Integrated miRNA–mRNA Analysis Reveals Obesity‐Driven Regulatory Networks in Human Visceral Adipose Tissue With and Without Type 2 Diabetes

ABSTRACT

Aims

Obesity is characterised by pathological alterations in visceral white adipose tissue (vWAT) that may contribute to the development of type 2 diabetes (T2D). While microRNAs (miRNAs) are key post‐transcriptional regulators, comprehensive human vWAT profiling across metabolic states remains limited. This study characterised vWAT miRNA expression in lean, obese and obese+T2D individuals to identify obesity‐driven regulatory networks associated with metabolic dysfunction.

Methods

Deep miRNA sequencing was performed on vWAT samples from a discovery cohort comprising lean controls and individuals with obesity (with and without T2D). Findings were validated via RT‐qPCR in an independent replication cohort. Differentially expressed miRNAs were bioinformatically integrated with matched mRNA transcriptomic data to construct putative functional regulatory associations and identify enriched pathways underlying metabolic impairment.

Results

The dominant transcriptomic signal was driven by obesity rather than T2D status, with substantial overlap between obese subgroups in principal component analyses. miR‐141‐3p, miR‐200b‐3p, miR‐12 136 and miR‐585‐3p showed consistent differential expression associated with obesity. miR‐141‐3p and miR‐200b‐3p were upregulated and inversely associated with metabolic stress–related genes, including TF and FBXO32. Integrated miRNA–mRNA analyses revealed putative regulatory associations involving inflammation, lipid metabolism, insulin signalling and iron homeostasis. These associations were robust across progressive covariate adjustment models for age and sex.

Conclusions

This study provides a comprehensive characterisation of the vWAT miRNA landscape predominantly shaped by obesity, with T2D contributing comparatively subtle additional variation. We identified putative miRNA–mRNA regulatory associations that may contribute to pathological adipose tissue dysfunction. These findings highlight candidate molecular regulators worthy of further functional investigation in the context of obesity and T2D.

1. Introduction

Obesity is a chronic disease with numerous metabolic, physical and psychosocial complications, including substantially increased risk for type 2 diabetes (T2D) [1] and cardiovascular disease [2]. Its prevalence continues rising globally [3], with projections suggesting 50% of people could be overweight or obese by 2030, with around 20% being obese [4].

White adipose tissue (WAT), once regarded merely as energy storage [5], is now recognised for its metabolic and endocrine functions [6, 7]. WAT is divided into visceral (vWAT) and subcutaneous (sWAT) adipose tissue [3, 8]. Accumulation of vWAT promotes inflammation and comorbidities, while sWAT loses its protective characteristics against insulin resistance during obesity [9, 10]. Studies show T2D patients have lower sWAT and higher vWAT levels, highlighting their distinct roles in obesity‐related diseases.

Obesity‐derived physiological changes produce modifications in molecular parameters [11], prompting research into genetics and epigenetics of obesity [12, 13]. MicroRNAs (miRNAs), non‐coding RNAs regulating post‐transcriptional gene expression [14], play essential roles in metabolism regulation [15], particularly in muscle and adipose tissue function [16, 17]. Recent evidence highlights miRNAs' central role in lipid metabolism [18]: miR‐33 and miR‐122 regulate fatty acid and cholesterol metabolism, miR‐34a affects obesity resistance and insulin sensitivity, while the miR‐30 family influences lipoprotein metabolism and LDL levels.

Ortega et al. [19] profiled nearly 800 miRNAs during adipogenesis, finding 11 significantly deregulated in subcutaneous fat from obese subjects with and without T2D. However, this study involved a small cohort over a decade ago. Despite some therapeutic target studies [20], the field remains cancer‐dominated, with most obesity‐related miRNA data from in vitro models, plasma samples, or small cohorts focusing on limited miRNAs. Comprehensive, large‐scale analyses of miRNA expression in human adipose tissue remain scarce.

Therefore, this project aims to explore adipose tissue miRNA expression profiles in depth to uncover specific miRNAs that could act as biomarkers explaining the onset and progression of metabolic diseases.

2. Methods

2.1. Study Population and Sample Acquisition

vWAT samples from obese and non‐obese volunteers were obtained during bariatric and hernia surgeries from the left upper quadrant. Patients were subdivided into a discovery cohort of 48 participants (10 controls, 19 with obesity, 19 with obesity and T2D) and a validation cohort of 100 participants (19 controls, 51 with obesity, 30 with obesity and T2D). Routine biochemical analysis was performed over 3 months prior to intervention. Eligibility criteria included age between 18 and 70 years and clinical stability at the time of recruitment. Exclusion criteria included known active inflammatory or autoimmune conditions, severe hepatic or renal impairment, active malignancy and ongoing immunosuppressive or corticosteroid treatments that could directly confound adipose tissue gene expression. All participants were ambulatory patients; no bedridden individuals were included. Body weight and height were measured under standardised conditions prior to surgery and BMI was calculated accordingly. All patients scheduled for bariatric surgery underwent a standardised preoperative very low‐calorie diet (VLCD) for 1–4 weeks prior to the intervention, with duration adjusted according to initial BMI, as per standard clinical practice. Written informed consent was obtained from all subjects and the study was approved by the Central University Hospital of Asturias Ethical Committee (CEImPA 2020.419). vWAT was collected under sterile conditions and frozen at −80°C immediately until use.

2.2. RNA Extraction, Library Preparation and Next‐Generation Sequencing

2.2.1. RNA Extraction and Quality Control

Total RNA was extracted using the SPLIT RNA Extraction Kit (Lexogen, Vienna, Austria). RNA integrity and concentration were assessed using an Agilent 2100 Bioanalyzer. Only samples with sufficient yield and RQN values were included.

2.2.2. Small RNA Library Preparation and Sequencing

Two synthetic spike‐in RNAs (cel‐miR‐39‐3p, ath‐miR‐159a) were added as internal controls. Small RNA libraries were prepared from 6 μL of total RNA using the Small RNA‐Seq Library Prep Kit (Lexogen). Adapter dimers were excluded by size selection (130–200 bp) using the Pippin Prep system. Libraries were grouped into pools, quantified using the KAPA Library Quantification Kit (Roche) and sequenced on an Illumina HiSeq X Ten with paired‐end 150 bp reads.

2.2.3. mRNA Library Preparation and Sequencing

Poly(A) + RNA was enriched using the Lexogen Poly(A) Selection Module. Libraries were prepared with the CORALL mRNA‐Seq Kit incorporating UMIs and dual indexes. Libraries were normalised, pooled at 80 nM and sequenced on an Illumina NovaSeq 6000 with paired‐end 150 bp reads.

