Dynamics of macrophage subset changes during skeletal muscle regeneration
To reveal the changes in macrophage heterogeneity and identify the specific subsets influencing skeletal muscle regeneration following injury, we intramuscularly injected CTX into the tibialis anterior (TA) of mice. The TA tissues were collected at the designated time points for multiple analyses (Fig. 1a). Consistent with previous reports25, these infiltrated immune cells were mainly CD68+ monocytes/macrophages (Fig. 1b). Cytofluorometric analysis confirmed the robust infiltration of circulating monocytes and the generation of macrophage subsets in the regenerating muscle. These cells were initially proinflammatory Ly6Chi cells but transformed into anti-inflammatory Ly6Clo cells by day 4 (Fig. 1c-e). For unbiased analysis that integrates temporal gene profiling of macrophages with analysis of their potentially varying roles in muscle regeneration, we enriched CD90.2−B220−CD45+ cell populations by using magnetic beads at the indicated time points during skeletal muscle regeneration. Subsequently, single-cell RNA-seq (scRNA-seq) analysis was performed using the 10X Genomics Chromium platform (Fig. S1a). After standard quality control and the removal of doublets, high-quality transcriptomes from 21,642 cells were revealed. We performed graph-based Leiden clustering and utilized uniform manifold approximation and projection (UMAP) embeddings for visualization. All clusters were annotated by utilizing the scMCA_MNN-muscle dataset and differentially expressed genes (DEGs). Fifteen cell clusters were revealed, including five subsets of monocytes/macrophages (Mo/Mφ), five subsets of neutrophils, and one subset each of dendritic cells, T cells, B cells, muscle progenitor cells, and stromal cells (Fig. S1b and c). As expected, macrophages and neutrophils were the most abundant cell populations after, with proportions of 67.9% and 15.3%, respectively (Fig. S1c).
Identification of five distinct macrophage subsets during skeletal muscle regeneration
We reclustered our scRNA-seq data of the macrophage subsets and identified five groups of macrophage subsets (Fig. 1f) whose distributions changed dynamically at different time points (Fig. 1g). Pseudotime analysis was performed on all five cell clusters along the injury-to-regeneration trajectory via Monocle (v3) to delineate the expression patterns of genes following muscle injury (Figs. 1h and S2a). These dynamic changes suggest that the unique macrophage subset distributions may have critical biological functions at specific time points. The five distinct macrophage clusters (Mo/Mφ clusters: 1, 2, 3, 4, and 5) shared a common core of expressed macrophage markers, including Adgre1, Cd68, Csf1r, Fcgr1, Lgals3, and Lyz2 (Fig. S1d). However, they exhibited distinct transcriptional profiles: Cluster 1 exhibited increased expression of the M2 macrophage activation markers Arg1 and Mrc1; Cluster 2 exhibited increased expression of genes associated with tissue regeneration, such as Igf1 and Gdf15, which is in line with the recently described finding that GDF15 is a critical effector during skeletal regeneration20; and Cluster 3 exhibited increased expression of genes associated with immune response activation and antigen presentation, including H2-Aa, H2-Ab1, and H2-Eb1. Moreover, the proportion of cells in Cluster 3 was notably greater than that of cells in other clusters in both the uninjured and regenerative stages; Cluster 4 exhibited high expression of genes involved in proinflammatory responses, such as Ifitm6, Gsr, and Hp; Cluster 5 exhibited increased expression of Acod1 (aconitate decarboxylase 1, also known as immunoresponsive gene 1), a key regulator of immunometabolism during infection and inflammation (Fig. S1e). During the acute inflammatory stage (days 1 and 2), most cells were cells from Clusters 1, 4, and 5, while there were only minor increases in the numbers of cells in Clusters 2 and 3. With the transition from the acute inflammatory stage to the regenerative stage (day 3), the proportions of cells in Clusters 1, 4, and 5 decreased, and there was a marked increase in the proportion of cells in Cluster 2, which then declined by day 7. In contrast, the proportion of cells in Cluster 3 gradually and consistently increased over 7 days, peaking at the regenerative stage. Thus, our findings identified five macrophage subsets with dynamic changes throughout the regenerative process. Although the Clusters 2 and 3 subsets were present in high proportions during the inflammatory-to-regenerative transition stage on day 3, Cluster 2 cells expressed relatively higher levels of genes involved in macrophage-mediated tissue regeneration than did Cluster 3 cells (Fig. S1e). For example, IGF1 is one of the best-characterized growth factors and has been shown to regulate muscle regeneration26. IGF1 binds its receptor IGF1R to phosphorylate the intracellular adapter protein insulin receptor substrate-1 (IRS-1), which in turn activates the PI3K/AKT pathway to facilitate skeletal muscle regeneration. Moreover, ablation of triggering receptor expressed on myeloid cells-2 (TREM2), a major macrophage sensor known for supporting immune cell responses, has been noted to impede hepatic reparative processes in response to metabolic disruptions27,28. In view of these findings, it is conceivable that the macrophage Cluster 2 we identified could play a role in facilitating muscle regeneration.
