Landscape of transcriptome‐wide m6A modification in diabetic liver reveals rewiring of PI3K‐Akt signaling after physical exercise

Type 2 diabetes mellitus (T2DM) is one of the most common diseases, and epigenetic modification N6‐methyladenosine (m6A) is essential for transcriptional modulation involved in its development. However, the precise role and landscape of transcriptome‐wide m6A alterations in molecular adaptations after physical exercise have yet to be fully elucidated.


| INTRODUCTION
Type 2 diabetes mellitus (T2DM), characterized primarily by insulin resistance (IR) and hyperglycemia, is a critical risk factor for impaired physical and mental health.IR is a serious condition in which the body lacks sufficient insulin to effectively regulate normal functioning of insulin-responsive organs. 1 The liver is a major organ in the endocrine system responsible for controlling glucose and lipid metabolism in physiological and pathological conditions, including obesity, diabetes, metabolic syndrome, and liver dysfunction. 2As the liver continuously manufactures glucose, excess lipids worsen liver IR, leading to an elevated fasting blood glucose, triggering increased insulin levels, and thus causing obesityrelated metabolic disorders. 2,3The inability to inhibit glucose production in the liver due to IR is the major etiological risk factor for hyperglycemia in T2DM. 4,5IR is regulated by several mechanisms, of which the PI3K-Akt pathway is the most prominent. 6This pathway plays a vital role in controlling insulin signaling and is considered an attractive therapeutic target. 7Abnormalities in this pathway can lead to improper glucose regulation and thus high blood glucose levels and IR. 8 Proper activation of this pathway is necessary for insulin to work properly, but kinase inhibitors directly targeting PI3K-Akt and its downstream effectors might disrupt the metabolic balance. 9Therefore, understanding the precise regulatory role of PI3K-Akt pathway activation in IR is significant for identifying pathogenic and therapeutically relevant target genes.
Regular physical activity improves chronic metabolic diseases caused by physiological and biological changes in the body. 10Along with a controlled diet and medication, physical exercise is part of an effective nonpharmacological approach for treating diabetes that also includes regulating body weight, controlling blood glucose, and ameliorating IR. 11,12 Physical exercise activates hepatocytes to trigger a wide variety of molecular adaptations in response to the regulation of energy metabolism, including normalization of abnormal lipid panels, elevated hepatic glucose disposal, suppression of inappropriate regulation of liver glucose production, and amelioration of peripheral IR. [13][14][15] Note that physical exercise would activate the PI3K-Akt signaling thus improve the utilization efficiency of insulin. 16Based on exercise-stimulated alterations to the hepatic transcriptome, adaptive gene in the PI3K-Akt pathway shows a more significant mechanism as a transducer, yet the detailed relationship between exercise and the PI3K-Akt pathway is poorly understood.An investigation of their relation to training stimuli is critical to improving the design of exercise-based interventions.
Epigenetic mechanisms can control gene expression patterns without changes to the DNA sequence. 17,18A kind of epigenome, N6-methyladenosine (m 6 A) emerges an abundant transcriptome modification with various aspects of RNA. 19The methylation of m 6 A modification has critical roles in various biological events, such as mRNA splicing, stability, export, localization, and translation. 20his reversible process can be modified by three types of functional protein: m 6 A methyltransferases, demethylases, and m 6 A-binding proteins. 213][24] However, little is known about the differential responses of RNA methylation to exercise during the formation and development of T2DM.
In this study, we subjected mice fed a high-fat diet to physical exercise to investigate the health-promoting properties of exercise under HFD conditions.Conjoint MeRIP-seq and RNA-seq was used to examine the effects of physical exercise on mRNA methylation and gene expression profiles in hepatic tissue.The study reveals the molecular adaptations to exercise in the m 6 A epigenetic regulatory network in the upstream nodes of the PI3K-Akt signaling pathway and, more important, establishes a rational theoretical basis for recommending physical exercise as a therapy for metabolic disorders associated with IR.Investigating the potential mechanism that underlies the relationship between physical exercise and epigenetic modification holds great promise positively regulated the expression of two upstream genes (Itga3 and Fgf21) in an m 6 A-dependent manner.for identifying novel therapeutic target genes and further developing the science of personalized exercise medicine.

