Altered m6A RNA methylation governs denervation-induced muscle atrophy by regulating ubiquitin proteasome pathway

The animal experiments involved in this project have been approved by the Animal Protection and Utilization Committee of Nantong University and the Jiangsu Provincial Animal Protection Ethics Committee (No. S20200312-003). All animals were kept in a specific pathogen free (SPF) level barrier system at the Experimental Animal Center of Nantong University. Animals were placed in an environment with optimal temperature (24 ± 2 ℃) and automatic control of light cycle, allowing for free access to food and water.

Thirty healthy adult male Sprague Dawley (SD) rats weighing 210 g ± 5 g were provided by the Experimental Animal Center of Nantong University. Rats were randomly divided into 10 groups (n = 3) to establish a sciatic nerve dissociation model. Preoperative weighing was performed and rats were anesthetized with 1% pentobarbital sodium at a dose of 0.3 ml/100 g body weight. Cut off 1.5–2 cm long nerves from the sciatic nerves of rats at the same location, then suture the skin with sutures soaked in 75% ethanol, and finally treat the surgical wound with iodophor. Starting from the time of nerve disconnection, the rats were euthanized using spinal dislocation method at 0 h, 0.25 h, 0.5 h, 1 h, 3 h, 6 h, 12 h, 24 h, 36 h, and 72 h, respectively. The tibialis anterior (TA) muscle was taken and weighed. The muscle tissue was soaked in liquid nitrogen and stored at − 80 °C for RNA extraction. The healthy adult male ICR mice weighing 30 g ± 2 g were provided by the Experimental Animal Center of Nantong University. All ICR mice were randomly divided into 5 groups (0–3 days), 6 groups (0–14 days) and 4 groups (drug treatment). The sciatic nerve dissociation models were established. Anesthetize mice with 1% pentobarbital sodium at a dose of 0.2 ml/10 g body weight. After preparing the corresponding samples, measure and calculate the muscle mass/body mass ratio (mg/g), the muscle wet weight ratio (denervated/contralateral mass ratio), and the muscle fiber cross-sectional area. Other methods refer to rats.

MeRIP-Seq completed at OE Biotech (Shanghai, China). In short, the total RNA (400 μg, Qualified quality inspection) was purified twice using magnetic beads carrying oligo dT to capture mRNA with ploy (A) tail (polyadenylate). Fragmentation of Poly (A) RNA was performed in a buffer containing 10 mM ZnCl₂. Fragmented RNA was divided into two parts, of which 10% was used as input control, and the rest RNA was subjected to m6A RNA immunoprecipitation (IP). RNA was first incubated with m6A polyclonal antibodies, followed by binding to pre-equilibrated m6A-Dynabeads. After multiple elutions, m6A-positive RNA was obtained using phenol–chloroform extraction and ethanol precipitation methods. Evaluate the quality of MeIP library using the BioAnalyzer 2100 system. Sequencing was performed on Illumina Hiseq to obtain a 150 bp double ended reading.

For raw data (raw reads) generated in high-throughput sequencing, firstly Trimmomatic software [35] was used to remove the connectors, then low-quality reads were filtered out, and finally high-quality clean reads were obtained. SortMeRNA software is used to remove ribosomal RNA [36]. Use the default parameters of HISAT2 software to compare clean reads to the reference genome of rats, while retaining the unique comparison reads for subsequent analysis. Using input samples as a control, MeTDiff software [37] was used for peak detection and differential analysis, with parameters of ‘FRAGMENT_LENGTH = 200, PEAK_CUTOFF_PVALUE = 0.01, PEAK_CUTOFF_FDR = 0.05’. The ChIPseeker software was used to annotate the detected peaks. MEME and DREME software were used to detect motifs in peak sequences, while Tomtom software was used to compare the obtained motif sequences with known motif databases and annotate them accordingly using known motifs. The GO database (www.​geneontology.​org) was used for GO analysis of differentially m6A genes. The rMATS software was used for variable splicing analysis.

According to instructions, total RNA was extracted from muscle tissue using TRIzol reagent (Vazyme, Nanjing, China), and the first strand cDNA was synthesized using a reverse transcription kit (Absin, Shanghai, China). Real time fluorescence quantitative PCR (qPCR) detection was performed using ChamQ Universal SYBR qPCR Master Mix (Vazyme, Nanjing, China) and BIO-RAD system (BIO-RAD-96CFX). Using 18 s as the internal reference, the relative mRNA expression level was detected using the 2^−∆∆Ct method. The mRNA primers were designed and synthesized by Shanghai Shenggong Biology Co., Ltd. The primer information is listed in Additional file 1: Table S1.

