Targeted Next‐Generation Sequencing of the Leptin‐Melanocortin Pathway in Severe Obesity

ABSTRACT

Objective

Pathogenic variants in five established leptin‐melanocortin pathway genes (LEP, LEPR, MC4R, PCSK1, POMC) are associated with severe early‐onset obesity and are targets for emerging treatments. However, these variants are rare in these patients, suggesting the involvement of additional genes interacting with this pathway.

Methods

Next‐generation sequencing (NGS) analysis was performed in 395 patients with severe obesity, including 213 children (mean BMI: 56.3 kg/m²; BMI‐z‐score: 4.6). The analysis targeted 20 genes, including the 5 established genes. Rare genetic variants were assessed for pathogenicity using prediction algorithms, genetic databases, and literature review. Phenotypic data were retrospectively collected, focusing on obesity severity, age of onset, familial history, eating behavior disorder, neurodevelopmental and endocrine‐associated diseases, and obesity complications.

Results

Pathogenic heterozygous variants were identified in 34 patients (8.6%), 18 of them harboring pathogenic variants in the 15 additional genes. In adults, early‐onset obesity was more frequent in potentially pathogenic variants carriers than in non‐carriers (83.3% vs. 55.0%, p = 0.04). No differences were observed in the other phenotypic characteristics.

Conclusions

This supports the relevance of expanded genetic testing in severe obesity. Early‐onset obesity remains a key clinical feature to guide genetic investigation and identify patients who may benefit from early personalized care and targeted treatments.

Study Importance.

• What is already known?

  • ○
    Pathogenic biallelic variants in LEP, LEPR, MC4R, PCSK1 or POMC are associated with severe obesity, indicating innovative targeted treatments.
• What does this study add?

  • ○
    Half of the pathogenic heterozygous variants found occur in additional genes that were not previously analyzed.
  • ○
    Phenotypes are similar between heterozygous carriers of pathogenic variants and non‐carriers in a cohort of patients with severe obesity.
• How might your results change the direction of research or the focus of clinical practice?

  • ○
    Improving phenotypic characterization of obesity may help to guide genetic investigations.
  • ○
    Assessing the clinical impact of heterozygous pathogenic variants for each gene will help therapeutic decisions.

1. Introduction

Obesity is a major global public health issue, with its prevalence rising sharply, particularly among children. In 2022, the World Health Organization (WHO) estimated that 160 million children and adolescents aged 5–19 years were living with obesity [1]. A recent forecasting study based on trends in the prevalence of obesity in children and adolescents between 1990 and 2021, which has tripled, shows that this substantial increase is set to accelerate further by 2050 [2].

Defined as an excess of body fat mass, obesity results from a positive energy balance driven by multiple interacting factors, including genetic predisposition and environmental influences [3].

The genetic contribution is particularly important in individuals with severe early‐onset obesity, with heritability estimates reaching 80% [4, 5]. Among the key genetic determinants, defects in the leptin‐melanocortin pathway, which play a fundamental role in the central regulation of energy balance, are strongly associated with the development of a severe early‐onset phenotype, typically manifesting before the age of 6 [6]. Biallelic variants, including homozygous and composite heterozygous variants, in five established genes (LEP, LEPR, MC4R, POMC and PCSK1) are linked to rapid weight gain, dysregulated hunger and satiety signals, and in some cases with endocrine and neurodevelopmental disorders [6, 7].

Heterozygous pathogenic variants, whether rare or common, are more prevalent and account for approximately 10% of patients with severe early‐onset obesity. However, their phenotypic impact remains partially characterized. The available data are limited and contradictory regarding the severity and clinical consequences of these variants. Depending on studied populations, they are associated with an attenuated phenotype or with no major impact [8, 9, 10, 11].

Beyond these five genes, several others such as KSR2, MRAP2, or NTRK2 were identified in case series of severe obesity in humans, and their biochemical relationship with the leptin–melanocortin pathway was further deciphered [12, 13, 14]. Their precise contribution in large cohorts of patients with severe early‐onset obesity and their associated phenotypic characteristics remain nonetheless poorly characterized.

Genetic diagnosis is crucial for these patients, as it provides prognostic insights, enables early detection of specific complications, and facilitates access to specialized care in expert centers. Furthermore, it enables access to interventions, including personalized targeted pharmacological treatments and clinical trials aimed at optimizing patient management. One notable example is setmelanotide, a MC4R agonist which restores melanocortin pathway function by reducing hunger with improved satiety and significant weight loss in patients with specific genetic defects [15].

The main objective of this study was to assess the diagnostic performance of next‐generation sequencing (NGS) targeting 20 genes involved in the regulation of hunger and satiety in the hypothalamus in a cohort of patients with severe obesity. The second objective was to identify specific clinical criteria to improve the diagnostic strategy.

