Stina Ingrid Alice Svensson1†
Anne Hoffmann2†Tobias Hagemann2
Akin Cayir3Sadia Saeed3
Adhideb Ghosh4
Christian Wolfrum4
Matthias Blüher2,5
Yvonne Böttcher1,3*- 1Department of Clinical Molecular Biology, EpiGen, Institute of Clinical Medicine, University of Oslo, Oslo, Norway
- 2Helmholtz Institute for Metabolic, Obesity and Vascular Research (HI-MAG) of the Helmholtz Zentrum München at the University of Leipzig and University Hospital, Leipzig, Germany
- 3EpiGen, Medical Division, Akershus University Hospital, Lørenskog, Norway
- 4Laboratory of Translational Nutrition Biology, Institute of Food, Nutrition and Health, ETH Zürich, Schwerzenbach, Switzerland
- 5Medical Department III – Endocrinology, Nephrology, Rheumatology, University of Leipzig Medical Center, University of Leipzig, Leipzig, Germany
Background: Obesity is a global health burden and recent evidence indicates that epitranscriptomic regulation is potentially involved in its etiology. The epitranscriptomic mark 5-methylcytosine (m5C) is implicated in cancer and recent data linked the gene expression of m5C writers, erasers and readers to diabetes, a well-known co-morbidity of obesity. Here, we tested whether gene expression of m5C regulators in paired samples of human visceral and subcutaneous adipose tissue is (i) adipose tissue depot-specific and (ii) correlates with important clinical variables of obesity.
Methods: Intra-individually paired adipose tissue samples from human subcutaneous adipose tissue (SAT) and omental visceral adipose tissue (OVAT) were utilized from three different cohorts from the Leipzig Obesity Biobank including a large cross-sectional cohort, a two-step bariatric surgery cohort and a cohort of metabolically healthy vs unhealthy individuals (LOBB, total N=962). Data analysis on intra-individual samples was performed by using the paired Wilcoxon signed-rank test, while in comparisons on independent groups the unpaired Wilcoxon rank-sum test was employed. Bonferroni correction method was used to adjust multiple testing of p-values and Spearman’s rank correlation was used to assess associations.
Results: We observed that multiple m5C regulators were differentially expressed between human subcutaneous and visceral adipose tissue depots. Interestingly, we found that for several regulators the effects were less pronounced after weight loss, whilst stronger in individuals with insulin resistance compared to their healthy counterparts. A strong correlation of m5C regulator expression with macrophages was observed in OVAT compared to its SAT counterpart. Correlations between m5C regulators with important clinical variables related to obesity were observed in all three cohorts.
Conclusion: Our findings provide evidence for adipose tissue depot-specific gene expression of m5C regulators that correlate with clinical variables of obesity.
Introduction
Obesity is a global health burden and the World Health Organization has predicted that 46% of the worlds adult population will be affected by obesity in 2025 (https://data.worldobesity.org/publications/?cat=22). It is well known that especially accumulation of fat in visceral depots rather than subcutaneous counterparts entails a higher risk of metabolic disorders and co-morbidities (1–3). There have been multiple studies reporting transcriptional and epigenetic differences between human subcutaneous adipose tissue (SAT) and omental visceral adipose tissue (OVAT), along with recent evidence for depot-specific differential chromatin landscapes showing that OVAT is more open and accessible compared to SAT (4–8). Moreover, multiple clinical variables related to obesity and type 2 diabetes such as body mass index (BMI), waist-to-hip ratio (WHR), fasting plasma glucose (FPG), glycosylated hemoglobin (HbA1c), homeostasis model assessment of insulin resistance (HOMA-IR) and triglyceride levels correlate with gene expression and epigenetic marks in adipose tissue (9–11).However, the exact molecular mechanisms leading to a higher risk for metabolic complications are not yet fully understood.
A growing area of interest is the understanding of post-transcriptional RNA modifications in metabolic diseases (12, 13). The best studied post-transcriptional modification on messenger RNA (mRNA) is m6A (N6-methyladenosine) that may impact on mRNA metabolism by mRNA decay, pre-mRNA processing, export and translation (14). There is evidence that gene expression of m6A modifiers is adipose tissue depot-specific and correlates with clinical variables related to obesity (7). Another RNA modification of high interest is m5C (5-methylcytosine) that was shown to have more than 10,000 potential sites for modification in the human transcriptome, including transfer RNA (tRNA), ribosomal RNA (rRNA), mRNA and non-coding RNAs (ncRNAs) (15).
m5C is a dynamic epitranscriptomic mark that is regulated by protein groups important for installing m5C methylation (methyltransferases- writers), removing m5C marks by demethylation (erasers) and for functional translation by binding proteins (readers) (14). m5C regulation has reportedly been linked to changes in gene expression, RNA stability and protein synthesis, impacting both cellular and physical processes (15).
A recent report described an association of m5C-regulators with type 2 diabetes mellitus (T2D) (16). Through an integrated analysis of gene expression, Song et al. (2022) showed an enrichment of GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways that were closely related to metabolism-related biological processes (16). The following genes encoding m5C regulators were investigated by Song et al. (2022): 3 erasers (TET1, TET2, and TET3), 11 writers (DNMT1, DNMT2, DNMT3A, DNMT3B, NSUN1, NSUN2, NSUN3, NSUN4, NSUN5, NSUN6 and NSUN7) and 16 readers (ALYREF, MBD1, MBD2, MBD3, MBD4, MECP2, NEIL1, NTHL1, SMUG1, TDG, UHRF1, UHRF2, UNG, YBX1, ZBTB4, and ZBTB33). The list of genes was generated by Song et al. (2022) through a review of the literature and based on publications from Du et al. (2020) and Zhang et al. (2021) (16–18). The authors observed that gene expression of m5C regulators significantly correlated with genes involved with T2D (16).
Additionally, recent studies support a role of m5C in adipose tissue biology. Liu et al. (2021) showed that the knockdown of NSUN2 (NOL1/NOP2/Sun domain family member 2), a well-known m5C writer, increased lipid accumulation in 3T3-L1 pre-adipocytes by interacting with CDKN1A (cyclin-dependent kinase inhibitor 1A) in early stages of adipogenesis (19). Further, Liu et al. (2022) discovered in adipocytes from pigs that the m5C reader ALYREF (Aly/REF export factor) was involved in recognizing and transporting YBX2 (Y-box binding protein 2) and SMO (smoothened, frizzled class receptor) into the cytoplasm, leading to inhibited adipogenesis and increased myogenesis (20).
