Increased body mass index is linked to systemic inflammation through altered chromatin co-accessibility in human preadipocytes

Obesity is associated with increased risk of cardiometabolic diseases and low-grade systemic inflammation. While macrophages have a known role, preadipocytes are increasingly implicated in the pro-inflammatory milieu of adipose tissue, yet the mechanisms remain unclear. The authors hypothesized that an obesogenic environment alters higher-order chromatin organization in human primary preadipocytes, specifically the co-accessibility of open chromatin within active nuclear compartments, thereby impacting gene regulation and systemic inflammation. To isolate environmental effects from genetic background, the study leveraged BMI-discordant monozygotic twin pairs and profiled chromatin accessibility and gene expression in subcutaneous preadipocytes, integrating promoter capture Hi-C to map regulatory interactions.

Prior work shows chronic low-grade inflammation as a hallmark of obesity contributing to insulin resistance and atherosclerosis. Adipose macrophages are important, but preadipocytes can secrete pro-inflammatory cytokines and adopt macrophage-like expression profiles, largely demonstrated under in vitro inflammatory stimuli. The genomic organization into A (active) and B (inactive) compartments relates to gene regulation and chromatin state. Genome-wide association studies implicate polygenic influences on BMI and comorbidities, but BMI loci often enrich for brain mechanisms. Gene–environment interactions are difficult to detect at scale but may be revealed by focusing on environmentally responsive genomic regions. There is a paucity of epigenomic resources for primary human preadipocytes, motivating direct profiling to understand their role in obesity-related inflammation.

Study design and cohorts: Primary subcutaneous preadipocytes were isolated from adipose biopsies of 10 BMI-discordant monozygotic twin pairs (ΔBMI ≥ 3 kg/m²) from Finnish cohorts; 9 pairs (n=18) passed ATAC-seq QC, and all 10 pairs (n=20) were used for RNA-seq. An independent primary preadipocyte sample (European origin) provided promoter Capture Hi-C (pCHi-C) data. UK Biobank (UKB) data (up to 372,652 unrelated Europeans; CRP GWAS n≈343,524) were used for heritability partitioning and SNP×BMI interaction analyses on CRP. Cell culture and assays: Preadipocytes were cultured at passages ≤5; ATAC-seq was performed on preadipocytes and on cells after 24 h of adipogenic induction (D1). RNA-seq was performed on preadipocytes. ATAC-seq library prep followed omni-ATAC; RNA-seq used Illumina TruSeq Stranded mRNA. Sequencing on Illumina HiSeq 4000. Data processing: ATAC-seq reads aligned to 1000 Genomes v37 with Bowtie2; duplicates and low MAPQ filtered; autosomes retained. Consensus peaks (MACS2, FDR<0.05) were defined across samples, excluding ENCODE blacklist regions; peak counts normalized (BPM) and corrected for family ID (random effect), age, sex, and FRiP. RNA-seq reads aligned with STAR 2-pass; genes with ≥1 CPM in ≥10% samples retained; TPMs corrected for family ID (random effect), age, sex, and 3' bias. A/B compartment inference: ATAC-seq coverage binned into 100-kb bins genome-wide across 18 preadipocyte samples; Spearman correlation matrices computed by chromosome; first eigenvector derived (nipals) and sign oriented using correlation with bin correlation level; eigenvectors smoothed (moving average, window 3) to define A (positive) and B (negative) compartments. Validation with ChromHMM 25-state imputed annotations across ENCODE/Roadmap and external features. pCHi-C integration: Preadipocyte promoter–enhancer interactions were overlapped with A/B compartments; enrichment tested via permutation of compartment locations. Co-accessibility metric: For each A compartment, co-accessibility computed as average adjacency (thresholded Spearman correlation) with all other A compartments. Associations tested with gene expression, chromatin states, ATAC peaks, and pCHi-C measures. Twin group comparison: Co-accessibility recomputed separately in lower-BMI and higher-BMI siblings (n=9 each). Genome-wide differences assessed with one-sample Wilcoxon; per-compartment differences evaluated via within-pair BMI-label permutations (512 permutations), defining altered compartments at permutation p<0.01 (121 compartments; ~88.5 Mb). Heritability partitioning and GxE: Partitioned LDSC regression estimated CRP heritability enrichment across B, A (unaltered), and A (altered) compartments, and across A compartment clusters. UKB SNP×BMI linear interaction tests on normalized CRP contrasted p-value distributions for SNPs in altered vs. unaltered A compartments and across clusters. Clustering of A compartments: ATAC signal within A compartments was dimensionally reduced (PCA→UMAP) and nearest-neighbor graphs constructed; Louvain clustering yielded 10 clusters. Cluster characterization used ChromHMM state coverage, super-enhancer (ROSE on MED1/H3K27ac ChIP-seq from adipogenic D1 in BM-hMSC-TERT4), number/type of pCHi-C interactions, and early differentiation ATAC dynamics (PAd vs D1 using limma-voom). Functional analyses: WGCNA on A-compartment genes produced co-expression modules used for NEAT GO-slim enrichment per cluster. Module–BMI associations assessed via mixed models. KEGG enrichment (WebGestalt) conducted for cluster 1 black-module genes. Candidate gene regulatory elements were linked by correlating local ATAC peaks (±250 kb) with DE genes. GWAS enrichment for the 52 DE genes (MAGENTA) used UKB BMI GWAS summary statistics. TF motif enrichment used HOMER on cluster-specific peaks and on peaks around the 52 genes.

