Fifty male Sprague–Dawley rats (6 weeks old) were obtained and housed under controlled environmental conditions with a 12 h light–dark cycle, ambient temperature of 22 °C–24 °C, and free access to food and water. All animals underwent a 1 week acclimatization period prior to experimental procedures. All experimental protocols were approved by the Biomedical Research Ethics Committee of Hunan Normal University (Approval No. 2021345) and were conducted in accordance with institutional guidelines for the care and use of laboratory animals.

After acclimatization, rats were randomly assigned to a normal control group (NC, n = 10) or a diabetes induction group (n = 40). Rats in the NC group were fed a standard laboratory diet, whereas rats in the diabetes induction group were fed a high-fat diet for 4 weeks. Following this period, rats were fasted overnight (∼12 h) and administered a single intraperitoneal injection of streptozotocin (STZ; 35 mg/kg body weight), freshly dissolved in citrate buffer. Rats in the NC group received an equivalent volume of citrate buffer.

Fasting blood glucose was measured from the tail vein on days 3 and 8 after STZ injection using a handheld glucose meter. Rats with fasting blood glucose ≥7.8 mmol/L on both occasions were considered successfully diabetic and were included in subsequent experiments.

Successfully modeled diabetic rats were randomly allocated into four groups: diabetes control (DC, n = 10), diabetes aerobic exercise (DA, n = 10), diabetes resistance exercise (DR, n = 10), and diabetes combined aerobic–resistance exercise (DAR, n = 10). Rats in the NC and DC groups did not undergo exercise intervention. All animals had ad libitum access to food and water throughout the experimental period. Body weight and fasting blood glucose were recorded weekly.

Exercise intervention was conducted using a motorized animal treadmill (ZH-PT, HuaiBei ZhengHua Biological Instrument Co., Ltd., China). Prior to formal training, rats in the DA, DR, and DAR groups completed a 3 day adaptation period at 10–15 m/min for 30 min/day to familiarize them with treadmill running.

For aerobic exercise, rats in the DA group ran at a speed of 17 m/min, corresponding to moderate-intensity aerobic exercise in rodents. For resistance exercise, rats in the DR group ran at 15 m/min while carrying progressively increased external loads attached to the tail. Load intensities were set at 10% of body weight during weeks 1–2, 30% during weeks 3–4, 50% during weeks 5–6, and 70% during weeks 7–8. Each resistance training session consisted of repeated running bouts separated by brief rest intervals. This protocol was designed to provide a load-dominant mechanical stimulus during locomotion and, consistent with rodent treadmill-based resistance models, necessarily retained an endurance component rather than representing an isolated resistance paradigm.

Rats in the DAR group alternated between aerobic exercise (Mondays, Wednesdays, Fridays) and resistance exercise (Tuesdays, Thursdays, Saturdays). Exercise intensity, duration, and progression were matched to those of the DA and DR groups. All exercise-trained rats performed 60 min of exercise per day, 6 days per week, for eight consecutive weeks. Rats in the NC and DC groups were handled similarly but did not perform exercise.

A schematic overview of the experimental design, diabetes induction, group allocation, and exercise intervention timeline is presented in Figure 1.

At the end of the 8 week intervention period, rats were fasted overnight (∼12 h) and anesthetized by intraperitoneal injection of 1% sodium pentobarbital (40 mg/kg). Blood samples were collected from the thoracic aorta, allowed to clot at room temperature for 30 min, and centrifuged at 4 °C and 4,000 rpm. Serum was separated and stored at −20 °C until analysis.

Bilateral gastrocnemius muscles were excised, rinsed with cold saline, blotted dry, and weighed. The right gastrocnemius muscle was fixed in 4% paraformaldehyde for histological analysis, and the left gastrocnemius muscle was snap-frozen and stored at −80 °C for biochemical and protein analyses.

