Tuning Macrophage Phenotype to Mitigate Skeletal Muscle Fibrosis Free

Ischemia can result secondary to trauma or vascular disease and is a surgical emergency requiring intervention to prevent ongoing muscle loss, rhabdomyolysis, and loss of extremity function (1, 2). Additionally, thousands of surgeries are performed daily in which ischemia is intentionally induced by the use of a tourniquet to limit blood loss (3). Ischemic muscle injury occurs across a wide variety of extremity injuries and often results in muscle fibrosis, which is the result of aberrant collagen deposition (1).

The response to skeletal muscle injury is a tightly coordinated process that balances the infiltration of neutrophils and macrophages, clearance of damaged myocytes, and satellite cell activation and differentiation to regenerate muscle fibers with reconstitution of the extracellular matrix (ECM) architecture (2, 4, 5). Perturbation of this process can result in skeletal muscle fibrosis, a pathological response to injury that results in replacement of damaged myocytes with stromal elements (collagen I, III, and IV) that do not contribute to muscle contraction and can result in significant motor impairment and loss of function (6, 7).

Following an acute muscular injury, elaboration of ECM elements serve as a temporary stop-gap measure to fill in damaged muscular elements (7, 8). Fibroadipogenic progenitor (FAP) transformation and fibroblast proliferation and collagen I and III deposition serve as templates for ingrowth of regenerated nerves, vasculature, and satellite cell differentiation into new myocytes (4, 5). In the remodeling phase of injury, this exuberant ECM is broken down by matrix metalloproteinases, allowing for reformation of normal muscle architecture (4, 9).

Macrophages have been implicated as mediators of both muscle regeneration and fibrosis. Within an injured muscle, resident and infiltrating macrophages clear necrotic myofibers and release proinflammatory cytokines that direct stromal remodeling and differentiation of muscle stem cells toward regeneration (10–14). After the initial inflammatory phase, macrophages shift to the anti-inflammatory and profibrotic M2c subtype directing resolution of the inflammatory response and tissue repair characterized by increased production of IL-10 and TGF-β1 (5, 10, 15). Targeting macrophages at each of these phases can significantly alter the regeneration/fibrosis balance following injury (16). Global macrophage depletion has been demonstrated to significantly impair muscle regeneration with markedly reduced clearance of necrotic myofibers and the subsequent development of severe fibrosis (17, 18). Deletion of the proinflammatory cytokine IL-6 prevents macrophage recruitment to the site of injury with impairment of muscle regeneration (19). Suppression of macrophage infiltration into injured muscle also results in increased fibrosis (10, 18, 20). Each of these findings clearly demonstrate a crucial role for the presence of macrophages within healing muscle for regeneration of injured myofibers.

FAPs within skeletal muscle are an important population that respond to signals from macrophages and are capable of differentiation into adipocytes or myofibroblasts (4, 21). FAPs proliferate in response to injury and differentiate into myofibroblasts to contribute to ECM deposition and wound contracture (22, 23). Their population wanes and returns to baseline levels as acute injury resolves (2). Urciuolo and coworkers demonstrated that the remodeling of temporary collagen I, III, and IV followed by organized collagen VI produced in this response is crucial for directing the proliferation and ingrowth of muscle satellite cells into a hospitable niche within the site of injury (24, 25). A close interplay between FAPs and satellite cells directs a coordinated ECM deposition and ECM remodeling exchange that allows differentiation of satellite cells into myocytes within an ECM envelope. In fibrotic conditions, however, FAPs differentiate into both myofibroblasts and adipocytes, and ECM produced during the acute phase is not infiltrated by regenerative cells due to satellite cell differentiation and proliferation being suppressed by TGF-β1 (26, 27). ECM elements persist and FAPs continue to proliferate and differentiate into myofibroblasts and adipocytes (4, 22).

The TGF-β family of signaling molecules is composed of a diverse group of cytokines and growth factors that contribute to wound healing and fibrosis in different tissues. TGF-β1 is the best characterized among them, functioning as an anti-inflammatory and profibrotic cytokine that has been demonstrated to shift the muscle injury response toward fibrosis (28). TGF-β1 binds to a complex of TGF-BRI and dimers of the TGFBRII receptor families, which activate SMADs 2/3 and connective tissue growth factor (CTGF) expression, along with NF-κB inhibition and anti-inflammatory effects downstream (28–32).This activates FAP transformation into fibroblasts, fibroblast proliferation, and deposition of ECM products like collagen I and III (28, 33). In skeletal muscle injury, increased myofiber TGF-β1 expression has been shown to result in increased fibrosis whereas myofiber-specific deletion of the TGFBRII receptor is protective against fibrosis (30, 34, 35). Although many tissues have basal level expression of TGF-β1, the source of TGF-β1 in muscle injury and fibrosis is unclear, with TGF-β1 production by both macrophages and myofibers (28). Li and coworkers demonstrated that TGF-β1 production by myofibers acts in autocrine positive feedback loop to increase their TGF-β1 expression causing satellite cells to differentiate into fibrotic progenitors (23). Their study did not investigate the initiation of this cascade nor the role of other cell types beyond muscle progenitors in this TGF-β1 autocrine loop. Persistent TGF-β1 signaling in injured muscle prevents apoptosis of FAPs and encourages their differentiation into myofibroblasts and adipocytes resulting in increased ECM deposition and fibrosis (2).