Inspired by these studies, we set out to better understand the role of m5C regulators in human obesity with a special emphasis on gene expression in different human adipose tissue depots (SAT vs. OVAT). To our knowledge, no studies exist exploring transcription of the m5C machinery in distinct adipose tissue depots and evaluating its relationship to clinical variables of obesity. To do so, we utilized the m5C regulator gene list as described in Song et al. (2022) and applied it to intra-individually paired samples of SAT and OVAT from the LOBB (Leipzig Obesity Biobank) cohort (21, 22). We aimed to (i) test whether gene expression of the m5C regulators is adipose tissue depot-specific; and (ii) whether gene expression of these genes correlates with clinical variables of obesity.
Method and materials
Cohort description
In this study, we used the Leipzig Obesity BioBank (LOBB; https://www.helmholtz-munich.de/en/hi-mag/cohort/leipzig-obesity-bio-bank-lobb)comprising intra-individually paired SAT and OVAT samples organized in several sub cohorts (total N=962). All patients were extensively clinically phenotyped regarding obesity, metabolic variables and anthropometric traits as described by Blüher (2020) and Klöting et al. (2010) (21, 22). Adipose tissue samples were collected during elective laparoscopic abdominal surgeries as described previously (23). Exclusion criteria comprised individuals under the age of 18, chronic substance or alcohol misuse, smoking within the last year leading up to surgery, acute inflammatory diseases, end-stage malignant diseases, uncontrolled thyroid disorder, and Cushing’s disease. Three sub cohorts of LOBB were considered: a cross-sectional cohort (CSC, N=839), a two-step bariatric surgery cohort (BSC, N=52) and a cohort comprising individuals with metabolically healthy vs. unhealthy obesity (MHUO, N=71). The CSC cohort comprises 818 subjects with obesity (72% women, 40% T2D, mean ± sd age 46.9 ± 11.8 years; mean ± sd BMI 48.9 ± 8 kg/m2) and 21 subjects without obesity (62% women, 17% T2D, mean ± sd age 49 ± 13.5 years; mean ± sd BMI 27 ± 2.5 kg/m2). The BSC cohort consists of 52 individuals that have undergone a two-step bariatric surgery at two time points and who were considered morbid obese at the first surgery. Only individuals who lost more than 10 kg of weight were considered. The first time point (first surgery) is labelled pre-surgery (70% women, 58% T2D, mean ± sd age 44.5 ± 9.6 years; mean ± sd BMI 54.3 ± 10.6 kg/m2), the second time point (second surgery) is labelled post-surgery (50% T2D, mean ± sd age 46.7 ± 9.6 years; mean ± sd BMI 40.2 ± 7.5 kg/m2). Individuals lost 41.1 ± 21.2 kg between the two surgeries. The MHUO cohort entails 30 obese individuals who are insulin sensitive (IS; 71% women, mean ± sd glucose infusion rate (GIR) clamp 89.5 ± 9.3 mmol/kg/min, mean ± sd age 38.5 ± 11.4 years; mean ± sd BMI 44.5 ± 7.1 kg/m2, 10% T2D) and 41 individuals who are insulin resistant (IR; 73% women, mean ± sd GIR clamp 28.7 ± 12–6 mmol/kg/min, mean ± sd age 50.4 ± 8 years; mean ± sd BMI 47.2 ± 6.9 kg/m2) 61% T2D). The cohort participants were stratified into metabolically healthy and unhealthy obesity groups according to previously published parameters. A detailed description of the MHUO cohort is provided in Reinisch et al., 2025 (31).
Library preparation and RNA-seq analysis
Library preparation and RNA sequencing (RNA-seq) were carried out as previously described (24). Briefly, RNA was isolated from adipose tissue samples using the SMARTseq (Switch Mechanisms at the 5’ End of RNA Templates Sequencing) protocol (25). Sequencing libraries were generated and subjected to single-end sequencing on a NovaSeq 6000 platform at the Functional Genomics Centre Zurich.
Raw sequencing reads were processed using Fastp (v0.23.4) for adapter removal and quality trimming (26). A sliding window approach was applied, trimming the read ends when the average quality within a four-base window dropped below a Phred quality score of 20. Reads shorter than 18 bases after trimming were discarded. Trimmed reads were aligned to the human reference genome (GRCh38, GENCODE release 47) using STAR (v2.7.11b), allowing up to 50 multiple alignments per read (27). Gene-level quantification was performed using featureCounts (v1.5.3), with fractional assignment for reads mapping to multiple genomic locations (28). Count data were transformed using variance-stabilizing transformation (VST) via the DESeq2 R package v1.48.1, and expression values were adjusted for age, sex and proportion of reads assigned to features (exon mapping rate) using the r package limma v 3.64.0 before further analysis (29, 30).
Statistical analyses
Analysis was conducted in R 4.5.0. Statistical significance was assessed using Wilcoxon tests implemented via the rstatix R package (v0.7.2). For comparisons involving intra-individually paired samples (pre- and post-intervention measures from the same individual; tissue comparisons), the paired Wilcoxon signed-rank test was applied. For comparisons between independent groups, the unpaired Wilcoxon rank-sum test was used. P-values were adjusted for multiple testing using the Bonferroni correction method. Spearman’s rank correlation coefficients were calculated to assess monotonic associations between variables. Confidence intervals for the correlation coefficients were computed using the standard Fisher z-transformation. All correlation analyses were performed using the RVAideMemoire R package (v0.9-83-11).
Results
Obesity- and depot-specific regulation of m5C regulators in adipose tissue
To assess whether genes encoding for the m5C regulatory machinery are expressed in an adipose tissue depot-specific manner, we evaluated whether their expression levels differ significantly among our individuals with obesity of the cross-sectional cohort (N=818). Multiple writers (DNMT1, DNMT3A, DNMT3B, DNMT2, NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, and NSUN7; all adj. p < 0.001; Figure 1A), readers (MBD2, MBD3, MBD4, NEIL1, NTHL1, SMUG1, TDG, UHRF1, UNG, YBX1, and ZBTB33; all adj. p<0.001; Figure 1B) and all of the known erasers (TET1, TET2, and TET3; all adj. p<0.01; Figure 1C) are indeed significantly different expressed in SAT vs. OVAT (Supplementary Table 1). Except NSUN5, the writers had a significantly lower expression in OVAT compared to SAT. This may hint at a potentially decreased deposition of the m5C mark on RNA molecules in OVAT. Moreover, in contrast, most of the genes encoding for reader proteins (MBD3, UNG, SMUG1, NEIL1, NTHL1, MBD4, and ZBZB33), which are generally important for RNA metabolism and function, were significantly upregulated in OVAT. The three erasers removing m5C from RNA molecules were downregulated in OVAT compared to SAT.
Figure 1. Differential gene expression of m5C regulators in intra-individually paired samples from subcutaneous and omental visceral adipose tissue in individuals from the cross-sectional cohort. Gene expression levels of all selected (A) m5C writers, (B) m5C readers and (C) m5C erasers are compared between subcutaneous adipose tissue (SAT; blue) and omental visceral adipose tissue (OVAT; red) within the same obese individuals (N=818) of the cross-sectional cohort (CSC). Statistical significance was assessed using paired Wilcoxon tests, with p-values corrected for multiple comparisons via Bonferroni correction. P value legend, <0.001 (***), <0.01 (**), <0.05 (*). VST, variance stabilizing transformation.