2.2.4. Bioinformatic Analysis of miRNA Data

Raw reads were quality‐controlled and adapter‐trimmed using Trim Galore! [21], then aligned to the human reference genome (GRCh38) using STAR [22]. Quantification was performed with featureCounts [23] using miRBase v22.1. miRNAs with base mean count > 5 were retained. Differential expression analysis was performed using DESeq2 [24], applying negative binomial distribution modelling and empirical Bayes shrinkage. Differentially expressed miRNAs were defined as those meeting p‐value < 0.05 and |log2FC| ≥ 1.5 across all comparisons. The DESeq2 design formula included only the group variable as the primary factor; age and sex were not included as covariates in this discovery analysis, which is acknowledged as a limitation and addressed through progressive covariate adjustment in the independent validation cohort.

All analyses were performed in a Linux‐based environment using the following software versions: STAR v2.7.3, Rsubread v2.6.4, DESeq2 v1.32.0, SAMtools v0.19.4, Bedtools v2.24.0, R v4.1.2 and Bioconductor v3.14.

2.2.5. Principal Component Analysis (PCA)

Principal component analysis (PCA) was performed to explore global expression patterns and visualise the separation among study groups based on miRNA and mRNA expression profiles. PCA was conducted using the prcomp function in R, based on centred and scaled expression values. The first two principal components (PC1 and PC2), explaining the largest proportion of total variance, were plotted to visualise clustering patterns among the study groups (Control, OBnoT2D and OBT2D). Confidence ellipses (95%) were drawn using the ggplot2 package to aid visual interpretation of group distributions.

2.2.6. Bioinformatic Analysis of mRNA Data

A differential gene expression (DGE) analysis was performed to compare individuals with and without obesity. This analytical focus was not part of the original pre‐specified plan but was guided by two convergent data‐driven observations: the clustering pattern observed in the PCA of mRNA expression profiles and the limited number of miRNAs exclusively differentially expressed in the OB_T2D vs. OB_noT2D comparison. Accordingly, the mRNA analysis was focused on the obesity vs. non‐obesity contrast as the primary comparison of interest. Gene‐level counts were normalised and processed using the limma–voom pipeline in R. Linear models were fitted adjusting for age and sex and complete‐case analyses were used. p‐values were corrected for multiple testing using the Benjamini–Hochberg false discovery rate (FDR). Quality control included library size assessment and multidimensional scaling (MDS) plots. Volcano plots were generated to visualise differentially expressed genes.

2.3. Putative Gene Targets, Pathway Enrichment Analysis and Related Diseases of the Differentially Expressed miRNAs

Target network, pathway enrichment analysis and related diseases were all performed using miRNet 2.0 [25] web‐based tool. miRbase IDs for the 10 modified miRNAs were uploaded and miRTarBase v9.0 database was selected to annotate experimentally supported target miRNA genes. For functional evaluation, enrichment analysis was conducted using the Kyoto Encyclopaedia of Genes and Genomes (KEGG). KEGG pathway categories annotated specifically as neoplastic diseases were excluded (e.g., pathways in cancer, organ‐specific cancers), while oncogenic signalling pathways with established roles in metabolic regulation (including MAPK, mTOR, Wnt, TGF‐β and p53 signalling) were retained.

2.4. miRNA Isolation and Quantification in the Validation Cohort

The expression levels from the 10 differentially expressed miRNAs (hsa‐miR‐146b‐3p, hsa‐miR‐342‐3p, hsa‐miR‐141‐3p, hsa‐miR‐100‐3p, hsa‐miR‐200b‐3p, hsa‐miR‐12136, hsa‐miR‐941, hsa‐miR‐369‐3p, hsa‐miR‐585‐3p and hsa‐let‐7g‐3p) were assessed in visceral adipose tissue samples collected from all validation cohort participants, as previously described [26]. Briefly, total RNA was extracted from 50–100 mg of tissue using TRIzol reagent (Invitrogen) and subsequently converted into cDNA using the TaqMan Advanced miRNA cDNA Synthesis Kit (Life Technologies, California, USA). Quantitative gene expression analysis was performed via RT‐PCR using TaqMan assays on the Applied Biosystems Prism 7900HT platform. hsa‐miR‐191‐5p served as the endogenous control and relative miRNA expression was calculated using the 2^−ΔCt method. The selection of the housekeeping miRNA was based on TaqMan technical documentation, NormFinder stability analysis using adipose tissue NGS data and validation from external sources [26, 27]. Ct values of hsa‐miR‐191‐5p showed no statistically significant differences across experimental groups in the discovery cohort (Kruskal–Wallis, p = 0.289; Figure S1; Table S1), confirming the stability of this endogenous control across metabolic states.

2.5. miRNA–mRNA Regulatory Network

To assess miRNA‐mRNA interactions, predicted mRNA targets from the differentially expressed miRNAs were extracted from the web‐based software miRNet [25] and miRDB [28, 29] and crossed with differentially expressed mRNAs from the NGS analysis. Venny diagrams were created for each miRNA target gene cluster. Specifically, target genes from up‐regulated miRNAs were crossed separately with both up‐ and down‐regulated mRNAs. Likewise, target genes from down‐regulated miRNAs were crossed separately with both up‐ and down‐regulated mRNAs. The miRNA‐mRNA interaction network was represented graphically with Cytoscape (v. 3.10.3). Interaction validation was performed by calculating the linear correlation between the expression of miRNAs and mRNAs.

2.6. Statistical Analysis

Statistical differences between the control group and the other two groups for demographical and biochemical characteristics were analysed using the Kruskal–Wallis test, followed by Dunn's post hoc, followed by Holm‐Bonferroni correction. Group differences in miRNA expression were initially assessed using the Kruskal–Wallis test, with post hoc pairwise comparisons performed using Dunn's test with Benjamini–Hochberg (BH) correction for multiple testing. Given the significant differences in age (Kruskal–Wallis, p < 0.001) and sex distribution (Fisher's exact test, p < 0.001) across groups, we subsequently performed linear regression models with progressive covariate adjustment to assess the robustness of the observed associations. Four models were fitted for each miRNA: M1 (unadjusted), M2 (adjusted for age), M3 (adjusted for sex) and M4 (adjusted for age and sex). For models M2–M4, all pairwise group comparisons were extracted using estimated marginal means (emmeans package, R), with BH correction applied across the three pairwise contrasts per miRNA. BMI was not included as a model covariate, as all obese participants met the inclusion criterion for obesity (BMI ≥ 30 kg/m²) by definition; residual BMI variation within the obese groups may nonetheless contribute to within‐group variability.

