Abstract
Monogenic forms of obesity have been identified in ≤10% of severely obese European patients. However, the overall spectrum of deleterious variants (point mutations and structural variants) responsible for childhood severe obesity remains elusive. In this study, we genetically screened 225 severely obese children from consanguineous Pakistani families through a combination of techniques, including an in-house–developed augmented whole-exome sequencing method (CoDE-seq) that enables simultaneous detection of whole-exome copy number variations (CNVs) and point mutations in coding regions. We identified 110 (49%) probands carrying 55 different pathogenic point mutations and CNVs in 13 genes/loci responsible for nonsyndromic and syndromic monofactorial obesity. CoDE-seq also identified 28 rare or novel CNVs associated with intellectual disability in 22 additional obese subjects (10%). Additionally, we highlight variants in candidate genes for obesity warranting further investigation. Altogether, 59% of cases in the studied cohort are likely to have a discrete genetic cause, with 13% of these as a result of CNVs, demonstrating a remarkably higher prevalence of monofactorial obesity than hitherto reported and a plausible overlapping of obesity and intellectual disabilities in several cases. Finally, inbred populations with a high prevalence of obesity provide unique, genetically enriched material in the quest of new genes/variants influencing energy balance.
Introduction
The monogenic forms of obesity have defined the current concepts of the central regulation of energy balance and have opened new avenues for precision medicine (1,2). Monogenic nonsyndromic obesity is due to pathogenic mutations in genes involved in leptin-melanocortin signaling, resulting in extreme, early-onset obesity with an insatiable craving for food (2). In addition to excessive adiposity, syndromic obesity associates with other abnormalities such as dysmorphic features, intellectual disability, and organ-specific anomalies (3).
Pathogenic variations causing severe obesity include not only point mutations but also copy number variations (CNVs) (4). We have recently developed a new strategy that is based on an augmented whole-exome sequencing (WES) method, named CoDE-seq, that enables accurate and cost-effective detection of point mutations and CNVs in one step, thus expediting comprehensive genetic diagnosis (5).
Although only 5–10% of severe, early-onset obesity cases in outbred populations have been evidenced to have a monogenic condition, we previously reported a high prevalence of homozygous mutations in LEP, LEPR, MC4R, and ADCY3 in a Pakistani consanguineous population (6–9). Compared with other countries where consanguinity is practiced, the inbreeding coefficient reported for the Pakistani population is among the highest, with 60–65% of consanguineous marriages (10,11). Since Pakistan’s population comprises several endogamous subethnic groups that are more or less isolated, the country has a very high genetic diversity, which optimizes comprehensive genetic studies, particularly in obesity because Pakistan has the ninth highest prevalence of obesity worldwide (12).
Here, we describe 110 (49%) genetically elucidated cases in the world’s largest cohort of consanguineous subjects with severe, early-onset obesity (n = 225). Another 10% (n = 22) of obese case subjects were found to carry potentially causative CNVs associated with intellectual disability. Altogether, 59% of cases in the studied cohort are likely to have a genetic cause, with 13% of these potentially as a result of CNVs.
Research Design and Methods
Participants
The investigation is based on 225 unrelated probands 0.2–22 years of age with severe, early-onset obesity and their family members from consanguineous families. Subjects were recruited on a voluntary basis from hospital pediatric units located in the province of Punjab, Pakistan. Patient/parent written informed consent was obtained for each subject. The study was approved by the institutional ethical committees. The selection criteria included a BMI SD score (SDS) for age ≥3 (using WHO Anthro version 3.2.2 and AnthroPlus), early-onset obesity and hyperphagia, and nonobese parents (with BMI of ≤30 kg/m2) of first- or second-degree relation. Family and medical history was obtained, and pedigrees spanning at least three generations were constructed. Anthropomorphic measurements were made, and a blood sample was obtained for subsequent genomic DNA extraction and hormone estimations.
Initial Screening of LEP and MC4R
DNA from all probands (n = 225) was screened for coding regions of LEP and MC4R by direct sequencing as previously described (7).