A/B compartment landscape and validation: In preadipocytes, A compartments are shorter (median ~300 kb) than B (median ~600 kb; Wilcoxon p≈3.08×10−76); B compartments comprise ~74.8% of the genome. A compartments show higher enhancer and promoter state coverage (Wilcoxon p=5.36×10−75 and 2.43×10−130, respectively), while B compartments are enriched for quiescent states (p≈8.34×10−95). Genes in A compartments are more highly expressed than in B (Wilcoxon p=8.85×10−46). pCHi-C interactions preferentially occur within the same compartment: 68.0% of interactions have both ends in the same compartment; after permutation, A-compartment interactions show 2.67-fold enrichment (p<1×10−10). Genes involved in pCHi-C interactions are more highly expressed only when located in A compartments (p=1.38×10−09). Co-accessibility and BMI: The A-compartment co-accessibility metric correlates with gene expression (Kruskal–Wallis p=1.27×10−15) and with active regulatory features (ChromHMM active states, ATAC, and pCHi-C). Comparing twins, A-compartment co-accessibility is significantly higher in lower-BMI siblings than in higher-BMI siblings (p=4.96×10−31), indicating genome-wide reduction in higher BMI. Permutation testing identified 121 A compartments with altered co-accessibility totaling ~88.5 Mb (BMI-responsive regions). Heritability and GxE: Partitioned LDSC for CRP shows B compartments are depleted (enrichment=0.840, p=1.99×10−09); A compartments enriched (enrichment=1.33, p=7.05×10−04); altered A compartments show stronger enrichment (enrichment=3.10, p=6.19×10−03). Across cardiometabolic traits, A compartments are enriched for heritability (enrichment=1.33–1.72; FDR<0.05) but not for BMI itself. UKB SNP×BMI interaction analysis on CRP reveals an excess of small p-values for SNPs in altered A compartments versus other A compartments (Wilcoxon p=8.72×10−05). A-compartment clustering: Ten clusters were defined; clusters 1, 2, 3, and 5 have the highest co-accessibility and active-state coverage. Cluster 5 is enhancer- and super-enhancer-rich (3.2-fold enrichment; FDR=7.8×10−04) and shows more promoter interactions per gene (mean 8 vs 3 in cluster 1; PKW≈8.85×10−46). Cluster 1 shows lower expression, more promoter–promoter interactions (PKW=1.05×10−24), GO enrichments for development, cell polarity/adhesion, and immune processes, and gains accessibility upon early differentiation, indicating a developmentally primed state. BMI-responsive regions are enriched in clusters 1 (2.65-fold; Padj=2.62×10−07) and 2 (3.47-fold; Padj=2.81×10−13). Cluster-level heritability and GxE: Cluster 1 is enriched for heritability of CRP and several obesity-related traits (BMI, TGs, ALT, glucose, systolic BP, FVC; enrichment=1.75–3.14; Padj<0.05), with no enrichment for LDL/total cholesterol. Cluster 1 also shows more SNP×BMI interaction signal for CRP than cluster 5 (post hoc p≈0.041). Candidate genes and regulatory circuits: WGCNA identified the black module enriched in cluster 1 (472/1054 genes; 44.8%; 1.3-fold; Padj=3.01×10−15; purple module also enriched: 116/228; 50.9%; 1.53-fold; Padj=2.92×10−06). Black-module PC1 associates with BMI status (Padj=5.5×10−03 overall; 3.1×10−03 in cluster 1 subset). KEGG enrichment implicates parathyroid hormone signaling. Among 380 black-module genes correlated with PC1 (FDR<0.05), 213 (56.1%) are differentially expressed (DE) between twins; 52 DE genes reside in reprogrammed (altered) A compartments. BMI GWAS enrichment around these 52 genes is significant (MAGENTA p=1.00×10−4), remaining significant versus alternative backgrounds (p<0.05). Motif analysis around these genes shows enrichment for KLF7 (implicated in adipogenesis and inflammatory signaling). At the MIR22HG/WDR81–INPPSK locus, a MIR22HG promoter ATAC peak harboring GWAS SNP rs11078597 (associated with CRP, TGs, ALT) correlates with distal INPPSK expression (but not WDR81), and inversely with MIR22HG expression; INPPSK is upregulated in higher-BMI siblings. An INPPSK exonic ATAC peak is inversely correlated with INPPSK expression and is differentially accessible between twins. The inverse correlation between the MIR22HG promoter peak and the INPPSK exonic peak present in lower-BMI twins is disrupted in higher-BMI twins (Fisher’s z p=0.027), illustrating BMI-dependent disruption of local co-accessibility consistent with altered gene regulation.

