S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

C E L L B I O LO G Y Copyright © 2023
The Authors, some
Single-cell and spatial transcriptomics identify a rights reserved;
exclusive licensee
macrophage population associated with skeletal American Association
for the Advancement
muscle fibrosis of Science. No claim to
original U.S. Government
Works. Distributed
Gerald Coulis1,2, Diego Jaime1,2, Christian Guerrero-Juarez3, Jenna M. Kastenschmidt1,2, under a Creative
Philip K. Farahat1,2, Quy Nguyen4, Nicholas Pervolarakis4, Katherine McLinden5, Lauren Thurlow5, Commons Attribution
Saba Movahedi1, Brandon S. Hughes1, Jorge Duarte1, Andrew Sorn1, Elizabeth Montoya1, License 4.0 (CC BY).
Izza Mozaffar1, Morgan Dragan4, Shivashankar Othy1,2, Trupti Joshi6, Chetan P. Hans7,
Virginia Kimonis8, Adam L. MacLean9, Qing Nie10, Lindsay M. Wallace11, Scott Q. Harper11,12,
Tahseen Mozaffar13,14, Marshall W. Hogarth15, Surajit Bhattacharya15, Jyoti K. Jaiswal15,
David R. Golann16, Qi Su16, Kai Kessenbrock4, Michael Stec16, Melissa J. Spencer17,
Jesse R. Zamudio5, S. Armando Villalta1,2,13*

Macrophages are essential for skeletal muscle homeostasis, but how their dysregulation contributes to the de-
velopment of fibrosis in muscle disease remains unclear. Here, we used single-cell transcriptomics to determine
the molecular attributes of dystrophic and healthy muscle macrophages. We identified six clusters and unex-
pectedly found that none corresponded to traditional definitions of M1 or M2 macrophages. Rather, the pre-
dominant macrophage signature in dystrophic muscle was characterized by high expression of fibrotic factors,
galectin-3 (gal-3) and osteopontin (Spp1). Spatial transcriptomics, computational inferences of intercellular
communication, and in vitro assays indicated that macrophage-derived Spp1 regulates stromal progenitor diff-
erentiation. Gal-3+ macrophages were chronically activated in dystrophic muscle, and adoptive transfer assays
showed that the gal-3+ phenotype was the dominant molecular program induced within the dystrophic milieu.
Gal-3+ macrophages were also elevated in multiple human myopathies. These studies advance our understand-
ing of macrophages in muscular dystrophy by defining their transcriptional programs and reveal Spp1 as a major
regulator of macrophage and stromal progenitor interactions.

INTRODUCTION functional heterogeneity of skeletal muscle macrophages (5). In

Macrophages have a central role in innate immunity and contribute acute muscle trauma, proinflammatory M1 macrophages initially
to tissue homeostasis by regulating tissue repair and remodeling of infiltrate injured muscle to phagocytose cellular debris and activate
the extracellular matrix (ECM) (1). Their diverse functions within muscle stem cells (6). The subsequent transition to M2 macrophag-
the tissue are matched by a high degree of molecular heterogeneity, es in the regenerative phase promotes muscle stem cell differentia-
reflecting microenvironmental adaptability (2). Skeletal muscle tion and the resolution of inflammation (3, 7). M1- and M2-like
macrophages comprise resident and monocyte-derived popula- macrophages have been described in Duchenne muscular dystrophy
tions, the latter infiltrating upon muscle injury or disease (3). The (DMD) (7). However, growing evidence indicates that the regula-
M1 and M2 macrophage paradigm (4) has been used to describe the tion and functional role of macrophages are more complex in the
context of chronic muscle degenerative diseases such as DMD.
Although macrophages are essential for muscle repair following
1
Department of Physiology and Biophysics, University of California Irvine, Irvine, acute trauma, their dysregulation promotes the pathogenesis of
CA, USA. 2Institute for Immunology, University of California Irvine, Irvine, CA, DMD, a lethal form of muscular dystrophy caused by mutations
USA. 3Carle Illinois College of Medicine, University of Illinois at Urbana-Champaign,
Champaign, IL, USA. 4Department of Biological Chemistry, University of California
in the DMD gene (8). The transition from M1 to M2 macrophages
Irvine, Irvine, CA USA. 5Department of Molecular Cell and Developmental Biology, seen in acute injury is disrupted in the X-linked muscular dystrophy
University of California Los Angeles, Los Angeles, CA, USA. 6Department of Health (mdx) mouse model of DMD by asynchronous bouts of muscle
Management and Informatics, University of Missouri, Columbia, MO, USA. injury and regeneration (9). Consequently, M1 macrophages are
7
Department of Cardiovascular Medicine, University of Missouri, Columbia, MO
USA. 8Department of Pediatrics, University of California Irvine, Irvine, CA, USA. chronically activated and promote muscle injury in an inducible
9
Department of Quantitative and Computational Biology, University of Southern nitric oxide synthase–dependent manner (10), and the reparative
California, Los Angeles, CA, USA. 10Department of Mathematics, Department of De- function of M2 macrophages is pathologically repurposed to
velopmental and Cell Biology, University of California Irvine, Irvine, CA, USA.
11
Center for Gene Therapy, The Abigail Wexner Research Institute at Nationwide promote fibrosis. This impairment is expected to induce macro-
Children’s Hospital, Columbus, OH, USA. 12Department of Pediatrics, The Ohio phage transcriptional programs that are distinct from those
State University, Columbus, OH, USA. 13Department of Neurology, University of induced in acute injury and contribute to muscle pathology. In
California Irvine, Irvine, CA, USA. 14Department of Pathology and Laboratory Med-
icine, University of California Irvine, Irvine, CA, USA. 15Children’s National Hospital,
support of this, several studies reported that perturbing macrophage
Research Center for Genetic Medicine, Washington, DC, USA. 16Regeneron Phar- function and/or activation impaired regeneration and promoted fi-
maceuticals Inc., Tarrytown, NY, USA. 17Department of Neurology, University of Cal- brosis in muscular dystrophy (11–13), Pompe disease (14), and dys-
ifornia Los Angeles, Los Angeles, CA, USA. ferlinopathy (15).
*Corresponding author. Email: [email protected]

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 1 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fibrosis is the aberrant accumulation of collagen and other ECM acute stage of disease (4 weeks of age) and age-matched, C57BL/
proteins in chronically inflamed tissues, leading to organ failure and 10 wild-type (WT) controls (Fig. 1A). This stage was selected
death (16). Fibro/adipogenic progenitors (FAPs) are stromal cells because muscle macrophage numbers are most elevated and their
that give rise to fibroblasts and adipocytes and regulate muscle depletion reduces muscle necrosis by ~80% in mdx mice (26), sug-
repair and fibrosis (17). Following acute injury, FAPs expand and gesting that a peak inflammatory state is achieved by this age. FACS
contribute to muscle repair by facilitating myogenesis and ECM for- yielded macrophage samples with greater than 92 to 96% purity.
mation (18). Fibrosis is mitigated by infiltrating macrophages that Uniform Manifold Approximation and Projection (UMAP) for
clear FAPs through tumor necrosis factor–α–mediated apoptosis dimensionality reduction of 11,367 single-cell profiles (WT = 5723;
(19). Conversely, transforming growth factor–β (TGF-β), which is mdx = 5644) was used to examine the data. The resulting analysis
highly up-regulated in dystrophic muscle (20), inhibits FAP apopto- partitioned single-cell profiles into eight clusters composed primar-
sis and guides their differentiation into matrix-producing myofi- ily of macrophages and a low proportion of contaminating cell
broblasts (17). In this setting, persistence of FAPs and their types, including endothelial cells, FAPs, satellite cells, and muscle-
skewed differentiation toward fibroblasts contributes to the devel- like cells (fig. S1, A and B). Clusters 0, 1, 2, and 3 made up more than
opment of muscle fibrosis. Unexpectedly, inhibition of TGF-β 90% of the single-cell profiles and were assigned a macrophage
does not fully restore FAP clearance (19), indicating that there are identity based on their expression of macrophage markers (fig.
additional factors that promote FAP differentiation and/or survival. S1C). We filtered out the contaminating cell types and reclustered
Osteopontin (Spp1) is a potential candidate because it is highly ex- the single-cell profiles corresponding to clusters 0 to 3 in fig. S1A,
pressed by macrophages (21), elevated in dystrophic muscle in pa- which resulted in the identification of six macrophage clusters
tients with DMD (22), and promotes fibrosis in muscular dystrophy (fig. S1D).
(22, 23). However, the repertoire of profibrotic factors, including The proportion of cluster 0 to 5 macrophages differed among
Spp1, expressed by dystrophic muscle macrophages and their cellu- healthy and dystrophic muscle (Fig. 1, B and C). Cluster 1 macro-
lar targets has not been fully defined. phages were almost exclusively present in healthy muscle, whereas
In this study, we used an unbiased single-cell RNA sequencing clusters 0 and 2 were predominantly found in dystrophic muscle.
(scRNAseq) approach to define the transcriptional profiles of mac- Clusters 3 to 5 were shared between healthy and dystrophic
rophages from normal and dystrophic muscle. The scRNAseq iden- muscle. Vast differences in transcriptional profiles were noted
tified several macrophage populations with transcriptomes not between macrophages isolated from dystrophic (red) and WT,
previously associated with muscular dystrophy. We focused on healthy muscle (blue) (Fig. 1D). Hierarchical cluster analysis of dif-
three populations that corresponded to resident macrophages, ferentially expressed genes (DEGs) revealed that each macrophage
monocyte-derived macrophages (MDMs), and a population charac- population expressed distinct transcriptional modules containing
terized by high expression of profibrotic factors, Lgals3 [galectin-3 cluster-specific genes with biomarker potential (Fig. 1E).
(gal-3)] (24) and Spp1 (25). Given the selective induction of Lgals3 Cluster 1 macrophages were characterized by increased expres-
and Spp1 in dystrophic muscle macrophages, we hypothesize that sion of Folr2, Mt2, Lyve1, Gas6, and Cbr2. This signature shared
this transcriptional profile defines a fibrogenic macrophage that similarities with a subset of skeletal muscle–resident macrophages
promotes fibrosis during muscular dystrophy. (SkMRMs) previously described in healthy muscle (27). Hereafter,
We demonstrate that gal-3+ macrophages are activated in re- cluster 1 macrophages are referred to as SkMRMs. Cluster 0 macro-
sponse to acute injury and are elevated in several muscle disorders. phages (hereafter referred to as gal-3+ macrophages) expressed high
Although this macrophage population is transient following acute levels of Spp1, Fabp5, Gpnmb, Trem2, Lgals3, and various cathepsin
injury, gal-3+ macrophages are chronically activated during muscu- genes. Lgals3 (gal-3) has been implicated in the development of fi-
lar dystrophy. Spatial transcriptomic analysis of dystrophic muscle brosis (28, 29), suggesting that gal-3+ macrophages promote fibrosis
revealed that areas enriched in gal-3+ macrophages and stromal cells during muscular dystrophy. In support of this, Spp1 (osteopontin)
expressed genes associated with muscle fibrosis. Furthermore, gal- also promotes muscle fibrosis in mdx mice (22) through a matrix
3+ macrophages colocalize with stromal cells in dystrophic lesions, metalloproteinase-mediated processing of TGF-β in stromal
and computational analysis with CellChat shows that Spp1 mediates cells (30).
communication between these cell types. Collectively, results from Cluster 2 macrophages, hereafter referred to as MDMs, were
this study identify a distinct transcriptional profile in dystrophic marked by Cd52, Plac8, Prdx5, and Hp. F4/80 (Adgre1), which is
muscle macrophages that is associated with fibrosis, revealing the lowly expressed in blood monocytes (fig. S2A) (31), was expressed
potential for therapeutic strategies that target fibrogenic macro- lower in MDMs relative to SkMRMs or gal-3+ macrophages (fig.
phages in muscular dystrophy. S2B). Ly6c2 and Cd52, which are highly expressed in blood mono-
cytes (fig. S2, C and E), were also highly expressed in MDMs, clus-
ters 3 and 5 (fig. S2, D and F). Flow cytometry analysis of healthy
RESULTS and dystrophic muscle macrophages confirmed these observations
scRNAseq reveals unique cell states within skeletal muscle (fig. S2, G to J). Further, MDMs expressed high levels of Ccr2 but
macrophages low levels of Cx3cr1 (fig. S2, K and L). Collectively, these findings
scRNAseq was performed to unbiasedly phenotype skeletal muscle suggest that cluster 2 is a monocyte-derived population that resem-
macrophage transcriptomes at single-cell resolution. We used a bles Ly6chiCCR2+ inflammatory monocytes.
droplet-based scRNAseq platform (Chromium, 10xGenomics) to Cluster 3 was defined by high expression of Cd74, major histo-
profile muscle macrophages (live F4/80+CD11b+Siglec-F− cells) pu- compatibility complex II genes (H2-Eb1, H2-DMb2, H2-Ab1, H2-
rified by fluorescence-activated cell sorting (FACS) from C57BL/ Aa, H2-DMb1, and H2-DMa) and the dendritic cells marker
10ScSn-Dmdmdx/J mice (B10.mdx) hindlimb muscle during the Itagx, suggesting that this population corresponds to dendritic

