Hypoxia-induced GRP78 activation disrupts the Fndc5/Irisin axis to accelerate skeletal muscle atrophy

Abstract

Hypoxia is a potent inducer of skeletal muscle atrophy; however, the underlying molecular mechanisms remain incompletely defined. Irisin, a myokine derived from Fndc5, plays a critical role in maintaining muscle mass and function, while endoplasmic reticulum (ER) stress has been implicated in muscle degeneration. Here, we investigated the interplay between hypoxia-induced ER stress and irisin regulation in skeletal muscle. Transcriptomic analyses and weighted gene co-expression network analysis (WGCNA) identified Fndc5 and Hspa5 (encoding GRP78) as key genes within hypoxia-related modules, displaying a strong negative correlation. In vivo, mice exposed to hypoxia showed reduced Fndc5/irisin expression accompanied by significant GRP78 upregulation. In vitro, chemical hypoxia and pharmacological induction of GRP78 by HA15 consistently suppressed Fndc5/irisin levels and impaired C2C12 myotube formation. Gene-miRNA network analysis suggested a shared post-transcriptional link between HSPA5-centered ER stress and FNDC5-associated atrophy programs under hypoxia, with miR-34a-5p as a candidate regulator. Collectively, these findings demonstrate that GRP78-driven ER stress under hypoxic conditions disrupts irisin production, thereby accelerating skeletal muscle atrophy. This work highlights a mechanistic axis linking ER stress to irisin deficiency in hypoxia-induced muscle wasting and provides new insights into potential therapeutic targets.

Background

Skeletal muscle constitutes approximately 40% of total body mass and nearly 50% of total protein content, making it one of the body’s largest and most metabolically active organs.¹ Skeletal muscle atrophy is a debilitating consequence of various physiological and pathological stressors,² manifesting as a loss of muscle mass and strength.3, 4, 5 Among these factors, hypoxic exposure—arising from high altitude or disease-related hypoxemia—is a well-recognized inducer of muscle wasting.⁶

The mechanisms underlying hypoxia-induced muscle atrophy are multifactorial, with endoplasmic reticulum (ER) stress emerging as a pivotal contributor.7, 8 In Citellus dauricus, hypoxia triggers severe muscle atrophy accompanied by upregulation of ER-stress markers, including GRP78, p-eIF2α, p-PERK, and ATF6, indicating activation of the unfolded protein response (UPR).9, 10, 11 Similarly, human placental tissue at ∼3100 m altitude exhibits elevated eIF2α phosphorylation, while rats exposed to simulated 8% O₂ display progressive muscle loss with increased GRP78 and PDI expression.¹² Severe hypoxia also induces CHOP/GADD153 and activates both proteasomal and apoptotic pathways.¹⁰ Notably, ER stress can promote hypoxia-induced muscle atrophy primarily through suppression of the Fndc5/Irisin axis without overt apoptosis, reflecting a tissue-specific response.13, 14

Irisin, a myokine synthesized and processed in the ER, functions as a biomarker of muscle strength and metabolic health.11, 15 Initially identified as an exercise-induced hormone,16, 17 Irisin facilitates myoblast proliferation, differentiation, and recovery from muscle disuse or denervation atrophy.18, 19, 20 Conversely, hypoxia significantly reduces circulating Irisin levels in mountaineers.¹⁸ The precursor protein Fndc5 requires N-glycosylation at Asn36/81 (mouse) or Asn7/52 (human) for proper processing by the sheddase ADAM10,19, 20 and loss of glycosylation destabilizes Fndc5 and impairs Irisin release.²¹ Given that the ER governs protein folding and post-translational modification, sustained ER stress may thus interfere with Irisin biosynthesis.

Our previous work demonstrated that hypoxia markedly decreased Fndc5/Irisin expression in mouse plasma and gastrocnemius muscle, while restoration of Fndc5 reversed myotube atrophy under 1% O₂.²² These findings led us to hypothesize that hypoxia-induced ER stress suppresses Fndc5/Irisin maturation, thereby promoting skeletal muscle atrophy. In this study, we show that hypoxia activates GRP78-dependent ER stress and downregulates Fndc5/Irisin, while pharmacologic GRP78 induction further exacerbates Irisin loss and impairs myogenic differentiation, identifying GRP78 as a key mediator linking ER stress to hypoxia-induced muscle degeneration.

Results

Gene co-expression network analysis and module characteristics of weighted gene co-expression network analysis related to irisin and endoplasmic reticulum stress

To identify coordinated gene expression programs associated with hypoxia, we applied weighted gene co-expression network analysis (WGCNA), a systems-level method that groups genes with similar expression patterns into functional modules. To determine the optimal soft-thresholding power, we evaluated the scale-free topology fit index (R²) and mean connectivity across a series of candidate β values ranging from 1 to 20. Based on the criterion that a higher R² indicates a closer approximation to scale-free topology while maintaining adequate connectivity, β = 3 was selected as the optimal soft threshold, as it achieved an R² greater than 0.9 with a satisfactory mean connectivity. This parameter setting was subsequently used for network construction. A soft-thresholding power of 3 was selected to approximate scale-free topology while preserving network connectivity (Figure 1a). Through this approach, we obtained several gene modules with similar co‑expression structures, which may reflect coordinated biological functions.

