|Home | About | Journals | Submit | Contact Us | Français|
Obesity is a highly heritable and genetically heterogeneous disorder1. Here we investigated the contribution of copy number variation to obesity in 300 Caucasian patients with severe early-onset obesity, 143 of whom also had developmental delay. Large (>500 kilobases), rare (<1%) deletions were significantly enriched in patients compared to 7,366 controls (P < 0.001). We identified several rare copy number variants that were recurrent in patients but absent or at much lower prevalence in controls. We identified five patients with overlapping deletions on chromosome 16p11.2 that were found in 2 out of 7,366 controls (P < 5 × 10−5). In three patients the deletion co-segregated with severe obesity. Two patients harboured a larger de novo 16p11.2 deletion, extending through a 593-kilobase region previously associated with autism2-4 and mental retardation5; both of these patients had mild developmental delay in addition to severe obesity. In an independent sample of 1,062 patients with severe obesity alone, the smaller 16p11.2 deletion was found in an additional two patients. All 16p11.2 deletions encompass several genes but include SH2B1, which is known to be involved in leptin and insulin signalling6. Deletion carriers exhibited hyperphagia and severe insulin resistance disproportionate for the degree of obesity. We show that copy number variation contributes significantly to the genetic architecture of human obesity.
Although the rise in obesity prevalence is driven by environmental factors, there is considerable evidence that weight is highly heritable1,7. Genome-wide association studies (GWAS) have identified common single nucleotide polymorphisms (SNPs) associated with increased body mass index (BMI)8-11; however, together these account for a small percentage of the inherited variation in BMI. Studies in patients with severe early-onset obesity have led to the detection of rare variants, many of which have an impact on the leptin–melanocortin system involved in energy homeostasis7,12,13. To explore the contribution of copy number variants (CNVs) to obesity, we studied 300 Caucasians with severe early-onset obesity. Rare CNVs can cause developmental diseases14; as such we enriched for patients with developmental delay in addition to severe obesity (n = 143) in this discovery set (Supplementary Information). Genomic DNA was hybridized to the Affymetrix 6.0 array and the frequency of CNVs in patients was compared to 7,366 controls of European ancestry. Data were analysed using Affymetrix Power Tools and Birdsuite software15. We performed extensive post-calling quality control (QC) and filtering to arrive at a final set of CNVs (Supplementary Information). We detected 15,407 CNVs in the 284 patients passing QC and 403,098 CNVs in the 7,366 controls passing QC. The median number of CNV calls per individual was similar between patients and controls (53 and 55, respectively), as was the median size of CNVs (22 kilobases (kb) and 23.5 kb, respectively).
Following criteria set previously16, we defined large, rare deletions as present in <1% of individuals and >500 kb in size. We observed a twofold enrichment of large, rare deletions in patients compared to controls (Table 1). We assessed the significance of this observation using conditional permutation tests controlling for differences in total number of CNV calls or data quality between patients and controls, and found it to be highly significant (P = 0.0005, Supplementary Information). Patients with developmental delay in addition to severe early-onset obesity were more likely to harbour rare, large deletions (Table 1). The trend towards enrichment of large, rare deletions remains in patients with severe obesity alone, but is no longer statistically significant.
We identified several rare CNVs recurrent among patients and enriched relative to controls (Table 2 and Supplementary Fig. 1). We confirmed these by multiplex ligation-dependent probe amplification (MLPA) and investigated co-segregation in families. We identified four patients with the 16p11.2 (29.5–30.1 megabases (Mb)) deletion reported in autism and mental retardation2-4 (Table 2). These patients had mild developmental delay requiring special educational needs support; two had autistic spectrum behaviour. This frequency (~1% of all obesity patients, ~2% of patients with obesity and developmental delay) is slightly higher than the frequency (0.5–1%) in published studies2,3,5, possibly due to the inclusion of patients with developmental delay in addition to severe obesity in our study (n = 143). The literature17 suggests that 16p11.2 (29.5–30.1 Mb) deletions may be associated with being overweight5 and with severe obesity in one patient4. Two of three patients harbouring duplications in this region were reported as underweight17. However, our study design did not permit detailed investigation of the contribution of 16p11.2 (29.5–30.1 Mb) deletions to altered body weight.