WES and CoDE-seq
The probands found negative for mutations in LEP and MC4R (n = 167) and their family members (n = 177) were screened using WES and DNA arrays or CoDE-seq. A first batch of 62 probands and 56 family members were screened by standard WES and DNA arrays (for CNV detection) as previously described by us (9). A second batch of 105 probands and 121 family members were screened with CoDE-seq (when the technology was available and validated). This technique combines standard capture targeting the whole exome (NimbleGen SeqCap EZ MedExome Target Enrichment) with an in-house–designed capture (NimbleGen SeqCap EZ Choice XL) (5). Sequencing was performed on an Illumina NovaSeq 6000 system. A mean sequencing depth of ∼100× was achieved for each individual using 150-bp paired-end reads. Computer analyses for variants detection and annotation have previously been described (5).
Variant Prioritization
The analysis, at first stage, was focused on homozygous variants. We excluded homozygous variants in obese probands that were also homozygous in family members with BMI ≤30 kg/m2 or BMI SDS for age <2. The variants with an allele frequency of >0.001% in the Genome Aggregation Database (gnomAD) were also ignored. The resulting list of rare mutations was searched in a list of known monogenic obesity genes (Supplementary Table 1). Subsequently, rare compound heterozygous variants (in genes with recessive inheritance) and heterozygous variants (in genes with known dominant inheritance) were also analyzed. All potentially causative variants were graded according to criteria of the American College of Medical Genetics and Genomics (13).
CNV Detection and Prioritization
For detection of CNVs from the genotyping data, integrated hidden Markov model algorithm (PennCNV) with its default method was used as described previously (8,14). For detection of CNVs from CoDE-seq data, we used eXome Hidden Markov Model (XHMM, version 1.0) as previously described (5,15). We excluded all CNVs with allele frequency >0.002 in the gnomAD structural variant data set. Furthermore, we looked for all well-known pathogenic CNVs on the basis of the literature, the Database of Chromosomal Imbalance and Phenotype in Humans Using Ensembl Resources (DECIPHER), and the Database of Genomic Variants, as listed in our previous work (5).
Biochemical Analysis
Serum leptin, insulin, and cortisol concentrations were determined by commercially available ELISA kits. Assays were performed in duplicate with an automated analyzer (Bio-Rad Laboratories, Hercules, CA). The intra- and interassay variations were <11%.
Data and Resource Availability
The data sets generated during the current study are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.
Results
The genetic screening of 225 unrelated obese subjects, initially by Sanger sequencing (for LEP and MC4R only) followed by WES or CoDE-seq, revealed 55 pathogenic genetic variants (including point mutations or CNVs) in 110 probands (49% of the overall cohort) (Tables 1 and 2 and Fig. 1).
Genetic and clinical data of probands with a pathogenic point mutation in genes associated with obesity
Genetic and clinical data of probands with CNVs causing obesity and/or intellectual disability
Genetic etiologies of obesity in a cohort of 225 patients from Pakistani consanguineous families.
LEP: 47% of Elucidated Cases
We identified seven homozygous pathogenic mutations in LEP carried by 52 probands (Table 1 and Fig. 1). These included two frameshift (p.G133Vfs*15 and p.Y140Tfs*8), one in-frame (p.I35del), one splice site (c.-29+1G>C), and three missense variations (p.D100N, p.N103K, and p.R105Q). Of these mutations, c.-29+1G>C (identified in six unrelated probands) and p.Y140Tfs*8 (identified in one proband) were novel (i.e., not reported in the literature or listed in gnomAD) (Supplementary Fig. 1). The remaining five LEP mutations have previously been reported (2). Of note, p.G133Vfs*15 was identified in 39 unrelated children and, by itself, accounted for obesity in 38% of the elucidated cases.
Besides excessive obesity and hyperphagia, the leptin-deficient probands often presented with respiratory infections, hepatomegaly, undescended testes, and delayed milestones. As anticipated, in leptin-deficient children, leptin levels were <0.2 ng or nondetectable (Supplementary Tables 2 and 3), with the exception of three probands with p.D100N and p.N103K variants with mean levels of 46 ± 12 ng/mL (Table 1 and Supplementary Table 4). Serum insulin and cortisol levels were variable but mostly in the higher range (Supplementary Tables 2–4).