| Physical exercise triggers a dynamic change in hepatic RNA m 6 A
To investigate the potential of m 6 A methylome related to exercise in obese mice, we evaluated the change in m 6 A modification in liver tissue.First, the effects of physical exercise on mice were examined.As shown in Figure S1A, 12 weeks of exposure body weight increased significantly in HFD mice compared to NFD mice.However, a small reduction in body weight (−10%) was found in HFD + EX group.At the end of feeding, an oral glucose tolerance test (OGTT) was examined during the 12th week of intervention.It revealed that mice subjected to exercise training exhibited greater glucose tolerance compared to HFD mice (Figure S1B).HFD mice showed an elevated fasting glucose concentration as compared with the control group, and this concentration was effectively blocked by exercise training (Figure 1A).Similar results were also observed for fasting insulin levels in the HFD + EX mice (Figure 1B).In contrast to the NFD and the HFD groups, the HFD + EX group showed an approximately 3-fold increase in relative hepatic TG, and exercise effectively reversed this increase (Figure S1C).Meanwhile, hepatic H&E staining displayed consistent changes among the three groups, further indicating the beneficial effects of exercise intervention on fatty liver (Figure S1D).
Next, total m 6 A-methylated RNA was detected.As shown in Figure 1C, RNA m 6 A modification was observed in all groups of liver tissue.The relative m 6 A level in HFD mice was significantly elevated as compared to the control mice, and it showed a slight increase in the HFD + EX group.To further validate the presence of m6A modifications on the mRNAs of mouse liver, the LC-MS/MS was applied to quantify the m 6 A methylome on mRNAs in different groups.As shown in Figure 1D, the percentage of mRNA m 6 A/A modifications in HFD mouse liver showed an upregulation as compared to the NFD mouse, and its level was also observed increased in HFD + EX mouse in contrast to HFD mouse, which was consistent with the m 6 A relative quantification method.To further uncover the dynamics of m 6 A factors, the relative mRNA level of several critical methyltransferases and demethylases was examined (Figure 1E).The relative mRNA level of Mettl3, Mettl14, and Alkbh5 in the HFD mice was markedly upregulated compared to the control mice, and only Fto expression was markedly downregulated.Compared to HFD mice, Rbm15 and Fto were significantly elevated in HFD + EX mice.Moreover, western-blot analysis showed that the protein expression of Mettl3, Mettl14 and Rbm15 in the HFD mice was markedly upregulated as compared to the NFD mice.Whereas exercise training mainly significantly promoted Rbm15 expression as compared with the HFD mice (Figure 1F).Together these results strongly support the idea that Rbm15-mediated m 6 A methylation may play a regulatory role in an exercise intervention.

| Physical exercise alters transcriptome-wide hepatic m 6 A modification profiling
Transcriptome-wide MeRIP-seq was used to investigate the deposition of m 6 A peaks and the dynamics of m 6 A. The MeRIP-seq analysis revealed a total of 15 239, 18 438, and 18 973 m 6 A peaks in the NFD, HFD, and HFD + EX groups, representing 8338, 9142, and 9231 genes, respectively.Consistent with the LC-MS/MS analysis, HFD can increase the level of hepatic m 6 A/A methylome, and exercise training further promote this level in a slight intensity.It should be noted that, exercise can indeed evoke hepatic m 6 A methylation modification changes in genes and further influence the corresponding cell signaling processes and physiological function, thus contributing to a number of adaptive responses in physical fitness.Here, the identified m 6 A peaks showed typical enrichment near the start codon (startC), stop codon (stopC), and coding sequence (CDS), although a slight difference existed among these three groups (Figure 2A).The abundance of m 6 A peaks in the startC region was 21.12%, 20.96%, and 20.55% in the NFD, HFD, and HFD + EX groups, respectively.The abundance of m 6 A peaks near the CDS region increased 0.75% from the NFD to the HFD group and then continued increased 0.84% from the HFD to the HFD + EX group.Finally, Abundance of m 6 A peaks in the stopC region decreased 0.62% in the HFD group (27.25%), followed by a 0.90% decrease in the HFD + EX group (26.35%).
In addition, we estimated the relative positions of m 6 Aassociated reads along the transcripts in the whole transcriptome (Figure S2).Consistent with m 6 A peaks, m 6 A reads were abundant throughout all mRNA transcripts that the reads promoted from the CDS region and reached at the 3′ untranslated region (UTR).Specifically, the density of m 6 A reads in the CDS region of NFD group was higher than that in HFD + EX group, followed by in HFD group.Moreover, in the 3′ UTR region, the density of m 6 A reads in HFD group was higher than that in NFD and HFD + EX groups (Figure S2).Obviously, these results suggest that hepatic m 6 A is present as a dynamic alternation during exercise training.Moreover, these m 6 A peaks with mRNA were characterized by consensus RRACH sequences.As shown in Figure 2B, GGACU was a top motif in all samples tested, which implies that the adopted RRACH motif was conserved during physical exercise. 25,26It is important to note that, as a top motif in the liver tissue of obese mice, GGACU is a prevalent m 6 A-modified sequence, as shown in a previous study. 27

