Diverse effector and regulatory functions of fibro/adipogenic progenitors during skeletal muscle fibrosis in muscular dystrophy

Summary

Fibrosis is a prominent pathological feature of skeletal muscle in Duchenne muscular dystrophy (DMD). The commonly used disease mouse model, mdx^5cv, displays progressive fibrosis in the diaphragm but not limb muscles. We use single-cell RNA sequencing to determine the cellular expression of the genes involved in extracellular matrix (ECM) production and degradation in the mdx^5cv diaphragm and quadriceps. We find that fibro/adipogenic progenitors (FAPs) are not only the primary source of ECM but also the predominant cells that express important ECM regulatory genes, including Ccn2, Ltbp4, Mmp2, Mmp14, Timp1, Timp2, and Loxs. The effector and regulatory functions are exerted by diverse FAP clusters which are different between diaphragm and quadriceps, indicating their activation by different tissue microenvironments. FAPs are more abundant in diaphragm than in quadriceps. Our findings suggest that the development of anti-fibrotic therapy for DMD should target not only the ECM production but also the pro-fibrogenic regulatory functions of FAPs.

Graphical abstract

Highlights

• •scRNAseq supports the effector and regulatory functions of FAPs in muscle fibrosis
• •FAPs are more abundant in the diaphragm than in quadriceps
• •Diaphragm and quadriceps FAPs contain different clusters in the steady state
• •Diaphragm and quadriceps FAPs respond differently to muscular dystrophy

Musculoskeletal medicine; Genetics

Introduction

Duchene muscular dystrophy (DMD) is the most common genetic muscle disease, which is lethal with no cure currently.^1,2 It is an X-linked recessive disease caused by the defective dystrophin gene. Fibrosis is a prominent pathologic feature of skeletal muscle, which directly contributes to muscle weakness.^3,4 Development of effective anti-fibrotic therapies relies on the understanding of molecular and cellular mechanisms mediating and regulating muscle fibrosis. The commonly used DMD mouse models, mdx and mdx^5cv, show progressive fibrosis in the diaphragm but not limb muscles.^5,6,7,8 Little is known why limb and respiratory muscles show such a difference, and understanding the cellular and molecular contributions to the difference may benefit the anti-fibrotic therapy development.

Fibrosis is characterized by excessive accumulation of extracellular matrix (ECM) proteins that consist of collagens, fibronectin, and proteoglycans.^9,10,11 Fibrosis development and progression in skeletal muscle can be promoted at different levels, including ECM gene upregulation stimulated by pro-fibrogenic factors such as transforming growth factor β1 (TGF-β1),¹² connective tissue growth factor (CTGF),¹³ and osteopontin,^14,15,16 reduced matrix metalloproteinases (MMPs)-mediated ECM protein degradation,¹⁷ increased tissue inhibitors of matrix metalloproteinases (TIMPs),¹⁷ and heightened collagen cross-linking catalyzed by lysyl oxidases (LOXs), which makes ECM stiff and resistant to proteolytic degradation.^18,19

Fibro/adipogenic progenitors (FAPs) are muscle interstitial mesenchymal progenitors, expressing mesodermal marker PDGFRα and stem cell markers Sca-1 and CD34. They are bipotent progenitors that can differentiate into fibroblasts and adipocytes.^20,21,22 They play important roles in skeletal muscle injury and repair.^22,23 Upon injury, they become activated, proliferating and expanding rapidly.^20,24,25 They provide a favorable tissue environment to promote satellite cell-mediated muscle regeneration.^{20,26,27,28,29} As muscle regeneration proceeds, excessive FAPs are cleared from the regenerative niche by apoptosis.²⁵ Failing to do so leads to the pathological accumulation of FAPs, contributing to fibro-fatty replacement of muscle.^21,25

In the present study, we performed single-cell RNA sequencing (scRNAseq) to determine the cellular expression of the genes involved in ECM production and degradation during skeletal muscle fibrosis in mdx^5cv, and to identify differences between diaphragm and quadriceps. Our findings demonstrate that FAPs are not only the key effector cells that express a high level of ECM genes, but also the predominant regulatory cells that express a high level of Ccn2, Ltbp4, Mmp2, Mmp14, Timp1, Timp2, and Loxs to regulate ECM production and degradation. The effector and regulatory functions are exerted by diverse FAP clusters, which are different between diaphragm and quadriceps, indicating different tissue microenvironments in these two types of skeletal muscles. FAPs are much more abundant in diaphragm than in quadriceps. FAPs are the key effectors and regulators of mdx^5cv diaphragm fibrosis.

Results

Fibro/adipogenic progenitors are the most abundant mononuclear cells in diaphragm but not quadriceps of both wild-type and mdx^5cv mice

Muscle necrosis, inflammation, and fibrosis are evident in both diaphragm and quadriceps muscles of mdx^5cv at 14 weeks of age. However, these changes subside afterward in the quadriceps with progressive fibrosis mainly seen in the diaphragm at 24 weeks of age (Figure S1A).^5,30 To uncover the cellular and molecular differences accounting for the divergent fibrosis progression, we performed scRNAseq using single-cell suspensions prepared from diaphragm and quadriceps of mdx^5cv and WT mice at 14 weeks of age. The single-cell suspensions contain mononuclear cells but not myofibers, as polynuclear myofibers cannot be included due to their large size. scRNAseq of the four muscle samples, each pooled from five male mice, were performed simultaneously (see Table S1 for quality control). By filtering out the cells of low quality,³¹ we obtained 16,766 genes from 3,814 cells of WT quadriceps, 17,677 genes from 6,988 cells of WT diaphragm, 18,398 genes from 6,302 cells of mdx^5cv quadriceps, and 18,152 genes from 6,089 cells of mdx^5cv diaphragm for analysis. Sequencing data were first analyzed using Uniform Mani-fold Approximation and Projection (UMAP) for dimension reduction to generate functionally enriched clusters in each sample. The identities of the generated clusters were determined by the expression of cell type-specific marker genes (Figure S2). Multiple cell types were identified in each sample (Figures 1A and S2, Tables S2 and S3). Most of the cell types were present in all four samples, but mesothelial cells (MCs) and invariant natural killer T cells (NKT) cells were only identified in diaphragm, while Schwann cells, pericytes, and tenocytes were only identified in quadriceps (Figure 1A). Mesothelium covers the surface of diaphragm but not quadriceps,³² which explains the detection of MCs only in diaphragm. The lack of other cell types in diaphragm or quadriceps is likely due to their very low abundance. Clusters with a high-level expression of Pdgfra, Ly6a, and Cd34 and lack of expression of Pecam1, Ptprc, Itga7, Tnmd, and Fmod were identified as FAPs (Figure S2B).

Figure 1.

FAPs are the most abundant among mononuclear cells in diaphragm but not quadriceps of both wild-type (WT) and mdx^5cv mice

(A) Uniform Mani-fold Approximation and Projection (UMAP) dimension reduction analysis showing different cell types in WT and mdx^5cv quadriceps and diaphragm muscles. FAP: Fibro/adipogenic progenitors, MO/MP: Monocyte/Macrophages, DC: Dendritic cells, EC: Endothelial cells, MuSC: Myogenic stem cells, B: B cells, T: T cells, NK: Natural killer cells, NKT: invariant Natural Killer T Cells, MC: Mesothelial cells, Neu: Neutrophils.

(B) Bar graph showing percentage of each cell type to total cells analyzed by scRNAseq in each indicated muscle sample, quantified by UMAP analysis.

(C) Bar graph showing percentages of FAP, EC, MuSC, and CD45⁺ cells to total mononuclear cells, quantified by FACS, in each indicated muscle sample (Qua.: quadriceps; Dia.: diaphragm).

(D) Scatterplot showing and comparing cell densities of FAPs in WT quadriceps, WT diaphragm, mdx^5cv quadriceps, and mdx^5cv diaphragm determined by FACS analysis. Data are represented as mean ± SEM Asterisks indicate significant differences (∗p < 0.05, ∗∗∗p < 0.001, ns: no significance, Kruskal-Wallis test followed by Dunn’s test for multiple comparisons).