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 2 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fig. 1. Identification of transcriptomic diversity in skeletal muscle macrophages by scRNAseq. (A) Muscle macrophages were isolated from 4-week-old WT (healthy)
and B10.mdx (dystrophic) mice and analyzed by scRNAseq. n = 1 (B) Dimensionality reduction via UMAP of healthy or dystrophic muscle macrophages. (C) Proportion of
WT or mdx muscle macrophage clusters. (D) Identities classified by genotype. (E) Differential gene expression analysis showing the top 10 most differentially expressed
genes (DEGs) for each cluster.

cells (Fig. 1E and fig. S2M). Cluster 4 macrophages expressed genes healthy tissues, Folr2+ macrophages were most abundant in skeletal
associated with chromatin, nucleosome (H2afv, Top2a, Hmgb2, and muscle (72% ±2.5) followed by the heart (30% ±3.4) (fig. S4) but
Hmgn2), and cell cycle (Birc5, Cks1b, and Cdca3) regulation. Cluster nearly absent in the brain, bone marrow, and blood. This expression
5 defined a macrophage population characterized by the high ex- pattern expands on the observations made in an earlier study exam-
pression of genes associated with interferon signaling (Irf7, Ifi47, ining Folr2 in tissue-resident macrophages (32). A gal-3hiFolr2lo
Ifit3, ifit2, and Isg20). macrophage population was also identified in dystrophic muscle
that corresponded to gal-3+ macrophages in the scRNAseq (fig.
Isolation and bulk transcriptome profiling of SkMRMs, S3, A and B), which was largely absent in several healthy tissues
MDMs, and gal-3+ muscle macrophages by bulk RNAseq (fig. S4). CD52 was most highly expressed in gal-3−Folr2− (WT)
A flow cytometry panel was developed to validate the scRNAseq or gal-3loFolr2− (mdx) macrophages and marked a population
macrophage populations that were predominantly associated with that likely corresponded to the MDMs in the scRNAseq analysis
either homeostasis or dystrophinopathy. Gal-3, Folr2, and CD52, (fig. S3, A and B). CD52+ monocytes or macrophages were
which were preferentially expressed by clusters 0, 1 and 2, respec- present in all WT tissues examined with the highest proportion in
tively, were used as markers to distinguish gal-3+ macrophages, the blood, lung, liver, and adipose tissue (fig. S4). Collectively, the
SkMRMs and MDMs (fig. S3, A and B). Macrophages in WT flow cytometry panel discriminated three muscle macrophage pop-
muscle did not express gal-3 but highly expressed Folr2, similar to ulations based on gal-3, Folr2, and CD52 expression.
SkMRMs identified in the scRNAseq analysis (fig. S3, A and B). In

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 3 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

To further link the flow cytometry macrophage populations to marker genes (fig. S3F). These observations suggest that the gal-3+
the scRNAseq profiles and establish their full transcriptomes, we population is a terminal transition state of either resident or MDMs.
performed bulk RNAseq on FACS-sorted populations. The macro- To further probe the macrophage transcriptomes, we performed
phage populations were sorted from 4-week-old WT (SkMRM) and principal components analysis (PCA) and classified DEGs. Using
dystrophic hindlimb muscle (gal-3hiFolr2lo and MDMs). The top the SkMRM dataset and publicly available microglia and bone
100 scRNAseq differentially expressed genes (scDEGs) from each marrow–derived macrophage (BMDMϕ) datasets (33, 34), the anal-
cluster were used to match transcriptomes to the sorted popula- ysis showed that ECM and development genes were enriched in
tions. The SkMRM, gal-3+, and MDM scDEGs nearly uniformly SkMRMs, suggesting a role in muscle homeostasis (fig. S5). The
distinguished the three transcriptomes and confirmed their match PCA was next used to classify the variation between the sorted
to the single-cell populations (Fig. 2, A to C, and fig. S3, C to E). In gal-3+ macrophages, MDMs, and SkMRMs (Fig 2D). The largest
addition, the gal-3+ scDEGs indicated partial overlapping profiles variance (PC1) separated normal and dystrophic macrophages
with the MDM scDEGs suggesting a cell state transition (Fig. 2B). and contained genes enriched in the innate immune response
Consistent with this, the gal-3+ population down-regulated resident (fig. S3G). The second largest (PC2) variation was enriched for
(Folr2, Gas6, Cbr2, and Lyve1) and MDM (Cd52, Ccr2, and Ly6c2) genes in pathways associated with the lysosome and lipid

Fig. 2. Muscle macrophages express transcriptomes distinct from M1 and M2 macrophages. (A to C) Heatmap showing the expression of the top 100 scRNAseq
DEGs from the scRNAseq analysis (scDEGs) in FACS-sorted SkMRM (A), gal-3+ Mϕ (B), and MDM (C). n = 3 per population. (D) PCA applied to the FACS-sorted macrophage
populations in (A) to (C). (E) Pathway analysis of top gene sets enriched in FACS-sorted gal-3+ Mϕ compared to SkMRM. (F) Pairwise comparison of gene expression
between FACS-sorted gal-3+ Mϕ and SkMRM. Colored points indicate DEGs from ECM-related gene sets. Lgals3 and Spp1 are highlighted by arrows. (G) Venn diagrams
of up-regulated and down-regulated DEGs in FACS-sorted gal-3+ Mϕ compared to SkMRM (Gal-3+/SkMRM) and MDMs compared to SkMRMs (MDMs/SkMRMs). (H) Cor-
relation analysis of DEGs from Gal-3+/SkMRM or MDM/SkMRM comparisons with M1 or M2 polarized Mϕ. (I) Expression of M1 and M2 macrophage markers in gal-3+ Mϕ
(0), SkMRMs (1), and MDMs (2) from scRNAseq data. (J) Violin plots of M2 markers in gal-3+ Mϕ, SkMRMs, and MDMs from the scRNAseq data. AU, arbitrary units.

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 4 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

metabolism, suggesting that gal-3+ macrophages exhibit enhanced A differential gene expression analysis between gal-3hi and gal-
lo
phagocytosis of muscle debris and lipid membranes (fig. S3H). By 3 spots revealed that 1365 and 125 genes were up- and down-reg-
directly comparing DEGs between the gal-3+ macrophages and ulated, respectively (Fig. 3C). The provided heatmap comprises 569
SkMRMs, ECM and inflammation-related pathways were further gal-3hi and 676 gal-3lo spots collected from five D2-mdx gastrocne-
identified as highly altered by the dystrophic environment (Fig. mius/plantaris muscle complexes. We assigned any spot with a
2E). We determined 959 and 2250 up-regulated and down-regulat- unique molecular identifier (UMI) count ≥ 3 as gal-3hi and spots
ed genes, respectively, in the gal-3+ macrophages compared to with a UMI count ≤ 1 as gal-3lo. A gene ontology (GO) analysis
SkMRMs (fig. S3I and table S1). Consistent with marker gene up- was performed on the DEGs, which showed that phagocytosis, en-
regulation, Lgals3 and Spp1 were among the highest up-regulated docytosis, and leukocyte migration were among the most enriched
and expressed genes in the gal-3+ state (Fig. 2F). terms (fig. S7A). Terms associated with regeneration and repair
The common transcriptional response of gal-3+ macrophages were also highly enriched in the gal-3hi areas (fig. S7B). Of particular
and MDMs to the dystrophic environment was indicated by an interest, GO terms associated with the ECM and fibroblasts (Fig. 3,
overlap of 727 up-regulated (41% of total) and 1442 down-regulated D and E) and multiple genes associated with fibrosis were enriched
(36% of total) DEGs compared to SkMRMs (Fig. 2G, P values > 1 × in the gal-3hi areas (Fig. 3F). Lgals3 and Spp1 were among the
10−16). Pathways enriched within the common DEGs indicated up- highest, revealing a greater than 30- and 25-fold increase, respec-
regulated genes involved in leukocyte activation and the inflamma- tively. The profibrotic matrix metalloproteinase-12 (Mmp12) was
tory response and down-regulated genes in glycosaminoglycan and also increased (35). Genes encoding ECM components were in-
ECM-related pathways (fig. S3J). Gal-3+ macrophages and MDMs creased, including fibronectin (Fn1), periostin (Postn), and several
also expressed unique DEGs, indicating specialized functional collagens, as well as components of growth factor pathways that
states for distinct dystrophic muscle macrophage populations. The induce fibrosis (e.g., Tgfβ and Pdgf ).
analysis of enriched pathways in the unique DEGs again indicated The spatial transcriptomics also revealed that platelet-derived
ECM and lysosomal gene up-regulation in the gal-3+ state com- growth factor receptor–α (PDGFRα), a growth factor receptor ex-
pared to the MDMs, potentially reflecting enhanced fibrotic and pressed on stromal cells, was increased in the gal-3+ areas (Fig.
phagocytic activity (fig. S3, K and L). 3G, highlighted in red). Immunofluorescence assays performed
To further characterize the inflammatory states of the dystrophic on mdx quadriceps revealed that PDGFRα+ stromal cells in patho-
macrophages, we compared them to well-defined M1 and M2 po- logical lesions were juxtaposed with gal-3+ macrophages (Fig. 3H).
larization signatures established by in vitro treatments of BMDMϕ Collectively, these findings suggest that gal-3+ macrophages interact
(34). The transcriptional profiles substantially differed between with stromal cells (e.g., FAPs) in degenerative lesions to promote
muscle and BMDMϕ macrophages (fig. S3M), and little to no cor- fibrosis.
relation was found between the MDM (C2/C1) or gal-3+ state (C0/
C1) and M1 or M2 polarization states (Fig. 2H). Further, several M1 Intercellular communication network analysis identifies
and M2 markers were heterogeneously expressed across the dystro- that FAPs and macrophages communicate via Spp1
phic macrophage populations in the scRNAseq analysis (Fig. 2I). To determine how gal-3+ macrophages and stromal progenitors in-
Unexpectedly, some M2 markers were expressed highest in teract, we performed a reference-based integration of skeletal
SkMRMs (Fig. 2J). Flow cytometry and quantitative polymerase muscle mononucleated cell datasets and surveyed intercellular com-
chain reaction (qPCR) analysis confirmed that M1 and M2 genes munication networks with CellChat (36). We used a publicly avail-
were not selectively enriched in any of the scRNAseq-defined able dataset of uninjured muscle (37) and referenced an additional
muscle macrophage populations (fig. S6). The transcriptomic pro- dataset independently prepared from muscle mononucleated cells
filing of muscle macrophages supports that a terminal differentia- pooled from three mdx mice. Three subpopulations of FAPs, in-
tion state marked by high gal-3 expression is a dominant signature cluding adipogenic, proremodeling, and stem populations, were
induced in muscular dystrophy. This gal-3 signature is associated identified, as reported previously (38), in uninjured and dystrophic
with regulation of the ECM and has little overlap with M1 and muscle. An increased proportion of proremodeling (53.2% versus
M2 macrophage activation signatures. 21.6%) and adipogenic FAPs (8.67% versus 5.70%) was noted in
dystrophic muscle, as well as an overall expansion of muscle mac-
Spatial transcriptomic analysis reveals that gal-3+ rophages (38.9% versus 14.4%), compared to healthy muscle (Fig.
macrophages are associated with stromal cells and 4A). Similar to the scRNAseq analysis of purified macrophages
ECM genes (Fig. 1), gal-3+ macrophages were the predominant macrophage
We used capture probe–based spatial transcriptomics (Visium, 10x population in dystrophic muscle (Fig. 4A). The probability of com-
Genomics) to gain insight on how gal-3+ macrophages functionally munication between two cell groups was visualized with circle plots
and spatially interfaced with the dystrophic environment. Spatial by designating FAPs as the central nodes of analysis. The highest
RNA sequencing was performed on the gastrocnemius/plantaris degree of inferred communication occurred with gal-3+ macro-
muscle complex of 6-week-old mdx mice in the DBA2/J back- phages, reflected by the thickness of the connecting edge (fig. S8,
ground (D2-mdx). This approach provides spatially resolved gene A to C). The CellChat analysis was also performed with gal-3+ mac-
expression analysis limited in resolution by 55-μm-diameter spot rophages, SkMRM, or MDMs as the central nodes of analysis. Circle
size. Interrogation of the spatial matrix revealed that areas with plots show a large probability of intermacrophage communication,
high gal-3 expression (Fig. 3A, highlighted in red) were confined followed by communication with FAP populations (Fig. 4, B to D).
to regions with active pathology that were densely populated by Little to no communication was found between macrophages and
mononuclear cells (Fig. 3B, highlighted in yellow). tenocytes.

