International Journal of Obesity (2020) 44:830–841 https://doi.org/10.1038/s41366-019-0357-5 ARTICLE Pediatrics Heterozygous rare genetic variants in non-syndromic early-onset obesity Clara Serra-Juhé1,2 Gabriel Á. Martos-Moreno3,4 Francesc Bou de Pieri1,2 Raquel Flores1,2 Julie A. Chowen ● ● ● ● 3,4 ● Luis A. Pérez-Jurado 1,2,5 Jesús Argente 3,4,6 ● Received: 3 November 2018 / Revised: 7 January 2019 / Accepted: 7 February 2019 / Published online: 29 March 2019 © The Author(s) 2019. This article is published with open access Abstract Background Obesity is a very heterogeneous disorder at both the clinical and molecular levels and with high heritability. Several monogenic forms and genes with strong effects have been identified for non-syndromic severe obesity. Novel therapeutic interventions are in development for some genetic forms, emphasizing the importance of determining genetic contributions. Objective We aimed to define the contribution of rare single-nucleotide genetic variants (RSVs) in candidate genes to non- 1234567890();,: 1234567890();,: syndromic severe early-onset obesity (EOO; body mass index (BMI) >+3 standard deviation score, <3 years). Methods Using a pooled DNA-sequencing approach, we screened for RSVs in 15 obesity candidate genes in a series of 463 EOO patients and 480 controls. We also analysed exome data from 293 EOO patients from the “Viva la Familia” (VLF) study as a replication dataset. Results Likely or known pathogenic RSVs were identified in 23 patients (5.0%), with 7 of the 15 genes (BDNF, FTO, MC3R, MC4R, NEGR1, PPARG and SIM1) harbouring RSVs only in cases (3.67%) and none in controls. All were heterozygous changes, either de novo (one in BDNF) or inherited from obese parents (seven maternal, three paternal), and no individual carried more than one variant. Results were replicated in the VLF study, where 4.10% of probands carried RSVs in the overrepresented genes. RSVs in five genes were either absent (LEP) or more common in controls than in cases (ADRB3, LEPR, PCSK1 and PCSK2) in both obese datasets. Conclusions Heterozygous RSVs in several candidate genes of the melanocortin pathway are found in ~5.0% patients with EOO. These results support the clinical utility of genetic testing to identify patients who might benefit from targeted therapeutic intervention. Introduction Obesity is the most prevalent chronic childhood disease in the occidental world and a risk factor for later obesity- and These authors contributed equally: Clara Serra-Juhé, Gabriel Á. metabolic-related disorders. It is a heterogeneous multi- Martos-Moreno, Francesc Bou de Pieri factorial disease with high heritability (50–75%) that is These authors jointly supervised this work: Luis A Pérez-Jurado, Jesús undoubtedly higher in severe early-onset cases [1–3]. The Argente genetic causes underlying obesity remain largely unknown with an important proportion of information missing Supplementary information The online version of this article (https:// doi.org/10.1038/s41366-019-0357-5) contains supplementary regarding its heritability. material, which is available to authorized users. A small percentage of obesity occurs as part of a syn- dromic entity [4–10], but in the majority, obesity is not * Jesús Argente accompanied by other specific phenotypes. Highly pene- jesus.argente@uam.es trant rare genetic variants, mainly autosomal recessive, * Luis A. Pérez-Jurado luis.perez@upf.edu affecting at least eight genes [LEP (MIM 164160), LEPR (MIM 601007), MC4R (MIM 155541), PCSK1 (MIM Extended author information available on the last page of the article 162150), POMC (MIM 176830), MC3R (MIM 155540), Heterozygous rare genetic variants in non-syndromic early-onset obesity 831 SIM1 (MIM 603128) or NTRK2 (MIM600456)] are and complementary tests to rule out syndromic, endocrine reportedly found in 2–5% of non-syndromic obese patients or secondary forms of obesity. These evaluations included [11–17]. However, multigenic or multifactorial inheritance an interview with a dietitian to determine eating habits such is the best model to explain the aetiology of most cases. as the frequency, speed and amount of food ingested. A Common genetic variants involved in these forms of obesity molecular karyotype by microarray and a methylation- have been identified by genome-wide and candidate-gene sensitive multiple-ligation probe amplification to discard association studies. Genetic variants in genes such as specific entities that might clinically overlap with isolated ADRB3 (MIM 109691), BDNF (MIM 113505), FTO (MIM obesity were also performed [5]. 610966), GHSR, NEGR1 (MIM 613173), PCSK2 (MIM A sample of 463 subjects with severe EOO was selected. 162151), PPARG (MIM 601487) and TMEM18 (MIM Blood was collected from patients and both parents when 613220) [18–22], among others, have been related to body possible to study the segregation of the genetic variants with mass index (BMI), but the strongest effect of a single the phenotype in the family. DNA was isolated from total common variant might explain about only 2% of the var- blood using the Gentra Puregene Blood kit (Qiagen) iance in BMI. Similar to other complex traits [23, 24], copy according to the manufacturer’s instructions. number variants (CNVs) also contribute to the genetic basis A total of 480 adult subjects with normal weight (BMI of obesity [25–28]. Identification of a potential genetic 80%) were discarded as likely false positives. 