In both WT and mdx^5cv diaphragm, FAPs were the most abundant cell type identified by scRNAseq (Figure 1A). Endothelial cells (ECs) and monocytes/macrophages (MOs/MPs) were the most abundant cell types in WT and mdx^5cv quadriceps, respectively (Figures 1A and 1B and Table S3). Since the cell representation could be altered by multiple processing steps of scRNAseq, we further determined the relative abundance of FAPs, ECs, myogenic stem cells (MuSCs), and CD45⁺ cells (contain MOs/MPs) by flow cytometry analysis (FACS) (Figures 1C and S1B). FACS and scRNAseq revealed similar percentages of these cells, except for the ECs in WT diaphragm, which showed a much lower percentage by scRNAseq than by FACS (Figures 1B and 1C). FAPs are therefore over-represented in WT diaphragm by scRNAseq. But FACS also showed that FAP was the most abundant cell type in WT diaphragm (Figure 1C). The FAP cell density, determined by FACS, was markedly higher in diaphragm than in quadriceps (Figure 1D). It increased dramatically in both quadriceps and diaphragm of mdx^5cv as compared with WT at 14 weeks of age, but it was then reduced in mdx^5cv quadriceps while remained high in mdx^5cv diaphragm at 24 weeks of age (Figure 1D), corresponding to the progressive fibrosis seen in mdx^5cv diaphragm but not quadriceps (Figure S1). The findings are consistent with previous studies reporting a higher level of PDGFRα protein expression in diaphragm than in limb muscles.^33,34,35

Fibro/adipogenic progenitors are the primary cells expressing extracellular matrix genes in the skeletal muscle of both wild-type and mdx^5cv mice

FAPs are known ECM-producing cells.³⁶ To determine the relative contribution of different cell types to ECM production, we compared the cellular expression of the genes encoding collagens, fibronectin, proteoglycans, and synthesizing enzymes for glycosaminoglycans (GAGs) which are the components of many proteoglycans¹⁰ (Figures 2, S3, and S4). For all the collagen genes with detectable expression by scRNAseq, the high expression was predominantly detected in FAPs, as well as tenocytes in quadriceps, which include Col1s and Col3s that encode collagen I and collagen III, the two major fibrillary collagen constituents in interstitial tissue of skeletal muscle³⁷ (Figures 2A and S3). FAPs and quadriceps tenocytes were also the major cell types that expressed a high level of fibronectin gene (Fn1) (Figure 2A) and many proteoglycan genes (Figures 2B and S4A). In addition, FAPs moderately expressed several genes encoding the synthesizing enzymes of GAGs (Figure S4B). MCs, Schwann cells, pericytes, and MuSCs also expressed multiple ECM genes but at a lower level (Figures 2A and 2B). Considering the relative low abundance, their contribution to the ECM production is likely limited. To determine the potential contribution by myofibers which cannot be included for scRNAseq, we performed quantitative reverse-transcription PCR (qRT-PCR) to compare FACS-sorted FAPs, FACS-sorted non-FAP cells, and whole muscle tissue samples from mdx^5cv mice. The mRNA expression of Col1a1, Col3a1, and Fn1 by the whole muscle tissue was significantly lower than that by the two sorted cell samples (Figure 2C), indicating that myofibers do not significantly contribute to ECM production in dystrophic muscles. Together, our findings are consistent with the notion that FAPs are the key ECM-producing cells in skeletal muscle in both steady state and mdx^5cv. Given the persistent high FAP density in mdx^5cv diaphragm (Figure 1C), FAPs are the main effector cells mediating the progression of fibrosis in this muscle.

Figure 2.

FAPs are the predominant cells expressing extracellular matrix (ECM) genes in both WT and mdx^5cv quadriceps and diaphragm

(A) Dot plot showing the expression of collagen and fibronectin genes by different cell types in different muscles. The dot color intensity reflects the average gene expression level, and the dot size represents the percentage of the cells expressing the gene.

(B) Dot plot showing the expression of proteoglycan genes.

(C) qRT-PCR showing and comparing Col1a1 and Col3a1, and Fn1 expression by FACS-sorted FAPs, FACS-sorted non-FAPs, and whole muscle of mdx^5cv quadriceps and diaphragm at 14 weeks of age. Data are represented as mean ± SEM Asterisks indicate significant differences (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ns: no significance, Kruskal-Wallis test followed by Dunn’s test for multiple comparisons). N = 5 mice/group.

Fibro/adipogenic progenitors are the predominant cells expressing many regulatory genes involved in skeletal muscle extracellular matrix production and degradation

To address which cell types are the key regulators of skeletal muscle fibrosis, we next analyzed the cellular expression of the genes encoding fibrogenic factors, MMPs, TIMPs, and LOXs.

Among the genes that regulate FAP activation and ECM production (Figures 3A and S5A), the CTGF gene (Ccn2), insulin-like growth factor-1 (IGF-1) gene (Igf1), and follistatin gene (Fst) were preferentially expressed by FAPs in both WT and mdx^5cv diaphragm. In quadriceps, tenocytes expressed a higher level of Ccn2 and Igf1 than FAPs, while FAPs were the major source of Fst. FAPs also expressed a high level of latent TGFβ binding protein 4 gene (Ltbp4). The FAP expression of these genes has also been reported by other studies.^{20,24,27,38,39} The findings indicate that FAPs regulate their own differentiation and ECM gene expression. FAPs in diaphragm and tenocytes in quadriceps expressed a high level of Tgfb3. Tgfb1 and Spp1, the two potent pro-fibrogenic genes, were mostly expressed by MOs/MPs in mdx^5cv, consistent with the notion that MOs/MPs promote muscle fibrosis in mdx.^40,41 MuSCs were the predominant cells expressing Tgfb2 and Pdgfa in mdx^5cv but not WT muscles, indicating their acquired fibrogenic potential in mdx^5cv. Pdgfb was mainly expressed by ECs in mdx^5cv muscles. These findings are consistent with the previous report.⁴² IFN-γ gene (Ifng) was mostly expressed by NK cells. Taken together, different fibrogenic factor genes are preferentially expressed by different cells, with FAPs, tenocytes, and immune cells being the most important contributors. The contribution of MCs to diaphragm fibrosis is likely limited due to their rarity.

Figure 3.

FAPs are the predominant cells expressing important regulatory genes of ECM production and degradation

(A-C) Dot plots showing the expression of fibrogenic factor genes (A), Mmps (B), Timps (B), and Loxs (C) by different cell types in quadriceps and diaphragm of WT and mdx^5cv mice.

(D) qRT-PCR showing and comparing the expression of selected ECM regulatory genes by FACS-sorted FAPs, FACS-sorted non-FAPs, and whole muscle of mdx^5cv quadriceps and diaphragm at 14 weeks of age. Data are represented as mean ± SEM Asterisks indicate significant differences (∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns: no significance, Kruskal-Wallis test followed by Dunn’s test for multiple comparisons). N = 5 mice/group.

Fibrosis is regulated not only at the level of ECM production but also at the level of ECM degradation. We next addressed which cells expressed Mmps, Timps, and Loxs to regulate ECM degradation (Figures 3B, 3C, and S5B). Among all the Mmps, Mmp2, Mmp9, and Mmp14 were strongly expressed (Figure S5B), with the expression of Mmp2 and Mmp14 mainly by FAPs and tenocytes while the expression of Mmp9 by neutrophils (Figure 3B). Timps were expressed by multiple cell types (Figure 3B), but Timp1 and Timp2 were preferentially expressed by FAPs and tenocytes. ECs were the main cells expressing Timp3 and Timp4, except for WT diaphragm where the Timp4 expression was diminished. Therefore, FAPs appear to play an active role in regulating ECM degradation. The genes that encode cross-linking enzymes, LOXs, were mainly expressed by FAPs and tenocytes (Figure 3C). FAPs may thus also regulate collagen cross-linking.