Furthermore, in OVAT, the DNMT3A writer gene (adj. p < 0.001) and the eraser enzymes TET3 and TET2 (adj. p<0.01) were significantly downregulated in individuals with obesity (N=818) compared to non-obese controls (N=21). Conversely, several reader genes (MBD2, SMUG1, and MBD4) were upregulated (adj. p < 0.05) in individuals with obesity (Supplementary Figure 1; Supplementary Table 1). In SAT, the eraser enzymes TET1 and TET2 (adj. p<0.001) were downregulated in obese individuals, along with the reader UHRF2 and the writer NSUN3 (adj. p<0.01). The writer DNMT1 (adj. p<0.01) and the reader MBD3 (adj. p<0.05) were upregulated in obese individuals (Supplementary Figure 2; Supplementary Table 1).
Taken together, these data from our large cross-sectional cohort provide sound evidence for adipose tissue depot-specific gene expression of m5C regulating genes. Some regulatory genes are differentially expressed between non-obese subjects and individuals with obesity, while certain patterns such as decreased eraser activity among obese individuals are observed across multiple depots.
m5C regulators correlate with fat accumulation, adipocyte size and metabolic traits in a depot-specific manner
We observed multiple correlations (p<0.05) between the expression levels of m5C methylation regulators and clinical variables related to obesity within the cross-sectional cohort (Figure 2, Supplementary Table 2). Most of these correlations with anthropometric variables were identified for body fat (N=292). In OVAT, NSUN2 (writer, ρ=-0.15, p=0.012) and MBD1 (reader, ρ=-0.16, p=0.007) exhibited weak negative correlations, while ALYREF (reader, ρ=0.12, p=0.041) showed a positive correlation with body fat (Figure 2A). In contrast, the SAT samples showed stronger negative correlations between body fat and several methylation regulators, including the writers NSUN2 (ρ=-0.14, p=0.016), NSUN4 (ρ=-0.12, p=0.048), and NSUN6 (ρ=-0.13, p=0.027), as well as the readers UHRF2 (ρ=-0.26, p<0.001), TDG (ρ=-0.14, p=0.014), and MBD1 (ρ=-0.13, p=0.028), along with the eraser TET2 (ρ=-0.15, p=0.009). Positive correlations with body fat were observed for the writer DNMT1 (ρ=0.2, p<0.001) and the readers SMUG1 (ρ=0.16, p=0.007), NEIL1 (ρ=0.17, p=0.005), and NTHL1 (ρ=0.2, p<0.001) (Figure 2B).
Figure 2. m5C regulators correlates with clinical variables of obesity in the CSC cohort. Spearman correlations of all selected m5C regulators included m5C writers, m5C readers and m5C erasers with obesity-related clinical variables are shown for (A) OVAT and (B) SAT. The color scheme representing the strength and direction of the correlation. P value (not adjusted for multiple testing) legend, <0.001 (***), <0.01 (**), <0.05 (*). BMI, body mass index; EBMIL, excess body mass index loss; FPG, fasting plasma glucose; FPI, fasting plasma insulin; FGF21, fibroblast growth factor 21; HOMA-IR, homeostatic model assessment of insulin resistance; HDL cholesterol, high-density lipoprotein cholesterol; HbA1c, Hemoglobin A1c; LDL cholesterol, low-density lipoprotein cholesterol; ssCRP, serum soluble C-reactive protein; VST, variance stabilizing transformation; WHtR, waist-to-height ratio.
Additionally, positive correlations were found between maximum adipocyte size in OVAT (N=75) and the erasers TET2 (ρ=0.26, p=0.027) and TET3 (ρ=0.26, p=0.024) suggesting that higher activity of these enzymes may be associated with larger fat cells. Conversely, SMUG1 showed a negative correlation (ρ=-0.24, p=0.041), indicating that increased SMUG1 expression might be linked to smaller adipocytes (Figure 2A). In SAT, ZBTB33 was positively correlated (ρ=0.28, p=0.016), while SMUG1 remained negatively correlated (ρ=-0.24, p=0.037) with the maximum adipocyte size (N=75) (Figure 2B).
In both adipose tissues, associations between metabolic parameters and m5C methylation regulators were observed. For example, in OVAT, DNMT3A (ρ=-0.16, p=0.007) and NEIL1 (ρ=-0.15, p=0.009) showed negative correlations with HOMA-IR (N=288), suggesting that higher activity of these genes may be linked to improved insulin sensitivity. Several methylation regulators also demonstrated correlations with circulating FGF21 (fibroblast growth factor 21) levels (N= 360). MBD3 (ρ=0.11, p=0.033) and SMUG1 (ρ=0.15, p=0.006) were positively associated with FGF21 levels, while DNMT2 (ρ=-0.13, p=0.014), MECP2 (ρ=-0.12, p=0.024), and NSUN6 (ρ=-0.11, p=0.034) exhibited weak negative correlations. These findings imply a potential epigenetic influence on FGF21 regulation, which could impact metabolic signaling pathways. Furthermore, resistin levels (N=420) were inversely correlated with several methylation regulators namely UHRF2 (ρ=-0.11, p=0.027), UNG (ρ=-0.17, p<0.001), NSUN2 (ρ=-0.14, p=0.004), and NSUN4 (ρ=-0.15, p=0.002), as well as positively associated with ALYREF (ρ=0.1, p=0.036), DNMT1 (ρ=0.11, p=0.028), DNMT3B (ρ=0.1, p=0.049), and TET3 (ρ=0.11, p=0.026) suggesting these enzymes may have a role in inflammation or adipokine regulation (Figure 2A). In SAT, several methylation regulators were also associated with HOMA-IR, including DNMT2 (ρ=-0.16, p=0.008), NEIL1 (ρ=-0.15, p=0.014), NSUN4 (ρ=-0.13, p=0.028), NSUN7 (ρ=-0.16, p=0.008), and YBX1 (ρ=0.14, p=0.022; N=288). Additionally, ALYREF (ρ=0.12, p=0.012) and ZBTB4 (ρ=-0.13, p=0.004) showed associations with HbA1c (N=450). Positive correlations with adiponectin levels were observed exclusively in SAT and involved MECP2 (ρ=0.22, p=0.047), MBD2 (ρ=0.24, p=0.031), and UHRF2 (ρ=0.26, p=0.018) (Figure 2B).
In summary, these findings highlight the potential role of specific m5C regulators in fat accumulation, adipocyte size, and obesity-related traits in a depot-specific manner, providing insights into potential epigenetic contributions to obesity and metabolic health.
