Identification and Functional Characterization of Metabolites for Skeletal Muscle Mass in Early Postmenopausal Chinese Women

Abstract

Low skeletal muscle mass (SMM) is a crucial component of the sarcopenia phenotypes. In the present study, we aim to identify the specific metabolites associated with SMM variation and their functional mechanisms of decreased SMM in early postmenopausal women. We performed an untargeted metabolomics analysis in 430 early postmenopausal women to identify specific metabolite associated with skeletal muscle mass indexes (SMIes). Then, the potential causal effect of specific metabolite on SMM variation was accessed by one-sample Mendelian randomization (MR) analysis. Finally, in vitro experiments and transcriptomics bioinformatics analysis were conducted to explore the impact and potential functional mechanisms of specific metabolite on SMM variation. We detected 65 metabolites significantly associated with at least one SMI (variable importance in projection > 1.5 by partial least squares regression and p < .05 in multiple linear regression analysis). Remarkably, stearic acid (SA) was negatively associated with all SMIes, and subsequent MR analyses showed that increased serum SA level had a causal effect on decreased SMM (p < .05). Further in vitro experiments showed that SA could repress myoblast’s differentiation at mRNA, protein, and phenotype levels. By combining transcriptome bioinformatics analysis, our study supports that SA may inhibit myoblast differentiation and myotube development by regulating the migration, adhesion, and fusion of myoblasts. This metabolomics study revealed specific metabolic profiles associated with decreased SMM in postmenopausal women, first highlighted the importance of SA in regulating SMM variation, and illustrated its potential mechanism on decreased SMM.

Sarcopenia is characterized by age-related loss of skeletal muscle mass (SMM) and function, leading to a substantially increased risk of falls, fractures, disability, and mortality (1). It was recognized as an independent condition with an International Classification of Diseases-10 code in 2016 (2). Low SMM is a key component of the sarcopenia phenotypes. SMM incipiently decreases as early as age 40 and can decline about 30%–50% between 40 and 80 years of age (3). One of the most striking phenomena marking women’s aging process is menopause, typically accompanied by relatively rapid metabolism and endocrine changes, which was associated with decreased SMM, and bone mineral density in women (4). Therefore, understanding the pathological mechanisms of decreased SMM during early menopause would improve its early prevention, diagnosis, and treatment of sarcopenia in elderly women.

Skeletal muscle approximately comprises 35% of body mass in adult women (5) and is one of the most critical sites for metabolic control. Metabolites are the end products of various metabolic pathways, and their levels reflect individuals’ metabolomic status (6). Metabolomic analysis can detect changes or differences in metabolite levels and is thus valuable for assessing SMM status in humans (7–9). A recent sarcopenia-related untargeted metabolomics study highlighted the importance of amino acids (aspartic acid and glutamic acid) and lipids (12S-HETRE, arachidonic acid, 12S-HETE, and glycerophosphocholine) in the regulation of body mass index adjusted appendicular lean mass (ALM/BMI) among 136 Caucasian women (7). Another untargeted metabolomics study in 319 older black men showed that 3 lipids and lipid-like molecules (C14:1 monoacylglycerol and C18:0 monoacylglycerol and C36:1 phosphatidylcholine plasmalogen), 1 benzenoid (4-hydroxymandelate), and 3 organic acids (lysine, methionine, and tryptophan) were independently associated with ALM (8). In addition, a study conducted among 305 elderly Taiwanese subjects indicated that glutamate levels in women and alpha-aminoadipate, Dopa, and citrulline/ornithine levels in men were related to ALM adjusted for height squared (ALM/height²) (9). However, metabolomics studies of sarcopenia traits are still rather limited, especially in postmenopausal women. Indeed, results from these previous studies were largely inconsistent in their significant findings, partially because different studies used different skeletal muscle indexes (SMIes, eg, ALM, ALM/BMI, ALM/height²) (7–9). Moreover, these previous studies did not provide an in-depth investigation into the potential mechanisms of the significant metabolites on SMM variation (7–9).

To address the questions and the gap in knowledge stated above, we performed an untargeted serum metabolomics analysis in 430 healthy early postmenopausal Chinese women to identify SMIes-associated metabolites. Furthermore, we performed molecular biology experiments and transcriptomics bioinformatics analysis to explore the functional mechanism of specific metabolites on SMM variation.