Fig. 1.

Gene co-expression network analysis and module characteristics of WGCNA related to irisin and endoplasmic reticulum stress. (a) Soft-threshold power analysis for network construction. The mean connectivity (left) and scale-free topology fit index (right) were evaluated across candidate powers; a soft-thresholding power of 3 was selected to approximate scale-free topology while maintaining sufficient connectivity. (b) Hierarchical clustering dendrogram of genes based on topological overlap, with module assignment indicated by color bars; each color represents a distinct co-expression module. (c) Module-trait relationship heatmap showing correlations between module eigengenes and the hypoxia phenotype (hypoxia vs. normoxia). Values indicate Pearson’s correlation coefficients (r) with corresponding P values; the turquoise module shows the strongest association with hypoxia (r = 0.98, P = 1e−05). (d) Eigengene adjacency/relationship heatmap illustrating inter-module correlations, where modules with similar eigengene expression patterns cluster together, suggesting shared regulatory programs. (e) Gene co-expression network heatmap (topological overlap matrix, TOM) for module structure visualization. Warmer colors indicate stronger gene–gene connection strength; the turquoise module displays dense intra-module connectivity, consistent with a tightly co-regulated program. (f) Scatter plot of module membership (MM; correlation between each gene and the turquoise module eigengene) versus gene significance (GS; correlation between each gene and the hypoxia trait). Candidate hub genes were screened using |MM= > 0.55 and |GS= > 0.6, yielding 6503 genes in the turquoise module. (g) Venn diagram showing the overlap between turquoise-module genes from WGCNA and differentially expressed genes (DEGs) identified between hypoxia and normoxia. Differential expression analysis detected 21,436 DEGs and 6210 genes overlapped with the WGCNA set, representing 28.6% of total DEGs.

Genes were grouped into distinct co-expression modules based on expression similarity (Figure 1b). The relationship between module eigengenes and hypoxia-related clinical parameters was assessed using correlation analysis. By applying a significance threshold based on P-values, we retained 29 modules that showed statistically significant correlations with at least one hypoxic trait. Each of these modules was subsequently annotated, with the annotation results provided in Supplementary file 1 and the corresponding raw data in Supplementary file 2. Correlation analysis between modules and the hypoxic phenotype identified the turquoise module as the most significantly correlated with hypoxia (r = 0.98, p = 1e^–05), suggesting that genes in this module are likely involved in the transcriptional response to low oxygen tension (Figure 1c). Hence, the turquoise module was designated as the key module for downstream functional exploration.

Inter-module relationships were visualized using an eigengene adjacency heatmap (Figure 1d), which showed that modules with similar expression profiles clustered together, implying shared regulatory pathways. The topological overlap matrix (TOM)-based heatmap demonstrated that genes within the turquoise module exhibited high co-expression connectivity, consistent with its identification as a central module implicated in the adaptive response to hypoxia. The gene co-expression network heatmap further confirmed that genes within the turquoise module were highly interconnected (Figure 1e), supporting its role as a functional hub in hypoxia adaptation.

To pinpoint key regulatory genes, we applied dual criteria of |MM= > 0.55 and |GS= > 0.6, identifying 6503 candidate genes within the turquoise module (Figure 1f, Original data seen in Supplementary file 3). Transcriptomic analyses further identified Fndc5 and HspA5 (heat shock protein family A member 5, also known as Hspa5/HSPA5, which encodes glucose-regulated protein 78, GRP78) as core genes in hypoxia-related modules. In parallel, differential expression analysis between the hypoxia and normoxia groups yielded 21,436 differentially expressed genes (DEGs) (Original data seen in Supplementary file 4). The intersection of these datasets revealed 6210 overlapping genes, accounting for 28.6% of total DEGs (Figure 1g). These overlapping genes likely represent the transcriptional core of the hypoxia response network in skeletal muscle, linking co-expression structure with differential regulation.

Analysis of core gene function enrichment and pathways

To further elucidate the biological processes linking Fndc5 downregulation and HspA5 activation under hypoxia, we performed gene set enrichment analysis (GSEA) and Gene Ontology (GO) functional annotation. As shown in Figure 2a, decreased Fndc5 expression was significantly associated with suppression of oxidative metabolism–related pathways, including the citrate cycle (TCA cycle), glycolysis/gluconeogenesis, and oxidative phosphorylation, genes within these pathways were preferentially expressed in samples with higher Fndc5 levels.

Fig. 2.

Distinct metabolic and proteostatic signatures associated with Fndc5 and HSPA5 expression under hypoxia. (a) Gene set enrichment analysis (GSEA) of KEGG pathways based on Fndc5 expression. Low Fndc5 expression was linked to the downregulation of oxidative metabolism-related pathways, including the TCA cycle, glycolysis/gluconeogenesis, and oxidative phosphorylation. (b) GSEA of KEGG pathways based on HSPA5 expression. High HSPA5 expression enriched the proteasome, ribosome biogenesis, and autophagy/mitophagy pathways, indicating ER stress-induced activation of protein quality control. (c) GO enrichment analysis of Fndc5-low-correlated genes. Significantly enriched terms included rRNA processing, proteasome complex, and ubiquitin ligase activity, consistent with translational repression and increased protein degradation. (d) GO enrichment analysis of HSPA5-high-correlated genes, highlighting extracellular matrix organization, sarcomere assembly, and cytoskeletal remodeling in muscle cells. The x-axis represents −log10(adjusted P value), and bubble size indicates the number of genes per term.