A formal a priori power calculation was not performed, as sample size was determined by the availability of surgical visceral adipose tissue specimens. Post hoc statistical power was estimated for the four miRNAs showing significant differential expression using the pwr package in R (α = 0.05), based on the partial f ² effect size derived from the fully adjusted model (M4), defined as the incremental R ² attributable to group membership after controlling for age and sex relative to the residual variance of the full model. Additionally, Spearman's correlation was performed to assess the relationship between the expression of different microRNAs and clinical parameters.

3. Results

3.1. Demographic and Biochemical Profile of Participants

Anthropometric and clinical profiles of patients included in both the discovery and the validation cohorts are presented in Table 1. As expected, significant changes were observed in BMI, glucose levels and HbA1c in both cohorts (p < 0.001). Additionally, obese patients with T2D were slightly older than those without T2D, which can be explained by less strict criteria for surgery in this group of patients. Finally, TG levels were also higher in the T2D group compared with the normoglycemic groups.

TABLE 1.

Demographical and biochemical characteristics of study cohorts.

	Discovery cohort				Validation cohort
	noOB_noT2D	OB_noT2D	OB_T2D		noOB_noT2D	OB_noT2D	OB_T2D
N (% male)	10 (70%)	19 (32%)	19 (47%)		19 (53%)	51 (14%)	30 (47%)
Age (years)	51.5 ± 6.6	45.05 ± 9.2	51.9 ± 6.6	*	58.3 ± 13	45.6 ± 9.7	52.9 ± 8.4	***
BMI (Kg/m2)	25.6 ± 2.1	46.0 ± 4.6	44.7 ± 6.1	***	25.2 ± 4	46.6 ± 4.9	46.5 ± 7.3	***
Glucose (mg/dL)	96.9 ± 11.8	114.9 ± 29.7	151.8 ± 46.8	***	103.8 ± 18.8	104.3 ± 20.6	137.5 ± 48.7	**
HbA1c (%)	5.0 ± 0.1	5.3 ± 0.4	6.2 ± 0.6	***	5.4 ± 0.5	5.4 ± 0.3	6.9 ± 1.3	***
Total cholesterol (mg/dL)	200.5 ± 36.2	167.0 ± 43.3	157.1 ± 44.9		185.8 ± 42.8	164.6 ± 37.3	159 ± 44.3
TG (mg/dL)	133.0 ± 69.1	99.7 ± 35.1	147.0 ± 61.9		157.0 ± 75.7	112.5 ± 50.2	157 ± 81.8	**

Note: Data expressed as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.

Abbreviations: BMI, body mass index; HbA1c, glycated haemoglobin; noOB‐noT2D, non‐obese non‐Type 2 diabetes individuals; OB‐noT2D, obese non‐Type 2 diabetes individuals; OB‐T2D, obese Type 2 diabetes individuals; TG, triglycerides.

3.2. Differential microRNA Profile in Visceral Adipose Tissue Associated With Obesity and Type 2 Diabetes

NGS analysis identified a total of 287 miRNAs in the discovery cohort. Differential expression was considered significant for miRNAs with a log2 fold change ≥ 1.5 or ≤ −1.5 and p‐value < 0.05. Thirteen miRNAs showed significant change between the group of individuals with obesity and controls (Figure 1A). By contrast, only 10 miRNAs showed significant differential expression in at least one of the comparisons (Figure 1B; Table S2). Additionally, the principal component analysis (PCA) of miRNA expression revealed partial segregation among the study groups (Figure 1C). Control subjects clustered distinctly from participants with obesity and/or T2D, indicating differential miRNA expression patterns associated with metabolic status. However, the two groups of obese individuals, those with and without T2D, showed overlapping distributions. This suggests that obesity exerts a dominant influence on the miRNA profile, while diabetes contributes additional, albeit less pronounced, variability. The first two principal components explained 57.6% of the total variance, demonstrating that these axes capture a substantial proportion of expression variability.

FIGURE 1.

(A) Volcano plot showing differential expression of miRNAs between lean subjects and obese individuals. Red dots represent significantly up‐regulated miRNAs, and green dots represent significantly down‐regulated miRNAs (B) Venn diagram illustrating the number of differentially expressed miRNAs in three comparisons: Ob_T2D vs. Control, Ob_NoT2D vs. Control, and Ob_T2D vs. Ob_NoT2D. (C) Principal Component Analysis (PCA) of miRNA profiles across Control, Ob_T2D, and Ob_NoT2D groups. Each dot represents an individual sample, and ellipses denote the 95% confidence interval for each group. [noOB_noT2D: People without obesity and without type 2 diabetes; OB_noT2D: People with obesity and without type 2 diabetes].

Technical validation was performed by RT‐PCR in the same cohort. Among the selected candidates, hsa‐miR‐100‐3p was not detectable by this method. hsa‐miR‐146b‐3p, hsa‐miR‐342‐3p, hsa‐miR‐369‐3p and hsa‐miR‐941 could not be validated. hsa‐miR‐12136 showed a trend towards differential expression, while hsa‐miR‐141‐3p, hsa‐let‐7g‐3p, hsa‐miR‐200b‐3p and hsa‐miR‐585‐3p exhibited the expected significant changes.

Then, the miRNAs that were either previously validated or showed a suggestive trend were quantified by RT‐PCR in an independent cohort. Unadjusted analysis (Figure S2, Tables S3 and S4) identified significant group differences in several miRNAs, including hsa‐miR‐12 136, hsa‐miR‐585‐3p, hsa‐miR‐141‐3p, hsa‐miR‐200b‐3p. However, given the significant imbalance in age and sex distribution across groups, we applied progressive covariate adjustment to evaluate whether these associations were independent of demographic confounders. We fitted progressive covariate adjustment models adjusting for age (Figure S3), sex (Figure S4) and full adjustment considering both age and sex (Figure 2). While miR‐200b‐3p and miR‐141‐3p only remained significant in Control versus OBnoT2D, miR‐12136 and miR‐585‐3p remained significant in both groups with Obesity vs. Control, independently of T2D status (Table S4). Moreover, while for miR‐200b‐3p and miR‐141‐3p age and sex were confounding differences between groups, in miR‐12136 and miR‐585‐3p these covariates revealed stronger associations. Additionally, post hoc power ranged from 0.593 to 0.997 across the four significant miRNAs (Table S5).

FIGURE 2.