Method

Study Cohort

Four hundred thirty unrelated early postmenopausal Chinese women were randomly recruited from Guangzhou City, China from June 2014 to January 2018. The inclusion criteria included (i) being aged 40 years or older, (ii) having lived in Guangzhou City for at least 3 months, (iii) being in the early postmenopausal stage (more than one year but less than 6 years since last menstruation) (10). We excluded all subjects with any use of antibiotics or estrogens medications in the past 3 months, as well as chronic liver disease, diabetes, and other diseases that could affect the risk of sarcopenia. All the participants had signed an informed consent form before they were recruited. This study was approved by the Third Affiliated Hospital of Southern Medical University (Guangzhou City, China) institutional review board.

Phenotype Measurements

Trained research staff administered face-to-face interviews with the study subjects to collect demographic characteristics and menopause time using a standardized questionnaire. All the questionnaire data were recorded. Body weight (kg) and height (cm) were measured twice in light indoor clothes without shoes to nearest 0.1 kg and 0.1 cm, respectively. Body mass index (BMI) was calculated as weight in kilograms divided by height in squared meters. Physical activity was categorized into low (without exercise), moderate (less than 2.5 hours per week), and high (more than 2.5 hours per week) groups, according to the average exercise time per week (11). For each subject, the lean mass in the arms and legs was measured by trained research staff using dual-energy X-ray absorptiometry (DXA) machine (GE Healthcare, Madison, WI, version 13.31.016). During the data collection, DXA was calibrated daily, and software and hardware were kept up-to-date. ALM (kg) was calculated as the sum of lean mass in the arms and legs. Individuals’ SMM was assessed by various SMIes, including ALM, ALM/height², ALM adjusted for weight (ALM/weight), and ALM/BMI.

Untargeted Metabolomic Assessment and Whole-Genome Sequencing

Blood samples were collected after an overnight fast for >8 h from all participants, transported to the laboratory with an ice pack, and immediately frozen at −80°C within 30 minutes. Whole blood samples were used for serum centrifugation and genomic DNA extraction. Genomic DNA was extracted by the SolPure DNA Kit (Magen, Guangzhou, China). Serum samples were used for extracting metabolites, and DNA samples were used for human whole-genome sequencing. Detailed information on untargeted metabolomic assessment and whole-genome sequencing were shown in Supplementary Material.

Statistical Analysis

Association analyses between SMIes and metabolites

Normalization and autoscaling were conducted for metabolomics abundance before metabolomics analysis. To identify metabolites that were significantly associated with variation in SMIes, we applied partial least squares (PLS) regression and multiple linear regression analysis to model the metabolomic profiles. PLS provides variable importance in projection (VIP) value for the descriptor data set X (metabolites) that best describe the response data set Y (SMIes), as it maximizes the covariance expressing the common structures between X and Y (12). Potential confounding factors, including age, menopause time, and physical activity, were adjusted in the PLS regression and multiple linear regression. Finally, metabolites with VIP > 1.5 in the PLS regression and p < .05 in multiple linear regression were considered statistically significant. The data analyses were performed using R packages, including “mixOmics,” “robustbase,” and “tidyverse.”

Pathway enrichment analysis of SMIes-associated metabolites

To identify potential pathways represented by the significant metabolites (associated with at least one SMI), we conducted a pathway enrichment analysis using the web tool MBROLE 2.0 (http://csbg.cnb.csic.es/mbrole2/analysis.php) (13). The R “ggplot2” package was used to visualize the enrichment results.

The causal effect of stearic acid on SMIes

Genome-wide association analysis

Genome-wide association (GWA) analysis was conducted to test associations between phenotypes (SMIes and stearic acid [SA]) and genotyped SNPs. Potential confounding factors, including age, menopause time, and physical activity, were adjusted in the GWA analysis. Quality control of genotype data and GWA analysis were implemented with Plink 1.9 software (14).

One-sample Mendelian randomization

Based on metabolomics analysis and GWA analysis results, we performed a one-sample Mendelian randomization (MR) to illustrate whether SA has a causal effect on SMM traits. Potential independent SNPs (linkage disequilibrium, r² < 0.001) with p-values < 5 × 10⁻⁵ in GWA analysis were considered as instrumental variables in one-sample MR. Inverse variance weighted, simple median, and weighted median were used for MR analysis (15). MR–Egger regression was performed to assess the horizontal pleiotropic pathway between genetic variants and outcomes (16). Intercept with p > .05 indicates no horizontal pleiotropic exists in M–Egger regression. The result with p < .05 in all 3 MR analysis methods was considered statistically significant. The MR analyses were performed using R packages, including “MRInstruments” and “MendelianRandomization.”