Depot-specific adipose tissue gene expression of m5C regulators is less pronounced in the two -step bariatric surgery cohort
Next, we tested whether transcriptional levels of the m5C regulatory machinery are affected by weight loss in the two-step bariatric surgery cohort and whether this relates to any correlation with clinical parameters. We found significant (adj. p < 0.05) differences in gene expression of four m5C regulators between the pre- and post-surgery timepoints (DNMT1, NSUN2, and YBX1 in OVAT, Figure 3A; TET3 in SAT, Figure 3B). All these regulators exhibited significantly lower expression levels after weight loss (post-surgery) compared to baseline (pre-surgery) values. An overview of all comparisons is provided in Supplementary Table 3. No other m5C regulators were found differentially expressed.
Figure 3. m5C regulator expression differences and their clinical correlations in intra-individually paired samples from SAT and OVAT in individuals following weight loss. Only the gene expression comparisons of m5C regulators that showed significant differences between pre- and post-surgery expression levels are presented for (A) OVAT and (B) SAT within the same individuals (N=52) of the bariatric surgery cohort. Statistical significance was assessed using paired Wilcoxon tests, with p-values corrected for multiple comparisons via Bonferroni correction. Spearman correlations (unadjusted p<0.05 only) between the delta (pre-post) expression levels of m5C regulators and the delta of clinical parameters are illustrated for (C) OVAT and (D) SAT. The color scheme representing the strength and direction of the correlation. P value legend, <0.001 (***), <0.01 (**), <0.05 (*). BMI, body mass index; EBMIL, excess body mass index loss; HbA1c, Hemoglobin A1c; FPG, fasting plasma glucose; LDL cholesterol, low-density lipoprotein cholesterol; VST, variance stabilizing transformation.
The correlation analysis between delta (pre – post) changes in m5C regulator expression and delta weight loss-related phenotypes in our bariatric cohort revealed several associations (p<0.05) (Supplementary Table 4, Figures 3C, D). In OVAT, several correlations were observed between the expression changes of several m5C regulators and measures of fat loss or lipid-related phenotypes. Specifically, decreases in NSUN4 (ρ=-0.33, p=0.018, N=52) and TET3 (ρ=-0.44, p=0.001, N=52) delta expression values were associated with greater reductions in EBMIL (excess body mass index loss), indicating that lower NSUN4 and TET3 levels post-surgery may be linked to more effective weight loss. Decreases in NSUN6 (ρ=-0.55, p=0.025, N=17) and TET2 (ρ=-0.49, p=0.049, N=17) expressions were associated with reductions in body fat. In addition, strong inverse correlations were observed between TET2 (ρ=-0.69, p=0.011, N=13), TDG (ρ=-0.56, p<0.05, N=13), and YBX1 (ρ=-0.62, p=0.029, N=13) expression and LDL (low-density lipoprotein) cholesterol. Regarding triglycerides, decreases in DNMT3B (ρ=-0.55, p=0.016, N=19) and TET2 (ρ=-0.6, p=0.008, N=19) delta expressions were linked to lower delta triglyceride levels. Conversely, an increase in UHRF2 (ρ=0.48, p=0.037, N=19) expression was positively correlated with triglyceride levels (Figure 3C).
In SAT, changes in ALYREF expression were positively correlated with measures of fat loss, including BMI (ρ=0.33, p=0.016, N=52), body weight (ρ=0.35, p=0.011, N=52), and EBMIL (ρ=0.29, p=0.036, N=52). These results suggest that increased ALYREF expression in SAT may be associated with greater reductions in body mass and enhanced metabolic activity during weight loss. In contrast to OVAT, the expression levels of several m5C regulators in SAT were associated with FPG and HOMA-IR. Specifically, NSUN1 (ρ=-0.36, p=0.016, N=44) and NTHL1 (ρ=-0.33, p=0.029, N=44) expression showed a negative correlation with FPG, indicating that higher NSUN1 and NTHL1 levels may be linked to lower FPG. Conversely, NSUN6 (ρ=0.35, p=0.022, N=44) and MBD2 (ρ=0.3, p=0.048, N=44) expression levels show a positive correlation with FPG. Further, UHRF1 expression showed a strong negative correlation (ρ=-0.59, p=0.049, N=12) with HOMA-IR, implying that higher UHRF1 levels might be associated with improved insulin sensitivity in SAT (Figure 3D).
This data illustrates that the gene expression profile of the m5C regulatory machinery is altered in the two-step bariatric surgery group and that such changes seem to be related to metabolic parameters. This inspired us to next evaluate the expression profiles in a cohort comprising metabolically healthy vs. unhealthy individuals with obesity.
Cell type–resolved m5C regulator profiles suggest stromal and immune involvement in IR-associated adipose tissue changes
To profile the expression landscape of m5C regulatory genes across various cell types, single-nucleus RNA sequencing (snucSeq) data from IS and IR morbidly obese individuals were analyzed (Figure 4) (31). These data were obtained from Reinisch et al., 2025 (31), and include a subset of participants from our metabolically healthy versus unhealthy obesity (MHUO) cohort. In brief, snucSeq was performed on pooled adipose tissue samples from metabolically healthy and unhealthy obese individuals, comprising 41 OVAT and 36 SAT samples. Despite an overall lower gene expression of m5C regulators in SAT compared to OVAT, several m5C regulator genes displayed markedly elevated expression levels and were broadly distributed across various immune cell types (e.g., macrophages, monocytes, and T cells) and stromal cell types (e.g., adipocytes and mesothelial cells). Specifically, MECP2, MBD2, TET3, TET2, and DNMT3A were strongly expressed in macrophages, a key immune cell type involved in adipose tissue inflammation and metabolic dysfunction. These genes also showed high expression in monocytes, suggesting their broader involvement in innate immune regulation. Among the regulators, MECP2 exhibited the strongest and most widespread expression across virtually all cell types, suggesting a potentially global role in transcriptional and epigenetic regulation within adipose tissue. MBD2 showed the highest expression in adipocytes in both tissues, as well as in meMesoCs (mesenchymal-like mesothelial cells) and cMesoCs (mesothelial cell subtypes) in OVAT, indicating a possible role in modulating both lipid storage and mesothelial cell identity or function. TET1 and ZBTB4 are also stronger expressed in adipocytes (Figure 4).