Taken together, FAPs actively express Ccn2, Ltbp4, Igf1, Fst, Mmp2, Mmp14, Timp1, Timp2, and Lox genes in skeletal muscle, regulating ECM production and degradation. Quantitative RT-PCR with FACS-sorted FAPs and non-FAP cells, as well as whole muscle samples, confirmed the major contribution of FAPs to the ECM regulatory gene expression in mdx^5cv quadriceps and diaphragm (Figure 3D). The findings support the notion that FAPs may play a key role not only in mediating but also in regulating skeletal muscle ECM homeostasis in the steady state and fibrosis in muscular dystrophy.^{21,24,25,33,35,43,44,45}

Fibro/adipogenic progenitors in mdx^5cv diaphragm and quadriceps display differences in transcriptomes

Since FAPs not only produce ECM but also regulate ECM production and degradation, we next determined whether the transcriptomes of FAPs in mdx^5cv quadriceps and diaphragm are different to account for the divergent fibrosis progression in these two muscles. To this end, transcriptome data of FAPs from all four muscle samples (WT quadriceps: 846 cells; WT diaphragm: 5,829 cells; mdx^5cv quadriceps: 1,140 cell; mdx^5cv diaphragm: 2,515 cells) were merged and analyzed for differentially expressed genes (DEGs, log₂FC ≥ 0.5 compared to the average expression level). UMAP analysis showed that FAP transcriptomes were different by the muscle type (quadriceps vs. diaphragm) and the disease state (WT vs. mdx^5cv) (Figure 4A), with 835 distinct DEGs identified in mdx^5cv quadriceps and 78 in mdx^5cv diaphragm (Figure 4B and Data S1). The DEGs in mdx^5cv quadriceps FAPs were functionally enriched in signaling pathways of TNF-α, IL-4 and IL-13, TGF-β, PDGF, and HGF, and in biological processes of cell migration, apoptosis, and angiogenesis. The DEGs in mdx^5cv diaphragm FAPs were functionally enriched in TGF-β response and ECM regulation. It is surprising that most of the ECM genes, fibrogenic factor genes, Mmps, Timps, and Loxs were not present in the distinct DEGs of mdx^5cv quadriceps or diaphragm FAPs (Data S1). Likewise, qRT-PCR showed no significant difference in Col1a1, Col3a1, Fn, Igf1, Mmp14, Timp1, Timp2, Lox, and Loxl1 expression between mdx^5cv quadriceps and diaphragm FAPs (Figures 2C and 3D). The expression of Ccn2, Ltbp4, and Mmp2 only showed a less than 2-fold increase in mdx^5cv diaphragm FAPs as compared with mdx^5cv quadriceps FAPs (Figure 3D). Therefore, there is no significant difference in the FAP expression of the ECM or ECM regulatory genes at a single-cell level between the two muscles. However, the gene expression of Col1a1, Col3a1, Igf1, Mmp2, Mmp14, Lox and Loxl1 (Figure 4C) and the protein expression of collagen I, collagen III, and fibronectin (Figures 4D and S3B) at a whole-muscle level was significantly higher in mdx^5cv diaphragm than in mdx^5cv quadriceps, which is most likely contributed by the higher FAP density in mdx^5cv diaphragm.

Figure 4.

FAPs in WT and mdx^5cv diaphragm and quadriceps display differences in transcriptomes

(A) UMAP dimension reduction analysis showing different FAP transcriptomes in different muscle samples.

(B) Venn diagram depicting differentially expressed genes (DEGs) shared by or unique in FAPs of different muscles.

(C) qRT-PCR showing and comparing the expression of selected ECM and ECM regulatory genes by different muscles. Data are represented as mean ± SEM Asterisks indicate significant differences (∗p < 0.05, ∗∗p < 0.01; ∗∗∗p < 0.001, ns: no significance, Kruskal-Wallis test followed by Dunn’s test for multiple comparisons). N = 5 mice/group.

(D) Western blot showing and comparing the protein expression of collagen 1α1 (COL1A1), collagen 3α1 (COL3A1), and fibronectin among different muscles. Results showed represent 3 independent experiments with a total of 6 mice for each group.

Fibro/adipogenic progenitors consist of diverse clusters which are different between quadriceps and diaphragm in wild-type and mdx^5cv mice

FAPs are functionally heterogeneous in both steady state and disease state.^{22,36,44,46,47,48,49} There are three types of tissue fibroblast clusters: universal clusters that are shared by all tissue fibroblasts in the steady state, specialized clusters that are induced by the tissue-specific microenvironment in the steady state, and activated clusters that are induced by a disease.⁵⁰ We next addressed whether quadriceps and diaphragm have different specialized FAP clusters in the steady state and different activated FAP clusters in mdx^5cv and whether different FAP clusters exert different effector and regulatory functions. To this end, we extracted transcriptome data of FAPs and re-clustered by UMAP analysis. Multiple clusters with differential transcriptomes were identified in each sample (Figures 5A and S6A).

Figure 5.

FAPs consist of diverse clusters which are different among quadriceps and diaphragm in WT and mdx^5cv mice

(A) UMAP clustering identifying FAP clusters in each muscle sample.

(B) Pseudotime analysis showing the trajectory of FAP differentiation starting from the Ly6c1^hiP16^hi cluster (labeled as 1) in each muscle sample.

Five FAP clusters were identified in WT quadriceps. The two universal clusters identified by the cross-tissue transcriptome analysis of fibroblasts,⁵⁰ including the adventitial stromal cell-like fibroblasts featured by Ly6c1, Pi16, Dpp4, and CD55 and the basement membrane-associated fibroblasts featured by Hsd11b1, Cxcl14, Col4a1, Col15a1, and Hspg2,⁵⁰ were identified in our WT quadriceps FAPs and named Ly6c1^hi and Hsd11b1 clusters, respectively (Figures 5A, 6A-6C, and S6A.1; Table S4). Similar FAP clusters were also identified by two previous transcriptome studies^36,46 and named Dpp4⁺ and Cxcl14⁺ clusters.⁴⁶ The findings confirm the reliability of our analysis. We named the clusters with different featured genes to best distinguish them from the other clusters in all four samples. Interestingly, we identified two adventitial Ly6c1^hi FAP clusters, which were distinguished by the expression level of Pi16: Ly6c1^hiPi16^hi and Ly6c1^hiPi16^lo (Figures 5A and 6B; Table S4). Compared to the Ly6c1^hiPi16^hi cluster, the Ly6c1^hiPi16^lo cluster expressed a lower level of Dpp4 but a higher level of Col4a1 (Figures 6B and 6C), suggesting that this is an intermediate cluster between the adventitia stromal cell-like and basement membrane-associated clusters. The Ly6c1^hi clusters expressed the highest level of mesenchymal progenitor markers Cd34 and Ly6a/Sca-1 (Figure 6A), indicating their higher stemness than the other clusters. Pseudotime analysis confirmed the progenitor status of the Ly6c1^hiPi16^hi cluster (Figure 5B), consistent with the finding by others.⁵⁰ The Ly6c1^hi clusters also expressed the highest level of TEK tyrosine kinase gene (Tek) (Figure 6B), an important molecule involved in vascular maintenance and homeostasis.^44,51 The Hsd11b1 cluster expressed a high level of homeostatic chemokine genes, Cxcl14 and Ccl11 (Figure 6C and Table S4), presumably helping maintain tissue homeostasis.^52,53 The two other clusters are specialized (Figures 5A, 6D, and S6A.1; Table S4). The Gdf10 cluster was enriched for genes involved in chondrogenesis, such as Gdf10, Mgp, Prg4, and Cilp (Figure 6D and Table S4). The pro-inflammatory cluster was very small (Figure S6B) and enriched for genes upregulated during inflammation, such as Cxcl1, Ccl7, and Spp1 (Table S4). As skeletal muscle suffers constant micro-injuries due to mechanical stretch, the pro-inflammatory cluster may represent the FAPs that respond to micro-injuries. To infer the progression trajectories among FAP clusters, we did Monocle pseudotime trajectory analysis (Figure 5B). The Ly6c1^hiPi16^lo cluster bridged the progression from the Ly6c1^hiPi16^hi cluster to the Hsd11b1 cluster, and the two specialized clusters were directly from the Hsd11b1 cluster. The findings are consistent with the analysis of cross-tissue fibroblasts.⁵⁰

Figure 6.

FAP clusters show diverse gene expression depending on the muscle type and disease state

Violin plots showing gene expression by different FAP clusters in different muscle samples.

(A) FAP marker genes.

(B) Featured genes of Ly6c1^hi cluster.

(C) Featured genes of Hsd11b1 cluster.

(D) Gdf10.

(E) Featured genes of Sfrp2 cluster.

(F) Featured genes of Postn cluster.

(G) Featured genes of Acta2 cluster.

(H) Ccl2.

(I) Ifit1.