Functional validation

Cell culture and differentiation

Mouse C2C12 myoblasts (purchased from Stem Cell Bank, Chinese Academy of Sciences) were evenly seeded in a 6-well plate at 1 × 10⁶ cells/well and placed in a 37°C constant temperature incubator containing 5% CO₂, and cultured in Dulbecco’s modified Eagle medium (DMEM, Basal Media, Shanghai, China) containing 10% fetal bovine serum (ZETA LIFE, San Francisco, CA) and 1% antibiotics (100 units/mL of penicillin) to 70%–80% confluence. 0.25% trypsin, penicillin, and streptomycin were purchased from New Cell and Molecular Biotech Co., Ltd. To induce myoblast differentiation, the control group was cultured with differentiation medium (DM) [DMEM + 2% horse serum (Hyclone, Logan, UT) + 1% antibiotics + 0.04% sterile absolute ethanol], and the treatment group was cultured with DM containing different concentrations (0, 5, 10, and/or 20 μM) of SA (Solarbio Life Science, Beijing, China). Three biological replicates were included in each group. After 5 consecutive days of induction (changing the medium every 24 h), myoblasts entered the end of differentiation, and mature myotubes spread throughout the cell culture dish. Unless otherwise specified, cell culture and differentiation are carried out according to this scheme.

Molecular biology experiments

CCK-8 assay (New Cell and Molecular Biotech, Suzhou, China) was used to evaluate cell toxicity of different concentrations of SA (0 μM + 4% absolute ethanol, 0, 5, 10, 20 μM). Molecular biology experiments, including immunofluorescence staining, Western blot, and quantitative real-time polymerase chain reaction (qRT-PCR), were performed to analyze the effects of different concentrations of SA on myoblasts differentiation (3 and 5 days) at morphological, protein, and mRNA levels, respectively. Myosin heavy chain (MyHC), myogenic differentiation (MyoD), and/or myogenin (MyoG) were applied as molecular markers of myoblasts differentiation for detection in molecular biology experiments. Three biological replicates were included in each group. The specific experimental steps are given in Supplementary Material.

Transcriptomics sequencing and analysis

Cultured cells from the treatment and control groups were collected at 0, 3, and 5 days after the treatment, respectively, and subject to RNA sequencing. Three biological replicates were set for each group (a total of 15 samples) and named as 0 d, 3d_0 μM, 3d_20 μM, 5d_0 μM, and 5d_20 μM, respectively. Detailed information on transcriptome RNA sequencing was described in Supplementary Material. Differentially expressed genes were screened using the “Limma” software package for R, and the significance threshold was set to ± 2 times fold changes. Unsupervised clustering was performed using the Fuzzy c-means algorithm as implemented in the “Mfuzz” package for R. STRING protein interaction database (http://string-db.org/) was used to analyze differential gene–protein interaction networks (medium confidence: 0.4). CytoHubba (Cytoscape plugin) was used to explore the hub genes network (top 10, degree). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed by the R package “clusterProfiler.” The false discovery rate (FDR) was corrected using the Benjamini−Hochberg method, and FDR ≤ 0.05 was significant enrichment. The R “ggplot2” package was used to visualize the enrichment results.

Results

Characteristics of the Study Subjects

In this study, 430 early postmenopausal Chinese women were recruited, ranging from 41.47 to 63.80 years old. The mean (SD) of menopause time was 2.18 (0.81) years, ranging from 1 to 4.90 years. Approximately half of the participants (49.07%) had a high level of physical activity in the present study. According to the Asian Working Group on Sarcopenia criteria (17), 20.7% of the subjects had decreased SMM (ALM/height² < 5.4 kg/m²), and 79.3% of the subjects had normal SMM (ALM/height² ≥ 5.4 kg/m²). The clinical characteristics of the samples were provided in Table 1.

Table 1.