Figure 4. Expression levels of m5C regulators across different cell types in single-nucleus RNA sequencing data from individuals with (insulin sensitivity) IS and (insulin resistance) IR. The dot plots display the expression intensity of m5C regulators, differentiated by cell type in (A) OVAT and (B) VAT of individuals with IS and IR of the metabolically healthy vs. unhealthy obesity cohort. Expression levels are indicated by color intensity, while the proportion of cells expressing each gene is represented by the size of the dots. Adipo, adipocytes; APCs, antigen-presenting cells; cMesoCs, classical mesothelial cells; EndoCs, endothelial cells; LECs, lymphatic endothelial cells; Macro, macrophages; MastCs, mast cells; meMesoCs, mesenchymal mesothelial cells; Mono, monocytes; NeurCs, neuronal cells; SMCs, smooth muscle cells; Peri, pericytes. Note that some genes are shown under their alternative names, NOP2 is listed as NSUN1, and TRDMT1 corresponds to DNMT2. The data were analyzed using the web applications available at https://github.com/WolfrumLab/MHUO/.
Analyses of m5C regulator RNAseq expression values between IS and IR samples of the MHUO cohort further (Figure 5, Supplementary Tables 5, 6) supported the functional relevance of selected m5C regulators identified by snucSeq. For example, MBD2 expression is significantly reduced in omental adipose tissue of IR individuals compared to IS individuals (Figure 5A), suggesting a potential loss of epigenetic regulation in key stromal cell types associated with insulin resistance. In IS individuals OVAT, DNMT3A expression is significantly negatively correlated with circulating leptin levels (ρ=-0.52, FDR=0.003), suggesting a potential anti-inflammatory or metabolically protective role. Additionally, TET3 expression in IS individuals in both tissues positively correlated with chemerin levels (OVAT: ρ=0.34, p=0.049; SAT: ρ=0.54, p=0.003) and positively with IL-6 levels in SAT (ρ=0.37, p=0.043) consistent with its stromal expression and involvement in adipokine regulation. Correlations observed for MECP2 and ZBTB4 in both IS and IR individuals, further support their roles in glycemic and inflammatory control, in line with their expression across adipocytes and immune cell types (Figure 5B).
Figure 5. m5C regulator expression differences and their clinical correlations in intra-individually paired samples from SAT and OVAT in individuals considered having healthy or unhealthy metabolic traits. Only the gene expression comparisons of m5C regulators that showed significant differences between expression levels of individuals which are insulin sensitive (IS; N=30) and insulin resistant (IR; N=41) are presented for (A) OVAT of the metabolically healthy vs. unhealthy obesity cohort. Statistical significance was assessed using paired Wilcoxon tests, with p-values corrected for multiple comparisons via Bonferroni correction. Spearman correlations (unadjusted p<0.05 only) for IS and IR expression levels of m5C regulators and clinical parameters are illustrated for (B) OVAT and (C) SAT. The color scheme representing the strength and direction of the correlation. P value legend, <0.01 (**), <0.05 (*). FPG, fasting plasma glucose; FPI, fasting plasma insulin; GIR, glucose infusion rate; HbA1c, Hemoglobin A1c; HOMA-IR, homeostatic model assessment of insulin resistance; IL-6, interleukin 6; MCP-1, monocyte chemoattractant protein 1; ssCRP, serum soluble C-reactive protein; VST, variance stabilizing transformation.
Discussion
Epitranscriptomic regulation in human adipose tissue through m5C methylation and its potential role in obesity is so far not well and systematically investigated. This work was inspired by the previous publication by Song et al. (2022) that shortlisted and identified multiple m5C regulators being related to type 2 diabetes through in-silico analyses (16). Here, we report that (i) m5C regulators are expressed in an adipose tissue depot-specific manner; (ii) and that gene expression levels of m5C regulators correlate with important clinical variables of obesity.
m5C regulators are adipose tissue depot-specific expressed
The role of the epitranscriptomic mark m5C has not been previously well described in the aetiology of obesity. A handful of studies reported a potential role of m5C in adipogenesis. Liu et al. (2021) provided insights into the molecular function of mRNA m5C in relation to the export of CDKN1A mRNA from the nucleus in controlling adipogenesis (19). They report the direct involvement of the writer and methyltransferase NSUN2 in regulating the m5C levels and the reader protein ALYREF to export CDKN1A mRNA to the cytoplasm (19). Furthermore, ALYREF was demonstrated to export YBX2 and SMO mRNA into the cytoplasm of porcine cells in a m5C-dependent manner to inhibit adipogenesis and promote myogenesis respectively (20). Hence, the findings reported by Liu et al. (2021) illustrate that known m5C regulators are involved in controlling adipogenesis, suggesting a possible link with obesity (19, 20). Our results provide novel insights in the gene expression profiles of writers, readers and erasers of m5C methylation in human adipose tissue and about depot-specific expression patterns in SAT vs. OVAT along with its adipose residing cell types in individuals with obesity, after weight loss and in individuals with metabolic alterations. In general, all except for NSUN5 of the m5C writers included in our study showed consistently lower gene expression in OVAT compared to SAT in the large cross-sectional cohort. The higher gene expression for m5C writers in SAT, especially the NSUN family with NSUN2, NSUN3, NSUN4, NSUN6 and NSUN7 all providing strong significances (p <0.001), is very interesting as higher levels of NSUN2 have been shown to inhibit adipogenesis (19, 20). This may indicate that depot-specific alterations in adipogenesis governed by m5C may be related to metabolic and anthropometric variables. Moreover, the writer NSUN3 was reported to catalyze the formation of m5C in mitochondrial tRNAs. (13). Given that mitochondrial dysfunction is known to be a consequence of obesity (32, 33), depot-specific NSUN3 gene expression may indicate at mitochondrial m5C potentially involved in obesity in a depot-specific manner (13). Moreover, one can speculate that a potentially higher activity of the writing machinery in SAT compared to OVAT may potentially lead to lower m5C deposition in OVAT. However, since our study lacks functional validation, more studies are warranted to test this. No causative conclusions can be drawn so far and the data presented in this study should therefore be interpreted with caution.
Moreover, our data indicate that weight loss (as demonstrated in our two-step bariatric surgery cohort) seems to be related to less pronounced depot-specific differences between SAT and OVAT. Four m5C regulators YBX1, DNMT1, NSUN2 in OVAT and TET3 in SAT, are differentially expressed between pre- and post-surgery, and their expression levels were all downregulated following weight loss. Interestingly, when analyzing cell-type specific gene expression by using snucSeq data, we discovered several genes strongly expressed in immune cells. The elevated expression of MECP2, MBD2, TET3, TET2, and DNMT3A in macrophages, particularly in omental tissue, supports the hypothesis of epigenetic reprogramming contributing to the development of insulin resistance, as macrophage-driven inflammation is a hallmark of IR-associated adipose remodeling.
