This cross-sectional study was conducted at Sichuan Provincial People’s Hospital from January to June 2025. The study protocol was approved by the Institutional Ethics Committee (Protocol No. 2025192), and written informed consent was obtained from all participants. Older adults aged 60–99 years who were capable of completing all physical and clinical assessments were eligible for inclusion. Exclusion criteria were as follows: (1) active inflammatory disease within the preceding 3 months; (2) presence of malignancy, autoimmune disease, liver disease, or renal dysfunction (estimated glomerular filtration rate [eGFR] <60 mL/min/1.73 m²); and (3) recent acute cardiovascular or cerebrovascular events. Ultimately, 139 older adults were enrolled and stratified into sarcopenic and non-sarcopenic groups based on established diagnostic criteria (Chen et al., 2020). In parallel, 20 younger adults (aged 18–30 years) matched for sex and body mass index (BMI) were also recruited. Fasting venous blood samples were collected in the early morning and processed for subsequent analyses.

Skeletal muscle mass was assessed using bioelectrical impedance analysis (BIA) (InBody 770, InBody, China). Assessments were performed in accordance with the manufacturer’s instructions: participants stood barefoot on the platform, aligning their heels and toes with the electrodes, and gripped the hand electrodes while keeping their arms slightly abducted to prevent torso contact. The appendicular skeletal muscle mass index (ASMI) was calculated as the sum of the lean mass of the four limbs (kg) divided by height squared (m²). Handgrip strength was quantified using a digital dynamometer (EH101, CAMRY, China). Participants performed a maximal isometric contraction using their dominant hand in a standing position with the elbow flexed at 90°. The maximum value from two trials was recorded. Gait speed was evaluated over a 6-m distance. To allow for acceleration and deceleration, participants walked along an 8-m course, and the time taken to traverse the central 6 m (1-m to 7-m marks) was recorded. The faster of two attempts was used for analysis. Sarcopenia was diagnosed according to the 2019 Asian Working Group for Sarcopenia (AWGS) consensus (Chen et al., 2020), defined as the presence of low muscle mass (ASMI <7.0 kg/m² for men, <5.7 kg/m² for women) combined with either low handgrip strength (<26 kg for men, <18 kg for women) or low gait speed (<1.0 m/s).

A post hoc power analysis was conducted utilizing G*Power software 3.1 to assess the statistical robustness of the primary outcome, circulating GDF11. At a significance level of α = 0.05, the comparison between younger adults (n = 20; 37.96 ± 18.61 pg/mL) and older adults (n = 139; 91.19 ± 46.05 pg/mL) demonstrated a statistical power exceeding 0.99. Within the older adult cohort, the comparison between sarcopenic (n = 98; 108.26 ± 35.55 pg/mL) and non-sarcopenic individuals (n = 41; 84.04 ± 48.23 pg/mL) resulted in a power of 0.86. These findings suggest that the study was adequately powered (power >0.80) to detect clinically significant differences in circulating GDF11 levels.

Information regarding socio-demographic characteristics (age, sex, and BMI), lifestyle factors (smoking status and alcohol consumption), and comorbidities (hypertension, diabetes, and hyperlipidemia) was collected. Physical activity was evaluated via the International Physical Activity Questionnaire (IPAQ). Participants were categorized into low, moderate, or vigorous activity groups according to their calculated weekly energy expenditure.

Serum GDF11 concentrations were measured utilizing a human GDF11 enzyme-linked immunosorbent assay (ELISA) kit (RAB1480, Sigma-Aldrich, United States) in accordance with the manufacturer’s protocol.

Serum 25-hydroxyvitamin D [25(OH)D] levels were quantified using a ELISA kit (SEKH-0341, Solarbio, China) following the manufacturer’s protocol. Vitamin D deficiency (VDD) was defined as a serum 25(OH)D concentration of <20 ng/mL.