These changes reflect a metabolic reprogramming from oxidative to glycolytic energy utilization, suggesting that Fndc5/Irisin deficiency is accompanied by impaired mitochondrial function and energy production. Conversely, Figure 2b demonstrates that high HspA5 expression correlated positively with enrichment of proteasome, ribosome biogenesis, and autophagy/mitophagy pathways, indicating coordinated upregulation of proteostasis-related pathways.This indicates activation of the endoplasmic reticulum stress (ERS) response and proteostasis remodeling, consistent with the UPR observed in hypoxic muscle cells (original data seen in Supplementary file 5).

Complementary GO analyses further supported these findings. In the Fndc5-low group (Figure 2c), enriched biological processes involved rRNA processing, ribosome biogenesis, and ubiquitin-proteasome system activity, suggesting that loss of Fndc5 triggers translational repression and enhanced protein degradation. In contrast, the HspA5-high group (Figure 2d) was enriched for extracellular matrix organization, sarcomere and myofibril structure, and collagen-containing extracellular matrix, reflecting maladaptive structural remodeling of myofibers under persistent ER stress (Original data seen in Supplementary file 6).

Analysis results of the correlation of key gene expression and regulatory network

To investigate the transcriptional alterations under hypoxic stress, a comparative analysis of gene expression was performed. Figure 3a shows that the expression of the Fndc5 gene was markedly downregulated in response to hypoxia, whereas HspA5 expression fluctuated without reaching statistical significance (P > 0.05).

Fig. 3.

Analysis results of the correlation of key gene expression and regulatory network. (a) The box plots of gene expression show the significantly differential expression of Fndc5 and Hspa5 genes between the control group and the hypoxia group. (b) Correlation plots of Fndc5 with key atrophy-related genes. Linear regression analysis showed significant negative correlations between Fndc5 and Trim63, Fbxo32, and Foxo3 (r = –0.94, –0.89, and –0.79, respectively; all P < .001). Shaded areas represent 95% confidence intervals. (c) Principal component analysis (PCA) plot, distinguishing the clustering distribution of samples in the control group and the hypoxia group based on gene expression profiles. (d) Scatter plot of gene expression correlation, showing the linear association and distribution trend of the expression levels of Fndc5 and Hspa5 genes. (e) Protein-protein interaction (PPI) network of core genes. The network showed a relatively dense interaction structure, with HSPA5 and XBP1 located in central positions, suggesting that they may serve as key nodes in the underlying molecular interaction network. (f) Transcription factor regulatory network. The red nodes represent core transcription factors, and the connecting lines show regulatory associations. The transcription factor regulatory patterns of key genes are mined. (g) Prediction of miRNA regulatory network and analysis of the miRNA regulation of genes such as Hspa5, Foxo3, and Fbxo32.

Because Trim63, Fbxo32, and Foxo3 are well-recognized canonical mediators of skeletal muscle atrophy, they were pre-specified candidate genes for focused analysis, rather than outputs of an unbiased gene-discovery pipeline. We then examined their correlations with Fndc5 expression to assess whether Fndc5-associated changes were linked to established atrophy-related transcriptional programs.23, 24 To explore potential gene interactions, a correlation heatmap was constructed (Figure 3b), revealing strong negative correlations between Fndc5 and Trim63 (r = –0.94, P < 0.001), Fbxo32 (r = –0.89, P < 0.001), and Foxo3 (r = –0.79, P < 0.001).

These findings suggest that Fndc5 may act in concert with key muscle atrophy–related genes to coordinate the hypoxic response. Principal component analysis was conducted on the expression profiles of core genes (Fndc5, HspA5) and muscle atrophy-related marker genes (Trim63, Fbxo32, Foxo3) (Figure 3c). The normoxic and hypoxic samples exhibited clear group-specific clustering (PC1 = 38.2%, PC2 = 21.5%), which directly demonstrated that hypoxia induced significant overall transcriptomic reprogramming in skeletal muscle cells, leading to a fundamental difference in the gene expression pattern between the two groups. Moreover, a linear regression plot (Figure 3d, R² = 0.68, P < .001) confirmed a significant negative correlation between Fndc5 and HspA5 expression, implying a potential antagonistic regulatory relationship under hypoxic conditions.

To elucidate the molecular network underlying this interaction, a protein-protein interaction (PPI) network centered on HspA5 and XBP1 was constructed. This network contained proteins associated with ER stress and oxidative stress, including ERN1 and SOD2, highlighting the linkage between hypoxia and cellular stress responses (Figure 3e). To further delineate the upstream regulatory architecture, a transcription factor (TF) regulatory network was built using the TRRUST database (Figure 3f). Incorporating Fndc5, HspA5, and the muscle atrophy markers (Fbxo32, Trim63, and Foxo3), a total of 55 TFs were identified (Original data seen in Supplementary file 7). Among them, 11 core transcription factors, including HspA5, ATF4, and XBP1, were highlighted through Cytoscape analysis. These core TFs exhibited extensive interconnectivity, suggesting a tightly coordinated regulatory module.

