Data merge and quality controlInformation for single-cell transcriptomes, including mouse SkM and SkM macrophages, was exhibited in Supplementary Table 1. Samples in each dataset were filtered based on original studies16,17,18,19, then merged using the “merge” function in the Seurat package, and the batch effect was removed using the “Harmony” method. Finally, 34,067 cells were obtained and further classified into 16 clusters (Fig. 1A,B).Figure 1Integration with Harmony method. (A) Skeletal muscle samples from 6 young, 7 old, and 4 geriatric mice were integrated with the Harmony method. (B) 16 clusters were recognized by Seurat 5.0.1. (C) Bubble plot showing the canonical markers of each cell type. (D) UMAP plot depicting the cell type in skeletal muscle. (E) Relative abundance of distinct cell types at different age in mice SkM.Single-cell profiling of SkM from young and old miceCanonical cell type markers (Supplementary Table 2) were implemented to identify different SkM cell types (Figs. 1C, S1, and S2)34. The 16 cell clusters were then divided into 11 cell types, which were labeled as endothelial cells (ECs), fibroadipogenic progenitors (FAPs), myonuclei, pericytes, smooth muscle cells, macrophages, neurons, tenocytes, satellite cells, neutrophils, B cells. A small group of cells was hard to identify and labeled as unknown (Fig. 1D). Furthermore, we calculated the proportion of these cell subpopulations between Y and O mice. ECs and FAPs were the most prevalent cell types, and satellite cell abundance was lower in O SkM, as previously reported16 (Fig. 1E).Lyve1-/MHCII+ is the main type of macrophage in the mouse SkMAs macrophages played a vital role in SkM regeneration and aging9,13, we took the macrophage cells group to sub-cluster analysis. Macrophages were divided into six clusters when the resolution was set as 0.3. We identified four cell groups from the six clusters (Fig. 2A). They were: (0) 294 macrophages highly expressed C1qa, C1qb, C1qc, H2-Aa and H2-Eb1; (1) 230 macrophages highly expressed Lyve1, Folr2 and Mrc1; (2) 213 macrophages highly expressed Gngt2, Hp and Gsr; (3) 200 macrophages highly expressed H2-Aa, H2-Eb1, H2-DMb1, Lyz1 and Fn1 (Fig. 2B–D). They were further classified into 4 cell groups based on the surface markers Lyve1 and MHC-II and 4 groups based on marker genes (Fig. 2E–F). We found that Lyve1−/MHCII+ was the largest group of macrophages, accounting for approximately 40% of the total macrophages (Fig. 2G). Another large single-cell database of SkM macrophages was used for further verification and investigation. After screening and quality control, 9,469 cells were classified into 11 clusters based on previous studies13. Marker genes for each cell cluster were exhibited in Supplementary Table 3. Finally, 11 types of SkM macrophages were identified; they were: (0) 2897 cells of highly expressed Lyve1, Folr2, and Mrc1, which is similar to cluster 1 of the former dataset; (1) 1565 cells of highly expressed Lyve1 and Cd209a; (2) 1494 cells highly expressed Lyz1 and Fn1, which corresponding to cluster 3 of the former dataset; (3) 1320 cells highly expressed Thbs1, Hp, and Gsr, which was homologous to cluster 2 of our dataset; (4) 498 cells highly expressed Gngt2; (5) 430 cells highly expressed Cxcl2 and Tnf. (6) 416 cells highly expressed Spp1, Fabp5 and Gpnmb; (7) 294 cells highly expressed Cd209a, H2-Ab1, H2-Eb1; (8) 283 cells highly expressed S100a8 and S100a9; (9) 226 cells highly expressed Top2a and Mki67. (10) 46 cells highly expressed Adamts1 and Sparcl1 (Fig. S3). Krasniewski et al. discovered that Lyve1−/MHCII+ macrophages comprise 39.97% of all macrophages, and they also elaborated the function and polarization of these 11 macrophage subclusters13, which further confirmed our conclusion (Fig. 2G). Overall, this large single-cell data of SkM macrophages from Krasniewki et al. identified 11 clusters of different macrophages, of which CI0, CI2, and CI3 correspond to cluster 1, cluster 3, and cluster 2 of our dataset, respectively (Fig. 2H). We also confirmed that Lyve1−/MHCII+ is the main type of macrophage in the mouse SkM. However, its function and role in SkM aging should be further elucidated.Figure 2Subclusters of macrophages and marker genes of each cluster in mice SkM. (A) The UMAP plot of the six identified clusters. (B) The heatmap, violin plot (C) and the dot plot (D) showing the representative markers across six clusters. (E) The UMAP plot illustrating the macrophages subgroups