Characteristics of the Study Population

Variables	Mean (SD) or Percentage	Minimum	Maximum
Age, y	52.94 (2.86)	41.47	63.80
Menopause time, y	2.18 (0.81)	1	4.90
Weight, kg	57.17 (7.55)	40.00	82.00
Height, m	1.58 (0.05)	1.42	1.72
BMI, kg/m2	22.93 (2.80)	16.42	33.73
ALM, kg	14.66 (1.79)	10.50	19.96
ALM/BMI, kg/BMI	0.64 (0.07)	0.43	1.00
ALM/hight2, kg/m2	5.88 (0.62)	4.32	7.78
ALM/weight	0.26 (0.02)	0.19	0.36
Physical activity, %
None	29.07
<2.5 h/wk	21.63
≧2.5 h/wk	49.07

Note: ALM, appendicular lean mass; BMI, body mass index.

Metabolomic Analysis of SMM

Using liquid chromatography–mass spectrometer (LC–MS), we detected 295 untargeted metabolic variables with known identities, including 142 in positive ion mode and 153 in negative ion mode. PLS and multiple linear regression analyses identified that 65 serum metabolites were significantly associated with at least one SMI (VIP > 1.5 and p < .05, Supplementary Tables S1–S4). Among these 65 significant metabolites, 23, 37, 26, and 21 metabolites were associated with ALM, ALM/height², ALM/weight, and ALM/BMI, respectively (Supplementary Tables S1–S4). In the pathway enrichment analysis, 20 of 65 SMM-associated metabolites were enriched in 8 significant metabolic pathways (p < .05, Figure 1). Of them, 3 pathways (biosynthesis of unsaturated fatty acids, linoleic acid metabolism, and primary bile acid biosynthesis) are fatty acid metabolism-related pathways, which are closely related to SMM metabolism. We identified that 2 saturated fatty acids (palmitic acid and SA) were significantly associated with SMM variation. Palmitic acid was identified to be negatively associated with both ALM/BMI and ALM/weight (Supplementary Tables S1–S4). Remarkably, SA was significantly associated with all SMIes in both PLS and multiple linear regression (Supplementary Tables S1–S4).

Figure 1.

Pathway enrichment analysis for the metabolites associated with skeletal muscle mass indexes.

The Causal Effect of SA on SMM Based on MR Analysis

MR analysis results indicated that SA had a negatively causal effect on all SMIes (p < .05, Supplementary Table S5) without detected pleiotropy bias (p-values of MR-Egger intercept > .05). Therefore, further in vitro experiment was performed to investigate the adverse causal effects and potential mechanisms of SA on decreasing SMM.

SA Significantly Repress Myoblast Differentiation

The measurement of the CCK-8 kit in the cell culture medium after incubating SAs (0, 5, 10, and 20 μM) for 24 hours showed that each SAs concentration and 0.04% absolute ethanol had no toxic effect on C2C12 cells (Supplementary Figure 1A). After continuous induction of C2C12 cell differentiation for 5 days, the length and width of mature myotubes on Day 3 and Day 5 of differentiation were measured using Image J software. Compared with the control group, different concentrations of SAs induced differentiation for 3 (Figure 2A) and 5 (Figure 2B) consecutive days, resulting in various degrees of decrease in the length and width of myotubes (Figures 2A and B), suggesting that SAs can repress skeletal muscle myogenic differentiation. In addition, the protein (Figure 2C) and mRNA (Figure 2D) expression levels of MyoD, MyoG, and MyHC were significantly decreased in the treatment group compared with those in the control group. Overall, these results demonstrated that SA could repress the differentiation of myoblasts and mature myotube development.

Figure 2.

Comparison of stearic acid (SA) in regulating the differentiation of C2C12 myoblasts. (A) Immunofluorescence staining of MyHC+ myotube after 3 days of C2C12 myoblast differentiation. MyHC+ myotube length and width statistics under 5 different visual fields after 3 days of C2C12 myoblast differentiation. (B) Immunofluorescence staining of MyHC+ myotube after 5 days of C2C12 myoblast differentiation. MyHC+ myotube length and width statistics under 5 different visual fields after 5 days of C2C12 myoblast differentiation. (C) Protein expression levels of MyHC, MyoD, and MyoG were compared with GAPDH as endogenous control at 3 and 5 days after myogenic differentiation of C2C12. (D) The relative expression of MyHC, MyoD, and MyoG mRNA at 3 and 5 days after myogenic differentiation of C2C12. *p < .05; **p < .01; ***p < .001; and ****p < .0001.