A large 15q11-q13 duplication is well-known to be associated with autism4,18; we found smaller, distinct 15q13.2-q13.3 duplications in two patients, which have previously been observed in controls19, as well as patients with mental retardation, autism and other phenotypes20 (Table 2). We also identified two patients with a 11q22-q23 deletion overlapping minimally with a CNV associated with mental retardation20,21 on a single brain-expressed gene GUCY1A2 (Table 2). Some of the CNV carriers in these studies were obese20. The finding of these rare recurrent CNVs in our cohort may point to a shared mechanistic basis for these neurodevelopmental diseases. Rare CNVs encompassing brain-expressed genes such as TSNARE1 and PDE2A were unique to patients with severe early-onset obesity and require further investigation.
The commonest CNV enriched in patients with severe obesity was identified in five unrelated patients harbouring deletions on 16p11.2 (Table 2) with a minimal overlapping segment of 220 kb (Fig. 1a). This 220-kb deletion (28.73–28.95 Mb) was found in 2 out of 7,366 controls (Supplementary Information); BMI data were unavailable. The minimal deleted region contains genes involved in neurological diseases (TUFM, ATP2A1), immunity (CD19, NFATC2IP, LAT) and genes of unknown function (ATXN2L, RABEP2, SPNS1), as well as SH2B1, which encodes an adaptor protein involved in leptin and insulin signalling. Disruption of Sh2b1 in mice results in obesity and severe insulin resistance22. As CNVs can unmask a recessive allele, we sequenced SH2B1 in these patients but did not find any coding/splice site mutations (Supplementary Information).
The 220-kb 16p11.2 deletion (28.73–28.95 Mb) was seen in three patients and co-segregated with severe early-onset obesity alone (Fig. 1b). A longer ~1.7-Mb deletion detected in two patients (Fig. 1c) encompassed the 220-kb deletion and extended through a 593-kb region (29.5–30.1 Mb) where deletions are associated with autism and mental retardation2-5,17 (Fig. 1c). These longer deletions occurred de novo; both these patients had mild developmental delay in addition to their severe obesity. These findings are consistent with a role for the SH2B1-containing 220-kb region (28.73–28.95 Mb) in severe obesity and the 29.5–30.1-Mb region in brain development. We used MLPA to screen for this rare SH2B1-containing (28.73–28.95 Mb) deletion in an independent set of 1,062 Caucasian patients with severe obesity alone from the same cohort, as no other comparable cohort of the same ancestry and size exists. We found two additional patients with this deletion. The deletion co-segregated with obesity in one family whose samples were available (Supplementary Information). The prevalence of the SH2B1-containing deletion was lower in the replication cohort (2 out of 1,062) compared to the discovery set with severe obesity alone (3 out of 157). Overestimation of case frequency in the primary screen is a recognized phenomenon especially with the investigation of rarer variants23. Overall the prevalence of the SH2B1-containing deletion in patients with severe early-onset obesity alone (5 out of 1,219; 0.41%) is significantly greater than in controls (2 out of 7,366; 0.027%) (P < 0.001).
The breakpoints of both classes of deletion are embedded within complex, segmentally duplicated regions of 16p11.2 containing directly-oriented, highly-similar (>98% sequence similarity) duplicated sequences greater than 15 kb in length (Fig. 1a). This observation strongly supports the hypothesis that these deletions arise through non-allelic homologous recombination between duplicated sequences14.
We characterized the phenotype of patients with the SH2B1-containing deletion. Patients rapidly gained weight in the first years of life (Fig. 2a); their excess weight was predominantly fat mass (Fig. 2b). They exhibited hyperphagia with increased ad libitum food intake (Fig. 2c). Fasting plasma insulin levels were disproportionately elevated compared to age- and obesity-matched controls and patients with other obesity syndromes including MC4R deficiency, where we have reported disproportionate hyperinsulinaemia12 (Fig. 2d). Fasting plasma glucose concentrations were in the normal range. Adult patients with the SH2B1-containing deletion had delayed, exaggerated insulin secretion in response to an oral glucose load compared to equally obese controls (Fig. 2e).
The similarities with the human leptin-receptor-deficient phenotype are striking13 and consistent with a role for central disruption of SH2B1 (ref. 6). Deletion of peripheral Sh2b1 in rodents impairs insulin signalling in muscle, liver and adipose tissue24, which might explain the severe and disproportionate insulin resistance in these patients. As heterozygous deletion of Sh2b1 in mice leads to obesity on a high-fat diet6, haploinsufficiency may be a plausible mechanism underlying the phenotype seen in humans (Supplementary Information).