Two-sample Mendelian randomization (MR) analyses were performed to investigate the potential causal effects of physical activity on sarcopenia-related traits using the TwoSampleMR package (Hemani et al., 2017). Summary-level genetic data were retrieved from the IEU OpenGWAS database (https://gwas.mrcieu.ac.uk). The exposure variable was physical activity, defined as strenuous sports participation (ukb-b-151; n = 335,599; 10,894,596 SNPs). Outcome variables included appendicular lean mass (ebi-a-GCST90000025; n = 450,243; 18,071,518 SNPs), usual walking pace (ukb-b-4711; n = 459,915; 9,851,867 SNPs), and low handgrip strength in individuals aged ≥60 years (ebi-a-GCST90007526; n = 256,523; 9,336,415 SNPs). Instrumental variables (IVs) were rigorously selected based on genome-wide significance (P < 5 × 10⁻⁹). To ensure independence, SNPs were pruned for linkage disequilibrium (LD) utilizing a strict threshold of r ² < 0.0001 within a 25,000 kb window. To mitigate weak instrument bias, only variants with an F-statistic >10 were retained. Causal estimates were calculated using the Inverse Variance Weighted (IVW) method as the primary analysis. Complementary sensitivity analyses were conducted to ensure robustness, including MR-Egger regression (to detect pleiotropy), weighted median, and weighted mode methods. Heterogeneity was assessed via Cochran’s Q test; if significant heterogeneity was observed (P < 0.05), a random-effects IVW model was employed; otherwise, a fixed-effects model was used. Horizontal pleiotropy was monitored using the MR-Egger intercept test. Additionally, the MR-PRESSO framework was applied to detect and correct for potential outliers. Finally, to account for multiple comparisons, the Benjamini–Hochberg false discovery rate (FDR) correction was applied across all exposure–outcome pairs.

To investigate the transcriptomic landscape associated with GDF11, we systematically searched the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and identified six datasets meeting the inclusion criteria (Table 1). For sarcopenia profiling, three RNA-sequencing datasets (GSE111016, GSE167186, and GSE226151) were selected using the terms “sarcopenia” and “Homo sapiens”, comprising skeletal muscle biopsies from older adults with sarcopenia and age-matched controls. Raw read counts were normalized to transcripts per million (TPM) to correct for sequencing depth and gene length–dependent biases. Differential expression analysis was performed using the limma package (Ritchie et al., 2015). Significant differentially expressed genes (DEGs) were defined as P < 0.05 and |log2FC| > 0.5. Functional enrichment of DEGs was assessed using Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses via the clusterProfiler package (Wu et al., 2021), applying an P < 0.05.

Given the limited availability of transcriptomic data from sarcopenic individuals undergoing structured exercise interventions, we extended the analysis to healthy young cohorts using the keywords “exercise” and “Homo sapiens”. Two RNA-sequencing datasets (GSE226973 and GSE235781) and one microarray dataset (GSE250122) were included. For RNA-sequencing datasets, TPM-normalized expression values were used; for the microarray dataset, normalized expression matrices provided by the original study were analyzed. All exercise-related datasets were derived from vastus lateralis muscle biopsies, a mixed-fiber muscle enriched in type II fibers that is particularly susceptible to age-related atrophy and critical for functional mobility (Nederveen et al., 2020; Prior et al., 2016). Exercise-induced changes in GDF11 expression were assessed using paired t-tests, with P < 0.05 considered statistically significant.

To construct the protein–protein interaction (PPI) network for GDF11, we integrated data from the STRING (https://string-db.org/) and GeneMANIA (http://www.genemania.org) databases, restricted to Homo sapiens. The resulting network was analyzed to identify key interacting proteins. Subsequently, Gene Ontology biological process (GO-BP) enrichment analysis was performed to elucidate the biological pathways and signaling cascades associated with GDF11.

The three-dimensional structure of the ACVR2B–GDF11 complex was predicted from amino acid sequences using AlphaFold 3 (https://alphafoldserver.com/; UniProt IDs: Q13705 and O95390). The predicted complex was processed using the Protein Preparation Wizard in Schrödinger Suite (2019-01) to correct bond orders and add hydrogen atoms. To resolve steric clashes, energy minimization was performed using the OPLS3e force field with an aqueous solvation model. This optimization followed a two-step protocol comprising steepest descent and conjugate gradient minimization (5,000 iterations each), followed by constrained geometry optimization. Finally, the binding free energy of the complex was calculated using the MM-GBSA method (Sampling: Minimize; Solvation: VSGB; Force Field: OPLS3e). Structural visualization was performed using PyMOL 2.1.