Macrophages are significant sources of TGF-β1 in injury response, as macrophages are the largest cell population at a muscle injury site during the first week after injury and could serve to initiate, sustain, or resolve the profibrotic environment (15, 20). M2 macrophages shift the immune response toward tissue repair and fibrosis with marked increase in TGF-β1 production (11). This study seeks to elucidate the interplay of macrophages and TGF-β1 in the response to muscle injury in both regeneration and fibrosis. We hypothesize macrophage Tgfb1 expression increases FAP proliferation with increased ECM deposition and that genetic myeloid cell specific Tgfb1 deletion will mitigate ischemia reperfusion (IR)–induced muscle fibrosis through more rapid muscle regeneration. We use a mouse model of IR muscle injury with cardiotoxin (CTX) injection to induce fibrosis and with macrophage specific Tgfb1 deletion mice and a novel murine TGF-β1/3 ligand trap to demonstrate critical role of TGF-β1 in skeletal muscle fibrosis.

All animals were housed in standard conditions: 22°C ± 2°C, receiving 12 h of light exposure each day, with no diet restrictions. Animal care was provided in accordance with the University of Michigan School of Medicine guidelines and policies for the use of laboratory animals (PRO00007930). Animals that underwent IR injury were C57BL/6J background at 8–10 wk of age. LysM-Cre;Tgfb1^fl/fl mice bred on a C57BL/6J background [as previously demonstrated (36)] were gender and age matched to C57BL/6J controls. Mice were euthanized at specified timepoints. Additionally, for our ligand trap experiments, C57BL/6J mice were used. Euthanasia was performed with carbon dioxide and cervical dislocation per University of Michigan animal use guidelines.

Mice received preoperative treatment with 0.06 mg/kg buprenorphine s.c. and were anesthetized with 2% isofluorane inhalation. The left leg and groin were de-epilated using Nair ointment and cleansed. Ten microliters of CTX (3 mg/ml) was injected from proximal to distal along the length of the left tibialis anterior (TA) muscle. An incision was made at the groin crease, and groin fat pad was retracted to exposed the femoral vessels. Superficial femoral artery and vein were isolated from the nerve just distal to the profunda femoris takeoff and occluded with an Ackland clamp. The incision was closed temporarily, and mice were kept in a warmed anesthetic chamber for 3 h, at which point clamps were removed and incisions formally closed. Postoperatively, mice were treated with 0.06 mg/kg buprenorphine every 12 h for 24 h.

TA muscles were harvested at the described timepoints and rinsed in PBS, weighed, and embedded in OCT mounting medium and flash frozen in liquid nitrogen-cooled methylbutane. Samples were stored in blocks at −80°C until sectioning using a cryotome at 5 μm thickness.

After inductive injury, the corresponding TA muscle was dissected at the indicated timepoints. Tissue was mechanically minced and digested for 45 min in 0.3% Collagenase Type I and 0.4% Dispase II (Gibco) in RPMI medium at 37°C under constant agitation at 120 rpm. Digestions were subsequently quenched with 10% FBS RPMI and filtered through 40 μm sterile strainers. Specimens were blocked with anti-mouse CD16/32 and subsequently stained using the following Abs: FITC:Ly6C, BV510:CD11b, APC-H7:Ly6G, BB700:F4/80 (Becton Dickinson), and BV650:MHCII (BioLegend). Stained and washed samples were run on a FACSAria II for cell sorting or LSRFortessa for analysis (Becton Dickinson) and analyzed using FlowJo software.

Baseline uninjured muscle (day 0) and postsurgery day 3 TA muscles were harvested from six IR CTX injury mice and eight uninjured mice. Two animals were pooled for each lane of IR CTX animals and four pooled for each lane of uninjured mice to maximize cell return. This resulted in 3 lanes of IR CTX (n = 2 mice per lane) and 2 lanes of uninjured (n = 4 mice per lane). The harvested tissue samples were digested for 45 min in 0.3% Type 1 Collagenase and 0.4% Dispase II (Life Technologies) in RPMI 1640 medium at 37°C under constant agitation at 120 rpm. Digestions were subsequently quenched with 10% FBS RPMI and filtered through 40-μm sterile strainers. Cells were then washed in PBS with 0.04% BSA, counted, and resuspended at a concentration of ∼1000 cells/ul. Cell viability was assessed with trypan blue exclusion on a Countess II (Thermo Fisher Scientific) automated counter, and only samples with >85% viability were processed for further sequencing.

Single-cell 3′ library generation was performed on the 10x Genomics Chromium Controller following the manufacturers protocol for the v2 reagent kit (10x Genomics, Pleasanton, CA). Following generation of single-cell gel bead-in-emulsions (GEMs), reverse transcription was performed and the resulting Post GEM-RT product was cleaned with DynaBeads MyOne Silane beads (Thermo Fisher Scientific, Waltham, MA). The cDNA was amplified, SPRIselect (Beckman Coulter, Brea, CA) quantified, then enzymatically fragmented and size selected using SPRIselect beads to optimize the cDNA amplicon size prior to library construction. Double-sided SPRI bead cleanup was performed after end repair and A-tailing, and single-sided cleanup is done after adapter ligation. Indexes were added during PCR amplification, and a final double-sided SPRI cleanup was performed. Libraries were quantified by Kapa quantitative PCR for Illumina Adapters (Roche), and size was determined by Agilent TapeStation D1000 tapes. Read 1 primer sequence are added to the molecules during GEM incubation. P5, P7, and sample index and read 2 primer sequence are added during library construction via end repair, A-tailing, adaptor ligation, and PCR. Libraries were generated with unique sample indices for each sample. Libraries were sequenced on a HiSeq 4000, (Illumina, San Diego, CA) using a HiSeq 4000 PE Cluster Kit (PN PE-410-1001) with HiSeq 4000 SBS Kit (100 cycles, PN FC-410-1002) reagents, loaded at 200 pM following Illumina’s denaturing and dilution recommendations. The run configuration was 26 × 8 × 98 cycles for Read 1, Index, and Read 2, respectively. Cell Ranger Single Cell Software Suite 1.3 was used to perform sample de-multiplexing, barcode processing, and single-cell gene counting (alignment, barcoding, and unique molecular identifier (UMI) count) at the University of Michigan Biomedical Core Facilities DNA Sequencing Core.