The m5C readers MBD3, UNG, SMUG1, NEIL1, NTHL1, MBD4 and ZBTB33 show higher expression in OVAT compared to SAT. TDG, YBX1, MBD2 and UHRF1 showed higher expression levels in SAT compared to OVAT. NTHL1, primarily known for its role in DNA repair, has also been identified alongside eight other m5C readers as part of a specific m5C modification signature used to detect early acute myocardial infarction (34, 35). Similarly, TDG has been proposed as a diagnostic m5C-associated marker in idiopathic pulmonary fibrosis (36). Additionally, NEIL1, a key component of the base excision repair pathway essential for genomic integrity, is also implicated in various human diseases. Defects in NEIL1 have been associated with increased risks of cancer, diabetes and obesity (37). In the weight-loss two-step bariatric surgery cohort, the reader YBX1, DNMT1, writer NSUN2 and the eraser TET3 were the only regulators showing a significant differential gene expression between pre-and post-surgery timepoints and were observed downregulated following surgery. This data may hint at a potential impact of weight loss on the expression of these genes and the downregulation of YBX1 could result in decreased levels of its target proteins as it is known to actively stabilize mRNA through directly binding to m5C (38). However, no causative conclusions are possible due to the lack of functional validation in our study. Importantly, given that our cross-sectional cohort comprised only a limited number of non-obese individuals, our comparisons between obese vs non-obese individuals showing in general lower expression levels for the m5C machinery in subjects with obesity cannot be overestimated, and more data and larger cohorts are warranted in the future to draw any definite conclusion.
Furthermore, in the MHUO cohort MBD2 had a lower expression in IR individuals in OVAT compared to the IS counterparts. Given that MBD2 was predominantly expressed in adipocytes and mesothelial cells in the snucSeq data, this downregulation may reflect a loss of epigenetic repression mechanisms in key stromal compartments. Such a reduction could lead to increased transcription of pro-inflammatory or fibrotic genes, contributing to the dysfunctional remodeling of visceral fat observed in insulin resistance. These findings suggest a potential protective role of MBD2 in maintaining adipose tissue homeostasis, particularly in visceral depots.
Moreover, we observed higher expression levels in SAT compared to OVAT for the erasers TET1, TET2, and TET3, while the expression is lower across tissue depots among subjects with obesity. This may potentially hint at a general reduction of the eraser activity suggesting the possibility that individuals with normal weight have lower m5C levels compared to obese patients. The TET family is known to be involved in demethylation of both DNA and RNA m5C modification, in addition to their ability to yield 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxycytosine (5caC) through oxidation of m5C in DNA (39, 40). TET2 has previously been identified as an antiadipogenic demethylase in adipocyte differentiation in 3T3-L1 cells (41). Additionally, TET3 expression was decreased in SAT following weight loss. These findings are supported by Jung et al. (2023), who showed that TET3 is important for white adipose tissue development and expansion. Our results from the CSC cohort are in line with this, as we observed a higher TET3 expression in SAT (42). Jung et al. (2023), further proposed that TET3 regulates adipogenesis and diet-induced tissue expansion by modulating DNA methylation at important target genes in adipose precursor cells (42). Lastly, due to the involvement of the TET family in multiple forms of modification in DNA and RNA the results should be carefully interpreted as future validation studies are needed to confirm m5C methylation levels in the two depots investigated (39, 40).
Collectively, these results demonstrate a distinct depot- and cell-type-specific pattern of m5C regulator expression, with omental macrophages showing particularly elevated expression levels. This suggests a potential epigenetic axis through which visceral adipose tissue may contribute to insulin resistance in morbid obesity.
Clinical variables important for obesity correlate with m5C regulators
We find expression of multiple m5C regulators correlated with several clinical traits related to obesity and metabolism. In the CSC cohort, anthropometric traits like body fat were significantly correlated with a multitude of genes in both OVAT and SAT. This correlation was more widespread across several regulators in SAT compared to OVAT. Interestingly, NSUN2 was found to be negatively correlated with body fat for both depots. In line with this, we observed lower NSUN2 gene expression post-surgery and among subjects with obesity across both adipose tissue depots. These results align with the findings from Liu et al. (2021) showing that overexpression of NSUN2 corresponds to lower adipogenesis (19). In SAT, two additional NSUN family members (NSUN4 and NSUN6) are negatively correlated to body fat. NSUN4 have been linked to recruiting the degrading machinery for dsRNA in mitochondria to inhibit activation of the immune response due to faulty mitochondrial action (43). This may suggest a potential link between NSUN4 expression and the increase of immune cells in obesity. Furthermore, the three reader proteins NEIL1, NTHL1, SMUG1 and the writer DNMT1 all had a strong positive correlation in SAT with percentage body fat. The reader proteins are mainly known to be involved in the DNA base excision repair pathway, an important mechanism to be functional in an active depot (34). Moreover, in addition we observed strong correlations of multiple m5C regulators with mean and maximum adipocyte size suggesting a potential functional mechanism linking the m5C machinery to adipose biology. Furthermore, gene expression of multiple m5C regulators is correlated with important metabolic variables such as fasting plasma glucose and HOMA-IR across depots. However, only the SAT depot showed positive correlations for the three m5C readers MBD2, MECP2 and UHRF2 with adiponectin expression. Adiponectin is an adipokine with insulin-sensitizing properties and mainly originates from the SAT depot (44).
Although this is to our knowledge the first study exploring the relationship of m5C regulator gene expression and clinical variables related to obesity, our study has several limitations. First, two of the cohorts (BSC and MHUO) have relatively small sample size, which may limit the statistical power and generalizability of the findings described here. Second, the number of subjects with normal weight is relatively low in our cross-sectional cohort and our observed findings need to be interpreted with caution. Third, some of the genes identified by Song et al. (2022) to be members of the m5C regulating machinery, are also described as DNA m5C regulators (such as DNMT family and MBD family). Therefore, the observed correlation of gene expression with clinical traits may concern both DNA methylation and RNA methylation and should be carefully interpreted. Moreover, other genes may be involved in the regulation of RNA m5C, that were not included in the gene set identified by Song et al. (2022) and are therefore not analyzed in our study. Lastly, more functional data are therefore warranted to confirm or reject our results.
Conclusion
In conclusion, this study is, to our knowledge, the first to establish a link between gene expression profiles of m5C regulators and clinical traits related to human obesity. Our results demonstrate that 5-methylcytosine regulators are expressed in an adipose tissue depot-specific manner and that gene expression levels in the different fat depots are differentially correlated with important clinical variables.
Data availability statement
Bulk RNA and snRNA-sequencing data from the LOBB reported in this study cannot be deposited in a public repository due to restrictions by patient consent. These restrictions are due to local data protection regulation in the written informed consent form tissue donors signed before taking part in the study. Access to human adipose tissue biobank data is regulated by the LOBB steering committee. Use of data is strictly limited to research purposes and not intended for commercialization. To request access, contact Matthias Blüher (bWF0dGhpYXMuYmx1ZWhlckBtZWRpemluLnVuaS1sZWlwemlnLmRl) or Anne Hoffmann (YW5uZS5ob2ZmbWFubkBoZWxtaG9sdHotbXVuaWNoLmRl). Web application links to explore the datasets from the MHUO cohort are available on GitHub (https://github.com/WolfrumLab/MHUO/).