2. Methods

2.1. Study Design and Population

We retrospectively included all patients with available medical records who had DNA analysis by NGS for the indication of severe and early‐onset obesity between 2018 and 2023 at the Functional Unit for the Genetics of Obesity and Dyslipidemia, Pitié‐Salpêtrière Hospital, Paris, France. All patients signed an informed consent form for the use of their retrospective clinical data and genetic results for research purposes (Comité de protection des personnes Ouest V‐number 2019‐A03201).

2.2. Clinical Data Collection

Phenotypic data were retrospectively retrieved from medical records and included the history of obesity and growth. We also collected as follows: family history of obesity, history of bariatric surgery for adults, eating disorders, central endocrine disorders, short stature, pubertal development disorders, cryptorchidism, neurodevelopmental disorders, and other relevant medical history and obesity‐related comorbidities (arterial hypertension [HT], type 2 diabetes [T2DM], obstructive sleep apnea [OSA], hepatic steatosis, insulin resistance). The precise definition of each phenotypic characteristic collected is provided in online Supporting Information Methods 1.

2.3. Genetic Analysis

Genetic analyses were conducted as part of routine medical care for patients with severe and/or early‐onset obesity (BMI > 35 kg/m² in adults, BMI > IOTF 35 in children; IOTF: International Obesity Task Force; the IOTF BMI curves extrapolate the BMI thresholds for obesity: 30, 35, 40 kg/m² to children) and associated features such as eating disorders, endocrine deficiencies, or neurodevelopmental disorders.

Targeted NGS was conducted as described by Gatta‐Cherifi et al. [16]. A panel of 20 genes was designed based on genes known to be involved in severe early‐onset obesity with eating disorders. In the following, we refer to the five genes LEP, LEPR, MC4R, POMC and PCSK1 as “the five established genes” as they are the best documented in terms of physiological involvement in the leptin‐melanocortin pathway and pathological impact in the event of mutation [6]. On the other hand, the 15 remaining genes ADCY3, ALMS1, ARNT2, BDNF, CEP19, KSR2, MAGEL2, MC3R, MRAP2, MYT1L, NCOA1, NTRK2, SH2B1, SIM1 and TUB are known to interact with the leptin‐melanocortin pathway or hypothalamus development or to cause specific syndromes with a high penetrance of associated obesity (see Table S1 for more details).

2.4. Variant Classification

To distinguish rare (minor allele frequency < 1%) pathogenic variants from benign ones, all detected variants were evaluated according to international experts' recommendations [17], using public genetic databases, scientific literature, and in silico prediction tools. The tools and databases and the associated threshold used are detailed in online Supporting Information Methods 2. The details on the data collected to classify each variant are available in Table S2.

Finally all variants identified in this cohort were classified into three categories (Figure 1):

1. “Benign” for variants with a low functional impact based on published data (probably benign variants) or predicted to be benign by in silico tools (predicted benign variants).

2. “Potentially pathogenic” for variants with a probable moderate or high functional impact (probably pathogenic variant) or predicted to be pathogenic by in silico tools (predicted pathogenic variants).

3. “Variants of unknown significance” (VUS), which included all variants that did not fit into the two previous categories.

FIGURE 1.

Classification scheme for variants based on available data. The functional impact of each rare variant (minor allele frequency < 1%) was systematically analyzed following this process. [Color figure can be viewed at wileyonlinelibrary.com]

2.5. Statistical Analysis

Quantitative variables were expressed as mean ± standard deviation, and qualitative variables as percentages. Variables were compared using the Wilcoxon test or t‐test for quantitative variables and the Fisher test or chi‐square test for qualitative variables, when appropriate. BMI comparisons were performed in relation to the maximum lifetime BMI for adult patients to include data from patients who had undergone bariatric surgery. The null hypothesis tested was that there was no difference between the group of patients carrying potentially pathogenic variants and non‐carriers and was rejected if p < 0.05. RStudio software (version 2023.03.0 + 386 using R version 4.2.3) was used for statistical analysis. The code used to perform the statistical analyses is provided in online Supporting Information Methods 3.

3. Results

3.1. Description of the Cohort of Patients With Severe Obesity

Between 2018 and 2023, a total of 395 patients including 213 children (< 18 years), underwent analysis of the 20 genes for severe obesity (Table 1). All individuals had severe obesity with a mean adult BMI of 56.3 ± 11.7 kg/m² and a BMI z‐score of 4.6 ± 1.8 in children. Early‐onset obesity (before the age of 6) was observed in most children and in more than half of the adults, although data are missing for 36 cases (19.7% of adults). Among the 61 patients who underwent bariatric surgery, 27 had gastric bypasses, 27 sleeve gastrectomy (including 3 patients under 18 years), and 7 adjustable gastric bands.

TABLE 1.

Phenotype of included patients.

