A genome-wide scan for human obesity genes reveals a major susceptibility locus on chromosome 10

Article metrics

Abstract

Obesity, a common multifactorial disorder, is a major risk factor for type 2 diabetes, hypertension and coronary heart disease1 (CHD). According to the definition of the World Health Organization (WHO), approximately 6-10% of the population in Westernized countries are considered obese2. Epidemiological studies have shown that 30-70% of the variation in body weight may be attributable to genetic factors. To date, two genome-wide scans using different obesity-related quantitative traits have provided candidate regions for obesity3,4. We have undertaken a genome-wide scan in affected sibpairs to identify chromosomal regions linked to obesity in a collection of French families. Model-free multipoint linkage analyses revealed evidence for linkage to a region on chromosome 10p (MLS=4.85). Two further loci on chromosomes 5cen–q and 2p showed suggestive evidence for linkage of serum leptin levels in a genome-wide context. The peak on chromosome 2 coincided with the region containing the gene (POMC) encoding pro-opiomelanocortin, a locus previously linked to leptin levels and fat mass in a Mexican-American population3 and shown to be mutated in obese humans5. Our results suggest that there is a major gene on chromosome 10p implicated in the development of human obesity, and the existence of two further loci influencing leptin levels.

Main

Monogenic animal models have shown that severe obesity can be caused by single gene mutations5,6,7,8,9. Although rare mutations in genes encoding leptin (LEP) and the leptin receptor (LEPR), leading to extreme obese phenotypes, have been identified10,11 in some human families, neither gene seems to have a major role in the development of human obesity. There have also been reports of suggestive evidence of linkage for candidate genes using extreme phenotypes or small, isolated populations13,14,15. The major genes conferring susceptibility to human obesity, however, remain mostly unknown. We therefore performed a genome-wide scan to determine chromosomal regions linked to obesity.

We selected 158 nuclear families comprising 514 individuals (Table 1). Families had a proband with a BMI>40 kg/m2 and at least one further affected sibling (BMI>27 kg/m2, Ntotal=264 sibpairs; Table 2). All individuals were genotyped for 380 microsatellite markers. The sex-averaged distance between markers was estimated at 9.1±2.5 cM (range 1.5-28.8 cM). As the mode of inheritance for obesity is complex, model-free methods were used for multipoint linkage analyses. Linkage to serum leptin levels was evaluated as the difference of the trait between the two sibs of a pair regressed on the IBD information.

Table 1 Phenotypic characteristics by affection status
Table 2 Family structures

Multipoint analyses revealed nine regions with at least marginal evidence for linkage (MLS>1.0; Table 3). Only one region, on chromosome 10p, showed strong evidence for linkage in a genome-wide context (MLS=4.85; Fig. 1). The borders (defined by MLS>1.0) of the positive peak lay between markers D10S548 and D10S208, comprising a region of approximately 15 cM. The maximum MLS values were near markers D10S197 and D10S611, which were less than 4 cM apart. Only two more regions, on chromosomes 6 and X, showed suggestive evidence for linkage (MLS>2.11)

Table 3 Markers with evidence for linkage
Figure 1: Multipoint analyses results with the most significant evidence for linkage on chromosomes 10, 2 and 5.
figure1

Red and blue curves correspond to affected sibpair analysis (BMI) and quantitative trait analysis (serum leptin values), respectively. The arrow denotes the approximate position of the POMC gene on chromosome 2.

Simulation tests showed that the result on chromosome 10 was robust to assumptions concerning marker allele frequency (data not shown). To establish a threshold of significance for the linkage results from the multipoint analyses, based on real data, we simulated the transmission of the alleles using observed allele frequencies and map order in our families. We computed and analysed 100 replicates of a whole genome scan by multipoint analysis using MapMaker. Only once did we observe an MLS value higher than the one obtained on chromosome 10 (MLS=5.28).

In addition to affected sibpair (ASP) analyses, we carried out quantitative trait analyses for serum leptin levels. Multipoint analyses revealed chromosomal regions with suggestive evidence for linkage on chromosomes 2p (lod score=2.68; Fig. 1) and 5 (lod score=2.93; Fig. 1). On chromosome 2, the maximum excess of allele sharing was observed for markers D2S165 and D2S367. Marker D2S367 was approximately 1.5 cM distal from marker D2S1788, which had previously been reported to be linked to leptin levels and fat mass in a genome-wide scan in Mexican-Americans3. Eleven further chromosomal regions showed at least nominal evidence for linkage (lod score>1.0) to leptin levels (Figs1,2).

