Towards a genetic obesity risk score in a single-center study of children and adolescents with obesity

Abstract

The objective of this study was to identify clinical and/or metabolic predictive factors of genetic obesity. Subjects aged ≤ 18 years with obesity (BMI ≥ 97 th percentile) followed-up at the Paediatric Endocrinology Clinic of Maggiore Hospital in Novara, Italy were screened for genes associated with obesity by next-generation sequencing. Anamnestic, anthropometric, and biochemical data were collected, and parents completed two questionnaires, to screen for hyperphagia and daytime sleepiness, respectively. The study included 50 patients. Six genetic variants (6/50 patients,12%) were classified as pathogenic/likely pathogenic, three (3/50 patients,6%) as polygenic, and 16 (13/50 patients,26%) as variants of uncertain clinical significance (VUS). Eight patients carried > 1 variant. All pathogenic mutations were in genes implicated in the hypothalamic melanocortin pathway or responsible for syndromic obesity. All subjects with definitive genetic diagnosis developed obesity before five years of age. There were no statistically significant differences in auxological nor metabolic parameters between the three genetic patterns of absent genetic mutations, VUS, and pathogenic/likely pathogenic mutations. Finally, a Genetic Obesity Risk Score was developed using logistic regression analysis, selecting Hyperphagia Questionnaire score, age of onset of obesity, and family history as variables. Genetic screening of our cohort of children and adolescents with severe obesity revealed pathogenic/polygenic variants in 18% of cases, with PCSK1 the most frequently mutated gene and with a definitive genetic diagnosis in 3 patients. Identifying clinical, behavioral, and metabolic features predictive of genetic obesity would facilitate early diagnosis and tailored management.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-96883-x.

Introduction

Obesity is a chronic disease characterized by the presence of excessive adipose tissue, reflected histologically by hyperplasic and hypertrophic adipocytes¹. Pediatric obesity is a global health problem that is increasing in prevalence in every country, regardless of wealth². According to the 2023 national census (https://www.epicentro.iss.it/okkioallasalute/indagine-2023-dati), 19% of Italian children were living with overweight and 9.8% with obesity (2.6% severe obesity), without a significant difference between sexes.

The early identification of any disease improves the effectiveness of preventive interventions^3–5. Childhood obesity prevention and treatment strategies have traditionally focused on behavioral and environmental risk factors. However, there is also a genetic component governing interindividual variations in body weight in response to “obesogenic” environments, as demonstrated in twin, family, and adoption studies^6,7. Therefore, genetic approaches such as genome-wide association studies (GWAS) or next generation sequencing (NGS) can be exploited to characterize the physiological and molecular mechanisms that control body weight^8,9. Genetic causes of obesity can be classified as follows¹⁰:

• monogenic obesity: a severe, early-onset form of obesity with a Mendelian inheritance, caused by a single-gene mutation (with most of the implicated genes involved in the leptin-melanocortin signaling pathway, which regulates appetite and satiety), with little or no influence of the environment.

• polygenic obesity: a common multifactorial form of obesity, resulting from an interaction between the obesogenic environment and hundreds of genetic variants.

• syndromic obesity: a more complex clinical phenotype in which obesity is the key aspect but also cognitive delay, dysmorphic features, organ-specific abnormalities, hyperphagia, and/or other signs of hypothalamic dysfunction are present^10,11.

To validate these findings, here we evaluated the frequency of pathogenic or potentially pathogenic obesity-related mutations in a single-center cohort of children and adolescents living with obesity. We then identified common clinical and/or metabolic features within this population to create a “Genetic Obesity Risk Score”.

Subjects and methods

Written informed consent or assent was obtained from the individuals or legal guardian/next of kin. This study was performed in accordance with the ethical standards and the 1964 Helsinki Declaration and its later amendments. The experimental protocol was approved by the local institutional Ethics Committee (“ROAD2023” Prot. n°634/CE 11/06/2024 - CE116/2024).

Study design and population

Data were collected from children and adolescents diagnosed with obesity (BMI ≥ 97 th percentile for age and sex using WHO charts) in 2022 and 2023 at University Hospital Maggiore della Carità of Novara, Italy. Other inclusion criteria were: age ≤ 18 years and exclusion of secondary causes of obesity (e.g., Prader Willi syndrome, Down syndrome, Cushing’s syndrome, hypothyroidism when restoration of euthyroidism resulted in BMI normalization, liver disease, gastrointestinal disorders, or other known chronic diseases). Medical history was assessed with a focus on birth weight, family history, age of onset of obesity (AoO), and neuropsychiatric disorders (autism spectrum disorder or cognitive delay). Patients underwent clinical and auxological examination (including waist circumference), blood chemistry, and genetic testing. Patients’ parents completed two questionnaires, the Hyperphagia Questionnaire (HQ)^11,12and the Epworth Sleepiness Scale (ESS)¹³, to assess hyperphagia and daytime sleepiness, respectively (see Supplementary Material) during first evaluation. Blood chemistry included glucose, fasting insulinemia, glycated hemoglobin (HbA1c), Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) index, triglycerides (TG), total cholesterol, high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), and TG/HDL-C ratio. Height, weight, and BMI SDS were calculated using WHO growth charts [https://www.who.int/tools/child-growth-standards/standards].