Total (n = 395)	Adults (n = 182)	Children (n = 213)
Female (n)	111 (61.0%)	102 (47.9%)
Age (years)	31.7 ± 11.2	10.8 ± 4.7
Maximal BMI (kg/m2)	56.3 ± 11.7
BMI (z‐score)		4.6 ± 1.8
≥ 1 family member with obesity	118/158 a (74.7%)	148/207 (71.5%)
Obesity before 6 years	86/147 a (58.5%)	192/206 (88.3%)
Bariatric surgery	58 (31.9%)	3 (1.4%)
Eating disorder	137/171 a (80.1%)	134/203 a (66.0%)
Endocrine comorbidity	31 (17.0%)	10 (4.7%)
Central hypogonadism	17 (9.3%)	6 (2.8%)
Neurodevelopmental disorder	36 (19.8%)	63 (29.6%)
Malformations	16 (8.8%)	25 (11.7%)
Obesity complications (≥ 1) b	146 (80.2%)	92 (43.2%)
Insulin resistance	43/53 (81.1%)	107/164 (65.2%)
Type 2 diabetes	46 (25.2%)	12 (5.6%)
Hypertension	63 (34.6%)	7 (3.3%)
OSA	114/155 a (73.5%)	38/72 a (52.8%)
Hepatic steatosis	86/116 a (74.1%)	68/124 a (54.8%)

Abbreviation: OSA: obstructive sleep apnea.

^a Values are shown as fractions for missing data.

^b Among the following: type 2 diabetes, hypertension, OSA, or hepatic steatosis.

Abnormal feeding behaviors were common in the cohort, with hyperphagia and food impulsivity being the most frequently reported. Endocrine disorders were present in 17% of adults, especially central hypogonadism (9.3%). Four adults and three children had a history of cryptorchidism with surgical relocation of one or both testicles. Neurodevelopmental disorders were present in one‐quarter of the cohort; the most frequent were intellectual disability.

Congenital malformations, including renal, cardiac, extremity malformations, or dysmorphic syndrome, were present in about 10% of patients. Spinal static disorders, such as scoliosis, were observed in five adults and three children. Six adults and two children suffered from epilepsy. Retinitis pigmentosa and nystagmus were found in two adults and six children. Four adults suffered from central deafness. Suckling disorders in the neonatal period were found in one adult and three children. Severe neonatal hypoglycemia was reported for on patient.

Complications of obesity, including type 2 diabetes, hypertension, OSA, and hepatic steatosis, affected more than 80% of adults and were less frequent in children.

3.2. Genetic Analysis

A total of 346 occurrences of 247 rare variants were identified in the cohort (Figure 2). Of these 247 variants, the vast majority (232/247) were nucleotide variants, 12 were deletions or insertions of less than 20 base pairs, and 3 were larger deletions. Thirty‐six variants were reported for the first time, as they were not referenced in the dbSNP, GnomAD, or EVS (Exome Variant Server from the National Heart, Lung, and Blood Institute exome sequencing project) databases.

FIGURE 2.

Flowchart of genetic variants by type. Nucleotide deletions and insertions were qualified as small when they were < 20 base pairs.

The 247 identified variants were classified as follows: benign 182/247 (73.7%), potentially pathogenic 32/247 (13.4%), or of uncertain significance (VUS) 33/247 (13.0%). Among these 247 variants, 16 were considered as probably pathogenic and 25 as probably benign based on functional or cohort studies. The remaining classified variants were classified with less certainty as predicted benign or predicted pathogenic.

The distribution of variants within the population was heterogeneous: 178 carried no variants (45.1%), while among carriers, the average number of variants per patient was 1.59 ± 0.77, with a maximum of 4 variants per patient (n = 5).

Rare probably pathogenic variants were detected in 18/395 patients (4.6%) (Figure 3) and were found in the following genes: MC4R (n = 7; 1.8%), NTRK2 (n = 3), POMC (n = 3), SH2B1 (n = 3), LEPR (n = 2), PCSK1 (n = 1), and TUB (n = 1).

FIGURE 3.

Frequency of genetic variants in the cohort (n = 395). VUS, variant of undetermined significance. [Color figure can be viewed at wileyonlinelibrary.com]

When including predicted pathogenic variants found in 16 patients (4.1%), the presence of rare potentially pathogenic variants was observed in 34/395 patients (8.6%) (Figure 3). These variants were found in MC4R (n = 10; 2.5%), NTRK2 (n = 6; 1.5%), POMC (n = 3), SH2B1 (n = 3), ALMS1 (n = 2), ADCY3 (n = 2), KSR2 (n = 2), LEPR (n = 2), TUB (n = 2), MRAP2 (n = 1), MYTL1 (n = 1), PCSK1 (n = 1), and SIM1 (n = 1). Thus, 18 of the 34 patients carrying rare potentially pathogenic genetic variants did so in the 15 newly analyzed genes.

Thirty‐two patients (8.1%) carried VUS (Figure 3). These patients were classified as “VUS carriers” whether they carried or did not carry a benign variant. The patients carrying a potentially pathogenic variant and VUS or a benign variant were classified as “potentially pathogenic variants carriers.”