The EpiQuik m6A RNA Methylation Quantification Kit (GEPIGENTEK, Farmingdale, New York, USA) was used to analyze the total amount of m6A in the TA muscle of mice. Follow the manufacturer’s instructions, in short, mix the total RNA with a special binding solution to ensure that m6A modified RNA was captured by specific antibodies. After signal enhancement, the m6A RNA-antibody complex was determined by colorimetric quantification at 450 nm using an enzyme-linked immunosorbent assay (Bio-Tek, Vermont, USA).

Mouse TA muscles were fixed with 4% paraformaldehyde (Beyotime, Shanghai, China) for 12 h, followed by sucrose gradient dehydration. Subsequently, the tissue was cut into 10 µm thick frozen sections using freezing microtome (CM3050S, Leica, Mannheim, Germany) and placed overnight in a 37 ℃ incubator. Slices were rinsed with PBS three times for 5 min each time. After adding an appropriate amount of blocking solution (Beyotime, Shanghai, China), the slices were sealed at room temperature for 1 h. They were incubated overnight with the first antibody anti-laminin antibody (1:1000, Abcam, Cambridge, UK) at 4 ℃, and then incubated with the second antibody Goat anti-mouse IgG H&L (1:500, Abcam, Cambridge, UK) at room temperature in dark for 2 h. Slices were sealed with sealing solution containing DAPI. An upright fluorescence microscope (Zeiss, Oberkochen, Germany) was used to capture fluorescence signals and obtain images. ImageJ software was used to measure the cross-sectional area of muscle fibers in various fields of view.

Tissue samples were lysed with protein lysis buffer (Beyotime, Shanghai, China) added with phosphatase inhibitor mixture (Beyotime, Shanghai, China) and protease inhibitor mixture (Beyotime, Shanghai, China). The BCA detection kit (Beyotime, Shanghai, China) was used to determine protein concentration. Extracted samples (20 μg total proteins per lane) was separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis, then transferred to PVDF membrane (Millipore, Massachusetts, USA), sealed with Tris buffer saline containing 5% skimmed milk powder, and incubated with the primary antibody overnight at 4 ℃. The primary antibodies used in this study include anti-Myosin Heavy Chain/MHC antibody (1:1000, Abcam, Cambridge, UK), anti-Fbx32 antibody (1:1000, Abcam, Cambridge, UK), anti-beta Tubulin (1:1000, Abcam, Cambridge, UK), anti-MuRF1 antibody (1:1000, Abcam, Cambridge, UK), anti-ALKBH5 antibody (1:1000, Proteintech, UK), and anti-FTO antibody (1:1000, Proteintech, UK). The PVDF membrane was eluted multiple times in buffer saline containing 0.1% Tween-20, and incubated with goat anti-rabbit IgG H&L as secondary antibodies at room temperature for 2 h. Immunoblotting was visualized by High-sig ECL Western Blotting Substrate (Tanon, Shanghai, China). Use ImageJ software to analyze contrast and normalize it with reference bands. ImageJ software was used to analyze the grayscale values of immunoblotting bands and normalize the target band with the reference band.

3-deazidenosine (Daa) is an inhibitor of S-adenosylhomocysteine hydrolase, which has been proved to reduce m6A methylation in vivo [38, 39]. R-2-hydroxyglutarate (R-2HG) is a tumor metabolite, which can increase intracellular m6A levels by inhibiting the activity of FTO [40, 41]. 5% DMSO was used to dissolve the drug, with a final concentration of 0.04 mg/μl for Daa and 0.0012 mg/μl for R-2HG. TA muscle intramuscular multipoint injection were performed immediately after the sciatic nerve transection in drug treatment group mice. Each mouse was injected with 15 μl per day for a total of 7 days.

All statistical analyses were performed using Prism 9 software (GraphPad, LaJolla, CA). All data are expressed as means ± standard deviation (SD) and analyzed using a one-way analysis of variance. Intergroup differences were detected using Tukey’s multiple comparisons test. A value of p < 0.05 was considered statistically significant.

The m6A characteristics of skeletal muscles have been analyzed in human, mouse and pig, but the data from rats is lacking [42‐47]. Therefore, we first focused on the characteristics of m6A modification in pre-atrophic rat skeletal muscles. Using m6A MeRIP, we constructed the m6A profile (IP) and conventional transcript profile (input) of rat skeletal muscles. The experiment setted up dense time gradients, including 0 h, 0.25 h, 0.5 h, 1 h, 3 h, 6 h, 12 h, 24 h, 36 h, 72 h after denervation, which would help to analyze the dynamic changes of m6A in detail (Fig. 1A).