To further explore the potential upstream regulation of Fndc5 and skeletal muscle atrophy-related core genes, we constructed a gene-miRNA regulatory network including Fndc5, HspA5, FOXO3, FBXO32, and TRIM63 (Figure 3g). The network showed extensive interactions between these five genes and multiple miRNAs, forming a relatively dense regulatory structure (Original data seen in Supplementary file 8). Topological analysis in Cytoscape was performed by calculating Degree and Neighborhood Connectivity for each node, and key miRNAs were screened using a cutoff of Degree ≥ 3, yielding 62 candidate miRNAs with potential regulatory importance. Among the highest-degree miRNAs (Degree = 5), representative nodes included hsa-miR-23a-3p, hsa-miR-26a-5p, hsa-miR-26b-5p, hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p, hsa-miR-34a-5p, and hsa-miR-106a-5p, indicating a shared post-transcriptional regulatory pattern across multiple core genes. In addition, FOXO3 and HspA5 displayed the highest Degree values and were positioned centrally in the network, supporting their hub-like roles in the miRNA-mediated regulatory network, whereas TRIM63 and Fndc5 showed relatively higher Neighborhood Connectivity among the five gene nodes.

Expression of endoplasmic reticulum stress-related genes and GRP78 protein under normoxia and hypoxia

To investigate whether hypoxia induces ER stress in skeletal muscle, we examined the expression of canonical ER stress markers in the gastrocnemius of mice exposed to 11.4% O₂. Quantitative analysis revealed a significant increase in GRP78 mRNA and protein levels under hypoxic conditions compared with normoxia (Figure 4a-c), indicating robust activation of the UPR. Among the downstream UPR branches, PERK expression was markedly elevated, whereas ATF6 and IRE1α remained unchanged, suggesting a selective activation pattern of ER stress signaling (Figure 4a). In addition, pro-apoptotic effectors such as s-XBP1 and CHOP did not exhibit significant alterations, implying that hypoxia predominantly triggers adaptive rather than apoptotic ER stress at this stage.

Fig. 4.

Expression of ER stress-related genes and GRP78 protein under normoxia and hypoxia. (a) Relative expression levels of ER stress-related genes (ATF6, GRP78, IRE1α, PERK, s-XBP1/total XBP1, and CHOP) under normoxia and hypoxia. (b) Representative Western blot showing GRP78 and GAPDH protein expression under normoxia and hypoxia. (c) Quantification of GRP78 protein levels normalized to GAPDH from the Western blot. Data are presented as mean ± SD (n = 4), with statistical significance (P = 0.048) indicated. (d) Representative images of cells treated with different concentrations of CoCl₂ (0 µM, 50 µM, and 100 µM). Red arrows indicate morphological changes induced by CoCl₂ treatment. Scale bar represents 50 µm. (e) Western blot analysis of HIF-1α and GRP78 expression levels in cells treated with 0 µM, 50 µM, and 100 µM CoCl₂. β-Tubulin was used as a loading control. (f) Quantification of GRP78 and HIF-1α protein levels from the Western blot data. Data are presented as mean ± SD (n = 3), with significant differences indicated (P < 0.05).

To further validate the effect of chemical hypoxia, C2C12 myoblasts were treated with cobalt chloride (CoCl₂), a hypoxia-mimetic agent. Phase-contrast imaging showed reduced cell density and altered morphology following CoCl₂ exposure in a concentration-dependent manner (Figure 4d). Consistently, Western blot analysis confirmed that both HIF-1α and GRP78 expression increased proportionally with CoCl₂ concentration (Figure 4e, f), further supporting that hypoxic stress activates GRP78-mediated ER stress in skeletal muscle cells.

Effects of HA15 on C2C12 myoblast viability, GRP78 and FNDC5/irisin expression, and myotube formation

To examine the effect of GRP78 activation on C2C12 myoblasts, the ER stress agonist HA15, known to target GRP78 and induce its overexpression, was applied at graded concentrations. Cell viability assessed by the CCK-8 assay showed that HA15 treatment reduced C2C12 myoblast viability at all tested concentrations compared with the DMSO control (Figure 5a). Although a modest rebound in cell viability was observed at 50 μM, viability remained consistently lower than that of the DMSO group, indicating an overall inhibitory effect of HA15 on cell survival. This concentration-dependent trend was similar with previous findings reported by Wu et al.²⁵

Fig. 5.