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 5 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fig. 3. Spatial transcriptomics reveals that gal-3+ macrophages are associated with stromal cells and ECM. (A and B) Spatially resolved gene expression of Lgals3
(gal-3) (A) and hematoxylin and eosin (H&E) staining of D2-mdx quadriceps. (B) Shown is one of five representative D2-mdx quadriceps. (C) Heatmap showing DEGs
between Lgals3hi and Lgals3lo spots. Shown are genes with a fold change ≥ 1.5 and false discovery rate < 0.01. All spots in a section, including those with and without
pathology, were unbiasedly analyzed. (D and E) Gene ontology (GO)/pathway analysis showing the enrichment of GO terms in gal-3hi spots associated with collagen/ECM
(D) and fibroblasts (E). (F) Expression of DEGs associated with fibrosis. (G) Spatially resolved gene expression of Pdgfra in mdx quadriceps. (H) Immunofluorescence
staining of 4-week-old mdx quadriceps with anti-PDGFRα (white), anticollagen (red), and anti–gal-3 (green) antibodies. Scale bars, 50 μm. DAPI, 4′,6-diamidino-2-
phenylindole.

Next, information flow for signaling pathways in the stromal cell The Spp1 pathway was further characterized given its known role
and macrophage network was assessed (Fig. 4E). Some pathways in promoting muscle fibrosis during muscular dystrophy (22). The
(e.g., CXCL, CHEMERIN, BMP, SEMA3, ANGPTL, and PERIOS- highest relative contribution to the Spp1 pathway was attributed to
TIN) maintain similar flow in healthy and dystrophic muscle, sug- the Spp1-CD44 ligand (L)–receptor (R) pair, followed by several in-
gesting that these pathways are equally important in regulating tegrin heterodimer receptors (Fig. 4F). Flow cytometry showed that
macrophage and FAP interaction during homeostasis and muscle Spp1 receptors—CD44, Itgav, and Itgb3—were up-regulated in
disease. In contrast, other pathways prominently change their infor- mdx PDGFRα+Sca1+ stromal cells, compared to WT stromal cells
mation flow in dystrophic muscle compared to uninjured muscle. (Fig. 4, G to J). RNAscope multiplexed with immunofluorescence
For example, MK and growth differentiation factor (GDF) are si- staining revealed that degenerative lesions with increased macro-
lenced in dystrophic muscle, and TWEAK, PTN, and GAS are de- phages and collagen contained elevated levels of Spp1 (Fig. 4, K to
creased. In contrast, GALECTIN, CCL and COMPLEMENT are M). In vitro culturing of gal-3+ muscle macrophages revealed that
increased in dystrophic muscle. Insulin-like growth factor (IGF), they secreted higher levels of Spp1 compared to bone marrow
VISFATIN, OSM, ANNEXIN and SPP1 were exclusively active in monocytes (Fig. 4N). Monocyte- and gal-3+ macrophage–condi-
dystrophic muscle. tioned media induced the expression of Acta2 in FAPs in a Spp1-

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 6 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fig. 4. Spp1 mediates FAP and macrophage interactions in dystrophic muscle. (A) Reference-based integration of skeletal muscle mononucleated cell datasets
prepared from 3-month-old WT and mdx mice. (B to D) Visualization and analysis of cell-cell communication using CellChat. Circle plots placing macrophage subsets
as the central nodes of analysis in the mdx dataset (B) to (D). An interaction between a pair of cell types is depicted by a line connecting two cell types. The thickness of the
line depicts the strength of that interaction. (E) Pathways enriched in the stromal cell and macrophage network of WT and mdx mice. (F) Relative contribution of Spp1
ligand (L)–receptor (R) pairs. (G) Expression of Spp1 receptors was measured in WT and mdx PDGFRα+Sca1+ FAPs by flow cytometry. Plots shown were gated on live
CD45−CD31− cells. (H to J) Representative histograms and quantification of the mean fluorescence intensity (MFI) of Spp1 receptors. Four-week-old mice were analyzed. n
= 3 to 4. *P < 0.05 and ***P < 0.001 using an unpaired Welch’s t test. (K to M) RNAscope multiplexed with immunofluorescence staining of adjacent section showing F4/
80+ macrophages (K, green), Spp1 mRNA (M, yellow), and areas enriched with collagen (L, red). Laminin is shown in white (K) to (M). Scale bar, 100 μm. (N) Spp1 secretion
was assessed by enzyme-linked immunosorbent assay. (O and P) The expression of Acta2 (O) and Col1a1 (P) was measured by reverse transcription qPCR (RT-qPCR) in
FAPs stimulated with monocyte (mono)–conditioned or gal-3+ macrophage–conditioned media. α-SPP1, neutralizing mouse Spp1 antibody. Shown is a representative of
two independent experiments with conditions done in duplicate.
Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 7 of 17
S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

dependent manner, but only gal-3+ macrophage–conditioned activation of gal-3+ macrophages was associated with increased col-
media increased the expression of Col1a1 (Fig. 4, O and P). Collec- lagen deposition at 52 weeks (Fig. 5D).
tively, these results suggest that the Spp1 pathway is a key regulator The induction of a gal-3+ transcriptional program was conserved
of gal-3+ macrophage and FAP interactions during muscular dys- in other forms of muscle disease. We performed a referenced-based
trophy and that Spp1 signals primarily through CD44 and a integration of muscle macrophage scRNAseq datasets prepared
subset of integrin heterodimers to control FAP differentiation. from 8-month-old B6A/J mice, a mouse model of limb girdle mus-
cular dystrophy 2B, and respective controls (fig. S10A). The mdx
Muscle damage expands a population of gal-3+ dataset from this study (Fig. 1) was used as the reference. Similar
macrophages that is chronically activated in muscular to mdx dystrophic muscle, the predominant muscle macrophage
dystrophy population in B6A/J mice corresponded to gal-3+ macrophages
We next examined the regulation of muscle macrophage popula- (fig. S10B). Immunofluorescence staining of 12-month-old
tions during muscular dystrophy. SkMRM and MDMs were elevat- muscle confirmed the presence of gal-3+ macrophages and their
ed in the early stages of disease but resolved by chronic stages (fig. proximity to PDGFRα+ stromal cells in B6A/J mice (fig. S10D).
S9, A and B). An elevation of gal-3+ (gal-3hiFolr2lo) macrophages Gal-3+ macrophages were largely absent in B6 healthy control
was observed as early as 3.5 weeks of age in mdx hindlimb muscle (fig. S10C). SkMRMs were similarly the dominant macro-
muscles and began to decline by 8 weeks but remained chronically phage population in control muscle. Although the lack of MDMs
elevated up to 52 weeks of age compared to controls (Fig. 5A). Gal-3 and cluster 5 macrophages in B6A/J and control samples could be
protein was up-regulated in dystrophic muscle macrophages as explained by disease-specific differences, we cannot rule out that
early as 4 weeks and remained elevated at 52 weeks of age (Fig. 5, differences in the stage of disease (i.e., age) could contribute to
B and C). Immunofluorescence staining showed that gal-3 was ex- this observation. Gal-3+ muscle macrophages were also elevated
pressed in a subset of F4/80+ macrophages in dystrophic muscle but in mouse models of valosin-containing protein (VCP)–associated
absent in WT muscle macrophages (fig. S9C). The chronic

Fig. 5. Chronic activation of gal-3+ macrophages in dystrophic muscle. (A) The number of gal-3+ macrophages in B10.mdx hindlimb muscle, normalized to muscle
mass (grams). n = 6 to 19 per time point. (B and C) Representative histograms and quantitative analysis of the geometric MFI of gal-3 in 4-week-old (B) and 52-week-old (C)
WT and mdx muscle macrophages. (D) Picrosirius red staining of WT and mdx quadriceps cryosections from 4- and 52-week-old mice. Scale bar, 100 μm. (E and F)
Enumeration of gal-3+ macrophages in the VCP-associated inclusion body myopathy mouse model (E) and in the facioscapulohumeral muscular dystrophy Tamoxifen
inducible Cre-DUX4 (TIC-DUX4) mouse model (F). n = 4 to 6, 10-month-old mice (E); n = 5, 10-week-old mice (F). (G and H) Regulation of gal-3+ macrophages frequency (G)
and number (H) after injury. n = 7 to 9 per time point (G) and (H). *P < 0.05, **P < 0.01, and ****P < 0.0001 using an unpaired Welch’s t test (B), (C), (E), and (F) or two-way
ANOVA with Sidak’s multiple comparisons test, for comparison with 2-week time point (A) or with day 0 (G) and (H). ****P < 0.0001 using Sidak’s multiple comparisons
test, for comparison with age-matched controls of 4, 8, 12, and 52 weeks, respectively.

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 8 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