Figure 2: Multipoint analyses results for all chromosomes.
figure2

Positions of candidate genes are indicated by arrows, chromosomal regions corresponding to obesity-related syndromes are indicated by filled rectangles. LEPR, leptin receptor; LEP, leptin; NPY1R, neuropeptide Y1 receptor; TUB, human tubby gene homologue; ASIP, agouti signalling protein; CPE, carboxypeptidase E; UCP2/3, uncoupling proteins 2 & 3; BBS 1-4, Bardet-Biedl syndrome 1-4; PWS, Prader-Willi syndrome. The total genetic distance covered by the analysed markers is given on the x-axis in cM.

Among the 380 markers tested, nine were located near or in known candidate genes for obesity (Figs1,2). With the exception of the POMC region on chromosome 2, none of these markers showed evidence for linkage consistent with criteria for a whole-genome scan and the phenotypes selected for analysis. A few other markers covered regions implicated in obesity-related syndromes, namely Bardet-Biedl syndrome (BBS) regions 1-4 on chromosomes 11q13, 16q21, 3p13 and 15q22, the Prader-Willi syndrome (PWS) region on chromosome 15q11 and the Cohen syndrome region on chromosome 8q22. Again, there was no strong evidence for linkage with markers in these regions, however, the BBS-4 region on chromosome 15 showed some evidence for linkage to serum leptin levels. Although our study showed no strong evidence for linkage with any of the above-mentioned genes or regions in a genome-wide context, it should be emphasized that this did not exclude them as candidate genes with either: (i) minor effects not detectable by linkage; (ii) effects on obesity related traits other than the ones studied here; or (iii) major effects in other unrelated populations.

Recent success in the genetic study of obesity in rodent models has increased our understanding of the mechanisms underlying fat and energy metabolism. However, none of the genes implicated in the monogenic forms of rodent obesity seem to have a major role in the development of human obesity16,17,18. Human obesity is the result of complex interactions between genetic and environmental factors. Although some candidate gene linkage studies have been reported, none give significant evidence for linkage in a genome-wide context19. Evidence for a major QTL influencing leptin levels has been described in a genome-wide screen in Mexican-Americans3, but no major locus has been identified for obesity.

Our results suggest that there is a major locus for obesity on chromosome 10p in our population. Few genes are known in that region, none of them an apparent candidate gene for obesity. The multipoint MLS value of 4.85 for the region exceeds the suggested reference threshold given as evidence for linkage20. Our own simulated 100 genome-wide scans showed that this value was exceeded only once, indicating that the probability of observing a higher MLS by chance anywhere in the genome was approximately 1 in 100.

Some investigators have suggested applying a correction for families with more than two affected sibpairs in the linkage analyses, however, it has been shown by other studies that this is often over-conservative, especially for moderate sibship sizes and if many families are included21 (as in our study; Table 1). Meunier et al. also noted that assuming all sibpairs as independent does not result in an inflation of the MLS, as is sometimes argued. Our own simulation of 100, whole genome, uncorrected multipoint analyses, based on real marker data and using the family structure of our collection, supports this view. Only twice did the value exceed an MLS of 4.0, showing that, using uncorrected statistics, the probability that the result on chromosome 10p is a chance finding is small.

An estimated contribution of the gene on chromosome 10 to obesity can be calculated from the locus specific λs, estimated from the parameters (z0, z1 and z2) for the maximum MLS value for this marker (λs=1/(4*z0). The best estimate for D10S197 (z0=0.15) is λs=1.66. An estimate for the total λs for morbid obesity in the French population has been calculated22 at λs=4.15. Therefore, in a non-additive model, the gene on chromosome 10p could account for as much as 36% (log 1.66/log 4.15) of the obesity in this population. Under an additive model, the contribution could be as large as 21% (1.66–1/4.15–1).