A tissue-regenerative macrophage subset exhibiting a resident macrophage gene signature
Resident and recruited macrophages play distinct roles in immune defense, with resident macrophages providing a constant level of immune surveillance, while recruited macrophages respond to acute infections or injuries. To visualize differential gene expression patterns, we summarized the origin of each monocyte/macrophage subset and examined the known marker genes associated with resident and recruited macrophages29. Resident macrophage signature genes, such as Axl, Cd74, and Cxcl16, were highly expressed in Clusters 2 and 3. Clusters 4 and 5 were characterized by the expression of the recruited macrophage markers Cxcr2, Ifitm1, and Sell. (Fig. S2a and b). However, it is noteworthy that Cluster 2 was conspicuously absent on day 0 (Fig. 1g), suggesting that Cluster 2 cells may not originate from resident macrophages. Our subsequent cell trajectory analysis demonstrated that Clusters 4 and 5 were the major infiltrating macrophage populations and likely contributed to Cluster 2 subset formation (Fig. S3a). Our findings revealed that clusters 1, 4, and 5 consisted of inflammatory macrophages. In contrast, clusters 2 and 3 were tissue-resident macrophages in the inflammatory-to-regenerative transition stage during skeletal muscle regeneration.
Identification of GPNMB-expressing macrophages as critical effectors in skeletal muscle regeneration
A key analysis in the investigation of the molecular mechanisms underlying changes in the state of macrophages is the identification of differentially expressed genes along the pseudotime trajectory, i.e., that determined by trajectory inference30 from single-cell RNA-sequencing data. This inferred trajectory highlights the key effectors within macrophage subsets that govern the biological processes of regeneration. By integrating the relative trajectory positions of the macrophage clusters with the distribution density of each identified group (Fig. 1h), Cluster 4 cells were identified as mainly present at the beginning of the trajectory; Cluster 3 cells were predominant at both ends of the trajectory, cells in Clusters 1 and 5 were identified at the early and middle positions, and Cluster 2 cells were identified at the end of the pseudotime axis. Notably, the positions of Cluster 3 cells at both ends of the pseudotime axis, together with their high proportions in uninjured muscle (Fig. 1g), suggest that Cluster 3 may represent a subset with steady-state characteristics. To further investigate the biological and functional roles of Cluster 2 macrophages during skeletal muscle regeneration, volcano plots were generated to visualize the DEGs in Cluster 2 cells vs. cells in Clusters 1, 3, 4, and 5. The top 10 up- and downregulated DEGs are labeled in the plots; among these DEGs, the Gpnmb gene was significantly upregulated, specifically in Cluster 2 (Fig. S3b). To further investigate the importance of macrophages with high GPNMB expression in tissue regeneration, we analyzed significantly activated macrophage marker genes31 using a pseudotime approach, focusing on critical factors involved in tissue regeneration and fibrosis. We found that Mertk, Igf1, and Nr1h3 exhibited the same expression pattern as Gpnmb (Fig. 2a); previous studies have highlighted the significance of these genes in tissue regeneration26,32–34. Nuclear receptor subfamily 1 group H member 3 (Nr1h3) is a gene that encodes a transcription factor involved in lipid metabolism and cholesterol homeostasis.