Genetic analysis

All patients underwent next-generation sequencing (NGS) for 79 obesity-related genes (most frequently mutated genes in syndromic/monogenic obesity) plus 1 chromosomal region included in the panel (Rhythm Pharmaceuticals Inc., Boston, MA, USA; see Supplementary Material) using buccal swab and/or peripheral blood samples. Samples were analyzed using Agilent SureSelect technology (Agilent, Santa Clara, CA, USA), and sequencing was performed using the Illumina HiSeq system (Illumina, San Diego, CA, USA). Variants with minor allele frequencies > 1% in public databases were excluded. Variants were classified and described according to ACMG (American College of Medical Genetics) international recommendations¹⁴, together with predictive scoring tools to assess variant pathogenicity, including CADD (Combined Annotation Dependent Depletion), REVEL (Rare Exome Variant Ensemble Learner), MetaSVM and MetaLR, PolyPhen- 2 and SIFT, GERP++ (Genomic Evolutionary Rate Profiling), PhyloP and PhastCons. We used HGMD (Human Genome Mutation Database), ClinVar, and dbSNP to search for reported cases, genetic association studies, and functional studies.

Statistical analysis

Continuous variables are expressed as means (± SD) or medians and interquartile ranges (IQR), and they were compared with the Kruskal-Wallis test with Dunn’s test and Bonferroni correction for multiple comparisons. Categorical variables are expressed as numbers and percentages and were compared with Fisher’s exact test. Variable distributions were evaluated according to three different genetic patterns: an absence of genetic mutations, variants of uncertain significance (VUS), and pathogenic/likely pathogenic mutations. Multivariable binary logistic regression was used to identify independent predictors of genetic obesity and to develop a quantitative scoring system to assist in identifying patients for genetic testing. Considering the low number of events, a limited number of covariates were entered into the final multivariable model to avoid overfitting. Results of logistic regression are reported as odds ratios (OR) with 95% confidence intervals (CI) (Table 1). Risk factors were organized into categories (with reference values for continuous variables), and the reference value W (e.g., mid-point) was determined for each category. We then established how far each category was from the base category in the regression (Table 2). The constant for the points system (B), or the number of regression units corresponding to each point, was defined. We let B reflect the increase in risk associated with a five-point increase in total HQ score (B = 5 × 0.070067 = 0.350335). Points were assigned to each risk factor category (Table 2). The goodness of fit of the final model was tested with the Hosmer-Lemeshow test. Receiver operating characteristics (ROC) curve analysis was used to investigate the diagnostic performance (sensitivity, specificity, positive likelihood ratio) of the model (see Supplementary Materials).

Table 1.

Multivariable logistic model for the genetic obesity risk score.

	Coefficient	Std. err.	Z	P > z	95% CI
Total hyperphagia score	0.070067	0.058435	1.2	0.231	− 0.04446	0.184596
Family history of obesity	1.116575	0.908106	1.2	0.219	− 0.66328	2.89643
Age of onset of obesity	0.337175	0.861241	0.39	0.695	− 1.35083	2.025175

Table 2.

Determination of risk factor profiles and regression coefficients.

	Categories		Reference values (Wij)	βi	X	Y	Points (Z)
Total HQ score	11	19	15 = W1REF	0.070067	0	0	0
	20	29	24.5		0.665637	1.900001	2
	30	39	34.5		1.366307	3.900001	4
	40	49	44.5		2.066977	5.900001	6
	50	59	54.5		2.767647	7.900001	8
	60	65	62.5		3.328183	9.500001	10
Family history of obesity	No		0 = W2REF	1.116575	0	3.187164	0
	Yes		1		1.116575		3
AoO > 5 years	No		0 = W3REF	0.337175	0	0	0
	Yes		1	1	0.337175	0.962436	1
Total score

Notes: X = βi (Wij − WiREF), Y = X/B, Z = Y rounded to the nearest integer.

Results

Of 54 patients tested, four were excluded because the DNA sample did not comply with the quantity/quality criteria to perform genetic testing.

Genetic analysis

Thirty-four genetic variants were detected in 27 patients (54% of cases); of these, 12% (six patients) were classified as pathogenic, 6% (three patients) as polygenic, and the remaining 30% (15 patients) as VUS. Of note, eight patients harbored more than one variant. Clinical and genetic features of patients with pathogenic/likely pathogenic or polygenic variants (18% collectively) are displayed in Table 3, and patients with VUS are listed in Table 4.

