Cellular composition and transcriptomics of subcutaneous adipose tissue linked to blood glycated haemoglobin

Abstract

Objective: Despite growing evidence, the mechanisms connecting adipose tissue (AT) function to type 2 diabetes (T2DM) remain incompletely understood. A detailed analysis of AT transcriptomes could offer valuable insights into this relationship. Here, we examined gene expression patterns in bulk subcutaneous AT, focusing on biological pathways and cellular composition associated with glycated haemoglobin (HbA1c) levels.

Methods: A transcriptomic dataset was obtained from subcutaneous AT samples of 901 adults collected during elective surgical procedures. We characterized cellular composition within subcutaneous AT in association with blood HbA1c levels by performing bulk adipose transcriptomes cell deconvolution analysis. We also conducted differential gene expression and overrepresentation analyses. We validated our cross-sectional study using two independent validation cohorts, performing further downstream analyses.

Results: Subcutaneous AT from subjects with increased HbA1c had lower adipocytes, smooth muscle, pericytes and other endothelial cell numbers. Pathways associated with HbA1c levels included cellular senescence and telomere-related pathways and extracellular matrix organisation. We identified the expression of RHO GTPases associated with HbA1c not previously linked to glucose homeostasis, with a possible sexual dimorphism shaped by the obesity state. The findings were confirmed in both longitudinal cohorts. At the gene level, HLA-DR, CCL13, and S100A4 mRNA levels were strongly correlated with HbA1c levels.

Conclusions: This study underscores the utility of AT transcriptome analysis in unravelling T2DM complexities. Our findings enhance knowledge of glucose homeostasis' molecular and cellular underpinnings, paving the way for potential therapeutic targets to mitigate the impact of AT dysfunction in metabolic diseases.

Keywords: Rho GTPases; glycated haemoglobin; immune system; subcutaneous adipose tissue; type 2 diabetes mellitus.

FIGURE 1

(A) Dotplot of significant Reactome pathways that are differentially expressed (q‐value < .01) in SAT samples of subjects with HbA1c levels above and below 5.7% (n = 740). Significant pathways resulted from an over‐representation analysis (ORA) of genes differentially expressed (DEGs, pSGoF < .05) in HbA1c levels above and below 5.7% while controlling for age, sex, BMI and sample origin, identified from the SAT RNA‐Seq data in the ADIPOMIT exploratory cohort. The ORA was performed with ConsensusPathDB, considering the Reactome database. Pathways are shown in the y‐axis; Counts refer to the number of significantly DEGs that belong to the given pathway (or gene‐set); Gene Ratio (x‐axis) is calculated as count/set size (being set size the number of genes in a given pathway). Dots are coloured according to pathways' q‐value. Pathways related to the immune system, inflammation and RHO GTPases are highlighted in colour. (B) Emapplot of significant pathways that are differentially expressed (Reactome database, q‐value < .05) in SAT samples of subjects with HbA1c levels above and below 5.7%. Significant pathways resulted from an ORA of DEGs (pSGoF < .05) in HbA1c levels above and below 5.7%. Dots are coloured according to q‐value; dot size represents the number of genes involved in each pathway; nodes link pathways with shared significant genes, with a minimum percentage of overlap genes of .2.

FIGURE 2

(A) Dotplot of Reactome pathways that are significantly associated (q‐value < .01) with HbA1c levels in women's SAT samples (n = 531). Significant pathways resulted from an ORA of DEGs (pSGoF < .05) significantly associated with HbA1c levels while controlling for age, BMI and sample origin, identified from the women's SAT RNA‐Seq data in the ADIPOMIT exploratory cohort. Pathways related to the immune system, inflammation and RHO GTPases are highlighted in colour. (B) Dotplot of Reactome pathways that are significantly associated (q‐value < .05) with HbA1c in men's SAT samples (n = 209). Significant pathways resulted from an ORA of DEGs (pSGoF < .05) significantly associated with HbA1c levels while controlling for age, BMI and sample origin, which were identified from the men's SAT RNA‐Seq data in the ADIPOMIT exploratory cohort. Pathways related to the immune system and inflammation are highlighted in colour.

FIGURE 3

(A) Dotplot of Reactome pathways that are significantly associated (q‐value < .001) with HbA1c levels in SAT samples from patients with obesity (n = 685). Significant pathways resulted from an ORA of DEGs (pSGoF < .05) significantly associated with HbA1c levels while controlling for age, sex, BMI and sample origin, identified from the SAT RNA‐Seq data. Pathways related to the immune system, inflammation and RHO GTPases are highlighted in colour. (B) Dotplot of Reactome pathways that are significantly associated (q‐value < .05) with HbA1c levels in SAT samples from patients without obesity (n = 55). Significant pathways resulted from an ORA of DEGs (pSGoF < .05) significantly associated with HbA1c levels while controlling for age, sex, BMI and sample origin, identified from the SAT RNA‐Seq data. Pathways related to the immune system and inflammation are highlighted in colour. (C) Dotplot of Reactome pathways that are significantly associated (q‐value < .05) with serum glucose levels in SAT samples from patients without obesity (n = 57). Significant pathways resulted from an ORA of DEGs (pSGoF < .05) significantly associated with serum glucose levels while controlling for age, sex and BMI, identified from the SAT RNA‐Seq data. Pathways related to the immune system and inflammation are highlighted in colour.

FIGURE 4

(A) Histogram showing subjects' count and blood HbA1c (%) levels, including Student's t‐test assessing differences between the high HbA1c group (above 5.7%) and the low (below 5.7%). (B) Spearman's correlation between BMI (kg/m²) and HbA1c (%). (C) HbA1c (%) versus GSVA scores regression analysis (β), after correcting for age, sex and BMI. (D) UMAP of cell type composition. Clustering indicates similarities in expression patterns based on cell deconvolution analysis. (E) Volcano plot of cell types from deconvolution analysis, showing fold change (FC) and p‐values comparing high versus low HbA1c levels.

FIGURE 5

(A) Validation cohort 1 study design. Women with obesity were recruited and SAT samples were obtained during bariatric surgery after informed consent. SAT was again extracted around 2 years after from the same patients. N = 30 from 16 patients with pre‐ and post‐BS data. (B–F) Boxplots showing change in HbA1c levels (%) (B), RAC2 (C), ARHGAP22 (D), ARHGAP30 (E) and ARHGAP4 expression levels as normalized intensities (F) between pre‐ and post‐bariatric surgery states. (G) Dotplot of significant Reactome pathways that are differentially expressed (q‐value<EXP‐05; 33 most significant pathways) in SAT samples taken before and after bariatric surgery from women within the validation cohort 1 (n = 16). Significant pathways resulted from an over‐representation analysis (ORA) of genes differentially expressed (DEGs, pSGoF < .05). No covariables were considered. The ORA was performed with ConsensusPathDB, considering the Reactome database.