The overall evidence from this study of populations with a majority of African ancestry adds to the growing body of knowledge supporting the role of FTO variants in obesity across multiple human populations. The location of most of these significant SNPs in intron 8 and the downstream region (rather than intron 1) suggests that there may be more than one genetic variant within FTO influencing obesity in humans. Further studies are needed to confirm, refine, and extend these findings.

The Howard University Family Study (HUFS) was supported by grants S06GM008016-380111 to A.A. and S06GM008016-320107 to C.R., both from the NIGMS/MBRS/SCORE Program. Support for the Africa-America Diabetes Mellitus (AADM) study was provided by National Institutes of Health (NIH) Grant 3T37TW0041-03S2 from the National Center for Minority Health and Health Disparities, as well as other NIH institutes (NHGRI and NIDDK through grant DK-54001). Participant enrollment for the HUFS was carried out at the Howard University General Clinical Research Center, which is supported by grant 2M01RR010284 from the National Center for Research Resources (NCRR), a component of the NIH. This research was supported in part by the Intramural Research Program of the Center for Research on Genomics and Global Health (CRGGH). The CRGGH is supported by the NHGRI, the NIDDK, the Center for Information Technology, and the Office of the Director at the NIH (Z01HG200362).

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

No potential conflicts of interest relevant to this article were reported.

Parts of this study were presented in abstract form at the 59^th American Society of Human Genetics Meeting, Honolulu, Hawaii, 24 October 2009.

The authors thank participants, physicians, and investigators in the HUFS and Africa America Diabetes Mellitus (AADM) study.