In the late stage of tissue regeneration, macrophages adopt an anti-inflammatory phenotype that helps to suppress inflammatory responses and restore normal tissue structure and function. However, a dysregulated response can result in persistent inflammation and maladaptive regeneration, ultimately leading to tissue-destructive fibrosis. Previous studies have indicated that, in a chronic inflammatory environment, the GPNMB secreted by macrophages can stimulate excessive deposition of ECM, ultimately leading to pulmonary fibrosis35. After skeletal muscle injury, macrophages play a key role in clearing apoptotic cells and aiding tissue regeneration, a process that involves the conversion of infiltrating monocytes to macrophages with inflammatory and regenerative phenotypes. Our results revealed time-dependent gene expression changes in Gpnmb and Ly6c, revealing that the GpnmbhiLy6clo macrophage population was predominantly enriched in Cluster 2 on day 3 (Fig. 2b). These findings highlight the importance of identifying specific GPNMB-expressing macrophage subsets during skeletal muscle regeneration. Based on the temporal dynamics of GPNMB expression and previous descriptions of tissue-resident macrophages36,37, we named this cell subset GPNMBhiLy6Clo “regenerative macrophages”. To validate the findings from single-cell RNA-seq analysis, we employed flow cytometry to assess the abundance of the GPNMBhiLy6Clo subset in the CD45+CD11b+ cell population in murine muscle postinjury and observed a peak on day 3, followed by a decrease to approximately 10.4%. Furthermore, our histological staining of tissue sections from various time points revealed cells with prominent GPNMB staining on day 3 (Fig. 2c). These findings suggest that macrophages with high GPNMB expression exhibit characteristics reminiscent of M2 macrophages. To validate the findings from single-cell RNA-seq analysis, we employed flow cytometry to identify the GPNMBhiLy6Clo subset in the CD45+CD11b+ cell population in murine muscle postinjury and observed a peak in the proportion on day 3, followed by a decrease to approximately 10.4%. Furthermore, our histological staining of tissue sections at various time points revealed prominent GPNMB-positive cells on day 3 (Fig. 2d). These findings suggest that macrophages with high GPNMB expression exhibit characteristics reminiscent of M2 macrophages.
CellChat identifies communication patterns and predicts the functions of macrophage subsets involved in skeletal muscle regeneration
Since a direct comparison of DEGs might not comprehensively capture the intricate signaling network, we conducted a thorough investigation of macrophage cellular communication dynamics. We performed CellChat24 analysis at distinct time points and revealed regulatory cell‒cell interactions. On day 3, we observed increased levels of tissue regeneration-related signaling factors, including IGF12,38, GAS39,40, GDF14, and nicotinamide phosphoribosyltransferase (NAMPT)7. On day 3, macrophages tended to send signals of IGF and GDF and receive signals of IGF, GAS, and GDF (Fig. S4a). Inflammatory signals, such as IL-1, IL-2, and TNF in macrophages, were abundant on days 1 and 2 (Fig. S4b). In addition to the signaling pathway network as a time-resolved signature, we also predicted the putative interactions among ligand and receptor pairs (Fig. 2e and f). On day 3, three critical skeletal muscle regeneration signaling pathways, GAS6/AXL, GAS6/MERTK, and IGF1/IGF1R, were highly enriched. IGF1/IGF1R has been demonstrated to promote skeletal muscle regeneration12,38. The TYRO3, AXL, and MERTK (TAM) receptor tyrosine kinases and their cognate glycoprotein ligands growth arrest-specific 6 (GAS6) and protein S (PROS1) are critical regulators of tissue homeostasis and inflammation41. Our results are consistent with the concept that TAM receptors are activated in macrophages in response to tissue injury42. The heightened activity of these three critical pathways confirms their roles in muscle regeneration and highlights their coordinated contribution to regenerative mechanisms.