| m 6 A-enriched genes participate in important biological functions and pathways
A total of 5024, 6757, and 5477 methylated genes were discovered in the liver of the NFD, HFD, and HFD + EX groups, respectively.We investigated these genes that contained altered m 6 A peaks to uncover deeper insight into m 6 A in an exercise intervention.In a comparison of the HFD and NFD groups, 7523 genes were identified as differentially methylated genes (DMGs), of which 3586 were upmethylated and 3937 were down-methylated (Figure 3A).In the HFD + EX and HFD groups, 7763 genes were identified as DMGs, of which 5723 were up-methylated and 2040 were down-methylated.An overlapped analysis of m 6 A-modified genes in the exercise intervention indicated that a total of 1109 genes were up-methylated during HFD feeding and then down-methylated after exercise training, whereas 2293 genes were down-methylated during HFD feeding and up-methylated after physical exercise.
Next, we used the hypermethylated genes for GO functional analysis and KEGG pathway analysis to uncover the possible functional significance of m 6 A-enriched genes in the mice exposed to the exercise intervention.The GO analysis indicated that metabolic process, biological regulation, and cellular process were the top three enriched biological processes (Figure 3B).The KEGG analysis uncovered that the up-methylated genes in the comparison of the HFD and NFD groups were mostly related to Rap1 pathway and PI3K-Akt pathway, whereas the downmethylated genes played significant roles in protein digestion and absorption (Figure 3C).In the comparison of the HFD + EX and HFD groups, the up-methylated genes participated primarily in PI3K-Akt pathway and cytoskeleton regulation, and the down-methylated genes were significantly related to the MAPK pathway and insulin secretion (Figure 3D).These overlapped m 6 A-methylated transcripts were preferentially enriched in these signaling pathway and are critical factors in the regulation after physical exercise. 28,29

| Hepatic transcriptome is enriched in several signaling pathways
To examine the regulatory role of m 6 A, we conducted RNA-seq analysis of tested liver tissue and further analyzed differentially expressed genes (DEGs) among the NFD, HFD, and HFD + EX groups.As shown in Figure 4A, 1453 DEGs were detected between the HFD and NFD groups, of which 764 were upregulated and 689 were downregulated.A comparison of the HFD + EX and HFD groups revealed 1586 DEGs, of which 1349 were upregulated and 237 were downregulated.A heat-map revealed the profile of DEGs in the NFD, HFD, and HFD + EX groups (Figure 4B).KEGG pathway analysis further uncovered that these upregulated genes in the HFD and NFD groups were significantly involved in osteoclast differentiation and P53 pathway, whereas the downregulated genes in these groups were mostly related to amino acid biosynthesis (Figure 4C).In the comparison of the HFD + EX and HFD groups, the upregulated genes were primarily associated with the PI3K-Akt and TNF signaling pathways, whereas the downregulated genes modulated the metabolic process (Figure 4D).Based on this analysis of m 6 A-mediated processes, we suppose that m 6 A modification might affect gene expression, thus regulating HFDinduced IR and metabolic abnormality.

| DMGs and DEGs are both involved in regulating insulin signaling
To investigate whether m 6 A modification affects gene expression, we evaluated DMGs and DEGs between two adjacent stages in a paired comparison analysis.Of the m 6 A up-methylated genes, 457 and 839 genes were upregulated from the NFD to the HFD group and from the HFD to the HFD + EX group, respectively, whereas 21 and 22 genes were downregulated from the NFD to the HFD group and from the HFD to the HFD + EX group, respectively (Figure 5A).Of the m 6 A down-methylated genes, 101 and 60 genes were downregulated from the NFD to the HFD group and from the HFD to the HFD + EX group, respectively, whereas 336 and 140 genes were upregulated from the NFD to the HFD group and from the HFD to the HFD + EX group, respectively (Figure 5B).Thus, m 6 A mainly activated correlated gene expression after physical exercise.
Furthermore, a KEGG pathway analysis was conducted on these genes that were both DEGs and DMGs.The overlapped m 6 A-methylated transcripts were mostly enriched in cancer and PI3K-Akt pathway, whereas overlapped expressed genes were related to the regulation of metabolism and the PI3K-Akt pathway (Figure 5C,D).
Note that these overlapped m 6 A-methylated transcripts and mRNA expression genes were both preferentially enriched in the PI3K-Akt pathway, a significant factor for the development of diabetes. 8We next collected these DEGs and performed heat map analysis to further probe their profiles in the NFD, HFD, and HFD + EX groups.Totally 87 DEGs were detected, of which 47 were upregulated and 40 were downregulated (Figure 5E).It is intriguing that these DMGs and DEGs were involved in all stages of the pathway, including core factors, upstream molecules, and downstream effectors.Next, we screened possible genes associated with the PI3K-Akt signaling pathway (Figure 5F).Activation of this pathway is related to certain upstream families, including the growth factors (GFs), receptor tyrosine kinase (RTK), extracellular matrix (ECM), and integrin; the core factors are involved in PI3K (PIK3CG), IRS1, and AKT; and the downstream effectors preferentially constitute the TSC2/mTOR, GSK3/ MYC, FOXO/Bim, and CREB/BCL2 pathways.Together these results reveal that m 6 A mediates the expression of gene involved in the PI3K-Akt pathway to control insulin signaling.

