Short-term and long-term high-fat diet promote metabolic disorder through reprogramming mRNA m⁶A in white adipose tissue by gut microbiota

Abstract

Background

Although short-term high-fat diet (S-HFD) and long-term high-fat diet (L-HFD) induce metabolic disorder, the underlying epigenetic mechanism is still unclear.

Results

Here, we found that both 4 days of S-HFD and 10 weeks of L-HFD increased mRNA m⁶A level in epididymal white adipose tissue (eWAT) and impaired metabolic health. Interestingly, S-HFD activated transposable elements (TEs), especially endogenous retroviruses (ERVs) in eWAT, while L-HFD activated long interspersed elements (LINEs). Subsequently, we demonstrated that both S-HFD and L-HFD increased m⁶A level of Ehmt2 and decreased EHMT2 protein expression and H3K9me2 level, accounting for activation of ERVs and LINEs. Overexpression of EHMT2 in eWAT or inhibition of ERVs and LINEs by antiviral therapy improved metabolic health under HFD feeding. Notably, we found that both short-term and long-term HFD feeding increased Fimicutes/Bacteroidota ratio and decreased the gut microbiome health index. Fecal microbiota transplantation (FMT) experiments demonstrated that gut microbiota from S-HFD and L-HFD was responsible for increased m⁶A level in eWAT, resulting in glucose intolerance and insulin insensitivity. Furthermore, we identified that both S-HFD and L-HFD increased the abundance of the gut microbial metabolite homogentisic acid (HGA), and HGA level was positively correlated with unclassified_f__Lachnospiraceae which was both increased in S-HFD and L-HFD feeding mice. Administration of HGA increased the m⁶A level of Ehmt2 and decreased the EHMT2 protein expression and H3K9me2 level in eWAT, leading to metabolic disorder in mice.

Conclusions

Together, this study reveals a novel mechanism that S-HFD and L-HFD induce metabolism disorder through gut microbiota-HGA-m⁶A-Ehmt2-ERV/LINE signaling. These findings may provide a novel insight for prevention and treatment of metabolism disorder upon short-term or long-term dietary fat intake.

Download video file ^{(106.1MB, mp4)}

Video Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-025-02047-4.

Background

White adipose tissue (WAT), including inguinal white adipose tissue (iWAT) and epididymal white adipose tissue (eWAT), traditionally thought to be simply a lipid storage site, is now widely recognized as a dynamic organ involved in energy metabolism [1]. In response to high-fat diet (HFD) feeding, WAT expands dramatically, involving both an increase in the size of adipocytes (hypertrophy) as well as an increase in de novo adipogenesis (hyperplasia) [2]. iWAT and eWAT differ in their developmental origin, anatomical location, and response to metabolic stimuli, in which eWAT serve as one of the most obesity-responsive tissues in mice. In response to obesity, eWAT expands dramatically and exhibits fibrosis and chronic inflammation, whereas these phenomena are not largely manifested in iWAT [3]. Numerous studies reported that even short-term (3–7 days) HFD feeding results in impaired glucose uptake in WAT as well as systemic insulin resistance [4–6]. In contrast, long-term intake of HFD feeding leads to chronic inflammation, insulin resistance, and obesity [7]. Despite studies demonstrated that both short-term HFD (S-HFD) and long-term of HFD (L-HFD) cause metabolism disorder, the underlying mechanism is still limited.

Accumulating evidence has demonstrated that environmental or nutritional effects can regulate systemic energy metabolism by different epigenetic mechanisms such as histone modifications, as well as DNA methylation [8]. More recently, it has been shown that mRNA m⁶A modification, the most abundant RNA methylation in eukaryote, plays a critical role in adipose tissue development and metabolic health [9, 10]. Maternal HFD intake could dynamically programmed mRNA m⁶A modifications in adipose and skeletal muscle tissues in offspring [11]. Supplementation of diet with nutrients, such as epigallocatechin-3-gallate and curcumin, have been shown to alter m⁶A RNA methylation of adipose and protect from obesity and metabolism disorder [12, 13]. However, the m⁶A profiles and function during S-HFD and L-HFD feeding have not been reported.

To explore the underlying mechanism for S-HFD and L-HFD-induced metabolism disorder, mice were fed with low-fat diet (LFD), 4 days of S-HFD or 10 weeks of L-HFD. We found that both S-HFD and L-HFD impaired metabolism health and increased m⁶A level in eWAT. Interestingly, compared to LFD, S-HFD decreased the RNA expression of genes related to response to virus as well as increased endogenous retroviruses (ERVs), while L-HFD increased more long interspersed elements (LINEs). Furthermore, S-HFD and L-HFD increased the mRNA m⁶A level of Ehmt2 and decreased EHMT2 protein expression as well as H3K9me2 level, thereby activating ERVs and LINEs, leading to glucose intolerance and insulin resistance. Finally, we revealed that both S-HFD and L-HFD increased the abundance of homogentisic acid, a key gut microbial metabolite, thereby increasing mRNA m⁶A level of Ehmt2 and decreasing the level of EHMT2 protein and H3K9me2 in eWAT, resulting in metabolic disorder. These findings provided a novel insight for prevention and treatment of metabolism disorder upon short-term or long-term dietary fat intake.

Methods

Animals

Six to 8 weeks old C57BL/6 mice in this study were purchased from Shanghai SLAC Laboratory Animal Co., Ltd (SLAC, China). Mice were housed and maintained at 22 ± 2 °C with a humidity of 35 ± 5% with 12 h light and dark cycles and free access to water and food. All research on animal study was carried out under approval of the Ethics Committee of Zhejiang University (ZJU20240321). All animal experiments were performed from March 21 to May 30, 2024.

Diet studies

For a low-fat diet (LFD) group, 6-week-old male C57BL/6 mice were fed LFD (10% fat in calories; Research Diets, D12450J, USA) for 70 days. For short-term high-fat diet (S-HFD) group, 6-week-old male C57BL/6 mice were fed LFD for 66 days and then fed a HFD (60% fat in calories; Research Diets, D12492, USA) for another 4 days. For long-term high fat (L-HFD) group, 6-week-old male C57BL/6 mice were fed a HFD for 70 days. At the end of 70 days of diet studies, GTT and ITT assays were performed to explore the role of gut microbiota on glucose tolerance and insulin sensitivity.

Glucose and insulin tolerance tests

For glucose tolerance test (GTT), mice fasted overnight were intraperitoneally injected with 2 g/kg glucose (Sangon Biotech, China). For insulin tolerance test (ITT), mice fasted overnight were intraperitoneally injected with 0.75 U/kg insulin (Beijing Solarbio Science & Technology, China). Blood glucose level was detected in tail blood at 0, 15, 30, 60, 90, and 120 min after glucose or insulin injection using a glucometer with glucose testing strips.