Ethics statement
The LOBB study was approved by the Ethics Committee of the University of Leipzig (approval no: 159-12-21052012) and performed in accordance with the Declaration of Helsinki. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.
Author contributions
SIAS: Writing – original draft, Writing – review & editing, Data curation, Investigation, Methodology, Visualization. AH: Writing – original draft, Writing – review & editing, Data curation, Formal analysis, Investigation, Methodology, Visualization. TH: Data curation, Writing – review & editing. AC: Writing – review & editing, Conceptualization. SS: Conceptualization, Writing – review & editing, Funding acquisition. AG: Writing – review & editing, Data curation. CW: Resources, Writing – review & editing. MB: Funding acquisition, Resources, Writing – review & editing. YB: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.
Funding
The author(s) declare financial support was received for the research and/or publication of this article. SIAS was funded by a research grant from Helse Sør-Øst to YB. SS was funded by an EU-Scientia postdoctoral Fellowship and project funding from the European Union's Horizon 2020 Research and Innovation program under the Marie Skłodowska-Curie Grant, (agreement No. 801133). MB received funding from grants from the DFG (German Research Foundation)—Project number 209933838—SFB 1052 (project B1) and by Deutsches Zentrum für Diabetesforschung (DZD, Grant: 82DZD00601).
Conflict of interest
MB received personal honoraria from Amgen, AstraZeneca, Bayer, Boehringer-Ingelheim, Lilly, Novo Nordisk, Novartis, and Sanofi as well as payments from Boehringer-Ingelheim to the institution.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declare that no Generative AI was used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2025.1647477/full#supplementary-material
References
1. Blüher M. Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol. (2019) 15:288–98. doi: 10.1038/s41574-019-0176-8
2. Idilman IS, Low HM, Gidener T, Philbrick K, Mounajjed T, Li J, et al. Association between visceral adipose tissue and non-alcoholic steatohepatitis histology in patients with known or suspected non-alcoholic fatty liver disease. JCM. (2021) 10:2565. doi: 10.3390/jcm10122565
3. Karpe F and Pinnick KE. Biology of upper-body and lower-body adipose tissue—link to whole-body phenotypes. Nat Rev Endocrinol. (2015) 11:90–100. doi: 10.1038/nrendo.2014.185
4. Saeed S, La Cour Poulsen L, Visnovska T, Hoffmann A, Ghosh A, Wolfrum C, et al. Chromatin landscape in paired human visceral and subcutaneous adipose tissue and its impact on clinical variables in obesity. eBioMedicine. (2025) 114:105653. doi: 10.1016/j.ebiom.2025.105653
5. Keller M, Kralisch S, Rohde K, Schleinitz D, Dietrich A, Schön MR, et al. Global DNA methylation levels in human adipose tissue are related to fat distribution and glucose homeostasis. Diabetologia. (2014) 57:2374–83. doi: 10.1007/s00125-014-3356-z
6. Keller M, Hopp L, Liu X, Wohland T, Rohde K, Cancello R, et al. Genome-wide DNA promoter methylation and transcriptome analysis in human adipose tissue unravels novel candidate genes for obesity. Mol Metab. (2017) 6:86–100. doi: 10.1016/j.molmet.2016.11.003
7. Rønningen T, Dahl MB, Valderhaug TG, Cayir A, Keller M, Tönjes A, et al. m6A regulators in human adipose tissue - depot-specificity and correlation with obesity. Front Endocrinol. (2021) 12:778875. doi: 10.3389/fendo.2021.778875
8. Rohde K, Keller M, la Cour Poulsen L, Rønningen T, Stumvoll M, Tönjes A, et al. (Epi)genetic regulation of CRTC1 in human eating behaviour and fat distribution. EBioMedicine. (2019) 44:476–88. doi: 10.1016/j.ebiom.2019.05.050
9. Keller M, Klös M, Rohde K, Krüger J, Kurze T, Dietrich A, et al. DNA methylation of SSPN is linked to adipose tissue distribution and glucose metabolism. FASEB J. (2018) 32:6898–910. doi: 10.1096/fj.201800528R
10. Clemente-Postigo M, Queipo-Ortuño MI, Fernandez-Garcia D, Gomez-Huelgas R, Tinahones FJ, and Cardona F. Adipose tissue gene expression of factors related to lipid processing in obesity. PLoS One. (2011) 6:e24783. doi: 10.1371/journal.pone.0024783
11. Wang Q, Wang J, Lin H, Huo X, Zhu Q, and Zhang M. Relationship between fat mass and obesity-associated gene expression and type 2 diabetes mellitus severity. Exp Ther Med. (2018) 15(3):2917–21. doi: 10.3892/etm.2018.5752
12. Balachander K, Priyadharsini JV, Roy A, and Paramasivam A. Emerging role of RNA m5C modification in cardiovascular diseases. J Cardiovasc Trans Res. (2023) 16:598–605. doi: 10.1007/s12265-022-10336-8
13. Li M, Tao Z, Zhao Y, Li L, Zheng J, Li Z, et al. 5-methylcytosine RNA methyltransferases and their potential roles in cancer. J Transl Med. (2022) 20:214. doi: 10.1186/s12967-022-03427-2
14. Trixl L and Lusser A. The dynamic RNA modification 5-methylcytosine and its emerging role as an epitranscriptomic mark. WIREs RNA. (2019) 10:e1510. doi: 10.1002/wrna.1510
15. Xue C, Zhao Y, and Li L. Advances in RNA cytosine-5 methylation: detection, regulatory mechanisms, biological functions and links to cancer. biomark Res. (2020) 8:1–13. doi: 10.1186/s40364-020-00225-0
16. Song Y, Jiang Y, Shi L, He C, Zhang W, Xu Z, et al. Comprehensive analysis of key m5C modification-related genes in type 2 diabetes. Front Genet. (2022) 13:1015879. doi: 10.3389/fgene.2022.1015879
17. Du E, Li J, Sheng F, Li S, Zhu J, Xu Y, et al. A pan-cancer analysis reveals genetic alterations, molecular mechanisms, and clinical relevance of m5 C regulators. Clin Trans Med. (2020) 10:e180. doi: 10.1002/ctm2.180
18. Zhang H, Xu P, and Song Y. Machine-learning-based m5C score for the prognosis diagnosis of osteosarcoma. J Oncol. (2021) 2021:1–11. doi: 10.1155/2021/4986227