Through strict screening criteria (fold enrichment ≥ 2, FDR ≤ 0.05), 12,479, 12,534, 12,173, 12,934, 11,773, 12,630, 10,136, 8953, 11,889, and 12,344 m6A peaks were defined at 0 h, 0.25 h, 0.5 h, 1 h, 3 h, 6 h, 12 h, 24 h, 36 h, and 72 h, respectively. They were associated with 7544, 7583, 7460, 7768, 7231, 7493, 6388, 5801, 7179, and 7458 transcripts (Additional file 1: Fig. S1A). For samples at different times, there was no significant difference in the median length (median and interquartile range) of m6A peak (Additional file 1: Fig. S1B). In skeletal muscle, the frequencies of GA-enriched motifs were higher among the fragments modified with m6A (Additional file 1: Fig. S1C). The distribution of detected m6A fragments and input fragments in the genes were different. The frequency of input was higher at the coding sequence (CDS), whereas the frequency of m6A fragment was higher at the junction of 5′ UTR and CDS (Additional file 1: Fig. S1D). The Principal Component Analysis (PCA) results showed that the data of m6A and input were gradually changed and could be clearly separated during skeletal muscle atrophy from early to mid to late stages (Additional file 1: Fig. S1E). Although the data at 12 h had a large degree of dispersion, according to general standards, expression differences could be distinguished based on expression levels, indicating that the data at 12 h did not affect differential gene recognition (Additional file 1: Fig. S1F).

The 7534 genes contained m6A modifications, accounting for approximately one-third of the total detected genes (Fig. 1B). The m6A peak was widely distributed in mRNA, with the higher distribution in CDS and 3′ UTR, and with the higher frequency at the junction of the two (Fig. 1C, D). Most genes were uniquely methylated by m6A, but 237 genes contained m6A methylation of 5′ UTR, 3′ UTR, and CDS simultaneously (Fig. 1E). Chr1, chr3, and chr10 emerged as the top three chromosomes with the highest frequency of modification, while chrY emerged as the least methylation modification (Fig. 1F).

It is known that m6A methylation can regulate the abundance of transcripts and also affect variable splicing [48, 49], both of which are closely related to denervation-induced muscle atrophy [11]. Here, we analyzed the correlation between m6A and gene expression, as well as variable splicing. Using rMATS, the numbers of variable splicing events were defined for the 0 h sample, including alternative 3′ splice site (A3SS, 3136), alternative 5′ splice site (A5SS, 1948), retained intron (RI, 29,354), skipped exon (SE, 26,171), and mutually exclusive exon (MXE, 283) (Additional file 1: Fig. S2A). The correlation between m6A and the two common variable splicing forms, SE and RI, was very low (r = 0.014, p = 0.25 and r = 0.0032, p = 0.6) (Additional file 1: Fig. S2B). According to distribution, m6A was divided into three types: 3′ UTR, CDS, and 5′ UTR, and correlation analysis of gene expression was conducted separately. There was a slight negative correlation between transcripts carrying m6A in 3′ UTR and expression levels (r = − 0.093, p = 3.7e⁻⁷), and a similar negative correlation (r = − 0.084, p = 6.7e⁻¹⁰) was also observed in transcripts carrying m6A in 5′ UTR and CDS (Fig. 1G, H). These data suggested that m6A methylation might negatively regulate gene expression but did not affect variable splicing. Gene Ontology (GO) analysis revealed the function of genes carrying m6A modification in healthy skeletal muscles. The genes modified by m6A in 3′ UTR were mainly related to biological processes such as positive regulation of cell migration, in utero embryonic development, protein autophosphorylation, and angiogenesis (Fig. 1I). The genes modified by m6A in 5′ UTR and CDS were mainly associated with negative regulation of transcription by RNA polymerase II, protein ubiquitination, DNA repair, regulation of transcription, DNA-templated, biological process, protein deubiquitination, and mRNA splicing via spliceosome, etc. (Fig. 1J).

