Genome-wide association study provides novel insight into the genetic architecture of severe obesity

Abstract

Severe obesity (SevO) is a primary driver of cardiovascular diseases (CVD), cardiometabolic diseases (CMD) and several cancers, with a disproportionate impact on marginalized populations. SevO is an understudied global health disease, limiting knowledge about its mechanisms and impacts. In genome-wide association study (GWAS) meta-analyses of the tail end of the BMI distribution (≥95th percentile BMI) and two SevO phenotypes [Obesity Class III BMI ≥ 40 kg/m2 and Obesity Class IV BMI ≥ 50 kg/m2] in 159,359 individuals across eleven ancestrally diverse population-based studies followed by replication in 480,897 individuals across six ancestrally diverse studies, we identified and replicated three novel signals in known loci of BMI [TENM2, PLCL2, ZNF184], associated with SevO traits. We confirmed a large overlap in the genetic architecture of continuous BMI and severe obesity phenotypes, suggesting little genetic heterogeneity in common variants, between obesity subgroups. Systematic analyses combining functional mapping, polygenic risk scores (PRS), phenome wide association studies (PheWAS) and environmental risk factors further reinforce shared downstream comorbidities associated with continuous measures of BMI and the importance of known lifestyle factors in interaction with genetic predisposition to SevO. Our study expands the number of SevO signals, demonstrates a strong overlap in the genetic architecture of SevO and BMI and reveals a remarkable impact of SevO on the clinical phenome, affording new opportunities for clinical prevention and mechanistic insights.

Fig 1. Schematic of the study design for genomic analysis and systematic comparisons of SevO in ancestrally diverse populations (Fig 1 created in biorender (2025, https://BioRender.com/f9g0w8p, ZT28HSBHBH.).

Fig 2. Forest-plot illustrating the direction of association for the four validated variants (discovery and replication) across the three obesity classes (95% class, Obesity Class 3 and Obesity Class 4) for each stratum (all, females and males).

Fig 3. Phenotypic variance explained (R2) by (A) using HapMap 3 SNPs in the BMI GWAS from the GIANT consortium and (B) using the PRS-CS method with HapMap3 consortium SNPs for each of the SevO GWAS traits (e.g., 95% Obesity, Obesity Class III and Obesity Class IV).

Using approximately 1/3 the sample size in our SevO PRS we show similar predictive power in explaining phenotypic variance of SevO traits when compared to the PRS generated from the BMI GWAS.

Fig 4. PheWAS analysis of secondary phenotypes (max 1668) within the UKBB Europeans with our Obesity Class III PRS derived from our discovery analyses.

Fig 5. Relationship of SevO PRS > 90^th percentile with < 10^th percentile across BMI categories in ancestry -combined samples from the UKBB.

Fig 6. Lifestyle modeling comparisons between sample groups defined by PRS upper and lower deciles and BMI categories.

We compared individuals having a low susceptibility PRS (<10^th percentile) and a BMI < 25 kg/m² with 1) “Resilient” group [PRS in the ≥ 90^th percentile and a BMI < 25 kg/m²], 2) < 10^th percentile for PRS and SevO, 3) >90^th percentile for PRS and SevO Class III and 4) all PRS and SevO Clas III with different lifestyle factors including dietary, physical activity and sleep patterns, alcohol and smoke servings and perceived body size at age 10.