LEPR: 15% of Elucidated Cases
We identified 14 homozygous pathogenic variants of LEPR in 17 probands (Tables 1 and 2 and Fig. 1). These included four splice-site (c.704–1G>A, c.2213–3C>G, c.2396–1G>T, and c.2396–2A>G), three frameshift (p.E580Kfs*37, p.A967Dfs*7, and p.S1090Wfs*6), one nonsense (p.W705*), and three missense mutations (p.E14K, p.N718S, and p.P876L); two copy-loss CNVs of 44.4 and 61 kb each; and a unique change at translation initiation site of LEPR (c.2T>C/p.?). Twelve of these LEPR variants were novel (Supplementary Fig. 2).
In contrast to previous findings (16), LEPR-deficient probands were phenotypically indistinguishable from leptin-deficient subjects. Apart from excessive obesity and hyperleptinemia (mean 41 ± 7.6 ng/mL, n = 15) (Supplementary Table 2), no other noticeable clinical problems except respiratory infection and/or delayed milestones in three subjects were reported.
MC4R: 11% of Elucidated Cases
We identified eight pathogenic homozygous mutations in MC4R in 12 probands that included three nonsense (c.47G>A/p.W16*, c.48G>A/p.W16*, and p.Y21*), one frameshift (p.Y212Sfs*5), one in-frame (p.F201_M204del), and three missense mutations (p.I69T, p.M161T, and p.R165W) (Table 1). Three of these variants were novel (Table 1 and Supplementary Fig. 3). In addition to excessive adiposity, hyperinsulinemia (44 ± 11 μIU/mL; n = 12) was recorded in most MC4R-deficient probands (Supplementary Table 2). No other significant clinical abnormalities were observed (Table 1).
Monogenic Syndromic Obesity (Bardet-Biedel Syndrome, ALMS1, and Prader-Willi Syndrome): 23% of Elucidated Cases
Twelve homozygous and two compound heterozygous mutations were identified in six genes linked to Bardet-Biedel syndrome (BBS) (BBS1, BBS2, BBS5, MKKS [BBS6], BBS9, and BBS10) in 17 probands (Table 1). The 12 homozygous mutations identified here included 6 frameshift, 2 splice-site, and 3 missense variants, and 1 was a substitution affecting the translation initiation codon (Table 1). Carriers of BBS gene mutations presented with central obesity and hyperphagia accompanied with other dysmorphic features (Table 1).
Four novel homozygous pathogenic, nonsense mutations in ALMS1 were identified in four male subjects with severe obesity (Table 1). The majority of probands with these mutations presented hyperinsulinemia (Supplementary Table 2).
Furthermore, we identified three probands carrying heterozygous deletions in the 15q11–13 Prader-Willi syndrome (PWS)–associated region (Table 2). Whereas one subject carried the typical 4.5-Mb deletion in the PWS-associated region, the other two subjects carried unique deletions of ∼3.3 and ∼3.8 Mb each (Supplementary Figs. 4–6). All three probands presented typical features of PWS (Table 1).
ADCY3: 4% of Cases
A girl with severe obesity was found with a unique 23-bp homozygous deletion (c.2173–10_2185del) in ADCY3 (Table 1). In the same cohort, we have previously reported three other homozygous loss-of-function mutations in this gene (9).
Variants of Uncertain Significance: Point Mutations in Obesity Genes
We identified a homozygous novel mutation in VPS13B (p.P2207T), CEP19 (p.M38T), and BBS9 (p.Q748L). These mutations were variants of uncertain significance (VUS) (Supplementary Table 5).
Pathogenic Copy-Loss CNV in Chromosome 16p11.2
A proband was identified with an ∼650-kb heterozygous deletion of chromosome 16p11.2 (Supplementary Fig. 7). Besides severe obesity, delayed milestones and intellectual disability were reported in the carrier.
CNVs Associated With Intellectual Disability and Potentially Causing Obesity (n = 22)
In addition to CNVs in known obesity-causing genes/regions, we found 47 CNVs, including 10 rare CNVs (allele frequency 0.001 in gnomAD) and 37 novel CNVs, in coding regions. These comprised 25 copy-loss and 11 copy-gain CNVs (Table 2). Of these, 36 CNVs have previously been associated with intellectual disability. Importantly, 28 of these 36 CNVs (21 copy loss and 7 copy gain) were found in 22 subjects who were negative for any other mutation (Table 2 and Fig. 1).