Table 3.

Clinical and genetic data of patients with pathogenic/likely pathogenic or polygenic gene variants.

Sex/age (years)	BMI SDS	Mutant gene	Variant protein change	Zygosity	ACMG interpretation	ClinVar/PMID	Clinical findings	AoO (years)	Family history obesity
M/1.7	6.6	LEPR PCSK1	NM_002303.6:c.1345 T > G p.(Trp449Gly) NM_002303.c.3G > A p.(Met1Ile) NM_000439.5:c.661 A > G p.(Asn221 Asp)	Compound Heterozygous Heterozygous	Pathogenic Polygenic risk variant	Novel Novel ClinVar ID: 14,040	Neonatal feeding problems, hyperphagia, dyslipidemia	< 1	No
M/5.7	3.6	PCSK1	NM_000439.5:c.661 A > G p.(Asn221 Asp)	Heterozygous	Polygenic risk variant	ClinVar ID: 14,040	Hyperphagia, hypertriglyceridemia, mild insulin-resistance	1	No
F/10.9	2.1	PCSK1	NM_000439.5:c.661 A > G p.(Asn221 Asp)	Heterozygous	Polygenic risk variant	ClinVar ID: 14,040	Hypertriglyceridemia, insulin-resistance	1	Yes
F/11.7	3.4	KIDINS220	NM_020738.3:c.2707G > A p.(Arg901Gln)	Heterozygous	Likely pathogenic	ClinVar ID: 930,551	No dysmorphic features, nor hyperphagia	6	Yes
M/15.9	2.8	NRP2	NM_201266.2:c.1715G > A p.(Arg572Gln)	Heterozygous	Likely pathogenic	Novel	Hypertension	11	Yes
M/7.4	3.4	ALMS1	NM_001378454.1:c.586 del p.(Thr196 Argfs* 7)	Heterozygous	Likely pathogenic	ClinVar ID: 1,419,935	Bilateral vesicoureteral reflux, autism	4	No
F/5.2	3.5	PCSK1	NM_000439.5:c.661 A > G p.(Asn221 Asp)	Heterozygous	Polygenic risk variant	ClinVar ID: 14,040	Dyslipidemia, mild insulin-resistance	4	Yes
M/1.5	5.33	ALMS1	NM_001378454.1:c.5923 del p.(Glu1975Serfs* 8)	Homozygous	Pathogenic	ClinVar ID: 92,193	Nystagmus, hyperphagia, hyperthyrotropinemia	< 1	No
F/11.8	2.3	MC4R	NM_005912.3:c.919 C > T p.(Gln307 Ter)	Heterozygous	Likely pathogenic	PMID: 19,284,607	Hypertension, insulin-resistance, hepatic steatosis	3	Yes

We highlighted in bold cases in which a definitive diagnosis was established.

Table 4.

Clinical and genetic data of patients with VUS.

Sex/age (years)	BMI SDS	Mutant gene	Variant protein change	Zygosity	ClinVar/PMID	Clinical findings	AGE of obesity onset (years)	Family history obesity
F/12	2.9	IFT74	NM_025103.4:c(587 + 1_588 - 1)_(726 + 1_727 - 1)dup	Homozygous	Novel	Hyperphagia	8	Yes
M/16.7	2.7	VPS13B CEP290	NM_017890.5:c.3811 A > T p.(Thr1271Ser) NM_025114.4:c(102 + 1_103 - 1)_(441 + 1_442 - 1)dup	Heterozygous Heterozygous	ClinVar ID: 195,900 Novel	Dyslipidemia, severe insulin-resistance	10	Yes
M/13.9	2.8	TBX3	NM_005996.4:c.1489 C > T p.(Pro497Ser)	Heterozygous	Novel	Autism, hyperphagia	3	Yes
F/9.1	2.9	RAI1	NM_030665.4:c.3065 A > G p.(Lys1022 Arg)	Heterozygous	Novel	Hyperphagia, insulin-resistance	10	Yes
F/12	2.7	BBS9	NM_198428.2:c.1993 C > T p.(Leu665Phe)	Homozygous	ClinVar ID: 166,740	Hypertension, insulin-resistance	5	No
M/5.7	3.9	HTR2 C	NM_000868.4:c.503G > A p.(Arg168Gln)	Hemizygous	Novel	Hyperphagia, hypertriglyceridemia, mild insulin-resistance	1	No
M/3.5	6.7	DNMT3 A	NM_175629.2:c.817 A > G p.(Asn273 Asp)	Heterozygous	Novel	Neonatal feeding problems, hyperphagia, autism	< 1	Yes
F/11	2.3	MC4R	NM_005912.3:c.757G > A p.(Val253Ile)	Homozygous	ClinVar ID: 492,862	Tall stature, insulin-resistance	1	No
M/10.6	3.8	SEMA3 F	NM_004186.5:c.1255 C > T p.(Pro419Ser)	Heterozygous	Novel	Hypertension	4	Yes
M/10.5	2.6	PPARG	NM_015869.5:c.1003G > C p.(Val335Leu)	Heterozygous	ClinVar ID: 436,394	Hyperphagia	6	No
F/8.4	3.8	EP300	NM_001429.4:c.5356 A > G p.(Ile1786 Val)	Heterozygous	Novel	Hepatic steatosis, hyperphagia	5	Yes
F/9.3	2.7	CREBBP	NM_004380.3:c.674 C > G p.(Pro225 Arg)	Heterozygous	Novel	Dyslipidemia, hypertension	2	No
F/5.8	3.5	SIM1	NM_005068.3:c.133 C > A p.(Leu45Ile)	Heterozygous	Novel	Hypertriglyceridemia	4	No
M/1.5	5.33	DYRK1B	NM_004714.3:c.970 C > T p.(Arg324 Cys)	Heterozygous	ClinVar ID: 2,636,132	Nystagmus, hyperphagia, hyperthyrotropinemia	< 1	No
F/13	2.7	POMC LEPR	NM_000939.4:c.706 C > G p.(Arg236Gly) NM_002303.6c.1968G > T p.(Lys656 Asn)	Heterozygous Heterozygous	PMID: 12,165,561	Insulin-resistance, hyperphagia	10	Yes