Seven FAP clusters were identified in WT diaphragm, of which five were shared by WT quadriceps, including Ly6c1^hiPi16^hi, Ly6c1^hiPi16^lo, Hsd11b1, Gdf10, and pro-inflammatory clusters (Figures 5A, 6A-6D, and S6A.2; Table S5). But they all showed some differences in featured gene expression compared to their counterparts in WT quadriceps (Tables S4 and S5). The differential FAP specialization was further demonstrated by the presence of two additional FAP clusters in WT diaphragm: the Sfrp2 cluster and the Postn cluster. The Sfrp2 cluster featured Sfrp2 expression and enriched pro-fibrotic genes, including Ccn2, Eln, and Mfap4 (Figure 6E and Table S5). The Postn cluster was also enriched for pro-fibrotic genes, including Ccn2, Postn, and C1qtnf3 (Figures 6E and 6F and Table S5). Interestingly, Sfrp2 was selectively expressed by the Ly6c1^hiPi16^hi cluster in quadriceps while by the Sfrp2 cluster in diaphragm, (Figure 6E). Sfrp2 gene encodes secreted frizzled-related protein 2 (SFRP2), which is an inhibitor of Wnt signaling and may regulate cell proliferation⁵⁴ and tissue fibrosis.^55,56 These findings suggest a generally higher pro-fibrotic microenvironment in diaphragm than in quadriceps, consistent with the higher FAP density in diaphragm than in quadriceps. In addition, the percentage of pro-inflammatory cluster increased dramatically in WT diaphragm as compared to WT quadriceps (Figure S6B), suggesting an increased level of micro-injuries in the steady state diaphragm. Pseudotime analysis showed that all the specialized clusters, in parallel with the Hsd11b1 cluster, were progressed from the Ly6c1^hiPi16^lo cluster which was directly from the Ly6c1^hiPi16^hi progenitor (Figure 5B).

Seven FAP clusters were identified in mdx^5cv quadriceps. Compared to WT quadriceps FAPs, the mdx^5cv quadriceps FAPs lost the Ly6c1^hiPi16^lo intermediate universal cluster but acquired three new activated clusters: Postn, Acta2, and Gm42418 clusters (Figures 5A, 6A-6G, and S6A.3; Table S6). The Postn cluster was also enriched for Ccn2, Postn, and C1qtnf3 as seen in WT diaphragm (Figures 6E and 6F and Table S6). The unique small Acta2 cluster featured the expression of Acta2, which encodes α-smooth muscle actin that is considered a marker of activated myofibroblasts.⁵⁷ This cluster was also enriched for genes involved in cell proliferation, including Stmn1, Hmgb2, and S100a4 but not Cdk genes (Figure 6G and Table S6). Therefore, it may represent a cycling population of activated fibroblasts. The small Gm42418 cluster featured high expression of Gm42418, Cd74, and several ribosome protein genes (Table S6). Pro-fibrotic Cthrc1, which was not detectable in WT FAPs, was enriched in the Postn and Acta2 clusters (Figure 6F). The FAP expression of Postn and Cthrc1 was also reported in acutely injured leg muscle.^36,46 Vcam1, a marker for pro-fibrotic FAPs,⁴⁴ was detected in the Hsd11b, Postn, and Acta2 clusters but not in WT FAPs (Figure 6B). The Ly6c1^hiPi16^hi, Hsd11b1, Gdf10, and pro-inflammatory clusters in mdx^5cv quadriceps also showed profound changes in transcriptomes when compared to their counterparts in WT quadriceps (Tables S4 and S6). For instance, Ccl2 expression was observed in the pro-inflammatory FAP cluster in mdx^5cv quadriceps but not in WT quadriceps or diaphragm (Figure 6H), and CCL2 mediates MO/MP infiltration in injured skeletal muscle.⁵⁸ Likewise, a subpopulation of FAPs in acutely injured muscle also expressed Ccl2.⁴⁶ Importantly, the relative abundance of the pro-inflammatory cluster was significantly increased, while the Ly6c1^hiPi16^hi progenitor cluster was significantly decreased (Figure S6B). Along with the loss of the Ly6c1^hiPi16^lo cluster, these changes indicate that the universal Ly6c1^hi clusters are greatly perturbed by muscular dystrophy, and they are differentiated into pro-inflammatory and pro-fibrotic clusters, corresponding to the inflammation and fibrosis seen in mdx^5cv quadriceps at 14 weeks of age (Figure S1). Pseudotime analysis showed that the Hsd11b1 cluster, progressed from the Ly6c1^hiPi16^hi cluster, is the source of all the specialized and activated clusters (Figure 5B).

Mdx^5cv diaphragm FAPs contained six clusters, four of which were also present in WT diaphragm FAPs, including the Ly6c1^hiPi16^hi, Sfrp2, Postn, and pro-inflammatory clusters (Figures 5A, 6, and S6A.4; Table S7). The Postn and pro-inflammatory clusters acquired the expression of pro-fibrotic Vcam1and Cthrc1 (Figures 6B and 6F). As seen in mdx^5cv quadriceps, FAPs in mdx^5cv diaphragm did not contain the Ly6c1^hiPi16^lo cluster. Compared to the other three samples, FAPs in mdx^5cv diaphragm showed substantial changes in clustering, including the loss of the universal Hsd11b1 cluster and the specialized Gdf10 cluster, and the presence of two distinct activated clusters: the stressed cluster and the interferon-activated cluster (Figures 5A, 6I, and S6A.4; Table S7). The stressed cluster was enriched for stress-response genes, including Apod, Egr1, Jun, and Hspa1b (Table S7), a sign of increased tissue stress in mdx^5cv diaphragm. The interferon-activated cluster uniquely expressed interferon-response genes, including Ifit1, Ifit3, Ligp1, and Cxcl9 (Figure 6I and Table S7), which was likely related to the significant Ifng expression by NK cells in mdx^5cv diaphragm (Figure 3A). Although this cluster is very small (Figure S6B), its unique presence suggests a different inflammatory microenvironment in mdx^5cv diaphragm than in mdx^5cv quadriceps. Pseudotime analysis showed a sequential progression: Ly6c1^hiPi16^hi → Stressed → Sfrp2 → Postn → pro-inflammatory clusters, with the interferon-activated cluster progressed from the stressed cluster (Figure 5B).

Taken together, quadriceps and diaphragm FAPs are transcriptionally different, each containing unique specialized clusters in the steady state and distinct activated clusters in mdx^5cv, with more pro-fibrotic features present in the diaphragm FAPs than in the quadriceps FAPs.

Extracellular matrix and extracellular matrix regulatory genes are expressed by multiple Fibro/adipogenic progenitor clusters in mdx^5cv skeletal muscles

To further address the contributions of individual FAP clusters to the effector and regulatory functions related to skeletal muscle fibrosis in mdx^5cv, we analyzed the expression of the genes involved in ECM production and degradation by different FAP clusters in mdx^5cv diaphragm and quadriceps.

The Postn cluster expressed the highest level of most of the collagen genes in both quadriceps and diaphragm, including Col1s and Col3s (Figure 7A). Fn1, however, was preferentially expressed by the Ly6c1^hiPi16^hi cluster (Figure 7A). Different proteoglycan genes were preferentially expressed by different FAP clusters (Figure 7B). Ccn2 was preferentially expressed by the Sfrp2 cluster in diaphragm while was similarly expressed by several clusters in quadriceps (Figure 7C). The Hsd11b1 cluster in quadriceps and the Sfrp2 cluster in diaphragm expressed the highest level of Mmp2 (Figure 7D). The quadriceps Acta2 cluster expressed a relatively high level of Loxs, especially Loxl3 (Figure 7F). The Postn cluster in both muscles expressed the highest level of Igf1, Mmp14, and Loxl1, as well as a moderate level of Lox, Loxl2, and Loxl3 (Figures 7C-7F). The pro-inflammatory cluster expressed the highest level of Fst and Timp1 in both muscles (Figures 7C and 7E) and the highest level of Lox in diaphragm (Figure 7F). Timp2 and Timp3 were highly expressed by the Ly6c1^hiPi16^hi cluster (Figure 7E). These findings suggest that the ECM production and degradation in mdx^5cv skeletal muscle may be regulated by multiple specialized and activated FAP clusters, and the Postn cluster appears the most pro-fibrotic one.

Figure 7.

ECM genes and ECM regulatory genes are expressed by multiple FAP clusters

(A–F) Dot plot showing the expression of collagen and fibronectin genes (A), proteoglycan genes (B), fibrogenic factor genes (C), Mmps (D), Timps (E), and Loxs (F) by FAP clusters of mdx^5cv quadriceps and diaphragm.

Discussion

Fibrosis is a prominent pathological feature of skeletal muscle in patients with DMD and animal models, which directly contributes to muscle weakness.^3,4 FAPs are the main fibrogenic cells in skeletal muscle.^20,21 By scRNAseq analysis of intramuscular mononuclear cells derived from limb (quadriceps) and respiratory (diaphragm) muscles of both WT and mdx^5cv mice, our present study has generated several important findings, which deepens our understanding of the cellular and molecular mechanisms mediating and regulating skeletal muscle fibrosis.