A total of ∼200 million reads were generated from the 10x Genomics sequencing analysis for each of the replicates. The sequencing data were first preprocessed using the 10x Genomics software Cell Ranger (10x Genomics) and aligned to mm10 genome (deposited to the Gene Expression Omnibus database: GSE144270 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE144270). For quality control, genes expressed in fewer than 10 cells and cells expressing fewer than 500 genes per cell, more than 20% mitochondrial UMI content, or UMI content >60,000 were filtered out. Replicates from the same group (condition) were pooled together for downstream analysis. The case group consisted of 7718 cells and 14,466 genes (before quality control, 7718 cells and 19,333 genes), whereas the control group consisted of 2760 cells and 12,636 genes (before quality control, 3520 cells and 17,730 genes). Downstream analysis steps were performed using Seurat pipeline (37). The downstream analysis steps for each sample type include normalization, scaling, dimensionality reduction (principal component analysis [PCA] and t-distributed stochastic neighbor embedding [t-SNE]), unsupervised clustering, cluster consolidation via centroid rank correlation analysis, discovery of differentially expressed cluster specific markers, and marker enrichment.

Presence of replicate batch effect was excluded by visual inspection of the contribution of each replicate to the PCA and t-SNE projections of the group. First, provisional clusters were assigned via unsupervised clustering (Seurat FindClusters, Louvain algorithm, k = 30, resolution = 0.4). This procedure led to 21 provisional clusters for the case group and 17 provisional clusters for the control group. Provisional clusters were aligned according to the rank correlation of their centroids, measured on the gene set derived from the intersection of the genes for each set (12,619 genes). Finally, (consolidated) clusters were obtained by aggregating similar clusters based on centroid rank correlation analysis. Clusters appear to be well distinguished in the PCA and t-SNE projections (Supplemental Fig. 1).

Gene ontology (GO) term and KEGG pathway analysis resulting from LRpath (38) (directional test, sign of the fold change), along with characteristic expression profiles, were used to identify the cell types of each cluster. Bonferroni adjustments were performed to determine statistical significance of gene expression fold changes within a cluster versus all other clusters in the same group. To characterize cluster phenotypes, heatmaps generalizing the degree of gene family expression were generated by considering the normalized expression for all individual genes of each gene family. These sums were calculated for each individual cell and averaged across all cells in each identified cluster. Resultant values were then scaled within gene families (e.g., “skeletal muscle, wasting, and atrophy”) across all clusters. Subanalyses within each gene family were performed by considering for each cluster the sum of the UMI fraction per cell, to reflect the contribution of each cluster to the overall signal of each specific gene in the injured and uninjured groups. Cluster subanalyses were performed by identifying a cohort of cells with positive gene expression for gene A and further stratification by gene B.

Consecutive tissue sections were stained using H&E, picrosirius red, and Masson’s trichrome stain protocols performed in biological triplicate and technical triplicate with sequential sections for each stain type. For quantification, slides were selected representing the midbelly of the TA muscle, avoiding the distal and proximal tendinous aponeuroses to standardize muscle architecture. Within this area of interest, proximal, middle, and distal sections were stained. Biological replicates of three animals each were used per group and stain type.

Fresh frozen sections were treated for 30 min in Bouin’s solution at 56°C, then rinsed in 70% ethanol for 5 min. Slides were stained for 10 min in Gill’s hematoxylin solution, rinsed in tap water, then differentiated for 5 s in acid alcohol. Slides were then stained for 4 min in eosin solution, followed by alcohol gradient and xylene clearing for subsequent mounting.

Slides were treated similarly to H&E with Bouin’s and alcohol rinse. They were then stained for 15 min in Weigert’s hematoxylin followed by acid alcohol differentiation. Slides were stained for 10 min in Biebrich scarlet, followed by phosphotungstic/molybdic acid bath for 15 min. Slides were rinsed and stained in aniline blue for 10 min, then rinsed in tap water and alcohol gradient, cleared with xylene, and mounted.

Slides were air dried for 20 min at 25°C, then placed in xylene for 10 min, followed by decreasing alcohol gradient (100, 95, 80, and 70%). Slides were set in PBS for 2 min to remove residual OCT embedding medium, then rinsed in tap water. Slides were laid flat (sample side up) in a flat dish and covered in picrosirius red stain solution and allowed to stain for 30 min. Slides were rinsed in 0.5% acetic acid solution for 2 min twice, then passed through increasing alcohol gradient, cleared in xylene, and mounted (39).