SA Significantly Repress Myoblast Differentiation via Myoblast Migration and Adhesion

Myogenic differentiation is a separable process with a clear timeline, divided into early, middle, and late differentiation, with distinct biological characteristics at each period (Figure 3A). To clarify the changes induced by SA on the key biological processes during myoblast differentiation, we collected undifferentiated (0 days), mid-differentiated (3 days), and end-differentiated (5 days) myoblasts for transcriptomics analysis (Figure 3A). After RNA seq analysis, strict filtering, and quality control, we quantified the expression value of 17 142 genes. Among them, the expression levels of MyHC, MyoD, and MyoG were consistent with the trend of our qRT-PCR detection results (Supplementary Figure 1B), which verified the reliability of transcriptome data. In addition, to account for observed differences due to biological variations, we compared the repeated data at 0, 3, and 5 days. The results showed that the reproducibility of each sample was good (Supplementary Figure 1C). In total, 1 329 differentially expressed genes were identified (“Limma” package for R, difference threshold ± 2 times), including 726 genes differentially expressed at 3 days, and 985 genes differentially expressed at 5 days, and 382 genes differentially expressed at both time points (Figure 3B).

Figure 3.

Time course of myogenic differentiation, the expression profile of differentially expressed genes, and differential gene co-expression time clustering in SA-induced differentiation. (A) The biological process is temporally separable during myogenic differentiation and transcriptomics experiment. (B) Volcano maps and Venn diagram of differentially expressed genes in the mid-differentiation stage (3d) and the end-differentiation stage (5d). (C) SA induced temporal clustering of 6 different expression patterns of all differential genes during myotube formation. The fuzzy c-means algorithm was used to perform unsupervised time clustering on all differentially expressed genes in the middle and end of differentiation. Each line represents the gene expression trend after homogenization. Membership value represents how well the gene profile fits the average cluster profile.

Unsupervised time clustering was performed on the combined set of differentially expressed genes (1 329 genes) using the fuzzy c-means algorithm, and all genes were divided into 6 temporal clusters (Figure 3C). Among them, cluster 1 (154 genes) and cluster 5 (162 genes) showed upregulation over time; cluster 2 (224 genes) and cluster 3 (347 genes) illustrated a downward adjustment with time; cluster 4 (148 genes) showed upregulation in the mid-differentiation phase and downregulation in the end-differentiation phase; cluster 6 (294 genes) showed downregulation in the mid-differentiation phase and upregulation in the end-differentiation phase (Figure 3C). With the same expression trend along the differentiation process, these gene clusters often have the same function and regulate the same biological process during the differentiation process.

Then, genes in each cluster were entered in the STRING database (http://string-db.org/) to predict the interaction network, and cytoHubba in Cytoscape was performed to identify the top 10 hub genes (according to connected node number—degree) in each interaction network (Supplementary Figure 2). Next, GO enrichment analysis was performed on the top 10 genes in each cluster (Supplementary Figure 2), and false discovery rate (FDR) < 0.05 indicated significant enrichment. The results showed that cluster 1 and cluster 2 were related to the cell migration, such as “response to wounding,” “wound healing”; cluster 3 was related to cell division and cell cycle; cluster 4 was related to extracellular structure and matrix organization; cluster 5 was related to mitochondrial respiratory chain; cluster 6 was associated with ATP metabolic process (Supplementary Figure 2). These GO terms were related to the biological process of myoblast differentiation. It is worth noting that clusters 1and2 closely related to cell migration, and the wound-healing scratch experiments (3 biological replicates) also indicated that SA inhibited C2C12 migration (Figure 4A). Then, KEGG enrichment analysis was performed on the significant hub genes in clusters 1and2. Among significantly enriched pathways in KEGG enrichment analysis, “regulation of actin cytoskeleton,” “focal adhesion,” “adherens junction,” and “cell adhesion molecules” were related to cell migration, adhesion, and fusion (Figure 4B). Skeletal muscle comprises multinucleated muscle fibers (myotubes), depend on myoblast migration, adhesion, and fusion during the differentiation of mononuclear myoblasts (Figure 4C).

Figure 4.

Stearic acid (SA) regulates the migration and adhesion of the C2C12 myoblast. (A) SA regulates the migration of the C2C12 myoblast. (B) KEGG enrichment analysis of top 10 hub genes in clusters 1 & 2. *p < .05. (C) SA regulates the working model of SMM.