First, despite more severe and progressive fibrosis seen in mdx^5cv diaphragm than in mdx^5cv quadriceps, the expression of most ECM genes by FAPs at a single-cell level is not different between these two muscles. But the ECM gene and protein expression at a whole-muscle level is significantly higher in mdx^5cv diaphragm than in mdx^5cv quadriceps. Since FAPs are the primary source of ECM in skeletal muscle, the relatively high ECM gene expression in mdx^5cv diaphragm is most likely contributed by the high abundance of FAPs in this muscle. It remains possible that the ECM protein degradation may also be different between mdx^5cv diaphragm and quadriceps.

The difference in the FAP density between mdx^5cv quadriceps and diaphragm is prominent. It could be in part intrinsic to the muscle type, as the FAP density is much higher in WT diaphragm than in WT quadriceps. Interestingly, the percentages of proliferating and apoptotic FAPs are both higher in mdx^5cv quadriceps than in mdx^5cv diaphragm (Figures S7A and S7B), indicating a higher FAP turnover in mdx^5cv quadriceps than in mdx^5cv diaphragm. The expression of the FAP senescence genes, including Cdkn1a, Cdkn2a, Trp53, and Fas,⁵⁹ is low in both mdx^5cv quadriceps and diaphragm FAPs (Figure S7C). Apoptosis, proliferation, and fibrogenic function of FAPs are regulated by TNF-α and TGF-β signaling. While TNF-α signaling induces FAP apoptosis, TGF-β signaling inhibits FAP apoptosis,²⁵ and it also promotes FAP proliferation and fibrogenic differentiation.^34,35,60 While TGF-β signaling is enriched in the DEGs of both mdx^5cv diaphragm and quadriceps FAPs, TNF-α signaling is enriched only in the mdx^5cv quadriceps FAPs. The difference might contribute, in part, to the persistent high number of FAPs in mdx^5cv diaphragm.

Second, our study supports the notion that FAPs are not only important effectors but also the master regulators of ECM homeostasis and fibrosis in skeletal muscle. FAPs are known primary ECM producers. Our study further suggests that FAPs may also be the critical regulators of ECM production and degradation, as they may regulate many aspects of skeletal muscle fibrogenesis by expressing Ccn2, Ltbp4, Mmp2, Mmp14, Timp1, Timp2, and Loxs.

Fibrosis is the end result of chronic inflammation, and the two potent pro-fibrogenic genes, Tgfb1 and Spp1, are mainly expressed by MOs/MPs in mdx^5cv muscles. But Ltbp4, an important regulatory gene of TGF-β bioavailability⁶¹ and a genetic modifier of DMD mice and humans,^62,63,64,65 is predominantly expressed by FAPs. FAPs also express TGF-β1, TGF-β2, and TGF-β3, which can promote their own fibrogenic differentiation.²⁴ Another important pro-fibrogenic growth factor, CTGF, is highly expressed by FAPs. CTGF stimulates Col1, Col3, and Fn expression. Ccn2 overexpression in mice caused dystrophic changes,⁶⁶ while CTGF inhibition in mdx reduced muscle fibrosis without affecting TGF-β signaling.⁶⁷ Interestingly, a recent study showed that CTGF, derived from myofibers but not fibroblasts, affected collagen content and organization, contributing to muscular dystrophy in δ-sarcoglycan-null mice.³⁸ The role of FAP-derived CTGF in mdx^5cv needs to be further determined. FAPs are also the major cells that express the IGF-1 and follistatin genes, regulating not only muscle fibrosis but also muscle regeneration.

MMPs and TIMPs are multifunctional proteins, regulating inflammation, fibrogenesis, myogenesis, and angiogenesis in skeletal muscle via controlling the degradation of ECM and the activation of growth factors and cell adhesion molecules.^17,68 Of the three Mmps that are highly expressed by skeletal muscle mononuclear cells, two are highly expressed by FAPs, including Mmp2 and Mmp14. Mmp9 is mainly expressed by neutrophils. These MMPs play a key role in collagen and fibronectin degradation, which can limit fibrosis. On the other hand, MMP-9 and MMP-14 can also promote fibrosis by cleaving and activating latent TGF-β1 and CTGF.^69,70 They are double-edged swords. Although MMP-14 has been implicated in promoting adipogenic differentiation of FAPs,^71,72 the expression of the adipogenic genes, including Pparg, Fabp4, Adipog, and Cebpa, is barely detectable in mdx^5cv FAPs (Figure S7D). MMP-14, along with TIMP2, activates MMP-2, and both MMPs are critically involved in skeletal muscle maturation.^{17,68,73,74,75} Ablation of MMP-2 in mdx impaired muscle regeneration and angiogenesis.⁷⁶ The role of MMP-14 in mdx has not been well studied. Given the diverse functions of this membrane-bound MMP, the overall role of the FAP-derived MMP-14 in mdx^5cv is difficult to predict and needs to be further determined.

TIMPs inhibit the proteolytic activity of MMPs. Among the four TIMP genes, Timp1 and Timp2 are mainly expressed by FAPs. FAPs are thus likely the significant contributors to the elevated TIMP-1 and TIMP-2 levels in the plasma and skeletal muscles of patients with DMD and animal models.^77,78,79,80 TIMP-1 and TIMP-2 can promote fibrosis not only by reducing ECM proteolysis but also by stimulating fibrogenic cell proliferation and collagen production.¹⁷ FAP-derived TIMPs may thus contribute significantly to the skeletal muscle fibrosis in mdx^5cv. Being the major cells that express Timp3 and Timp4, ECs may regulate FAP differentiation, as Timp3 inhibits adipogenic differentiation of FAPs.^72,81,82

The important regulatory functions of FAPs are further exemplified by their significant expression of Loxs for collagen cross-linking. FAPs, therefore, may regulate the mechanical properties of ECM and contribute to the increased collagen cross-linking detected in skeletal muscles of patients with DMD and animal models.¹⁹ Taken together, FAPs appear critically involved in the regulation of skeletal muscle fibrosis in mdx^5cv. FAPs may also regulate inflammation, regeneration, and angiogenesis associated with muscular dystrophy via their predominant expression of important growth factors, Mmps, and Timps.

Third, our study demonstrates the presence of FAP clusters specialized by muscle type in the steady state, limb (quadriceps) muscle vs. respiratory (diaphragm) muscle. While FAPs in both muscles contain the universal adventitia-associated Ly6c1^hi clusters and basement membrane-associated Hsd11b1 cluster and share the specialized Gdf10 and pro-inflammatory clusters, the diaphragm FAPs also contains the uniquely specialized Postn and Sfrp2 clusters. Based on the enriched genes in the muscle type-specific FAP clusters, the diaphragm appears to have an increased level of micro-injury and a generally high pro-fibrotic microenvironment compared to the quadriceps. Likewise, muscle type-specific clusters have also been identified in the steady state resident macrophages, which also suggests an increased level of micro-injury in diaphragm as compared to quadriceps.⁸³ The intrinsic differences in tissue microenvironment may be related to the different structural origins that limb and diaphragm muscles arise from during embryonic development,²² and they may contribute, in part, to the differential responses to a variety of muscle diseases. While some muscle diseases affect both limb and respiratory muscles, others preferentially affect limb or respiratory muscles.

Fourth, mdx^5cv diaphragm and quadriceps FAPs respond differently to muscular dystrophy, with more profound changes seen in the FAP clustering in the diaphragm than in the quadriceps. Both universal and specialized clusters are greatly perturbed by the muscular dystrophy in mdx^5cv. While FAPs in both muscles lose the intermediate Ly6c1^hiPi16^lo cluster, FAPs in mdx^5cv diaphragm also lose the universal Hsd11b1 cluster and the specialized Gdf10 cluster. Hsd11b1 encodes 11b-Hydroxysteroid Dehydrogenase 1, which has been implicated in suppressing fibrosis, as deficiency or inhibition of this enzyme promotes fibroblast activation and enhances fibrosis in liver and skin.^84,85 The Hsd11b1 cluster is thus likely differentiated into more pro-fibrotic clusters in mdx^5cv diaphragm. The selective loss of the Hsd11b1 cluster in mdx^5cv diaphragm also suggests more severe perturbation of tissue homeostasis in mdx^5cv diaphragm than in mdx^5cv quadriceps, as this cluster is enriched for homeostatic chemokine genes. PDGFRα⁺ mesenchymal progenitors-derived GDF10 has been shown to maintain skeletal muscle integrity⁸⁶ and suppress adipogenic differentiation of FAPs.⁴⁹ The loss of this cluster may suggest its differentiation into pathogenic clusters to contribute to muscular dystrophy in mdx^5cv diaphragm. Mdx^5cv diaphragm acquires two uniquely activated clusters: the stressed cluster and the interferon-activated cluster. The presence of the stressed cluster suggests exceptionally high stress in the microenvironment of mdx^5cv diaphragm. Since many stress-responsive genes are also potent pro-inflammatory genes, such as Egr1 and Jun,^87,88 the presence of the stressed cluster also indicates a high pro-inflammatory microenvironment in mdx^5cv diaphragm. The unique presence of the interferon-activated FAP cluster further supports a special inflammatory microenvironment in mdx^5cv diaphragm. These two uniquely activated clusters likely contribute to the persistent inflammation and progressive fibrosis in mdx^5cv diaphragm.

