Copy number variants (CNVs) are structural variants that include duplications or deletions of genomic regions ranging from 50 bp to several megabases (Zarrei et al., 2015). These structural variants can arise by different mutational mechanisms such as replication, repair of double-strand breaks and recombination (Gu et al., 2008; Carvalho and Lupski, 2016). CNVs contribute to human diversity but also to several diseases, including neurodevelopmental disorders and severe obesity (Wang et al., 2010; Wheeler et al., 2013; D'Angelo et al., 2018; Saeed et al., 2020).

Deletion in 16p11.2 region is the most established CNV linked to early-onset severe obesity (Bochukova et al., 2010; Walters et al., 2010). CNVs also contribute to syndromic obesity (Vuillaume et al., 2014; Pettersson et al., 2017; D’Angelo et al., 2018; Selvanayagam et al., 2018; Saeed et al., 2020; Madeo et al., 2021). The identification of rare CNVs in obesity has led to the discovery of several novel candidate genes. However, in many cases the disease-causing gene within the region of CNVs is unclear. Indeed, unraveling the pathogenicity of CNVs remains a major challenge (Nowakowska, 2017). Pathogenic CNVs are usually large and contain genes with high evolutionary copy number conservation. It has been proposed that dosage sensitivity of individual genes within the region of CNVs is a frequent cause of CNV pathogenicity. Haploinsufficiency is commonly used as a model of gene dosage sensitivity and predictions of haploinsufficiency have been used to predict the impact of CNVs (Huang et al., 2010; Lek et al., 2016; Rice and McLysaght, 2017).

We previously identified rare CNVs enriched in subjects with severe childhood obesity (Pettersson et al., 2017). In a case-control study of 90 obese subjects and 67 normal-weight controls, using array comparative genomic hybridization, we identified rare CNVs in 17 obese subjects. However, the pathogenicity of several of the identified CNVs remained unclear and the potential phenotypic contribution of the genes within the CNVs is not known. We hypothesized that rare variants in the genes within the CNVs might contribute to the development of obesity. We designed a custom-made targeted exome sequencing panel, including all genes located in the regions of the rare CNVs identified in our cohort of patients with childhood obesity, in search for rare variants and novel candidate genes.

CNVs are causing several diseases. However, in many cases it is challenging to determine the pathogenicity and clinical impact of the CNVs and to pinpoint the potential contribution of specific genes to the phenotype. In addition, the clinical relevance of rare CNVs can be difficult to determine because of incomplete penetrance and variable phenotypic expression. We performed targeted exome sequencing to identify rare variants in 117 genes located in previously identified CNV regions in subjects with severe early-onset obesity (Pettersson et al., 2017) in search for potential candidate genes. We found rare heterozygous variants with allele frequencies <0.1% in twelve of the CNV genes.

The 16p11.2 deletion is the most common CNV linked to severe obesity. However, the mechanism by which the proximal 16p11.2 deletion causes the phenotype is poorly understood and none of the genes in this region have so far been identified as a solely causative gene for the complex phenotype associated with proximal 16p11.2 CNVs. The patients with 16p11.2 deletions often present with various neurodevelopmental and neurobehavioral conditions, including autism spectrum disorders (ASD) and speech and language problems (Zufferey et al., 2012; Chung et al., 2021). The link between obesity and neurodevelopmental disorders may be due to common genes involved in the neurodevelopment that affect pathways regulating energy balance. Recently, new approaches for fine-mapping the genes in 16p11.2 region have been developed using GWAS and biobank data. Vysotskiy et al. performed an association analysis between imputed gene expression levels and BMI and they found four genes in the 16p11.2 region (TMEM219, SPN, TAOK2 and INO80E) associated with BMI (Vysotskiy et al., 2021). We identified rare heterozygous variants in four genes in the proximal 16p11.2 region; a frameshift variant in TAOK2 and heterozygous missense variants in SEZ6L2, ALDOA and KIF22. These four genes are potential candidate genes for the phenotype in patients with 16p11.2 CNVs. TAOK2 has a pLI of 1, suggesting rare loss-of-function variants may be deleterious (Lek et al., 2016). TAOK2 has been implicated in neurodevelopmental disorders and TAOK2 mutations found in patients with ASD showed impaired dendrite and synaptic development; the BMIs in the affected patients were not reported (Richter et al., 2019).TAOK2 heterozygous and knockout mice show gene brain morphological and behavioral abnormalities (Richter et al., 2019). SEZ6L2 has also been proposed as a candidate gene for ASD because of high expression in the brain (Kumar R. A. et al., 2009; Konyukh et al., 2011). Subcutaneous adipose tissue expression of ALDOA has been correlated with weight regain after dietary intervention (Roumans et al., 2017). ALDOA encodes an enzyme involved in glycolysis and the upregulation of glucose intake may be related to the risk of weight regain (Roumans et al., 2017). Furthermore, analysis of human fetal gene expression data found that KIF22 and ALDOA are significantly enriched in progenitors compared to post-mitotic cells making them candidates for having specific roles in neurogenesis in the developing human fetal cerebral cortex. It has been suggested that neurogenesis is disrupted in 16p11.2 CNVs (Morson et al., 2021). The participants in our study with rare variants in ALDOA, KIF22, TAOK2 and SEZ6L2 did not present with ASD or neurodevelopmental disorders (Table 1). However, the subjects with the missense variants in KIF22 and SEZ6L2 had learning difficulties. To our knowledge, no variants in these four genes have been reported in monogenic obesity. We consider the contribution of the identified rare variants to development of obesity unclear.