inclusion body myopathy and facioscapulohumeral muscular dys- expressed in a subset of CD68+ macrophages (Fig. 7A). An immu-
trophy (Fig. 5, E and F). nohistochemical examination of gal-3 showed that the number of
The regulation of gal-3+ macrophages following BaCl2-induced gal-3+ macrophages in interstitial or perivascular regions did not
acute injury was also examined. A large increase in the proportion differ between control and myopathic patients, except for an in-
and number of gal-3+ muscle macrophages occurred 5 days after crease in gal-3+ perivascular macrophages in limb girdle muscular
injury (Fig. 5, G and H). By day 9, gal-3+ macrophages began to con- dystrophy 2A (LGMD2A) (Fig. 7, B to D). However, the number of
tract and returned to their homeostatic levels by day 14. The expan- gal-3+ myofiber-invading macrophages was significantly elevated in
sion of gal-3+ macrophages at day 5 coincided with the transition of DMD, antisynthetase syndrome (ASS), and LGMD2A (Fig. 7, B and
muscle toward the repair phase and when collagen deposition was E). A nonsignificant trend for an increase in patients with inclusion
most apparent, suggesting that they are involved in remodeling of body myositis (IBM) was also noted. Consistent with the increase in
the ECM during muscle regeneration (fig. S11, A and B). gal-3+ macrophages in DMD and LGMD2A, the expression of SPP1
in whole muscle was elevated (Fig. 7F). RNA was not available for
Gal-3+ macrophages are derived from SkMRMs and IBM, necrotizing autoimmune myopathy (NAM), and ASS groups
peripheral monocytes to measure transcript levels of SPP1. Further, COL1A mRNA was
The capacity of SkMRMs and monocytes to differentiate into significantly elevated in DMD muscle, suggesting that a similar
gal-3hiFolr2lo macrophages was determined through adoptive gal-3+ macrophage and Spp1 pathway promotes fibrosis in DMD
transfer assays. SkMRMs or bone marrow monocytes from (Fig. 7, F and G).
CD45.1+ congenic WT mice were injected into the quadriceps of
4-week-old mdx.CD45.2+-recipient mice. Before the adoptive trans-
fer (day 0), bone marrow monocytes did not express Folr2 or gal-3 DISCUSSION
(Fig. 6A). CD45.1+-transferred monocytes up-regulated gal-3 and Macrophage diversity is regulated by the integration of tissue-in-
Folr2, leading to an increased proportion of cells that acquired the trinsic cues with inflammatory signals to induce unique transcrip-
gal-3hiFolr2lo phenotype as early as day 2 and remained elevated 7 tional programs. Single-cell transcriptomics has helped uncover an
days after transfer (Fig. 6, A and B). The adoptive transfer of unprecedented degree of macrophage heterogeneity attributed to
CD45.1+ SkMRMs from WT muscle resulted in a down-regulation the integration of these complex signals in several organs, including
of Folr2 and an up-regulation of gal-3, resembling the gal-3hiFolr2lo the heart, brain, and adipose tissue (42–44). Recently, comparative
phenotype by the second day after transfer (Fig. 6, C and D). single-cell transcriptomic studies of tissue-resident macrophages
CD45.1+ SkMRMs were substantially declined 7 days after transfer, isolated from distinct anatomical sites, including skeletal muscle,
suggesting that gal-3hiFolr2lo emerging from the SkMRM pool are demonstrated that muscle macrophages expressed a transcriptome
short-lived (Fig. 6D). Intriguingly, mdx gal-3+ macrophages that differs from those of lung and peritoneal macrophages (27).
transferred into healthy WT muscle did not revert to a SkMRM Transcriptomic analysis of muscle macrophages isolated from
phenotype, suggesting that they exhibit little plasticity in this acutely injured muscle has also shown that macrophages acquire
setting (Fig. 6, E and F). Despite their inability to resolve, gal-3+ distinct transcriptional programs adapted to promoting muscle re-
macrophages did not promote collagen deposition in WT muscle generation (45, 46). However, the diversity of muscle macrophages
or induce the expression of genes associated with muscle fibrosis and how their function and transcriptional states are influenced by
(fig. S12) (39–41). muscle disease in an immunocompetent setting have not
The regulation of SkMRMs, MDMs, and gal-3+ macrophage been defined.
marker genes in endogenous macrophages FACS-sorted from 4- Here, we performed an unbiased scRNAseq analysis of muscle
week-old WT and mdx muscle was assessed by reverse transcription macrophages from healthy and dystrophic muscle to define the
qPCR (RT-qPCR) (Fig. 6, H and I). The following marker genes transcriptional programs induced in macrophages that promote
were interrogated: spp1, Fabp5, and Gpnmb; Folr2, Mt2, and muscle fibrosis during muscular dystrophy. We found six distinct
Lyve1; and Plac8, Chil3, and Prdx5, because of their preferential ex- macrophage populations, none of which resembled polarized M1
pression in gal-3+ macrophages, SkMRMs, and MDMs, respectively or M2 macrophages. Rather, the six populations expressed
(fig. S13). Folr2hi macrophages isolated from dystrophic muscle varying degrees of M1 and M2 macrophage marker genes. The
(mdx SkMRM) began to lose expression of SkMRM genes (Fig. gal-3hiFolr2lo (gal-3+) signature emerged as the predominant tran-
6G) but gained expression of gal-3+ macrophage genes (Fig. 6I). scriptional state in dystrophic muscle macrophages. Gal-3+ macro-
Similarly, MDMs in dystrophic muscle up-regulated gal-3+ macro- phages expressed profibrotic factors, including Spp1 and gal-3, and
phage genes (Fig. 6I), whereas MDM genes were lowly expressed in were chronically activated in muscular dystrophy. A recent scRNA-
gal-3+ macrophages (Fig. 6H). This regulation of SkMRM, MDM, seq analysis of immunodeficient mdx mice (mdxNSG) described a
and gal-3+ macrophage genes is consistent with the adoptive trans- M2c-like macrophage population that was marked by high expres-
fer assays and supports the interpretation that SkMRMs and recruit- sion of Spp1, suggesting that adaptive immunity likely has a
ed CD52+ monocytes are activated and differentiate into gal- minimal role in the induction of the gal-3+/Spp1+ macrophage
3hiFolr2lo macrophages within the dystrophic niche. state (47). Gal-3+ macrophages interacted with stromal cells (e.g.,
FAPs) in regions enriched with genes associated with muscle fibro-
Gal-3+ macrophages are elevated in human sis, and computational inferences predicted that communication
diseased muscle between these cells is partly mediated by Spp1. Gal-3+ macrophages
To determine whether gal-3+ macrophages are also present in were also elevated in several human myopathies, including DMD,
human dystrophic muscle, we quantified their numbers in archived and were juxtaposed with PDGFRα+ stromal cells. Further, SPP1
muscle biopsies. Immunofluorescence assays showed that gal-3 was and collagen were up-regulated in DMD biopsies, suggesting that

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 9 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fig. 6. Peripheral monocytes and skeletal muscle-resident macrophages give rise to gal-3+ macrophages. (A and B) Adoptive transfer of monocytes into 4-week-
old mdx mice. Graphical abstract of the workflow and representative flow plots of monocytes before and after transfer (A). Frequency of the donor monocytes that
converted to gal-3+ macrophages at 2 and 7 days after transfer (B). n = 3 to 13 per group. (C and D) Adoptive transfer of SkMRMs into 4-week-old mdx mice. Schematic
of the workflow and representative flow plots (C). Frequency of SkMRMs that converted to gal-3+ macrophages (D). n = 3 to 5 per group. (E and F) Schematic of the transfer
of gal-3+ macrophages into healthy WT muscle (E) and their quantification (F). n = 3 to 7 per group. (G to I) RT-qPCR quantification of the expression of cluster 1 (E), 2 (F),
and 0 genes (G) in FACS-sorted SkMRMs from WT and mdx muscle, and gal-3+ Mϕ and MDMs from mdx muscle. ***P < 0.001, and ****P < 0.0001 using a one-way ANOVA
with Sidak’s multiple comparisons test (B) and (D) for comparison with the day 0 mean or an unpaired Welch’s t test (F). A two-way ANOVA with Sidak’s multiple com-
parisons test was used for the gene expression assays (G) to (I). IM, intramuscular; ns, not significant.

a Spp1-mediated gal-3+ macrophage and stromal cell interaction is mdx mice (10, 23). Spp1 and gal-3 are also associated with tissue
conserved in human to promote muscle fibrosis. fibrosis (48–52) and adipogenic differentiation of FAPs in skeletal
Consistent with a potential role for gal-3+ macrophage–derived muscle (53), suggesting a cooperative role for gal-3 and Spp1 in in-
Spp1 in the pathogenesis of muscular dystrophy, Spp1 enhances the ducing fatty fibrosis in DMD. In support of this, Spp1 and gal-3
matrix metalloproteinase 9–mediated processing of TGF-β into its induce fibroblast proliferation and differentiation of myofibroblasts
active form to promote muscle fibrosis in mdx mice (22, 23, 30). (54, 55). It is essential to note, however, that gal-3 may have a po-
Further, Spp1 skews macrophages toward a proinflammatory phe- tential role in repair as demonstrated by the emergence of gal-3+
notype, which promote muscle damage in acute stages of disease in macrophages in damaged tissues (56–58) and the impairment of

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 10 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Fig. 7. Gal-3+ macrophages are elevated in human chronic muscle disease. (A) Immunofluorescence staining of gal-3+ macrophages in human IBM muscle. CD68
(red), gal-3 (green), nuclei (blue). Scale bars, 100 μm. (B) Representative images of immunohistochemical staining of gal-3 in control and myopathic patients. Scale bars,
100 μm. (C to E) Quantification of gal-3+ cells in the interstitial space (C), the perivascular area (D), or infiltrating the myofiber (E). n = 3 to 8 frozen sections per patient type.
(F and G) Expression of human SPP1 (F) and COL1A (G) mRNA in control, DMD, and LGMD biopsies. n = 6 to 8 patients were used to measure RNA expression. *P < 0.05, **P
< 0.01, and ***P < 0.001 using a two-way ANOVA with Kruskal-Wallis multiple comparisons test (C) to (G).

tissue repair in gal-3–deficient mice (59, 60). Studies of barium and IL-13. Given that the cytokines that induce M2 activation (e.g.,
chloride–induced muscle injury or autologous muscle grafting in IL-4 and IL-13) are increased with muscle injury or disease and are
mice showed that gal-3 and Spp1 promote regeneration in acute set- typically low or absent in healthy muscle (39, 66), it is unlikely that
tings (61–63). These findings suggest that muscle regeneration and these factors promote the M2-like phenotype of SkMRMs. Rather,
fibrosis share common pathways of regulation, and a temporal-de- tissue-associated or metabolic signals are more likely factors to
pendent mechanism partly determines whether Spp1 and gal-3 induce this phenotype (67). Consistent with this interpretation,
promote regeneration or fibrosis. This perspective is consistent bulk RNAseq analysis revealed that SkMRMs were associated with
with the view that fibrosis, to some extent, reflects a dysregulated ECM and muscle-associated pathways, suggesting that these ho-
repair process (7, 64). While experimental approaches targeting a meostatic functions induce a molecular phenotype sharing some
single factor (e.g., Spp1 or gal-3) will advance our knowledge on features with M2 macrophages. Although the function of
how they individually contribute to muscle fibrosis, resetting the SkMRMs remains to be defined, a recent study reported a popula-
transcriptional program of gal-3+ macrophages to its homeostatic tion of muscle-resident macrophages marked by high expression of
state is likely a more effective therapeutic strategy for DMD. Lyve1 and Timd4 that promoted the clearance of acutely damaged
In this regard, the present study defines the homeostatic signa- muscle and promoted metabolic reprogramming of dystrophic
ture of SkMRMs and establishes a posteriori knowledge to further muscle (65). Given that Folr2hi SkMRMs also expressed high
investigate how this transcriptional state is regulated. The transcrip- levels of Lyve1 and Timd4, it is likely that these are similar popula-
tional profile of SkMRMs identified in this study overlapped with tions with conserved function.
resident macrophages identified by scRNAseq analysis of healthy A major advancement in this study was the observation that the
skeletal muscle (27, 65). Wang and colleagues described a resident dystrophic environment converts muscle-resident macrophages
population in the diaphragm that was characterized by high expres- and peripheral monocytes into gal-3+ macrophages. Prior studies
sion of Folr2, Lyve1, Ltc4s, Fxyd2, and Fcgrt, genes that were ex- implicated a role for recruited monocytes in the muscle pathology
pressed the highest in SkMRMs in the present study. of mdx mice at 12 weeks of age (68). Although Ly6chi monocytes
Unexpectedly, we found that markers associated with M2 macro- were reduced out to 6 months of age in CCR2-deficient dystrophic
phages were most highly expressed in SkMRMs. This observation mice, myonecrosis and fibrosis returned to control levels by this
is consistent with the high expression of M2 markers in the point (69), suggesting that other monocyte or macrophage popula-
CD209+ muscle–resident macrophages described by Wang et al. tions promote dystrophinopathy at later stages of disease. In this
(27). Although SkMRMs expressed some markers of M2 macro- regard, Zhou and colleagues concluded that Ly6clo monocytes,
phages, their overall transcriptional signature substantially differed which returned to control levels, were responsible for the lack of a
from M2 macrophages polarized in vitro with interleukin-4 (IL-4) sustained protective effect in CCR2-deficient dystrophic mice.