Among the remaining patients who did not carry any potentially pathogenic variants or VUS, 143/395 (36.2%) carried a benign variant (Figure 3). Of them, 24/394 = 6.1% carried probably benign variants, and 119/395 = 30.1% carried a predicted benign variant.

No patient in this cohort carried a biallelic variant. Two patients were compound heterozygous for two rare potentially pathogenic variants in different genes: one had a pathogenic complete deletion of one SH2B1 allele and a predicted pathogenic TUB c.1375C>T variant, and the other had the probably pathogenic NTRK2 c.611C>A variant [18] and the predicted pathogenic variant POMC c.116C>T.

The ALMS1 gene was the most frequently involved within this cohort with 74 different variants identified in 83 patients (21%). Of these, 75 carried benign variants, 6 carried VUS and 2 carried potentially pathogenic variants. Variants in at least one of the 5 established genes (LEP, LEPR, MC4R, PCSK1, and POMC) were detected in 51 patients (12.9%), including 16 with potentially pathogenic variants, and no VUS.

3.3. Phenotypes of Patients Carrying Potentially Pathogenic Variants

The detailed phenotypic characteristics of the 34 patients carrying rare potentially pathogenic variants are presented in Table 2.

TABLE 2.

Phenotypic description of the 34 patients carrying rare potentially pathogenic variants.

	Gene	Type of variant	DNA	Protein	Sex	Age	BMI a	Age obesity onset	EBD	NDD	Endocrine disorder	Bariatric surgery	Obesity complications	Other comorbidity	Familial history of obesity
1	ADCY3 c	Missense	c.422A>C	p.Tyr141Ser	M	14	38.5 (+3.8)	2	Y	N	N	N	HS, IR, OSA	Hereditary spherocytosis	N
2	ADCY3 c	Missense	c.2870C>T	p.Ser957Leu	F	42	54.1	< 12	N	N	N	N	IR, OSA	N	Mother & 2 S
3	ALMS1	Nonsense	c.7051G>T	p.Glu2351X	F	22	35.1	3	Y	Y	CD	N	HS, IR	Dysmorphia, associated DDX3X pathogenic variant	N
4	ALMS1	Large deletion	Exon 7	—	F	48	40.6	NA	Y	N	N	N	HS, OSA, T2D	Spontaneous retinal detachment	N
5	KSR2	Frameshift deletion	c.429_430 delGA	p.Thr144Argfs*22	F	20	51.2	3	Y	Y	N	N	IR, OSA	Brachymetacarpia, brachymetatarsia	Mother BS, 1 S
6	KSR2	Missense	c.2720G>A	p.Arg907His	F	20	53.1	2	N	N	N	N	IR	N	1 S
7	LEPR b	Missense	c.1835G>A	p.Arg612His	M	12	38.0 (+4.3)	5	Y	N	N	N	OSA	N	N
8	LEPR b	Large deletion	Exons 6–8	—	F	6	21.1	1	N	N	PP	N	N	N	Mother (BMI 43.6 before BS) and father BMI 42
9	MC4R b	Missense	c.380C>T	p.Ser127Leu	F	24	25.7 (45.9)	6	Y	N	N	GBP: −45% of initial weight and further +8% regain in 5 years	N	N	Mother
10	MC4R	Nonsense	c.466C>T	p.Gln156*	F	18	44.7	2	Y	N	N	SG at 18 years, no further data	HS, IR	Consanguinity	Mother (BMI 42 before BS), 1 S
11	MC4R b	Missense	c.494G>A	p.Arg165Gln	F	11	42 (+4.7)	1	N	N	N	N	HS, OSA	N	Father (BMI 35)
12	MC4R c	Missense	c.632 T>C	p.Leu211Pro	M	3	30.4 (+9.5)	1	Y	Y	N	N	N	N	Mother (BMI 38), father (BMI 32) and 2 S
13	MC4R b	Missense	c.731C>A	p.Ala244Glu	M	20	55.5	< 3	Y	N	N	N	N	N	Mother and father
14	MC4R b	Missense	c.757G>A	p.Val253Ile	M	16	60.3 (+6.7)	1	Y	N	N	N	HS, IR, OSA	N	Mother and father, mother BS
15	MC4R b	Missense	c.811 T>C	p.Cys271Arg	F	7	29.9 (+4.7)	2	N	N	N	N	N	N	Mother and 2 S
16	MC4R b	Missense	c.883 T>C	p.Ser295Pro	F	19	31.6	4	N	N	N	N	HS, IR	Hypothyroidism of unknown etiology	N
17	MC4R	Missense	c.892G>A	p.Asp298Asn	F	26	38.4 (48.9)	5	Y			SG: −20% of initial weight in 1 year	HS, IR	N	Father
18	MC4R b	Missense	c.914G>A	p.Arg305Gln	F	6	23.2 (+3.4)	2	Y	N	N	N	N	N	Mother and father
19	MRAP2 c	Missense	c.154G>C	p.Gly52Arg	M	27	40.4 (67.4)	8	N	N	N	SG: −27% of initial weight in 18 months	HS, HT, OSA	N	N
20	MYT1L	Nonsense	c.1776C>A	p.Cys592*	M	40	56.6	< 3	Y	Y	GD	N	HS, HT, IR, OSA	Neonatal hypotonia, dysmorphia, cryptorchidism	Mother (BMI 31), father (BMI 46)
21	NTRK2	Missense	c.420G>C	p.Leu140Phe	M	4	19.6 (+2.8)	2	N	N	N	N	N	N	N
22					M	17	54.1 (+5.9)	4	Y	N	N	N	IR, OSA	N	Mother BS
23	NTRK2 b	Missense	c.611C>A	p.Pro204His	F	17	40.4 (+3.6)	8	Y	N	N	N	IR, OSA	N	Mother and 1 S, mother BS
24					M	44	68.5	< 12	Y	N	N	N	HS, HT, OSA, T2D	N	Mother
25	NTRK2 b POMC	Missense Missense	c.611C>A c.116C>T	p.Pro204His p.Thr39Met	M	14	40.2 (+3.9)	5	N	Y	CD	N	HS	N	Father (BMI 55), carrying the same POMC mutation
26	NTRK2 c	Missense	c.2283G>T	p.Glu761Asp	F	29	48.4	2	Y	N	N	N	IR, OSA	N	N
27	PCSK1 b , c	Missense	c.160G>A	p.Gly54Ser	F	20	39.8 (55.9)	NA	NA	N	N	SG: −29% of initial weight in 1 year	N	N	NA
28	POMC b	Missense	c.706C>G	p.Arg236Gly	F	6	25.1 (+3.9)	4	N	N	N	N	N	N	Mother and father
29					M	25	22.4 (41.6)	< 6	N	N	N	SG: −46% of initial weight in 18 months	OSA	N	3 S (BMI < 35)
30	SH2B1 b	Missense	c.269C>A	p.Pro90His	M	21	52.6	2	N	N	N	N	HS, HT, OSA, T2D	N	Mother (BMI 35)
31	SH2B1 b	Missense	c.449A>G	p.Lys150Arg	F	18	46.6	1	Y	Y	N	N	HS, IR, OSA	Neonatal suction disorder, small for gestational age	1 S
32	SH2B1 b TUB	Complete deletion			M	46	58.0	< 6	Y	Y	N	N	OSA	N	Father (BMI 37)
		Missense	c.1375C>T	p.Arg459Cys
33	SIM1	Missense	c.596C>G	p.Ser199Cys	F	19	43.6	4	Y	Y	N	N	IR	Epilepsia	Mother (BMI 48), father (BMI 40) and 1 S
34	TUB	Missense	c.943C>A	p.Pro315Thr	M	5	24.5 (+5.2)	3	N	Y	N	N	IR	Brachymetacarpia	N