19. Liu Y, Zhao Y, Wu R, Chen Y, Chen W, Liu Y, et al. mRNA m5C controls adipogenesis by promoting CDKN1A mRNA export and translation. RNA Biol. (2021) 18:711–21. doi: 10.1080/15476286.2021.1980694
20. Liu Y, Yang Y, Wu R, Gao C, Liao X, Han X, et al. mRNA m5C inhibits adipogenesis and promotes myogenesis by respectively facilitating YBX2 and SMO mRNA export in ALYREF-m5C manner. Cell Mol Life Sci. (2022) 79:481. doi: 10.1007/s00018-022-04474-0
21. Blüher M. Metabolically healthy obesity. Endocr Rev. (2020) 41:bnaa004. doi: 10.1210/endrev/bnaa004
22. Klöting N, Fasshauer M, Dietrich A, Kovacs P, Schön MR, Kern M, et al. Insulin-sensitive obesity. Am J Physiology-Endocrinology Metab. (2010) 299:E506–15. doi: 10.1152/ajpendo.00586.2009
23. Langhardt J, Flehmig G, Klöting N, Lehmann S, Ebert T, Kern M, et al. Effects of weight loss on glutathione peroxidase 3 serum concentrations and adipose tissue expression in human obesity. Obes Facts. (2018) 11:475–90. doi: 10.1159/000494295
24. Hagemann T, Czechowski P, Ghosh A, Sun W, Dong H, Noé F, et al. Laminin α4 expression in human adipose tissue depots and its association with obesity and obesity related traits. Biomedicines. (2023) 11:2806. doi: 10.3390/biomedicines11102806
25. Picelli S, Faridani OR, Björklund ÅK, Winberg G, Sagasser S, and Sandberg R. Full-length RNA-seq from single cells using Smart-seq2. Nat Protoc. (2014) 9:171–81. doi: 10.1038/nprot.2014.006
26. Chen S, Zhou Y, Chen Y, and Gu J. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics. (2018) 34:i884–90. doi: 10.1093/bioinformatics/bty560
27. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. (2013) 29:15–21. doi: 10.1093/bioinformatics/bts635
28. Liao Y, Smyth GK, and Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. (2014) 30:923–30. doi: 10.1093/bioinformatics/btt656
29. Love MI, Huber W, and Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. (2014) 15:550. doi: 10.1186/s13059-014-0550-8
30. Ritchie ME, Phipson B, Wu D, Hu Y, and Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. (2015) 43(7):e47. doi: 10.1093/nar/gkv007
31. Reinisch I, Ghosh A, Noé F, Sun W, Dong H, Leary P, et al. Unveiling adipose populations linked to metabolic health in obesity. Cell Metab. (2025) 37:640–655.e4. doi: 10.1016/j.cmet.2024.11.006
32. Paramasivam A, Meena AK, Venkatapathi C, Pitceathly RDS, and Thangaraj K. Novel biallelic NSUN3 variants cause early-onset mitochondrial encephalomyopathy and seizures. J Mol Neurosci. (2020) 70:1962–5. doi: 10.1007/s12031-020-01595-8
33. De Mello AH, Costa AB, Engel JDG, and Rezin GT. Mitochondrial dysfunction in obesity. Life Sci. (2018) 192:26–32. doi: 10.1016/j.lfs.2017.11.019
34. Das L, Quintana VG, and Sweasy JB. NTHL1 in genomic integrity, aging and cancer. DNA Repair. (2020) 93:102920. doi: 10.1016/j.dnarep.2020.102920
35. Guo Y, Jiang H, Wang J, Li P, Zeng X, Zhang T, et al. 5mC modification patterns provide novel direction for early acute myocardial infarction detection and personalized therapy. Front Cardiovasc Med. (2022) 9:1053697. doi: 10.3389/fcvm.2022.1053697
36. Tian L, Song W, Wu J, Lan Y, and Chen L. Diagnostic and predictive values of m5C−associated genes in idiopathic pulmonary fibrosis. Mol Med Rep. (2024) 31:53. doi: 10.3892/mmr.2024.13418
37. Mohajeri Khorasani A, Raghibi A, Haj Mohammad Hassani B, Bolbolizadeh P, Amali A, Sadeghi M, et al. Decoding the role of NEIL1 gene in DNA repair and lifespan: A literature review with bioinformatics analysis. Advanced Biol. (2024) 8:2300708. doi: 10.1002/adbi.202300708
38. Zhao T, Zhang Z, Chen Z, Xu G, Wang Y, and Wang F. Biological functions of 5-methylcytosine RNA-binding proteins and their potential mechanisms in human cancers. Front Oncol. (2025) 15:1534948. doi: 10.3389/fonc.2025.1534948
39. Zhang X, Zhang Y, Wang C, and Wang X. TET (Ten-eleven translocation) family proteins: structure, biological functions and applications. Sig Transduct Target Ther. (2023) 8:297. doi: 10.1038/s41392-023-01537-x
40. Fu L, Guerrero CR, Zhong N, Amato NJ, Liu Y, Liu S, et al. Tet-mediated formation of 5-hydroxymethylcytosine in RNA. J Am Chem Soc. (2014) 136:11582–5. doi: 10.1021/ja505305z
41. Hou Y, Zhang Z, Wang Y, Gao T, Liu X, Tang T, et al. 5mC profiling characterized TET2 as an anti-adipogenic demethylase. Gene. (2020) 733:144265. doi: 10.1016/j.gene.2019.144265
42. Jung BC, You D, Lee I, Li D, Schill RL, Ma K, et al. TET3 plays a critical role in white adipose development and diet-induced remodeling. Cell Rep. (2023) 42:113196. doi: 10.1016/j.celrep.2023.113196
43. Recazens E and Jourdain AA. Cytosine methylation flags mitochondrial RNA for degradation. Trends Biochem Sci. (2024) 49:843–5. doi: 10.1016/j.tibs.2024.08.001
Keywords: adipose tissue, obesity, RNA methylation, epitranscriptomics, m5C regulators
Citation: Svensson SIA, Hoffmann A, Hagemann T, Cayir A, Saeed S, Ghosh A, Wolfrum C, Blüher M and Böttcher Y (2025) Regulators of RNA m5C methylation are adipose tissue depot-specific expressed and correlate with clinical variables of obesity in humans. Front. Endocrinol. 16:1647477. doi: 10.3389/fendo.2025.1647477
Received: 15 June 2025; Accepted: 12 August 2025;
Published: 15 September 2025.
Edited by:
Claire Joanne Stocker, Aston University, United KingdomReviewed by:
William V. Trim, Harvard Medical School, United StatesJiekai Yin, University of California, Riverside, United States
Carine Jaafar, American University of Beirut, Lebanon
Gunjan Sharma, University of California, Los Angeles, United States
Copyright © 2025 Svensson, Hoffmann, Hagemann, Cayir, Saeed, Ghosh, Wolfrum, Blüher and Böttcher. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Yvonne Böttcher, eXZvbm5lLmJvdHRjaGVyQG1lZGlzaW4udWlvLm5v
†These authors have contributed equally to this work