based on the Lyve1/MHCII expression levels and (F) their marker genes. (G) The relative abundance of macrophages subgroups in young and old mice SkM based on Lyve1/MHCII classification. (H) The UMAP plot representing 11 distinct SkM macrophages subgroups based on marker genes in GSE195507.Functional enrichment between young and old skeletal muscle from the same subset of MacDifferently expressed genes between various Mac subpopulations in young and old SkM were analyzed by selecting the top 200 up-regulated genes and the top 200 down-regulated genes within the same cell group for enrichment analysis using the Metascape platform22. SPP1+ Mac, S100a8/9+ Mac, and Gngt2+ Mac were ruled out because they mainly disturbed in old skeletal muscle, and Adamts1+ Mac was excluded because of its small number13. Most up-regulated genes in old mice SkM macrophages were enriched in terms related to lipid metabolism, including phospholipid dephosphorylation, membrane lipid metabolic process, Ether lipid metabolism, etc., inflammation including regulation of interleukin − 6 production, positive regulation of chemokine production, etc. Most down-regulated genes in old SkM macrophages are mainly enriched in terms such as mitochondrial biogenesis, immune response-regulating signaling pathway, etc. Meanwhile, some signaling pathways, including Selenium metabolism selenoproteins, were shown in almost all cell clusters; however, several pathways, like positive regulation of cholesterol efflux and white fat cell differentiation, were only found in certain Mac subgroups, which demonstrated the homogeneity and heterogeneity of response to signals in distinct Mac subgroups of SkM (Fig. 3).Figure 3The dot plot exhibiting the homogeneity and heterogeneity in SkM macrophages subpopulations in Y and O mice.A landscape of enrichment analysis between different subpopulations of SkM macrophageEnrichment analysis was implemented on 11 SkM macrophage subpopulations using AUCell, UCell, singscore, and ssgsea to further demonstrate their homogeneity and heterogeneity. Adipogenesis, which has been proven to be increased in aged SkM5, was found to be up-regulated in SPP1+ macrophages but down-regulated in Cxcl2+ macrophages, which suggests SPP1+ Mac might be involved in the formation of intermuscular adipogenesis during skeletal muscle aging. Angiogenesis, a biological process associated with aging35, was discovered to be up-regulated in SPP1+ Mac and Addmts1+ macrophages but down-regulated in Lyve1−/Cd209a+ Mac and Thbs1+ Mac. In addition, the TNFA-SIGNALING-VIA-NFKB pathway, the INTERFERON-ALPHA-RESPONSE pathway, and the INTERFERON-GAMMA-RESPONSE pathway were revealed to be down-regulated in SPP1+ Mac. In Lyve1+ /Cd209a+ Mac, the TNFA-SIGNALING-VIA-NFKB pathway and the IL6-JAK-STAT3-SIGNALING pathway were shown to be markedly up-regulated (Fig. 4A). Figure 4B exhibited that the results explored by these four methods were principally compatible. Figure 4C suggested a markedly increased activity of adipogenesis, angiogenesis, oxidative phosphorylation, and protein secretion pathways in SPP1+ Mac. Our study found that in the subpopulations of Mac primarily distributed in aging mice SkM, angiogenesis, adipogenesis, and senescence-associated pathways were up-regulated.Figure 4Comprehensive enrichment analysis of different subpopulations of mice SkM macrophages. (A) Functional enrichment analysis of gene signature from GSEA. (B) The bar plot illustrating the number and proportion of up, down and no significant genes with 4 distinct methods. (C) Density scatterplots and density heatmaps exhibiting representative up or down-regulated pathways.SPP1+ Mac owned more senescent characteristics and stronger adipogenesis abilityAdipogenesis has been verified as a potential factor in SkM aging5,35. We, therefore, attempted to calculate the adipogenesis and senescence score of 11 SkM macrophage subpopulations in young and old mice SkM. Macrophages in the skeletal muscle of old mice had a greater adipogenesis score than those in young mice, although there was no significant difference in senescence score. Among all macrophage subgroups, SPP1+ Mac scored the highest. (Figs. 5A,B, S4A,B). Correlation analysis confirmed that adipogenesis had a strong linkage with senescence (R = 0.35, p < 2.2e−16, Fig. 5C). Cell cycle analysis of the young and old skeletal muscle Mac was then performed, respectively. Compared with Mac in young mice SkM, macrophages in old mice SkM were prone