RNA isolation and m⁶A dot blot

Total RNA was extracted with RNAiso Plus reagent (Takara, Japan), then mRNA was isolated using the Dynabeads® mRNA DIRECT™ kit (Thermo Scientific, USA). After denaturing at 65 °C for 5 min, 200 ng mRNA was loaded on a hybond-N⁺ membrane (GE Healthcare, USA) and crosslinked by a UV Stratalinke at 1200 μJ twice. The membrane was washed with PBST for 5 min and blocked with 5% non-fat milk in PBST for 1 h. The membrane was incubated with anti-m⁶A antibody (Synaptic Systems, #202 003, Germany) overnight at 4 °C and then incubated with secondary antibody at room temperature for 1 h. The membrane was visualized using chemiluminescence (ECL Plus detection system) and stained by 0.1% methylene blue.

Western blot

Total protein was extracted with RIPA buffer (Beyotime Biotechnology, China) containing 1 × protease and phosphatase inhibitor (Beyotime Biotechnology, China). A total of 30 μg protein was loaded into 4–20% PrecastProteinPlusGel (Yeasen Biotechnology, China) and transferred to PVDF membranes (Life Technologies, USA). After blocking in 5% skim milk, the membranes were incubated with primary antibodies overnight at 4 °C, and then incubated with corresponding secondary antibodies for 1 h at room temperature. The antibodies were summarized below: rabbit monoclonal anti-EHMT2 antibody (ABclonal, #A19288, China); rabbit polyclonal anti-Histone H3 antibody (ABclonal, #A2348); rabbit polyclonal anti-H3K9me2 antibody (ABclonal, #A2359, China); rabbit monoclonal anti-beta Actin antibody (Servicebio, #GB15003, China).

Quantitative real‑time PCR analysis

Total RNA from cells or tissues were extracted using TRIzol reagent (TAKARA, Japan) and reverse-transcribed into cDNA using M-MLV reverse transcriptase (Thermo Fisher Scientific, USA). qPCR was performed using the SYBR Green PCR Master Mix (VAZYME, China) with the Rio-rad Real-Time PCR System (Bio-rad, USA). The data were analyzed following the 2^−ΔΔCt method. The primers were listed as follows. Ehmt2-qPCR-F: 5′-AGCCAAGAGGGGTCTCCAAT-3′; Ehmt2-qPCR-R: 5′-CTCGCTGATGCGGTCAATCT-3.

MeRIP-qPCR

Twenty-microgram RNAs were fragmented at 30 s on, 30 s off, 30 cycles in Bioruptor® Pico. One-tenth of the fragmented RNAs were saved as input control. A total of 0.5 μL m⁶A antibody (NEB, USA) was incubated with 20 μL Protein G Magnetic beads (NEB, USA) in 200 μL reaction buffer (10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.1% IGEPAL in nuclease free water) with orbital rotation at 4 °C for 1 h. After washing beads twice with 250 μL reaction buffer, resuspended with beads with 200 μL reaction buffer and added nine-tenth of the fragmented RNAs incubated. The beads-antibody-RNAs complex was incubated with orbital rotation at 4 °C for 4 h. Washed the beads-antibody-RNAs complex twice with 200 μL reaction buffer, 200 μL low-salt buffer (10 mM Tris–HCl, pH 7.5, 50 mM NaCl, 0.1% IGEPAL in nuclease free water) and 200 μL high-salt buffer (10 mM Tris–HCl, pH 7.5, 500 mM NaCl, 0.1% IGEPAL in nuclease free water) respectively. Finally, RNAs were eluted with RLT buffer (Qiagen, Germany) following ethanol precipitation. The IP and input control RNA were reverse-transcribed with random hexamers, and m⁶A enrichment was determined by qPCR. The data were analyzed following the 2^−ΔΔCt method, and the relative enrichment of m⁶A in each sample was calculated by normalizing to input. The primers were listed as follows. Ehmt2-MeRIP-qPCR-F: 5′-GGGTGGACTCTGACAGCAAG-3′; Ehmt2-MeRIP-qPCR-R: 5′-GACGTGTCATTGGAGACCCC-3′.

In vivo transfection

pcDNA3.1-Ehmt2-3 × Flag plasmid was injected into the eWAT of 8-week-old male C57BL/6 mice (20 μg/mouse) in the presence of in vivo DNA transfection reagent (Entranster-in vivo; Engreen, China).

Antivirus treatment of mice

Eight-week-old male C57BL/6 mice were provided a 60 mg/kg Emtricitabine (Aladdin, China) as described previously [14]. In detail, antiretroviral drug Emtricitabine was dissolved in water to 7.5 mg/mL, and Emtricitabine was administrated daily by gavage to mice in a total volume 200 μL starting at HFD feeding. After 1 week, GTT and ITT assays were performed to explore the role of gut microbiota on glucose tolerance and insulin sensitivity.

Antibiotic treatment and fecal microbiota transplantation

Antibiotic (Abx) treatment was performed as described previously [15]. Briefly, 8-week-old male C57BL/6 mice were administered an antibiotic cocktail containing ampicillin, neomycin, metronidazole, and vancomycin via oral gavage for 5 days. Fecal microbiota transplantation was performed as described previously [15]. Briefly, fresh feces were collected from LFD, S-HFD, and L-HFD feeding mice at the end of 10 weeks of LFD feeding, 4 days of HFD feeding or 10 weeks of HFD feeding. A total of 100 mg feces were suspended in 1 ml sterile PBS. The suspension was filtered by 80 mesh, 200 mesh and 400 mesh sterile gauze to obtain fecal microbiota suspensions. Abx-treated mice were daily gavaged with 200 μL of fecal microbiota suspensions. All mice were fed with NCD. After 1 week, GTT and ITT assays were performed to explore the role of gut microbiota on glucose tolerance and insulin sensitivity.