Baseline characteristics of the study population

Of the 50 enrolled subjects, 26 (52%) were male. The mean age was 9.2 ± 4.1 years, with a range of 1.5 to 16.7 years at the time of clinical observation. Most patients (29/50, 58%) were prepubertal, as defined by the Tanner scale. The mean weight was 2.89 ± 1.1 SDS and the mean height was 0.97 ± 1 SDS, with a median difference from mid-parental target height (delta target) of 0.1 SDS (IQR 1). The mean BMI was 3.26 ± 1 SDS and mean waist circumference was 84.04 ± 14.6 cm. Waist-to-height ratio was > 0.5 in all subjects but one. Most subjects (45/50, 90%) were adequate for gestational age (AGA) at birth, and only two subjects were LGA (large for gestational age)¹⁵. Twenty-nine subjects (58%) had a paternal and/or maternal family history of obesity. The median age of onset of obesity was 3.25 (IQR 3.75 years), with 34 subjects (68%) developing obesity before the age of five years. Only five subjects (10%) had neuropsychiatric disorders (autism spectrum disorder and/or cognitive delay).

The population was divided into three groups according to genetic test results: subjects with a negative test or with benign mutations (28/50, 56%), subjects with VUS (13/50, 26%), or subjects with polygenic/pathogenic mutations (9/50, 18%). Table 5 shows the auxological and metabolic characteristics of each group. There were no significant differences between the three genetic patterns. Most subjects had family history of obesity (57.1%, 61.5%, and 55.5% in the three groups, respectively) and were prepubertal, except in the group with VUS (46.1% prepubertal, p = 0.58). The lipid profiles (particularly triglycerides and TG/HDL-C ratio) were altered in all groups, with median triglyceride levels of 89 mg/dL (IQR 54 md/dL) in the entire cohort.

Table 5.

Auxological and biochemical characteristics of the entire study cohort and the three groups according to genetic test result.

	Entire cohort	Negative/benign (N = 28)	VUS (N = 13)	Polygenic/pathogenic mutations (N = 9)	p-value
Age (years)	9.2 (4.1)	9.39 (5.85)	10.5 (4.78)	7.4 (6.6)	0.57
Weight (SDS)	2.8 (1.3)	2.9 (1.3)	2.7 (1.5)	3.1 (1.8)	0.25
Height (SDS)	0.97 (1.47)	0.93 (1.56)	0.8 (0.95)	1.3 (1.42)	0.22
Delta target	0.1 (1)	0.23 (1.4)	0.1 (0.9)	0.1 (0.8)	0.78
BMI (kg/m2)	28.2 (7)	26.7 (8.06)	31 (6.6)	28.2 (4.8)	0.88
BMI (SDS)	3.3 (1)	3.35 (0.85)	2.9 (0.67)	3.5 (1.19)	0.22
Waist circumference (cm)	81 (20.5)	82 (18.7)	86 (22.5)	79.9 (23)	0.68
Waist-to-height ratio	0.58 (0.08)	0.58 (0.09)	0.58 (0.09)	0.58 (0.09)	0.81
Prepubertal (%)	58	64.3	46.1	55.5	0.58
Family history of obesity (%)	58	57.1	61.5	55.5	> 0.99
Age of onset of obesity (years)	3.25 (3.75)	3 (3.5)	3.5 (4.25)	3.5 (5)	0.46
Blood glucose (mg/dl)	85.5 (9)	85.5 (9)	85 (7)	86 (10)	0.99
Insulin (mU/l)	12.7 (18.6)	15 (16.2)	12.6 (26.3)	12.5 (18.6)	0.96
Hba1c (%)	5.4 (0.3)	5.3 (0.4)	5.5 (0.4)	5.45 (0.6)	0.47
HOMA-IR	2.5 (4.3)	3.3 (3.9)	2.54 (3.5)	2.3 (2.47)	0.54
Total cholesterol (mg/dl)	152 (27)	153.6 (29.3)	150.9 (29.4)	166.4 (31.5)	0.3
HDL-C (mg/dl)	48 (12)	44.9 (11.3)	49 (11.2)	52.4 (12.3)	0.44
Triglycerides (mg/dl)	89 (54)	100 (46.1)	90 (45.4)	116 (48.9)	0.31
Tg/HDL-C ratio	2 (1.9)	2.5 (1.4)	2 (1.4)	2.4 (1.5)	0.72