to be in the S phase, macrophages in the G2M phase remained almost unchanged, and the G1 phase was decreased, suggesting that cell cycle arrest might be involved in SkM aging (Fig. 5D). Further study verified that SPP1 was up-regulated remarkedly in SPP1+ Mac of old mice SkM compared with that in SKM of young mice (Fig. 5E), and SPP1 expression was significantly increased in macrophages of old mice SkM and mainly expressed at SPP1+ Mac (Fig. 5F). Eventually, scMetabolism was implemented to speculate the metabolism activity of SkM macrophage subpopulations at single-cell resolution. SPP1+ Mac showed the most active metabolic pathways compared to other macrophage subgroups (Fig. S5). Terms linked to lipid metabolism, including Glycerophospholipid metabolism, Ether lipid metabolism and Sphingolipid metabolism, and carbohydrate metabolism, including Glycolysis/Gluconeogenesis and energy metabolism like Oxidative phosphorylation, were found to be significantly up-regulated in macrophages of old mice SkM (Figs. 5G, S6A). Correlation analysis was then conducted between Glycerophospholipid metabolism score, Senmayo score, and adipogenesis score. Our findings proved that Senmoyo Score significantly correlated with the Glycerophospholipid metabolism positively (R = 0.38, p < 0.001, Fig. S6B). Further study uncovered that SPP1+ Mac had an increased activity in lipid metabolism like Glycerolipid metabolism, Sphingolipid metabolism, and Steroid metabolism (Figs. 5H, S7). Our study unveiled that SPP1+ Mac holds significant senescent features and enhanced adipogenesis ability via bioinformatic analysis. Additionally, the differences in metabolic pathways were also explored between SkM macrophages in young and old mice.Figure 5Characterization of SPP1+ macrophages in mice SkM. (A) The violin plot and the UMAP plot showing the distinction in Senmayo score in subgroups of mice young and old SkM macrophages. (B) The violin plot and the UMAP plot showing the distinction in adipogenesis score in subgroups of mice Y and O SkM macrophages. (C) The correlation plot exhibiting relation between adipogenesis and cellular senescence. (D) The bar plot showing the proportion of cell cycle in SkM macrophages of Y and O mice. (E) Comparison of SPP1 expression level in SPP1+ macrophages of Y and O mice SkM using scatterplot. (F) SPP1 expression level in SkM macrophages in Y and O mice using UMAP plot. (G) The heatmap showing metabolic pathway between Y and O macrophages of mice SkM. (H) The heatmap showing metabolic pathways in distinct macrophages subgroups of mice SkM. Y, young. O, old. SkM, skeletal muscle.Pseudo-time analysis of macrophages from young and old mice SkMPrevious studies have reported the polarization of SkM macrophages under different circumstances, for example, inflammation, muscle regeneration, and aging5,36. Pseudo-time analysis with Monocle 3 provides a robust tool to explore the state transition of macrophages during aging, inferring cell polarization based on gene expression patterns. Using this method, the polarization trajectories of young and old macrophages between these subgroups were then further investigated, respectively. Most of the macrophages in young mice SkM went through a polarization trajectory from Lyve1+/Mrc1+ Mac to Lyve1+/Cd209a+ Mac then to Lyz1+ Mac finally to Thbs1+ Mac (Fig. 6A,B, S8A). Top ten genes that had the most significant changes during the trajectory were exhibited in Fig. 6C. To study further, Lyve1+/Mrc1+ Mac, Lyve1+/Cd209a+ Mac, Lyz1+ Mac, and Thbs1+ Mac were extracted to be re-analyzed. The immune response-regulating signaling pathway was shown to be activated along with the pseudo-time trajectory. Additionally, the immune response-regulating signaling pathway-related genes were down-regulated in aging mice SkM macrophages, which was in line with previous enrichment analysis (Fig. 6D,E), proving that the immune response-regulating signaling pathway participated important role in the polarization of SkM macrophages in Y mice. Following that, the relationship between Lyve1+/Mrc1+ Mac, Lyve1+/Cd209a+ Mac, Lyz1+ Mac and Thbs1+ Mac was further analyzed by the pseudo-time method. The results show Lyz1+ Mac served as a state between Lyve1+ /Mrc1+ Mac and Thbs1+ Mac (Fig. 6F,G).Figure 6Pseudo-time analysis of SkM macrophages in mice. (A) UMAP plot exhibited SkM macrophage subpopulations of Y mice. (B) UMAP presenting pseudo-time trajectory of macrophages from Y mice SkM using Monocle 3. (C) Changes