Two of the CNVs (16p11.2 deletion and 22q11.21 duplication) that we previously identified in our obese cohort include over 20 genes. In large CNVs, several genes within the locus may contribute to the phenotype. Disruption of multiple genes may be responsible for the complex phenotype seen in patients with 16p11.2 deletions, explaining the difficulty of identifying a single causative gene in the 16p11.2 region. Jensen et al. proposed that haploinsufficiency of certain genes may be altered by haploinsufficiency of other genes in the same CNV region and that these genes interact with each other via shared biological pathways (Jensen and Girirajan, 2019).

Interestingly, we identified heterozygous missense variants in SORCS1 in two subjects. Another study investigating genetic variants in 72 individuals with obesity also found two rare missense SORCS1 variants of unknown significance (Ginete et al., 2021). SORCS1 is highly expressed in the central nervous system and encodes a vacuolar protein sorting 10 domain-containing receptor that binds neuropeptides. SORSC1 has been associated with insulin secretion and diabetes risk (Goodarzi et al., 2007) and loss of both Sorcs1 and Sorcs3 in mutant mice caused increased adiposity and enhanced food intake, but not increased body weight (Subkhangulova et al., 2018). The participants in our study with the rare variants in SORSC1 did not present with insulin resistance or diabetes. In each family the variants were inherited from a parent with overweight, but overweight/obesity was present also in family members without the variant and therefore the contribution of these variants to the phenotype remains uncertain. More studies are needed to investigate the molecular mechanisms of SORCS1 and the possible contribution of this gene to early-onset obesity.

Another gene of interest is XRN1 located in 3q23 region. XRN1 is highly constrained for both loss-of-function and missense variants. This gene encodes 5′-3′ Exoribonuclease 1, which is involved in replication-dependent histone mRNA degradation. No XRN1 variants in humans have been associated with obesity, but a recent study investigating forebrain-specific Xrn1 knockout mice found that lack of Xrn1 in neurons led to obesity, hyperphagia, leptin resistance and hyperglycemia. The researchers proposed that Xrn1 in the hypothalamus is important for the mRNA degradation in regulating gene expression in the leptin signaling pathway (Takaoka et al., 2021). In our study, we identified a heterozygous missense variant in XRN1 of uncertain significance in a participant with isolated obesity. Further studies are required to explore the clinical relevance of this variant and biological function of XRN1.

Our study was limited by the lack of a control group of normal-weight subjects. However, we compared the allele frequencies of the variants we found in this study with the Finnish population in gnomAD and SISu project, with genotype data of more than 10,000 Finns. Because of the methodological limitation of targeted exome sequencing we may have missed variants in non-coding parts of the selected genes and variants in novel genes that were not included in the panel but may contribute to obesity. We may also have missed more common variants potentially contributing to the phenotype. Other limitations of this study were unavailability of parental DNA samples in part of the cohort and the lack of functional evaluation of the genetic variants. Further studies are needed to explore the possible role of the identified variants in the development of obesity.

In conclusion, the genes and mechanisms mediating the obesity phenotype in the previously identified CNVs remains uncertain and more studies are needed to provide insight to the potential contribution of the rare CNVs and the specific genes to severe obesity.