Data are expressed as percentages or median (IQR).

Questionnaires

All enrolled subjects’ parents completed the Hyperphagia Questionnaire (HQ) and Epworth Sleepiness Scale (ESS) questionnaire (Table 6). The ESS was normal (total score < 10) in all subjects, and there were no significant differences in HQ or ESS scores between the three genetic groups.

Table 6.

Results from the hyperphagia questionnaire and Epworth sleepiness scale for the entire cohort and in the three genetic categories.

	All (n = 50)	Negative/benign (n = 28)	VUS (n = 13)	Polygenic/pathogenic mutations (n = 9)	p-value
Hyperphagic behavior	8.5 (2.8)	9.2 (2.5)	9.7 (4.6)	11.1 (5.3)	0.79
Hyperphagic impulse	8.5 (2.8)	8.3 (2.4)	9.3 (3.8)	10.7 (5)	0.64
Hyperphagia severity	3.7 (1.6)	4.1 (1.6)	2.9 (1.6)	4.7 (2.1)	0.15
Total hyperphagia	21.6 (7.1)	21.6 (5.2)	21.9 (9.2)	26.5 (11.8)	0.75
ESS	4.1 (4.9)	2.7 (3.4)	7 (5.6)	5.7 (4.5)	0.13

All data are expressed as median (IQR).

Genetic obesity risk score

Logistic regression analysis was performed to develop a “Genetic Obesity Risk Score”. Considering the small sample size and the absence of significant differences in univariable analyses, we decided to include only variables of high clinical relevance in the model: total HQ score, a positive family history of obesity, and early AoO of obesity (5 years). Other variables such as neuropsychiatric disorders and BMI were not selected because of their asymmetric distributions. A multivariable logistic model (Table 1) and ROC curve (AUC = 0.71) were obtained and, consistently with the univariable analyses, none of the selected covariates was statistically significant. As depicted in Fig. 1, we estimated the risk associated with the total score using multiple logistic regression. In the suggested system, the total score ranged from 0 to 14.

Fig. 1.

Risk probability for the Genetic Obesity Risk Score.

Discussion

In this Italian single-center cohort of 50 children and adolescents living with obesity, we detected thirty-four genetic variants in approximately half of the patients. A total of six variants (12%) were classified as pathogenic/likely pathogenic, whereas 6% and 30% of subjects had polygenic gene variants and VUS, respectively. Similar studies have been conducted in other countries. For instance, in their analysis of 105 Turkish children living with severe early-onset obesity, Akinci et al. screened for 41 known obesity-related genes and found a genetic mutation in 10.4% of cases¹⁶. A national study from Slovenia, which analyzed 11 genes, revealed that 1.4% of enrolled subjects had pathogenic heterozygous variants in leptin-melanocortin signaling pathway genes and 4.1% of participants carried rare VUS¹⁷. In a retrospective French study of 1,486 adults and children with obesity screened for mutations in LEP, LEPR, POMC, and PCSK1, the authors found at least one variant in 8.5% of the cohort¹⁸. Similarly to our study, the diagnostic yield for pathogenic mutations in a Dutch cohort was 13%¹⁹. Our NGS gene list included 79 genes plus 1 chromosomal region, which may explain the higher positive detection rate in our population, together with possible selection bias since we analyzed children attending a tertiary pediatric endocrine center. Our panel incorporates both genes that are well-established in the literature as being associated with genetic forms of obesity and genes described in a limited number of reports. The latter were deliberately included to explore their potential contribution to the genetic architecture of obesity. The variants identified in these genes are predominantly classified as VUS. This classification reflects the current limitations in available functional and clinical data, meaning that their precise impact on obesity risk remains unclear. Further studies, including functional assays, segregation analysis, and expanded cohort evaluations, will be essential to clarify their pathogenicity and clinical relevance.