Effects of HA15 on C2C12 myoblast viability, GRP78 and Fndc5/irisin expression, and myotube formation. (a) CCK-8 assay showing the viability of C2C12 myoblasts treated with different concentrations of HA15 (1, 5, 10, 50, and 100 µM). Data are presented as mean ± SD (n = 6), with significant differences indicated (P < 0.05). (b) Western blot analysis of GRP78 and Fndc5/irisin protein levels in C2C12 myoblasts after treatment with HA15 at the indicated concentrations. β-Tubulin is used as a loading control. (c) Quantification of GRP78 protein expression normalized to β-Tubulin from the Western blot. Data are presented as mean ± SD (n = 3), with significant differences indicated (P < 0.001). (d) Immunofluorescence staining showing the expression of MyHC (green) and nuclear staining with Hoechst (blue) in C2C12 myotubes treated with various concentrations of HA15 (1 µM, 5 µM, 10 µM, 50 µM, and 100 µM). (e) Quantification of the myotube fusion index, data presented as mean ± SD (n = 3). (f) Quantification of myotube diameter, data presented as mean ± SD (n = 3). (g) Quantification of myotube area, data presented as mean ± SD (n = 3). Significant differences are indicated by P values.

Western blot analysis further confirmed that exposure to 50 and 100 μM HA15 markedly increased GRP78 protein expression (Figure 5b, c). In contrast, these same concentrations significantly reduced Fndc5/Irisin levels (Figure 5d). These data indicate that excessive GRP78 induction by HA15 activates ER stress and concurrently suppresses Fndc5/Irisin expression, implicating a negative regulatory relationship between ER stress and Irisin synthesis.

To assess the functional consequences of HA15-induced ER stress on myogenic differentiation, C2C12 myotubes were stained with anti-MyHC antibody after 24 h of treatment (Figure 5e). Marked inhibition of myotube formation was observed at 50 and 100 μM, where MyHC-positive fibers were virtually absent. Quantitative analysis showed that the fusion index, myotube area, and myotube diameter (Figure 5f-h) all decreased to baseline at these concentrations, while lower doses (1-10 μM) produced no significant alterations. Collectively, these results demonstrate that high-dose HA15-induced GRP78 overexpression triggers ER stress, leading to suppression of Fndc5/Irisin and complete blockade of myogenic differentiation.

Discussion

Hypoxia represents a major physiological challenge to skeletal muscle, initiating a cascade of adaptive and maladaptive responses that ultimately impair contractile and metabolic function. In the present study, we identified that hypoxia markedly reduces Fndc5 expression, accompanied by upregulation of ER stress markers such as HspA5 (GRP78) and HSP90B1 (GRP94). Consistent with previous observations, activation of the UPR through GRP78 can suppress protein synthesis and disrupt myofibrillar structure.26, 27, 28 Our data extend this concept by demonstrating that GRP78 overexpression induced by HA15 further aggravates Fndc5 suppression and impairs myogenic differentiation, supporting a direct inhibitory link between ERS and Irisin biosynthesis.

Irisin is processed in the ER and secreted into circulation after proteolytic cleavage of Fndc5. Perturbations in ER homeostasis can therefore interfere with proper protein folding and glycosylation, leading to defective Irisin maturation. Previous reports showed that the GRP78-ATF4-XBP1 axis coordinates the transcriptional response to hypoxia and oxidative stress, and prolonged activation of this pathway accelerates muscle protein degradation.29, 30, 31 The negative correlation between Fndc5 and HspA5 observed here suggests that excessive GRP78 activation may shift the UPR from an adaptive to a pro-degenerative program. This mechanism provides a plausible explanation for the concurrent decline in Irisin and increase in ERs during hypoxia.

The gene-miRNA regulatory network constructed in this study revealed a relatively dense pattern of miRNA-mediated post-transcriptional interactions among Fndc5, HSPA5, FOXO3, FBXO32, and TRIM63, suggesting potential crosstalk between skeletal muscle atrophy-related programs and ER stress/proteostasis regulation under hypoxic conditions. Topological network analysis further identified 62 key miRNAs, among which several high-degree miRNAs (including miR-23a-3p, miR-26a/b-5p, the miR-29 family, miR-34a-5p, and miR-106a-5p) were connected to multiple core genes, indicating coordinated regulation through a shared miRNA network and supporting a complex “multi-gene-multi-miRNA” regulatory architecture. Notably, previous studies have shown that miR-34a can directly bind to the 3′UTR of Fndc5 and suppress its expression, thereby regulating the Fndc5/irisin axis,³² which is consistent with the high-degree feature of miR-34a-5p observed in our network.

In parallel, as a central mediator of ER stress, GRP78/HspA5 has been reported to be directly targeted by multiple miRNAs (such as miR-30d, miR-181a, and miR-199a-5p),³³ and miR-30c has also been shown to target HspA5 in renal tubular epithelial cells,³⁴ further supporting the biological plausibility of miRNA-mediated regulation of HspA5. Taken together, our findings support the notion that miRNAs may serve as a critical intermediate layer linking ER stress-proteostasis and atrophy-associated transcriptional programs. However, the specific miRNA-target relationships proposed in this study are still primarily based on database prediction and network topology analysis and therefore require further validation through luciferase reporter assays, miRNA mimic/inhibitor interventions, and protein-level experiments.