HGA treatment of mice

To determine the effects of HGA, 8-week-old male C57BL/6 mice were injected intraperitoneally with PBS or 250 mg/kg HGA (Sigma) daily for 4 days as described previously [16].

m⁶A-seq

m⁶A library construction was performed as previously reported [17]. Briefly, 1 μg mRNA in 100 μL nuclease-free water was fragmented by Bioruptor (30 s on/off for 30 cycles; Diagenode, Belgium). Ten percent of the fragmented mRNA was used as input and the remaining RNA was used for m⁶A IP using the EpiMark N6-Methyladenosine Enrichment Kit (NEB, USA). Both input and IP samples were used for library construction by using the VAHTS Universal V8 RNA-seq Library Prep Kit (Vazyme, USA). Library sequencing was performed at Azenta Life Sciences (China) on an Illumina NovaSeq machine in pair-end mode with 150 bp per read.

m⁶A-seq data analysis

Analysis was performed as described previously [18]. Raw reads were trimmed by Trimmomatic (v0.39) [19], and then aligned to the mouse genome (mm10) using STAR (v2.7.10a) [20]. m⁶A peaks were called using MACS2 (v2.1.1) [21] with the parameter “—nomodel.” The m⁶A peaks were annotated using mouse gene GTF file downloaded from the ENSEMBL database or transposable elements (TEs) GTF file downloaded from TEtranscripts website. Differentially analysis of m⁶A level was performed using QNB (v1.1.11) [22]. Motif enrichment analysis of m⁶A peak was performed with HOMER software (v4.11.1) in RNA mode [23], and GO enrichment of m⁶A-modified genes was performed using clusterProfiler R package [24]. TCseq R package (v1.30.0) was used to perform an unsupervised clustering analysis of m⁶A-modification as previous study reported [25]. Briefly, m⁶A level of each peak identified from LFD, S-HFD, and L-HFD were used for unsupervised clustering analysis by TCseq R packages. “cm” was set as the clustering method to generate clusters by using “timeclust” function of TCseq R packages. “z-score” was set to plot the clusters generated from “timeclust” by using timeclustplot function of TCseq R packages.

RNA-seq data analysis

For RNA-seq data analysis, raw reads were trimmed by Trimmomatic (v0.39), and then aligned to the mouse genome (mm10) using STAR (v2.7.10a) with default option. Reads count was calculated by featureCounts (v2.0.1) [26] and differentially analysis was performed by DESeq2 (v1.46.0) [27] with p value < 0.05 and |log2Foldchange|> 0.58. TEs analysis was performed as previously described [28]. Briefly, cleaned reads were aligned to mm10 using STAR (v2.7.10a) with the option –runMode alignReads, –winAnchorMultimapNmax 100, and –outFilterMultimapNmax 100. The abundance of TEs were identified by TEtranscript software (v 2.2.1) [29] with the option –mode multi, –minread 1, and -i 10. GTF files for TE annotation were downloaded from the TEtranscripts website. DESeq2 R package (v1.46.0) was used to perform differentially analysis of TEs with p value < 0.05 and |log2Foldchange|> 1.

16S rRNA sequencing and data analysis

Fresh feces (~ 200 mg) were collected from LFD S-HFD and L-HFD-fed mice. Fecal DNA was extracted from fresh fecal samples using E.Z.N.A.® soil DNA Kit according to the manufacturer’s protocols. Specific primers (338F, 5′-ACTCCTACGGGAGGCAGCAG-3′; 806R, 5′-GGACTACHVGGGTWTCTAAT-3′) were used to amplify V3–V4 region of 16S rRNA gene, and the amplification products were sequenced on Illumina MiSeq platform by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

Bioinformatic analysis of the gut microbiota was carried out using the Majorbio Cloud platform (https://cloud.majorbio.com). Sequencing reads were clustered into operational taxonomic units (OTUs) using UPARSE (v7.1) with 97% sequence similarity level [30]. RDP Classifier was used to analyze the taxonomy of OTU representative sequence. Based on the OTU information, alpha diversity indices were calculated with Mothur (v1.30.1) [31]. The similarity among the microbial communities in different samples was determined by principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity using Vegan (v2.5–3) package. The linear discriminant analysis (LDA) effect size (LEfSe) (http://huttenhower.sph.harvard.edu/LEfSe) was performed to identify the significantly abundant taxa of bacteria among the different groups (LDA score > 3.5, p < 0.05). The gut microbiota health index (GMH) was calculated as previously reported [32]. Run MetaPhlAn2 on stool metagenome(s) using the “–tax_lev s” argument and merge the MetaPhlAn2 outputs. Open the GMHI.R script and load the merged MetaPhlAn2 output file by providing the appropriate path. Then, run GMHI.R in its entirety.

Untargeted metabolomics analysis

Metabolites from fresh fecal samples collected from LFD S-HFD and L-HFD-fed mice were extracted with buffer (methanol:water, 4:1). The LC–MS/MS analysis of the metabolites was performed by Thermo UHPLC-Q Exactive HF-X system (Thermo Fisher Scientific, USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Data analysis was processed by using Progenesis QI (Waters). Bioinformatic analysis of the gut metabolites was carried out using the Majorbio Cloud platform (https://cloud.majorbio.com).

Statistical analysis

Data was presented as means ± SD unless otherwise indicated. Either Student’s t-test or ordinary one-way ANOVA methods were utilized to determine differences among multiple groups by using GraphPad software. p < 0.05 is considered as statistically significant.

Results

Both short-term and long-term HFD feeding induce metabolism disorder and increase m⁶A level in eWAT

To investigate the role of S-HFD and L-HFD feeding on metabolic health, 4 days of S-HFD and 10 weeks of L-HFD feeding assay (Fig. 1A) were performed. Body weight and body weight gain were increased in L-HFD compared to LFD, and S-HFD (Fig. 1B, C, and S1A). Consistently, the results of food intake showed that daily calorie intake was increased in L-HFD compared to LFD, and S-HFD (Fig. S1B). Moreover, L-HFD, not S-HFD, significantly increased fat ratio and tissue weight of iWAT, eWAT, and liver of mice (Fig. 1D–E and S1C). In addition, both S-HFD and L-HFD feeding increased the adipocyte size of iWAT and eWAT (Fig. 1F). Notably, both S-HFD and L-HFD induced glucose intolerance and insulin insensitivity of mice (Fig. 1G–J), even though the plasma insulin level increased in L-HFD instead of S-HFD compared to LFD (Fig. S1D). Our previous studies indicated that mRNA m⁶A play important role in metabolism regulation [9, 33]. To explore whether S-HFD and L-HFD feeding could affect the mRNA m⁶A level of WAT, m⁶A dot-blot was performed. The results showed that both S-HFD and L-HFD feeding increased the mRNA m⁶A level in iWAT and eWAT, especially in eWAT (Fig. 1K), suggesting a regulatory role of m⁶A on eWAT.

Fig. 1.