1000 genomes, Kaviar, Spanish variant server) [37] higher Custom-made Perl scripts were used to automate sample than 1/1000 or described in a homozygous state in at least identification crossing the pools where the same variant was one individual, assuming that variants with a relatively high detected. The proportion of variants in each pool was also frequency in the general population are unlikely high considered in the pipeline as an estimation of allele fre- penetrant variants related to the severe phenotype of EOO. quency. Although obesity is a highly heterogeneous dis- All nonsense or frameshift, canonical splicing and mis- order, the same RSV might appear twice (present in two sense variants predicted as damaging (Condel >0.522 and alleles from either the same or different samples), expected CADD >20) [38, 39] were considered as highly likely to be in 5% of the reads. Consequently, the first step was filtering pathogenic (from now on called likely pathogenic variants). the variants present in between 1 and 10% of the mapped All previously reported mutations with experimentally reads. Variants with a lower percentage were probably validated functional consequences were also included [40]. artefacts and variants with a higher percentage (present in To define pathogenicity, we also considered the constraint three or more samples in the same pool) were thought too metrics for variation tolerance of each gene. Variation- Heterozygous rare genetic variants in non-syndromic early-onset obesity 833 intolerant genes were considered those with pLi (probability variants detected with the pooled DNA approach were of loss-of-function intolerance) >0.50 and missense Z-score selected and all (23/23, 100%) validated by Sanger sequen- >0 [37]. cing. Although the pooled DNA approach may not detect a small percentage of real variants, it does not capture false Variant validation and co-segregation positives. These results demonstrate that the strategy used is efficient to identify RSVs in large cohorts of samples. To validate RSVs detected by pooled DNA-sequencing, PCR amplicons from individual DNA samples were Burden of RSVs in EOO obtained and sequenced by Sanger technology. The same technique was used to analyse parental samples when A total of 73 different RSVs were selectively identified in available to define inheritance pattern and possible co- only one group (Table 1): 43 variants in 48 subjects of the segregation of the RSV with the phenotype in each family. patient group and 30 variants in 31 controls. This difference was statistically significant (odds ratio (OR) = 1.61; p = Replication cohort 0.0342; Table 1 and Supplemental Tables 1 and 2). We used very stringent criteria of pathogenicity to better define the Exomes from 293 unrelated patients with EOO from the putative clinically relevant variants: all loss-of-function Viva la Familia (VLF) study [41] were used to replicate the (nonsense, frameshift and canonical splice-site variants) and results found in our initial cohort. VLF comprises obese missense variants predicted as pathogenic by two stringent probands with BMI >95th percentile, between 4 and 19 algorithms (Condel and CADD) were considered as years old, from Hispanic families in Houston. Sequence likely pathogenic. Likely or known pathogenic (previously read archive (SRA) files were downloaded from publicly validated) RSVs were found in 23 patients and 8 controls available servers (dbGAP: phs000616.v2.p2). Variant call- (OR = 2.98; p = 0.0055). Most RSVs were present in a ing in the 15 candidate genes was done with GATK [42] single sample, but a few cases with two samples harbouring and filtering was performed using the same criteria as the same RSV were found and three patients shared a var- above. iant in FTO. A subset of sequenced genes accumulated the differential Ethics statement mutational load found in patients compared to controls: BDNF, FTO, MC3R, MC4R, NEGR1, PPARG and SIM1. The project was approved by the Medical Ethical Com- Analysing this group of genes separately, a total of 30 mittee of Hospital Infantil Universitario Niño Jesús in patients with RSVs were identified in patients versus four Madrid, Spain and is in accordance with the “Ethical controls (OR = 7.78; p < 0.0001; Table 1) and four of these Principles for Medical Research Involving Human Sub- seven overrepresented genes (BDNF, NEGR1, PPARG and jects” adopted in the Declaration of Helsinki by the World SIM1) are variation intolerant (Supplemental Table 3). Medical Association (64th WMA General Assembly, For- Analysing the likely pathogenic variants in the subset of taleza, Brazil, October 2013) and Spanish data protection genes, RSVs were identified in 17 patients (3.67%) and no act (Ley Orgánica 15/1999 de Protección de Datos). Written controls (p < 0.0001; Table 1). Two of the variants identi- informed consent was obtained from all patients or their fied were present in two unrelated patients (Table 2). All 17 legal guardian after a complete description of the study. likely pathogenic variants identified in the overrepresented subset of genes heterozygous and no individual carried more than one likely pathogenic variant (Tables 1 and 2). Results In the remaining subset of genes, 18 RSVs were identi- fied in patients and 27 in controls (not statistically sig- RSV detection sensitivity and specificity nificant), including 6 likely pathogenic variants in patients and 8 in controls (not statistically significant; Supplemental The mean sequence coverage of patient pools was 2396× Tables 1 and 2). (SD: 574) and 3428× (SD: 660) in controls, with >98% of targeted sites showing coverage above 400×. The sensitivity Heterozygous RSVs in patients with EOO and specificity of this approach were assessed. Six of the seven known variants in samples included in the study were Among the genes with higher mutational load in patients, blindly identified with the filters mentioned above, yielding a MC4R was the most frequent with seven RSVs in patients sensitivity of 86%. The only variant not identified had low- (1.51%). Of these, five have been reported as pathogenic by quality scores. As proof of concept for high specificity, 23 demonstration of partial (H76R) or complete (S127L, 834 C. Serra-Juhé et al. Table 1 RSVs identified per All RSVs Likely pathogenic RSVs gene in each one of the analysed groups: EOO-Sp, n = 463; VLF, EOO-Sp (%) VLF (%) Controls (%) EOO-Sp (%) VLF (%) Controls (%) n = 293; and controls, n = 480 ADRB3 1 (0.22) 1 (0.34) 4 (0.83) – – 1 (0.21) BDNF 4 (0.86) 1 (0.34) – 3 (0.65) 1 (0.34) – FTO 4 (0.86) 4 (1.37) 1 (0.21) 1 (0.22) – – GHSR 2 (0.43) 6 (2.05) 3 (0.63) – 1 (0.34) – LEP – – – – – – LEPR 7 (1.51) 5 (1.71) 5 (1.04) 2 (0.43) 2 (0.68) 3 (0.63) MC3R 3 (0.65) 3 (1.02) – 2 (0.43) 1 (0.34) – MC4R 7 (1.51) 6 (2.05) 2 (0.42) 6 (1.30) 4 (1.37) – NEGR1 3 (0.65) – – 1 (0.22) – – NTRK2 2 (0.43) 3 (1.02) 4 (0.83) 1 (0.22) 2 (0.68) 1 (0.21) PCSK1 4 (0.86) – 5 (1.04) 3 (0.65) – 1 (0.21) PCSK2 2 (0.43) – 5 (1.04) – – 2 (0.42) PPARG 3 (0.65) – – 2 (0.43) – – SIM1 6 (1.30) 2 (0.68) 1 (0.21) 2 (0.43) – – TMEM18 – 1 (0.34) 1 (0.21) – 1 (0.34) – 48 (10.37) 32 (10.92) 31 (6.46) 23 (4.97) 12 (4.10) 8 (1.67) Selected