We also used quantitative trait analyses to identify loci influencing serum leptin levels. Our results showed suggestive evidence for linkage with serum leptin levels for loci on chromosomes 2p and 5. The region identified on chromosome 2 correlates with the results of a previous study, which identified this region in Mexican-Americans3. This result suggests that there may be common genes contributing to obesity-related phenotypes in populations with ethnically and culturally (environmentally) distinct backgrounds. The interval on chromosome 2 encompasses the POMC locus.

The region on chromosome 5 is more difficult to evaluate because of the size of the interval. This region harbours a number of candidate genes for obesity, moreover, the region also seems to have undergone extensive duplication processes which might explain the broadness of the peak (that is, more than one gene). Some of these genes are ordered in clusters on 5p and 5q (J.L. Nahon, pers. comm.). Therefore, the region on chromosome 5 shows some similarity to the HLA region on chromosome 6 in type 1 diabetes, which has a comparable organization and gene density and a similarly broad peak for linkage.

As in most studies, we do not have 100% power of detection for loci linked to obesity. Using models that differ with regard to assumptions of prevalence, distance (θ) between markers and disease locus, we determined that the power to detect linkage is 60-80% in our sample. This indicates that we may have missed other loci contributing to obesity in this population.

The results of our genome-wide scan have revealed three putative gene loci for obesity and leptin levels on chromosomes 10, 5 and 2. One of these loci has already been shown to be linked to leptin levels in another set of families, and the locus on chromosome 10p may account for 21-36% of the obesity in our study population. Once these genes have been identified, it will be possible to determine if and how these different loci may interact to lead to the complex phenotype obesity.

Methods

Subjects.

Our sample consisted of 158 nuclear families collected through the Department of Nutrition, Hôtel-Dieu Hospital in Paris (46%) or by a multimedia campaign at the Institut Pasteur de Lille (54%) in France. Families were ascertained through a proband with a BMI>=40 kg/m2 and were included in the study if the proband had at least one additional sibling with a BMI>=27 kg/m2. Additional family members were included and genotyped when available. This collection comprised 514 individuals. Both parents were available in 30% of the families, and one parent in 37% of the families. For the remaining 33%, parents were not available. The maximum number of possible affected sibpairs was 264 ((n*(n–1))/2), with n being the number of sibs in the sibship).

Genotyping.

Genotyping was carried out using a fluorescence-based semi-automated technique on automated DNA sequencing machines (ABI 377, PE ABI). The basis for the genome-screen was the ABI genome-scan set of 358 microsatellite markers (PE ABI). A set of 100 markers was added to this set to either replace markers giving poor amplification products or to fill in gaps of more than 20 cM. PCR for the amplification of markers were carried out as multiplex reactions following protocols developed in our laboratory (conditions available on request). Analyses and assignment of the marker alleles were done with GENOTYPER (PE ABI). All genotyped markers were checked for incompatibilities using a customized version of the program UNKNOWN from the LINKAGE package. Incompatibilities were either resolved unambiguously or families were discarded from linkage analyses.

Linkage analyses.

As the mode of inheritance for obesity is unclear, model-free sibpair-based methods were used to detect linkage. Allele frequencies were estimated using data from all independent, unrelated individuals (founders) by the program DOWNFREQ from the ANALYSE package. Two-by-two recombination fractions were estimated to establish map order and genetic distances between markers for all chromosomes, using the MLINK program implemented in the VITESSE linkage package24. Markers showing a significant deviation of their position from the GENETHON map location (odds ratio in favour of GENETHON position of less than 1000:1) were discarded.

Threshold estimations for the genome-scan were carried out using observed phenotypes, allele frequencies and map order, and simulation of allele transmission for all markers from all chromosomes was done using SIMULATE. One-hundred replicates were generated for each chromosome. Subsequent multipoint analyses for each chromosome were performed using unweighted MAPMAKER ( ref. 25) procedures.