C2C12 murine myoblasts (ZQ0092, Zhongqiaoxinzhou Biotech, China) were seeded in culture plates and maintained in growth medium consisting of Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C in a humidified atmosphere containing 5% CO₂. When cells reached approximately 70%–80% confluence, the growth medium was replaced with differentiation medium (DMEM supplemented with 2% horse serum and 1% penicillin–streptomycin) to induce myogenic differentiation. The differentiation medium was refreshed every 48 h C2C12 myoblasts were continuously induced to differentiate for 5 days, during which elongated, multinucleated myotubes were formed, consistent with established criteria for myotube maturation (Wang M. Y. et al., 2024; Kim et al., 2025). Physiological circulating GDF11 levels in rodent models are typically reported to range from 2 to 10 ng/mL (Egerman et al., 2015; Kraler et al., 2023; Garbern et al., 2019), and in vitro studies commonly employ recombinant GDF11 (rGDF11) at concentrations between 10 and 200 ng/mL (Frohlich et al., 2022; Zhao et al., 2020; Sutherland et al., 2020; Hammers et al., 2017; Sharma and McNeill, 2009; Idkowiak-Baldys et al., 2019; Su et al., 2019). Based on these established conditions, differentiated C2C12 myotubes were treated with rGDF11 (120-11-20UG, PeproTech, United States) at final concentrations of 0–100 ng/mL for 48 h.

Total cellular proteins were extracted from C2C12 cells using RIPA lysis buffer supplemented with protease inhibitors. Cell lysates were incubated on ice for 30 min and centrifuged to collect the supernatants. Protein concentrations were determined using a standard protein quantification assay. Equal amounts of protein were separated by 10% SDS–PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were blocked with 5% skimmed milk for 1 h at room temperature and subsequently incubated with primary antibodies against Smad3 (ab40854, Abcam, United States), phosphorylated Smad3 (p-SMAD3; #9520, CST, United States), Atrogin-1 (67172-1-Ig, Proteintech, China), MuRF1 (55456-1-AP, Proteintech, China) and GAPDH (1E6D9, Proteintech, China), with GAPDH serving as the loading control. After incubation with HRP-conjugated secondary antibodies, protein bands were detected using enhanced chemiluminescence (ECL) reagents and visualized by autoradiography.

Total RNA was isolated using TRIzol reagent (R0016, Beyotime, China) and reverse-transcribed into cDNA using the PrimeScript RT reagent kit (R233-01, Vazyme, China). Quantitative real-time PCR (RT-qPCR) was conducted using SYBR Green Master Mix (Q341-02, Vazyme, China) on a QuantStudio 5 Real-Time PCR System (Applied Biosystems). Transcript levels of Atrogin-1 and MuRF1 were quantified using the 2^−ΔΔCt method after normalization to Gapdh, with the control group set to 1. The primer sequences used in this study are listed in Table 2.

Continuous variables are presented as mean ± standard deviation (SD), unless otherwise specified, and categorical variables as frequencies and percentages. Intergroup comparisons utilized Student’s t-test or the Mann–Whitney U test for continuous data, and the chi-square test for categorical variables. To minimize potential confounding, propensity score matching (PSM) was performed using a 1:1 nearest-neighbor algorithm based on relevant covariates. Variables with a P < 0.1 in univariate analysis were advanced to a multivariate logistic regression model. Significant independent predictors (P < 0.05) were subsequently used to construct a nomogram for sarcopenia risk prediction. Mediation analysis was executed using Model 4 of Hayes’ PROCESS macro 3.4 for SPSS. To evaluate the significance of indirect effects, bootstrapping was performed with 5,000 resamples to generate 95% bias-corrected confidence intervals (CIs); effects were deemed significant if the 95% CI excluded zero. All analyses were conducted using SPSS 23.0 and R 4.1, with a two-sided P < 0.05 defining statistical significance.