Abbreviations: BMI, body mass index (kg/m²); BS, bariatric surgery; CD, corticotropic deficiency; EBD, eating behavior disorder; F, female; GBP, gastric bypass; GD, gonadotropic deficiency; HS, hepatic steatosis; HT, hypertension; IR, insuline resistance; M, male; NDD, neurodevelopmental disorder; OSA, obstructive sleep apnea; PP, precocious puberty; S, sibling; SG, sleeve gastrectomy; T2D, type 2 diabetes.

^a Values in parentheses show maximum BMI in case of bariatric surgery or BMI z‐score in patients < 18 years.

^b Available data in the literature about the functional impact of the variant.

^c Variants reported for the first time.

Maximal BMI in adults did not differ (Figure 4) between carriers of rare potentially pathogenic variants (n = 20; 51.6 ± 10.3 kg/m²) and non‐carriers (n = 162; 54.1 ± 11.8 kg/m², p = 0.06). Similarly, the BMI z‐score in children was comparable between carriers (n = 14; 4.1 ± 1.8) and non‐carriers (n = 199; 4.3 ± 1.8, p = 0.90). There were no differences in other phenotypic characteristics between patients with and without potentially pathogenic variants (Table 3). However, early onset obesity (< 6 years) was more frequent in adults carrying rare potentially pathogenic variants (15/18; 83.3% of carriers with childhood BMI data) than in non‐carriers (71/129; 55.0%, p = 0.04). The repartition of rarer phenotypes (epilepsy, retinitis pigmentosa and nystagmus, central deafness, neonatal suckling disorders and severe neonatal hypoglycemia) between carriers and non‐carriers of potentially pathogenic variants is detailed in Table S3.

FIGURE 4.

Comparison of BMI according to variant carrying status. Maximum BMI was used in adults to include patients who underwent bariatric surgery. BMI‐z‐score was used for children < 18 years. Variable distributions were compared using the Wilcoxon test. NS, nonsignificant. [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 3.