Picrosirius slides were imaged on a Nikon 11,530 light microscope with standardized exposure of 2.9 s on ×10 and 11.7 s on ×20 magnifications. Within a sample, 5 separate noncontiguous fields were imaged. TIFF files were analyzed using ImageJ software. Images were binarized into 8-bit images and autothresholding performed using the Huang threshold, which accurately represented positive staining and contrast levels. Using scaled measures from a standardized scale bar, stained area was measured in μm². These were averaged in GraphPad Prism and subjected to two-way ANOVA.

Brightfield images of Picrosirius-stained histological specimens were quantified using a novel image processing pipeline. First, Picrosirius images were processed by a color deconvolution algorithm to yield separate images of muscle fibers and collagen. Next, the separated collagen images were denoised and processed through shape-based binarization filters to yield “skeletonized” tracings of individual fibers and branchpoints. A panel of 13 relevant fiber and branchpoint parameters (e.g., fiber length, width, persistence, global alignment, branchpoint density) was then quantified. Comparison of injured and uninjured wild-type (WT)/knockout and untreated/treated specimens in two-dimensional plots was achieved by t-SNE clustering.

Immunofluorescence images stained for Laminin (AbCam, Boston, MA), embryonic myosin H chain (eMHC) (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), and DAPI were obtained in three separate noncontiguous fields per image at ×20 magnification. Image J software was used to split RGB channels. The laminin channel was analyzed for fiber detection using Myovision muscle analysis software provided by the University of Kentucky (40). Cross-sectional area and Feret diameter measurements were exported and graphed as frequency of distribution using GraphPad Prism. This was performed for n = 3–5 animals per group using three separate fields per slide. Positive eMHC staining was quantified using the separated red channel of the same images used for laminin analysis. ImageJ software was used to threshold using the Yen protocol, and fluorescent area was measured.

Histologic evaluation was performed at indicated time points in TA muscles from WT C57BL/6J (both injured and contralateral uninjured TAs), LysM-Cre;Tgfb1^fl/fl mutants on a C57BL/6J background, and TGFBRII-FC treated (WT) mice following inductive ischemia/reperfusion injury and concomitant CTX at indicated timepoints. Muscle specimens were flash frozen and sectioned as described above, subsequently mounted on Superfrost Plus slides (Thermo Fisher Scientific), and stored at −20°C. Sections were blocked and permeabilized with 2% serum, 1% BSA, 0.1% fish skin gelatin, 0.001% Triton X-100, and 0.0005% Tween 20. Sections were incubated overnight with the following Abs: F4/80 (Abcam), PDGFRα (R&D), TGF-β1 (Novus), pSMAD3 (Novus), Ly6G (Abcam), laminin 1/2 (Abcam), eMHC (Developmental Studies Hybridoma Bank), perilipin-1 (Cell Signaling Technology), anti-15 Lipoxygenase 1 (ALOX15) (Abcam), and KI67 (Abcam). Fluorophore conjugated secondary Abs were diluted to 1:200. Nuclear counterstain was performed with Hoechst 33342 (Life Technologies). Appropriate primary Ab-negative controls were run simultaneously with each tested sample.

Cryosections for immunolabeled samples were performed on a Leica TCS SP8 Laser Scanning Confocal Microscope with tunable laser whose excitation and emission parameters were set to the fluorophore manufacturer’s instructions. Each site was imaged in all channels and quantified in ImageJ. Images were adjusted linearly only in brightness and contrast identically across comparison groups for clarity where necessary.

For quantification of positively fluorescence, images were automatically thresholded via Triangle, Yen, or Otsu method and then measured across 1–3 separate high-power fields per biologic specimen (ImageJ, n = 1–3 per specimen, n = 3–5 specimens per treatment group). Images were then analyzed for area via the standard ImageJ measurement panel.

C57BL/6J mice (n = 9) received murine TGFBRII-mFC (Acceleron), consisting of the TGFBRII extracellular domain attached to murine IgG2a-Fc group, at 10mg/kg s.c. immediately following clamp removal after IR/CTX procedure described above (postoperative day [POD]0) and every 3 d thereafter (POD3, POD6, POD9, and POD12). Control mice were treated with PBS injection. Three mice from each group were sacrificed at POD3, POD7, and POD14.

Bone marrow macrophages were harvested from six age- and sex-matched LysM-Cre;R26^{mTmG +/−} mice through tibia and femur harvest. Long bones were placed vertically oriented in microcentrifuge tubes and briefly spun at 500 rpm to allow centripetal egress of medullary bone marrow to the base of the tubes. Lympholyte gradient was used to separate buffy coat cells, and these were counted then incubated with FITC-conjugated anti-CD3, CD19, Ly6G, and CD11c Abs. These were then incubated with magnetic beads conjugated to anti-FITC Ab and passed through magnetic columns for negative selection. The flow-through cells were then washed and incubated with FITC-conjugated anti-CD11b Abs followed by magnetic anti-FITC beads. These were passed through new columns and the flow through was discarded. Columns were removed from the magnet bar and washed to isolate bound cells. These were washed and resuspended in PBS at 15 million cells/ml. 100 uL of cell suspension (1.5 million cells) were injected by tail vein into LysM-Cre;Tgfb1^fl/fl age- and sex-matched mice (n = 3). The animals were allowed to equilibrate for 24 h, then underwent IR/CTX injury. Animals were sacrificed at POD7 and TA muscles were harvested as described above. Histological sections were stained for light microscopy as described above, and immunofluorescence sections were stained with only DAPI.