genes 30 (6.48) 16 (5.46) 4 (0.83) 17 (3.67) 6 (2.05) – Other genes 18 (3.89) 16 (5.46) 27 (5.63) 6 (1.30) 6 (2.05) 8 (1.67) Bold represents the selected genes. The number and percentage of individuals carrying rare and probably pathogenic genetic variants per gene are shown. The total burden of RSVs is significantly higher in EOO-Sp (48/463) than controls (31/480) (OR = 1.61; p = 0.0342), and in VLF compared to controls (32/293; OR = 1.69; p = 0.0306). The proportion of individuals carrying likely pathogenic RSVs was also significantly higher in EOO-Sp than in controls (23/463 vs. 8/480; OR = 2.98; p = 0.0055). Genes accumulating the RSV load in patients compared with controls are in bold: BDNF, FTO, MC3R, MC4R, NEGR1, PPARG and SIM1. The proportion of individuals carrying RSVs was significantly higher in EOO-Sp than in controls (30/463 vs. 4/480; OR = 7.78; p < 0.0001). This difference was also statistically significant in VLF (16/293; OR = 6.55; p = 0.0002). Considering only probably pathogenic variants, the proportion in patients (17/463 in EOO-Sp and 6/293 in VLF) was significantly higher than in controls (p < 0.0001 and p = 0.0029, respectively), where none were found (0/480). OR odds ratio, RSV rare sequence variant, OR odds ratio, EOO-Sp early-onset obesity-Spain, VLF Viva la Familia T150I, A259V and P272L) loss of function [13, 40]. The protein (Q206K) found in two siblings concordant with the two RSVs not previously described include a nonsense phenotype. These RSVs were inherited from an obese father change interrupting the protein in the 22nd codon (R22X) (splicing variant) and obese mother (missense variant) and a missense change (V52A) predicted as likely benign. (Table 2). The variant detected in a control subject and four In the five cases with parental samples available, the RSV of the RSVs in patients are not predicted to be pathogenic was inherited from obese mothers. Two missense variants in by our criteria. MC4R were detected in the control group, both predicted as In the BDNF gene, three RSVs were found in four benign and not previously described in patients. Three patients and none in controls. Two of the variants are pre- patients were also found to have RSVs in a related gene, dicted to be pathogenic. One of these RSVs was not found MC3R, with the likely pathogenic RSV (R302W) identified in parental samples, indicating that it occurred de novo in in two unrelated patients (Table 2). the patient. Parenthood was proven using microsatellite SIM1 also showed a high mutational load in patients markers. The other RSV (I231V) was found in two unre- (n = 6) compared to controls (n = 1). Two of the six var- lated cases, paternally inherited in the case with parental iants identified in patients are strongly predicted to be samples available (Table 2). pathogenic. One affects a canonical splice site in intron 3 We detected three RSVs (two likely pathogenic) in (c.352 + 1G>A) and the other is a missense variant on a PPARG in patients, as well as one single likely pathogenic highly conserved residue of the first Perl–Arnt–Sim of the missense RSV in FTO and NEGR1, and none in controls. Table 2 Main clinical features of the patients harbouring likely pathogenic RSVs in the overrepresented subset of genes Gender Gene Type gDNA cDNA Protein Inheritance Age at BMI Height Target Bone age Newborn Cognition Hyperphagia Metabolic Other comorbidities Familial obesity exam (SDS) (SDS) height acceleration weight and disturbances background (years) (SDS) (months) (SDS) behaviour Male MC4R Missense* Chr18:58038768 815C>T P272L* Maternal 6.84 +6.5 +4.3 +0.10 26 +3.2 NR + IR Precocious pubarche Mother (infancy) not adult Male Missense* Chr18:58038807 776C>T A259V* Maternal 6.92 +7.0 +1.2 −1.52 −11 +2.0 NR + None Macroorchydism Both parents and families Male Missense* Chr18:58039134 449C>T T150I* – 9.58 +5.8 +1.8 −0.90 42 −1.0 NR + IR + IGT + Liver steatosis Both parents HU and families Male Missense* Chr18:58039203 380C>T S127L* – 14.25 +14.0 −1.3 −2.33 14 −0.4 NR +++ IR + HT + None Both parents HU and families Male Missense* Chr18:58039356 227A>G H76R* Maternal 14.00 +5.8 +1.1 −0.81 33 +0.1 NR +++ IR + HU Liver steatosis Both parents and paternal family (grandfather and 3 aunts) Male Nonsense Chr18:58039519 64A>T R22X Maternal 5.00 +18.8 +1.0 −2.06 51 +5.7 NR +++ IR + HT + Liver steatosis Both parents HU and maternal family (morbid) Female MC3R Missense Chr20:54824803 904C>T R302W – 11.16 +4.4 +0.8 +0.57 10 −0.2 NR + IR + HU None Both parents and families Male Missense Chr20:54824803 904C>T R302W – 12.00 +3.8 −0.9 −1.00 4 −0.4 NR + None None Both parents and families Female SIM1 Missense Chr6:100896482 616C>A Q206K Maternal 8.75 +5.6 +1.8 +0.48 12 −1.1 NR + IR + IGT None Father and sister Male Splicing Chr6:100898138 352+1G>A – Paternal 8.66 +4.3 +0.7 −0.01 12 +1.16 NR + IR + IGT None Father and Heterozygous rare genetic variants in non-syndromic early-onset obesity paternal family (grandmother, 2 uncles and 1 aunt) Male PPARG Missense Chr3:12422920 326T>A I109N Maternal 9.33 +5.5 +4.2 +1.06 24 +5.1 NR + IR None Father Male Missense Chr3:12434173 457C>T R153W – 11.00 +4.1 −0.4 −1.57 0 −2.0 NR + IR + IGT + None Father and sister HT Male BDNF Missense Chr11:27679421 691A>G I231V Paternal 6.00 +3.9 +0.3 −0.76 -6 −0.4 NR + None None Mother and maternal family (grandfather) Female Missense Chr11:27679421 691A>G I231V – 11.75 +3.3 −0.1 −1.18 3 +0.1 NR +++ IR + HU Liver steatosis Mother Female Missense Chr11:27679691 421T>G C141G De novo 11.16 +6.1 +2.0 −0.41 17 +0.1 NR + IR + HU Liver steatosis Brother and paternal family (aunt, grandfather) Male NEGR1 Missense Chr1:72400858 313 A > G I105V Maternal 2.92 +8.8 +1.7 −0.11 3 +0.8 SP+BP +++ HU None Both grandmothers Female FTO Missense Chr16:53859890 238 C > T R80W Paternal 6.66 +3.8 +1.1 −0.57 14 −1.7 NR +++ IR + HT + Liver steatosis Sister, both HU parents and maternal family (grandmother, 2nd degree aunt) In columns “Type” and “Protein” the * indicates functional information available SDS standard deviation score, BMI body mass index, gDNA genomic DNA, cDNA complementary DNA, NR nothing relevant, SP speech delay, BP behavioural disturbances, IR insulin resistance, IGT impaired glucose tolerance, HU hyperuricaemia, HT hypertriglyceridaemia 835 836 C. Serra-Juhé et al. Clinical