Comparison of patients with and without pathogenic variants.

Total (n = 395)	Non‐carriers (n = 361)	Rare pathogenic variant (n = 34)	p
Female (n)	181 (50.1%)	19 (55.9%)
Children (n)	192 (53.2%)	14 (41.2%)
Age (years)	20.1 ± 13.4	20.2 ± 12.1
≥ 1 family member with obesity	222/316 (70.3%)	21/32 (65.6%)	0.81
Obesity before 6 years	233/303 a (76.9%)	28/32 (87.5%)	0.30
Eating disorder	237/322 a (73.6%)	20/33 (60.6)	0.16
Endocrine comorbidity	36 (10.0%)	3 (8.8%)	1
Central hypogonadism	21 (5.8%)	1	0.71
Neurodevelopmental disorder	85 (23.5%)	9 (26.5%)	1
Malformations	36 (10.0%)	4 (11.8%)	0.77
Obesity complications (≥ 1) b	202 (56.0%)	22 (64.7%)	0.71
Type 2 diabetes	51 (14.1%)	3 (8.8%)	0.45
Hypertension	60 (16.6%)	4 (11.8%)	0.48
OSA	126/189 (66.7%)	17/23 (73.9%)	0.64
Hepatic steatosis	131/204 (64.2%)	14/25 (56.0%)	0.38

Abbreviation: O: obstructive sleep apnea.

^a Values are shown as fractions for missing data.

^b Among the following: insulin resistance, type 2 diabetes, hypertension, OSA, or hepatic steatosis.

4. Discussion

This study evaluated the performance of NGS to identify potentially pathogenic variants in 20 genes of the leptin‐melanocortin pathway in a large cohort of subjects with severe obesity. The cohort was thoroughly phenotyped, and the functional impact of each variant was systematically assessed using available literature and in silico prediction tools. Potentially pathogenic heterozygous variants were identified in 8.6% of patients, with more than half of them in newly analyzed genes included in this expanded panel. These findings reinforce the strong genetic contribution to severe obesity and underline the clinical importance of broad genetic screening to identify patients who may benefit from targeted therapies. The 8.6% prevalence of potentially pathogenic variants in this cohort aligns with previous studies that estimate the frequency of rare variants in patients with severe and/or early‐onset obesity varying from 5% to 9.3% [19, 20, 21]. As expected, MC4R was the most frequently affected gene (1.8% of patients) in this cohort, consistent with prior studies reporting 1% to 6% of heterozygous variants in patients with severe or early‐onset obesity [19, 21, 22].

This high prevalence of potentially pathogenic variants supports the usefulness of NGS panels in genetic diagnosis. Compared to Sanger direct sequencing, NGS reduces both time and cost, while allowing for the simultaneous analysis of multiple genes. In this study, 18 of the 34 patients harbored a rare potentially pathogenic variant in genes beyond the five established genes of the leptin‐melanocortin pathway. However, since the selection of genes was predefined, there is a risk of missing potentially pathogenic variants in nontargeted genes, which may limit the comprehensiveness of the analysis. Additionally, comparing results across literature remains challenging, as the reported frequency of rare variants varies between 5% and 10%, depending on the genetic background of the studied populations and the inclusion criteria used. Another important limitation is the heterogeneity of gene panels, which range from 15 to 52 genes [19, 20, 21]. While our gene panel included well‐studied genes such as LEP, LEPR, MC4R, PCSK1, POMC, BDNF, MC3R, NTRK2, and SIM1, other studies have considered additional genes identified through animal models or genome wide association study (GWAS), such as the GHSR, NEUROG3, or OTP genes. This highlights the need for standardized and comprehensive genetic screening approaches to enhance the consistency and comparability of findings across cohorts. The emergence of whole‐exome sequencing (WES) and whole‐genome sequencing (WGS) may help overcome these challenges by enabling broader genetic analyses, including coding and non‐coding regions [23].

One of the main challenges in genetic studies of obesity is the classification of the VUS which were found in 8.1% of the cohort. Due to the lack of functional data, these variants could not be classified as pathogenic, possibly leading to an underestimation of the true impact of pathogenic variants despite the severe phenotype of the studied population. Functional characterization of newly identified variants remains a major challenge. While ACMG classification and in vitro functional tests provide more accurate assessments of variant pathogenicity, most studies still rely on in silico prediction tools due to their ease of use and low cost [24]. However, these tools have limitations, as they do not always reflect the true biological impact of a variant. A large‐scale functional study of over 12,000 missense variants in the LEPR, PCSK1, and POMC genes showed that a substantial proportion of VUS resulted in at least partial loss of function of the encoded protein [25]. Similar findings also have been reported for newly identified PCSK1 variants, demonstrating variable pathogenic effects [10]. To improve the classification of these variants, further studies integrating functional analysis are essential. Machine learning‐based approaches may also provide innovative tools to predict variant effects more accurately and efficiently in the next years [26].