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 11 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Similar to Ly6c+ bone marrow monocytes, SkMRMs that were Spp1+ macrophages and FAP interactions, the present findings only
adoptively transferred into dystrophic muscle differentiated into provide a correlative association with fibrosis and require further
gal-3+ macrophages, an activation state with putative fibrogenic ac- study. We note that a potential role for gal-3+ macrophages in fibro-
tivity. We propose that the sustained differentiation of dystrophic sis is further supported by studies showing that macrophages
SkMRMs cooperates with the continuous recruitment of monocytes undergo a similar activation program in volumetric muscle
to promote fibrosis throughout the course of muscular dystrophy. loss (74).
In the absence of Ly6chi inflammatory monocyte recruitment, Prior studies have shown that macrophages can also inhibit fi-
SkMRMs and Ly6clo monocytes become the dominant populations brosis. Macrophage depletion using CD11b–diphtheria toxin recep-
that promote fibrosis by adopting the gal-3+ program. Our observa- tor (DTR) transgenic mice impairs clearance of vascular cell
tions that the transcriptional state of SkMRMs is associated with the adhesion molecule (VCAM) positive FAPs and exacerbates collagen
ECM (fig. S5) and this program is retained in gal-3+ macrophages deposition (75). Because CD11b is expressed by all myeloid cells
(fig. S3L), suggest that, provided the appropriate niche, SkMRMs (76), these findings suggest that an undefined myeloid subset, dis-
are poised to differentiate into a population with an intrinsic tinct from gal-3+ macrophages, promotes clearance of VCAM+
quality to promote fibrosis. The development of lineage tracing FAPs to inhibit fibrosis. The enrichment of GO sets associated
systems to parse out the contribution of SkMRMs and MDMs to with ECM and developmental programs suggest that SkMRMs are
the gal-3+ macrophage pool and muscle fibrosis will be required a population with the functional capacity to inhibit fibrosis (fig. S5).
in future studies. Future studies will require specific deletion of SkMRMs to assess
Collectively, this study identified diverse subsets of muscle mac- their role in inhibiting fibrosis.
rophages with distinct functions and transcriptional profiles. The Collectively, these findings underscore the complexity of macro-
gal-3+Spp1+ signature reflected the predominant transcriptional phage and FAP interactions, which are regulated by cellular hetero-
state of a dystrophic muscle macrophage. Colocalization of gal-3+ geneity and differences in the inflammatory milieu between acutely
macrophages with stromal progenitors, and the observation that injured versus diseased muscle. In the present study, we interrogat-
Spp1 mediates communication between these cell types in vitro, re- ed macrophage and FAP interactions at the acute stages of muscular
inforces the importance of a macrophage-FAP fibrogenic axis in dystrophy. However, the persistence of this interaction with disease
promoting the pathogenesis of muscle disease. The translational progression has yet to be assessed. To elucidate the dynamics of
significance of this interaction is highlighted by the observation macrophage and FAP interactions and how they regulate the pro-
that a similar fibrogenic axis exists in human, as gal-3+ macrophages gression of fibrosis, future studies will require an assessment of con-
were elevated in several human muscle diseases. Further, gal-3+ ditional knockouts of Spp1 and its receptors throughout multiple
macrophages were identified in three different models of chronic stages of muscular dystrophy.
muscle disease and acute muscle injury, suggesting that a canonical
mechanism associated with muscle damage triggers differentiation
into the gal-3+ state. A likely candidate is phagocytosis of muscle cell MATERIALS AND METHODS
debris, which has been documented as a key contributor to macro- Study design
phage activation (70). However, this represents only one of the This study aimed to determine how skeletal muscle macrophages
many macrophage-FAP fibrogenic circuits that have been previous- promote fibrosis by using transcriptomics to define their homeo-
ly documented in muscle (11, 19). Mechanistic approaches relying static signature and how this state is altered with muscle disease.
on mouse genetics to study macrophage and stromal progenitor in- scRNAseq was used to define macrophage diversity in healthy
teractions will advance our understanding of how this axis pro- and diseased muscle. Bulk RNAseq analysis was performed on the
motes muscle fibrosis during muscular dystrophy. In summary, predominant macrophage populations identified in the scRNAseq
by defining the transcriptional heterogeneity of muscle macrophag- studies, to determine the regulation of the transition from homeo-
es, this study has advanced the understanding of macrophage acti- static to the diseased state. Spatial transcriptomics was used to un-
vation and function during muscle homeostasis and degenerative derstand how dystrophic macrophages interfaced with the
disease. The defined transcriptional states open a path for develop- dystrophic environment to promote fibrosis. Multiple mouse
ing groundbreaking therapeutic approaches to inhibit immune-me- models were used to demonstrate the significance of a gal-3+ mac-
diated muscle fibrosis. rophage population in muscular disease. Further, adoptive transfer
experiments were used to determine whether resident macrophages,
Limitations of the study and peripheral monocytes gave rise to gal-3+ macrophages. The
Studies of acute muscle injury suggest that the functional role of translation of these studies was assessed by quantifying gal-3+ mac-
FAP and gal-3+ macrophage interactions in muscle regeneration rophages in human muscle disease through an immunohistochem-
and fibrosis is complex. Nawaz et al. (71, 72) demonstrated that ical examination of archived, deidentified muscle biopsies. Prior
CD206+ cells, presumably macrophages, inhibit the promyogenic approval for collecting muscle tissue and its use in research was
activity of FAPs through a TGF-β–mediated repression of follistatin. given by the Institutional Review Board at the University of Califor-
TGF-β, whose processing depends on Spp1 (30), also promotes the nia Irvine (UCI; HS 2019-5134). All participants provided written
fibrogenic differentiation of FAPs and fibrosis (17, 73). Here, we informed consent and Health Insurance Portability and Account-
report that CD206 is expressed in gal-3hi macrophages (fig. S6D), ability Act authorization for data collection and the use of muscle
which promoted the differentiation of FAPs in a Spp1-dependent tissue for research.
manner. Whether gal-3+ macrophage–derived Spp1 promotes the
processing of latent TGF-β to its bioactive form remains to be ad-
dressed. Given the lack of an in vivo functional assessment of gal-3+/

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 12 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Experimental animals integration. For neighbor and cluster identification, the integrated
C57BL/10 (no. 000476), mdx mice (C57BL/10ScSn-Dmdmdx/J) object was scaled, and significant PCs were identified via statistical
(no. 001801), and mdx mice in the DBA2/J background (D2- and heuristic testing as recommended in Seurat. Clustered cells were
mdx) (no. 013141) were obtained from The Jackson Laboratory. visualized using UMAP. Before anchoring and integration, macro-
CD45.1 congenic mice (no. 002014) were also obtained from The phage identities from both datasets were renamed to match macro-
Jackson Laboratory and crossed with mdx mice at the UCI. B6A/J phage identities from the current study. Briefly, cells from WT
(no. 012767) and C57BL/6 mice (no. 000664) were bred in vivari- (monocyte_patrolling, monocyte_inflammatory, monocyte_mixed,
ums at Children’s National Hospital. VCP (77) and double homeo- M2 macrophage_Cx3cr1 lo, M2 macrophage_Cx3cr hi, and dendritic
box 4 (DUX4) (78) mice were provided by collaborators at the UCI cells) and mdx (Lyz2, Ctss, Cd68, Fcgr2b, Cd14, and Adgre1-express-
and Ohio State University, respectively. Animal experiments were ing cells) were subclustered and subjected to reference-based
approved by the Institutional Animal Care and Use Committee of mapping using the macrophage identities described in this study
UCI and performed under Institutional Animal Care and Use Com- (i.e., reference cells – macrophage 0 to 5) (81). Following a similar
mittee guidelines. approach, WT FAP_adipogenic, FAP_proremodeling, FAP_stem,
and tenocytes (37) were used as reference to assign identities to
Single-cell RNA sequencing Pdgfra + FAPs and tenocytes in the mdx dataset. All other previously
Single-cell preparation and analysis of muscle macrophages assigned cell identities described in Oprescu et al. (37) were kept in
Macrophages were sorted from 4-week-old WT and B10.mdx mice, the final clustering except for capillary, mixed, and vein endothelial
washed, and resuspended at a concentration of ~1000 cells/μl. Using cells, which were collapsed into “endothelial cells”; T and natural
the 10x Genomics platform, libraries for WT and mdx muscle mac- killer (NK) cells, which were separated into T and NKT cells; and
rophages were generated following the Chromium Single Cell 3′ Re- cycling cells, which were identified by expression of Mki67.
agents Kits v2 User Guide: CG00052 Rev. B. Quantification of Reference-based mapping of B6AJ and limb girdle
cDNA libraries was performed using the Qubit dsDNA HS Assay macrophages
Kit (Life Technologies, Q32851) and high-sensitivity DNA chips We performed reference-based mapping, anchoring, and integra-
(Agilent, 5067-4626). Quantification of library construction was tion of muscle macrophage scRNA datasets prepared from two
performed using KAPA qPCR (Kapa Biosystems, KK4824). 10x Ge- healthy and two dystrophic mice using Seurat (version 3.2.2, R
nomics libraries were sequenced on the Illumina HiSeq 4000 plat- Studio version 3.6.1) (81). Datasets were prepared from 8-month-
form to achieve an average of ~50,000 reads per cell according to the old B6A/J (LGMD2B, n = 2) and B6 (WT, n = 1). The annotated
recommendations in the Chromium Single Cell 3′ Reagents Kits v2 macrophage dataset in Fig. 1 was used as the reference. WT B10,
User Guide: CG00052 Rev. B. Sequencing reads were processed mdx, WT B6, and B6A/J objects were merged, SCTransform-nor-
using the 10x Genomics Cell Ranger 2.1.0. Each library was malized, anchored, and integrated as described above. Clustered
aligned to an indexed mm10 genome using Cell Ranger Count. cells were visualized using two-dimensional UMAP.
To generate an aggregated matrix of WT and mdx cells and to Modeling cell-cell communication networks
prepare data for downstream analysis, the Cell Ranger Aggr func- Intra- and intercellular communication networks were modeled
tion was used to normalize the number of confidently mapped based on the basis of abundance of known ligand-receptor (L-R)
reads per cell in each library. transcript pairs with CellChat (version 1.1.3) (36). To identify con-
The Seurat pipeline (version 3.0.2) was applied to the aligned cell served and perturbed FAP-macrophage communication networks
matrix using R (version 3.6.1) to identify cell clusters. Quality in WT and mdx muscles, we lifted cells from a WT-mdx integrated
control filtering was first performed to remove genes that were object. We performed joint manifold and classification learning
not expressed (>0) in at least three cells and cells that had less analyses as described in CellChat.
than 200 genes. The trimmed expression count matrix was log-
transformed for downstream processing, and highly variable Bulk RNAseq and analysis
genes were detected. PCA was combined with the elbow method Total RNA was monitored for quality control using the Agilent Bio-
to determine loadings for the generation of UMAP with 10 PCs in- analyzer Pico RNA chip (Agilent Technologies, Santa Clara, CA)
cluded. Using these same PCs, Seurat’s default clustering was per- and NanoDrop (Thermo Fisher Scientific, Waltham MA) absor-
formed and was followed up with marker gene detection to bance ratios for 260/280 and 260/230 nm. Library construction
elucidate gene expression signatures corresponding to the resultant was performed according to the SMARTer Stranded Total RNA-
clusters. Seq Kit v2-Pico Input Mammalian (Takara Bio, Mountain View,
Reference-based mapping of macrophages, FAP, and CA). The input quantity for total RNA was 2 ng. The total RNA
tenocytes was fragmented for 3 min at 94°C. SMART (Switching Mechanism
We performed anchoring and integration of single-cell datasets of At 5′ end of RNA Template) cDNA synthesis technology was used
uninjured WT [i.e., day 0, publicly available dataset from Oprescu et to synthesize cDNA from the fragmented total RNA. Illumina
al. (37)] and mdx mouse muscle that served as a negative control adapters were ligated to the ends and enriched by 5 cycles of
group in a separate study of mice treated with tamoxifen (dataset PCR. R-probes v2 (mammalian specific) are then hybridized to
provided by M.J.S.). Seurat (version 3.2.2, R Studio version 3.6.1) the cDNA that contains ribosomal RNA and human mitochondrial
was used for anchoring and integration (79). Briefly, merged ribosomal RNA sequences. The R-probe v2 hybridized cDNAs are
Seurat objects were normalized, and highly variable genes (features) then cut by ZapR v2. The leftover library fragments are further en-
and scaling were performed with SCTransform (80). The top 2000 riched by 14 cycles of PCR. The final libraries are purified via
highly variable features were selected and used for anchoring. Inte- AMPure XP beads. The resulting libraries were validated by qPCR
gration anchors (30 dimensions) were computed and used for (Kapa library quantification kit, Kapa Biosystems, Wilmington MA)

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 13 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