of canonical genes expressions in Y SkM macrophages along the pseudo-time trajectory. (D) Scatter plot showing the expression trend of genes related to the immune response-regulating signaling pathway. (E) Violin plot showing the expressions of Bcl2a1d, malt1, Lgals3, Clec4e in SkM macrophages of aging mice. (F) 3D Scatterplot representing Lyve1+/Mrc1+ Mac, Lyve1+/Cd209a+ Mac, Lyz1+ Mac, and Thbs1+ Mac. (G) 3D Scatterplot showing pseudo-time trajectory in Lyve1+/Mrc1+ Mac, Lyve1+ /Cd209a+ Mac, Lyz1+ Mac, and Thbs1+ Mac. (H) UMAP illustrates different SkM macrophage subpopulations in O mice with Monocle 3. (I) UMAP presents the pseudo-time trajectory of macrophages from aging mice SkM using Monocle 3. (J) Changes of canonical genes expressions in O SkM macrophages along the pseudo-time trajectory. (K) PPI network of SPP1 co-regulated genes. 3D, three-dimension; Y, young; O, old.In old SkM macrophages, Lyve1+/Mrc1+ Mac distributed at the beginning of the trajectory, Lyz1+ Mac followed, SPP1+ Mac and Thbs1+ Mac were seen throughout the trajectory, and Gngt2+ Mac arrived at the end of the trajectory. These results were in compliance with previous observations from Krasniewski et al. (Figs. 6H, S7B)13. Meanwhile, the top 10 genes related to the pseudo-time were identified (Fig. 6J). To explore the complicated association between Lyve1+/Mrc1+ Mac, Lyz1+ Mac, SPP1+ Mac, Thbs1+ Mac, and Gngt2+ Mac, these cells were figured out and re-analyzed. Lyve1+/Mrc1+ Mac was located at the beginning of the trajectory, followed by Lyz1+ Mac, then SPP1+ Mac and Thbs1+ Mac, and Gngt2+ Mac was at the end of the trajectory (Figs. 6I, S8A). Modules of co-regulated genes were then analyzed to illustrate the role of SPP1 better. Finally, 92 modules were identified (Supplementary Table 4). PPI network analysis was then conducted to explore co-regulated genes in highly related modules involving SPP1 (Fig. S8B). The results indicated that SPP1-related genes are involved in osteoclast signaling, antigen processing-cross presentation, ROS and RNS production in phagocytes, metabolic pathways, etc. These results exhibited that SPP1 might participate in amino sugar and nucleotide sugar metabolism and proteasome degradation as well as membrane protein complex (Fig. 6K). Briefly speaking, the polarization trajectories in SkM macrophages of young mice were identified, suggesting that the immune response-regulating signaling pathway took a vital role in keeping the normal polarization of macrophages in young mice SkM. In the meantime, the function and co-regulated association of SPP1 have also been explored.Spp1 promoted a unique cell–cell communication in aging SkM macrophagesTo unveil and compare the intercellular communications networks of different types of macrophages in young and old mice SkM, cell–cell interaction analysis was implemented using CellChat29. In the SkM macrophages from Y mice, most signaling patterns, which included CCL, GALECTIN, VISFATIN, etc., were sent from Lyve1+/Cd209a+ Mac. S100a8/9+ Mac received most signaling patterns, including MIF, CXCL, ANNEXIN, etc. (Fig. 7A,B). In aging SkM macrophages, SPP1+ Mac performed a significant role in the majority of outgoing signaling pathways, comprising SPP1, TNF, MIF, etc. Lyve1+/Cd209a+ Mac and S100a8/9+ Mac received most of the signaling. Compared to young SkM macrophages, a unique SPP1-mediated signaling was unveiled (Fig. 7C,D). Ligand receptors regulated by Lyve1+/Cd209a+ Mac were investigated as they sent most signals in SkM of young mice, and ccl2-ccr2 was finally identified (Fig. 7E). Equally, the signaling network activated by SPP1+ Mac was illustrated, and Spp1-Cd44 was present in nearly all SkM macrophage subpopulations (Fig. 7F). SPP1+ Mac played a key part as the sender, mediator, and influencer in the SPP1 signaling pathway network. Lyve1+/Cd209a+ Mac and Gngt2+ Mac mainly acted as receivers (Fig. 7G,H). In addition, Spp1-Cd44 made the greatest contribution to the SPP1 signaling pathway network (Fig. 7I). Almost all aging SkM macrophage subgroups held high levels of Spp1 induced receptors (Fig. 7J). Further, CD44 expression was significantly decreased in skeletal muscle Mac of old mice (Fig. 7K). Lastly, the communication patterns were shown in aging SkM macrophages for comprehending the synergistic effect of signaling pathways among the target cells and the secreting cells. By this means, we discovered a synergy among the SPP1 signal network, CCL, MIF, and other signals in the