Several susceptibility loci for high BMI and obesity have recently been identified by genome-wide association (GWA) analyses of both large population-based samples and samples including extreme obese phenotypes [1–11]. Analysis of the latter group is based on the assumption that subjects with an extreme phenotype, such as morbid obesity, may be enriched for variants that influence the risk of developing obesity [7]. It is possible that extreme phenotypes of common disorders are due to rare. but highly penetrant, alleles that are difficult to identify by GWA analysis based on common single nucleotide polymorphisms (SNPs). However, based on GWA analysis of extreme obesity, Cotsapas et al found no evidence for rare highly penetrant alleles and concluded that most cases of severe obesity are extremes of the phenotypic spectrum rather than a distinct condition [6]. However, this does not exclude the existence of rare variants underlying specific cases of morbid obesity [12–14].

Susceptibility genes for human obesity are believed to act primarily on the central regulation of food intake. Three reported susceptibility loci for obesity harbour genes that are known to be involved in catabolic hypothalamic pathways (MC4R, PCSK1, and BDNF) and many others contain genes that are highly expressed in the central nervous system [11, 15, 16]. However, studies in mice mutated for Fto, the homolog of the human obesity susceptibility gene FTO, have demonstrated peripheral metabolic effects of this gene [17, 18]. Overall, the mechanisms of action of most obesity genes are not well understood and adipose tissue, as well as skeletal muscle, which are sites for storage, release, and metabolism of fatty acids, may be involved.

Overall, the reported genetic variation in obesity explains only a minor fraction of the total genetic variation expected to be present in the population. In the most recent GWA meta-analysis, 32 obesity loci explained together only about 2-4% of the genetic variation in BMI [11]. Thus, much of the genetic variation controlling obesity remains to be discovered [11, 15, 19]. In this study we report the identification of a novel susceptibility gene for obesity by GWA analysis of morbidly obese cases and controls that have a lifelong history of leanness. The association for this gene was confirmed in five independent case versus control cohorts, as well as in population based samples. Finally, we quantified the expression of the novel gene contributing to obesity in human adipose tissue samples.

The G-allele of rs2116830, located in intron 28 of the gene Potassium channel, calcium activated, large conductance, subfamily M, alpha member (KCNMA1) on chromosome 10, showed consistent effect direction and significant allelic association with obesity in the French cohort 5 comprising morbidly obese adults and population-based adult control subjects (P = 0.0017), but not in the German cohort 6 comprising morbidly obese youth and adult lean controls (P = 0.35) (Table 3, Figure 2A). The overall meta-analysis p-value for cases vs controls showed genome-wide significance (P = 2.82 × 10^-10) with an odds ratio (OR) of 1.26 [95% C.I. 1.12-1.41], with no statistical evidence for heterogeneity in impact on obesity between cohorts (Figure 2A). The impact of the G-allele of rs2116830 on obesity under different genetic models was tested in a joint analysis of all case-control cohorts. The recessive, P = 2.5 × 10 ^-9 (OR 1.30 [95% C.I. 1.19-1.42]), but not the additive or dominant genetic model reached genome-wide significance (Table 4).

rs988712*G, which is located 113,058 base pairs downstream of the established obesity locus BDNF on chromosome 11 [8], showed consistent effect direction and was significantly associated with obesity in the French cohort 5 (P = 0.0020), but not in the German cohort 6 (P = 0.135). The overall meta-analysis p-value for association with obesity showed genome-wide significance (P = 5.2 × 10^-17) with an OR of 1.36 [95% C.I. 1.20-1.55] (Figure 2B). There was statistical evidence for heterogeneity in impact on obesity between cohorts for rs988712 (P = 0.027). Additive, dominant and recessive genetic models reached genome-wide significance and the G-allele of rs988712 was associated with obesity in both men and women (Table 4). Only two of seven obesity-associated SNPs in the BDNF region in [8], rs6265 and rs10501087, were genotyped in our GWA cohort. Neither showed strong LD with rs988712 in our cohort, r2 0.57-0.59.

For the new obesity susceptibility locus rs2116830 (KCNMA1) we performed quantitative trait analysis of BMI in three adult population-based samples, n = 4133 in total (Table 1). One of these cohorts (controls in French cohort 5) was also used in the case-control studies described above. For BMI we could detect 1.0 kg/m² difference between genotypes with 93% power with nominal P of 0.001. There was neither significant association with BMI in any population alone, nor in joint analysis of all three populations (additional file 1, Table S3). (The same results were obtained when we included gender, age, and ethnicity in the analysis (results not shown). When we compared allele frequencies between obese cases and non-obese controls in the population based samples we observed consistent effect direction (additional file 1, Table S3) as in the case-control analysis of obesity in cohorts 1-5 described above (Table 2). The results were non-significant; however, this was expected as the total number of obese cases in the population-based samples was small. 48 subjects had BMI > 35 kg/m² and 267 subjects had BMI > 30 kg/m².

We investigated to what extent our GWA replicated published obesity or BMI loci that had been identified by GWA analysis. For each of the more than 30 published loci, we selected one, or sometimes two, SNPs for which the strongest association with obesity or BMI had been reported. 13 of these SNPs, which represented eleven loci, were analyzed for association with obesity in cohort 1. The remaining SNPs were not represented on the Affymetrix 500K Gene chip arrays used for the GWA or the SNP did not pass our quality control. Besides BDNF, our GWA study confirmed allelic association for MC4R with nominal P< 0.05 (additional file 1, Table S4).