In this cohort, when combining potentially pathogenic variants and VUS, 16.7% of patients carried a variant in the leptin–melanocortin pathway, confirming its major role in energy balance and severe early‐onset obesity. A recent systematic review estimated that 13.2% of patients with severe obesity harbor rare pathogenic or likely pathogenic variants, with 12.3% being heterozygous carriers and 0.93% homozygous or composite heterozygous carriers [24]. Altogether these data strongly support expanded genetic analyses to refine the genetic diagnosis of early‐onset severe obesity and deepen our understanding of its pathophysiology.

The severe phenotype of patients included might partly explain the high frequency of heterozygous carriers in this cohort. Indeed, early‐onset of obesity, before the age of 6, was particularly frequent in adults carrying pathogenic variants. This result may not have been observed in children because of a selection bias. Patients under the age of 18 attending a tertiary pediatric center for obesity have most frequently developed obesity since childhood. Comparison with previously published cohorts of patients with suspected genetic obesity is limited, as the precocious age of onset is often an inclusion criterion in studies [19, 20, 21]. Furthermore, Renard et al. recently reported that early‐onset obesity before 6 years of age or earlier was an expected phenotype in case of a pathogenic variant in genes of the leptin–melanocortin pathway [27], confirming the early impact of the impaired leptin–melanocortin pathway on the evolution of corpulence.

The severity of obesity was not different between carriers and non‐carriers of potentially pathogenic variants, consistent with a finding in a cohort of 1230 patients, which also found no link between genetic status and BMI [20]. This may be explained by the multifactorial origin of obesity, where genetic predisposition interacts with environmental factors such as socioeconomic status or sedentary lifestyle. Large‐scale GWAS showed polygenic obesity score can have an impact on BMI similar to a rare pathogenic variant in the leptin–melanocortin pathway [28]. In addition, the limited number of patients carrying rare variants in this study may have reduced statistical power to detect phenotypic differences. The presence of undetected pathogenic variants in other genes interacting with the leptin–melanocortin pathway may contribute to the severe phenotype of some patients classified as non‐carriers. These findings highlight the need for broader genetic testing, either by expanding gene panels or using other methods such as WGS, particularly in extreme phenotypes [29].

Abnormal feeding behaviors were reported in most patients, regardless of their genetic status concerning pathogenic variants. This finding is likely influenced by the retrospective nature of data collection and the variability in how eating disorders are assessed in routine care.

Although all patients were evaluated by a specialist dietitian and/or physician, standardized questionnaires could improve the characterization of eating behavior especially in the younger patients [30]. Moreover, eating behavior is highly dynamic and can be influenced by various factors throughout life, particularly in adults. Emotional environment, psychological state, psychotropic treatments, restrictive diets, and other environmental factors can modulate eating patterns, making it difficult to establish a direct and stable relationship between genetic predisposition and eating behavior.

The endocrine disorders classically described in biallelic carriers of LEP, LEPR, PCSK1, and POMC variants were rare in this cohort, occurring in only three patients with potentially pathogenic variants (8.8%) compared to 10.5% of non‐carriers. This confirms previous results [8] showing a lower frequency of endocrine disorders in heterozygous carriers compared to biallelic carriers.

The absence of phenotypic differences in patients carrying pathogenic variants may also be due to incomplete penetrance, as this study did not include a control group to analyze the effect of the identified variants in healthy patients. It is therefore possible that these variants may be found in nonobese patients. Previous large‐scale studies showed that MC4R variants substantially influence BMI without systematically causing obesity [28, 31]. Further studies are also needed to decipher the role of heterozygous variants in genes associated with recessive inheritance, such as ADCY3, ALMS1, ARNT2, CEP19, KSR2, LEPR, MAGEL2, MC3R, POMC, NCOA1, and TUB, which may still contribute to obesity risk in a heterozygous state.