Both short-term and long-term high fat diet feeding induce metabolic disorder and elevate m⁶A level in WAT. A Mice were divided into three groups. Low fat diet (LFD): mice fed low fat diet for 10 weeks. Short-term high fat diet (S-HFD): mice fed low fat diet for 66 days, then fed high fat diet for 4 days. Long-term high fat diet (L-HFD): mice fed high fat diet for 10 weeks. B Representative photographs of mice in LFD, S-HFD and L-HFD group. C The growth curve of mice in LFD (n = 11), S-HFD (n = 11), and L-HFD group (n = 12). One-way ANOVA for multiple comparisons were used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. D The ratio of lean mass and fat mass in mice from LFD (n = 10), S-HFD (n = 10), and L-HFD group (n = 12). One-way ANOVA for multiple comparisons were used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. E Weight of iWAT and eWAT from LFD, S-HFD, and L-HFD group. One-way ANOVA for multiple comparisons were used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. F H&E staining of iWAT and eWAT from LFD, S-HFD, and L-HFD group. Scale bars, 100 μm. G Blood glucose levels of mice in LFD (n = 7), S-HFD (n = 7), and L-HFD (n = 8) group during an intraperitoneal injection glucose tolerance test (GTT). ^#P < 0.05, ^##P < 0.01, and ^###P < 0.001 were used to compare S-HFD and LFD. *P < 0.05, **P < 0.01, and ***P < 0.001 were used to compare L-HFD and LFD. H The area under the curve (AUC) was calculated based on GTT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. I Blood glucose levels of mice in LFD (n = 8), S-HFD (n = 8), and L-HFD (n = 12) group during an insulin tolerance test (ITT). ^#P < 0.05, ^##P < 0.01, and ^###P < 0.001 were used to compare S-HFD and LFD. *P < 0.05, **P < 0.01, and ***P < 0.001 were used to compare L-HFD and LFD. J The AUC was calculated based on ITT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. K mRNA m⁶A modification level of iWAT and eWAT from LFD, S-HFD, and L-HFD mice were detected by dot-blot. Methylene blue staining was used as a loading control. The quantitative data of dot-blot results was calculated by Image J

Activation of TEs is related to metabolic disorder phenotypes upon short-term and long-term HFD feeding

To investigate gene expression on transcriptome level in eWAT upon LFD, S-HFD, and L-HFD feeding, RNA-seq was performed. Differential gene expression analysis identified 3144 upregulated and 889 downregulated genes in S-HFD group when compared with LFD group (Fig. S1 E). Comparing with LFD group, 869 upregulated and 1165 downregulated genes in L-HFD group were identified (Fig. S1 F). Comparing with S-HFD group, 1577 upregulated and 4133 downregulated genes in L-HFD group were identified (Fig. S1 G). To explore the function of differentially expressed genes, Gene Ontology (GO) enrichment analysis of upregulated and downregulated genes was performed respectively. The results showed that S-HFD increased the expression of genes involving in protein localization to cilium, smoothened signaling pathway and cell cycle regulation (Fig. 2A) when compared with LFD, indicating that abundantly ciliated preadipocytes were activated. Compared with LFD, L-HFD increased the expression of genes involving in cytokine production and extracellular structure organization (Fig. 2B), suggesting an enhanced extracellular remodeling and proinflammatory response. Compared with S-HFD, L-HFD increased the expression of immune response process related genes (Fig. 2C). These findings were similar to previous study that inflammation in eWAT is not apparent after 1 week of HFD [5]. Interestingly, when compared with LFD, S-HFD decreased the expression of genes involving in phagocytosis, response to virus and defense response to bacterium (Fig. 2D), while L-HFD decreased the expression of genes associated with defense response to bacterium, negative regulation of proteolysis, and cofactor metabolic process (Fig. 2E). Not surprisingly, L-HFD impaired cilium-related process when compared with S-HFD (Fig. 2F).

Fig. 2.

Activation of TEs is related to metabolic disorder phenotypes upon short-term and long-term HFD feeding. A GO enrichment results of up regulated genes in S-HFD compared with LFD. B GO enrichment results of upregulated genes in L-HFD compared with LFD. C GO enrichment results of upregulated genes in L-HFD compared with S-HFD. D GO enrichment results of downregulated genes in S-HFD compared with LFD. E GO enrichment results of downregulated genes in L-HFD compared with LFD. F GO enrichment results of downregulated genes in L-HFD compared with S-HFD. G Differential TEs in eWAT between S-HFD and LFD. H Differential TEs in eWAT between L-HFD and LFD. I Number and ratio plot of differential TEs in eWAT from L-HFD and SHFD compared to LFD. J Transcription factors significantly associated with the TEs of interest. The associations were analyzed by RTFAdb database. K HuGE Score of EP300 in the T2D-related diseases and traits. HuGE Score < 1 indicates no evidence, ≥ 3 indicates moderate evidence, ≥ 10 indicates strong evidence, and ≥ 30 indicates very strong evidence. L HuGE Score of MAFK in the T2D-related diseases and traits. HuGE Score < 1 indicates no evidence, ≥ 3 indicates moderate evidence, ≥ 10 indicates strong evidence, and ≥ 30 indicates very strong evidence

It has been reported that HFD promotes endogenous retrovirus (ERVs) expression in keratinocytes [14], promoting us to ask whether S-HFD would induce transposable elements (TEs), such as ERVs, accounting for the observed response to virus. To characterize the TE expression in eWAT, TE expression analysis based on RNA-seq data was performed. The result showed that S-HFD activated the RNA expression of ERVs, such as MER50, LTR28, and LTR68 (Fig. 2G and S1H). In contrast, L-HFD specificity activated long interspersed elements (LINEs), such as L1Md_Gf, L1Md_F, and L1M3f (Fig. 2H and S1I). In total, 20 and 8 TEs with upregulation and downregulation in eWAT upon S-HFD feeding were identified. Twenty-three and 13 TEs with upregulation and downregulation in eWAT upon L-HFD feeding were found. Notably, LTR and LINE were the majority upregulation TEs in S-HFD and L-HFD fed mice, respectively (Fig. 2I). Recently, TEs have been gaining increasing attention due to their regulatory roles in gene expression, in which TEs influence transcriptional activity of nearby genes by providing binding sites for transcription factors (TFs) [34]. To identify the potential TFs interacted with those increased TEs, TEs–TFs association analysis by using RTFAdb, a public repository of the overrepresented TFs in the binding sites of the human and mouse TFs [35], was performed. The results indicated that TFs, such as EP300, TAL1, TCF12, and ZNF384, were significantly associated with MER50, an overrepresented retrotransposon in S-HFD fed mice, while MAFK and ZNF384 were significantly associated with L1Md, which was upregulated in L-HFD fed mice (Fig. 2J). To ask whether those TFs were related to metabolism regulation, Human Genetic Evidence (HuGE) Score of EP300 and MAFK were calculated in T2D Knowledge Portal (T2DKP) website [36]. The results showed that EP300 showed very strong evidence to type 2 diabetes (T2D), height and weight, and strong evidence to total cholesterol and PC3 dietary pattern (Fig. 2K), indicating that increased ERVs may contribute to metabolic disorder by interacting with TFs, such as EP300. In contrast, MAFK showed very strong evidence to serum albumin and moderate evidence to adiponectin, fasting insulin, γ-glutamyl transferase, and height (Fig. 2L). Previous studies reported that serum albumin plays a role in estimating inflammatory activity [37], and adiponectin was correlated to obesity and insulin resistance [38], suggesting that increased LINEs may contribute to inflammation and insulin resistance during obesity by interacting with TFs, such as MAFK. Together, these results suggested that S-HFD and L-HFD respectively induced ERVs and LINEs which may contribute to different signaling pathway leading to metabolism disorder.