and sized by an Agilent Bioanalyzer DNA high-sensitivity chip. The treated with the EasySep reagents, and monocytes were isolated
concentrations for the libraries were normalized and then multi- by depletion using an EasyPlate magnet (STEMCELL Technolo-
plexed together. The multiplexed libraries were sequenced using gies). Muscle immune cells were isolated from hindlimb muscles
paired-end 100 chemistry on the NovaSeq 6000 (Illumina, San as described previously and enriched using the EasySep Release
Diego, CA). Mouse Allophycocyanin (APC) selection kit (STEMCELL Technol-
Public data for microglia (33) (GSE132877) and bone-derived ogies). Briefly, the muscle single-cell suspension was resuspended in
macrophage (34) (provided by authors) RNAseq were processed 0.25 ml of recommended medium and incubated with the Folr2-
from original fastq read files. All reads were mapped to the mouse APC antibody (2 μg/ml) for 5 min on ice. Following incubation
genome (mm10) (82) with TopHat (83) (version 2.0.14) with refer- with 25 μl of APC selection cocktail for 5 min at 4°C, APC+ cell se-
ence GENCODE transcript annotation (84) (M9). Overlapping lection is obtained by incubating with RapidSpheres magnetic beads
reads were counted and summarized by gene using HTSeq (85) on ice for 3 min. Folr2+ macrophages bound to the magnetic beads
(1.99.2). The R package DESeq2 Field (85) (version 1.28.1) was were eluted after adding the ice-cold release buffer on ice for 3 min,
used to determine DEGs between datasets. From DESeq2 output, counted, and saved for downstream experiments. Isolated mono-
DEGs were classified for twofold changes up or down with a false cytes (20,000 cells/μl in PBS) or Folr2+ macrophages (15,000 cells/
discovery rate < 0.01. Gene sets from the Molecular Signature Da- μl in PBS) were injected in 10 μl intramuscularly. Briefly, mdx mice
tabase (86) (mSigDB) were downloaded from the Gene Set Enrich- were anesthetized with isoflurane while placed over a heating pad to
ment Analysis webpage (http://software.broadinstitute.org/gsea) to maintain thermoregulation. Anesthetized mice were sterilized with
determine gene set enrichment. Plots were generated in R using 70% ethanol, and 10 μl of cell suspension was injected into the
Venn diagram (version 1.6.20), ggrepel (version 0.9.0), Complex- quadriceps using a Hamilton syringe. When appropriate, the
Heatmap (version 2.4.3), plyr (version 1.8.6), and ggplot2 other quadriceps was used as a control where no cells were injected
(version 3.3.3). into the muscle. Flow analysis of the quadriceps of mdx mice was
performed 2 and 7 days following the adoptive transfer.
Spatial transcriptomics
The 10x Genomics Visium Spatial Gene Expression platform was Statistical analyses
used for spatial transcriptomics analysis of muscle from dystrophic Data were expressed as mean ± SEM. Statistical analyses were per-
mice (D2-mdx, stock no. 013141 from The Jackson Laboratory) ac- formed using GraphPad Prism version 9.2. Statistical comparisons
cording to the manufacturer’s guidelines. Briefly, 6-week-old male between the two groups were performed using an unpaired t test
mice were euthanized via cervical dislocation under isoflurane an- with Welch’s correction. One-way or two-way analysis of variance
esthesia. The gastrocnemius/plantaris muscle complex was immedi- (ANOVA) with a post hoc Bonferroni test, Kruskal-Wallis multiple
ately dissected and frozen in optimal cutting temperature (OCT) comparisons test, or Sidak’s multiple comparisons tests was con-
embedding media in liquid nitrogen–cooled isopentane. Muscle ducted when comparing multiple groups. P values δ of 0.05 were
tissue was cryosectioned at −20°C at 10-μm thickness onto considered significant.
Visium Spatial Gene Expression slides (10x Genomics) and stored
at −80°C until processing. Sections were fixed in prechilled metha-
nol for 30 min. Hematoxylin and eosin (H&E) staining was per- Supplementary Materials
formed per the published protocol from 10x Genomics, with This PDF file includes:
Supplementary Materials and Methods
imaging performed using a Zeiss Axio Observer microscope.
Figs. S1 to S13
H&E images were stitched and processed using Zen 2.0 software. Legend for table S1
Following imaging, tissues were permeabilized for 12 min, which Table S2
was predetermined as the optimal time for 10-μm mouse muscle
sections using the 10x Genomics Visium Tissue Optimization Kit. Other Supplementary Material for this
Spatially tagged cDNA libraries were built using the 10x Genomics manuscript includes the following:
Table S1
Visium Spatial Gene Expression Library Construction Kit. Sequenc-
ing was performed on an Illumina NextSeq 500/550 using 150-cycle
high output kits (read 1 = 28, read 2 = 120, index 1 = 10, and index 2
REFERENCES AND NOTES
= 10). Alignment to the mouse reference genome mm10 (Ensembl 1. L. C. Davies, S. J. Jenkins, J. E. Allen, P. R. Taylor, Tissue-resident macrophages. Nat. Immunol.
93) was done using the Space Ranger 1.0.0 pipeline to derive a 14, 986–995 (2013).
feature spot-barcode expression matrix (10x Genomics). Alignment 2. Y. Okabe, Molecular control of the identity of tissue-resident macrophages. Int. Immunol.
of H&E images was done using Loupe Browser. 30, 485–491 (2018).
3. J. G. Tidball, S. A. Villalta, Regulatory interactions between muscle and the immune system
during muscle regeneration. Am. J. Physiol. Regul. Integr. Comp. Physiol. 298,
Adoptive transfer
R1173–R1187 (2010).
Monocytes were isolated from bone marrow using the EasySep 4. D. M. Mosser, J. P. Edwards, Exploring the full spectrum of macrophage activation. Nat. Rev.
Mouse Monocyte Enrichment kit (STEMCELL Technologies). Immunol. 8, 958–969 (2008).
Briefly, femurs and tibias were harvested and flushed through 5. M. Saclier, H. Yacoub-Youssef, A. L. Mackey, L. Arnold, H. Ardjoune, M. Magnan, F. Sailhan, J.
with cold phosphate-buffered saline (PBS). Bone marrow cell sus- Chelly, G. K. Pavlath, R. Mounier, M. Kjaer, B. Chazaud, Differentially activated macrophages
pensions were passed through a 70-μm cell strainer to obtain a orchestrate myogenic precursor cell fate during human skeletal muscle regeneration. Stem
Cells 31, 384–396 (2013).
single-cell suspension. Red blood cells were removed using the
6. M. L. Novak, E. M. Weinheimer-Haus, T. J. Koh, Macrophage activation and skeletal muscle
red blood cell lysis buffer according to the manufacturer’s instruc- healing following traumatic injury. J. Pathol. 232, 344–355 (2014).
tions (Sigma-Aldrich). The bone marrow cell suspension was

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 14 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

7. C. J. Mann, E. Perdiguero, Y. Kharraz, S. Aguilar, P. Pessina, A. L. Serrano, P. Munoz-Canoves, 28. L. Calvier, M. Miana, P. Reboul, V. Cachofeiro, E. Martinez-Martinez, R. A. de Boer, F. Poirier, P.
Aberrant repair and fibrosis development in skeletal muscle. Skelet. Muscle 1, 21 (2011). Lacolley, F. Zannad, P. Rossignol, N. Lopez-Andres, Galectin-3 mediates aldosterone-
8. M. Koenig, E. P. Hoffman, C. J. Bertelson, A. P. Monaco, C. Feener, L. M. Kunkel, Complete induced vascular fibrosis. Arterioscler. Thromb. Vasc. Biol. 33, 67–75 (2013).
cloning of the Duchenne muscular dystrophy (DMD) cDNA and preliminary genomic or- 29. E. Martinez-Martinez, L. Calvier, A. Fernandez-Celis, E. Rousseau, R. Jurado-Lopez, L. V.
ganization of the DMD gene in normal and affected individuals. Cell 50, 509–517 (1987). Rossoni, F. Jaisser, F. Zannad, P. Rossignol, V. Cachofeiro, N. Lopez-Andres, Galectin-3
9. S. Dadgar, Z. Wang, H. Johnston, A. Kesari, K. Nagaraju, Y. W. Chen, D. A. Hill, T. A. Partridge, blockade inhibits cardiac inflammation and fibrosis in experimental hyperaldosteronism
M. Giri, R. J. Freishtat, J. Nazarian, J. Xuan, Y. Wang, E. P. Hoffman, Asynchronous remod- and hypertension. Hypertension 66, 767–775 (2015).
eling is a driver of failed regeneration in Duchenne muscular dystrophy. J. Cell Biol. 207, 30. I. Kramerova, C. Kumagai-Cresse, N. Ermolova, E. Mokhonova, M. Marinov, J. Capote, D.
139–158 (2014). Becerra, M. Quattrocelli, R. H. Crosbie, E. Welch, E. M. McNally, M. J. Spencer, Spp1 (os-
10. S. A. Villalta, H. X. Nguyen, B. Deng, T. Gotoh, J. G. Tidball, Shifts in macrophage phenotypes teopontin) promotes TGFβ processing in fibroblasts of dystrophin-deficient muscles
and macrophage competition for arginine metabolism affect the severity of muscle pa- through matrix metalloproteinases. Hum. Mol. Genet. 28, 3431–3442 (2019).
thology in muscular dystrophy. Hum. Mol. Genet. 18, 482–496 (2009). 31. S. Gordon, J. Hamann, H. H. Lin, M. Stacey, F4/80 and the related adhesion-GPCRs. Eur. J.