| The PI3K-Akt signaling pathway mediates exercise-induced improvements in insulin sensitivity
Convincing evidence proves that the PI3K-Akt pathway is required for normal metabolism because of its multiple functions, and an imbalance in this pathway leads to obesity and T2DM. 30To determine whether m 6 A mediates gene expression in the PI3K-Akt pathway in an exercise intervention, we used qRT-PCR to evaluate the genes regulated by m 6 A modification.Upstream molecules in the PI3K-Akt pathway were screened by conjoint MeRIPseq and RNA-seq and found to be mainly involved in four families (Figure 6A).In the whole transcriptome of these genes, several m 6 A reads were detected throughout the mRNA transcripts (Table S1), and their m 6 A peak distribution throughout the transcript body in different samples were shown by Integrative Genomics Viewer (IGV) (Figure S3).Of these genes, Reln, Thbs1, Lama4, and Lamb3 belong to the ECM family that is critical for βcell survival and function in T2DM. 31The mRNA level of Reln was markedly downregulated and that of Thbs1, Lama4, and Lamb3 was upregulated in the HFD as compared with the NFD group, and Reln and Lama4 reversed after physical exercise.3][34] Itga3, Itga5, Itgb3, and Itgb7 were markedly suppressed in the HFD group in contrast to the NFD group, and only Itga3, Itgb3, and Itgb7 were markedly promoted in a comparison of the HFD + EX and HFD groups.Egf, Vegfb, Pdgfb, and Fgf21 are GFs and participate in a wide array of physiological processes.Egf, Vegfb, and Fgf21 were markedly upregulated in the HFD group in contrast to the NFD group, whereas Egf was markedly downregulated and Fgf21 drastically upregulated in a comparison of HFD + EX and HFD groups.Insr, Pdgfrb, and Fgfr4 are RTKs, which are critical regulators of Akt.Pdgfrb markedly inhibited in the HFD as compared with the NFD group but markedly elevated after the exercise intervention.][37] Given the fact that research on adaptation to exercise in upstream nodes remains limited, we subsequently examined the protein expression of these significantly changed genes associated with the upstream node of the PI3K-Akt pathway.Here, we included the NFD plus exercise group in order to effectively distinguish the effects of exercise adaptations between normal-diet and highfat-diet status.As shown in Figure 6B, compared with the NFD group, exercise training only suppressed Lama4 expression and induced Fgf21 expression under physiological condition.It can be seen that exercise training only altered some genes expression.Moreover, the protein level of Lama4, Egf, and Fgf21 was consistently markedly promoted compared to the HFD and NFD groups, whereas the level of Itga3, Itgb3, and Pdgfrb was markedly suppressed (Figure 6B).In a comparison of the HFD + EX and HFD groups, the expression of these genes was reversed, except for Fgf21.Expression of Fgf21 was drastically increased after physical exercise, as in previous studies. 2,38ollectively, these results imply that physical exercise positively regulates the upstream nodes of the PI3K-Akt pathway in the HFD status, which exerts an important role in modulating hepatic IR triggers T2DM.