Violin plots showing relative miRNA expression in visceral adipose tissue across metabolic groups, adjusted for age and sex (linear model with estimated marginal means, emmeans package; Benjamini‐Hochberg correction applied across pairwise contrasts). Validation cohort (n = 100: 19 controls, 51 OB_noT2D, 30 OB_T2D). (A) hsa‐miR‐200b‐3p; (B) hsa‐miR‐141‐3p; (C) hsa‐miR‐12 136; (D) hsa‐miR‐585‐3p. Expression is shown as −2^ΔCt, normalised to hsa‐miR‐191‐5p. *p < 0.05, **p < 0.01, ***p < 0.001; ns: Not significant. [Control: Individuals without obesity and without type 2 diabetes; OB_noT2D: Individuals with obesity and without type 2 diabetes; OB_T2D: Individuals with obesity and type 2 diabetes].

Then, correlations between miRNA expression and the different biochemical and demographic variables (Figure 3) were performed. Both hsa‐miR‐200b‐3p and hsa‐miR‐141‐3p positively correlate with BMI; furthermore, hsa‐miR‐200b‐3p also negatively correlates with HDL cholesterol. Additionally, hsa‐miR‐12136 positively correlates with Triglycerides. These correlations were assessed using Spearman's rank correlation test. Given the exploratory nature of these analyses, no correction for multiple testing was applied; results should be interpreted as hypothesis‐generating.

FIGURE 3.

Spearman matrix correlation plot between miRNAs expression in the visceral adipose tissue and biochemical variables in the validation cohort (n = 100). Correlation coefficients (rho) are shown within each cell; no correction for multiple testing was applied given the exploratory nature of the analysis. Significant correlations (p < 0.05) are indicated. *p < 0.05, **p < 0.01, ***p < 0.001.

3.3. Gene Target Prediction and Functional Enrichment Analysis

To explore the potential biological functions and pathways regulated by the differentially expressed miRNAs, we predicted which genes are affected by them. As shown in Figure 4A, multiple genes were shared by the different miRNAs. Then, we used the miRNet website to analyse the Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathways and after excluding cancer‐related pathways, 43 pathways showed significant changes. Top affected pathways are shown in Figure 4B, highlighting the insulin signalling pathway and the Type 2 diabetes pathway.

FIGURE 4.

(A) miRNA interaction network with predicted target genes. Target genes were predicted using miRNet and miRDB databases and intersected with differentially expressed mRNAs identified by mRNA sequencing. Network constructed using discovery cohort data (n = 48). (B) Pathway enrichment analysis was performed using miRNet; pathways annotated specifically as neoplastic diseases were excluded, while oncogenic signalling pathways with established roles in metabolic regulation (including MAPK, mTOR, Wnt, TGF‐β, and p53 signalling) were retained. Top 20 significantly affected pathways are shown.

3.4. Differential mRNA Profile in Visceral Adipose Tissue Associated With Obesity

To evaluate real interactions between miRNAs and gene expression, mRNA‐seq was also performed in the same discovery cohort. The PCA showed moderate separation among the three metabolic groups. Though some overlap persisted, samples from individuals with obesity and with T2D tended to cluster together, reflecting transcriptomic signatures associated with metabolic dysfunction. The first principal component accounted for 46.8% of the total variation, suggesting that differences in mRNA expression are largely driven by metabolic alterations rather than by individual variability. These two observations, the overlapping clustering of both obese groups in the PCA and the finding that non miRNA was differentially expressed in the OB_T2D vs. OB_noT2D comparison, informed the decision to focus the mRNA analysis on the obesity vs. non‐obesity contrast rather than on a T2D‐specific comparison. Since only one exclusive miRNA was found in the OB_T2D vs. OB_noT2D comparison and due to the similar grouping shown in the PCA representation (Figure 5A), mRNA analysis focused exclusively on the changes associated with obesity (obese vs. non‐obese). Finally, 23 205 genes were identified by NGS analysis in the discovery cohort (Figure 5B). Considering differential expression, a log₂ fold change ≥ 1.0 or ≤ −1.0 and an adjusted p‐value < 0.05, 63 genes showed a differential expression, being 51 downregulated and 12 upregulated in the group of people with obesity (Table 2).

FIGURE 5.

(A) Principal Component Analysis (PCA) of mRNA expression profiles in visceral adipose tissue samples across three metabolic groups [discovery cohort, n = 47: 10 controls, 19 OB_noT2D, 18 OB_T2D; one sample excluded due to quality control]. Each point represents an individual sample and ellipses indicate the 95% confidence interval for each group. (B) Volcano plot showing differential gene expression between obese (n = 37) and non‐obese (n = 10) individuals. Differential expression was assessed using limma‐voom. Red dots indicate significantly dysregulated genes (adjusted p < 0.05) and green dots represent non‐significant genes. [Control: Individuals without obesity and without type 2 diabetes; OB_noT2D: Individuals with obesity and without type 2 diabetes; OB_T2D: Individuals with obesity and type 2 diabetes].

TABLE 2.

Differentially expressed mRNAs detected by NGS in the adipose tissue of people with obesity compared to lean controls.