The study demonstrates that increased BMI is associated with a broad reduction in the co-accessibility of active chromatin compartments in primary human preadipocytes, indicating altered higher-order genomic coordination. These BMI-responsive A-compartment regions are enriched for CRP heritability and contain a concentration of SNP×BMI interaction effects on CRP in the UK Biobank, linking preadipocyte chromatin reprogramming to systemic inflammation. Clustering reveals that developmentally primed genomic neighborhoods (cluster 1), rather than enhancer super-enriched regions (cluster 5), disproportionately harbor BMI-responsive changes and genetic architecture for cardiometabolic traits, suggesting that poised promoter-rich regions may mediate immunomodulatory responses to obesogenic environments. The identification of 52 DE genes within reprogrammed compartments, enrichment for BMI GWAS signals, and a representative regulatory circuit at the MIR22HG–INPPSK locus provide mechanistic candidates for how BMI perturbs preadipocyte gene regulation contributing to inflammation. Overall, the findings support a model in which obesity perturbs higher-order chromatin coordination in preadipocytes, sensitizing regulatory regions to gene–environment interactions that influence systemic inflammatory markers.

This work maps A/B chromatin compartments in primary human preadipocytes and shows that increased BMI is linked to widespread disruption of co-accessibility across active genomic regions. These BMI-responsive regions contribute disproportionately to CRP heritability and harbor excess SNP×BMI interaction effects on CRP, implicating preadipocyte-origin mechanisms in obesity-related inflammation. Clustering pinpoints developmentally primed compartments as key mediators, and integrative analyses nominate 52 BMI-responsive genes, including a mechanistic example at the MIR22HG–INPPSK locus. Future research should expand sample sizes to increase power for genome-wide peak-level analyses, assess visceral adipose preadipocytes, perform in vivo validations and functional perturbations of candidate genes and regulatory elements, and refine environmental exposure measurements to dissect specific drivers within the obesogenic milieu.

Key limitations include a modest sample size (9 MZ pairs for ATAC-seq, 10 pairs for RNA-seq), which constrained power for genome-wide peak-level differential analyses and necessitated focusing on higher-order compartment metrics. BMI is a heterogeneous exposure, and twins may differ in multiple environmental factors (diet, activity, smoking, medications), limiting attribution to specific stimuli. Primary preadipocytes were cultured in vitro without ex vivo comparisons, raising uncertainties about in vivo persistence of observed changes. Analyses were performed on subcutaneous preadipocytes; contributions from visceral preadipocytes may differ. Gene–environment interaction detection remains challenging and underpowered even in large cohorts, and heritability enrichments reflect marginal effects rather than direct causality. The study largely interrogates higher-order coordination, potentially underestimating enhancer-level BMI responses.