Such dual regulation may exacerbate proteostasis collapse and reinforce the progression of muscle atrophy. Supporting this, Irisin has been shown to alleviate ER stress and apoptosis in cardiomyocytes³⁵ and to protect against hypoxia/reoxygenation injury via AMPK-dependent signaling.³⁶ Therefore, reduced Irisin availability under hypoxic stress likely removes an important cytoprotective buffer against ERS. This hypothesis is in agreement with findings from chronic kidney disease, chronic obstructive pulmonary disease, and cardiovascular pathologies, where concomitant GRP78 upregulation and Irisin downregulation predict muscle wasting.37, 38, 39

From a broader perspective, our results highlight a previously underappreciated crosstalk between ER proteostasis and myokine signaling. The reciprocal regulation of GRP78 and Fndc5/Irisin under hypoxia suggests that the ER stress response not only determines the cellular fate of muscle fibers but also shapes systemic metabolic adaptation. Bioinformatic integration based on GEO datasets and WGCNA-limma analytical pipelines reinforced the robustness of these associations,40, 41 emphasizing that the GRP78-Irisin axis constitutes a conserved hypoxia-responsive module across multiple datasets.

Conclusion

In summary, our study identifies HspA5 (GRP78)-mediated ER stress as a pivotal mechanism leading to the suppression of Fndc5/Irisin and consequent skeletal muscle atrophy under hypoxia. Overactivation of the UPR disrupts protein folding and translation, while decreased Irisin levels weaken cellular resistance to stress, together amplifying muscle degeneration. Our gene-miRNA network analysis identifies a shared post-transcriptional regulatory layer linking ER stress/proteostasis (HspA5-centered) and atrophy-associated transcriptional programs (Fndc5/FOXO3/FBXO32/TRIM63) under hypoxic conditions, with miR-34a-5p emerging as a key candidate regulator for further validation. These findings reveal a mechanistic link between ER stress and myokine deficiency and suggest that therapeutic modulation of the GRP78-Irisin pathway could offer new avenues for preventing hypoxia-associated muscle wasting.

Materials and methods

Mice and hypoxic exposure

Sixty male C57BL/6J mice, aged 11-13 weeks, were obtained from the Air Force Military Medical University for this study. The mice were randomly divided into two groups: a control group and a hypoxia group. The control group, consisting of 30 mice, was housed in a normoxic environment at an altitude of 399 m. The remaining 30 mice were initially acclimated in a hypoxic chamber (China, FLYDWC50-ⅡC) set at 15.4% oxygen concentration, simulating an altitude of 3000 m, for a period of 3 days. Following acclimatization, the mice were then exposed to a lower oxygen concentration of 11.6% (equivalent to an altitude of 5300 m). Throughout the experiment, all mice had ad libitum access to food and water, with alternating 12-h light and dark cycles maintained for the duration of the study.

Cell culture and hypoxic treatment

C2C12 murine myoblasts were cultured in growth medium containing high-glucose Dulbecco’s Modified Eagle Medium (Gibco 12800-017), supplemented with 10% fetal bovine serum (FBS; BI C04001-500) and 1% penicillin-streptomycin (Beyotime C0222). Cells were incubated at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. Upon reaching 70%-80% confluence, the growth medium was replaced with differentiation medium (DM), comprising Dulbecco’s Modified Eagle Medium supplemented with 2% (v/v) horse serum (HS; Gibco, 26050088), as described by Valle-Tenney et al.⁴² To induce chemical hypoxia, CoCl₂ was added to the DM at concentrations of 50 and 100 μM.

Cell viability assay

The impact of HA15 on cell viability was assessed using the Cell Counting Kit-8 (CCK-8; TargetMol, Boston, MA, USA; C0005). Initially, 2.5 × 10⁴ C2C12 murine myoblasts were seeded into a 96-well plate. Upon reaching 70%-80% confluence, the cells were incubated in DM for 48 h prior to treatment with varying concentrations of HA15 (1, 5, 10, 50, and 100 μM) for 24 h. Subsequently, 10 μl of CCK-8 reagent was added to each well, and the luminescence signal was measured using a multifunction microplate reader (BioTek Synergy HT).

RNA extraction and real-time quantitative PCR

Total RNA was extracted from gastrocnemius muscle tissue using Trizol reagent (AG; AG21102), following the manufacturer's protocol. Subsequently, 1 μg of RNA was reverse transcribed into complementary DNA (cDNA) using the HiScript II QRT SuperMix (Vazyme; R223-01). Quantitative real-time polymerase chain reactions (qPCR) were conducted in a final reaction volume of 10 μl, comprising 4.6 μl of cDNA template, 0.2 μl of specific primers, and 5 μl of 2× ChamQ SYBR qPCR Master Mix (Vazyme; Q311-02). GAPDH was employed as the internal housekeeping gene for normalization. The relative mRNA expression levels were calculated using the comparative Ct method (2^-ΔΔCt). Primer sequences were synthesized by Sangon Biotech and are detailed in Table 1.

Table 1.

Primer sequences.