Discussion

Our study generated a list of 65 SMM-associated metabolites that showed potential for improving the risk prediction, understanding the pathogenesis, and providing new targets for preventative and therapeutic treatment for sarcopenia in elderly women. Unfortunately, our results did not confirm these findings in previous sarcopenia-related metabolomics studies (7–9). The reason for the discrepancy is probably because of the difference of lifestyle, diets, populations, and ethnicity. More independent participate samples, in vivo, and in vitro experiments are needed to further validate the relationships between these SMM-associated metabolites and SMM identified in this present study. Among these 65 significant metabolites, SA was negatively associated with all SMIes and was a potential causal candidate for decreasing SMM, as determined by MR analysis. Subsequent in vitro functional experiments suggested that SA could significantly inhibit differentiation of myoblasts and mature myotube development at mRNA, protein, and phenotype levels. Additionally, the transcriptomic analysis showed that SA might regulate myoblast migration, adhesion, and fusion to inhibit development of mature myotubes. The collective results support and highlight the potential importance of SA in regulating SMM (Figure 4C).

Emerging evidence supported the role of lipids and their derivatives in regulating SMM through their ability to modulate muscle cell growth (18). It has been reported that saturated fatty acids are directly associated with liver fat and liver fat markers (19,20), and a high dietary intake exacerbates the development of sarcopenia (21). Previous studies showed that dietary saturated fatty acids as major risk factors for the amplification of skin inflammation, via an amplified proinflammatory immune response to toll-like receptor-like stimuli (22). The proinflammatory factors induce infiltration of macrophages and other immune cells into muscle and other tissues, which secrete a large amount of proinflammatory cytokines and chemokines, thus broadening the local chronic inflammation into a systemic inflammation condition in muscles and other tissues (23). Skeletal muscle tissue responds to proinflammatory cytokines with atrophy and sarcopenia (24).

Excessive delivery of palmitic acid, a major long-chain saturated fatty acid, could result in lipid oversupply to skeletal muscle and loss of muscle mass (25). A clinical randomized controlled trial recruited 39 young and normal-weight individuals were randomized to eat muffins containing either sunflower oil (high in the major dietary polyunsaturated fatty acids linoleic acid) or palm oil (high in the major saturated fatty acids palmitic acid) for 7 weeks found that the weight gain of the 2 diet groups was the same (+1.6 kg) (25). However, the saturated fatty acids group gained more liver fat, total fat, and visceral fat, but less lean tissue than subjects in the polyunsaturated fatty acids group. Conversely, compared with subjects in the saturated fatty acids group, the polyunsaturated fatty acids group had a nearly 3-fold larger increase in lean tissue (25). Consistently, our study found that palmitic acid was negatively associated with both ALM/weight and ALM/BMI, and linoleic acid was positively associated with ALM/BMI.

In the present study, we also identified that another significant saturated fatty acid, SA, was negatively associated with all SMIes. Additionally, we validated that SA was causally negatively associated with SMM by the one-sample MR analysis. An excessive intake of SA leads to metabolic disorders of glucose and lipids (26). In C2C12 myotubes, glucose oxidation was impaired by 30% and 34% by SA in the absence or presence of insulin, respectively (27). Additionally, glycogen synthesis in primary culture of rat skeletal muscle cells and ATP generation in response to pyruvate was decreased by 45% and 34% by SA, respectively (27). These results suggested that SA might induce musculoskeletal cell injuries. As C2C12 cells possess features that resemble the differentiated muscles tissue of human myotubes (28), this enables the use of such cell lines for muscle mass and function studies, such as muscle atrophy and contraction/exercise-like stimuli (29,30). The differentiation of muscle stem cells effectively alleviates the loss and atrophy of skeletal muscle, repair skeletal muscle damage, and promote skeletal muscle development (31,32). It starts with activated muscle stem cells differentiating into myoblasts, which migrate and fuse into multinuclear myotubes. MyoD expression is a sign of muscle stem cell activation in the early stage of muscle differentiation, promoting the transformation of quiescent muscle satellite cells into myoblasts (33,34). It plays a role as a master switch in the regulation of muscle-specific gene transcription. MyoG is responsible for initiating the terminal differentiation of myoblasts, prompting the cells to change from oblate to fusiform, fuse into myotubes, and mature into muscle fibers (35). MyHC is the basic unit of myosin and is often used as a marker of muscle mass. We found that SA inhibited C2C12 myoblast differentiation by observing the adverse changes in the myotube phenotype and expression of myogenic molecules (MyoD and MyoG). Additionally, the protein and mRNA levels of MyHC were significantly decreased by SA at the different differentiation times. Further transcriptome RNA sequencing data analysis also confirmed that SA could inhibit MyHC expression at both mid-differentiation and end-differentiation times of myoblasts. In the metabolomics and MR analysis, we also observed that increased serum SA might have a causal effect on decreased SMM in early postmenopausal women. These findings imply that SA is a risk factor for sarcopenia, which motivates us to explore further the potential mechanisms of SA on decreasing SMM variation.