The detected mutations (LEPR, MC4R, NRP2, and PCSK1) are implicated in the hypothalamic melanocortin pathway, which is responsible for appetite and satiety regulation. LEPR encodes the leptin receptor, a key hormone controlling energy and feeding control in the hypothalamus. LEPR deficiency causes an autosomal recessive, non-syndromic monogenic form of obesity characterized by severe early-onset obesity, marked hyperphagia, and endocrine abnormalities²⁰. To date, 34 different pathogenic/likely pathogenic variants in LEPR have been reported (https://www.ncbi.nlm.nih.gov/clinvar/). The compound heterozygous variants detected here have not been described in the literature and are not reported in the gnomAD database. They are located in a highly conserved residue, with bioinformatic analysis predicting them to have a deleterious effect. Our affected patient (OBG13) developed obesity in the first months of life; his birth weight was AGA, and he had no facial dysmorphisms, neurodevelopmental delay, hypertension, hypogonadism, nor other endocrinopathies. His blood tests showed hypertriglyceridemia and a high TG/HDL-c ratio (> 2.2).

MC4R encodes the MC4 receptor, to which α-MSH binds to promote satiety and reduce food intake. Subjects with MC4R variants develop early-onset severe obesity, hyperphagia, increased linear growth and bone density, but normal pubertal development, fertility, and thyroid function^21,22. We detected two different MC4R variants: the NM_005912.3:c.919 C > T p.(Gln307Ter) heterozygous variant (patient OBG50) was classified as pathogenic according to ACMG guidelines, since it causes a premature stop codon. This variant has been already described in an Italian girl living with obesity, and its deleterious effect was confirmed with functional bioassays²³. The homozygous NM_005912.3:c.757G > A p.(Val253Ile) variant in exon 1 of MC4R (patient OBG23) is located in a highly conserved residue and is described in the literature in individuals with obesity²⁴. However, a previous study failed to show segregation of this variant in affected family members²⁵, so, with the available information, this variant should be classified as a VUS. Several features of the patient (e.g., hyperphagia, tall stature, hyperinsulinemia) closely resemble those reported previously in MC4R knock-out mice²⁶. Mutations in MC4R are the most relevant cause of monogenic obesity, with a variable prevalence of 0.5–8.5%^27,28.

Nevertheless, PCSK1 was the most frequently mutated gene in our population (four cases). PCKS1 encodes pro-hormone convertases 1 and 3 (PC1 - 3), proteolytic enzymes that play a key role in the generation of substrates such as POMC and α-MSH, which increase energy expenditure and regulate glucose metabolism via MC4R in the hypothalamic paraventricular nucleus^29,30. The detected variant NM_000439.5:c.661 A > G p.(Asn221 Asp) (OBG13, 17, 22, 44) is described in GWAS as a common polygenic risk variant for obesity^31–33, with functional validation studies showing potential pathogenicity through a 10–30% reduction in enzymatic activity³⁴.

Interestingly, we found a novel, likely pathogenic heterozygous variant NM_201266.2:c.1715G > A p.(Arg572Gln) in NRP2, which encodes neuropilin 2, a transmembrane receptor that, after binding with semaphorin 3 (SEMA3) in the hypothalamus, promotes the development of neuronal projections to the paraventricular nucleus and indirectly stimulates MC4R⁸. NRP2 is not yet associated with a disease in OMIM. However, heterozygous variants in this gene have been reported in patients with obesity and have also been implicated in the development of the hypothalamic melanocortin circuit³⁵.

The clinical and therapeutic implications of heterozygous variants in genes in the leptin-melanocortin pathway in severe obesity are still a matter of debate. Courbage et al. speculated a possible cumulative effect of combined heterozygous variants, since the BMI of patients with this profile was significantly higher than the BMI of patients with a single heterozygous variant and similar to the BMI of patients with homozygous variants¹⁸. The potential benefit of treatment with setmelanotide in these subjects has not been yet explored. In our cohort, eight patients carried more than one heterozygous variant. For example, patient OBG49 carried combined heterozygous VUS variants in POMC and LEPR; she was a 13-year-old girl with a three-year history of obesity (BMI 2.7 SDS), mild hyperphagia, and insulin resistance, in whom lifestyle interventions were ineffective and only liraglutide treatment allowed a sufficient weight control.