Adipose tissue expression of KCNMA1 is increased in obesity

To elucidate potential mechanisms of action of KCNMA1 rs2116830 in obesity we determined mRNA levels of in human abdominal subcutaneous adipose tissue from obese and lean subjects. KCNMA1 mRNA was increased about fourfold in adipose tissue (P = 0.0001) as well as isolated fat cells (P = 0.04) derived from obese subjects compared to controls (Figure 3).

We have identified one new susceptibility locus for obesity near KCNMA1 (rs2116830) and confirmed association with BDNF (rs988712) [8], by GWA analysis in a limited sample of morbidly obese and lean adults This study was followed up by genotyping of five additional European case-control cohorts. Both loci reached genome-wide significant association with obesity in meta-analysis of investigated cohorts [33].

Our exploratory GWA study was performed in morbidly obese subjects raising the question to what extent the results are applicable to other groups of obese patients? Cotsapas et al proposed that severe obesity in most cases is a condition at the extreme of the phenotypic spectrum rather than a distinct condition, which suggests that the same genes are involved in all obese patient groups of the same ethnic origin [6]. However, the susceptibility loci for morbid obesity identified by Meyre et al and Scherag et al seemed to have limited or no impact on BMI in the general population [7, 10]. This may also be true for the novel obesity gene KCNMA1 identified in this study. KCNMA1 rs2116830 was not associated to BMI in three population based samples. It is tempting to speculate that KCNMA1 is of minor importance for the development of a moderate increase in fat mass, but contribute to excessive accumulation of adipose tissue in obesity. However, the results from the population-based cohorts should be interpreted with caution as the gender-ratio differed widely between samples. The differences in gender-ratios between the different case-control cohorts were smaller and are thus less likely to confound the results.

Many longitudinal studies show a strong continuity from mild obesity in childhood to more severe obesity in adulthood [34]. In addition, most obesity-susceptibility loci identified by GWA studies in adults are already associated with anthropometric traits in children/adolescents [35]. Together, this supports a shared genetic background for early onset and morbid obesity. There are only few large cohorts available for studies of distinct forms of obesity. In order to obtain genome-wide significance we had to include cohorts with different forms of obesity in the case control analyses [adults with morbid obesity (BMI > 40 kg/m²), adult obesity (BMI 30-40 kg/m²), and childhood obesity]. The two susceptibility loci for obesity reported here, KCNMA1 and BDNF displayed nominal allelic association with obesity in each investigated adult case-control cohort. However, the associations of KCNMA1 and BDNF with childhood obesity are less evident and need to be comfirmed in additional cohorts; with both genes being associated with obesity in French but not in German children. Joint analysis of our GWA data on morbid obese and lean adults, and the GWA data from Hinney's et al on obese children and adult lean controls may help to clarify to what extent morbid and childhood obesity have a similar genetic background [3].

Among obesity and BMI associated loci reported in other GWA studies, we could only confirm two, BDNF and MC4R. One weakness in our analysis was that we, due to a small sample sizes and different ethnicity, did not perform imputation of all BMI associated loci in published GWAs. Our analysis was therefore limited to published obesity-associated SNPs that produced high quality genotypes in our study, and encompassed only eleven out of more than 30 reported loci. The inability to confirm published obesity loci could be due to small sample size, different ethnicity, differences in sample selection and definition of phenotype. Obesity genes are not necessarily universally detected. Several obesity genes in previous GWAs could not be replicated in large European populations [36, 37]. All published obesity loci detected by GWA confer a modest to small risk for obesity. This is also true for KCNMA1 and BDNF in this study. The odds ratio for developing obesity is 1.26 and 1.36, respectively, for these two loci. The heterogeneity between cohorts for impact of BDNF rs988712 on obesity may be due to differences in ethnicity, age, or severity of obesity, as well as low power.

In order to further evaluate mechanisms by which KCNMA1 could contribute to the pathogenesis of obesity we performed gene expression studies in adipose tissue and observed increased KCNMA1 mRNA expression in obesity. KCNMA1 encodes one subunit of the large-conductance voltage- and Ca(2+)-activated K+ channel (BK channel), which is implicated in human epilepsy, blood pressure regulation, and the risk of myocardial infarction [38, 39]. In addition, at cellular levels stimulation of KCNMA1 channels enhances proliferation of human pre-adipocytes in vitro [40]. The latter is intriguing since it has recently been shown that there is a high rate of adipocyte turnover in vivo; with about 10% of fat cells being renewed annually [41]. Furthermore, adipocyte number is a major determinant for the fat mass in adults [41]. Thus, KCNMA1 could hypothetically contribute to obesity by increasing number of fat cells. We did not have access to tissue samples to study KCNMA1 expression in other organs, including the organ strongest implicated in regulation of food intake and obesity, the brain. We therefore cannot exclude a change of KCNMA1 expression y in other organs in association with obesity.

Besides small sample size there are other weaknesses in our GWA., e.g. we did not calculate power and ancestry principal components, nor did we formally test for relatedness between subjects.

In conclusion, we identified KCNMA1 as a novel susceptibility locus for obesity, which may promote obesity at least in part by acting in adipose tissue. Furthermore, we confirmed the previously described obesity locus BDNF. Further studies of KCNMA1 may highlight new targets for treating obesity.