The workflow diagram for this study is presented in Figure 1. A total of 159 participants were recruited, consisting of 20 younger adults (22.25 ± 3.01 years) and 139 older adults (72.75 ± 6.89 years). Notably, circulating GDF11 levels were significantly elevated in the older cohort compared to the younger group (91.19 ± 46.05 pg/mL vs. 37.96 ± 18.61 pg/mL; P < 0.001; Table 3). Subsequently, the older participants were stratified into sarcopenia (SG, n = 41) and non-sarcopenia (NSG, n = 98) groups. The SG group demonstrated expected deficits in clinical characteristics such as gait speed, grip strength, and ASMI (all P < 0.001), along with differences in age, sex, BMI, and physical activity (all P < 0.05). Importantly, GDF11 levels were significantly higher in the SG compared to the NSG (108.26 ± 35.55 vs. 84.04 ± 48.23 pg/mL; P = 0.004). To mitigate potential confounding bias, a 1:1 PSM was performed. Post-matching analysis confirmed that the elevation in GDF11 levels remained significant (107.04 ± 36.83 vs. 74.52 ± 45.53 pg/mL; P = 0.003) (Table 4).

To identify the clinical determinants of sarcopenia, both univariate and multivariate logistic regression analyses were conducted, as detailed in Table 5. The univariate analysis revealed that advanced age, male sex, and elevated circulating GDF11 levels were significantly associated with increased odds of sarcopenia (all P < 0.05), while physical activity was linked to decreased odds (P < 0.05). Variables with a P-value less than 0.1 in the univariate analysis were included in the multivariate model. Upon adjustment, age (OR = 1.11, P = 0.002), moderate physical activity (OR = 0.28, P = 0.008), and circulating GDF11 levels (OR = 1.01, P = 0.013) remained independently associated with sarcopenia.

A prognostic nomogram was developed based on these independent predictors (Figure 2A). The model exhibited satisfactory discriminative ability, with an area under the receiver operating characteristic curve (AUC) of 0.794 (Figure 2B). Calibration analysis demonstrated a strong concordance between predicted and observed risks (Figure 2C), with a calibration slope of 1.000 and an intercept of 0.000. The model’s fit was further corroborated by a low Brier score (0.144) and a non-significant Spiegelhalter’s p-value (P = 0.793). These findings indicate satisfactory internal performance; however, external validation in independent cohorts is warranted to assess generalizability and clinical applicability.

In order to identify factors associated with circulating GDF11 levels, a multivariable linear regression analysis was performed, adjusting for variables including age, sex, BMI, smoking status, alcohol consumption, hypertension, diabetes, hyperlipidemia, physical activity, and VDD (Table 6). Among the covariates analyzed, physical activity emerged as the only factor significantly associated with circulating GDF11 levels (P = 0.037). The analysis revealed negative values for both the standardized (β = −0.18) and unstandardized (B = −13.16) coefficients, suggesting an inverse relationship between physical activity and circulating GDF11 levels.

Physical exercise is a well-recognized intervention for the prevention of muscle atrophy (Lu et al., 2021). To evaluate the potential causal effects of physical activity on traits associated with sarcopenia, a MR analysis was conducted, focusing on appendicular lean mass, usual walking pace, and low hand grip strength (aged ≥60 years). The analysis revealed a significant positive causal relationship between physical activity and appendicular lean mass (IVW OR = 3.24, 95% CI 3.07–3.44, FDR <0.001; Figures 3A,B) as well as usual walking pace (IVW OR = 3.35, 95% CI 2.97–3.77, FDR <0.001; Figures 3C,D). In contrast, increased physical activity was causally linked to a decreased risk of low hand grip strength (IVW OR = 0.18, 95% CI 0.15–0.22, FDR <0.001; Figures 3E,F). Throughout the analyses, no substantial heterogeneity or horizontal pleiotropy was detected after correction for MR-PRESSO–identified outliers, supporting the robustness of the causal estimates.