Genetic screening in our cohort eventually led to diagnosis of Alström syndrome, an autosomal recessive ciliopathy caused by mutations in ALMS1³⁶, in one case with the fitting phenotype (congenital nystagmus, hyperphagia), and it was therefore fundamental for starting tailored multidisciplinary follow-up. We identified VUS in 30% of the patients. Unfortunately, we did not perform any functional tests, which could better define the impact of VUS. The majority of mutations classified as VUS have also been related to syndromic obesity, including Rubinstein Taybi syndrome (EP300), Bardet Biedl syndrome (BBS9, IFT74), and Tatton-Brown Rahman syndrome (DNMT3A); however, these patients did not display the typical phenotype associated with the syndromes despite the presence of homozygous variant (for example polydactyly for Bardet Biedl Syndrome). Syndromic obesity is a more complex clinical phenotype in which obesity is the key aspect but that also features cognitive delay, dysmorphic features, organ-specific abnormalities, hyperphagia, and/or other signs of hypothalamic dysfunction. Over 80 different syndromes presenting with pediatric obesity have been described³⁷.

Fifty-six% of patients have no identified variant. Exome sequencing could certainly help identify new genes implicated in obesity, although previous studies found a low number of new candidate genes by screening hundreds of patients³⁸. Whole-genome sequencing could certainly enhance detection in cases as it enables to discover variants located in deep intronic or regulatory regions, which are rare but can play a role. Additionally, copy number analysis would allow the identification of genomic rearrangements. In the future, rare genetic variants that are currently classified as VUS can be systematically reanalyzed as new scientific evidence, functional studies, and population data become available. Advances in genomic research, machine learning algorithms, and large-scale biobanks will enable a more accurate interpretation of these variants, potentially leading to their reclassification as either benign or pathogenic. Therefore, we are planning to reanalyze the sequences of patients with VUS or negative tests in the future. This continuous reassessment is crucial for improving diagnostic accuracy, guiding personalized treatment decisions, and enhancing genetic counseling, ultimately contributing to better patient care and disease management.

We found no statistically significant differences in auxological and metabolic parameters between the three genetic patterns (negative test/benign variant, VUS, pathogenic/polygenic variant), probably due to small sample sizes. In particular, BMI-SDS values were not higher in patients with genetic obesity disorders than those with negative tests (p = 0.22). Notably, AoO was similar between the three groups, which might be related to selection bias, since most of the subjects in the cohort started gaining weight before the age of five. Early-onset obesity (< 5 years) is a recognized red flag for suspected genetic obesity³⁹. Numerous studies set five years as the cut-off age for the definition of monogenic or syndromic forms of obesity^19,40. Recently, some authors identified an optimal AoO cut-off of ≤ 3.9 years for non-syndromic and ≤ 4.7 years for syndromic genetic obesity. Nevertheless, some forms of genetic obesity show variable and gradual BMI increases during childhood⁴¹, so a more specific and tailored approach may be advisable. In addition, the early AoO seen in subjects with negative genetic tests reflects a global trend in childhood obesity, with a dramatically increasing prevalence in very young children, primarily due to “obesogenic” environments, overnutrition, and sedentary lifestyles⁴². These environmental influences may also impact subjects with genetic obesity.

A positive family history of obesity was more common in patients with VUS or a negative genetic test, but this was not statistically significant. Unfortunately, parents’ genetics were not investigated in this study, so we could not perform segregation studies to better define VUS. A family history of obesity could potentially select families with a high polygenic score, whereas the absence of a family history suggests a de novo monogenic origin of obesity. Nevertheless, it is also known that children who have parents with overweight/obesity are at high risk of developing the disease themselves, even in the absence of a genetic background, as the family environment exerts an important influence on the development of the children’s habits⁴³. To note, the patient with MC4R pathogenic variant had a positive family history of obesity.

Several studies have highlighted that birth weight may be associated with the development of several childhood-onset pathological conditions, including obesity. Nam et al. demonstrated that LGA infants are at increased risk of becoming children, adolescents, and young adults with overweight or obesity, as well as developing metabolic syndrome and giving birth to LGA offspring⁴⁴. Similarly, Saggese et al. found metabolic and body composition alterations with accumulation of adipose tissue especially in the abdomen in SGA infants. These conditions are more evident in SGA subjects who show rapid and excessive weight regain after birth⁴⁵. In subjects with genetic obesity, birth weight is often AGA, followed by a rapidly increasing BMI within the first 2–5 years of life⁴¹. Nearly all of our population was AGA at birth, and no patient was small for gestational age.

Previous studies have shown that children with obesity are four times more likely to develop type 2 diabetes than children with a normal BMI⁴⁶. Other studies also describe the presence of insulin resistance in syndromic (e.g., Alström syndrome) and monogenic (leptin, MC4R, and POMC deficiency) forms of obesity. As expected, nearly all the patients in our cohort displayed insulin resistance and some of them also dyslipidemia.

Hyperphagia is commonly observed in children with genetic obesity⁴⁷. We also screened for hyperphagia using the Hyperphagia Questionnaire (HQ), specifically designed by Dykens et al. to measure symptoms of food-related concerns and behavior in subjects with Prader Willi syndrome¹¹; the HQ is already validated for Italian children with this condition¹². The highest total score was found in subjects with pathogenic/polygenic mutations, although without statistical significance. Recently, in a study by Arnouk et al. the same questionnaire was used to explore eating behavior in a cohort of children with severe, early-onset obesity with or without genetic diagnosis: the authors concluded that HQ could help diagnose severe hyperphagia in children, regardless of genetic origin of the disease⁴⁸.