HFD-induced m⁶A level increase of Ehmt2 is responsible for activation of TEs and impaired metabolism.

Recent studies suggested that m⁶A modifications on TE RNAs can regulate their stability and transcription [17]. Since mRNA m⁶A level in eWAT was increased upon S-HFD and L-HFD feeding (Fig. 1K), thus we guessed that m⁶A modification may participate in ERV or LINE regulation. To gain the transcriptome-wide m⁶A profiling of eWAT, m⁶A-seq of eWAT from LFD, S-HFD, and L-HFD groups were performed. Distribution analysis of m⁶A peaks at distinct region (5′ untranslated region, 5′UTR; coding sequences, CDS; 3′ untranslated region, 3′UTR) showed that the majority of m⁶A peaks was found in CDS region (Fig. 3A), and m⁶A peaks were especially abundant in the vicinity of start and stop codons (Fig. 3B). The most enriched consensus motif identified from m⁶A peaks was GGACU (Fig. 3C), which was conserved in the published consensus motif RRACH (where R = A/G, A = m⁶A, and H = A/C/G) [17, 18]. To test whether m⁶A regulated the TE expression, differential m⁶A peaks were annotated to TEs. However, the RNA abundance of most TEs with differential m⁶A level in S-HFD and L-HFD group was not differentially expressed when compared with LFD group (Fig. 3D, E). Consistently, there seems not obviously correlation between mRNA abundance and mRNA m⁶A level (Fig. 3F, G). These results suggested that m⁶A modification may not directly regulate TE activation in eWAT upon HFD feeding.

Fig. 3.

HFD-induced m⁶A level increase of Ehmt2 is responsible for activation of TEs and impaired metabolism. A Transcriptome-wide distribution of mRNA m⁶A peak. The bar chart shows the percentages of m⁶A peak within distinct regions: CDS, 5′UTR, and 3′UTR. B The m⁶A peak distribution pattern within mRNA in different regions: 5′UTR, CDS, and 3′UTR. C The most enriched consensus motif identified from m⁶A peaks on exons by HOMER software. D Veen plot of differentially expressed TEs and differentially methylated TEs between S-HFD and LHFD. E Veen plot of differentially expressed TEs and differentially methylated TEs between L-HFD and LFD. F Four-quadrant plots to show the mRNAs with a significant change both in m⁶A level and mRNA levels between S-HFD and LHFD. G Four-quadrant plots to show the mRNAs with a significant change both in m⁶A level and mRNA levels between L-HFD and LFD. H PCA results of based on m⁶A level in eWAT from LFD, S-HFD, and L-HFD group. I Unsupervised clustering analysis of m⁶A level in eWAT from LFD, S-HFD, and L-HFD group. J GO enrichment results of genes with increased m⁶A level in eWAT from S-HFD and L-HFD group compared to LFD group. K Veen plot of genes with increased m⁶A level in S-HFD or L-HFD group and genes involving in covalent chromatin modification. L m⁶A peak of Ehmt2 in eWAT from LFD, S-HFD, and L-HFD group was visualized by integrative genomics viewer. M MeRIP-qPCR results of m.⁶A level of Ehmt2 mRNA in eWAT from LFD, S-HFD, and L-HFD group. N qPCR results of Ehmt2 mRNA expression in eWAT from LFD, S-HFD, and L-HFD group. O Western blot results of EHMT2 protein expression and H3K9me2 level in eWAT from LFD, S-HFD, and L-HFD group. P HuGE Score of EHMT2 in the T2D-related diseases and traits. HuGE Score < 1 indicates no evidence, ≥ 3 indicates moderate evidence, ≥ 10 indicates strong evidence, and ≥ 30 indicates very strong evidence. Q Western blot results of EHMT2 protein expression in eWAT transfected with or without EHMT2 plasmid. R Blood glucose levels of mice in NCD (n = 6), S-HFD (n = 6), S-HFD + OE-EHMT2 (n = 6), and S-HFD + antivirus (n = 6) groups during GTT assay. S AUC was calculated based on GTT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. T Blood glucose levels of mice in NCD (n = 6), S-HFD (n = 6), S-HFD + OE-EHMT2 (n = 6), and S-HFD + antivirus (n = 6) groups during ITT assay. U The AUC was calculated based on ITT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001

Notably, principal component analysis (PCA) of m⁶A level of m⁶A peaks showed that S-HFD and L-HFD induced more similar m⁶A profile compared with LFD (Fig. 3H), which was consistent with the observed higher m⁶A level upon HFD feeding by dot-blot (Fig. 1K). Furthermore, a cluster of genes with higher m⁶A level both in S-HFD and L-HFD compared with LFD were identified (Fig. 3I), and these genes were significantly enriched in mRNA processing and covalent chromatin modification (Fig. 3J), in which chromatin modification, such as histone modification, was reported important for silencing TE expression [39, 40]. To identify whether there is a differential m⁶A-modified gene which could regulate the expression of TEs through chromatin modification, genes with increased m⁶A level upon HFD feeding and genes associated with covalent chromatin modification were overlayed, and Ehmt2, Kat2b, Cdh5, and Huwe1 with higher m⁶A level both in S-HFD and L-HFD group were screen out (Fig. 3K). It was noting that Ehmt2, a gene encoding protein as a H3K9me2 methyltransferase, was reported which could decrease the expression of ERVs [41] and LINE1 [42]. To determine the potential role of m⁶A modification in Ehmt2, IGV viewer was subsequently employed to check the m⁶A modification in Ehmt2. The results showed that the m⁶A modification level in CDS region of Ehmt2 in eWAT was much higher in S-HFD and L-HFD group compared to LFD group (Fig. 3L). MeRIP-qPCR results further demonstrated that the m⁶A level of Ehmt2 was significantly increased both in S-HFD and L-HFD group (Fig. 3M). Although the mRNA expression of Ehmt2 was not changed (Fig. 3N), EHMT2 protein expression were remarkably decreased regardless of S-HFD and L-HFD (Fig. 3O). In line with the decreased EHMT2 expression, H3K9me2 level in eWAT was significantly decreased upon S-HFD and L-HFD feeding when compared with LFD feeding (Fig. 3O). To illustrate the role of EHMT2 on metabolism regulation, huge score of EHMT2 was calculated in T2DKP website. The results suggested that EHMT2 showed strong evidence to total cholesterol, waist-hip ratio, HbA1c, and T2D and moderate evidence to obesity (Fig. 3P), indicating a potential role of EHMT2 on metabolism regulation. To explore the role of EHMT2 in eWAT, EHMT2 plasmid was transfected into eWAT of mice in vivo, and mice were fed normal chow diet (NCD). Western blot results suggested that EHMT2 was successfully overexpressed in eWAT (Fig. 3Q). The results of GTT and ITT assays showed that overexpression of EHMT2 in eWAT could alleviate glucose intolerance and insulin insensitivity upon S-HFD feeding (Fig. 3R–U).