features of RSV carriers likely pathogenic variants (6/293 VLF patients versus 0/480 controls; p = 0.0029; Supplemental Table 5). The main phenotypic features of patients harbouring likely pathogenic RSVs in the selected genes are detailed in Table 2. The six patients harbouring pathogenic RSVs in Discussion MC4R were males with a BMI above + 5.5 standard deviation score (SDS), orexigenic impulsivity, hyper- Analysis of an elevated number of individuals is necessary insulinism, some overgrowth (height >+1 SDS over target to identify genetic factors with strong effects in hetero- height) and advanced bone age (14 months on average), a geneous diseases with important environmental influence, phenotype previously defined in patients with MC4R such as obesity. Pooled DNA-sequencing reduces the mutations [43, 44]. The patient with the A259V variant had number of sequencing reactions, thus reducing costs with- macroorchydism (8 cc volume testes despite Tanner stage out decreasing the number of samples. This approach must I), which is not a common feature of MC4R deficiency, be optimized to identify all relevant variants and discern given the lack of MC4R expression in the testis [45]. The between real variants and artefacts. Sequencing coverage patient with the novel nonsense mutation (R22X) had an above 400 reads in most positions would yield >10 reads of extremely high BMI (+18.8 SDS), uncontrollable orexi- any single RSV diluted in a pool of 20 different samples. genic impulsivity, overgrowth (+3 SDS), advanced bone Since sequence artefacts are detected in most mapped age (51 months over chronological age) and hyperinsulin- positions, quality filters were required to discriminate ism. His mother, also a carrier of the R22X mutation, had a between real variants and false positives. The percentage of BMI of 51.3 kg/m2. The patient with a predicted benign reads with the variant, strand bias and Phred score had the RSV at MC4R (V52A) was a girl with milder obesity (+4.2 greatest discriminatory ability. Full validation rate by San- BMI-SDS), no overgrowth (height at −1 SDS), average ger sequencing (23 of 23 variants) proved the high speci- bone age, and hyperinsulinism. ficity and suitability of the pooling strategy. The two patients with likely pathogenic RSVs in MC3R We failed to detect one of seven7 previously defined had milder obesity (BMI around +4 SDS) and no over- RSVs (sensitivity 86%). This incomplete detection rate growth. One had insulin resistance. might invalidate using pooling strategies for diagnostic Patients harbouring SIM1 RSVs had BMIs of 4.3 and 5.6 purposes, but it should not affect the results of this study. SDS, increased longitudinal growth (+0.7 and +1.3 SDS The average coverage of the patient pools was lower than over standardized target height), mild hyperphagia, insulin that of the control pools (2396× vs. 3428×). If this differ- resistance, and impaired glucose tolerance. ence were to generate bias when detecting RSVs, it would Patients with RSVs in the BDNF gene had BMIs ranging favour RSV detection in controls. A more similar coverage from +3.3 to +6.1 SDS. Patients with PPARG RSVs between groups might even increase the difference in bur- including one sibling (BMI +4.1 to +5.5 SDS) showed den between patients and controls. insulin resistance and three out of four had dyslipidaemia. We focused on RSVs in 15 candidate genes that could have a strong effect in a small subgroup of patients Replication cohort according to information available at the onset of this study. This strategy might have missed some variants with higher The results of the VLF study are shown in Table 1 and frequencies that also contribute to obesity. RSVs were Supplemental Table 4. The frequency of RSVs in VLF identified in 10.36% of patients with half of them (5.0%) replicated the results for BDNF, FTO, MC3R, MC4R and most likely being pathogenic, a burden significantly higher SIM1. No RSVs with the established criteria were detected in patients than in controls. This burden was attributable to a in NEGR1 or PPARG in the VLF cohort, but likely patho- subset of seven genes (BDNF, FTO, MC3R, MC4R, genic RSVs were identified in GHSR and TMEM18, absent NEGR1, PPARG and SIM1). The replication study corro- in controls. The increased rate of RSVs between VLF borated these results indicating that highly penetrant RSVs subjects (32/293, 10.92%) and controls (31/480, 6.46%) in in these genes contribute to part of the missing heritability all genes was similar to that of the Spanish EOO dataset of EOO. (OR = 1.69; p = 0.0306). The separate analysis of the When patients are grouped according to the gene muta- overrepresented subset of genes (BDNF, FTO, MC3R, ted, most groups are too small to establish conclusive MC4R, NEGR1, PPARG and SIM1) yielded a significantly clinical molecular correlations, but some common features higher proportion of VLF patients with RSVs (16/293) than can be discerned. Among the overrepresented genes, MC4R controls (4/480; OR = 6.55; p = 0.0002). This difference was the most frequent (1.51% of patients), similar to reports was also statistically significant when considering only revealing heterozygous and homozygous mutations in Heterozygous rare genetic variants in non-syndromic early-onset obesity 837 MC4R accounting for 1–6% of severe obesity in humans larger deletions encompassing BDNF along with the other [40, 43, 44]. Patients with RSVs in MC4R had hyperphagia critical genes (WAGRO syndrome) [10]. Disruption of with compulsive eating, very severe obesity (mean BMI + BDNF expression in a patient with a paracentric inversion 8.9 SDS), overgrowth, accelerated skeletal maturation and and deletion of the entire BDNF locus in a mother and child metabolic disturbances including insulin resistance. Beha- were associated with hyperphagia, severe obesity and vioural disturbances occurred mainly associated with food mildly impaired cognition with or without attention deficits seeking. It is of note that all patients described here with and hyperactivity [50–52]. However, no point mutations in RSVs at MC4R were male. The literature does not support BDNF have been reported to date in obese patients. In our the possibility that in obese subjects this gene is exclusively series, three patients (0.65%) harboured two different