Finally, we explored the possibility of developing a “Genetic Obesity Risk Score”, although the number of positive genetic tests was limited, and the model was not accurate. Three risk factors were selected based on their clinical importance: the total score recorded in the HQ questionnaire, the presence of a family history of obesity, and AoO of obesity. After categorizing each risk factor, a score was established based on the values recorded in each category. The sum of the points provided a total score between 0 and 14. In this way, using Fig. 1, it is possible to estimate the probability of a positive genetic test. For example, a point score of 8 corresponds to an 80% probability of finding an obesity-related genetic variant. No risk scores for monogenic obesity have been published so far to our knowledge. Khera et al. proposed a polygenic risk score (PRS) by analyzing 2.1 million SNPs in the UK Biobank that increased the prediction of BMI alterations up to 23% in adults and children⁴⁹. Despite these strong associations with BMI and obesity, the predictive performance of the PRS is weak⁵⁰. Differently from monogenic diseases, polygenic risk scores aggregate the small effects of many common genetic variants to estimate genetic susceptibility. Our pivotal “Genetic Obesity Risk Score”, on the contrary, is meant to be used as a guide for clinicians to suspect a monogenic/syndromic cause of obesity in children and to select patients to undergo genetic screening. It will need to be applied to a larger population for validation.

A major limitation of our study is the small sample size. A recent review reported that genetic screening was only performed in 8% of patients in whom it would be indicated in the USA⁵¹. An important reason for underdiagnosis might be limited access to genetic diagnostics. Although NGS is now easily accessible in clinical practice and is becoming more affordable, it is not yet part of routine care in many countries. Moreover, it may take weeks to months for the results, due to increasing demand, and this limits its use and diagnostic potential. Turnaround time is known to impact outcomes in many clinical settings. In the context of pediatric genetic obesity, protracted diagnostic interval means prolonging the diagnostic odyssey and delaying specific treatment and support for patients and families. A systematic review calculated the ‘genetic prevalence’ of LEPR deficiency as 1.34 per 1 million people in Europe (998 predicted patients), higher than expected based on the literature⁵². Missing these diagnoses could have important clinical consequences, since specific treatments and management strategies could not be started. Our clinical management of patients changed when we received the results of the genetic test: for instance, patient diagnosed with Alström syndrome underwent renal and ophthalmological screening and patient with LEPR deficiency was candidate to setmelanotide treatment with benefit on his weight. This underlines the importance of performing genetic test when appropriate.

Conclusion

This study emphasizes the importance of performing genetic testing when appropriate in the context of severe, early-onset pediatric obesity. Genetic screening of our cohort of children with obesity revealed pathogenic/polygenic variants in 18% of cases, with PCSK1 the most frequently mutated gene. We established a definitive genetic diagnosis for obesity in 3 patients. Identifying patients with genetic obesity is necessary to acquire more information on etiology, define clinical-phenotypic-metabolic evolution, and to optimize treatment. As more obesity-associated variants are being discovered, there is a growing expectation that genetic information will soon be used to identify individuals at risk, especially for monogenic/syndromic forms of obesity. The future of prevention, diagnosis, and treatment should focus on a greater understanding of the genetic determinants of childhood obesity.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conception and design of work: I.R and C.P. Data acquisition, analysis, and interpretation: G.M., M.G. and S.A. Drafting of work: G.M. and C.P. Critical revision and final approval of version: I.R., M.G, F.P., S.M. and S.B. All authors have responsibility for all aspects of the work, ensuring that issues relating to the accuracy or integrity of any part of the work are investigated and resolved appropriately.

Funding

Buccal swab kits for genetic tests were provided by Rhythm Pharmaceuticals Inc. (222 Berkeley Street, Suite 1200, Boston, MA 02116, USA) and analyzed in an ISO15189-accredited lab in Porto, Portugal, by Unilabs/CGC Genetics (Nygaardsvej 32, 2100 Copenhagen, Denmark) for the project “Rare Obesity Advanced Diagnosis™ (ROAD)”. Rhythm Pharmaceuticals did not participate in nor support the present study.

Data availability

Data is provided within the manuscript or supplementary information files. The raw data collected in this study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

Written informed consent was obtained from the individual(s), and minor(s)’ legal guardian/next of kin. This study was performed in accordance with the ethical standards and the 1964 Helsinki Declaration and its later amendments. The experimental protocol was approved by the local institutional Ethics Committee (“ROAD2023” Prot. n°634/CE 11/06/2024 - CE116/2024).