Previous studies demonstrated that antivirus medicine, such as Emtricitabine, could inhibit the expression of ERVs [14] and LINEs [43]. To illustrate whether the increased ERVs or LINEs was the reason of impaired metabolic health upon S-HFD and L-HFD feeding, antivirus treatment of mice was performed. The results showed that inhibited ERVs or LINEs could improve the glucose tolerance and insulin sensitivity under S-HFD feeding (Fig. 3R–U). Together, these results revealed that S-HFD and L-HFD feeding increased the m⁶A level of Ehmt2 and decreased the EHMT2 protein expression and H3K9me2 level, thereby activating ERVs and LINEs, resulting in metabolic disorder.

Gut microbiota is responsible for increased m⁶A level in eWAT and metabolism disorder

Recently, numerous studies demonstrated that gut microbiota are involved in diet-induced obesity and metabolism disorder. To explore whether gut microbiota would play a role in S-HFD- and L-HFD-induced metabolism disorder, 16S rDNA sequencing of fecal samples from LFD and S-HFD and L-HFD group were performed (Figs. S2A and S2B, Tables S1 and S2). The principal coordinate analysis (PcoA) on family level of gut microbiota revealed a distinctly difference among LFD, S-HFD, and L-HFD fed mice (Fig. 4A). S-HFD and L-HFD changed the microbial composition on phylum level (Fig. 4B), especially increased Fimicutes/Bacteroidota (F/B) ratio (Fig. 4C), a potential marker indicating metabolism disorder in individuals and mice [44]. Besides, both S-HFD and L-HFD decreased the gut microbiome health index compared with LFD (Fig. 4D, E). To explore the differential gut microbiota, linear discriminant analysis (LDA) of gut microbiota was performed. Compared with LFD, S-HFD increased the abundance of uncultured_bacterium_g__norank_f__Desulfovibrionaceae and unclassified_f__Lachnospiraceae (Fig. 4F), while L-HFD increased abundance of Lachnospiraceae_bacterium_28-4 and Romboutsia_ilealis (Fig. 4G). These results suggested that both S-HFD and L-HFD disrupted the gut microbiota composition.

Fig. 4.

Gut microbiota is responsible for increased m⁶A level in eWAT and metabolism disorder. A PcoA plot on family level of 16S rDNA data from LFD (n = 9), S-HFD (n = 9), and L-HFD (n = 10) groups. B Phylum level taxonomy and relative abundance in 16S rDNA data from LFD, S-HFD, and L-HFD groups. C Fimicutes/Bacteroidota ratio in LFD, S-HFD, and L-HFD groups. D Gut microbiome health index between LFD and S-HFD groups. E Gut microbiome health index between LFD and L-HFD groups. F LDA scores of gut microbiota at species level in LFD and S-HFD groups. G LDA scores of gut microbiota at species level in LFD and L-HFD groups. H Representative photographs of iWAT and eWAT in Abx-LFD, Abx-S-HFD, and Abx-L-HFD groups. I Tissue weight of eWAT from Abx-LFD, Abx-S-HFD, and Abx-L-HFD groups. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. J H&E staining of eWAT in Abx-LFD, Abx-S-HFD, and Abx-L-HFD groups. Scale bars, 20 μm. K Mean adipocyte size of eWAT from Abx-LFD, Abx-S-HFD, and Abx-L-HFD groups. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. L Dot-blot analysis of mRNA m.⁶A level in eWAT from Abx-LFD, Abx-S-HFD, and Abx-L-HFD groups. Methylene blue staining was used as a loading control. The quantitative data of dot-blot results was calculated by Image J. M Blood glucose levels of mice in Abx-LFD (n = 5), Abx-S-HFD (n = 6), and Abx-L-HFD (n = 6) groups during GTT assay. N The AUC was calculated based on GTT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. O Blood glucose levels of mice in Abx-LFD (n = 5), Abx-S-HFD (n = 6), and Abx-L-HFD (n = 6) groups during ITT assay. P The AUC was calculated based on ITT results. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001

Since numerous studies demonstrated that gut microbiota could regulate the m⁶A modification in specific tissues [45, 46]. To study whether gut microbiota could affect the mRNA m⁶A modification in eWAT and impaired metabolism health, fecal microbiota from LFD, S-HFD, and L-HFD group were transplanted into antibiotic-treated (Abx) mice, and all mice were fed with NCD. Compared with LFD fed mice fecal microbiota recipient mice (Fig. S2C), mice received fecal microbiota from S-HFD and L-HFD increased the size of iWAT and eWAT (Fig. 4H), tissue weight, and adipocyte size in eWAT (Fig. 4I–K). More importantly, transplantation of fecal microbiota from S-HFD and L-HFD increased m⁶A level in eWAT (Fig. 4L), as well as induced glucose intolerance and insulin insensitivity (Fig. 4M–P). These results suggested that S-HFD and L-HFD increased m⁶A level in eWAT and impaired metabolism health by disrupting gut microbiota.

Gut microbiota-derived homogentisic increased m⁶A level in eWAT and impaired metabolism

Generally, the gut microbiota produces a myriad of metabolites that affect host physiology and metabolic health [47]. To identify potential gut microbial metabolites which could regulate m⁶A, untargeted metabolomics analyses of fecal samples from LFD, S-HFD, and L-HFD fed mice were performed. The results of partial least square discriminant analysis (PLS-DA) showed that metabolites were distinctly divided into LFD, S-HFD, and L-HFD group, in which metabolites from S-HFD and L-HFD clustered more closer (Fig. 5A). To gain the key metabolites in each group, metabolite variable importance in the projection value (VIP) analysis was performed, and top 30 differential metabolites were identified (Figs. S3A and S3B). To screen out potential metabolites related to metabolism disorder, the top 30 differential metabolites in this study were overlapped with previously reported obesity and T2D-related metabolites [48]. Interestingly, only homogentisic acid (HGA) was identified (Fig. 5B), and the abundance of HGA was much higher both in S-HFD and L-HFD group compared to LFD group (Fig. 5C). Consistently, the results of HGA concentration in feces showed that HGA was increased both in S-HFD and L-HFD compared to LFD (Fig. S3C). To explore which gut microbiota was responsible for HGA production, correlation analysis between gut microbiota abundance and HGA level was performed. The results showed that HGA level was positively correlated with unclassified_f__Lachnospiraceae which was increased in S-HFD and L-HFD feeding mice. Besides, Lachnospiraceae_bacterium_28-4 and Romboutsia_ilealis from L-HFD feeding mice were positively correlated with HGA level, and uncultured_bacterium_g__Dubosiella from LFD feeding mice was negatively correlated with HGA level (Fig. 5D).