For multipoint analyses of qualitative traits (BMI>27 as cut-off point for obese individuals), two different methods were used. The SibPal program from the SAGE package26 is based on methods first proposed by Haseman and Elston27. For linkage testing, the mean proportion of alleles shared IBD among affected sibpairs (π) is estimated and tested against the null hypothesis of no linkage (π = 0.50), with significant excess of sharing taken as evidence for linkage. We also used a maximum likelihood method. The test statistic is a likelihood ratio maximized over several parameters. Likelihood is maximized as a function of the IBD probabilities p(ibd=0), p(ibd=1) and p(ibd=2) (z0, z1 and z2, respectively), and the results are compared with the likelihood values under the null hypothesis (0.25, 0.5 and 0.25, respectively). This ratio follows a mixture of χ2 with 1 and 2 degrees of freedom. Significant deviation from the expected values is taken as evidence for linkage. This method is implemented in the MAPMAKER/SIBS package. Where parental genotypes are missing, estimates of the probabilities for all possible parental genotypes conditioned on the sibship genotype information were calculated. No differences were obeserved using the two linkage analysis packages. In the text, only the MAPMAKER results are shown.

Quantitative traits were analysed using the Haseman-Elston method as well as estimating a likelihood ratio using MAPMAKER/SIBS. The Haseman-Elston quantitative trait linkage analysis method uses the squared phenotypic difference (Y) for each sibpair in a family which is regressed on the estimated proportion of marker allele IBD (π). If a marker is near the susceptibility locus, the difference is expected to decrease (that is, is negatively correlated) with the increase in the estimated IBD sharing. A one sided t-test is used to evaluate if the regression coefficient differs significantly from the expected value. Analyses were carried out with SIBPAL of the SAGE package (R.C. Elston, J.E. Baily-Wilson, G.E. Bonney, B.J. Keats and A.F. Wilson, unpublished, available from the Department of Biometry and Genetics, Louisiana State University Medical Center).

The second test statistic uses the calculation of IBD sharing probabilities implemented in MAPMAKER/SIBS. Likelihood is maximized on the parameters s 20, s 21 and s 22, the variances of Y for the pairs sharing 0, 1 and 2 alleles IBD, respectively. Under the null hypothesis, these three variances are equal. The resulting maximum lod score is the ratio of the maximum likelihood and the likelihood under the null hypothesis.

To estimate the influence of allele frequencies on the results, we generated random sets of allele frequencies through a Monte-Carlo procedure. The original genotype data was then analysed using MapMaker-Sibs and the results were retrieved and analysed in the same program. In a first approach, we randomly determined 10,000 sets of alleles with no constraints to upper or lower limits. In a similar approach, we created a set of pseudo-random allele frequencies in which the frequencies for rare alleles (that is, frequency<0.01) were artificially increased to be more than 0.05 to account for mis-estimations of rare alleles from the sample data. In a third test, we increased the allele frequencies for the most prevalent alleles to different values.

References

  1. 1

    Frayn, K.N. & Coppack, S.W. Insulin resistance, adipose tissue and coronary heart disease. Clin. Sci. Colch. 82, 1–8 (1992).

  2. 2

    de Onis, M. & Habicht, J.P. Anthropometric reference data for international use: recommendations from a World Health Organisation Expert Committee. Am. J. Clin. Nutr. 64, 650– 658 (1996).

  3. 3

    Comuzzie, A.G. et al. A major quantitative trait locus determining serum leptin levels and fat mass is located on human chromosome 2. Nature Genet. 15, 273–276 (1997).

  4. 4

    Norman, R.A. et al. Autosomal genomic scan for loci linked to obesity and energy metabolism in Pima Indians. Am. J. Hum. Genet. 62, 659–668 (1998).

  5. 5

    Krude, K. et al. Severe early onset obesity, adrenal insufficiency and red hair pigmentation caused by POMC mutations in humans. Nature Genet. 19, 155–157 (1998).

  6. 6

    Zhang, Y. et al. Positional cloning of the mouse obese gene and its human homologue. Nature 372, 425– 434 (1994).

  7. 7

    Chen, H. et al. Evidence that the diabetes gene encodes the leptin receptor: identification of a mutation in the leptin receptor gene in db/db mice. Cell 84, 491–495 (1996).

  8. 8

    Naggert, J.K. et al. Hyperproinsulinemia in obese fat/fat mice associated with a carboxypeptidase E mutation which reduces enzyme activity. Nature Genet. 10, 153–142 (1995).