11. G. Juban, M. Saclier, H. Yacoub-Youssef, A. Kernou, L. Arnold, C. Boisson, S. Ben Larbi, M. Immunol. 41, 2472–2476 (2011).
Magnan, S. Cuvellier, M. Theret, B. J. Petrof, I. Desguerre, J. Gondin, R. Mounier, B. Chazaud, 32. S. A. Dick, A. Wong, H. Hamidzada, S. Nejat, R. Nechanitzky, S. Vohra, B. Mueller, R. Zaman, C.
AMPK activation regulates LTBP4-dependent TGF-β1 secretion by pro-inflammatory mac- Kantores, L. Aronoff, A. Momen, D. Nechanitzky, W. Y. Li, P. Ramachandran, S. Q. Crome, B.
rophages and controls fibrosis in Duchenne muscular dystrophy. Cell Rep. 25, Becher, M. I. Cybulsky, F. Billia, S. Keshavjee, S. Mital, C. S. Robbins, T. W. Mak, S. Epelman,
2163–2176.e6 (2018). Three tissue resident macrophage subsets coexist across organs with conserved origins
12. M. Wehling-Henricks, M. C. Jordan, T. Gotoh, W. W. Grody, K. P. Roos, J. G. Tidball, Arginine and life cycles. Sci. Immunol. 7, eabf7777 (2022).
metabolism by macrophages promotes cardiac and muscle fibrosis in mdx muscular 33. A. Crotti, H. R. Sait, K. M. McAvoy, K. Estrada, A. Ergun, S. Szak, G. Marsh, L. Jandreski, M.
dystrophy. PLOS ONE 5, e10763 (2010). Peterson, T. L. Reynolds, I. Dalkilic-Liddle, A. Cameron, E. Cahir-McFarland, R. M. Ransohoff,
13. B. Vidal, A. L. Serrano, M. Tjwa, M. Suelves, E. Ardite, R. De Mori, B. Baeza-Raja, M. Martinez BIN1 favors the spreading of tau via extracellular vesicles. Sci. Rep. 9, 9477 (2019).
de Lagran, P. Lafuste, V. Ruiz-Bonilla, M. Jardi, R. Gherardi, C. Christov, M. Dierssen, P. Car- 34. C. P. Hans, N. Sharma, S. Sen, S. Zeng, R. Dev, Y. Jiang, A. Mahajan, T. Joshi, Transcriptomics
meliet, J. L. Degen, M. Dewerchin, P. Munoz-Canoves, Fibrinogen drives dystrophic muscle analysis reveals new insights into the roles of Notch1 signaling on macrophage polariza-
fibrosis via a TGFβ/alternative macrophage activation pathway. Genes Dev. 22, tion. Sci. Rep. 9, 7999 (2019).
1747–1752 (2008). 35. M. Giannandrea, W. C. Parks, Diverse functions of matrix metalloproteinases during
14. G. J. Schaaf, T. J. M. van Gestel, E. Brusse, R. M. Verdijk, I. F. M. de Coo, P. A. van Doorn, A. T. fibrosis. Dis. Model. Mech. 7, 193–203 (2014).
van der Ploeg, W. W. M. Pijnappel, Lack of robust satellite cell activation and muscle re- 36. S. Jin, C. F. Guerrero-Juarez, L. Zhang, I. Chang, R. Ramos, C. H. Kuan, P. Myung, M. V. Plikus,
generation during the progression of Pompe disease. Acta Neuropathol. Commun. 3, Q. Nie, Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12,
65 (2015). 1088 (2021).
15. J. H. Baek, G. M. Many, F. J. Evesson, V. R. Kelley, Dysferlinopathy promotes an intramuscle 37. S. N. Oprescu, F. Yue, J. Qiu, L. F. Brito, S. Kuang, Temporal dynamics and heterogeneity of
expansion of macrophages with a cyto-destructive phenotype. Am. J. Pathol. 187, cell populations during skeletal muscle regeneration. iScience 23, 100993 (2020).
1245–1257 (2017). 38. D. W. McKellar, L. D. Walter, L. T. Song, M. Mantri, M. F. Z. Wang, I. De Vlaminck, B. D. Cos-
16. T. A. Wynn, T. R. Ramalingam, Mechanisms of fibrosis: Therapeutic translation for fibrotic grove, Large-scale integration of single-cell transcriptomic data captures transitional
disease. Nat. Med. 18, 1028–1040 (2012). progenitor states in mouse skeletal muscle regeneration. Commun. Biol. 4, 1280 (2021).
17. A. Uezumi, T. Ito, D. Morikawa, N. Shimizu, T. Yoneda, M. Segawa, M. Yamaguchi, R. Ogawa, 39. J. M. Kastenschmidt, G. Coulis, P. K. Farahat, P. Pham, R. Rios, T. T. Cristal, A. H. Mannaa, R. E.
M. M. Matev, Y. Miyagoe-Suzuki, S. Takeda, K. Tsujikawa, K. Tsuchida, H. Yamamoto, S. Ayer, R. Yahia, A. A. Deshpande, B. S. Hughes, A. K. Savage, C. R. Giesige, S. Q. Harper, R. M.
Fukada, Fibrosis and adipogenesis originate from a common mesenchymal progenitor in Locksley, T. Mozaffar, S. A. Villalta, A stromal progenitor and ILC2 niche promotes muscle
skeletal muscle. J. Cell Sci. 124, 3654–3664 (2011). eosinophilia and fibrosis-associated gene expression. Cell Rep. 35, 108997 (2021).
18. A. W. B. Joe, L. Yi, A. Natarajan, F. Le Grand, L. So, J. Wang, M. A. Rudnicki, F. M. V. Rossi, 40. H. Khalil, O. Kanisicak, R. J. Vagnozzi, A. K. Johansen, B. D. Maliken, V. Prasad, J. G. Boyer, M. J.
Muscle injury activates resident fibro/adipogenic progenitors that facilitate myogenesis. Brody, T. Schips, K. K. Kilian, R. N. Correll, K. Kawasaki, K. Nagata, J. D. Molkentin, Cell-specific
Nat. Cell Biol. 12, 153–163 (2010). ablation of Hsp47 defines the collagen-producing cells in the injured heart. JCI Insight 4,
19. D. R. Lemos, F. Babaeijandaghi, M. Low, C.-K. Chang, S. T. Lee, D. Fiore, R.-H. Zhang, A. e128722 (2019).
Natarajan, S. A. Nedospasov, F. M. V. Rossi, Nilotinib reduces muscle fibrosis in chronic 41. S. W. Lee, J. Y. Won, W. J. Kim, J. Lee, K. H. Kim, S. W. Youn, J. Y. Kim, E. J. Lee, Y. J. Kim, K. W.
muscle injury by promoting TNF-mediated apoptosis of fibro/adipogenic progenitors. Nat. Kim, H. S. Kim, Snail as a potential target molecule in cardiac fibrosis: Paracrine action of
Med. 21, 786–794 (2015). endothelial cells on fibroblasts through snail and CTGF axis. Mol. Ther. 21,
20. P. Bernasconi, E. Torchiana, P. Confalonieri, R. Brugnoni, R. Barresi, M. Mora, F. Cornelio, L. 1767–1777 (2013).
Morandi, R. Mantegazza, Expression of transforming growth factor-β1 in dystrophic 42. S. A. Dick, J. A. Macklin, S. Nejat, A. Momen, X. Clemente-Casares, M. G. Althagafi, J. Chen, C.
patient muscles correlates with fibrosis. Pathogenetic role of a fibrogenic cytokine. J. Clin. Kantores, S. Hosseinzadeh, L. Aronoff, A. Wong, R. Zaman, I. Barbu, R. Besla, K. J. Lavine, B.
Invest. 96, 1137–1144 (1995). Razani, F. Ginhoux, M. Husain, M. I. Cybulsky, C. S. Robbins, S. Epelman, Self-renewing
21. M. Scatena, L. Liaw, C. M. Giachelli, Osteopontin: A multifunctional molecule regulating resident cardiac macrophages limit adverse remodeling following myocardial infarction.
chronic inflammation and vascular disease. Arterioscler. Thromb. Vasc. Biol. 27, Nat. Immunol. 20, 29–39 (2019).
2302–2309 (2007). 43. M. P. Emont, C. Jacobs, A. L. Essene, D. Pant, D. Tenen, G. Colleluori, A. Di Vincenzo, A. M.
22. S. A. Vetrone, E. Montecino-Rodriguez, E. Kudryashova, I. Kramerova, E. P. Hoffman, S. D. Liu, Jorgensen, H. Dashti, A. Stefek, E. McGonagle, S. Strobel, S. Laber, S. Agrawal, G. P. Westcott,
M. C. Miceli, M. J. Spencer, Osteopontin promotes fibrosis in dystrophic mouse muscle by A. Kar, M. L. Veregge, A. Gulko, H. Srinivasan, Z. Kramer, E. De Filippis, E. Merkel, J. Ducie,
modulating immune cell subsets and intramuscular TGF-β. J. Clin. Invest. 119, C. G. Boyd, W. Gourash, A. Courcoulas, S. J. Lin, B. T. Lee, D. Morris, A. Tobias, A. V. Khera, M.
1583–1594 (2009). Claussnitzer, T. H. Pers, A. Giordano, O. Ashenberg, A. Regev, L. T. Tsai, E. D. Rosen, A single-
23. J. Capote, I. Kramerova, L. Martinez, S. Vetrone, E. R. Barton, H. L. Sweeney, M. C. Miceli, M. J. cell atlas of human and mouse white adipose tissue. Nature 603, 926–933 (2022).
Spencer, Osteopontin ablation ameliorates muscular dystrophy by shifting macrophages 44. H. Keren-Shaul, A. Spinrad, A. Weiner, O. Matcovitch-Natan, R. Dvir-Szternfeld, T. K. Ulland,
to a pro-regenerative phenotype. J. Cell Biol. 213, 275–288 (2016). E. David, K. Baruch, D. Lara-Astaiso, B. Toth, S. Itzkovitz, M. Colonna, M. Schwartz, I. Amit, A
24. L. C. Li, J. Li, J. Gao, Functions of galectin-3 and its role in fibrotic diseases. J. Pharmacol. Exp. unique microglia type associated with restricting sevelopment of Alzheimer’s disease. Cell
Ther. 351, 336–343 (2014). 169, 1276–1290.e17 (2017).
25. P. Zahradka, Novel role for osteopontin in cardiac fibrosis. Circ. Res. 102, 270–272 (2008). 45. A. Patsalos, L. Halasz, M. A. Medina-Serpas, W. K. Berger, B. Daniel, P. Tzerpos, M. Kiss, G.
26. M. Wehling, M. J. Spencer, J. G. Tidball, A nitric oxide synthase transgene ameliorates Nagy, C. Fischer, Z. Simandi, T. Varga, L. Nagy, A growth factor-expressing macrophage
muscular dystrophy in mdx mice. J. Cell Biol. 155, 123–131 (2001). subpopulation orchestrates regenerative inflammation via GDF-15. J. Exp. Med. 219,
27. X. Wang, A. A. Sathe, G. R. Smith, F. Ruf-Zamojski, V. Nair, K. J. Lavine, C. Xing, S. C. Sealfon, L. e20210420 (2022).
Zhou, Heterogeneous origins and functions of mouse skeletal muscle-resident macro- 46. T. Varga, R. Mounier, A. Patsalos, P. Gogolak, M. Peloquin, A. Horvath, A. Pap, B. Daniel, G.
phages. Proc. Natl. Acad. Sci. U.S.A. 117, 20729–20740 (2020). Nagy, E. Pintye, S. Poliska, S. Cuvellier, S. B. Larbi, B. E. Sansbury, M. Spite, C. W. Brown, B.
Chazaud, L. Nagy, Macrophage PPARγ, a lipid activated transcription factor controls the
growth factor GDF3 and skeletal muscle regeneration. Immunity 45, 1038–1051 (2016).