Fig. 5.

Gut microbiota-derived homogentisic increased m⁶A level in eWAT and impaired metabolism. A PLS-DA for gut metabolites in LFD (n = 9), S-HFD (n = 9), and L-HFD (n = 10) group. B Overlap results between T2D-related metabolites and differential metabolites from S-HFD and L-HFD compared to LFD. C Abundance of HGA based on untargeted metabolomics results from LFD, S-HFD, and L-HFD groups. One-way ANOVA for multiple comparisons was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. D Heatmap plot of Spearman correlation results between gut microbiotas and HGA. E Dot-blot analysis of mRNA m⁶A level in eWAT from mice treated with or without HGA. Methylene blue staining was used as a loading control. F MeRIP-qPCR results of m.⁶A level of Ehmt2 mRNA in eWAT from mice treated with or without HGA. G qPCR results of Ehmt2 mRNA expression in eWAT from mice treated with or without HGA. H Western blot results of EHMT2 and H3K9me2 expression in eWAT from mice treated with or without HGA. I GTT results of 8-week-old mice treated with or without HGA for 3 days (n = 5 per group). J The AUC was calculated based on GTT results. Student’s t-test was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. K ITT results of 8-week-old mice treat with or without HGA (n = 5 per group). L The AUC was calculated based on ITT results. Student’s t-test was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001

To determine whether HGA plays a role in m⁶A regulation, 8-week-old male mice were injected intraperitoneally with 250 mg/kg HGA daily for 4 days as previously reported [16, 49, 50], and mice were fed with NCD. eWAT was isolated to analyze m⁶A level, and the results showed that HGA treatment increased the overall mRNA m⁶A level in eWAT (Fig. 5E) and the m⁶A level of Ehmt2 mRNA (Fig. 5F). Although HGA treatment did not affect the mRNA expression of Ehmt2 (Fig. 5G), HGA treatment increased the protein expression of EHMT2 and H3K9me2 level in eWAT (Fig. 5H), which was consistent with the results observed in eWAT from LFD, S-HFD, and L-HFD group (Fig. 3M–O). To validate the role of HGA on metabolism health, GTT and ITT assays were performed after 4 days HGA treatment. The results showed that HGA treatment significantly impaired glucose tolerance and insulin sensitivity (Fig. 5I–L). Taken together, these findings suggested that HGA served as the potential metabolite responsible for HFD-induced m⁶A level increase and EHMT2 expression decrease, leading to metabolism disorder regardless of S-HFD and L-HFD feeding (Fig. 5M).

Discussion

It has been demonstrated that acute lipid overload for 5 h or short-term (3–7 days) HFD intake results in impaired glucose uptake of adipose tissue and systemic insulin resistance [4, 5, 51]. The systemic insulin resistance steadily worsened with increasing adipose tissue inflammation and obesity upon L-HFD intake [52]. Nonetheless, our knowledge of the underlying mechanism of S-HFD- and L-HFD-induced metabolism disorder is still limited. Here, we demonstrated that both 4 days S-HFD and 10 weeks L-HFD feeding significantly impaired whole body glucose tolerance and insulin sensitivity. Moreover, we found that both S-HFD and L-HFD increased the m⁶A level in WAT, especially in eWAT. It has been reported that 3 weeks of mouse maternal HFD intake increased mRNA m⁶A modifications in eWAT in offspring, but 8 or 15 weeks of mouse maternal HFD intake showed opposite results [11]. Previous study suggested that the administration of curcumin significantly increased m⁶A level in iWAT and prevents HFD-induced obesity [13]. Thus, the alteration of m⁶A level in eWAT upon S-HFD and L-HFD feeding suggested a potential role of m⁶A on HFD-induced metabolism disorder.

Previous study suggested that the initial stage of HFD-induced insulin resistance is independent of inflammation, whereas the more chronic state of insulin resistance in established obesity is largely mediated by proinflammatory actions [4]. Ji et al. reported that S-HFD feeding induced remarkably pronounced immune responses and inflammation in eWAT, which may account for impaired metabolic health [53]. In this regard, our finding revealed that S-HFD did not induce obvious proinflammatory at transcript level, but increase nutrition sensing ability of adipose tissue by promoting cilium-related function and cell cycle. S-HFD was reported to promote proliferation capacity of preadipocyte and adipose tissue hyperplasia [54]. As for L-HFD, a proinflammatory response was found in eWAT at transcript level, which was consistent with previous studies [55, 56]. To our surprise, a function of response to virus was found in eWAT after S-HFD feeding, which was not reported before. Recently, HFD was reported to promote ERV expression involving in response to virus in keratinocytes [14]. Here, we revealed that S-HFD-induced expression of ERVs, whereas L-HFD activated expression of LINEs. In fact, numerous studies suggested that retrotransposons contained cis-regulatory sequences binding to TFs to regulate gene expression in diseases and tissue development [57]. Fresquet et al. found that epigenetic therapy killed cancer cells by rewiring mitochondrial metabolism upon ERV activation [58]. Another study reported that specific ERV function as active enhancers to drive germline genes during mitosis-to-meiosis transition in male mice [59]. Hansen et al. revealed that several waist-to-hip ratio-associated variants map within primate-specific Alu retrotransposons harboring a DNA motif associated with adipocyte differentiation [60]. Our findings suggested that S-HFD-induced ERVs showing potential interaction with TFs, such as EP300 involving in T2D disease. EP300 functions as histone acetyltransferase that regulates transcription via chromatin remodeling and is important in various biological processes. EP300-mediated acetylation of histones alters global chromatin structure and gene expression, promoting the development and progression of fibrosis [61]. Additionally, knockdown of EP300 in human aortic endothelial cells suppressed diabetes-related genes including CD58, SIGIRR, CD200, ICAM2, and CCL2 [62]. However, L-HFD induced LINEs showing potential association with TFs, such as MAFK which was related to inflammation. It was known that MafK positively regulates NF-κB activity by enhancing CBP-mediated p65 acetylation facilitating recruitment of p65 to NF-κB promoters such as IL-8 and TNFα [63]. Thus, EP300 and MAFK may promote the expression of obesity-related genes to induce metabolic disorder through transcriptional regulation. These results indicated that activation of specific TEs may play an important role in metabolism upon HFD feeding.