| Physical exercise supports Rbm15 in regulating upstream genes in an m 6 A-dependent manner
To examine whether m 6 A mediates expression of upstream molecules after physical exercise, we performed MeRIP-qPCR to analyze m 6 A modification patterns on Lama4, Itga3, Itgb3, Fgf21, and Pdgfrb transcripts.We found that Itga3 and Pdgfrb were significantly downregulated from the NFD to the HFD group and then upmethylated from the HFD to the HFD + EX group.In contrast, Fgf21 was markedly upregulated from the NFD to the HFD group and continuously up-methylated from the HFD to the HFD + EX group (Figure 7A).The findings reveal that the expression of these three upstream genes in the PI3K-Akt pathway is strongly positively correlated with m 6 A dynamic modification after physical exercise.
To further confirm the regulatory role of m 6 A, we knocked down an m 6 A methyltransferase Rbm15 in HepG2 cells by a siRNA.MeRIP-qPCR revealed that in vitro transfection of Rbm15 siRNA led to a significant decrease in m 6 A in mRNA (Itga3 and Fgf21) in contrast to the scrambled control (Figure 7B).A similar result was observed for qRT-PCR and indicated that the relative mRNA of these two genes showed significant downregulation in HepG2 cells after Rbm15 knockdown (Figure 7C).Moreover, we conducted an RNA stability assay to examine the relationship between m 6 A and their mRNA stability.The mRNA level of Itga3 and Fgf21 was inhibited in Rbm15-silenced cells after actinomycin D treatment, which suggests that Rbm15 knockdown resulted in a reduction in their mRNA stability (Figure 7D).Together these results suggest that Rbm15 regulates upstream genes of the PI3K-Akt pathway after physical exercise in an m 6 A-dependent manner.

| DISCUSSION
With continuous advances in research technology, evidence related to m 6 A modification mounts.As the most prevalent internal mRNA modification, m 6 A is considered a new layer of epitranscriptomic gene regulation in various biological processes. 21,39Convincing evidence demonstrates that physical exercise remodels the hepatic transcriptome and improves insulin sensitivity and the transcriptome-wide m 6 A methylome responses that lead to such adaptations.However, the role of early transcriptional control in regulating hepatic adaptation to exercise has not been characterized.To elucidate the mechanism of m 6 A in an exercise intervention, we combined the m 6 A methylome and the transcriptome and found a dynamic change in m 6 A in the transcripts of liver organs.An investigation of the potential mechanism of exercise might provide a novel perspective on this intervention as a new strategy for treating HFD-induced T2DM and related metabolic disorders.
Here, we used high-throughput sequencing to investigate genome-wide profiling of m 6 A modification in mouse liver.MeRIP-Seq revealed an abundance of m 6 A throughout mRNA transcripts in the exercise intervention.m 6 A reads were distributed predominantly on the consensus GGACU motif, and m 6 A peaks in the NFD and HFD groups were mostly enriched around the CDS region and then decreased in the 3′ UTR, which is consistent with previous reports. 27This distribution pattern also existed after the exercise intervention, and it slightly promoted the density of reads in the CDS.In general, m 6 A in the CDS region is closely related to post-translational gene regulation, and m 6 A in the 3′ UTR preferentially interacts with factors affecting splicing, translocation, stability, and translation of RNA. 25,26Thus, dynamic changes in the m 6 A methylome across landmarks would regulate specific transcript outputs in a stage-specific manner, which would reveal the dynamic characteristics of m 6 A modification and differences in mRNA expression in the exercise intervention.Given this dynamic role of m 6 A modification in an exercise intervention, correcting the alteration in m 6 A modification and m 6 A-related genes might be a valuable strategy.
To examine critical genes regulated by m 6 A modification, we used GO and KEGG analyses to uncover possible regulatory roles of differentially methylated transcripts.According to the data of conjoint MeRIP-seq and RNAseq, several differentially methylated mRNAs were closely associated with the key signaling pathways in the development and progression of T2DM.As in previous studies, DMGs and DEGs were significantly enriched in several processes associated with energy metabolism, insulin regulation, glucose homeostasis, and inflammation response, which indicates the conservative and fundamental roles of m 6 A in regulating the development of T2DM. 40,41The most significant pathway was involved in the PI3K-Akt signaling, in which quite a few key genes were differentially methylated.In obesity and diabetes, activation of the PI3K-Akt signaling cascade by exercise stimuli modulates metabolism and further improves IR. 16 Given the important role of PI3K-Akt signaling, targeting this pathway might be a valuable strategy for treating IR and related metabolic disorders.Based on our analysis, we speculate that m 6 A might play critical roles in mediating a regulatory effect on gene expression in the PI3K-Akt pathway in an exercise intervention.
Mounting evidence reveals a close link between m 6 A modification and the PI3K-Akt pathway, especially for the important regulatory role of m 6 A in PI3K-Akt signaling. 22,24,42Here we found genes with altered m 6 A methylation enriched in all stages of the PI3K-Akt pathway, including upstream genes, core factors, and downstream effectors.Among them, core factors and downstream effectors are key factors in the response to exercise. 43owever, at present, little is known about how upstream genes respond to physical exercise, especially their epigenetic changes.In the study, we found that exercise can alter the expression level of more genes to counteract the HFD-induced impaired metabolism in mice as compared with the NFD + EX mice.Here we focused on two upstream molecules (Itga 3 and Fgf21), which were markedly regulated in the liver of mice with HFD-induced diabetes in an exercise intervention.Of them, Itga3 has been screened and identified as a new target gene for treating diabetic nephropathy. 44Fgf21 is a promising exerciseinduced hepatokine because of its pleotropic effects on maintaining energy homeostasis in rodents and humans. 45These findings imply that Itga3 and Fgf21 might be possible target genes for preventing and treating IR and related metabolic disorders.In addition, we provide evidence that m 6 A mediates Itga3 and Fgf21 expression.Based on the results, Rbm15 is present as a critical m 6 A factor during exercise training and may exert a regulatory role.Downregulation of the screened methyltransferase Rbm15 in the exercise intervention markedly decreased m 6 A and mRNA levels of Itga3 and Fgf21 as well as their mRNA stability.For Fgf21, the expression of Fgf21 and Rbm15 showed a significant upregulation under HFD status, and exercise training further promoted their expression levels.Whereas, for Itga3, the expression of Itga3 exhibited a significant downregulation in the HFD mice, and exercise training reversed it.These results suggested that Itga3 may also be affected to a greater or lesser extent by many other factors, except Rbm15.Collectively, we surmise that these upstream genes related to m 6 A modification influence the expression, resulting in alterations to overall translation and modulating PI3K-Akt signaling and its downstream effectors, which might be a critical layer of mRNA transcripts in the progression of T2DM.

