DNA methylation and gene expression analysis in adipose tissue to identify new loci associated with T2D development in obesity

Obesity, driven by chronic energy imbalance, causes adipose tissue expansion and dysfunction that contribute to insulin resistance and type 2 diabetes (T2D). Visceral adipose tissue (VAT) accumulation is particularly linked to T2D risk, yet individuals vary in susceptibility that is not fully explained by lifestyle, environment, or genetics. Epigenetic mechanisms, especially DNA methylation, may mediate interactions between genetic background and environmental exposures and help explain differential T2D risk in obesity. Prior studies have found obesity-associated DNA methylation changes, some linked to T2D or glucose homeostasis, but paired analyses of DNA methylation and gene expression in the context of T2D, particularly in VAT, are scarce. This study investigates VAT DNA methylation differences between obese individuals with and without T2D, examines correlations between methylation and gene expression changes, and evaluates relationships with T2D-related traits (fasting glucose and HbA1c) to better understand mechanisms underlying T2D development in obesity.

The authors note that multiple studies have reported DNA methylation alterations in obesity, with some loci associated with T2D and glucose homeostasis. However, the interplay between DNA methylation and gene expression is complex and region-dependent, and few studies have paired methylation and gene expression analyses in T2D, particularly in VAT. Prior work has implicated methylation changes in insulin resistance, adipogenesis, inflammation, and mitochondrial processes, and highlighted dynamic methylation in CpG shores and shelves. Evidence for epigenetic involvement of the MHC locus in T2D is limited but emerging. Overall, the literature supports a role for epigenetic regulation in metabolic dysfunction, while emphasizing gaps in integrative methylation–expression analyses directly in adipose tissue in T2D.

Design: Cross-sectional analysis of VAT from obese women with or without T2D. Participants: 19 female adults (BMI ≥ 35 kg/m²) undergoing bariatric surgery in Mexico City; 9 without diabetes (OND) and 10 with T2D (OD) per ADA criteria. Exclusions: other endocrine diseases or dysregulated hypertension. Medication: in OD, 4 metformin+insulin, 4 metformin only, 1 insulin only; one OND on metformin. Ethics: informed consent; approved by institutional committees. Sample collection: VAT biopsies collected during surgery, stored in RNAlater at −70°C. DNA methylation: DNA extracted from 50 mg VAT (QIAamp kit). Illumina Infinium Human Methylation EPIC BeadChip (850K) used. Quality control via NanoDrop ratios >1.8 and gel electrophoresis; all samples passed QC. Data processing in R using ChAMP: probe filtering (low bead count, non-CpG, multi-hit, SNP-related per Zhou 2016; Y-chromosome probes removed), leaving 781,385 probes. Normalization: beta-mixture quantile normalization. Sources of variation assessed by SVD including age, batch, and treatment; ComBat used to adjust significant covariates (age and array batch). Probe annotation from EPIC v1.0 B5 manifest; promoter defined as TSS1500, TSS200, 5'UTR, or first exon; gene body included Body, exon-bound, 3'UTR. Differential methylation: delta β = mean OND β − mean OD β. Limma used with Benjamini–Hochberg FDR control; DMCs defined at FDR < 0.05. DMRs identified with Bumphunter (default: clusters min 7 probes, max 300 bp gap), area p < 0.05. Gene expression: RNA extracted from ~150 mg VAT (RNeasy Lipid Tissue kit). Quality via Bioanalyzer (RIN > 8). Expression profiled with Affymetrix Clariom S Human arrays. Preprocessing with oligo Bioconductor; raw expression deposited (E-MTAB-11841). Normalization improved by co-processing with available blood-sample data from same array; robust multi-array average used. Batch and age corrected with ComBat. Y-chromosome probes removed; 19,872 probe sets retained. Differential expression assessed with Limma; after BH correction no DEGs remained significant; for downstream analyses uncorrected p < 0.05 with |logFC| > 0.5 threshold applied. Correlation analyses: For DMC–DEG pairs, Pearson correlation with bootstrap (R = 1000) using Boot package; significance at p < 0.05. Associations of DMC β values with fasting glucose and HbA1c analyzed similarly. Pathway analysis: WebGestalt for KEGG pathway overrepresentation of genes with DMCs, DMRs, and DEGs; p values BH-adjusted. Validation and extended analyses: Independent VAT methylation datasets from 14 Chinese (OND=8, OD=6) and 7 German (OND=3, OD=4) women (EPIC arrays; GSE162166 and E-MTAB-10999) combined for validation (11 OND, 10 OD). DMC CpGs from discovery tested for direction and significance in validation (nominal p < 0.05). Extended multi-ethnic analysis merged public datasets with discovery, totaling 20 OND and 20 OD; genome-wide methylation re-analyzed as above. Data availability: methylation E-MTAB-11037; expression E-MTAB-11841.