All statistical parameters are reported in the 20Results section and figure legends giving the numbers of animals per sample, the mean, and the p values of statistical tests. Statistical tests were performed in GraphPad Prism 8.0 software. Histological quantification of stained images was subject to Student t test, or ANOVA with multiple comparisons posttest. Cross-sectional area and Feret diameter were represented as frequency of distribution graphed in percent frequency and compared using Komolgorov–Smirnov analysis. Single-cell sequencing statistical methods are described in their respective methods sections.

IR is one of the most common injuries to which skeletal muscle is exposed, given the high prevalence of tourniquet usage and large number of patients with vascular disease and injury. However, most patients who are exposed to an IR insult also have a concomitant muscle injury. Thus, we first set out to characterize the difference between an IR injury and an IR injury plus a muscle injury. In mice that were exposed to a IR injury, we observed centralized nuclei and rounded myofiber shape, indicating a prolonged regeneration with moderate inflammatory cell infiltrate (Fig. 1A). With combined IR/CTX injury, however, we observed a stark contrast with decrease in myofiber size as well as rich inflammatory cell infiltration around the muscle fibers, indicating a more permanent damage (Fig. 1A).

This damage was further characterized by a profound increase in the amount of disorganized ECM, as demonstrated by positively staining collagen by picrosirius red, demonstrating a distinct, robust fibrotic response unique to the CTX injury with concomitant ischemia and reperfusion to the TA muscle. Thus, to better assess the role of myeloid cells on muscle fibrosis and regeneration, we demonstrate that IR/CTX is a more representative model.

Polytrauma in the form of IR/CTX injury generated a fibrotic response absent in CTX alone. Given the profound enrichment of inflammatory infiltrate we observed by histology, we aimed to characterize the inflammatory response that occurs after polytrauma via flow cytometry to quantify the recruitment of myeloid circulatory cells to the site of injury. We observed that neutrophils and monocytes (CD11b⁺Ly6G⁺ and CD11b⁺Ly6G⁻ respectively, Fig. 1B) were differentially recruited between the injury groups, with IR/CTX inducing a markedly protracted course of inflammation than either injury alone.

Neutrophil infiltration peaked at day 3 and rapidly decreased thereafter, with more prolonged presence in the combined IR/CTX group (Fig. 1C). Similarly, monocytes peaked at day 3 in all groups. However, whereas monocytes were almost entirely gone in the CTX only group by 2 wk, we observed a significantly prolonged presence of monocytes in the IR/CTX group (Fig. 1D). This corresponded to increased presence of F4/80+ mature macrophages in the muscle after polytrauma versus either injury alone (Fig. 1E). These findings suggest that polytrauma sustains a prolonged inflammatory environment that likely contributes to regulation of the regenerative pathway following injury.

Having established a unique fibrotic phenotype and sustained intense inflammation in the IR/CTX muscle, we performed a directed single-cell RNA sequencing experiment on this injury model to further characterize the cellular constituents and behavior at the transcriptional level postpolytrauma. We analyzed TA muscle harvested from C57BL/6J mice that were uninjured (n = 6) and IR/CTX (n = 6) at 3 d postinjury.

We first performed a completely blinded analysis of TA tissues by surveying the global expression of Tgfb1 for uninjured versus IR/CTX injured muscles. We observed a marked enrichment of Tgfb1 in IR/CTX muscles compared with uninjured control (Fig. 2A). Population identification was performed using previously described marker enrichment.

To characterize the perturbation from baseline caused by inductive IR/CTX, clusters were first generated unsupervised (Fig. 2B), and subsequently rank correlated by centroids to compare clusters between treatment groups (Fig. 2C). These comparisons suggested putative cluster relationships or intermediate phenotypes that were determined by enrichment terms generated via GO terms and KEGG pathway analyses.

The resulting 11 and 8 cluster families from uninjured and IR/CTX muscles, respectively, were identified via these GO terms and KEGG pathway analyses in addition to characteristic gene expression (Supplemental Fig. 2): macrophages (Ccr2, Adgre1, Csf1r, Mrc1), dendritic cells (Cd209a, Kldr1), lymphocytes (Ms4a1), granulocytes (Ccr1, Csf3r), endothelial cells/vascular progenitor cells (Kdr, Cdh5, Pecam1, Tek), smooth muscle cells (Acta2, Mylk, Myh1), Schwann cells/neural progenitors (Plp1, Cnp, Dhh, Mbp, Mpz, Sox10), FAPs (Pdgfra, Prrx1, Osr2, Col1a1), scleraxis expressing mesenchymal stem cells (Scx, Tnmd), Satellite cells (Pax7, Chodl, Sdc4), skeletal muscle (Myf6, Actn3, Pgam2), and myoblasts (Myod1, Myf5). Notably, Col1a1 and Pdgfra coexpression highlighted FAPs with a higher number seen in IR/CTX FAPs compared with uninjured controls (Fig. 2D). This overlap suggested a responding cell within the injury niche likely responsible for the fibrosis observed above. Furthermore, given the extensive role of TGF-β1 in the regulation of tissue fibrosis, we surveyed and found that the Tgfb1 expression observed previously was present within the predominant myeloid (macrophage [Adgre1 i.e., F4/80], dendritic cell, and granulocyte [Csf3r]) clusters of the IR/CTX groups at significantly higher levels than in the uninjured muscle clusters (Fig. 2E).