GPNMB promotes M2 macrophage polarization via the upregulation of specific transcription factors
To validate this hypothesis, we isolated murine bone marrow-derived cells (BMDCs) and induced their differentiation into M1 and M2 macrophages in vitro (Fig. 3a). mRNA and protein expression analyses revealed a significant increase in GPNMB expression in M2 macrophages (Fig. 3b and c). These results indicate that GPNMB is a marker of M2 macrophages involved in muscle tissue regeneration and led us to hypothesize that the overexpression of GPNMB promotes M2 macrophage polarization. We conducted GPNMB overexpression experiments to verify this hypothesis in murine bone marrow-derived macrophages (mBMDMs). Our findings demonstrated that ectopic GPNMB expression increased the expression of M2 macrophage markers without affecting the expression of M1 macrophage markers (Fig. 3d). The overexpression of GPNMB was found to activate key transcription factors linked to M2 macrophage polarization, such as Irf4 and Pparg. In contrast, the expression of the critical regulators Irf5, Nfkb1, and Stat1 in M1 macrophages was unaffected by ectopic GPNMB expression (Fig. 3e), indicating that GPNMB may serve as part of the M2 gene expression program in macrophages.
GPNMB knockout impairs skeletal muscle regeneration
To investigate the role of GPNMB in muscle regeneration, we generated GPNMB-knockout mice and subjected them to CTX-induced muscle injury. The temporal evolution of muscle regeneration was monitored and compared with that of wild-type (WT) C57BL/6 controls (Fig. 4a). Western blot analysis of muscle tissue extracts confirmed the absence of the GPNMB protein in the KO group, confirming that the observed phenotypic differences were attributable to GPNMB deficiency (Fig. 4b). The gross morphological analysis of the TA muscles from the GPNMB-KO mice revealed marked impairment in muscle regeneration at days 4 and 7 postinjury (Fig. 4c). Compared with those from the WT mice, the injured muscles from the GPNMB-KO mice were visibly less striated. Histological evaluations conducted through hematoxylin and eosin (H&E) staining provided further insights into the compromised regenerative response in GPNMB-KO mice. Without GPNMB, the injured muscles exhibited exacerbated inflammatory responses and a notable delay in myofiber regeneration (Fig. 4d). The injured sites in the GPNMB-KO mice were characterized by increased cellular infiltration and a lack of newly formed myofibers. This phenotype starkly contrasted with the organized regenerative patterns observed in WT controls. Quantitative analyses were performed to measure the cross-sectional area (CSA) of muscle fibers, a key indicator of regenerative progress. On day 4 postinjury, the mean CSA of regenerating myofibers in GPNMB-KO mice was significantly reduced, indicating a failure to initiate the regenerative process properly. This defect persisted through day 7, with KO mice displaying markedly smaller myofibers than WT mice (Fig. 4e). The statistical significance of these differences was confirmed, underscoring the necessity of GPNMB for efficient muscle regeneration. Our findings unveil a previously unappreciated role for GPNMB in facilitating muscle tissue repair, suggesting that GPNMB may act as a modulator of the cellular and molecular events that orchestrate the complex process of muscle regeneration following acute injury.