mis- affected in males. However, there may be a global bias in sense RSVs in this gene predicted as pathogenic, with none the parental transmission of RSVs. To determine this pos- detected in controls. Two unrelated cases carried the same sibility, a larger sample size is required, as well as addi- variant (I231V) and another carried a de novo variant tional investigation. It is also of note that there was a lack of (C141G9). Obese patients with BDNF variants presented significant differences in gene burden between the two mild to severe hyperphagia and insulin resistance with relatively close ethnic groups compared, European Spa- dyslipidaemia in two cases. No behavioural or learning niards and Hispanics in the USA. issues detected. Like MC4R, MC3R has a critical role in regulating Six RSVs in SIM1 were found in our cohort and two energy balance [46]. It was recently reported to regulate the were predicted as likely pathogenic. A third case was the upper and lower limits of an individual’s homeostatic set- affected brother of case 1782, concordant for the Q206K points and the response to external metabolic challenges RSV. The phenotype of all three cases included a BMI-SDS [47]. MC3R-knockout mice were shown to have dysregu- of +4 to +5, no intellectual or behavioural abnormalities, lation of energy expenditure, feeding responses and neu- and insulin resistance with impaired glucose tolerance. Sim1 roendocrine responses to metabolic challenges such as haploinsufficiency in mice induces hyperphagia, obesity fasting, high fat diet intake, pregnancy and the loss of and central nervous system developmental abnormalities estrogens. For example, these mice lost more weight during [53]. Deletions [54, 55] and translocations [56] affecting fasting than wild-type mice, but gained more weight when SIM1 cause severe obesity in association with intense put on a high fat diet [47]. The three patients (0.65%) with orexigenic impulsivity and other features resembling Pra- missense mutations in MC3R had a similar degree of obe- der–Willi syndrome. Loss-of-function point mutations in sity, which was milder than those harbouring RSVs in SIM1 with variable expressivity and incomplete penetrance MC4R, and no overgrowth or advanced skeletal matura- also produce a “Prader–Willi-like” phenotype [57], but also tion. These patients did not exhibit measurable hyperphagia, non-syndromic obesity [58]. Thus the described RSVs similar to what was observed in MC3R-deficient mice [47]. could contribute to the postulated increased intra-family risk However, as with mice, an inadequate diet could contribute for non-syndromic obesity [58]. to the excess weight gain. Only one of our patients had Significant differences in RSV load between patients and hyperinsulinism and hyperuricaemia; this patient also har- controls were detected in NEGR1 (3 patients, 0 controls), boured a variant in the other allele that did not fulfil the FTO (4 patients, 1 control) and PPARG (3 patients, 0 defined criteria as its frequency is 1/945 in the European controls), with five of the RSVs predicted as likely patho- population. However, if this second variant is functionally genic. Although association studies have documented their relevant, inheritance would be compatible with a recessive role in obesity, no highly penetrant RSVs have been or codominant model as reported for MC4R. In the codo- described to date in these genes. Our data indicate that some minant mode of inheritance, both monoallelic and biallelic RSVs at NEGR1, FTO and PPARG may behave as highly mutations can cause the disorder and patients with biallelic penetrant alleles causative of EOO, although the number of mutations are more affected than their heterozygous rela- patients is too small to define distinctive phenotypes. The tives [48]. patient with a likely pathogenic RSV at NEGR1 had speech BDNF is a pro-survival factor in the brain that also and behavioural problems and hyperuricaemia, while the participates in appetite regulation. Heterozygous BDNF- three patients harbouring RSVs in PPARG (including one knockout mice are obese, hyperphagic and hyperactive [49] sibling) had insulin resistance and associated dyslipidaemia, and disruption and deletion of BDNF have been described in agreement with the role of PPARG variants in predis- in patients with obesity. WAGR syndrome is a contiguous position to metabolic syndrome [59]. gene deletion syndrome on chromosome 11p13 character- Weaker evidence for definite implication was obtained ized by Wilms’ tumour, Aniridia, Genitourinary anomalies for GHSR, NTRK2 and TMEM18. Likely pathogenic RSVs and Mental retardation, with obesity present in patients with at NTRK2 were identified in one of our patients and two 838 C. Serra-Juhé et al. from the VLF study, but also in one control. Single patients In summary, we documented a higher burden of with RSVs at GHSR and TMEM18, absent in controls, were likely pathogenic heterozygous RSVs in several candidate found in the VLF cohort. Whether these monoallelic RSVs genes in patients with severe EOO compared to controls. contribute to the obesity phenotype awaits studies in larger The yield for RSV detection is relatively high, around 5% series and functional studies. De novo NTRK2 mutations, a of patients with EOO. Identification of a potential genetic highly intolerant gene to both loss-of-function (pLI = 1) cause of the phenotype in each case is very important for and missense variants (miss-z = 4.35), have been previously individualized follow-up and genetic counselling to the associated with severe obesity and developmental delay family. In addition, definition of the specific defect at the [17]. Thus, some heterozygous partially functional variants molecular level may be instrumental, as this group of could cause a phenotype of non-syndromic obesity without individuals may be relatively refractory to weight loss significant developmental delay. through diet and exercise and novel therapeutic approaches RSVs in five genes were either absent (LEP) or more for genetically defined obesities are becoming available. For common in controls than in cases (ADRB3, LEPR, PCSK1 instance, highly potent second-generation MC4R agonists and PCSK2) in both datasets. Biallelic mutations in LEP, lead to weight loss in individuals with MC4R deficiency