| Animals
Four-week-old male C57BL/6J mice (SPF) were maintained in an animal room at 24 ± 0.5°C with a stable humidity (60%) and a controlled light/dark cycle (12 h).After adapting for 1 week, mice were randomly divided into four groups: mice fed a normal-fat diet (NFD; Cat.D12492, 10 kcal%; Research Diets, NJ, USA), NFD mice intervened with exercise training (NFD + EX), mice fed a high-fat diet (HFD; Cat.D12450, 60 kcal%), and HFD mice subjected to physical exercise (HFD + EX).All mice had freely access to food and drink for 12 weeks.The exercise training followed a previous report. 46In brief, for the first 4 weeks, the mice in the exercise group got a mild intensity run on treadmill with 5 m/min for first 5 min, 10 m/min for the subsequent 30 min, and 5 m/min for the final 5 min.For the last 8 weeks, the mice in the exercise group had a moderate intensity on treadmill with 5 m/min for the first 5 min, 13 m/min for the subsequent 30 min, and 5 m/min for the final 5 min.After the experiment, we collected blood samples and stored them at −80°C.In addition, we dissected liver tissues and divided into two parts: one portion was subjected to fixation using a 10% formalin solution to facilitate histological analysis, and another portion was cryopreserved at −80°C for RNA and protein extraction.All animal housing and experiments have been conducted in strict accordance with the guidelines of the Animal Ethics Committee of Fujian University of Traditional Chinese Medicine (2020092).

| Biochemical parameters
Blood glucose was detected at the time of blood sampling with an Accu-Chek® monitoring system (Roche Diagnostics, Switzerland).After 12 weeks of feeding and intervention, the mice were fasted overnight and gavage with glucose (1 g/kg body weight).Blood was collected from the tail vein at 30, 60, 120 and 180 min, and the concentration of plasma glucose was directly determined.Insulin was measured with a commercial ELISA kit (Mercodia, Sweden), and triglycerides were measured with a biochemical kit (Nanjing Jiancheng, China).

| Histology
A small portion of liver tissue was collected, washed with ice-cold phosphate-buffered saline (PBS), fixed in a 10% paraformaldehyde medium, and embedded.A section thick (~ 4 μm) was then stained by H&E for histopathology and assessed with an Olympus BX71 microscope (Hachioji, Japan).

| qRT-PCR assay
The prepared RNA was reversed with PrimeScript RT-PCR kit (Otsu, Japan).Then a relative mRNA was examined with FastStart Essential DNA Green Master (Roche, Germany) with beta-actin mRNA as a housekeeping gene.The result was evaluated via the 2 −ΔΔCT method.

| Global m 6 A measurement
Global RNA m 6 A content of the tested liver was assayed with an EpiQuik™ m 6 A RNA Methylation Quantification Kit (Colorimetric, Epigentek).Briefly, a total of 200 ng RNA was mixed with negative and diluted positive samples for 90 min, incubated with various antibodies, detected at the 450 nm wavelength, and then examined with a standard curve.

| Liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based mRNA modification
Liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based mRNA modification was performed by Aksomics Biotech Inc. (Shanghai, China).mRNA was isolated from total RNA by NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB, E7490), and quantified using Qubit RNA HS Assay kit (ThermoFisher Scientific, Q32855).mRNA was next hydrolyzed into single nucleosides, and the nucleosides were then dephosphorylated via the enzyme mix.The mixture was deproteinized by Satorius spin filter and analyzed by Agilent 6460 QQQ mass spectrometer with an Agilent 1260 HPLC system.The data was obtained from Agilent Qualitative Analysis software, and each modified nucleoside was acquired and normalized to quantity of mRNA.The level of m 6 A was examined as a percentage of total unmodified A.

| MeRIP-qPCR
The fragmented mRNA was prepared in a GenElute™ mRNA Miniprep Kit (Sigma) and incubated with anti-m 6 A antibody-bound protein G for 1 h.Then the incubated mRNA was washed three times, blocked with BSA, and rotated overnight at 4°C.Finally, the mRNA was purified with elution buffer.Subsequently, RT-PCR was performed to quantify the amount to m 6 A enrichment of target gene.
4.9 | MeRIP-seq and data analysis MeRIP-seq has been performed by Cloud-Seq Biotech Ltd. Co. (Shanghai, China).Briefly, total RNA was purified using a TRIzol kit (Invitrogen, #15596026).The concentration and purity of each RNA sample were evaluated by a NanoDrop ND-1000 (ThermoFisher Scientific, MA, USA).The integrity and contamination were measured using a denatured agarose gel electrophoresis.The mRNA was prepared using an Seq-Star™ poly(A) mRNA isolation kit (MD, USA), and m 6 A immunoprecipitation was further prepared by a GenSeq™ m 6 A RNA IP Kit (Shanghai, China).Subsequently, the library with/without immunoprecipitation was constructed by a NEBNext® Ultra II Directional RNA Library Prep Kit (MA, USA), and its quality was assessed via a BioAnalyzer 2100 system (Agilent, CA, USA), and then library sequenced on an Illumina Novaseq 6000 sequencer (Illumina, CA, USA) with 150-bp paired-end reads.The reads were collected using Q30, and the 3′ adaptor-trimming and low-quality reads were quality controlled on a Cutadapt software (v1.9.3).
After raw data process, the clean reads were aligned to reference genome on Hisat2 software (v2.0.4).The methylated sites on RNAs (peaks) were screened by MACS software (v1.4), and the differentially peaks were identified with a fold change cutoff of ≥2 and false discovery rat cutoff of ≤0.0001 by diffReps package.Fifty nucleotides on every side of top 1000 peaks in each sample were applied for motif enrichment.The analysis of gene ontology was examined using R-package topGO (v3.2), and the dot-blot presents the enrichment score of the top 10 most significant terms.

| RNA-seq and data analysis
Total RNA was prepared as above, and the rRNAs were removed using NEBNext rRNA Depletion kit (New England Biolabs, #E6310).The RNA library was constructed using NEBNext UltraTM II Directional RNA Library Prep kit (New England Biolabs, #E7760), the quality was evaluated using BioAnalyzer 2100 system (Agilent Technologies, #G2939BA).Library sequencing was carried on an Illumina Novaseq 6000 sequencer with 150-bp paired end reads, and the quality was controlled by Q30 and cutadapt software.The high-quality trimmed reads were applied for mRNAs analysis.
Then, the high-quality reads were aligned to the mouse reference genome (USCS mm10) on Hisat2 software (v.2.0.4).Based on the Ensembl gtf gene annotation file, the Cuffdiff software (v2.2.1) was performed to get the FRKM as the mRNA expression profile.The fold-change and P-value were examined based on the FPKM value, and the differentially expressed-mRNA was detected.GO and Pathway enrichment analysis were performed based on the differentially expressed-mRNA, and the R-package heatmap.2 was applied for heatmap drawing.

| Statistical analyses
GraphPad Prism-8 software was used for the statistical analysis and means ± SEMs are presented.Statistical significance was assessed via one-way analysis of variance followed by post-hoc tests and two-tailed unpaired Student's t-test.The p-value (<0.05) was considered as statistically significant.

| CONCLUSIONS
In summary, we have described an m 6 A transcriptomewide map of exercise-induced m 6 A function.Our findings suggest a significant link between m 6 A and upstream genes of PI3K-Akt signaling after physical exercise, which could allow for a fuller understanding of exercise-induced epigenetic mechanisms and guide future research on the clinical treatment of T2DM.