Primers	Forward primer sequences	Reverse primer sequences
GRP78	5′-TTGGAATGACCCTTCGGTGC-3′	5′-GTGCCAGCATCTTTGGTTGC-3′
GRP94	5′-AACACACTAGGTCGTGGAACA-3′	5′-TCATAAGTTCCCAATCCCACAC-3′
IRE1α	5′-TTGACCTTCATGTCTGGGGAA-3′	5′-AGCTTGCTCTTCCCTTTGAGT-3′
ATF6	5′-CAGCTGTAATAGCCCCTCCTC-3′	5′-GCAAAACAGTCTGGCCTTTGG-3′
CHOP	5′-AACAGAGGTCACACGCACAT-3′	5′-ACTTTCCGCTCGTTCTCCTG-3′
PERK	5′-GACGAATCGCTGCACTGGAT-3′	5′-GGAAGTTTTGTGGGTGCCCT-3′
s-xbp1	5′-GCTGAGTCCGCAGCAGGT-3′	5′-CTGGGTCCAAGTTGAACAGAAT-3′
t-xbp1	5′-TGAAAAACAGAGTAGCAGCTCAGA-3′	5′-CCCAAGCGCTGTCTTAACTC-3′
GAPDH	5′-CGGTGCTGAGTATGTCGTGG-3′	5′-ATGAGCCCTTCCACAATGCC-3′

Western blotting

Gastrocnemius muscle tissues or C2C12 myotubes were lysed by using RIPA lysis buffer with an enhancer (Beyotime; P0013B), supplemented with a protease inhibitor cocktail (Beyotime; P1005; 1:100). Following electrophoresis and transfer, the PVDF membrane was incubated with primary antibodies: anti-HIF-1α (Cell Signaling Technology; #36169; 1:1000), anti-Fndc5 (Abcam; ab174833; 1:2000), anti-β-tubulin (Abcam; ab6046; 1:2000), anti-GAPDH (Abcam; ab8245; 1:1000). The membrane was incubated overnight at 4°C. Subsequently, the HRP-conjugated secondary antibody (Immunoway; RS0001; 1:10000) was applied for 1 h at room temperature. Protein bands were visualized using a chemiluminescence detection system (Tanon, Shanghai; T5200) with the Clarity™ Western ECL Substrate (BIO-RAD; #170-5060).

Immunofluorescence staining

C2C12 myotubes were treated with HA15 for 24 h followed by rinsing with phosphate-buffered saline (PBS; 1×) and fixation with 4% (w/v) paraformaldehyde at room temperature for 10 min. The myotubes were subsequently washed three times with PBS containing 0.1% Tween 20 and permeabilized using Triton X-100 (Beyotime; P0096) for 10 min at room temperature. After additional rinsing, the cells were incubated with 1% bovine serum albumin (BSA) for 30 min. Primary antibody, mouse anti-myosin heavy chain (MF-20; DSHB), diluted 1:200 in PBST, was applied to the myotubes and incubated overnight at 4°C. The cells were washed three times with PBS for 5 mins each before incubation with Alexa Fluor™ 488 goat anti-mouse antibody (Invitrogen; R37120) diluted in 1% BSA in a light-protected environment at room temperature for 1 h.

Nuclei were stained with Hoechst 33258 (Beyotime; C1011) for 10 min. Image analysis was performed using ImageJ software to calculate the myotube fusion index (defined as the total number of nuclei in myotubes with more than two nuclei divided by the total number of nuclei in the same field of view), as well as myotube diameter (Feret’s diameter) and area.

Data source and construction of co-expression network

This study conducted multi-dimensional analysis based on the transcriptome dataset GSE9400 from the GEO database.⁴³ The limma software package (version 3.62.2) was used to identify DEGs, with the selection criteria being |logFC| > 1 and corrected Padj < 0.05. The gene co-expression network was constructed using WGCNA:⁴⁴ the soft threshold (power) was set to 3 (scale topological index R² ≈ 0.9, the average connectivity curve was smooth). Co-expression networks were constructed using WGCNA to identify hypoxia-associated modules. A total of 60 modules were obtained. Modules significantly related to the hypoxic trait (|module-trait correlation| > 0.3 and P < 1e-5) were selected, and the "turquoise" module was determined as the core module (correlation = 0.98, P = 1e-5). Based on the module membership degree (MM > 0.55) and gene significance (GS > 0.6), 6503 key genes were extracted from this module.

Gene expression verification and interaction analysis

Expression validation of candidate genes (Fndc5, Hspa5) was conducted. After calculating the expression Z-score, independent sample t tests were used to evaluate the differences between groups (Fndc5, P < 0.05; Hspa5, P = 0.125). The results were visualized using ggplot2 (version 3.5.2) as box plots and violin plots. Gene interaction relationships were explored through Pearson correlation analysis:¹ A correlation coefficient heatmap of Fndc5, Hspa5, and skeletal muscle atrophy marker genes (Fbxo32, Trim63, FoxO3) was drawn using ggcorrplot (version 0.1.4.1);² A scatter plot of Fndc5-Hspa5 expression was drawn using ggpubr (version 0.6.1) (R = −0.69, P = 0.058), with a linear regression line and 95% confidence interval added.