In summary, our study supports the critical roles of FAPs in skeletal muscle ECM homeostasis and fibrosis, both as predominant effectors and crucial regulators. The persistent high abundance of FAPs appears the main contributor to the progressive fibrosis in mdx^5cv diaphragm. Limb and diaphragm muscles have different embryonic origins, locate at different anatomical positions, and work under different mechanical stress. These differences likely drive the differential FAP specialization in the steady state and differential FAP activation in muscular dystrophy. The high pro-fibrotic microenvironment likely contributes to the persistent high FAP abundance and progressive fibrosis in mdx^5cv diaphragm. The anti-fibrotic therapy development for DMD should target not only the ECM production but also the pro-fibrogenic regulatory functions of FAPs.

Limitations of the study

Since polynuclear myofibers cannot be included in single-cell suspensions for scRNAseq or FACS sorting due to their large size, their expression of the ECM and ECM regulatory genes cannot be directly assessed or compared with FAPs. Future snRNAseq can help circumvent the difficulty. scRNAseq cannot assess locations of FAPs in dystrophic muscle, while spatial transcriptomics can (https://www.biorxiv.org/content/10.1101/2022.03.17.484699v1.full.pdf). Future spatial transcriptomics studies will be helpful to localize different FAP clusters and correlate with histopathological changes in dystrophic muscle. Our scRNAseq is performed using the mice at 14 weeks of age, after which mdx^5cv diaphragm and limb muscles undergo divergent fibrosis progression. Future studies at later time points may help identify additional differences in FAP transcriptomes that underlie the divergent fibrosis progression. Our study has identified several regulatory genes of fibrosis that are predominantly expressed by FAPs. Future studies with FAP-specific deletion are needed to verify the functional importance of the FAP expression of these genes in muscular dystrophy.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti- a/b-tubulin	Cell Signaling Technology	Cat# 2148S; RRID: AB_2288042
Rabbit anti-collagen 1a1	Cell Signaling Technology	Cat# 72026S; RRID: AB_2904565
Rabbit anti-fibronectin	Cell Signaling Technology	Cat# 63779S
Rabbit anti-collagen III	Abcam	Cat# ab7778; RRID:AB_306066
IRDye 800CW anti-rabbit IgG	LI-COR Biosciences	Cat# 926-32211; RRID: AB_621843
PerCP-Cy5.5 Rat anti-mouse CD45	BD Biosciences	Cat# 550994; RRID:AB_394003
PE Rat anti-mouse CD45	BD Biosciences	Cat# 553081; RRID:AB_394611
APC Rat anti-mouse CD31	BD Biosciences	Cat# 551262; RRID:AB_398497
PE Rat anti-mouse CD31	BD Biosciences	Cat# 553373; RRID:AB_394819
PE-Cy7 Rat anti-mouse Ly6-A/E (Sca-1)	BD Biosciences	Cat# 558162; RRID:AB_647253
BV421 Rat anti-mouse CD140A (PDGFRa)	BD Biosciences	Cat# 562774; RRID:AB_2728781
PE Rat anti-mouse Integrin α7	R&D Systems	Cat# FAB3518P; RRID:AB_2249296
APC Rat IgG2a, κ Isotype Control	BD Biosciences	Cat# 551139; RRID: AB_10054921
PE Rat IgG2a, κ Isotype Control	BD Biosciences	Cat# 559317; RRID: AB_10050484
PE-Cy7 Rat IgG2a, κ Isotype Control	BD Biosciences	Cat# 552784; RRID: AB_394465
BV421 Rat IgG2a, k Isotype Control	BD Biosciences	Cat# 562602; RRID:AB_11153860
PerCP-Cy5.5 Rat IgG2b, κ Isotype Control	BD Biosciences	Cat# 550764; RRID: AB_393874
PE Rat IgG2b, k Isotype Control	BD Biosciences	Cat# 555848; RRID:AB_396171
Chemicals, peptides, and recombinant proteins
Mayer’s Hematoxylin solution	Millipore Sigma	Cat# MHS16
Eosin Y solution	Millipore Sigma	Cat# HT110116
4% paraformaldehyde solution	Thermo Scientific	Cat# J19943.K2
Poly Mount Xylene	Polysciences	Cat# 24716-120
Collagenase B	Roche Diagnostics	Cat# 11088831001
Dispase II	Roche Diagnostics	Cat# 04942078001
Fetal bovine serum	Corning	Cat# 35-011-CV
Bovine serum albumin	Millipore Sigma	Cat# A7030
Normal mouse serum	Millipore Sigma	Cat# S7273
Lympholyte-M solution	Cedarlane	Cat# CL5035
TRIzol reagent	Invitrogen	Cat# 15596026
RPMI 1640	Gibco	Cat# 11-875-093
RIPA buffer	Pierce Biotechnology	Cat# 89901
cOmplete (protease inhibitor)	Roche Diagnostics	Cat# 04.693.116.001
Bolt 4–12% Bis-Tris mini-gel	Invitrogen	Cat# NW04122
Immobilon PVDF membrane	Millipore	Cat#I SEQ00010
Acridine Orange and Propidium Iodide staining solution	Nexcelom	Cat# CS2-0106
Critical commercial assays
Vybrant™ DyeCycle™ Violet/SYTOX™ AADvanced™ Kit	Invitrogen	Cat# A35135
EdU Click 647 Kit + EdU	Baseclick GmbH	Cat# BCK647
Rneasy Micro Kit	Qiagen	Cat# 74004
Super-Script III Reverse Transcriptase Kit	Invitrogen	Cat# 18080051
PowerTrack™ SYBR Green Master Mix	Applied Biosystems	Cat# A46109
BCA protein assay Kit	Pierce Biotechnology	Cat# 23225
Single Cell 3′ Reagents Kits V3.1	10x Genomics	https://assets.ctfassets.net/an68im79xiti/1eX2FPdpeCgnCJtw4fj9Hx/7cb84edaa9eca04b607f9193162994de/CG000204_ChromiumNextGEMSingleCell3_v3.1_Rev_D.pdf
Deposited data
Single cell-based RNA sequencing data	https://www.ncbi.nlm.nih.gov/geo/	GEO:GSE218201
Experimental models: Organisms/strains
C57BL/6J mice	The Jackson Laboratory
mdx5cv mice	The Jackson Laboratory; Chapman VM, et al. PNAS 86(4): 1292–6.
Oligonucleotides
Gapdh (Forward): 5′-GTGTTCCTACCCCCAATGTGTC-3′	IDT DNA Oligos	N/A
Gapdh (Reverser): 5′-GTAGCCCAAGATGCCCTTCAGT-3′	IDT DNA Oligos	N/A
Col1a1 (Forward): 5′-GCTCCTCTTAGGGGCCACT-3′	IDT DNA Oligos	N/A
Col1a1 (Reverse): 5′-CCACGTCTCACCATTGGGG-3′	IDT DNA Oligos	N/A
Col3a1 (Forwad): 5′-AACCTGGTTTCTTCTCACCCTTC-3′	IDT DNA Oligos	N/A
Col3a1 (Reverse): 5′-ACTCATAGGACTGACCAAGGTGG-3′	IDT DNA Oligos	N/A
Fn1 (Forward): 5′-AAACTTGCATCTGGAGGCAAACCC-3′	IDT DNA Oligos	N/A
Fn1 (Reverse): 5′-AGCTCTGATCAGCATGGACCACTT-3′	IDT DNA Oligos	N/A
Igf1 (Forward): 5′-CTACAAAAGCAGCCCGCTCT-3′	IDT DNA Oligos	N/A
Igf1 (Reverse): 5′-CTTCTGAGTCTTGGGCATGTCA-3′	IDT DNA Oligos	N/A
Ccn2 (Forward): 5′-AGGACTGCAGCGCGCAATGT-3′	IDT DNA Oligos	N/A
Ccn2 (Reverse): 5′-GAGGCCCTTGTGTGGGTCGC-3′	IDT DNA Oligos	N/A
Mmp2 (Forward): 5′-GCCCCGAGACCGCTATGTCCACT-3′	IDT DNA Oligos	N/A
Mmp2 (Revers): 5′-GCCCCACTTCCGGTCATCATCGTA-3′	IDT DNA Oligos	N/A
Timp1 (Forward): 5′-ACTCGGACCTGGTCATAAGGGC-3′	IDT DNA Oligos	N/A
Timp1 (Reveres): 5′-TTCCGTGGCAGGCAAGCAAAGT-3′	IDT DNA Oligos	N/A
Timp2 (Forward): 5′-GGCAACCCCATCAAGAGGA-3′	IDT DNA Oligos	N/A
Timp2 (Reverse): 5′-CCTTCTGCCTTTCCTGCAATTAG-3′	IDT DNA Oligos	N/A
Mmp14 (Forward): 5′-CTTCAAAGGAGATAAGCACTG-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Mmp14 (Reverse): 5′-AGAAGTAGGTCTTCCCATTG-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Ltbp4 (Forward): 5′- CCCATTCTTCGAAATATCACC-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Ltbp4 (Reverse): 5′- GAAAACCCTCTGAACCATAAG-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Lox (Forward): 5′-CACCGTATTAGAAAGAAGCC-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Lox (Reverse): 5′-GTCCTTCCTACTTAAGCTAATC-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Loxl1 (Forward): 5′-TGCTATGACACATACAATGC-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Loxl1 (Reverse): 5′-GAACAATGTACTTGGGGTTC-3′	Millipore Sigma	KiCqStart® SYBR® Green Primers: Cat# KSPQ12012
Software and algorithms
GraphPad Prism 8	GraphPad	https://www.graphpad.com/scientific-software/prism/
FlowJo V10	BD Biosciences	https://www.bdbiosciences.com/en-us/products/software/flowjo-v10-software
Image Studio software	LI-COR Biosciences	https://www.licor.com/bio/image-studio/
Image J	NIH	https://imagej.nih.gov/ij/
R package Seurat version 4.0	Satija Lab at NYGC	https://satijalab.org/seurat/
g:Profiler		https://biit.cs.ut.ee/gprofiler/gost
Monocle3		https://cole-trapnell-lab.github.io/monocle3/