LEPR and PCSK1 with a recessive model of inheritance [40]. Thus, several patients in this study could possibly have been described in severely obese patients. In a study of benefit from this therapy. Our results reinforce the role of 300 obese subjects, biallelic LEP mutations were identified the leptin–melanocortin pathway in obesity and bring to with a frequency of 3% [48], but 3/4th of the patients with light other genes that may carry highly penetrant obeso- biallelic mutations were homozygous, with mutations in genic single allele variants, such as FTO, NEGR1 and identical-by-descent regions in consanguineous families PPARG. The availability of an evidence-based algorithm for [48]. Thus, the prevalence of biallelic mutations in recessive diagnostic analysis of early-onset obesity would be of great genes in populations that are not inbred, such as the two clinical interest. However, the obese phenotype of the cohorts studied here, might be quite low [60]. POMC was children reported here, as well as in other studies, is non- not included in this study based on the assumed adrenal specific, and an algorithm that could accurately suggest insufficiency associated to pathogenic variants in this gene. which genes should be studied in each case is difficult even However, POMC is now being analysed in this cohort in if phenotype–genotype relationships lead to a better defi- another study. nition of the different clinical conditions. Instead, given the It is clear that functional analyses at the cellular, tissue currently available and affordable tools for diagnostic and organism levels are required to define the physio- genomic analysis, a genotype-first approach using expanded pathogenic mechanisms of all variants reported here. panel of genes with effective copy number analyses should However, we show strong epidemiological and statistical be indicated in these children with early-onset obesity. evidence by defining the genes harbouring heterozygous deleterious changes in our large cohort of EOO patients and Acknowledgements JA was funded by the Spanish Ministry of Health (FIS-PI13/02195 and PI16/00485, co-funded by FEDER), the Fun- then validating the results in a replication dataset. More- dación de Endocrinología y Nutrición and the “Centro de Investiga- over, there is experimental evidence for the abnormal ción Biomédica en Red” for obesity and nutrition (CIBEROBN) of the function of some of the reported RSVs [13, 40, 61]. Many Instituto de Salud Carlos III, Spain. LAP-J was funded by the Spanish genes involved in appetite and metabolic control appear to Ministry of Health (FIS-PI1302481, co-funded by FEDER), the Generalitat de Catalunya (2014SRG1468), the Institució Catalana de be sensible to dosage [50, 62, 63], with rescue of hap- Recerca i Estudis Avançats (ICREA Academia programme), the loinsufficiency of Mc4r and Sim1 recently shown to revert Spanish Ministry of Economy and Competiveness “Programa de the obese phenotype in mice [64]. Thus, although the Excelencia María de Maeztu” (MDM-2014-0370) and the Centro de molecular mechanisms underlying how these newly Investigación Biomédica en Red for rare diseases (CIBERER) of the Instituto de Salud Carlos III, Spain. JAC was funded by grants from described RSVs affect body weight remain to be estab- the Spanish Ministry of Science and Innovation (BFU2017-82565- lished, there is precedence for heterozygous affectation of C21-R2). We would like to thank Francisca Díaz and Sandra Canelles these genes being functionally important. Moreover, the for their excellent technical assistance. overall prevalence of monoallelic RSVs found in our cohort is similar to that of a recent publication [65] where a Compliance with ethical standards diagnosis rate of genetic obesity was reported in 3.9% of patients (48/1230) and 7.3% of the paediatric subgroup (12/ Conflict of interest LAPJ is a partner and scientific advisor of qGe- nomics Laboratory. The remaining authors declare that they have no 164). It should also be emphasized that the interaction of conflict of interest. these genetic variants with the environment is of utmost importance with morbid obesity being more likely to occur Publisher’s note: Springer Nature remains neutral with regard to in these cases when dietary habits are inadequate. jurisdictional claims in published maps and institutional affiliations. Heterozygous rare genetic variants in non-syndromic early-onset obesity 839 Open Access This article is licensed under a Creative Commons 15. Vaisse C, Clement K, Guy-Grand B, Froguel P. A frameshift Attribution 4.0 International License, which permits use, sharing, mutation in human MC4R is associated with a dominant form of adaptation, distribution and reproduction in any medium or format, as obesity. Nat Genet. 1998;20:113–4. long as you give appropriate credit to the original author(s) and the 16. Yeo GSH, Farooqi IS, Aminian S, Halsall DJ, Stanhope RG, source, provide a link to the Creative Commons license, and indicate if O’Rahilly S. A frameshift mutation in MC4R associated with changes were made. The images or other third party material in this dominantly inherited human obesity. Nat Genet. 1998;20:111–2. article are included in the article’s Creative Commons license, unless 17. Yeo GS, Connie Hung CC, Rochford J, Keogh J, Gray J, Sivar- indicated otherwise in a credit line to the material. If material is not amakrishnan S, et al. A de novo mutation affecting human TrkB included in the article’s Creative Commons license and your intended associated with severe obesity and developmental delay. Nat use is not permitted by statutory regulation or exceeds the permitted Neurosci. 2004;7:1187–9. use, you will need to obtain permission directly from the copyright 18. Willer CJ, Speliotes EK, Loos RJF, Li S, Lindgren CM, Heid IM, holder. To view a copy of this license, visit http://creativecommons. et al. Six new loci associated with body mass index highlight a org/licenses/by/4.0/. neuronal influence on body weight regulation. Nat Genet. 2009;41:25–34. 19. Herbert A, Gerry NP, McQueen MB, Heid IM, Pfeufer A, Illig T, References et al. A common genetic variant is associated with adult and childhood obesity. Science. 2006;312:279–83. 1. Friedman JM. Obesity in the new millennium. Nature. 