Ensembl gene id	logFC	adj. p	chr	Gene symbol
ENSG00000221288.1	−2.8751	0.0048	chr2	MIR663B
ENSG00000270956.1	−2.1908	0.0493	chr2	RP11‐65L3.4
ENSG00000264384.2	−2.1074	0.0254	chr1	RN7SL431P
ENSG00000178821.13	−1.9700	0.0366	chr1	TMEM52
ENSG00000277543.1	−1.9179	0.0465	chr16	RP11‐452L6.8
ENSG00000108878.5	−1.9049	0.0126	chr17	CACNG1
ENSG00000265142.9	−1.7912	0.0479	chr18	MIR133A1HG
ENSG00000161896.12	−1.7163	0.0473	chr6	IP6K3
ENSG00000250105.1	−1.6736	0.0392	chr11	CTD‐3074O7.2
ENSG00000181856.15	−1.5933	0.0011	chr17	SLC2A4
ENSG00000164591.14	−1.5531	0.0493	chr5	MYOZ3
ENSG00000185028.4	−1.5211	0.0479	chr5	LRRC14B
ENSG00000204460.3	−1.5149	0.0479	chr2	LINC01854
ENSG00000130222.11	−1.5144	0.0322	chr9	GADD45G
ENSG00000198881.10	−1.5010	0.0384	chrX	ASB12
ENSG00000112183.15	−1.4419	0.0473	chr6	RBM24
ENSG00000274904.1	−1.4162	0.0443	chr16	CTD‐2515O10.5
ENSG00000087085.16	−1.3824	0.0480	chr7	ACHE
ENSG00000231739.2	−1.3812	0.0342	chr2	GAPDHP59
ENSG00000167549.19	−1.3721	0.0487	chr17	CORO6
ENSG00000233125.3	−1.3217	0.0202	chr1	ACTBP12
ENSG00000147573.17	−1.2882	0.0096	chr8	TRIM55
ENSG00000101892.12	−1.2877	0.0461	chrX	ATP1B4
ENSG00000178596.10	−1.2812	0.0403	chr1	GAPDHP29
ENSG00000214563.2	−1.2801	0.0351	chr6	GAPDHP15
ENSG00000126778.12	−1.2778	0.0191	chr14	SIX1
ENSG00000248626.1	−1.2749	0.0482	chr5	GAPDHP40
ENSG00000229001.1	−1.2444	0.0461	chr10	ACTBP14
ENSG00000166816.15	−1.2421	0.0021	chr16	LDHD
ENSG00000091513.16	−1.2301	0.0021	chr3	TF
ENSG00000140284.11	−1.2276	0.0017	chr15	SLC27A2
ENSG00000280114.1	−1.2224	0.0352	chr1	RP5‐875O13.7
ENSG00000286143.1	−1.1942	0.0091	chr2	RP11‐316O14.3
ENSG00000231072.1	−1.1816	0.0369	chr1	GAPDHP64
ENSG00000164434.12	−1.1539	0.0430	chr6	FABP7
ENSG00000156804.7	−1.1492	0.0322	chr8	FBXO32
ENSG00000234700.1	−1.1423	0.0361	chr7	MTCO1P8
ENSG00000037965.6	−1.1391	0.0101	chr12	HOXC8
ENSG00000181016.9	−1.1246	0.0473	chr7	LSMEM1
ENSG00000124374.9	−1.1111	0.0427	chr2	PAIP2B
ENSG00000180818.5	−1.1059	0.0468	chr12	HOXC10
ENSG00000198892.7	−1.1053	0.0230	chr1	SHISA4
ENSG00000213171.3	−1.0975	0.0456	chr1	LINGO4
ENSG00000163050.18	−1.0904	0.0064	chr1	COQ8A
ENSG00000249930.1	−1.0856	0.0322	chr4	AC007016.3
ENSG00000235162.9	−1.0775	0.0365	chr12	C12orf75
ENSG00000233764.1	−1.0692	0.0151	chr22	MTCO1P20
ENSG00000239873.2	−1.0413	0.0478	chr1	GAPDHP27
ENSG00000160862.13	−1.0277	0.0006	chr7	AZGP1
ENSG00000269883.1	−1.0048	0.0289	chr14	RP11‐463C8.7
ENSG00000236211.1	−1.0017	0.0350	chr2	MTCO1P7
ENSG00000155659.15	1.0046	0.0470	chrX	VSIG4
ENSG00000102359.9	1.0593	0.0091	chrX	SRPX2
ENSG00000272149.1	1.0864	0.0032	chr3	RP11‐627J17.1
ENSG00000154277.13	1.1083	0.0066	chr4	UCHL1
ENSG00000173267.14	1.1481	0.0016	chr10	SNCG
ENSG00000198759.12	1.1756	0.0453	chrX	EGFL6
ENSG00000251548.1	1.1851	0.0046	chr5	RP11‐215G15.4
ENSG00000181019.13	1.2143	0.0070	chr16	NQO1
ENSG00000174697.5	1.2206	0.0011	chr7	LEP
ENSG00000277587.1	1.2279	0.0413	chr19	CTD‐3116E22.8
ENSG00000261602.1	1.3439	0.0122	chr16	CTD‐2033A16.1
ENSG00000000005.6	1.7466	0.0024	chrX	TNMD

3.5. Visualisation of the miRNA–mRNA Regulatory Network

Venny comparisons were performed between the target genes of each differentially expressed miRNAs and the differentially expressed mRNA from the NGS analysis. Target genes from up‐regulated miRNAs were crossed with both down‐regulated mRNAs (Figure S5A–G) and up‐regulated mRNAs (Figure S5H–N). Additionally, target genes from down‐regulated miRNAs were crossed with both up‐regulated mRNAs (Figure S6A–D) and down‐regulated mRNAs (Figure S6E–H). A miRNA‐mRNA network was also performed (Figure 6A) and correlations were analysed. No correction for multiple testing was applied to these exploratory correlation analyses; accordingly, borderline p‐values should be interpreted with caution. Specifically, hsa‐miR‐200b‐3p was inversely correlated with FBXO32 (p‐value = 0.035, Sp_rho = −0.273; Figure 6B), hsa‐miR‐141‐3p was inversely correlated with TF and NQO1 (p‐value = 0.025, Sp_rho = −0.293 and p‐value = 0.033, Sp_rho = −0.280 respectively; Figure 6C,D), hsa‐miR‐342‐3p was inversely correlated with NQO1 and UCHL1 (p‐value = 0.023, Sp_rho = −0.299 and p‐value = 0.034, Sp_rho = −0.274 respectively; Figure 6E,F) and hsa‐let‐7g‐3p inversely correlated with RBM24 and COQ8A (p‐value = 0.040, Sp_rho = −0.258 and p‐value = 0.027, Sp_rho = −0.283 respectively; Figure 6G,H).

FIGURE 6.

(A) Predicted miRNA–mRNA interaction network. Nodes represent miRNAs (triangles) and mRNAs (hexagons). Orange indicates downregulated molecules and blue indicates upregulated molecules in obesity. Edges denote predicted regulatory interactions. Target predictions were derived from miRNet and miRDB databases and intersected with differentially expressed mRNAs from the discovery cohort (n = 48). Only overlapping interactions are shown. (B–H) Scatter plots showing significant putative correlations between selected miRNAs and the expression of their predicted mRNA targets. Correlations were assessed using Spearman's rank test; no correction for multiple testing was applied given the exploratory nature of the analysis. Spearman's rho and p‐values are indicated within each panel. Expression values are shown as −2^ΔCt, normalised to hsa‐miR‐191‐5p (miRNA) and as normalised counts (mRNA).