Generally, TEs were silenced by deposit of DNA and histone modification, such as DNA methylation and H3K9me2, at TEs loci, which was mediated by TE regulators [59]. In this study, we found that both S-HFD and L-HFD feeding could significantly increase m⁶A level of Ehmt2 mRNA and decrease EHMT2 protein expression as well as H3K9me2 level in eWAT. It has been reported that EHMT2, the H3K9me2 methyltransferase, suppressed both ERVs [41] and LINE1 elements [42]. Thus, the specific TE expression in this study may activate by the removal of H3K9me2 mediated by downregulated EHMT2 expression. Other studies reported that EHMT2 was markedly decreased in the liver of db/db mice and HFD-fed mice, thereby impairing hepatic insulin signaling [64]. Hepatic Ehmt2 knockdown exacerbated Dex-induced glucose and insulin intolerance [65]. Muscle-specific KO of Ehmt2 protected against obesity under HFD stress in female but not in male mice [66]. Removal of H3K9me2 in adipocytes by EHMT2 deletion enhanced chromatin opening and binding of the early adipogenic transcription factor Cebpb to Pparg promoter, thus promoting adipogenesis [67]. Loss of EHMT2 in adipocytes exacerbated TNFα’s deleterious effects on inflammatory gene expression and lipolysis [68]. In immune cells, such as macrophages, EHMT2-dependent H3K9me2 was associated with gene repression during endotoxin tolerance [69], suggesting that EHMT2 plays a critical role in immune cell function. More in-depth research about the specific role of EHMT2 in adipocytes and immunes will be helpful to understand the underlying mechanism contributing metabolic disorder upon HFD feeding. Nonetheless, our results suggested that overexpression of EHMT2 in eWAT could alleviate HFD-induced metabolism disorder. Together, these results provided a potential evidence linking dietary fat, metabolism, and TEs, where HFD-induced m⁶A reprogramming modulate the TEs activation by EHMT2 resulting in metabolic disorder.

Recently, more and more studies demonstrated that HFD-feeding induces obesity and the epigenetic modification of host tissues by altering gut microbiota. Wang et al. found transcriptome-wide reprogramming of mRNA m⁶A modification by the mouse microbiome [45]. Another study showed the impact of the gut microbiota on the mRNA m⁶A of mouse cecum and liver [46]. In this study, we found that both S-HFD and L-HFD could significantly shape the gut microbiota composition and increase F/B ratio which was previously reported serve as a biomarker of obesity and metabolic disorder [44]. Furthermore, we identified HGA, a gut microbial metabolite which was significantly increased both in S-HFD and L-HFD group, was responsible for increased mRNA m⁶A level of Ehmt2 and decreased EHMT2 expression as well as H3K9me2 level, leading to impaired metabolic health. However, the underlying mechanism of HGA on m⁶A level regulation is unclear. It has been demonstrated that HGA could be transformed to the TCA cycle intermediates such as fumaric acid and succinic acid [70] which was reported to inhibit m⁶A demethylase including ALKBH5 and FTO, providing a potential mechanism of HGA on m⁶A regulation [71]. In human, the expression of m⁶A regulators correlated with human obesity [72]. TEs, such as LINE-1, in human visceral adipose tissue is associated with metabolic syndrome status and related phenotypes [73]. Besides, HGA showed the strongest positive associations with body mass index, waist circumference, and increased T2D risk [74]. Despite the advancement we found in mice, more in-depth research in human samples needs to be conduct to study the integrated mechanisms about diet, m⁶A, gut microbiome in regulating human metabolic health.

Conclusions

In summary, we demonstrated that both short-term and long-term HFD feeding increased m⁶A level in WAT and systemic metabolic disorder. Moreover, we provided an evidence that HFD increased m⁶A level of Ehmt2, thereby downregulated EHMT2 expression and H3K9me2 level, and activated TE expression, resulting in metabolic disorder. Finally, we found that gut-microbiota derived HGA serve as the key factor for HFD-induced m⁶A epitranscriptome reprogramming and impaired metabolic health. Our findings provided a novel mechanism and function of m⁶A-mediated TE activation in metabolic disorder, highlighting the interaction role of host mRNA modification and gut microbiota on metabolic health regulation.

Abbreviations

Abx

Antibiotic treated

CDS

Coding sequences

EHMT2

Euchromatic histone lysine methyltransferase 2

ERVs

Endogenous retroviruses

eWAT

Epididymal white adipose tissue

GO

Gene ontology

GTT

Glucose tolerance test

HFD

High-fat diet

HGA

Homogentisic acid

HuGE

Human genetic evidence

ITT

Insulin tolerance test

iWAT

Inguinal white adipose tissue

LEfSe

Linear discriminant analysis effect size

LFD

Low-fat diet

L-HFD

Long-term HFD

LINEs

Long interspersed elements

m⁶A-seq

M⁶A sequencing

PCA

Principal component analysis

PcoA

Principal coordinate analysis

PLS-DA

Partial least squares discriminant analysis

qPCR

Quantitative real‑time PCR

RNA-seq

RNA sequencing

S-HFD

Short-term HFD

T2D

Type 2 diabetes

T2DKP

T2D knowledge portal

TEs

Transposable elements

TFs

Transcription factors

UTR

Untranslated region

Authors’ contributions

LYH., LJQ., RRT., XZM., LYJ., CYS., HCQ., LYX. and YTYD. performed experiments under the supervision of WXX. LYH. and LJQ. performed bioinformatics analysis of sequence data. LYH. wrote the manuscript under the supervision of WXX. WXX. designed the project and provided the final approval of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (32330098), the Science and Technology Innovation Leading Talent Project of Zhejiang Province (2022R52023), the National Natural Science Foundation of China (U21A20249), and the National Key R&D Program of China (2023YFD1301303).

Data availability

The m6A-seq datasets (CRA011918, https://ngdc.cncb.ac.cn/gsa/s/f0FDR4S1) have been deposited in the Genome Sequence Archive linked to the project PRJCA018477. The 16S rDNA datasets (CRA011935, https://ngdc.cncb.ac.cn/gsa/s/r0dY3Uu3) have been deposited in the Genome Sequence Archive linked to the project PRJCA018487. Untargeted metabolomic datasets (OMIX004571, https://ngdc.cncb.ac.cn/omix/preview/ImckzEo4) have been deposited in OMIX linked to the project PRJCA018487.

Declarations

Ethics approval and consent to participate

All experimental procedures used in this study were approved by the Committee on Animal Care and Use and Committee on the Ethics of Animal Experiments of Zhejiang University.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.