Rare ALMS1 variants were found in 21% of patients, but only two patients carried a potentially pathogenic variant. ALMS1 encodes a ubiquitous protein involved in centrosome and primary cilium function [32]. The first patient, a woman aged 22, carried a nonsense variant in exon 8/23 of ALMS1, leading to a loss of over 40% of the protein's amino acids, as well as a pathogenic heterozygous variant of DDX3X, identified elsewhere. She presented with early‐onset severe obesity (BMI = 35.1 kg/m², onset at age 3) associated with impulsive eating, neurodevelopmental disorders, corticotropic deficiency, dysmorphic features, insulin resistance, and fatty liver. No family history of T2DM or obesity was reported. The precise contribution of these variants and their potential interaction in her phenotype remain uncertain. Indeed, DDX3X syndrome, an X‐linked dominant disorder, is typically associated with neurodevelopmental disorders and dysmorphic features, while Alström syndrome, an autosomal recessive disorder, is characterized by obesity with eating disorder, insulin resistance, fatty liver, progressive vision loss leading to blindness around age 20, and central deafness, but without intellectual disability. The patient had no signs of visual or auditory impairment. Furthermore, corticotropic insufficiency is not a classical feature of either syndrome. Genetic analysis of her relatives is ongoing to clarify the pathogenicity of these two variants and their possible interaction. The second patient carried a complete heterozygous deletion of exon 7 of ALMS1, resulting in a frameshift mutation. Her clinical picture combined severe obesity (BMI = 40.6 kg/m²), eating disorder, T2DM, OSA, and fatty liver. The age of obesity onset was not available in medical records, and no family history of obesity was reported. This patient did not exhibit the neurosensory impairments associated with Alström syndrome. Previous studies have reported pathogenic heterozygous ALMS1 variants in patients suffering from severe early‐onset obesity, but their frequency and associated phenotype remain poorly characterized [20, 33]. To date, the role of heterozygous ALMS1 variants in obesity remains uncertain and requires further investigation.

This study has several limitations, including the absence of a control population, which limits the interpretation of genotype–phenotype associations. The selection of variants with low frequency in the general population helped mitigate this bias. Incomplete clinical data in medical records highlight the need for standardized phenotypic characterization specialized in centers managing early‐onset severe obesity. The assessment of eating behaviors should be standardized using validated tools, as methodological variability may bias the interpretation of their genetic determinants. The diagnosis of endocrine deficiencies in this cohort was also challenging. Obesity itself induces hypogonadism, particularly in men [34], which may explain why the most observed endocrine deficits concern the gonadotropic axis. In addition, this deficiency is typically detected after puberty and may have been underestimated in the pediatric cohort, given its median age of 10 years. Finally, segregation data and functional analysis were unavailable for most variants, limiting the ability to confirm their causal role.

In conclusion, this study defines the prevalence of pathogenic variants in 20 genes of the leptin‐melanocortin pathway and their associated phenotype in severe obesity. It reinforces the clinical relevance of expanded genetic analyses to identify patients who may benefit from personalized multidisciplinary management and targeted therapies. Although identifying a likely pathogenic variant or VUS does not necessarily grant access to targeted treatment, it can provide patients and their family with a potential explanation for the disorder, reducing uncertainty and emotional burden. To optimize genetic diagnostics, we propose performing NGS panels including at least 20 genes in patients with severe early‐onset obesity. In cases where no relevant variant is identified or when findings do not fully explain the phenotype, further genetic analysis such as WGS should be considered. A detailed phenotypic characterization is crucial for refining genetic investigations and improving clinical decision‐making. Tools such as Obsgen (http://obsgen.nutriomics.org) developed by our research team could assist clinicians in identifying patients who would benefit most from comprehensive genetic screening.

Funding

N. Faccioli benefited from a fellowship from the French Pediatric Society (French Pediatric Society (Société Française de Pédiatrie, 18000 €) cofounded by Novo Nordisk Laboratory) cofounded by Novo Nordisk Laboratory. This work was supported as part of the national plan for rare diseases by the French Ministry of Health via the defiscience network.

Conflicts of Interest

A. Azar‐Kolakez participated in Rhythm Pharmaceuticals' board and received its support for attending meetings. C. Carette received grants or contracts; consulting fees; payment or honoraria for lectures, presentations, manuscript writing, or educational events; and support for attending meetings from Lilly, Novo Nordisk, Astra Zeneca, Boehringer, Rhythm Pharmaceuticals, and PFIZER. She also participated in BIOPHARMA board. K. Clément is involved as an investigator in clinical trials funded by Rhythm Pharmaceuticals, Novo Nordisk, Boeringher, and Bioprojet, unrelated to the present work. She also received payment or honoraria for lectures, manuscript writing, or educational events from Rhythm Pharmaceuticals and MSD. J. Le Beyec‐Le Bihan participated in Rhythm Pharmaceuticals' board and received from it payment or honoraria for lectures, manuscript writing, or educational events and support for attending meetings. C. Poitou is involved as an investigator in clinical trials funded by Rhythm Pharmaceuticals and Novo Nordisk, unrelated to the present work. She also received payment or honoraria for lectures, manuscript writing, or educational events and support for attending meetings from Rhythm Pharmaceuticals. A. Linglart received consulting fees, payment, or honoraria for lectures, presentations, manuscript writing, or educational events and support for attending meetings from Novo Nordisk, Pfizer, Alexion, Kyowa Kirin, Biomarin, Inozyme. B. Dubern is involved as an investigator in clinical trials funded by Rhythm Pharmaceuticals, Lilly, and Novo Nordisk, unrelated to the present work. She also received payment or honoraria for lectures, manuscript writing, or educational events and support for attending meetings from Rhythm Pharmaceuticals and Novo Nordisk.

Supporting information

Data S1: Supporting Information.

Table S2: Variants details and data compiled for their classification.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.