Fasting blood glucose was measured using the glucose oxidase method with a commercial assay kit (Nanjing Jiancheng Bioengineering Institute, China). Serum insulin, glycated hemoglobin (HbA1c), and irisin levels in serum and muscle tissue were quantified using enzyme-linked immunosorbent assay kits (Shanghai Enzyme-Linked Technology Bio-Tech Co., Ltd., China) according to the manufacturers’ instructions. Insulin resistance was estimated using the homeostasis model assessment of insulin resistance (HOMA-IR), calculated as insulin (mIU/L) × fasting blood glucose (mmol/L)/22.5.

Gastrocnemius muscles from three randomly selected rats per group were dehydrated, embedded in paraffin, and sectioned at 3 μm thickness. Sections were stained with hematoxylin–eosin (HE) to assess muscle fiber morphology and with Masson’s trichrome to evaluate collagen deposition and fibrosis. Images were captured using an Eclipse Ci-L light microscope (Nikon, Japan). Muscle fiber cross-sectional area was quantified using Image-Pro Plus 6.0 software. Collagen volume fraction was calculated from Masson-stained sections using standard image analysis procedures.

Total skeletal muscle protein concentration was determined using a bicinchoninic acid protein assay kit (Nanjing Jiancheng Bioengineering Institute, China) following homogenization of muscle tissue in lysis buffer.

Protein extraction from gastrocnemius muscle was performed using samples from six randomly selected rats per group. Tissues were homogenized in RIPA lysis buffer, and equal amounts of protein were separated by SDS–PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were blocked with 5% non-fat milk at room temperature and incubated overnight at 4 °C with primary antibodies against MuRF1, Atrogin-1, PI3K, AKT, mTOR, integrin β1 (Abcam), and integrin α7 (Bio-Rad). After washing, membranes were incubated with horseradish peroxidase–conjugated secondary antibodies. Protein bands were visualized using enhanced chemiluminescence and quantified by densitometry using Image-Pro Plus 6.0 software. GAPDH or β-actin was used as the internal loading control, as appropriate.

Statistical analyses were structured around three biological questions: whether induction of diabetes produced a measurable pathological phenotype relative to normal controls (family 1: NC vs DC), whether exercise intervention rescued the diabetic phenotype relative to diabetic controls (family 2: DC vs DA, DR, DAR), and which exercise modality conferred greater efficacy among trained groups (family 3: DA vs DR, DAR; DR vs DAR). For each planned pairwise contrast, group differences were evaluated using two-sample tests (Welch’s t-test or Wilcoxon rank-sum test, as appropriate based on distributional characteristics). In addition, baseline-to-week-8, within-group changes for body weight were assessed using two-sample tests. P values were adjusted for multiple testing within each contrast family using the Holm method. Effect sizes were reported as Hedges’ g for parametric contrasts and Cliff’s δ for nonparametric contrasts. All tests were two-tailed, and P < 0.05 was considered statistically significant. Statistical analyses were performed using R. The data that support the conclusions of the study can be accessed at Figshare (DOI: 10.6084/m9.figshare.31286356).

Rats subjected to high-fat feeding followed by streptozotocin injection developed a stable diabetic phenotype. Compared with NC, DC exhibited significantly elevated fasting blood glucose, glycated hemoglobin, and HOMA-IR values, confirming successful induction of hyperglycemia and insulin resistance (Figure 2). Circulating insulin levels were also significantly lower in DC than in NC.

Exercise training improved selected metabolic parameters in diabetic rats. Compared with DC, fasting blood glucose and HOMA-IR were significantly reduced in DA, DR, and DAR (Figures 2A,D). In contrast, glycated hemoglobin levels did not differ significantly between DC and any exercise condition (Figure 2C). Circulating insulin levels were not significantly altered by exercise training, and no significant differences were observed among DA, DR, and DAR (Figure 2B).

No significant differences were detected among exercise modalities for fasting blood glucose, glycated hemoglobin, insulin, or HOMA-IR.

DC exhibited a progressive reduction in body weight over the experimental period compared with NC (Figure 3A). Within-group analysis showed that body weight increased significantly from week 0 to week 8 in NC (P = 0.026, Cliff’s δ = 1.00). In contrast, DC showed a reduction in body weight over the same period that did not reach statistical significance (P = 0.056, Cliff’s δ = 0.86). No significant within-group changes in body weight were detected in DA, DR, or DAR across the intervention period.