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xingyu Wang (wangx20@bu.edu).

Materials availability

This study did not generate unique new reagents or mouse lines.

Experimental model and subject details

Animals

C57BL/6J (WT) and mdx^5cv mice (B6 background) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Only males were used in this study because DMD only affects males. For scRNAseq analysis, WT and mdx^5cv mice used were at 14 weeks of ages with body weight between 25 and 30 grams. For FACS analysis determining FAP density, an additional time point at 24 weeks of age was used for mdx^5cv mice with body weight between 30 and 35 grams. Our study protocols were approved by the Institutional Animal Care and Use Committee at the Boston University School of Medicine (Boston, MA, USA.).

Method details

Muscle sample collection

Quadriceps and diaphragm were collected from 14-week-old male WT mice, and 14-week-old and 24-week-old male mdx^5cv mice. Tendons were removed from the tissue specimens. Collected muscle samples were thoroughly washed in 1× phosphate-buffered saline (PBS, pH 7.4) to remove blood contamination.

Histopathological analysis

Collected muscle samples were frozen in liquid nitrogen-chilled isopentane for 30 seconds. Cross sections were then obtained on a Leica cryostat at a thickness of 10 μm. Sections were air-dried for 30 minutes at room temperature and then fixed in 4% (w/v) of paraformaldehyde (Thermo Scientific) for 10 minutes followed by washing in 1× PBS for 5 minutes. Hematoxylin and Eosin (H&E) staining was then performed as following: 1) stained in Mayer’s Hematoxylin solution (Millipore Sigma, Burlington, MA, USA.) for 10 minutes, 2) rinsed in warm running tap water for 15 minutes, 3) washed in distilled water for 30 seconds, 4) placed in 95% (v/v) reagent alcohol for 30 seconds, 5) stained in Eosin Y solution (Alcoholic, Millipore Sigma) for 1 minute, 6) dehydrated and cleared through 2 changes each of 95% reagent alcohol, absolute reagent alcohol, and xylene for 2 minutes each, 7) mounted with Poly Mount Xylene (Polysciences, Warrington, PA, USA). Stained sections were viewed and photographed under a bright-field microscope (Nikon, Y-THR-L).

Muscle single-cell suspension preparation

Single-cell suspension was prepared by collagenase/dispase digestion.⁸³ Quadriceps and diaphragm samples (less than 250 mg) were minced in 2.5 mL digestion solution (1 U/ml collagenase B and 1 U/ml dispase II (Roche Diagnostics, Indianapolis, IN, USA) in 1× PBS) and incubated at 37°C for 1 hour. The reaction was terminated by adding 10 mL PBS containing 10% (v/v) fetal bovine serum (FBS, Corning, Glendale, AZ, USA). The mixture was then filtered through a 70-μm cell strainer and centrifuged at 250 g for 5 minutes. The pellet was collected, and the supernatant was centrifuged again at 250 g for 5 minutes at 4°C. The pellet was combined with the pellet from the first centrifugation, washed with PBS, and centrifuged at 670 g for 10 minutes at 4°C. The pellet was re-suspended in 3 mL PBS and filtered through a 40-μm cell strainer. Cell suspension was layered on equal volume of the Lympholyte-M solution (Cedarlane, Burlington, NC, USA) and centrifuged at 2,100 g for 45 minutes at room temperature. Cells at the interface were collected in 10 mL PBS containing 10% FBS, pelleted by centrifugation at 670 g for 10 minutes at 4°C, and re-suspended in appropriated buffer for following analysis.

Flow cytometry analysis and cell sorting

Single-cell pellets were washed once in ice-cold FACS buffer (1× PBS, 0.5% (w/v) bovine serum albumin (BSA, Millipore Sigma), 0.1% (w/v) sodium azide), centrifuged at 500 g for 3 minutes at 4°C, and re-suspended in ice-cold FACS staining buffer (1× PBS, 2% (w/v) BSA, 2% (v/v) normal mouse serum (Millipore Sigma).

For FACS analysis of different cell types, cells were stained on ice for 20 minutes with fluorescence-labelled antibodies targeting cell-type specific markers: Fibro/Adipogenic progenitors (FAPs) were identified as Sca-1⁺PDGFRα⁺CD45⁻CD31^-α7-integrin^- cells, endothelial cells (ECs) were identified as CD31⁺CD45⁻ cells, and muscle satellite cells (MuSCs) were identified as α7-integrin⁺CD45⁻CD31⁻Sca-1^- cells. Fluorescence-labelled corresponding normal IgG isotypes were included as negative controls for gating. All antibodies were from commercial providers (See key resources table for details) and diluted for staining following manufacturer’s recommendations. After staining, cells were washed twice with FACS buffer and analyzed on a LSR II (BD Bioscience, San Jose, CA, USA) with BD FACSDiva™ software. Collected data were then analyzed using FlowJo software (Tree Star, Inc., Ashland, OR, USA).

To clean up single cell suspension for subsequent single-cell based analysis, cell sorting based on FSC/SSC was performed to exclude dead cells, cell duplexes, and tissue debris. Sorting FAPs and non-FAPs for qRT-PCR analysis were performed with antibodies listed above. Cell sorting was done by the Flow Cytometry Core of Boston University School of Medicine. Collected cells from 5 individual male mice were combined, and pelleted by centrifuging at 300 g for 10 minutes. For scRNAseq analysis, pellets were re-suspended in FBS supplemented with 10% of DMSO, and cryopreserved in liquid nitrogen. For qRT-PCR analysis, pellets were dissolved in TRIzol reagent (Invitrogen, Waltham, MA, USA) for RNA preparation.