Although Tgfb1 was observed also in macrophages, lymphocytes, and a large population of endothelial/vascular progenitor cells, the relatively smaller captured endothelial/vascular progenitor cell cluster in the injured muscle demonstrate the large influx of recruited myeloid cells, and denote a much larger pool for Tgfb1 expression than observed in quiescent muscle. This represents an important role for myeloid lineage derived Tgfb1 expression within postpolytrauma muscle.

Our initial analysis demonstrated an important role for TGF-β1 signaling, and we sought to perform more detailed analyses of the cellular constituents of the IR/CTX injury with RNA sequencing. The overexpression for genes in each gene family (fibrosis and TGF-β1 signaling) were calculated for all clusters. Two major regions of interest were identified from the resultant heatmap (Fig. 3A): 1) TGF-β1 signaling gene families between uninjured (cluster ctrl.A) and IR/CTX (clusters case.A and case.C) myeloid cells and 2) fibrosis gene families between uninjured (cont.I) and IR/CTX (case.G) FAPs.

For further resolution, we subsequently performed subanalyses of each individual gene within a given family. Across all analyzed terms, both injured macrophages and FAPs were enriched in fibrotic (Fig. 3B) genes as well as TGF-β1 signaling (Fig. 3C) genes.

Malecova and coworkers (41) have demonstrated that two populations of FAPs are present within developing and injured muscle based on Vcam-1 and Tie2 expression profiles. Namely, they found that in mature muscle, Vcam-1–positive FAPs indicate a shift to a fibrotic response to injury. In our sequencing data, we identified Tie2 expression–positive FAPs in both the uninjured and IR/CTX muscles (128 versus 91 cells, respectively). Among these groups we found that 74.7% of the IR/CTX Tie2-positive FAPs were Vcam-1–positive compared with only 5.4% in the uninjured muscle (Fig. 3D), demonstrating that in addition to increased population, Pdgfra-positive FAPs also exhibit a phenotypic shift after IR/CTX injury toward a profibrotic phenotype.

The sustained pool of inflammation serves as a source of copious TGF-β1 that is likely acting on macrophages both in an autocrine and paracrine fashion. The nearby macrophages and extracellular TGF-β1 are associated with a phenotypic shift in FAPs as well as increased fibrotic signaling, contributing synergistically to the fibrotic phenotype.

Having established the presence of macrophage Tgfb1 signaling and FAP fibrosis, we next evaluated the effect of conditional myeloid-specific deletion of Tgfb1 using the LysMCre flox system, which specifically targets circulating myeloid populations (42), including macrophages, dendritic cells, and granulocytes, the same cells expressing Tgfb1 in our transcriptomic data. We performed IR/CTX injury in C57BL/6J mice and LysM-Cre;Tgfb1^fl/fl mice on a C57BL/6J background and harvested left (injured) and right (uninjured) TA muscles for flash freezing and histological analysis at 3 and 7 d postinjury.

Brightfield microscopy demonstrated complete resolution of muscle fibrosis in the conditional deletion (LysM-Cre;Tgfb1^fl/fl) mice, rescuing the muscle to noninjured morphology by 7 d (Fig. 4A): picrosirius red staining quantification shows a significantly higher mean area of collagen staining within the injured age-, strain-, and gender-matched control muscle compared with the LysM-Cre;Tgfb1^fl/fl animals (Fig. 4B). We further validated this finding using a machine learning artificial intelligence algorithm, which found that the LysM-Cre;Tgfb1^fl/fl mice appeared similar to the uninjured mice in the amount of fibrosis (Supplemental Fig. 3A).

To assess if this antifibrotic phenotype in LysM-Cre;Tgfb1^fl/fl mice was driven by an alteration in the inflammatory response, we performed additional histology at day 3. These sections demonstrated robust inflammatory cell infiltrate within both deletion and control injured muscle at the earlier timepoint, with prominent muscle edema distorting normal muscle architecture (Fig. 4C) with comparable macrophage infiltration (Fig. 4D) and similar fiber morphology (Fig. 4E).

Although immunologically the two treatment groups appeared similar, the conditional deletion mice were found to express much higher levels of embryonal myosin H chain (eMHC) versus control suggesting a more rapid regenerative state (Fig. 4F) and potentially faster healing. Indeed, in re-examination of day 7 histology, we observed complete resolution of F4/80⁺ macrophages (Fig. 4G) along with cessation of eMHC expression to near-uninjured levels, whereas the WT mice began exhibiting positive eMHC labeling (Fig. 4H). Morphologically, the median cross-sectional area of WT injured myofibers was significantly reduced compared with both control uninjured muscle and LysM-Cre;Tgfb1^fl/fl injured muscle. Median Feret diameter was similarly decreased in control injured muscle compared with WT uninjured and LysMCre;Tgfb1^fl/fl injured muscle (Fig. 4I), consistent with a more rapid and robust regeneration in the early phase following injury to complete regeneration in the conditional deletion compared with WT animals.