Despite the absence of significant within-group weight gain, exercise training significantly attenuated diabetes-associated weight loss (Figure 3A). At week 8, body weight was significantly higher in DA, DR, and DAR compared with DC. No significant differences in body weight were observed among DA, DR, and DAR at any time point.

Of note, food intake was not directly measured in the present study; therefore, differences in body weight should be interpreted as reflecting net energy balance rather than isolated differences in caloric intake. Accordingly, exercise-associated changes in body weight may arise from differences in energy expenditure, substrate utilization, or tissue partitioning.

Assessment of skeletal muscle mass revealed significant reductions in gastrocnemius muscle weight, gastrocnemius-to-body weight ratio, and muscle fiber cross-sectional area in DC compared with NC (Figures 3B–D). Exercise training significantly increased absolute gastrocnemius muscle mass in DAR relative to DC, whereas no significant increases were observed in DA or DR (Figure 3B). The gastrocnemius-to-body weight ratio was significantly higher in DA and DAR compared with DC, but not in DR (Figure 3C). Muscle fiber cross-sectional area did not differ significantly between DC and any exercise condition (Figure 3D).

No significant differences in muscle mass indices or fiber cross-sectional area were observed among DA, DR, and DAR.

Histological analysis revealed uniform muscle fiber size and orderly arrangement in NC, whereas DC displayed greater fiber size heterogeneity and disrupted organization (Figure 4A). Exercise training did not result in statistically significant differences in these histological features relative to DC.

Masson’s trichrome staining showed increased interstitial collagen deposition in DC compared with NC (Figure 4A). Quantitative analysis confirmed a significantly higher collagen volume fraction in DC (Figure 4B). Exercise training did not significantly reduce collagen volume fraction in any exercise condition, and no significant differences were observed among DA, DR, and DAR.

Total skeletal muscle protein content was significantly reduced in DC compared with NC (Figure 5A). Exercise training significantly increased muscle protein content in DA, DR, and DAR relative to DC. Muscle protein levels were significantly higher in DR and DAR than in DA, and were also significantly higher in DAR than in DR.

Protein expression levels of the ubiquitin–proteasome markers MuRF1 and atrogin-1 did not differ significantly between NC and DC (Figures 5B–D). Exercise training did not significantly alter the expression of either marker relative to DC, and no significant differences were observed among DA, DR, and DAR.

Both serum and skeletal muscle irisin levels were significantly reduced in DC compared with NC (Figures 6A,B). Exercise training significantly increased serum irisin and skeletal muscle irisin levels in DA, DR, and DAR relative to DC. No significant differences in irisin levels were observed among DA, DR, and DAR.

Expression of integrin α7 did not differ significantly between NC and DC, whereas integrin β1 expression was significantly reduced in DC (Figures 6C–E). Exercise training significantly increased integrin α7 and integrin β1 expression in DR and DAR relative to DC, but not in DA. Integrin α7 expression was significantly higher in DR and DAR than in DA, whereas no significant difference was observed between DR and DAR.

PI3K expression did not differ significantly between NC and DC and was not significantly altered by exercise relative to DC (Figures 6F, H). However, PI3K expression was significantly higher in DR and DAR than in DA, with no significant difference between DR and DAR.

Expression levels of AKT and mTOR were significantly reduced in DC compared with NC (Figures 6F,G,I,J). Exercise training significantly increased AKT and mTOR expression in DA, DR, and DAR relative to DC. No significant differences were detected among DA, DR, and DAR for AKT or mTOR expression.

Consistent with established models of T2DM, diabetic rats displayed pronounced metabolic dysfunction, including hyperglycemia and insulin resistance, alongside reductions in skeletal muscle mass and total muscle protein content. Exercise training improved selected metabolic indices across all exercise modalities; however, glycated hemoglobin and circulating insulin levels were not significantly altered, indicating that metabolic rescue was partial rather than complete. Importantly, these metabolic improvements were not accompanied by consistent increases in muscle fiber cross-sectional area or reductions in fibrosis, underscoring that metabolic and structural adaptations can be dissociated in diabetic skeletal muscle (Espino-Gonzalez et al., 2024).