Given the observed associations among physical activity, circulating GDF11, and sarcopenia, mediation analysis was conducted. Results indicated that physical activity exerted a significant direct effect on sarcopenia (B = −0.87, P = 0.009). Notably, a significant indirect effect mediated by decreased circulating GDF11 levels was also identified (B = −0.13; 95% CI, −0.37 to −0.01). This mediation effect accounted for 13.0% of the total effect, raising the possibility that the downregulation of circulating GDF11 might contribute to the beneficial impact of physical activity on sarcopenia (Table 7).

To assess GDF11 expression in the context of sarcopenia, three independent transcriptomic datasets (GSE111016, GSE167186, and GSE226151) were analyzed. No significant differences in GDF11 mRNA expression were observed between sarcopenic and non-sarcopenic individuals across all datasets (Figures 4A–C). Intersection analysis of differentially expressed genes (DEGs) revealed minimal overlap among the three datasets, indicating substantial molecular heterogeneity associated with sarcopenia (Figure 4D). Given this heterogeneity, DEGs were subsequently integrated across datasets, yielding a total of 452 genes, including 238 upregulated and 214 downregulated genes. Functional enrichment analysis further showed that upregulated genes were predominantly enriched in inflammatory pathways, such as cytokine–cytokine receptor interaction, TNF signaling, and JAK–STAT signaling, whereas downregulated genes were mainly associated with metabolic pathways and oxidative phosphorylation (Figures 4E,F).

To further examine whether exercise influences GDF11 expression in skeletal muscle, three exercise-intervention transcriptomic datasets (GSE226973, GSE235781, and GSE250122) were analyzed. Consistently, no significant changes in GDF11 mRNA levels were detected before and after exercise interventions (Figures 4G–I). Collectively, these findings suggest that the elevated circulating GDF11 observed in sarcopenia is unlikely to originate primarily from skeletal muscle, and that muscular GDF11 expression appears largely unresponsive to exercise.

Given the elevated circulating GDF11 levels observed in patients with sarcopenia, we sought to explore the mechanistic basis underlying its potential pro-atrophic effects. PPI network analysis identified a prominent interaction cluster centered on GDF11, featuring strong associations with activin receptors (ACVRs), particularly activin A receptor type 2B (ACVR2B), a well-established regulator of skeletal muscle mass and muscle atrophy (Hulmi et al., 2018; Lee et al., 2020) (Figures 5A,C). Consistent with this, GO-BP enrichment analysis revealed that GDF11-associated proteins were significantly enriched in ACVRs, the TGF-β pathway, and SMAD phosphorylation cascades (Figures 5B,D). To further substantiate these findings at the structural level, protein–protein docking simulations demonstrated a high-affinity interaction between GDF11 and ACVR2B (binding free energy: −275.86 kcal/mol), stabilized by multiple hydrogen bonds, salt bridges, and hydrophobic interactions at the binding interface (Figure 5E).

Given that ACVR2B is a well-established upstream receptor that predominantly transduces signals via SMAD2/3 phosphorylation (Abazarikia et al., 2025), and that SMAD3 activation is a well-established driver of muscle atrophy through transcriptional induction of the E3 ubiquitin ligases MuRF1 and Atrogin-1 (Frohlich et al., 2022; Peris-Moreno et al., 2021; Roh et al., 2019; Lan et al., 2024), we next sought to experimentally validate the functional impact of GDF11 on myotube catabolic signaling. Differentiated myotubes were exposed to recombinant GDF11 (rGDF11; 0–100 ng/mL) for 48 h. Immunoblotting revealed a robust and dose-dependent increase in SMAD3 phosphorylation, without appreciable changes in total SMAD3 levels (Figures 5F,G). Consistently, rGDF11 treatment led to a progressive upregulation of the muscle-specific ubiquitin ligases MuRF1 and Atrogin-1 at both the protein and mRNA levels (Figures 5H–K). Together, these findings demonstrate that exogenous GDF11 activates the ACVR2B–SMAD3 signaling cascade and promotes a transcriptional program characteristic of muscle atrophy, implicating enhanced ubiquitin–proteasome–mediated proteolysis as a downstream effector mechanism.