4. Discussion

This study aimed to identify and validate microRNAs (miRNAs) differentially expressed in visceral white adipose tissue (vWAT) across individuals with varying degrees of obesity and type 2 diabetes (T2D). Initially, a discovery cohort underwent next‐generation sequencing (NGS) to explore global miRNA expression profiles. Technical validation of statistically significant candidates was subsequently performed by RT‐qPCR in the same cohort. Among these, hsa‐miR‐100‐3p was not detectable, while hsa‐miR‐146b‐3p, hsa‐miR‐342‐3p, hsa‐miR‐369‐3p and hsa‐miR‐941 could not be confirmed. In contrast, hsa‐miR‐12136 showed a trend towards differential expression and hsa‐miR‐141‐3p, hsa‐let‐7g‐3p, hsa‐miR‐200b‐3p and hsa‐miR‐585‐3p were successfully validated, exhibiting consistent and significant differences between experimental groups. The observed validation attrition is consistent with well‐documented technical differences between NGS and RT‐qPCR platforms, including differences in sensitivity, dynamic range and susceptibility to sequencing or adapter ligation biases affecting low‐abundance miRNAs [30, 31] Mean normalised NGS counts and mean ΔCt values by group are provided in Tables S2 and S3, allowing direct comparison of expression trends across platforms.

Following this initial technical validation, a larger independent cohort was used to assess the biological relevance of the validated miRNAs. This step was designed to confirm whether the observed expression patterns were reproducible in a broader population and to explore potential associations with biochemical and metabolic parameters. This two‐stage design strengthens the reliability of miRNA biomarkers and reduces the risk of false‐positive findings that can arise from small‐scale sequencing studies.

An important conceptual consideration emerging from our data is the dominant role of obesity over T2D status in shaping the vWAT miRNA and mRNA landscape. The results consistently indicate that the most robust transcriptomic differences occur between lean and obese individuals regardless of T2D status, with no miRNA exclusively differentially expressed between obese subgroups and substantial PCA overlap between them. T2D‐specific molecular changes in vWAT, if present, appear largely masked by the pervasive transcriptional reprogramming driven by obesity itself. This pattern is consistent with our previous pilot study in a smaller cohort, where T2D‐related differences were observed for only two miRNAs in visceral adipose tissue [32] and with earlier findings by Ortega et al. [19] in subcutaneous adipose tissue of obese women. Together, these observations suggest that larger cohorts with more granular metabolic phenotyping, including measures of insulin resistance, beta‐cell function and disease duration, may be needed to unmask T2D‐specific regulatory networks in vWAT.

Among the miRNAs assessed in the biological validation, let‐7g‐3p no longer showed differences between groups. This contrasts with published literature reporting downregulation of let‐7 family members in patients with Metabolic Syndrome [33], as well as associations with obesity, T2D [34] and adipocyte differentiation [35]. However, this apparent discrepancy may be partially explained by the biological heterogeneity within the let‐7 family. Although all members share a common seed region and are generally assumed to exert similar functions [36], evidence from rodent models suggests that individual isoforms may have distinct regulatory roles [37, 38]. Therefore, associations attributed to let‐7 g‐3p in previous studies may reflect the activity of the let‐7 family as a whole rather than this specific isoform. Additionally, many miRNA studies are conducted in small cohorts [20, 34] and effects that reach significance in limited sample sizes may not replicate in larger, more representative populations, as observed here.

miR‐141‐3p was upregulated in individuals with obesity as well as for miR‐200b. This upregulation was statistically significant across all covariate adjustment models, including full adjustment for age and sex simultaneously, although the magnitude of statistical significance was attenuated after adjustment (from p < 0.001 to p < 0.05), suggesting that demographic factors partially contribute to the observed group differences. These two miRNAs are part of a bigger miRNA family, the 200 miR family, which consists of five members divided into two chromosomal clusters and two functional clusters [39, 40]. What is surprising is that in this case both have a similar tendency even though they do not share either chromosomal cluster or functional group. This suggests that all the members of this family tend to have higher expression in obesity independently of which cluster they belong to. The miR‐200 family members have been studied independently and predominantly in cancer, where studies have provided evidence of the action of miR‐200 in Epithelial‐Mesenchymal transition [40]. But a recent study from our group has shown that this family is very important in obesity and type 2 diabetes [26], where there is a positive correlation with BMI and in silico pathways related to diabetes have been found. These findings support the notion that members of the miR‐200 family may play an important role in the molecular pathways associated with obesity and the development of T2D, potentially serving as biomarkers of metabolic dysfunction. In our cohort, both miR‐200b‐3p and miR‐141‐3p were more highly expressed in individuals with obesity, independent of diabetes status. This pattern suggests that their dysregulation may occur early in the metabolic alteration process and could contribute to impaired insulin signalling or to the disruption of protective regulatory pathways that normally maintain metabolic homeostasis. However, mechanistic studies are still required to determine whether these miRNAs exert a causal influence on the progression towards T2D or reflect compensatory responses to the altered metabolic environment.

Further exploration of our findings points towards a potential regulatory relationship between miR‐141, miR‐200b and HOXC10. This gene encodes a transcription factor implicated in cell differentiation and proliferation [41], which indicates a poor prognosis in cancer when it is upregulated [42]. However, it is important to note that in our dataset, HOXC10 expression did not show significant correlations with miR‐141 or miR‐200b and this gene is also predicted to be regulated by additional miRNAs whose expression did not differ significantly among groups. Therefore, these observations should be interpreted with caution. The apparent trends may reflect complex regulatory networks rather than direct relationships.

In addition to HOXC10, our analysis identified FBXO32 and TF as potential targets showing negative correlations with miR‐200b and miR‐141, respectively. FBXO32 is a key E3 ubiquitin ligase involved in protein degradation and metabolic remodelling in skeletal muscle [43]. It has been seen that obesity favours muscle atrophy with an upregulation of FBXO32 [44] and reciprocally compromised musculoskeletal health consistently emerges as a common hallmark in the progression of this metabolic disorder, instigating a vicious cycle that further deteriorates the metabolic status [45]. Also, this gene is key in epithelial‐mesenchymal transition (EMT) as it stabilises proteins that take part in epigenetic remodelling needed for a suitable environment for EMT progression [46].

Interestingly, in our cohort, FBXO32 expression displayed a negative correlation with miR‐200b, with higher miR‐200b levels observed in individuals with obesity. This finding appears contrary to the previously reported oncogenic context in which FBXO32 promotes EMT and metastasis [46]. Considering that miR‐200b is widely recognised as an EMT suppressor [47], these opposing trends may reflect tissue‐specific regulatory dynamics in adipose tissue rather than canonical tumour mechanisms. Furthermore, context‐dependent regulation of miRNAs under metabolic stress has been reported [48], suggesting that inflammatory and metabolic cues in visceral adipose tissue could modify the miR‐200b–FBXO32 interaction.