In order to further investigate how m6A changes during denervation-induced muscle atrophy, we defined the differential m6A peaks based on strict criteria (fold change ≥ 2, FDR ≤ 0.05). The differential peaks were less in the initial stage of denervation, but significantly increased after 6 h. From 0.25 h to 72 h, 28, 61, 105, 68, 438, 653, 683, 771, and 855 differential m6A peaks were identified (Fig. 2A). At 6 h after denervation, the number and amplitude of down-methylated m6A peaks were higher than those of up-methylated m6A peaks in skeletal muscle, which was consistent with the results of decrease in m6A content at 72 h (Figs. 1F and 2B). About 50% of genes contained only one m6A peak, while the remaining 50% contained more than one m6A peak (Fig. 2C). For differential peaks, most of them only corresponded to one transcript, especially at 6 h (Fig. 2D). At 0.25 h and 0.5 h of atrophy, differential peaks were mainly distributed in the CDS region of mRNA. Over time, differential peaks gradually decreased in the CDS region, while differential peaks gradually increased in the 3′ UTR (Fig. 2E).

During skeletal muscle atrophy, there was only one host gene, Myh9, which underwent differential m6A modification at all time points. The 458 differential m6A peaks were 72 h specific and belonged to 402 host genes. The 137 differential m6A peaks were common at 36 h and 72 h, and they could be mapped to 95 genes (Fig. 2F, Additional file 2: File S1).

We examined the relationship between differential m6A modifications and gene expression during denervation-induced muscle atrophy. The results showed that the expression levels of differential m6A methylated genes were generally higher than those of constitutive m6A methylated genes (Fig. 3A). To further analyze the relationship between m6A and gene expression, we counted differentially expressed genes and differential m6A genes at various time points and extracted their intersection genes (Fig. 3B). At early time points (0.25 h, 0.5 h, 1 h), the proportion of intersection genes in the differential m6A genes approached 0, indicating that m6A changes in the early stage of denervation did not affect gene expression (Fig. 3B, C). At later time points (24 h, 36 h, 72 h), this proportion of intersection increased to about 40%, indicating that about 40% of m6A modified genes were differentially expressed in the late stage of denervation (Fig. 3B, C). The correlation analysis results showed that there was a negative correlation between differential m6A gene expression and m6A modification at different time points (Fig. 3D). There were two genes, Erfe and Hspa1l, which exhibited both differential m6A modification and differential expression at various time points in the late stage of denervation (Fig. 3E). Erfe (also known as Ctrp15) encodes a skeletal muscle-derived myonectin, which can improve acute myocardial injury by regulating energy metabolism [50, 51]. Hspa1l encodes a heat shock protein involved in autophagy regulation [52, 53]. Although both them are related to skeletal muscle physiology, they have not been reported in denervation-induced muscle atrophy.

By clustering host genes labeled with differential m6A into biological pathways, we observed the function of the differential m6A genes after 72 h of denervation. The 1106 differential m6A genes were mainly related to biological processes such as zinc ion binding, ubiquitin protein ligase activity, transcription coactivator activity, ATP binding, RNA polymerase II cis-regulatory region sequence-specific DNA binding, protein phosphatase binding, nucleic acid binding, glucocorticoid receptor binding, DNA binding, and mannosyl-oligosaccharide 1,2-alpha-mannosidase activity (Fig. 3F). The 313 differential m6A intersection genes were mainly related to biological processes such as zinc ion binding, tau protein binding, histone acetyltransferase binding, sequence-specific DNA binding, GPI anchor binding, glucocorticoid receptor binding, transcription coactivator activity, ATP binding, and ubiquitin protein ligase activity (Fig. 3G). It was worth noting that common pathways included zinc ion binding, ubiquitin protein ligase activity, ATP binding, sequence-specific DNA binding, and transcription activator activity, suggesting that m6A might affect the related gene expressions of ubiquitin–proteasome pathway and transcriptional activation during denervation-induced muscle atrophy.