Functional enrichment and pathway annotation

Functional enrichment was performed based on the intersection of DEGs and key genes from WGCNA. GO/KEGG analysis: Using clusterProfiler (version 4.14.6) and org.Mm.eg.db (version 3.20.0), significant thresholds were set at p value < 0.05 and q value < 0.05. The top 15 GO entries (biological process, cellular component, molecular function) and the top 20 KEGG pathways were displayed. GSEA analysis： Grouped by the median expression of HspA5/Fndc5, differential analysis was performed using the limma package,⁴¹ sorted genes by logFC in descending order, and the pathway enrichment was executed using gseKEGG, and the top 5 enriched pathways were visualized using enrichplot. Genes were ranked according to their correlation with the indicated gene expression level (Fndc5 or HspA5), and the running enrichment score (ES) is shown as the cumulative enrichment statistic across the ranked gene list. The x-axis indicates the rank position in the ordered dataset. The short vertical ticks mark the positions of genes belonging to each KEGG gene set, and the bottom panel shows the ranking metric distribution. The peak (or trough) of each running ES curve represents the ES, indicating where the corresponding pathway gene set is most concentrated in the ranked list.

Construction and topological analysis of the transcription factor regulatory network

To identify potential upstream transcription factors regulating Fndc5, HspA5, and skeletal muscle atrophy-related genes (e.g., Trim63, Fbxo32, and Foxo3), we queried the TRRUST v2 database to retrieve transcription factor predictions and TF-target regulatory relationships (http://www.grnpedia.org/trrust/). Specifically, within TRRUST, the organism was set to Homo sapiens, and the target gene list (Fndc5, HspA5, Trim63, Fbxo32, Foxo3) was submitted for analysis. The predicted transcription factors and their corresponding TF-target regulatory pairs were exported. The retrieved interactions were then curated by removing duplicated TF-target pairs and standardizing gene symbols. The finalized TF-target interaction file was imported into Cytoscape (v3.10.0) for network visualization and topological analysis.

In Cytoscape, transcription factors and target genes were represented as network nodes, and TF-target regulatory relationships were represented as edges to construct the transcription factor regulatory network. Network topology metrics were computed using the built-in network analysis functions, with a focus on degree (the number of directly connected edges), which reflects the breadth of regulation for TF nodes or the extent of regulatory input for target nodes. Transcription factors with degree ≥ 4 were defined as core regulators and were retained for downstream presentation and biological interpretation.

Gene-miRNA regulatory network construction and topological analysis

To investigate the potential upstream post-transcriptional regulatory relationships between Fndc5 and skeletal muscle atrophy-related core genes, we constructed a gene-miRNA regulatory network for Fndc5, HspA5, FOXO3, FBXO32, and TRIM63. First, predicted miRNAs targeting these genes were retrieved from the Target Scan Human database (https://www.targetscan.org/vert_80/; species: Human) by querying each gene symbol using the default prediction settings and ranking criteria. Target Scan predicts miRNA targets based on conserved seed-matching sites (e.g., 8mer, 7mer, and 6mer) within the 3′ untranslated region (3′UTR) of target genes and provides prediction rankings based on targeting efficacy-related models. The resulting gene-miRNA interaction pairs were then organized and imported into Cytoscape (v3.10.0) for network visualization and topological analysis. Information on the Target Scan algorithm and Release 8.0 is available in the official documentation.

Within Cytoscape, topological parameters for each node were calculated, including degree and neighborhood connectivity. Degree represents the number of direct connections of a given node, whereas neighborhood connectivity represents the average connectivity of its neighboring nodes. To identify key miRNAs with potential regulatory importance, a threshold of Degree ≥ 3 was applied. In addition, node colors were mapped according to neighborhood connectivity, with deeper red indicating higher local connectivity. Based on these topological features, we further evaluated hub-like properties of the core genes and analyzed potential shared regulatory patterns within the gene-miRNA network.

Statistical analyses

All assays were conducted in at least three independent replicates, and the quantitative data are expressed as the meanstandard deviation (SD). Statistical analyses were conducted using GraphPad Prism 8.0 (GraphPad Software, United States). The Student’s t-test or one-way analysis of variance (ANOVA) was employed to evaluate statistical significance, with a P-value of less than 0.05 considered statistically significant.

Funding and support

This work was supported by the Major Project of the Wuhu Municipal Health and Wellness Commission (Grant No. WHWJ2023z002); the "Hengrui Research Fund" Collaborative Project on Industry-Academia Integration (Grant No. XQHR202412); the Talent Introduction Fund Project of the Yiji-shan Hospital, Wannan Medical College (Grant No. KY29020689), the Anhui Provincial Clinical Medical Research Transformation Special Program (Grant No. 202527c10020017).

CRediT authorship contribution statement

Shiqiang Liu: Writing – review & editing, Writing – original draft. Linyao Xu: Visualization, Resources. Xinru Song: Validation. Zhenhao Zhu: Software. Yuxin Zhang: Methodology. Yumeng Sun: Formal analysis. Sharon Nyoja: Conceptualization. Qin Wang: Formal analysis. Jialin Gao: Supervision. Lizhuo Wang: Supervision. Huiyun Xu: Supervision.

Declarations of interest

All authors declare that they have no financial or non-financial competing interests that could be perceived as influencing the content or evaluation of this manuscript. This declaration is made on behalf of all contributing authors.

Appendix A. Supplementary material

Supplementary material

.

Supplementary material

.

Supplementary material

.

Supplementary material

.

Supplementary material

.

Supplementary material

.

Supplementary material

.

Supplementary material

.

Data availability

Data will be made available on request.

• Gene Expression OmnibusHypoxic gene regulation on mice Quadriceps muscle