- Participants: OD and OND groups had similar BMI, blood pressure, and lipid levels; OD had higher HbA1c (6.3 ± 1% vs 5.4 ± 0.2%) and fasting glucose (7.49 ± 4.3 mmol/L vs 4.53 ± 0.52 mmol/L). - Differential methylation: 11,120 DMCs (48.4% hypo-, 51.6% hypermethylated in OD) across all chromosomes; highest density at the MHC locus on chr6 (132 DMCs across 70 genes). 71% intragenic (39.6% gene body, 31.4% promoter). Most located in low CpG density regions: shore 17.6%, shelf 7.1%, open sea 58.5%, CpG islands 16.8%. - DMRs: 96 DMRs, mostly hypermethylated (74% in OD), largely overlapping CpG islands; 92 intragenic, 80 near TSS (some extending into gene body), 12 within gene bodies. Overlap between DMCs and DMRs in 54 genes (e.g., BLCAP, SLC25A24, PM20D1, PAX8, LCLAT1). - Genes with largest methylation changes included both novel T2D-related candidates (TRANK1, TEX2, SH2D3C, ATAD1, ANKEF1, MIR138-2, OR10A5, SIM1, PRRC2C, TECRL, ZDHHC14, PHTF1, C11orf66, SH3TC2, MRGPRX1, RNF212, FLJ16171) and known T2D-related genes (FSD1L, NSF, SLIT3, PTPRN2, PSMD10, MAD1L1, MIR572, ATM, LCLAT1, TNFRSF8). - Pathways enriched among DMC genes included fatty acid metabolism, aldosterone synthesis/secretion, oxytocin signaling, GABAergic and dopaminergic synapses. - Differential expression: 252 DEGs between OD and OND (55.6% upregulated in OD). - Overlap DMC–DEG: 68 DEGs harbored 88 DMCs (35 promoter, 53 gene body). Nominal pathway enrichments included PPARG and Hippo signaling, ketone body metabolism, BCAA degradation, butanoate metabolism, and fatty acid metabolism. - Correlations: 26 DMC–DEG pairs (24 genes) showed significant methylation–expression correlations. Top correlations in ATP11A, LPL, PRRX1, ABCC9, EHD2. - Trait associations: Of 88 DMCs in DMC–DEG genes, 38 (35 genes) correlated with HbA1c (p < 0.05), 11 also with fasting glucose. Among the 24 genes with DMC–DEG correlations, 16 (including ATP11A, LPL, EHD2, ACVR1C, MAP4) showed methylation associated with glucose/HbA1c. Example ATP11A: cg25043602 methylation correlated inversely with ATP11A expression (r = −0.533, p = 0.037) and positively with HbA1c (r = 0.686, p = 0.002) and glucose (r = 0.539, p = 0.031); cg16762784 similarly (expression r = −0.766, p = 0.003; HbA1c r = 0.548, p = 0.038). LPL promoter hypermethylation associated with increased HbA1c. EHD2 body methylation correlated with overexpression and with glucose and HbA1c. - Validation: In combined validation cohort (11 OND, 10 OD), 233 CpGs from discovery showed same directional differential methylation at nominal p < 0.05; enriched pathways included glutamatergic synapse, long-term depression, and Hippo signaling. - Extended multi-ethnic analysis (20 OND, 20 OD): 9,648 DMCs in 5,135 genes; 2,092 genes and 945 DMCs overlapped with discovery with consistent directionality (except cg25140607 at TFAP2A). DMC-based clustering separated OND and OD across ethnicities. LCLAT1 and GSTTP2/GSTT1 showed highest delta β (>25%). Twenty-six pathways shared with discovery (oxytocin signaling, GABAergic/glutamatergic synapses, Hippo, MAPK, circadian entrainment, aldosterone synthesis/secretion). Thirty-two DMRs occurred in the same genes as in discovery (e.g., GALR1, LCLAT1, SLC25A24, SLC1A2, GRIK2, TDRD12, MIR886, GSTO2, LRCOL1). - Clustering: Unsupervised clustering of DMCs clearly separated OD from OND; expression profiles did not discriminate groups as effectively.

The study addressed whether epigenetic alterations in VAT distinguish obese individuals who develop T2D from those who do not and whether such alterations relate to gene expression and glycemic traits. Extensive differential methylation, particularly in low CpG density regions and at the MHC locus, differentiated OD from OND and suggests immune and metabolic regulatory pathways are epigenetically perturbed in T2D with obesity. Although only a subset of methylation changes corresponded to expression changes, the identified DMC–DEG pairs mapped to biologically relevant pathways (PPARG/Hippo signaling and metabolic processes) that influence adipocyte differentiation and insulin sensitivity, supporting a functional role for specific epigenetic alterations. Key genes (ATP11A, LPL, EHD2) showed strong methylation–expression relationships and correlations with fasting glucose and HbA1c, linking epigenetic status to clinically relevant traits. The prominent involvement of the MHC locus and LCLAT1 underscores roles for immune regulation and mitochondrial lipid remodeling in T2D pathophysiology within adipose tissue. Validation and multi-ethnic analyses confirmed many methylation alterations and demonstrated that methylation profiles can classify individuals independent of ethnicity, highlighting methylation’s potential as a biomarker. Overall, findings indicate that specific epigenetic modifications in VAT contribute to T2D susceptibility in obesity and can inform risk stratification and mechanistic understanding.

This study identifies novel and known candidate genes and pathways implicated in T2D development among individuals with obesity by integrating VAT DNA methylation and gene expression with glycemic traits. Methylation profiles robustly discriminate OD from OND, outperforming expression profiles, and specific methylation–expression–trait relationships (e.g., ATP11A, LPL, EHD2) link epigenetic variation to metabolic dysfunction. Cross-cohort validation and multi-ethnic analyses support the generalizability of key methylation signals (e.g., MHC locus, LCLAT1) while also revealing population-specific marks. Future research should include larger cohorts, male participants, longitudinal designs to assess causality, and functional studies to delineate mechanisms by which identified epigenetic changes influence adipose biology and T2D risk.

- Small sample size (n=19) limited statistical power, particularly for gene expression where no DEGs survived multiple testing correction, necessitating use of uncorrected p values with an effect-size threshold. - Female-only cohort limits generalizability; authors note the need to analyze male samples. - Potential confounding by medications (metformin and insulin) despite inclusion in variation assessment; final adjustments applied to age and batch only. - Cross-sectional design precludes causal inference. - Complexity of the MHC locus hindered clear pairing of methylation and expression changes, suggesting more exhaustive locus-specific studies are needed.