To verify the relationship between m6A methylation levels and denervation-induced muscle atrophy, we established the sciatic nerve transection mouse model and dissected the tibialis anterior muscles. Morphological analysis of skeletal muscle showed significant decreases in the muscle mass/body mass ratio and muscle fiber cross sectional area after 72 h of denervation (Den 3d), compared to 0 h of denervation (Ctrl) (Fig. 4A–D). Then, we used RT-qPCR to analyze the expression changes of m6A regulators at 0 h, 6 h, 24 h, 36 h and 72 h after denervation (Fig. 4E). One of the erasers of m6A, Fto, showed significant increased expression, and the expression of another eraser, Alkbh5, gradually increased and was upregulated about sixfold at 72 h. The expressions of m6A writers, Mettl3 and Mettl14, did not show significant changes at 6 h, but were upregulated at 24 h, 36 h and 72 h, respectively. The expression of m6A writers, Virma, was gradually increased, and upregulated about ninefold at 72 h. The expression changes of m6A writers, Wtap, exhibited fluctuations during atrophy. These results indicated that the changes of m6A modification were complex during denervation-induced muscle atrophy. The m6A RNA Methylation Quantification Kit was used to detect the content of m6A in skeletal muscle. Compared with the control group, the proportion of m6A decreased by about 70% at 72 h after denervation, indicating that the demethylation of mRNA was dominant in the late stage of denervation-induced muscle atrophy (Fig. 4F). Based on the qPCR results, we speculated that reduction of m6A might be mediated by Alkbh5 in the late stage of atrophy.

Then, we further examined the expression changes of demethylase during denervation-induced muscle atrophy at six time points: 0 day, 1 day, 3 day, 5 day, 7 day, and 14 day after denervation. The results showed that there were significant decreases in muscle mass/body mass ratio and wet weight ratio after 3 days of denervation (Additional file 1: Fig. S3A, B). The expression levels of Alkbh5 and Fto showed upward trends with prolonged denervation, with both significantly higher levels at 7 days of denervation (Additional file 1: Fig. S3C). These indicated that the expression of Alkbh5 and Fto were negatively correlated with the degree of skeletal muscle atrophy.

To further confirm whether m6A methylation played a role in denervation-induced muscle atrophy, we used two drugs to alter the content of m6A in skeletal muscle. We injected 0.04 mg/μl Daa into the target muscle of normal mice (innervation group) once a day for 7 consecutive days. At the same time, we constructed denervated animal model (denervation group), and the target muscle were injected with 0.0012 mg/μl R-2HG once a day for 7 consecutive days (Fig. 5A). Two experiments were conducted by injecting equal volumes of DMSO as control, and all mice were euthanized on the eighth day for subsequent experiments. We first tested the effectiveness of two drugs. The results showed that compared to the control group, Daa reduced the content of m6A by about 40% in normal skeletal muscles, while R-2HG increased the content of m6A by about onefold in denervated skeletal muscles (Fig. 5B). Morphological and statistical results showed that Daa significantly reduced the muscle fiber cross sectional area, the muscle mass/body mass ratio, and wet weight ratio of normal mouse muscles (Fig. 5C–E). The effect of R-2HG was opposite to that of Daa, as it significantly increased the muscle fiber cross sectional area, the muscle mass/body mass ratio, and wet weight ratio of denervated mouse muscles (Fig. 5C–E). In addition, we detected the expression of myosin heavy chain (MHC). The results showed that denervation resulted in a significant downregulation of MHC in skeletal muscles. Daa significantly inhibited the expression of MHC in normal skeletal muscles, while R-2HG significantly increased the expression of MHC in denervated skeletal muscles, consistent with morphological results (Fig. 5H, I). At the same time, we detected the expression changes of Alkbh5 and Fto. The results showed that the expression levels of both them were positively correlated with the degree of muscle atrophy, and this result was consistent with the relative m6A levels (Fig. 5J). Our results suggested that whether skeletal muscle atrophy was controlled by the level of m6A. On the other hand, R-2HG might be a potential drug for treating denervation-induced muscle atrophy.

We next explored the mechanism of m6A regulation in denervation-induced muscle atrophy. The above results suggested that the function of differential m6A methylated genes was mainly related to the ubiquitin–proteasome system (Fig. 3F, G). We analyzed the correlation between the expression of m6A methylated genes of and m6A modification during denervation-induced muscle atrophy. The results showed a negative correlation between m6A modification and expression of most ubiquitin–proteasome related genes (Fig. 6A). Based on the discovery of a decrease in total m6A content during denervation-induced muscle atrophy, we speculated that the decreased m6A methylation might activate the ubiquitin–proteasome system, ultimately leading to the atrophy phenotype.

To validate this, we detected the effect of m6A changes on expressions of two key E3 ubiquitin ligases. The results showed that Daa significantly increased the expressions of Trim63 and Fbx32 in normal skeletal muscles, while R-2HG significantly downregulated the expression of them in denervated skeletal muscles (Fig. 6B). The expressions of Trim63 and Fbx32 were significantly negatively correlated with the content of m6A in skeletal muscles, and also matched the phenotype of muscle atrophy. Our results indicated that m6A modification regulated denervation-induced muscle atrophy through ubiquitin–proteasome system.