A defining outcome of this study is that exercise-induced increases in muscle protein content occurred largely in the absence of overt hypertrophy. Absolute gastrocnemius muscle mass increased significantly only with combined aerobic–resistance exercise, whereas muscle fiber cross-sectional area remained unchanged across all exercise modalities. These findings demonstrate that protein accretion can occur independently of detectable fiber enlargement, supporting the existence of a non-hypertrophic anabolic state (Joanisse et al., 2013). Such an adaptation may reflect increased myofibrillar protein density, altered intracellular protein composition, or expansion of non-contractile protein pools rather than classical hypertrophic growth. This distinction is particularly relevant in diabetic muscle, where structural remodeling may be constrained by extracellular matrix accumulation, impaired satellite cell activity, or altered neuromuscular activation (Orlando et al., 2016; Han et al., 2022; Wu and Coletta, 2025).

Notably, exercise-induced protein accretion was not accompanied by suppression of canonical ubiquitin–proteasome system markers. Expression of MuRF1 and atrogin-1 did not differ significantly between normal and diabetic controls and was not altered by any exercise modality. These results indicate that diabetes-related muscle protein loss in this model is not characterized by sustained activation of these E3 ubiquitin ligases and that exercise-mediated increases in muscle protein content occur independently of detectable changes in these proteolytic pathways. This observation challenges the prevailing assumption that exercise preserves muscle mass in diabetes primarily through inhibition of ubiquitin–proteasome–mediated degradation and instead suggests that enhanced anabolic processes dominate muscle adaptation under these conditions (Bodine and Baehr, 2014; O’Neill et al., 2018).

At the molecular level, these tissue-level dissociations are paralleled by coordinated but non-uniform changes across the irisin–integrin–PI3K/AKT/mTOR signaling axis. Both serum and skeletal muscle irisin levels were reduced in diabetic rats and increased with exercise training, with skeletal muscle irisin exhibiting an exercise-general response and circulating irisin showing greater sensitivity to resistance-containing interventions. Downstream, AKT and mTOR expression increased across all exercise modalities, whereas integrin α7, integrin β1, and PI3K displayed modality-dependent changes, with greater engagement in resistance and combined exercise conditions. Importantly, PI3K expression was not reduced in diabetic muscle relative to controls, nor was it uniformly “rescued” by exercise, indicating that exercise-related differences in PI3K signaling reflect patterning among exercise modalities rather than normalization from a diseased baseline.

Collectively, these findings support a model in which exercise induces patterned activation of anabolic signaling networks rather than uniform engagement of a linear signaling cascade. While irisin has been proposed as a myokine linking muscle contraction to anabolic and metabolic adaptation (Boström et al., 2012; Guo et al., 2023), and integrin α7β1 is a key mediator of mechanotransduction in skeletal muscle (Boppart et al., 2006), the present data demonstrate coordinated associations rather than receptor-mediated causality. Accordingly, the results should be interpreted as evidence of signaling alignment within an exercise-responsive network, not as proof of a direct irisin–integrin–PI3K–mTOR pathway.

An important implication of the present findings is that exercise modality confers selective, endpoint-specific advantages rather than universal superiority across outcomes. Combined aerobic–resistance exercise produced the largest increases in muscle protein content and was associated with greater engagement of selected integrin- and PI3K-related signaling components, suggesting heightened sensitivity to mechanical loading–related cues. However, modality differences were not observed for metabolic indices, muscle fiber cross-sectional area, fibrosis, or core AKT/mTOR responses. These results indicate that exercise modality exerts domain-specific effects, reinforcing the need to evaluate exercise adaptations across multiple biological levels rather than inferring overall efficacy from isolated endpoints.