Novel CNVs Overlapping Genes Potentially Associated With Obesity (n = 4)
Four novel CNVs affecting candidate genes for energy balance impairment were also identified. This includes a small 8-kb heterozygous copy-loss CNV in SLC5A4 (Table 2). SLC5A4 is involved in the neuronal glucose sensing mechanism and control of food intake (17). Furthermore, we found a novel heterozygous 473-kb copy-loss CNV involving ATRNL1. This gene, mainly expressed in the brain, modulates the melanocortin signaling pathway (18). A 366.3-kb copy-gain CNV encompassing FGFR2 and ATE1 was identified in a proband also deficient for BBS5 (Table 1). ATE1 is reported to affect adipogenesis and adipocyte function (19). Another 858-kb copy-gain CNV affecting the LIPI gene, which is involved in regulation of fat metabolism (20), was identified in a severely obese proband (Table 2).
Discussion
The main result of this largest genetic study of patients with childhood-onset, severe obesity is genetic elucidation of obesity in 110 (49%) probands from a consanguineous population. This unexpectedly high percentage includes loss-of-function point mutations or CNVs in loci classically known to cause obesity. We confirm our earlier findings that leptin deficiency alone explains obesity in one-fifth of patients, making leptin deficiency the predominant etiology of monogenic obesity in this Pakistani population (7,8,21). The proportion of elucidated variants may even be much higher because we also identified an additional 22 cases with rare or novel CNVs linked to intellectual disabilities often associated with a severely obese phenotype. These cases provide credence to the notion that loci causing intellectual disability may also be involved in obesity (4). Thus, altogether, we elucidate up to 132 (59%) obesity cases, excluding severely obese patients in which VUS were identified in obesity genes (Supplementary Table 5).
This investigation highlights the importance of CNVs in the diagnosis of obesity. Although suggested by some of the pioneering studies (4,22), this hypothesis has so far received little attention from the clinical viewpoint (23) possibly because of the high cost of microarray technology, thus restricting its use to syndromic phenotypes only. In the current study, CoDE-seq has allowed us to directly detect CNVs in addition to point mutations (5). This comprehensive genetic screening in obesity is powerful because our retrospective clinical investigations unraveled the yet-undiagnosed intellectual disability in the probands carrying CNVs.
In conclusion, we show that next-generation sequencing approaches make it possible to uncover the genetic causality of severe obesity in a large proportion of subjects from the consanguineous population of Pakistan. An unrelenting quest for the discovery of new genes and variants, and associated pathways predisposing to obesity, is crucial in the development of specific and effective pharmacotherapies for the treatment of obesity.
Article Information
Acknowledgments. The authors thank the patients and their families for participation in the study. The authors are grateful to H. Crouch (Imperial College London) for help in single nucleotide polymorphism array experiments and I. Qureshi (University of Lahore) for technical assistance. The authors also thank F. Allegaert and N. Larcher (CNRS-UMR 8199–European Genomic Institute for Diabetes) for DNA extraction and storage.
Funding. This study was supported by the Fédération de Recherche 3508 LabEx European Genomics Institute for Diabetes (ANR-10LABX-46), the ANR EquipEx 2010 session (ANR-10-EQPX-07-01, “LIGAN-PM”), the European Community (FEDER), and the Region Hauts-de-France. The research leading to this study was also supported by funding from the Pakistan Academy of Sciences (PAS003 to M.A.) and the European Research Council (GEPIDIAB 294785 to P.F.).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.S., M.A., S.M.D., Q.M.J., Q.A., and L.I. collected samples and performed biochemical analysis. S.S., M.A., A.Bo., and P.F. designed the study and wrote the first draft of the manuscript. S.S., E.V., E.D., M.D., S.A., A.Ba., L.B., and A.Bo. performed whole-exome sequencing and analyzed the genetic data. J.M., H.A., W.I.K., and T.A.B. identified and recruited obese families. S.L. carried out microarray experiments. S.G. managed the database. All authors contributed to the final version of the manuscript. S.S. and P.F. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Footnotes
This article contains supplementary material online at https://doi.org/10.2337/figshare.12200843.
- Received December 16, 2019.
- Accepted April 25, 2020.
- © 2020 by the American Diabetes Association
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.