  9. 9

    Noben-Trauth, K., Naggert, J.K., North, M.A. & Nishina, P.M. A candidate gene for the mouse mutation tubby. Nature 380, 534–538 (1996).

  10. 10

    Klebig, M.L., Wilkinson, J.E., Geisler, J.G. & Woychik, R.P. Ectopic expression of the agouti gene in transgenic mice causes obesity, features of type II diabetes and yellow fur. Proc. Natl Acad. Sci. USA 92, 4728–4732 (1995).

  11. 11

    Montague, C.T. et al. Congenital leptin deficiency is associated with severe early-onset obesity in humans. Nature 387, 903–908 (1997).

  12. 12

    Clement, K. et al. Mutation in the human leptin receptor causes obesity and pituitary dysfunction. Nature 392, 398– 401 (1998).

  13. 13

    Clement, K. et al. Indication for linkage of the human OB gene region with extreme obesity. Diabetes 45, 687– 690 (1996).

  14. 14

    Reed, D.R. et al. Extreme obesity may be linked to markers flanking the human OB gene. Diabetes 45, 691– 694 (1996).

  15. 15

    Norman, R.A., Bogardus, C. & Ravussin, E. Linkage between obesity and a marker near the tumour necrosis factor-* locus in Pima Indians. J. Clin. Invest. 96, 158–162 (1995).

  16. 16

    Considine, R.V. et al. Evidence against either a premature stop codon or the absence of obese gene mRNA in human obesity. J. Clin. Invest. 95, 2986–2988 (1995).

  17. 17

    Francke, S. et al. Genetic studies of the leptin receptor gene in morbidly obese French Caucasian families. Hum. Genet. 100, 491–496 (1997).

  18. 18

    Norman, R.A. et al. Absence of linkage of obesity and energy metabolism to markers flanking homologues of rodent obesity genes in Pima Indians. Diabetes 45, 1229–1232 (1996).

  19. 19

    Chagnon, Y.C., Perusse, L. & Bouchard, C. The human obesity gene map: the 1997 update. Obes. Res. 6, 76–92 (1998).

  20. 20

    Lander, E. & Kuglyak, L. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nature Genet. 11, 241–247 (1995).

  21. 21

    Meunier, F., Philippi, A., Martinez, M. & Demenais, F. Affected sib-pair tests for linkage: Type 1 errors with dependent sib-pairs. Genet. Epidemiol. 14, 1107– 1111 (1997).

  22. 22

    Tiret, L. et al. Segregation analysis of height-adjusted weight with generation- and age-dependent effects: the Nancy Family Study. Genet. Epidemiol. 9, 389–403 (1992).

  23. 23

    Considine, R.V. et al. Serum immunoreactive-leptin concentrations in normal-weight and obese humans. N. Eng. J. Med. 334, 292–295 (1996).

  24. 24

    O'Connell, J.R. & Weeks, D.E. The VITESSE algorithm for rapid exact multilocus linkage analysis via genotype set-recording and fuzzy inheritance. Nature Genet. 11, 402–408 (1995).

  25. 25

    Kruglyak, L. & Lander, E.S. Complete multipoint sib-pair analysis of qualitative and quantitative traits. Am. J. Hum. Genet. 57, 439–454 (1995).

  26. 26

    Tran, L.D., Elston, R.C., Keats, B.J.B. & Wilson, A.F. Sib-pair linkage program (SIBPAL). in S.A.G.E. Statistical Analysis for Genetic Epidemiology, Release 2.2 (Louisiana State University, New Orleans, 1994).

  27. 27

    Haseman, J.K. & Elston, R.C. The investigation of linkage between a quantitative trait and a marker locus. Behav. Genet. 2, 3–19 (1972).

Download references

Acknowledgements

We thank the patients for their participation to this study. We also thank the Assistance Publique/Hôpitaux de Paris (AP/HP) and the Programme Hôspitalier de Recherche Clinique (PHRC) for their support. We are grateful for the constructive discussions and help with the statistical analyses by M. Lathrop and M. Farrall. J.H. and E.V. were supported through European grant program BIOMED2 BMH4-CT950662.

Author information

Correspondence to Jörg Hager or Philippe Froguel.

Rights and permissions

Reprints and Permissions

About this article

Further reading