Interpretation of the present muscle-level adaptations requires consideration of the broader neuromuscular system and the inherent translational boundaries of the experimental model. In humans, resistance training–induced improvements in strength and muscle mass are strongly influenced by neural and integrative neuromuscular adaptations, particularly during early phases of training (Lecce et al., 2026a). Extensive physiological evidence demonstrates that increases in supraspinal drive, spinal excitability, motor unit recruitment and discharge behavior, and neuromuscular junction stability contribute substantially to exercise responsiveness and can precede, or occur independently of, overt muscle hypertrophy (Egan and Sharples, 2023; Furrer et al., 2023).

In parallel, T2DM-related sarcopenia in humans is increasingly recognized as a mixed myogenic and neurogenic condition. Beyond intrinsic alterations in muscle protein metabolism, diabetes is associated with impairments in neural activation, motor unit remodeling, excitation–contraction coupling, neuromuscular junction integrity, and extracellular matrix organization, all of which contribute to reduced muscle quality and diminished adaptive capacity (Lecce et al., 2026b). These neuromuscular deficits are considered major determinants of muscle weakness and functional decline in diabetic populations and may constrain both the magnitude and nature of exercise-induced adaptations.

By contrast, rodent treadmill- and tail-load–based exercise paradigms primarily emphasize peripheral mechanical loading and metabolic engagement of skeletal muscle and do not capture the full spectrum of supraspinal, cortical, or fine motor unit–level remodeling characteristic of human resistance training. In the present study, no direct indices of neural adaptation—such as motor unit number estimation, nerve conduction properties, neuromuscular junction morphology, or electromyographic recruitment patterns—were assessed. Consequently, the observed associations between exercise training, muscle protein accretion, and anabolic signaling should be interpreted as muscle-level contributors within a broader neuromuscular adaptive process, rather than as sufficient explanations for exercise-induced mitigation of diabetic sarcopenia.

Within this context, the present findings identify a reproducible pattern of exercise-responsive muscle adaptations—namely increased protein accretion and coordinated activation of anabolic signaling pathways—that may translate across species. At the same time, they underscore that recovery of neural, motor unit, and functional adaptations central to resistance training efficacy in humans cannot be inferred from muscle-intrinsic molecular outcomes alone. These considerations emphasize the importance of integrating molecular, neural, and functional endpoints in future translational studies and caution against overextension of muscle-level findings when extrapolating to human neuromuscular performance.

The present findings have direct implications for exercise-based strategies aimed at mitigating diabetic sarcopenia. Notably, the results demonstrate that exercise can promote muscle protein accretion and anabolic signaling without inducing overt hypertrophy or reversing fibrosis. This pattern of adaptation may be particularly relevant for older adults with T2DM, individuals with early or moderate sarcopenia, or patients with limited tolerance for high-load resistance training, in whom hypertrophic responses are often blunted (Hansen et al., 2025).

Resistance-containing exercise modalities preferentially engaged integrin-associated and PI3K-related signaling, whereas exercise-general effects predominated for AKT/mTOR activation, suggesting that multimodal exercise programs may provide complementary benefits at the molecular level. Rather than prioritizing hypertrophy as the sole therapeutic goal, exercise prescriptions for diabetic populations may benefit from emphasizing protein accretion, signaling responsiveness, and metabolic support, particularly during early or intermediate stages of intervention.

With respect to exercise dose and feasibility, the rodent protocol used here represents a high-frequency, moderate-intensity stimulus relative to established rodent endurance and resistance paradigms, designed to ensure consistent mechanical and metabolic engagement. Although direct quantitative translation to humans is inappropriate, this paradigm is physiologically more analogous to regular moderate-intensity aerobic activity combined with low-to-moderate load resistance exercise performed on most days of the week. Such patterns align with current clinical recommendations for individuals with T2DM (Kanaley et al., 2022; American Diabetes Association Professional Practice Committee, 2024) and are feasible for many, though not all, patient populations. Importantly, diabetes-related neuromuscular impairments may limit tolerance to high-volume or high-load exercise in frail individuals, underscoring the need for progressive, individualized, and multimodal exercise prescriptions (Lecce et al., 2026b).