Similarly, the inverse association between miR‐141 and TF (transferrin) in visceral adipose tissue may reflect an adaptive response to metabolic stress in obesity. TF plays a crucial role in iron transport and homeostasis and its dysregulation has been linked with oxidative stress and chronic inflammation in metabolic disorders [49, 50, 51]. Under normal physiological conditions, TF expression correlates positively with adipogenic differentiation and insulin action. However, when adipose tissue expansion reaches its maximum and becomes metabolically exhausted, as occurs in advanced obesity, TF biosynthesis in both visceral and subcutaneous adipose depots is significantly reduced [51]. Therefore, the elevated expression of miR‐141 observed in individuals with obesity could represent a mechanism through which local adipose tissue iron homeostasis is altered in obesity, consistent with previous reports describing its involvement in iron regulation [52]. Whether these local changes translate to systemic iron dysregulation remains to be determined, as systemic iron markers were not assessed in the present study.

Taken together, these findings suggest that members of the miR‐200 family may contribute to a broader network of post‐transcriptional regulation in visceral adipose tissue, potentially balancing gene expression associated with both metabolic stress and tumorigenic pathways. Nevertheless, functional validation experiments will be essential to confirm these interactions and to clarify their biological relevance in the context of obesity and type 2 diabetes.

For miR‐12136 and miR‐585‐3p, a similar statistically significant pattern of differential expression was observed, with both miRNAs showing reduced levels in individuals with obesity. Notably, the statistical significance of these associations increased after progressive covariate adjustment for age and sex, with both miRNAs showing stronger effects in the fully adjusted model (from p < 0.01 to p < 0.001 for most comparisons), indicating that demographic factors were partially masking rather than inflating the true magnitude of the associations. This decrease may suggest their involvement in pathways that favour obesity, as lower miRNA expression typically leads to higher target gene expression. For miR‐12136, little is known about its function; it has only been identified in adipose‐derived stem cells from diabetic patients as well as in extracellular vesicles from patients with metabolic syndrome [53]. Interestingly, obesity has been increasingly recognised as a driver of neurological dysfunction through visceral adipose tissue‐derived inflammatory mediators and dyslipidemia [54], raising the possibility that miRNAs dysregulated in visceral adipose tissue, including hsa‐miR‐12 136, may participate in adipose‐brain crosstalk. However, direct evidence for such a role remains absent and further studies are needed to elucidate its biological function in obesity‐related pathologies.

miR‐585‐3p follows a similar trend in literature; only a few articles talk about this miRNA in association with cell growth and proliferation in cancer, where it seems to act as a protector of cell and tumour proliferation [55, 56]. The observed lower expression of miR‐585‐3p in individuals with obesity is of potential interest given the well‐established association between obesity and cancer risk [57]. However, the present dataset does not allow us to confirm this association.

Taken together, both miR‐12136 and miR‐585‐3p show marked differential expression due to obesity, but limited functional data exist for either in this setting, establishing these miRNAs as promising candidates for further research, potentially serving as biomarkers or targets for therapeutic interventions in obesity‐related pathologies.

Although several studies have investigated miRNA expression in obesity, the majority have relied on in vitro models or small patient cohorts and have typically focused on only one or a few miRNAs rather than broad expression profiles [20]. In contrast, the present work integrates a genome‐wide approach with experimental and biological validation in human adipose tissue, providing a more comprehensive overview of miRNA dysregulation in the context of obesity and T2D. This strategy has allowed us not only the confirmation of previously described miRNAs but also the identification of novel candidates potentially implicated in adipose tissue dysfunction and metabolic alterations.

5. Limitations

Several limitations of the present study should be acknowledged. First, the small RNA‐seq discovery analysis was performed in a relatively small cohort (n = 48) and the DESeq2 differential expression model did not include age and sex as covariates, in contrast to the limma‐voom pipeline applied for mRNA analysis. While the biological robustness of the key findings was subsequently confirmed in the independent validation cohort using RT‐PCR with progressive covariate adjustment, the exploratory and unadjusted nature of the initial sequencing‐based discovery should be considered when interpreting those results. Second, comprehensive data on liver and kidney function, concomitant diseases, diabetes duration and ongoing pharmacological treatments, including metformin, insulin and GLP‐1 receptor agonists, were not systematically collected as part of the study protocol. Although standard preoperative screening excluded individuals with severe hepatic or renal impairment and active inflammatory or immunosuppressive conditions, the potential influence of these factors on the transcriptomic profiles reported here cannot be excluded and future studies should incorporate more comprehensive clinical phenotyping. Third, as bulk tissue RNA sequencing was performed, the cellular origin of the detected miRNA and mRNA signals cannot be determined. Visceral adipose tissue is a heterogeneous tissue and increased immune cell infiltration associated with obesity may contribute to the observed transcriptomic differences; future studies incorporating cell‐type deconvolution or single‐cell sequencing approaches will be required to resolve the contribution of individual cell populations. Fourth, post hoc power analysis based on partial f ² effect sizes from the fully adjusted model indicated adequate statistical power for hsa‐miR‐12136 (power = 0.997) and hsa‐miR‐585‐3p (power = 0.958), but more limited power for hsa‐miR‐141‐3p (power = 0.593) and hsa‐miR‐200b‐3p (power = 0.603); findings for these latter miRNAs should be interpreted with caution and replicated in larger cohorts. Fifth, although the differentially expressed miRNAs showed biologically meaningful fold changes (1.5‐ to 3.5‐fold), formal diagnostic performance analysis was not conducted given the exploratory nature of this study; validation of biomarker potential would require prospective evaluation in an independent cohort with appropriate sample size for classification analyses. Finally, the correlative nature of the miRNA–mRNA associations identified in this study precludes causal inference and functional validation experiments will be necessary to confirm the biological relevance of the proposed regulatory interactions in the context of obesity and type 2 diabetes.