Having demonstrated that selective deletion of TGF-β1 from LysM myeloid cells can mitigate fibrosis, we hypothesized that this effect was regulated directly by the circulating macrophages recruited to the site of injury, instead of a more systemic resistance to inflammation. To confirm this, we performed adoptive transfer experiments in which endogenously labeled bone marrow macrophages from LysM-GFP mice were harvested and enriched and injected into LysM-Cre;Tgfb1^fl/fl mice who previously were demonstrated to be rapid healers, with complete resolution of fibrosis by the day 7 timepoint. These mice injected with macrophages of full fibrotic potential (i.e., intact Tgfb1 expression) were observed to recapitulate the degree of fibrosis seen in WT animals (Fig. 5A), with systemically injected GFP⁺ myeloid cells found at the site of injury in the recipient mouse (Fig. 5B), confirming the phenotype being governed by the myeloid infiltrate at the cellular level.

Our RNA sequencing data suggested Tgfb1 activity exerts effects at least partly in an autocrine fashion. Given the findings of our genetic lineage deletion and adoptive transfer experiments, we hypothesized that sequestration of TGF-β1 would also demonstrate protective effects. We set out to analyze if a clinically relevant ligand trap could similarly decrease traumatic muscle fibrosis, given that no macrophage specific therapies are currently available. Although this therapeutic is not cell specific, the majority of the TGF-β1 present at the injury site during the treatment period derived from macrophages.

Mice were treated with TGFBRII-Fc at the time of IR/CTX injury and every 3 d until euthanasia. Similar to LysM-Cre;Tgfb1^fl/fl mice, ligand trap treated mice also demonstrated significantly less muscle fibrosis by H&E and picrosirius stain (Fig. 6A, 6B) and appeared more like an uninjured muscle by artificial intelligence (Supplemental Fig. 3B). Additionally, we observed increased Feret diameter similar to uninjured muscle and similar resolution of eMHC labeling regeneration (Fig. 6C, 6D), akin to what we observed in the conditional deletion mice, verifying the utility of a clinical trap for TGF-β1 in mitigating posttraumatic muscular fibrosis.

Interestingly, we demonstrated through flow cytometry analysis that the recruited cell populations remained the same between vehicle treated and ligand trap treated animals (Supplemental Fig. 4A). Neutrophil, macrophage, Ly6C, and F4-80 expression remained the same between vehicle-treated and ligand trap–treated animals at day 0, 3, 7, and 14 after injury, suggesting that TGF-β1 exerts its fibrotic effects within the end organ tissue rather than preventing macrophage recruitment of migration.

We found previously that the PDGFRα FAPs had significant increase in profibrotic gene profiles and ECM element gene expression coupled to increased downstream TGF-β1–responsive gene expression after IR/CTX injury compared with control (Fig. 3B, 3C). Coupled with increased Tgfβ1 expression by infiltrating macrophages, this suggests that FAPs are responding to macrophage-produced TGF-β1 and increasing their ECM deposition in response. With this in mind we looked histologically at PDGFRα cells in our LysM-Cre;Tgfb1^{fl/fl l} and control mice and found relatively normal architecture in the LysM-Cre;Tgfb1^fl/fl mice without surrounding ECM deposition and fibrosis.

In contrast, we found that PDGFRα was much more prominent and disorganized in the control mice with increased collagen deposition surrounding these cells.

We noted increased PDGFRα staining in WT injured muscle compared with control uninjured muscle, LysM-Cre;Tgfb1^fl/fl injured muscle and TGFBRII-FC treated muscle at 1 wk (Fig. 6E) indicating proliferation of FAPs within the control injured muscle. The pattern of PDGFRα staining is also markedly more organized within the interstitial space of LysM-Cre;Tgfb1^fl/fl and TGFBRII-FC–treated injured muscle compared with control injured muscle, which demonstrates disorganized broad swaths of PDGFRα staining (Fig. 6E). We then compared positive PDGFRα staining with Ki67 staining; control injured animals demonstrated clear association between the two stains, indicating that FAPs were the proliferating cell population whereas there was poor concordance in the FAP/proliferation correlation plots of LysM-Cre;Tgfb1^fl/fl and TGFBRIIFC–treated animals (Fig. 6F).

This is an important indicator that proliferation of FAPs can be significantly reduced following injury in the absence of TGF-β1.This difference in fibrosis was not accompanied by a change in adipogenesis as quantified by perilipin immunofluorescence (IF) staining (Supplemental Fig. 4B). Thus, Tgfb1 deletion in myeloid cells does not alter FAP differentiation but rather their proliferation and consequent ECM deposition.

In our study we demonstrate that the IR and CTX model of muscle injury produces a robust inflammatory reaction that results in ECM deposition, myonecrosis, and ultimately fibrosis. Our use and analysis of single-cell RNA sequencing within the context of muscular polytrauma is to our knowledge novel, facilitating the identification of myeloid cells that were an important source of TGF-β1 following muscle injury and FAPs as the likely recipient cells responding to TGF-β1. We demonstrated TGF-β1 produced by myeloid cells is crucial to the immune response to muscle injury. Using a myeloid-specific deletion of TGF-β1, we were able to demonstrate that conditional deletion of this signal in animals experience a robust inflammatory and regenerative response that resolves more quickly than control animals with restoration of muscle architecture and myofiber integrity. This expedited response results in less collateral inflammatory damage to the muscle and subsequent ECM deposition following IR/CTX. Importantly, we found that the absence of myeloid TGF-β1 reduces the proliferation of fibro/adipogenic progenitors within the muscle responsible for the elaboration of ECM elements that become fibrosis. Finally, we showed that treatment with TGFBRII-FC, a TGF-β1/3 ligand trap, results in a similar improved resolution of inflammation and expedited regeneration with a concomitant decrease in ECM deposition.