20. Turcot V, Lu Y, Highland HM, Schurmann C, Justice AE, Fine 2000;40:632–4. RS, et al. Protein-altering variants associated with body mass 2. Walley AJ, Asher JE, Froguel P. The genetic contribution to non- index implicate pathways that control energy intake and expen- syndromic human obesity. Nat Rev Genet. 2009;10:431–42. diture in obesity. Nat Genet. 2018;50:26–41. 3. Llewellyn CH, Trzaskowski M, Plomin R, Wardle J. Finding 21. Meyre D, Delplanque J, Chèvre JC, Lecoeur C, Lobbens S, the missing heritability in pediatric obesity: the contribution Gallina S, et al. Genome-wide association study for early-onset of genome-wide complex trait analysis. Int J Obes. 2013;37: and morbid adult obesity identifies three new risk loci in European 1506–9. populations. Nat Genet. 2009;41:157–9. 4. Hatada I, Ohashi H, Fukushima Y, Kaneko Y, Inoue M, Komoto 22. Thorleifsson G, Walters GB, Gudbjartsson DF, Steinthorsdottir V, Y, et al. An imprinted gene p57KIP2 is mutated in Beckwith- Sulem P, Helgadottir A, et al. Genome-wide association yields Wiedemann syndrome. Nat Genet. 1996;14:171–3. new sequence variants at seven loci that associate with measures 5. Martos-Moreno GÁ, Serra-Juhé C, Pérez-Jurado LA, Argente J. of obesity. Nat Genet. 2009;41:18–24. Underdiagnosed Beckwith–Wiedemann syndrome among early 23. Glessner JT, Wang K, Cai G, Korvatska O, Kim CE, Wood S, onset obese children. Arch Dis Child. 2014;99:965–7. et al. Autism genome-wide copy number variation reveals ubi- 6. Nicholls RD, Knoll JHM, Butler MG, Karam S, Lalande M. quitin and neuronal genes. Nature. 2009;459:569–73. Genetic imprinting suggested by maternal heterodisomy in non- 24. Stefansson H, Rujescu D, Cichon S, Pietiläinen OP, Ingason A, deletion Prader–Willi syndrome. Lett Nat. 1989;342:281–5. Steinberg S, et al. Large recurrent microdeletions associated with 7. Buiting K, Gross S, Lich C, Gillessen-Kaesbach G, El-Maarri O, schizophrenia. Nature. 2008;455:232–6. Horsthemke B. Epimutations in Prader–Willi and Angelman 25. Jarick I, Vogel CIG, Scherag S, Schäfer H, Hebebrand J, Hinney syndromes: a molecular study of 136 patients with an imprinting A, et al. Novel common copy number variation for early onset defect. Am J Hum Genet. 2003;72:571–7. extreme obesity on chromosome 11q11 identified by a genome- 8. Sanz-Fernández M, Muñoz-Calvo MT, Pozo-Román J, Martos- wide analysis. Hum Mol Genet. 2011;20:840–52. Moreno GÁ, Argente J. Clinical and radiological findings in a 26. Serra-Juhé C, Martos-Moreno GÁ, Bou de Pieri F, Flores R, case of pseudohypoparathyroidism type 1a: Albright hereditary González JR, Rodríguez-Santiago B, et al. Novel genes involved osteodystrophy. An Pediatr. 2015;82:439–41. in severe early-onset obesity revealed by rare copy number and 9. Katsanis N, Ansley SJ, Badano JL, Eichers ER, Lewis RA, sequence variants. PLoS Genet. 2017;13:1–19. Hoskins BE, et al. Triallelic inheritance in Bardet–Biedl Syn- 27. Wheeler E, Huang N, Bochukova E, Keogh JM, Lindsay S, Garg drome, a Mendelian recessive disorder. Science. 2001;293: S, et al. Genome-wide SNP and CNV analysis identifies common 2256–9. and low-frequency variants associated with severe early-onset 10. Han JC, Liu QR, Jones M, Levinn RL, Menzie CM, Jefferson- obesity. Nat Genet. 2013;45:513–7. George KS, et al. Brain-derived neurotrophic factor and obesity in 28. Bochukova EG, Huang N, Keogh J, Henning E, Purmann C. the WAGR syndrome. N Engl J Med. 2008;359:918–27. Large, rare chromosomal deletions associated with severe early- 11. Ranadive S, Vaisse C. Lessons from extreme human obesity: onset obesity. Nature. 2010;463:666–70. monogenic disorders. Endocrinol Metab Clin North Am. 29. Grandone A, Di Sessa A, Umano GR, Toraldo R, Miraglia Del 2008;37:733–51. Giudice E. New treatment modalities for obesity. Best Pract Res 12. Farooqi IS, Matarese G, Lord GM, Keogh JM, Lawrence E, Agwu Clin Endocrinol Metab. 2018;32:535–49. C, et al. Beneficial effects of leptin on obesity, T cell hypor- 30. Mardis ER. A decade’s perspective on DNA sequencing tech- esponsiveness, and neuroendocrine/metabolic dysfunction of nology. Nature. 2011;470:198–203. human congenital leptin deficiency. J Clin Invest. 2002;110:1093– 31. Zhang J, Chiodini R, Badr A, Zhang G. The impact of next- 103. generation sequencing on genomics. J Genet Genom. 2011;38:95– 13. Granell S, Serra-Juhé C, Martos-Moreno GÁ, Díaz F, Pérez- 109. Jurado LA, Baldini G, et al. A novel melanocortin-4 receptor 32. Zuzarte PC, Denroche RE, Fehringer G, Katzov-Eckert H, Hung mutation MC4R-P272L associated with severe obesity has RJ, McPherson JD. A two-dimensional pooling strategy for rare increased propensity to be ubiquitinated in the ER in the face of variant detection on next-generation sequencing platforms. PLoS correct folding. PLoS ONE. 2012;7:e50894. ONE. 2014;9:e93455. 14. Montague CT, Farooqi IS, Wareham NJ, Soos MA, Rau H, 33. Hernández M, Castellet J, Narvaiza JL, Rincón JM, Ruiz I, Sán- Wareham NJ, et al. Congenital leptin deficiency is associated chez E, et al. Curvas y tablas de crecimiento. 2nd edn. Madrid: with severe early-onset obesity in humans. Nature. 1997;387: Instituto de Investigación sobre Crecimiento y Desarrollo, Fun- 903–8. dación Faustino Orbegozo, Editorial Garsi; 1988. 840 C. Serra-Juhé et al. 34. Ristow M, Müller-Wieland D, Pfeiffer A, Krone W, Kahn R. 50. Gray J, Yeo GSH, Cox JJ, Morton J, Adlam AL, Keogh JM, et al. Obesity associated with a mutation in a genetic regulator of adi- Hyperphagia, severe obesity, impaired cognitive function, and pocyte differentiation. N Engl J Med. 2015;339:953–9. hyperactivity associated with functional loss of one copy of the 35. Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, brain-derived neurotrophic factor (BDNF) gene. Diabetes. Sougnez C, et al. Sensitive detection of somatic point mutations in 2006;55:3366–71. impure and heterogeneous cancer samples. Nat Biotechnol. 51. Jouan L, Gauthier J, Dion PA, Rouleau GA. Rare variants in 2013;31:213–9. complex traits: novel identification strategies and the role of de 36. Robinson JT, Thorvaldsdóttir H, Winckler W, Guttman M, Lander novo mutations. Hum Hered. 2013;74:215–25. ES, Getz G, et al. Integrative Genomics Viewer. Nat Biotechnol. 52. Harcourt BE, Bullen DVR, Kao KT, Tassoni D, Alexander EJ, 2011;29:24–26. Burgess T, et al. Maternal inheritance of BDNF deletion, with 37. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, phenotype of obesity and developmental delay in mother and Fennell T, et al. Analysis of protein-coding genetic variation in child. Am J Med Genet A. 2018;176:194–200. 60,706 humans. Nature. 2016;536:285–91. 53. Michaud JL, Boucher F, Melnyk A, Gauthier F, Goshu E, Lévy E, 38. González-Pérez A, López-Bigas N. Improving the assessment of et al. Sim1 haploinsufficiency causes hyperphagia, obesity and the outcome of nonsynonymous SNVs with a consensus deleter- reduction of the paraventricular nucleus of the hypothalamus. iousness score, Condel. Am J Hum Genet. 2011;88:440–9. Hum Mol Genet. 2001;10:1465–73. 39. Kircher M, Witten DM, Jain P, O’Roak BJ, Cooper GM, Shen- 54. Villa A, Urioste M, Bofarull J, Martínez-Frías M. De novo dure J. A general framework for estimating the relative patho- interstitial deletionq16.2q21 on chromosome 6. Am J Med Genet. genicity of human genetic variants. Nat Genet. 2014;46:310–5. 1995;55:379–83. 40. Collet TH, Dubern B, Mokrosinski J, Connors H, Keogh JM, 55. Faivre L, Cormier-Daire V, Lapierre JM, Colleaux L, Jacquemont Mendes de Oliveira E, et al. Evaluation of a melanocortin-4 S, Geneviéve D, et al. Deletion of the SIM1 gene (6q16.2) in a receptor (MC4R) agonist (Setmelanotide) in MC4R deficiency. patient with a Prader–Willi-like phenotype. J Med Genet. Mol Metab. 2017;6:1321–9. 2002;39:594–6. 41. Butte NF, Cai G, Cole SA, Comuzzie AG. Viva la Familia Study: 56. Holder JL, Butte NF, Zinn AR. Profound obesity associated with a genetic and environmental contributions to childhood obesity and balanced translocation that disrupts the SIM1 gene. Hum Mol its comorbidities in the Hispanic population. Am J Clin Nutr. Genet. 2000;9:101–8. 2006;84:646–54. 57. Bonnefond A, Raimondo A, Stutzmann F, Ghoussaini M, 42. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Ramachandrappa S, Bersten DC, et al. Loss-of-function mutations Kernytsky A, et al. The Genome Analysis Toolkit: A MapReduce in SIM1 contribute to obesity and Prader-Willi – like features. J framework for analyzing next-generation DNA sequencing data. Clin Invest. 2013;123:3037–41. Genome Res. 2009;20:254–60. 58. Ramachandrappa S, Raimondo A, Cali AMG, Keogh JM, Hen- 43. Govaerts C, Srinivasan S, Shapiro A, Zhang S, Picard F, Clement ning E, Saeed S, et al. Rare variants in single-minded 1 (SIM1) are K, et al. Obesity-associated mutations in the melanocortin 4 associated with severe obesity. J Clin Invest. 2013;123:1–9. receptor provide novel insights into its function. Peptides. 59. Rocha RM, Barra GB, ÉCCC Rosa, Garcia ÉC, Amato AA, 2005;26:1909–19. Azevedo MF. Prevalence of the rs1801282 single nucleotide 44. Martinelli CE, Keogh JM, Greenfield JR, Henning E, van der polymorphism of the PPARG gene in patients with metabolic Klaauw AA, Blackwood A, et al. Obesity due to melanocortin 4 syndrome. Arch Endocrinol Metab. 2015;59:297–302. receptor (MC4R) deficiency is associated with increased linear 60. Paolini B, Maltese PE, Del Ciondolo I, Tavian D, Missaglia S, growth and final height, fasting hyperinsulinemia, and incomple- Ciuoli C, et al. Prevalence of mutations in LEP, LEPR, and MC4R tely suppressed growth hormone secretion. J Clin Endocrinol genes in individuals with severe obesity. Genet Mol Res. Metab. 2011;96:181–8. 2016;19:15. https://doi.org/10.4238/gmr.15038718. 45. Chhajlani V. Distribution of cDNA for melanocortin receptor 61. Moore BS, Mirshahi T. Genetic variants help define the role of the subtypes in human tissues. Biochem Mol Biol Int. 1996;38:73–80. MC4R C-terminus in signaling and cell surface stability. Sci Rep. 46. Zegers D, Beckers S, Hendrickx R, Van Camp JK, Van Hoor- 2018;8:10397. enbeeck K, Desager KN, et al. Prevalence of rare MC3R variants 62. Lee YS, Poh LK, Kek BL, Loke KY. The role of melanocortin 3 in obese cases and lean controls. Endocrine. 2013;44:386–90. receptor gene in childhood obesity. Diabetes. 2007;56:2622–30. 47. Ghamari-Langroudi M, Cakir I, Lippert RN, Sweeney P, Litt MJ, 63. Drabkin M, Birk OS, Birk R. Heterozygous versus homozygous Ellacott KLJ, et al. Regulation of energy rheostasis by the phenotype caused by the same MC4R mutation: novel mutation melanocortin-3 receptor. Sci Adv. 2018;4. https://doi.org/10.1126/ affecting a large consanguineous kindred. BMC Med Genet. sciadv.aat0866. 2018;19:135. 48. Farooqi IS, Wangensteen T, Collins S, Kimber W, Matarese G, 64. Matharu N, Rattanasopha S, Tamura S, Maliskova L, Wang Y, Keogh JM, et al. Clinical and molecular genetic spectrum of Bernard A, et al. CRISPR-mediated activation of a promoter or congenital deficiency of the leptin receptor. N Engl J Med. enhancer rescues obesity caused by haploinsufficiency. Science. 2007;356:237–47. 2018. pii: eaau0629. 49. Fox EA, Biddinger JE, Jones KR, McAdams J, Worman A. 65. Kleinendorst L, Massink MPG, Cooiman MI, Savas M, van der Mechanism of hyperphagia contributing to obesity in brain- Baan-Slootweg OH, Roelants RJ, et al. Genetic obesity: next- derived neurotrophic factor knockout mice. Neuroscience. generation sequencing results of 1230 patients with obesity. J Med 2013;229:176–99. Genet. 2018;55:578–86. Heterozygous rare genetic variants in non-syndromic early-onset obesity 841 Affiliations Clara Serra-Juhé1,2 Gabriel Á. Martos-Moreno3,4 Francesc Bou de Pieri1,2 Raquel Flores1,2 Julie A. Chowen ● ● ● ● 3,4 ● Luis A. Pérez-Jurado 1,2,5 Jesús Argente 3,4,6 ● 1 Genetics Unit, Universitat Pompeu Fabra, Hospital del Mar 4 Research Institute (IMIM), C/Doctor Aiguader, 8, 08003 CIBER de Fisiopatología de la Obesidad y Nutriciόn Barcelona, Spain (CIBEROBN), Instituto de Salud Carlos III, C/Sinesio Delgado, 4, 28029 Madrid, Spain 2 Centro de Investigación Biomédica en Red de Enfermedades Raras 5 (CIBERER), Instituto de Salud Carlos III, C/Sinesio Delgado, 4, Women’s and Children’s Hospital, South Australia Medical and 28029 Madrid, Spain Health Research Institute (SAMHRI) and University of Adelaide, 72 King William Road, North Adelaide, SA 5006, Australia 3 Hospital Infantil Universitario Niño Jesús, Department of Endocrinology, Instituto de Investigación La Princesa, 6 IMDEA Food Institute, CEIUAM + CSI, Crta. de Cantoblanco, 8, Universidad Autónoma de Madrid, Department of Pediatrics, 28049 Madrid, Spain Avenida Menéndez Pelayo, 65, 28009 Madrid, Spain