Impaired macrophage efferocytosis and muscle regeneration following GPNMB knockout and MERTK inhibition
GPNMB overexpression in primary murine macrophages led to the upregulation of the expression of efferocytosis-related genes, including Mertk and Axl, suggesting that GPNMB plays a regulatory role in the genetic orchestration of the efferocytosis process (Fig. 5a). We further investigated the functional role of GPNMB in macrophage-mediated efferocytosis. GPNMB-deficient macrophages exhibit a marked reduction in the phagocytosis of apoptotic cells. Fluorescence microscopy and flow cytometry analyses revealed that compared with their wild-type counterparts, mBMDMs from GPNMB-KO mice exhibited significantly decreased uptake of CFSE-labeled apoptotic C2C12 myoblasts, underscoring the importance of GPNMB in the clearance of apoptotic cells (Fig. 5b). To understand the broader implications of GPNMB deficiency for muscle regeneration, we further explored the impact of disruption of efferocytosis on muscle regeneration by employing a pharmacological approach to inhibit MERTK. The administration of a MERTK inhibitor resulted in significant deficits in muscle tissue architecture and repair, as evidenced by histopathological evaluations (Fig. 5c-f). Notably, administering the inhibitor led to a dose-dependent exacerbation of muscle repair impairment, particularly at higher dosages (Fig. 5c). Histological analysis of muscle regeneration revealed that MERTK inhibition, particularly at a higher dosage, severely compromised the muscle repair process on days 4 and 7 postinjury. These morphological defects were confirmed by measuring muscle fiber CSA and length (Fig. 5d–f). These results underscore the importance of the GPNMB-MERTK axis in muscle regeneration and highlight a potential therapeutic target for promoting tissue repair following injury.
GPNMB stimulation promotes muscle regeneration and the myogenic differentiation of murine myoblasts
Previous research has indicated that the GPNMB expressed on macrophages undergoes enzymatic processing by disintegrin and metalloproteinase domain-containing protein 10 (ADAM10)43 to generate soluble GPNMB. Soluble GPNMB has critical functions; for example, it interacts with CD44 to promote cancer cell stemness and metastasis44. Via a similar pathway, soluble GPNMB promotes mesenchymal stromal cell survival, proliferation, and migration45. Moreover, the binding of soluble GPNMB to syndecan-4 impedes the extravasation of activated T cells into inflamed skin46. To investigate the potential of GPNMB to promote myogenic differentiation of myoblasts, we performed myotube differentiation assays using C2C12 cells with or without the addition of recombinant GPNMB (rGPNMB). C2C12 cells treated with or without rGPNMB treatment were collected at several time points during the differentiation process, including 0 (before differentiation), 12, 24, and 72 hours. The expression of Myod and Myog gradually increased during myoblast differentiation, and the addition of rGPNMB significantly increased their expression. However, rGPNMB only temporarily elevated the expression of the early differentiation marker Myf2a at 12 hours and did not affect the expression of the myoblast proliferation marker Myf5 at any time point (Fig. 6a). Immunofluorescence staining revealed that treatment with rGPNMB resulted in increased MyHC expression (Fig. 6b) and the formation of larger myotubes (Fig. 6c) containing a greater number of nuclei (Fig. 6d) in C2C12 cells than in the rGPNMB withdrawal group. Furthermore, our in vitro data indicate that rGPNMB facilitates myoblast differentiation. In vivo, we examined the role of exogenous GPNMB in facilitating skeletal muscle regeneration by administering 10 or 20 µg of rGPNMB to injured skeletal muscle. The results showed that the delivery of rGPNMB restored muscle architecture at the injury site (Fig. 6e). Specifically, on day 4, compared to injury alone, a single dose of rGPNMB led to a significant increase in the CSA of the TA muscle (225 ± 110.2 vs. 128 ± 43.2 µm2, p = 0.019, in the 10 µg group; 408 ± 139.5 vs. 128 ± 43.2 µm2, p < 0.0001, in the 20 µg group) Fig. 6f. Furthermore, on day 7, the effect of rGPNMB was sustained, as shown by the marked increase in the CSA of the TA muscle (728 ± 246.6 vs. 471 ± 195.3 µm2, p = 0.0013 in the 10 µg group; 1051 ± 340.9 vs. 471 ± 195.3 µm2, p < 0.0001 in the 20 µg group) Fig. 6g.