FAP apoptosis assay

Muscle single-cell suspension was first stained with antibodies identifying FAPs as described in the section of flow cytometry analysis and cell sorting. After being washed twice with FACS buffer, cells were again re-suspended in FACS buffer and stained with Vybrant™ DyeCycle™ Violet/SYTOX™ AADvanced™ reagents (Invitrogen) following manufacturer’s instruction. Stained cells were analyzed by flow cytometry immediately. Apoptotic FAPs were identified as Vybrant⁺SYTOX⁻.

FAP proliferation assay

Mdx^5cv mice at 14 weeks of age received intraperitoneal injection of EdU (100 mg/mouse, Baseclick GmbH, Germany). Quadriceps and diaphragm were then collected 24 hours later and single-cell suspension was prepared as described in the section of muscle single-cell suspension preparation. FAP proliferation was then analyzed by determining the incorporation of EdU using EdU Click 647 Kit (Baseclick GmbH, Germany) following manufacturer’s instruction, combined with FACS staining and analysis of FAPs as described in the section of flow cytometry analysis and cell sorting.

RNA preparation and qRT-PCR

Total RNA samples were prepared using TRIzol reagent (Invitrogen). Samples of FAPs and non-FAPs were prepared by cell sorting as stated above and dissolved in TRIzol reagent. To prepare whole-muscle RNA, quadriceps and diaphragm, collected from 14-week-old male WT and mdx^5cv mice, were homogenized in TRIzol reagent using Bead Mill 24 homogenizer (Fisher Scientific). Total RNA was then prepared following manufacturer’s instruction. TRIzol-prepared total RNA was further cleaned up using RNeasy Micro Kit (Qiagen, Germantown, MD, USA) following manufacturer’s instruction. Concentration of cleaned RNA was measured using NanoDrop 2000 Spectrophotometer (Thermo Scientific). 1 μg of total RNA was then reverse-transcribed into cDNA using Super-Script III Reverse Transcriptase system (Invitrogen) following manufacturer’s instruction. Expression of genes at mRNA level was then analyzed by real-time PCR with PowerTrack™ SYBR Green Master Mix (Applied Biosystems, Waltham, MA, USA) using StepOne™ Real Time PCR System (Applied Biosystems). Sequences and providers of the primers used are listed in key resources table. The sequences and target genes of the primers were listed in key resources table. Data was calculated by ΔΔCt method. For sorted cell analysis, data was presented as bar graph with error bar showing standard deviation of triplicated PCR results. For whole muscle analysis, data were presented as scatter plot showing results of each individual samples.

Western blot

Quadriceps and diaphragm, collected from 14-week-old male WT and mdx^5cv mice, were homogenized in RIPA buffer (Pierce Biotechnology, Waltham, MA, USA) supplemented with cOmplete protease inhibitor cocktail (Roche Diagnostic). After centrifugation (10,000 ×g for 15 minutes), the protein concentration was quantitated by Pierce BCA protein assay kit (Pierce Biotechnology). 50 μg of protein lysate were run on 4–12% Bis-Tris mini-gel (Invitrogen) under reducing condition then transferred to PVDF membranes (Immobilon). Blots were blocked with 5% (w/v) of non-fat milk in TBS-T (1× Tris-buffered saline (pH7.6) with 0.1% (v/v) Tween 20) and then incubated with primary antibodies include α/β-tubulin (1:1000 dilution), collagen 1α1 (1:500), fibronectin (1:250) (all purchased from Cell Signaling Technology, Danvers, MA, USA), and collagen III (1:500) (Abcam, Waltham, MA, USA). Signals were detected by IRDye 800 or 680 (LI-COR Biosciences, Lincoln, NE)-conjugated secondary antibodies with appropriate species specificity. Immunoblots were visualized by the Image Studio software (LI-COR Biosciences) that accompanies the LI-COR Odyssey infrared system (LI-COR Biosciences).

Single-cell cDNA library preparation and sequencing

Cryopreserved cells were thawed in a 37°C water bath for two minutes. The thawed content was added dropwise into 9 mL of warm freshly prepared RPMI 1640 (Gibco) with 20% of FBS, spun 300 g for 5 minutes, and re-suspended in 5 mL of RPMI with10% of FBS. The cells were counted, and viability was assessed using Acridine Orange and Propidium Iodide staining followed by counting on a K2 Cellometer (Nexcelome, Lawrence, MA, USA). Next, scRNAseq was performed following the Single Cell 3′ Reagents Kits V3.1 User Guidelines (10x Genomics, Pleasanton, CA, USA). All samples were above 70% viability. Cells were loaded to target 5,000 cells final for generating the single-cell emulsion (Chromium Single Cell 3′ Chip kit A v2 PN-12036 or v3 chip kit B PN-2000060). Reverse-transcription (RT) was performed in the emulsion, cDNA amplified, and library constructed with v3.1 chemistry. Libraries were indexed for multiplexing (Chromium i7 Multiplex kit PN-12062). Following preparation, libraries were quantified by Qubit 3 fluorometer (Invitrogen) and quality was assessed by Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Equivalent molar concentrations of libraries were pooled, and the reads were adjusted after sequencing the pools in a Miseq (Illumina, San Diego, CA, USA). The libraries were then sequenced in a Novaseq 6000 (Illumina) at the New York Genome Center (NYGC, New York, NY, USA) following 10× Genomics recommendations.

Single cell-based mRNA sequencing analysis

The gene × cell expression matrices were loaded to the R package Seurat version 4.0 (https://cran.r-project.org/web/packages/Seurat/index.html) for downstream analyses.³¹ Cells of low quality, defined by the detectable expression of fewer than 200 genes per 1000 Unique Molecular Identifiers (UMI) and the presence of more than 7.5% mitochondria content, were filtered out.³¹ The Seurat default global-scaling normalization method LogNormalize was used to normalize the feature expression measurements for each cell by the total expression, multiplies this by a scale factor and log-transforms the result. The Seurat default global-scaling normalization method, “LogNormalize”, was used for scaling the data.⁸⁹ The method Vst was used to identify highly variable features in function FindVariableFeatures.

A shared-nearest neighbor (SNN) graph–based clustering method was constructed to identify discrete cell populations. Then the obtained clusters were optimized by using the “FindClusters” function.

Linear dimensional reduction PCA was performed on variable genes. A clear Elbow in the plot of the PC standard deviation was used to define the significant PCs. Based on the identified PCs above, non-linear dimensional reduction technique UMAP was performed.^89,90,91 Differentially Expressed Genes (DEGs) were calculated and obtained by the function of FindAllMarkers based on the Wilcoxon rank sum test.³¹ Top marker genes in each cluster shown in Heatmap were ordered by log fold-change of the average expression. Differential gene expression analysis was performed in each of the cell clusters, and the log transformed fold changes (log₂FC) were used to perform hierarchical clustering between genes.

Functional enrichment analysis

We used g:Profiler webserver (https://biit.cs.ut.ee/gprofiler/gost) to perform functional enrichment analysis for the identified DEGs in individual clusters or subclusters. Top DEGs (adj. p-value < 0.05) were tested for enrichment of biological pathways with Gene Ontology (GO), Reactome, CORUM and KEGG. Results from different databases were compared to ensure the reliability of the analysis.

Pseudotime trajectory analysis

The cells of FAP clusters were further analyzed using Monocle for pseudotime trajectory.⁹² Seurat objects were converted to CellDataSet objects using the as.cell_data_set function from SeuratWrappers. The trajectories were built using Monocle 3 (https://cole-trapnell-lab.github.io/monocle3/docs/citations/, https://satijalab.org/signac/articles/monocle.html). Each sample was analyzed separately by the same procedure. Interactive function order cells was used to select cells as the root nodes in the graph. The function get_earliest_principal_node was applied to identify the root principal points.

Venn diagram

Differential gene expression among FAPs of 4 muscles (WT diaphragm, mdx^5cv diaphragm, WT quadriceps, and mdx^5cv quadriceps) were determined by Seurat FindMarkers function, using a log fold-change threshold of 0.5 and an adjusted p value cutoff of 0.05. Venn diagram was prepared using Venny2.1.

Quantification and statistical analysis

Data were analyzed using GraphPad Prism 8 software (GraphPad Software, San Diego, CA, USA). The Mann-Whitney test was performed to compare between two groups; the Kruskal-Wallis test followed by Dunn’s test was performed to compare multiple (>=3) groups. A p value of <0.05 was considered statistically significant.

Published: January 20, 2023

Data and code availability

Single-cell based RNA sequencing (scRNAseq) data have been deposited to GEO repository (GEO: GSE218201) and are publicly available as of the date of publication. The accession number is also listed in the key resources table. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.