incoming communication pattern (Fig. 7L). When SPP1+ Mac served as secreting cells, the SPP1 and VEGF signaling were synergetic and mostly dominated by SPP1+ Mac (Fig. 7M). Conclusively, our results displayed that the abundance and peculiarity of SPP1 signaling network communication patterns in aging SkM macrophages compared to macrophages of young mice SkM, and the Spp1-Cd44 receptor-ligands binding pattern contributed mostly to this network.Figure 7Developmental relationships and communication of SkM macrophages in Y and O mice. (A) The quantity and strength of signals for SkM macrophage subgroups in Y mice. (B) Comparison of the intensity of enriched outgoing and incoming signals in Mac subgroups from Y mice SkM. (C) The quantity and strength of signals for SkM macrophage subgroups in O mice. (D) Comparison of the intensity of enriched outgoing and incoming signals in Mac subgroups from O mice SkM. (E) L-R interaction contributions in signals emitted by Lyve1+/Cd209a+ Mac. (F) L-R interaction contributions in signals emitted by SPP1+ Mac. (G) Exhibition of SPP1 signaling pathway between distinct macrophage subgroups of mice SkM using circle plot. (H) Investigation of the function of distinct SkM macrophage subgroups in the SPP1 signaling pathway. (I) Exhibition of each L-R pair’s contribution in SPP1 signaling pathway using the bar plot. (J) Illustration of SPP1 signaling pathway-associated genes expressions in SkM macrophage subgroups via the violin plot. (K) Comparison of CD44 expression level in SPP1+ macrophages of Y and O mice SkM using a scatterplot. (L, M) Alluvial plot of targeted cells’ incoming and outgoing communication patterns. L-R, ligand-receptor. Y, young. O, old. Mac, macrophages.SPP1+ Mac increased in old mice SkM, and a senotherapeutic drug decreased the proportion of SPP1+ macrophage in vivo modelsDespite the fact that we have exhibited through elaborate bioinformatics analysis that SPP1+ Mac was significant during skeletal muscle aging, confirmation through in vivo experiments is indispensable. Of note, previous studies suggested Senolytics have great potential to eliminating senescent cells in multiple diseases37,38, however, their effect on aging macrophages is unknown. Therefore, the young, old-vehicle, and old-DQ mice SkM models were then constructed to verify the role of SPP1+ Mac (Fig. 8A). We first evaluated body composition and quadriceps weight in three mice groups. Old-Veh mice showed higher body weight and quadriceps weight but lower lean mass when compared to young mice; however, there was no statistical difference in these parameters between old-Veh and old-DQ mice (Fig. 8B–E). Furthermore, we observed an increased proportion of centrally nucleated fibers in old-Veh mice compared with young mice but significantly decreased in old-DQ mice compared to old-Veh mice (Fig. 8F). These results indicated significant senescence-associated phenotypes were exhibited in aging mice SkM and could partially attenuate by D + Q treatment. Finally, we calculated and compared the proportion of SPP1+ Mac in young, old-Veh, and old-DQ mice groups. Confocal IF results revealed that SPP1+ Mac was significantly more abundant in the Old-Veh mice group compared with that in the young mice, while the reduced abundance of SPP1+ Mac in Old-DQ was found when compared to the Old-Veh group (Fig. 8G). In conclusion, the proportion of SPP1+ Mac is elevated in vivo aging mice model, and D + Q contributes to the reduction of SPP1+ Mac, which might suggest that administration of D + Q in mice could rescue SkM aging by the clearance of senescent macrophages.Figure 8SPP1+ macrophages were increased in SkM of the aging mice model, and senolytics decreased its proportion in vivo. (A) The diagram presents the experimental process. (B–E) Illustration of the gastrocnemius, quadriceps, tibialis anterior, and soleus in young, Old-Veh, and Old-DQ male mice group (B). Comparison of body weight (C), lean mass percentage (D), and quadriceps muscle weight (E) between distinct groups (n = 3 per group). (F) IF pictures of quadriceps cross-sections stained for Laminin in different groups. (G) Confocal microscopy of quadriceps IF staining exhibited F4/80 (green) positive and SPP1 (red) double positive (yellow arrows) SPP1+ macrophages in quadriceps of different groups (n = 3 per group). IF, immunofluorescence. SkM, skeletal muscle. Ns, no significant *, **, *** represent p value less than 0.05, 0.01, 0.001 respectively. Y, young; O, old.