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 15 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

47. K. K. Saleh, H. Xi, C. Switzler, E. Skuratovsky, M. A. Romero, P. Chien, D. Gibbs, L. Gane, M. R. 69. W. Zhao, X. Wang, R. M. Ransohoff, L. Zhou, CCR2 deficiency does not provide sustained
Hicks, M. J. Spencer, A. D. Pyle, Single cell sequencing maps skeletal muscle cellular di- improvement of muscular dystrophy in mdx5cv mice. FASEB J. 31, 35–46 (2017).
versity as disease severity increases in dystrophic mouse models. iScience 25, 70. L. Arnold, A. Henry, F. Poron, Y. Baba-Amer, N. van Rooijen, A. Plonquet, R. K. Gherardi, B.
105415 (2022). Chazaud, Inflammatory monocytes recruited after skeletal muscle injury switch into
48. Y. Matsui, N. Jia, H. Okamoto, S. Kon, H. Onozuka, M. Akino, L. Liu, J. Morimoto, S. R. Rittling, antiinflammatory macrophages to support myogenesis. J. Exp. Med. 204,
D. Denhardt, A. Kitabatake, T. Uede, Role of osteopontin in cardiac fibrosis and remodeling 1057–1069 (2007).
in angiotensin II-induced cardiac hypertrophy. Hypertension 43, 1195–1201 (2004). 71. A. Nawaz, M. Bilal, S. Fujisaka, T. Kado, M. R. Aslam, S. Ahmed, K. Okabe, Y. Igarashi, Y.
49. I. Abdelaziz Mohamed, A. P. Gadeau, A. Hasan, N. Abdulrahman, F. Mraiche, Osteopontin: A Watanabe, T. Kuwano, K. Tsuneyama, A. Nishimura, Y. Nishida, S. Yamamoto, M. Sasahara, J.
promising therapeutic target in cardiac fibrosis. Cell 8, (2019). Imura, H. Mori, M. M. Matzuk, F. Kudo, I. Manabe, A. Uezumi, T. Nakagawa, Y. Oishi, K. Tobe,
50. R. J. Slack, R. Mills, A. C. Mackinnon, The therapeutic potential of galectin-3 inhibition in Depletion of CD206+ M2-like macrophages induces fibro-adipogenic progenitors activa-
fibrotic disease. Int. J. Biochem. Cell Biol. 130, 105881 (2021). tion and muscle regeneration. Nat. Commun. 13, 7058 (2022).
51. F. Takahashi, K. Takahashi, T. Okazaki, K. Maeda, H. Ienaga, M. Maeda, S. Kon, T. Uede, Y. 72. K. M. Jansen, G. K. Pavlath, Mannose receptor regulates myoblast motility and muscle
Fukuchi, Role of osteopontin in the pathogenesis of bleomycin-induced pulmonary growth. J. Cell Biol. 174, 403–413 (2006).
fibrosis. Am. J. Respir. Cell Mol. Biol. 24, 264–271 (2001). 73. J. R. Fallon, E. M. McNally, Non-Glycanated biglycan and LTBP4: Leveraging the extracel-
52. T. Wolak, H. Kim, Y. Ren, J. Kim, N. D. Vaziri, S. B. Nicholas, Osteopontin modulates angio- lular matrix for Duchenne muscular dystrophy therapeutics. Matrix Biol. 68-69,
tensin II-induced inflammation, oxidative stress, and fibrosis of the kidney. Kidney Int. 76, 616–627 (2018).
32–43 (2009). 74. J. A. Larouche, E. C. Wallace, B. D. Spence, E. Buras, C. A. Aguilar, Spatiotemporal mapping
53. N. Takada, M. Takasugi, Y. Nonaka, T. Kamiya, K. Takemura, J. Satoh, S. Ito, K. Fujimoto, S. of immune and stem cell dysregulation after volumetric muscle loss. JCI Insight 8, (2023).
Uematsu, K. Yoshida, T. Morita, H. Nakamura, A. Uezumi, N. Ohtani, Galectin-3 promotes 75. B. Malecova, S. Gatto, U. Etxaniz, M. Passafaro, A. Cortez, C. Nicoletti, L. Giordani, A. Torci-
the adipogenic differentiation of PDGFRα+ cells and ectopic fat formation in regenerating naro, M. De Bardi, S. Bicciato, F. De Santa, L. Madaro, P. L. Puri, Dynamics of cellular states of
muscle. Development 149, dev199443 (2022). fibro-adipogenic progenitors during myogenesis and muscular dystrophy. Nat. Commun.
54. N. C. Henderson, A. C. Mackinnon, S. L. Farnworth, F. Poirier, F. P. Russo, J. P. Iredale, C. 9, 3670 (2018).
Haslett, K. J. Simpson, T. Sethi, Galectin-3 regulates myofibroblast activation and hepatic 76. R. S. Larson, T. A. Springer, Structure and function of leukocyte integrins. Immunol. Rev.
fibrosis. Proc. Natl. Acad. Sci. U.S.A. 103, 5060–5065 (2006). 114, 181–217 (1990).
55. Y. Lenga, A. Koh, A. S. Perera, C. A. McCulloch, J. Sodek, R. Zohar, Osteopontin expression is 77. A. Nalbandian, K. J. Llewellyn, M. Kitazawa, H. Z. Yin, M. Badadani, N. Khanlou, R. Edwards, C.
required for myofibroblast differentiation. Circ. Res. 102, 319–327 (2008). Nguyen, J. Mukherjee, T. Mozaffar, G. Watts, J. Weiss, V. E. Kimonis, The homozygote
56. C. Jaquenod De Giusti, A. E. Ure, L. Rivadeneyra, M. Schattner, R. M. Gomez, Macrophages VCP(R155H/R155H) mouse model exhibits accelerated human VCP-associated disease pa-
and galectin 3 play critical roles in CVB3-induced murine acute myocarditis and chronic thology. PLOS ONE 7, e46308 (2012).
fibrosis. J. Mol. Cell. Cardiol. 85, 58–70 (2015). 78. C. R. Giesige, L. M. Wallace, K. N. Heller, J. O. Eidahl, N. Y. Saad, A. M. Fowler, N. K. Pyne, M. Al-
57. J.-D. Lin, H. Nishi, J. Poles, X. Niu, C. McCauley, K. Rahman, E. J. Brown, S. T. Yeung, N. Kharsan, A. Rashnonejad, G. A. Chermahini, J. S. Domire, D. Mukweyi, S. E. Garwick-
Vozhilla, A. Weinstock, S. A. Ramsey, E. A. Fisher, P. Loke, Single-cell analysis of fate-mapped Coppens, S. M. Guckes, K. J. McLaughlin, K. Meyer, L. R. Rodino-Klapac, S. Q. Harper, AAV-
macrophages reveals heterogeneity, including stem-like properties, during atherosclerosis mediated follistatin gene therapy improves functional outcomes in the TIC-DUX4 mouse
progression and regression. JCI Insight 4, e124574 (2019). model of FSHD. JCI Insight 3, e123538 (2018).
58. E. G. O’Koren, C. Yu, M. Klingeborn, A. Y. W. Wong, C. L. Prigge, R. Mathew, J. Kalnitsky, R. A. 79. T. Stuart, A. Butler, P. Hoffman, C. Hafemeister, E. Papalexi, W. M. Mauck 3rd, Y. Hao, M.
Msallam, A. Silvin, J. N. Kay, C. Bowes Rickman, V. Y. Arshavsky, F. Ginhoux, M. Merad, D. R. Stoeckius, P. Smibert, R. Satija, Comprehensive integration of single-cell data. Cell 177,
Saban, Microglial function is distinct in different anatomical locations during retinal ho- 1888–1902.e21 (2019).
meostasis and degeneration. Immunity 50, 723–737.e7 (2019). 80. C. Hafemeister, R. Satija, Normalization and variance stabilization of single-cell RNA-seq
59. F. O. Quenum Zangbede, A. Chauhan, J. Sharma, B. B. Mishra, Galectin-3 in M2 macro- data using regularized negative binomial regression. Genome Biol. 20, 296 (2019).
phages plays a protective role in resolution of neuropathology in brain parasitic infection 81. Y. Hao, S. Hao, E. Andersen-Nissen, W. M. Mauck 3rd, S. Zheng, A. Butler, M. J. Lee, A. J. Wilk,
by regulating neutrophil turnover. J. Neurosci. 38, 6737–6750 (2018). C. Darby, M. Zager, P. Hoffman, M. Stoeckius, E. Papalexi, E. P. Mimitou, J. Jain, A. Srivastava,
60. D. C. Humphries, R. Mills, R. Dobie, N. C. Henderson, T. Sethi, A. C. Mackinnon, Selective T. Stuart, L. M. Fleming, B. Yeung, A. J. Rogers, J. M. McElrath, C. A. Blish, R. Gottardo, P.
myeloid depletion of galectin-3 offers protection against acute and chronic lung injury. Smibert, R. Satija, Integrated analysis of multimodal single-cell data. Cell 184,
Front. Pharmacol. 12, 715986 (2021). 3573–3587.e29 (2021).
61. D. K. Wasgewatte Wijesinghe, E. J. Mackie, C. N. Pagel, Normal inflammation and regen- 82. V. A. Schneider, T. Graves-Lindsay, K. Howe, N. Bouk, H. C. Chen, P. A. Kitts, T. D. Murphy, K. D.
eration of muscle following injury require osteopontin from both muscle and non-muscle Pruitt, F. Thibaud-Nissen, D. Albracht, R. S. Fulton, M. Kremitzki, V. Magrini, C. Markovic, S.
cells. Skelet. Muscle 9, 6 (2019). McGrath, K. M. Steinberg, K. Auger, W. Chow, J. Collins, G. Harden, T. Hubbard, S. Pelan, J. T.
62. K. Uaesoontrachoon, D. K. Wasgewatte Wijesinghe, E. J. Mackie, C. N. Pagel, Osteopontin Simpson, G. Threadgold, J. Torrance, J. M. Wood, L. Clarke, S. Koren, M. Boitano, P. Peluso, H.
deficiency delays inflammatory infiltration and the onset of muscle regeneration in a Li, C. S. Chin, A. M. Phillippy, R. Durbin, R. K. Wilson, P. Flicek, E. E. Eichler, D. M. Church,
mouse model of muscle injury. Dis. Model. Mech. 6, 197–205 (2013). Evaluation of GRCh38 and de novo haploid genome assemblies demonstrates the en-
63. D. G. Cerri, L. C. Rodrigues, V. M. Alves, J. Machado, V. A. F. Bastos, I. D. Carmo Kettelhut, L. C. during quality of the reference assembly. Genome Res. 27, 849–864 (2017).
Alberici, M. C. R. Costa, S. R. Stowell, R. D. Cummings, M. Dias-Baruffi, Endogenous galectin- 83. C. Trapnell, L. Pachter, S. L. Salzberg, TopHat: Discovering splice junctions with RNA-Seq.
3 is required for skeletal muscle repair. Glycobiology 31, 1295–1307 (2021). Bioinformatics 25, 1105–1111 (2009).
64. T. A. Wynn, Common and unique mechanisms regulate fibrosis in various fibroproliferative 84. A. Frankish, M. Diekhans, A. M. Ferreira, R. Johnson, I. Jungreis, J. Loveland, J. M. Mudge, C.
diseases. J. Clin. Invest. 117, 524–529 (2007). Sisu, J. Wright, J. Armstrong, I. Barnes, A. Berry, A. Bignell, S. Carbonell Sala, J. Chrast, F.
65. F. Babaeijandaghi, R. Cheng, N. Kajabadi, H. Soliman, C. K. Chang, J. Smandych, L. W. Tung, Cunningham, T. Di Domenico, S. Donaldson, I. T. Fiddes, C. Garcia Giron, J. M. Gonzalez, T.
R. Long, A. Ghassemi, F. M. V. Rossi, Metabolic reprogramming of skeletal muscle by res- Grego, M. Hardy, T. Hourlier, T. Hunt, O. G. Izuogu, J. Lagarde, F. J. Martin, L. Martinez, S.
ident macrophages points to CSF1R inhibitors as muscular dystrophy therapeutics. Sci. Mohanan, P. Muir, F. C. P. Navarro, A. Parker, B. Pei, F. Pozo, M. Ruffier, B. M. Schmitt, E.
Transl. Med. 14, eabg7504 (2022). Stapleton, M. M. Suner, I. Sycheva, B. Uszczynska-Ratajczak, J. Xu, A. Yates, D. Zerbino, Y.
Zhang, B. Aken, J. S. Choudhary, M. Gerstein, R. Guigo, T. J. P. Hubbard, M. Kellis, B. Paten, A.
66. J. E. Heredia, L. Mukundan, F. M. Chen, A. A. Mueller, R. C. Deo, R. M. Locksley, T. A. Rando, A.
Reymond, M. L. Tress, P. Flicek, GENCODE reference annotation for the human and mouse
Chawla, Type 2 innate signals stimulate fibro/adipogenic progenitors to facilitate muscle
genomes. Nucleic Acids Res. 47, D766–D773 (2019).
regeneration. Cell 153, 376–388 (2013).
85. S. Anders, P. T. Pyl, W. Huber, HTSe–A Python framework to work with high-throughput
67. R. Mounier, M. Theret, L. Arnold, S. Cuvellier, L. Bultot, O. Goransson, N. Sanz, A. Ferry, K.
sequencing data. Bioinformatics 31, 166–169 (2015).
Sakamoto, M. Foretz, B. Viollet, B. Chazaud, AMPKα1 regulates macrophage skewing at the
time of resolution of inflammation during skeletal muscle regeneration. Cell Metab. 18, 86. A. Liberzon, C. Birger, H. Thorvaldsdottir, M. Ghandi, J. P. Mesirov, P. Tamayo, The molecular
251–264 (2013). signatures database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015).
68. K. Mojumdar, F. Liang, C. Giordano, C. Lemaire, G. Danialou, T. Okazaki, J. Bourdon, M. Rafei,
J. Galipeau, M. Divangahi, B. J. Petrof, Inflammatory monocytes promote progression of Acknowledgments: We thank the UCI Institute for Immunology flow cytometry core for access
Duchenne muscular dystrophy and can be therapeutically targeted via CCR2. EMBO Mol. to equipment and the UCI Institute for Clinical and Translational Science for resources that
Med. 6, 1476–1492 (2014). supported this study. Part of the RNAseq studies was made possible through access to the
Genomics High Throughput Facility Shared Resource of the Cancer Center Support Grant

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 16 of 17

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

(P30CA-062203) at UCI. We also thank the NSF-Simons Center for Multiscale Cell Fate Research advisory role for and/or receiving research funds from Alexion, Amicus, Argenx, Arvinas,
for their support of the scRNAseq studies through a student opportunity award. Funding: Audentes, AvroBio, Horizon Therapeutics, Immunovant, Maze Therapeutics, Momenta (now
Research reported in this publication was supported by the National Institutes of Neurological Janssen), Sanofi-Genzyme, Sarepta, Spark Therapeutics, UCB, and Modis/Zogenix. I.M. also
Disorders and Stroke (NINDS) grant R01NS120060 and National Center for Advancing serves on the data safety monitoring board for Acceleron, Avexis, and Sarepta. D.R.G., Q.S., and
Translational Sciences (NCATS) grant KL2TR001416 (to S.A.V.); National Institute of General M.S. are employees and shareholders of Regeneron Pharmaceuticals. M.J.S. is a cofounder of
Medical Sciences (NIGMS) grant R01GM143536 (to J.R.Z.); National Institute of Arthritis and MyoGene Bio and SkyGene Bio. All other authors declare that they have no competing interests.
Musculoskeletal and Skin Diseases (NIAMS) grant U54 AR052646-07 (to M.J.S.); NIAMS grant Data and materials availability: All data needed to evaluate the conclusions in the paper are
R01AR055686 (to J.KJ.); NIAMS grant R01AR078340 (to T.M.); Department of Defense (DoD) present in the paper and/or the Supplementary Materials. scRNAseq datasets for mdx muscle
Congressionally Directed Medical Research Program (CDMRP) Discovery Award mononuclear cells (Fig. 4) and healthy and dystrophic muscle macrophage (Fig. 1 and fig. S10)
W81XWH1910012 (to G.C.); National Institute of Allergy and Infectious Diseases (NIAID) grant have been deposited in the Gene Expression Omnibus (GEO) database under accession number
1R01AI168063-01 (to S.O.); NIAMS grant P30AR075047, National Science Foundation (NSF) GSE226173. The WT muscle dataset, GSE138826, was used for reference-based mapping of
grant DMS11763272, and a Simons Foundation grant 594598 (to Q. Nie.); and NIAID training macrophages, FAP, and tenocytes (Fig. 4). The spatial RNAseq datasets accession number is
grant award T32AI060573 (to J.M.K.). Author contributions: Conceptualization: G.C., J.M.K., GSE225766. PCA was performed by comparing microglia RNAseq dataset, GSE132877, bone
J.R.Z., and S.A.V. Methodology: G.C., J.M.K., D.J., Q.N., N.P., Q.S., T.M., M.S., J.R.Z., and S.A.V. marrow–derived macrophages RNAseq dataset (http://kbcommons.org/system/browse/diff_
Investigation: G.C., D.J., C.G.J., J.M.K., P.K.F., Q.N., N.P., K.M., L.T., S.M., B.S.H., J.D., A.S., E.M., M.D., exp/MusMusculus) with our muscle macrophage bulk RNAseq datasets (GSE225593).
S.O., T.J., C.P.H., A.L.M., Q. Nie, I.M., V.K., L.M.W, S.Q.H., T.M., M.W.H., S.B., J.K.J., K.K., Q.S., D.R.G.,
M.S., M.J.S., J.R.Z., and S.A.V. Visualization: G.C., J.M.K., M.S., J.R.Z., and S.A.V. Funding acquisition: Submitted 18 July 2022
T.M., M.J.S., J.R.Z., and S.A.V. Project administration: S.A.V. Supervision: J.R.Z. and S.A.V. Data Accepted 5 June 2023
curation: J.M.K., C.G.J., J.R.Z., and Q.S. Statistical and other formal analysis: G.C., J.M.K., C.G.J., Published 7 July 2023
A.L.M., Q.S., and J.R.Z. Writing—original draft: G.C., M.S., J.R.Z., and S.A.V. Writing—review and 10.1126/sciadv.add9984
editing: G.C., D.J., P.K.F., T.M., M.S., M.J.S., J.R.Z., and S.A.V. Competing interests: I.M. discloses an

Coulis et al., Sci. Adv. 9, eadd9984 (2023) 7 July 2023 17 of 17