Author Contributions

Conceptualisation: Miguel García‐Villarino, Carmen Lambert. Methodology: Elsa Villa‐Fernández, Ana Victoria García, Laura Gallardo‐Nuell, Miguel García‐Villarino, Judit Fernández‐García, Aldara Martin Alonso, Pedro Pujante, Jessica Ares, Tomás González‐Vidal, María Moreno Gijón, Lourdes María Sanz Álvarez, Estrella Olga Turienzo Santos, José Manuel Fernández‐Real, Mario Fernández Fraga, Carmen Lambert. Formal analysis: Elsa Villa‐Fernández, Ana Victoria García, Carmen Lambert. Investigation: Elsa Villa‐Fernández, Ana Victoria García, Laura Gallardo‐Nuell, Miguel García‐Villarino, Judit Fernández‐García, Aldara Martin Alonso, Pedro Pujante, Jessica Ares, Tomás González‐Vidal, Lorena Suárez Gutiérrez, Claudia Lozano‐Aida, Raquel Rodríguez Uría, Sandra Sanz Navarro, María Moreno Gijón, Elías Delgado, Carmen Lambert. Resources: Lorena Suárez Gutiérrez, Claudia Lozano‐Aida, Raquel Rodríguez Uría, Sandra Sanz Navarro, María Moreno Gijón. Data curation: Miguel García‐Villarino, Carmen Lambert. Writing – original draft preparation: Elsa Villa‐Fernández, Ana Victoria García, Carmen Lambert. Writing – review and editing: Elsa Villa‐Fernández, Ana Victoria García, Laura Gallardo‐Nuell, Miguel García‐Villarino, Judit Fernández‐García, Aldara Martin Alonso, Claudia Lozano‐Aida, Lorena Suárez Gutiérrez, Pedro Pujante, Jessica Ares, Tomás González‐Vidal, Raquel Rodríguez Uría, Sandra Sanz Navarro, María Moreno Gijón, Lourdes María Sanz Álvarez, Estrella Olga Turienzo Santos, José Manuel Fernández‐Real, Mario Fernández Fraga, Elías Delgado, Carmen Lambert. Validation: Elsa Villa‐Fernández, Ana Victoria García, Carmen Lambert. Supervision: Lourdes María Sanz Álvarez, Estrella Olga Turienzo Santos, José Manuel Fernández‐Real, Mario Fernández Fraga, Elías Delgado. Funding acquisition: Elías Delgado.

Funding

This work was supported by the Instituto de Salud Carlos III (CD23/00037, CM24/00080, PI19/01162), the Gobierno del Principado de Asturias (IDE/2024/000705) and the Agencia Estatal de Investigación (MCIU‐24‐FPU23‐03765).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: hsa‐miR‐191‐3p expression among groups.

Table S2: Differentially expressed miRNAs detected by NGS in the discovery cohort.

Table S3: Mean ΔCT by group in the validation cohort.

Table S4: Pairwise group comparisons of differentially expressed miRNAs across four progressive adjustment models in visceral adipose tissue.

Table S5: Post hoc statistical power estimates for the four miRNAs showing statistically significant differential expression in visceral adipose tissue.

Figure S1: Boxplot showing Ct values of hsa‐miR‐191‐5p across experimental groups in the discovery cohort (n = 48: 10 controls, 19 OB_noT2D, 19 OB_T2D). Statistical differences were assessed using the Kruskal–Wallis test, followed by Dunn's post hoc test with Bonferroni correction. No statistically significant differences were observed across groups, confirming the stability of hsa‐miR‐191‐5p as an endogenous control across metabolic states. [Control: individuals without obesity and without type 2 diabetes; OB_noT2D: individuals with obesity and without type 2 diabetes; OB_T2D: individuals with obesity and type 2 diabetes].

Figure S2: Violin plots showing relative miRNA expression in visceral adipose tissue across metabolic groups in the unadjusted model (Kruskal–Wallis test followed by Dunn's post hoc test with Benjamini–Hochberg correction). Validation cohort (n = 100: 19 controls, 51 OB_noT2D, 30 OB_T2D). (A) hsa‐miR‐200b‐3p; (B) hsa‐miR‐141‐3p; (C) hsa‐miR‐12 136; (D) hsa‐miR‐585‐3p; (E) hsa‐let‐7g‐3p. Expression is shown as −2^ΔCt, normalised to hsa‐miR‐191‐5p. *p < 0.05, **p < 0.01, ***p < 0.001; ns: not significant. [Control: individuals without obesity and without type 2 diabetes; OB_noT2D: individuals with obesity and without type 2 diabetes; OB_T2D: individuals with obesity and type 2 diabetes].

Figure S3: Violin plots showing relative miRNA expression in visceral adipose tissue across metabolic groups in the model adjusted for age (linear model with estimated marginal means, emmeans package; Benjamini‐Hochberg correction applied across pairwise contrasts). Validation cohort (n = 100: 19 controls, 51 OB_noT2D, 30 OB_T2D). (A) hsa‐miR‐200b‐3p; (B) hsa‐miR‐141‐3p; (C) hsa‐miR‐12 136; (D) hsa‐miR‐585‐3p. Expression is shown as −2^ΔCt, normalised to hsa‐miR‐191‐5p. *p < 0.05, **p < 0.01, ***p < 0.001; ns: not significant. [Control: individuals without obesity and without type 2 diabetes; OB_noT2D: individuals with obesity and without type 2 diabetes; OB_T2D: individuals with obesity and type 2 diabetes].

Figure S4: Violin plots showing relative miRNA expression in visceral adipose tissue across metabolic groups in the model adjusted for sex (linear model with estimated marginal means, emmeans package; Benjamini‐Hochberg correction applied across pairwise contrasts). Validation cohort (n = 100: 19 controls, 51 OB_noT2D, 30 OB_T2D). (A) hsa‐miR‐200b‐3p; (B) hsa‐miR‐141‐3p; (C) hsa‐miR‐12 136; (D) hsa‐miR‐585‐3p. Expression is shown as −2^ΔCt, normalised to hsa‐miR‐191‐5p. Colours indicate sex (orange: female; blue: male). *p < 0.05, **p < 0.01, ***p < 0.001; ns: not significant. [Control: individuals without obesity and without type 2 diabetes; OB_noT2D: individuals with obesity and without type 2 diabetes; OB_T2D: individuals with obesity and type 2 diabetes].

Figure S5: Venn diagrams showing the overlap between in silico–predicted target genes of selected miRNAs and differentially expressed mRNAs identified by RNA‐seq analysis. Blue circles represent predicted target genes for each up‐regulated miRNA, and yellow circles indicate significantly (A) downregulated or (B) upregulated mRNAs. Overlapping areas indicate genes shared between miRNA target predictions and RNA‐seq results, with the number of genes and representative gene names displayed in each intersection.

Figure S6: Venn diagrams showing the overlap between in silico–predicted target genes of selected miRNAs and differentially expressed mRNAs identified by RNA‐seq analysis. Blue circles represent predicted target genes for each down‐regulated miRNA, and yellow circles indicate significantly (A) upregulated or (B) downregulated mRNAs. Overlapping areas indicate genes shared between miRNA target predictions and RNA‐seq results, with the number of genes and representative gene names displayed in each intersection.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.