We performed a genome-wide analysis of common copy number variations (CNVs) in two discovery samples ascertained for early-onset (extreme) obesity. We observed 244 CNVRs that covered ~1.56% of the genome-most of these regions were already known from databases. Among the 244 CNVRs, we detected 20 regions that were directionally consistently associated with obesity in both discovery samples.
Among these 20 associated regions, we identified a new CNVR 11q11 associated with early-onset (extreme) obesity. While the presence of copy number variability at CNVR 11q11 is well established (28
), its relationship to obesity has not been described before. CNVR 11q11 covers the three protein coding genes OR4P4
(according to RefSeq of UCSC hg18; http://genome.ucsc.edu/
) of the OLR family 4. OLRs interact with odorant molecules in the nose, to initiate a neuronal response that triggers the perception of a smell. They form the largest mammalian protein superfamily. There is a high percentage (~55%) of human pseudogenes (29
). The OLR proteins are members of the family of G-protein-coupled receptors arising from single-coding exon genes.
It was hypothesized that CNVs play an important role in the evolution of the human olfactory repertoire (29
). About half of the CNVs affecting the human OLR repertoire involve more than one OLR. It was also observed that CNVs were more frequent among pseudogenes of OLRs than among functional genes (29
). Comparison to the chimpanzee, reference genome revealed that all of the detected deletion alleles are human derived. This indicated a profound effect of human-specific deletions on the individual OLR gene content. It was suggested that these deletion alleles may be used in future genetic association studies of olfactory inter-individual differences (29
). Evidence was provided for OLR enrichment in CNVs not being due to positive selection but due to the dominance of OLR in segmentally duplicated regions (30
). Additionally, purifying selection against CNVs is lower in regions containing OLRs than in regions containing essential genes (30
). A possible link between OLRs and obesity has previously been suggested based on the observation of altered olfactory acuity in morbidly obese patients (31
Additionally, within the 20 associated regions, we confirmed two previously published obesity associations: CNVR 1p31.1 20 kb upstream of NEGR1
) and CNVR 10q11.22 (1) covering the four genes SYT15
) with PPYR1
being the most interesting candidate, given its role in energy homeostasis and regulation of food intake (27
null animals have, for instance, a reduced body weight. PP reduces food intake predominantly via stimulation of the anorexigenic melanocortinergic pathway. This effect is mediated by direct action on local PPYR1
within the arcuate nucleus (32
). PP binds to the PPYR1. In our study as well as in the previous study which initially described the association of CNVR 10p11.22 (1
) with BMI (19
), the loss of a PPYR1
gene copy number was associated with obesity. Among those individuals (n
= 57) of our two GWAS discovery samples that were called for the loss of a PPYR1
gene copy number, only one was homozygous for the deletion. One can only speculate about these seemingly contradictory findings between studies in human and mice. Several mechanisms might explain the difference: (i) mice may not be the most appropriate model to understand the function of the human PPYR1, as mice express a functional Y6 receptor (in humans only an non-functional pseudo gene exists) and this receptor could explain some of the results of the Ppyr1 knockout mice. (ii) Sequence identity between mice Ppyr1 amino acid sequence is only 76% compared with human PPYR1. The high variability of PPYR1 across species might explain its different roles among species. (iii) Ppyr1 knockout mice are hyperphagic but nonetheless lean (27
). If the gene has a slightly altered function in humans, hyperphagia is potentially retained but not the lean phenotype. (iv) Compared with the 56 studied subjects who are hemizygous for the PPYR1
knockout mice have neither copy of the gene nor corresponding gene expression products. Thus, subjects with only one copy of PPYR1
may produce less protein, which will regulate energy homeostasis through agonists or other pathways to inhibit obesity. In contrast, knockout mice have neither copy of the gene nor the corresponding gene expression products, which may also explain the difference between human studies and mice experiments.
While the replications of previous findings underline the strength of our study, the focus on consistently associated common CNVs for early-onset (extreme) obesity using an agnostic genome-wide approach is also accompanied by weaknesses. First of all, our follow-up/replication step using tagging SNPs has focused on only 2 of all 20 candidate CNVRs-albeit those with the strongest association signals. Thus, it is conceivable that additional regions could be true positive findings. However, given their weaker association signals in both of our deletion samples, larger replication samples than the one used by us will be required to rule out false-negative findings. This is underlined by the results of our array-based replication analyses supporting four additional CNVRs of all 20 candidate CNVRs. Secondly, our study clearly has limited detection power to comprehensively assess CNV at a genome-wide level. None of the tested CNVs achieved a stringent genome-wide significance level and for those that could be replicated only moderate effects (odds ratio of ~1.2) were observable. Thus, our observation for early-onset (extreme) obesity underlines the recent finding that common CNVs are unlikely candidates to explain larger parts of the ‘missing heritability’ (26
). Finally, it has to be underlined that the focus on CNVRs is accompanied by (a) the difficulty to define the CNV markers that make up a CNV and (b) the difficulty to assess when and if association signals in different samples pertain to the same source. Here both problems have been dealt with by providing upper limits of correlations among associated CNVs of independent samples and by omitting combined P
-values across samples.
Since we replicated two CNVRs using tagging SNPs, their association could have already been detected by genome-wide SNP association studies. Indeed, rs2815752 (tagging CNVR 1p31.1) near NEGR1
is a well-known and well-replicated BMI locus (8
). In contrast, for rs9804659 (tagging CNVR 11q11), no association to obesity, BMI or waist–hip ratio has yet been reported in the primarily population-based genetic association studies of adults (33
Very recently, another 21-kb CNV that lies 50 kb upstream of the gene GPRC5B
has been reported to be associated with BMI (33
). The identification of this CNV was based on the tagging SNP rs12444979 in a large meta-analysis for population-based samples, whereas the deletion allele of the CNV is tagged by the non-risk allele of the SNP. We could also identify this CNV as one of the 244 common CNVRs analysed here. However, a significant association with early-onset (extreme) obesity was observed only in our case–control GWAS discovery sample (one-sided min. PCC
= 0.0066 and min. Pfamily-based
= 0.0408 for the non-deletion direction).
Owing to our focus on common CNVs, the role of rare deletions or duplications has only been addressed here, by exploring selected loci harbouring known rare CNVs in extreme obese patients (20
). In order to comprehensively investigate whether certain rare CNVs [e.g. large (>1 Mb) CNVs (cf. 22
)] predispose to obesity, a more comprehensive quality control (QC) and validation of each single CNV call as well as a much larger adequately powered analysis sample is needed. It is conceivable that such assessments will become feasible within large-scale consortia such as GIANT [Genetic Investigation of ANthropometric Traits (8
In summary, we provide evidence for a new common CNVR for early-onset (extreme) obesity on chromosome 11q11 covering the coding regions of OLR genes. Moreover, we confirmed two previously described common CNVRs for obesity. Overall, however, common CNVRs seem to be of minor importance to account for the ‘missing heritability’ of early-onset (extreme) obesity.