:
These findings highlight the pivotal role of the exercise-induced m 6 A epigenetic network and contribute insights into the intricate epigenetic mechanism underlying insulin signaling.K E Y W O R D S PI3K-Akt signaling, RNA m 6 A modification, transcriptome, type 2 diabetes mellitus | 3 of 17 CHEN et al.

F I G U R E 1
Characterization of m 6 A-methylated RNA in liver tissue in the exercise intervention.(A) Fasting serum glucose in NFD, HFD, and HFD + EX mice for 12 weeks.(B) Fasting serum insulin in three groups.(C) Global m 6 A-methylated RNA relative content of the tested liver in NFD, HFD, and HFD + EX groups.(D) Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis showing the percentage of m 6 A/A in mRNAs of NFD, HFD, and HFD + EX mouse livers.(E) The relative mRNA level of critical methylation (Mettl3, Mettl14, Rbm15, and Wtap) and demethylation (Fto and Alkbh5) by RT-qPCR assay and (F) The protein expression of m6A methylation and demethylation by western-blot assay in NFD, HFD, and HFD + EX groups.Data are means ± SEMs, *p < 0.05, † p < 0.01, ‡ p < 0.001.HFD, male C57BL/6 mice fed a high-fat diet; HFD + EX, male C57BL/6 mice fed a high-fat diet supplemented with exercise training; NFD, male C57BL/6 mice fed a control diet.

F I G U R E 2
Distribution of m 6 A-modified peaks in transcripts of NFD, HFD and HFD + EX mice.(A) Distribution of m 6 A peaks in different gene contexts.(B) Top three detected motifs among m 6 A peaks in the NFD, HFD, and HFD + EX groups.3′ UTR, 3′ untranslated region; 5′ UTR, 5′ untranslated region; CDS, coding sequence; HFD, high-fat diet; HFD + EX, high-fat diet supplemented with exercise training; NFD, control diet; startC, start codon; stopC, stop codon.F I G U R E 3 m 6 A modification in the exercise intervention.(A) Venn diagram of m 6 A modified DMGs.(B) Top 10 GO terms for DMGs with fold changes.(C) Enriched biological processes of DMGs between NFD and HFD.(D) Enriched biological processes of DMGs between HFD and HFD + EX.DMG, differentially methylated gene; GO, Gene Ontology; HFD, high-fat diet; HFD + EX, high-fat diet supplemented with exercise training; NFD, control diet.| 7of 17 CHEN et al.

F I G U R E 4
Gene expression in the exercise intervention.(A) Venn diagram of DEGs.(B) Heat map of DEGs among NFD, HFD, and HFD + EX. (C) Enriched biological processes of DEGs between NFD and HFD.(D) Enriched biological processes of DEGs between HFD and HFD + EX.DEG, differentially expressed gene; HFD, high-fat diet; HFD + EX, high-fat diet supplemented with exercise training; NFD, control diet.

F
I G U R E 5 m 6 A-mediated gene expression in the exercise intervention.(A) Number of up-or downregulated genes in the exercise intervention stratified by up-methylated genes.(B) Number of up-or downregulated genes stratified by down-methylated genes.(C) Enriched biological processes of overlapped methylated genes among NFD, HFD, and HFD + EX. (D) Enriched biological processes of overlapped expressed genes among NFD, HFD, and HFD + EX.The red arrow shows the same process between overlapped methylated and expressed genes.(E) Heat map of m 6 A-regulated genes involved in the PI3K-Akt signaling pathway among NFD, HFD, and HFD + EX. (F) Schematic representation of the PI3K-Akt signaling pathway for the probable mechanism of m 6 A modification in the exercise intervention.HFD, high-fat diet; HFD + EX, high-fat diet supplemented with exercise training; NFD, control diet.

F I G U R E 7
Rbm15 positively regulates the expression and mRNA stability of three upstream genes involved in the PI3K-Akt signaling pathway.(A) m 6 A level of Lama4, Itga3, Itgb3, Fgf21, and Pdgfrb in the exercise intervention.(B) m 6 A level of Itga3 and Fgf21 in HepG2 cells transfected with siRbm15.(C) Relative mRNA level of the targeted genes in control and Rbm15 knockdown cells.(D) mRNA stability of the targeted genes.Data are means ± SEMs.*p < 0.05, † p < 0.01, ‡ p < 0.001.HFD, high-fat diet; HFD + EX, high-fat diet supplemented with exercise training; NFD, control diet.| 13 of 17 CHEN et al.