Common SNPs near the SH2B1 locus have been associated with BMI in GWAS10,11. We investigated CNVs upstream or downstream of known obesity genes or other BMI-associated loci. We identified a 153-kb deletion 60 kb downstream from MC4R (chromosome 18: 55,975,710–56,128,721) in a single patient (Supplementary Fig. 2a), inherited from an obese father (Supplementary Fig. 3). This deletion removes a BMI association interval25 (Supplementary Fig. 2a) and its distal end partially overlaps with a CNV found in a single control (chromosome 18: 55,804,160–56,063,854). Further studies will be required to ascertain the significance of this finding. A 100-kb duplication encompassing part of LEPR was found in a single patient but also in 16 controls (Supplementary Fig. 2b). Although we did not find any rare CNVs close to FTO8,9, we detected the common 45-kb deletion upstream of NEGR110,11 at comparable prevalence in patients and controls (Supplementary Fig. 2c).
We report here the first rare CNVs associated with severe early-onset obesity. We show that deletion of 16p11.2 is associated with highly penetrant familial severe early-onset obesity and severe insulin resistance. Although the contribution of other genes or non-coding genetic material cannot be excluded, the phenotype is consistent with a role for SH2B1 in human energy homeostasis and glucose metabolism. We did not identify any coding/splice site mutations in SH2B1 in 500 Genetics of Obesity Study (GOOS) patients (Supplementary Information). Segmental duplications on 16p11.2 and concomitant increase in deletion rate through non-allelic homologous recombination may explain why deletions are a relatively more frequent source of loss of function in SH2B1 than at other monogenic obesity loci. Novel CNVs unique to patients with severe early-onset obesity and the increased CNV burden (which is not fully accounted for by SH2B1 deletions) suggest a role for additional genes in the aetiology of severe obesity. Our findings indicate the presence at the same loci of both common variants influencing susceptibility to common obesity and more highly penetrant rare variants, including CNVs, associated with severe early-onset forms of the disease. Strategies aimed at looking for rare variants near common susceptibility loci may well prove to be fruitful in other common complex diseases.
A discovery set of 300 UK Caucasian patients from the Genetics of Obesity Study (GOOS) cohort were randomly selected for this study (143 had developmental delay). All patients had severe obesity defined as a BMI standard deviation score (s.d.s.) >3 and onset of obesity before 10 years of age (Supplementary Information). Mutations in LEPR, POMC and MC4R were excluded by direct nucleotide sequencing and a karyotype performed. DNA samples were run on Affymetrix Genome-Wide Human SNP Array 6.0 by Aros, Inc., and compared to control data on the same platform obtained on over 7,000 controls recruited from the Wellcome Trust Case Control Consortium 2 (WTCCC2) and the Genetic Association Information Network (GAIN). The WTCC2 data set contains 6,000 individuals, 3,000 from the 1958 British Birth Cohort and 3,000 from the UK Blood Service Collection, used as common controls in genome-wide association studies of 13 disease conditions undertaken by Wellcome Trust Case Control Consortium 2 (http://www.wtccc.org.uk/ccc2). The GAIN data set contains 1,442 individuals of European ancestry used as part of the control cohorts in genome-wide association studies of schizophrenia and bipolar disorders. The data set was obtained from the database of Genotype and Phenotype (dbGaP) found at http://www.ncbi.nih.gov/gap through dbGaP accession numbers phs000021.v1.p1, phs000021.v2.p1 and phs000017.v1.p1. Further methods for CNV calling, analysis and MLPA are included in the Supplementary Information. All clinical studies were conducted in accordance with the principles of the Declaration of Helsinki using methods that have previously been reported13.
We thank the patients, their families and all the Physicians who have referred patients to the Genetics of Obesity Study (GOOS). E.G.B., N.H., M.E.H. (grant no. 077014/Z/05/0Z) and I.S.F. (grant no. 082390/Z/07/Z) were funded by the Wellcome Trust. I.S.F. and S.O.R. are funded by the MRC Centre for Obesity and Related Disorders and NIHR Cambridge Biomedical Research Centre. We would like to thank L. Mavrogiannis for advice on MLPA analysis and V. Trowse for help with clinical studies. This study makes use of control data provided by the Genetic Association Information Network (GAIN) and the Wellcome Trust Case Control Consortium 2 (WTCCC2), through work funded by NIH and the Wellcome Trust (under award 085475). The GAIN General Research Use data sets used for the analyses described in this manuscript were obtained from the database of Genotype and Phenotype (dbGaP; http://www.ncbi.nlm.nih.gov/gap, dbGaP accession numbers phs000017 and phs000021). A full list of the investigators who contributed to the generation of the WTCCC2 data is available from http://www.wtccc.org.uk/ccc2.
The authors declare no competing financial interests.