To further investigate the potential mechanisms underlying the effect of SA on decreased SMM, we conducted in vitro experiment and transcriptomics bioinformatics analyses. The fusion of myoblasts was greatly affected by aging and was the main reason for the irreversible loss of skeletal muscle function in the elderly adults (36,37). Myoblast fusion is one of the crucial steps in myoblast differentiation (38), the defect in myoblast fusion is the leading cause of muscle diseases (39). The fusion process follows an ordered set of cellular events, including cell migration, alignment, adhesion, and membrane fusion (38). As a critical link in the fusion and regeneration of myoblasts, cell migration plays an essential role in developing skeletal muscle and its related diseases (40). In the present study, the significant hub genes of clusters 1 and 2 were significantly enriched in the biological process of “response to wounding” and “wound healing,” and molecular function of “growth factor binding” and “growth factor receptor binding,” which were closely related to cell migration (41). Our wound-healing scratch experiment further confirmed that SA significantly inhibits cell migration during the induction of myogenic differentiation. Impaired myoblast migration could reduce myoblast fusion (42,43) and impaired SMM. Among the significant signaling pathways enriched in KEGG enrichment analysis, “regulation of actin cytoskeleton,” “focal adhesion,” “adherens junction,” and “cell adhesion molecules” were closely related to cell adhesion, indicating that SA inhibits myoblast fusion might occur via cell adhesion.

Our study has several strengths. First, the nonhypothesis-driven untargeted metabolomics approach enabled the study to discover novel metabolites and pathways involved in muscle regulation. Second, we used a series of SMIes to evaluate SMM comprehensively and selected metabolites that were significantly associated with all SMIes for the follow-up study to increase the reliability of the results. Third, we integrated multiple omics (metabolomics and genomics), in vitro experiments, and transcriptomics and bioinformatic analyses and illustrated that increased serum SA is a risk factor for decreased SMM, providing essential knowledge for understanding the pathogenesis of sarcopenia in elderly women. However, there were still several limitations in our study. First, our study did not assess skeletal muscle function (such as measuring grip strength and walking pace), which limited our understanding of the relationships between sarcopenia and serum metabolites. In the present study, approximately half of early postmenopausal women had a high level of physical activity in Guangzhou, China, which might not represent the whole population of postmenopausal women, and more data are needed to confirm this conclusion. Another potential limitation is that no detailed information on nutritional factors was assessed and adjusted, such as protein intake which is an important factor for sarcopenia. An additional limitation of this study is that we performed only in vitro experiments to illustrate the potential machine of SA on decreased SMM that cannot fully recapitulate in vivo functionality. For future clinical applications, in vivo feasibility should be determined.

Conclusion

This study used an untergeted metabolomics approach, followed by functional studies, to assess metabolic changes associated with SMM variation in early postmenopausal women and the potential mechanism of action. Our collective results suggested that SA plays a vital role in SMM metabolism by repressing the release of myogenic factors and regulating the migration, adhesion, fusion, and development of myoblasts. These findings provided novel insights into new biomarkers and potential pathophysiological mechanisms for preventing or alleviating SMM loss in elderly women.

Funding

This work was partially supported or benefited by grants from the National Institutes of Health [U19AG05537301, R01AR069055, R01AG061917, U54MD007595 to H.W.D. and H.S.] and the National Key R&D Program of China (2017YFC1001100 to H.M.X.).

Conflict of Interest

None declared.

Author Contributions

H.W.D. conceived, initiated, and directed the project. H.M.L. conducted the data analysis, cell experiments and drafted the manuscript. H.M.X. managed the study. X.L. and J.S. collected samples and clinical phenotypes. R.G. contributed to the data analysis. Z.H.Q. contributed to the cell experiments. H.W.D. revised and finalized the manuscript. Q.Z. and H.S. contributed to manuscript revision and discussion. H.M.L., H.W.D., H.M.X., J.S., and X.L. in writing of the review and editing.