Our model of muscle injury and ischemia is notably unique in its generation of a robust and protracted inflammatory response that results in ECM deposition and fibrosis that persists. Additionally, to our knowledge, we are also the first to characterize interesting cellular populations with this resolution in the muscle such as Schwann cells/neural progenitors, scleraxis positive MSCs, vascular progenitor cells, and myoblasts that all seem to demonstrate interrelatedness (per ranked correlation comparisons), suggesting either fluidity and free transition in the muscular context and at the very least, highlighting less colloquial muscular subpopulations that may play a regulatory role in aberrant muscle healing. Clinically, ischemic injury to the extremities is often accompanied by traumatic muscle injury. CTX offers a muscle injury of reproducible severity we can apply to an easily isolated and well-characterized muscle group. Whereas other models of muscle injury (crush, volumetric muscle loss, cryoinjury) may represent injury close to that seen clinically with IR and CTX is easily standardized and applied to specific muscle groups to allow targeted study of muscle injury.

Our single-cell RNA sequencing data reveals that infiltrating macrophages become a substantial population within the injured muscle at 3 d and they are expressing high levels of Tgfb1 in the IR/CTX model. Of note, although granulocytes (cluster IRCTX.C) on average exhibited resemblance to macrophage populations (cluster IRCTX.A) in the injured muscle (Fig. 3A), when examining total transcripts on aggregate, macrophages were clearly the major myeloid source of fibrotic signaling and Tgfb1 signaling (Fig. 3B, 3C), further reinforcing the critical regulatory role of the postinjury macrophage. We also found a corresponding rise in downstream gene expression both within the macrophages and the TGF-β1–responsive genes in FAPs. This suggests an interplay between the myeloid-produced TGF-β1 and the FAP response seen in our fibrosis injury model. Malecova and coworkers (41) recently used single-cell sequencing to identify the interaction between FAPs and macrophage populations that separate FAPs into a profibrotic lineage and regenerative lineage. In our study, we use subpopulation network analysis to follow the signaling pathways from our myeloid cells to the FAP cells within the muscle and demonstrate that in the IR/CTX model FAPs are responding to TGF-β1. We also found that the profibrotic FAP lineage expressing Tek (Tie2) and Vcam1 are significantly elevated in IR/CTX injury. A limitation of single-cell sequencing in muscle is the missing myocyte population. As myocytes are too large to sequence in the current systems, the myocyte response and production of TGF-β1 is absent in our single-cell data (43). However, the dramatic effects we saw from myeloid-specific deletion of Tgfb1 do not rely on myocyte produced TGF-β1, as myocyte expression of Tgfb1 is not affected by this conditional deletion (42). Additionally, as we did not sequence the conditional deletion or ligand trap mice, it is important to note that there likely is a transcriptionally relevant augmentation of myeloid cell phenotype that we did not capture. This is especially interesting given the enrichment in autocrine signaling gene families in the IR/CTX model that warrants further study.

Nevertheless, in our study, we were able to demonstrate a critical role of myeloid lineage–derived TGF-β1 in muscle healing and fibrosis. Our myeloid-specific Tgfb1 deletion animals were able to maintain muscle architecture following injury and appeared essentially uninjured at 7 d despite sustaining a significant insult in our IR/CTX model of muscle injury. Using a two-hit muscle injury model, we generated both fibrosis and regeneration within the same muscle. Our LysM-Cre;Tgfb1^fl/fl animals have a robust macrophage infiltrate and elevated regenerating muscle fibers present at 3 d following injury, which goes on to entirely resolve by day 7. Of note, at day 3, the control and LysM-Cre;Tgfb1^fl/fl muscle looks largely the same between the two groups, implying that the deletion animals are able to de-escalate the immune response and more efficiently repair the muscle following injury.

As Tgfb1 is a crucial mediator of the shift from macrophage M1 to M2 phenotype, the absence of Tgfb1 may serve to prevent this transition and resolve the persistence of M2 macrophages that go on to contribute to a profibrotic response. At day 3, the macrophages present are largely M1 and represent the acute inflammatory response rather than the more gradual fibrotic response. With early termination of this response, the macrophages present can effectively clear necrotic debris but not excessively efface or damage neighboring myocytes, allowing them to regenerate more efficiently. Importantly, our adoptive transfer experiments support this model as they show that macrophages capable of expressing Tgfβ1 are sufficient to restore fibrosis and muscle injury within the deletion animal muscle at 1 wk. We further corroborate this importance in this phenotypic change of macrophages at the locus of injury, with our TGFBRII-Fc experiments. Notably, TGFBRII-Fc ligand trap treatment facilitates rapid muscle degeneration even with similar numbers of inflammatory infiltrate (Supplemental Fig. 4A), whereas conditional knockout appeared to markedly attenuate inflammation. Interestingly, the reperfusion and CTX injury is a context markedly devoid of typical signals marking the resolution of inflammation including 15-lipoxygenases (ALOX15) (44). Although TA muscles regenerated in the absence of TGF-β1, there were no commensurate increases in ALOX15 (Supplemental Fig. 4C). This suggests that it is indeed the immune response to the IR and CTX injury that induces much of the muscle injury and fibrosis rather than a response intrinsic to the muscle tissue or accelerated resolution of inflammation. With full reconstitution of fibrosis and muscle injury following administration of pure macrophages, this experiment demonstrates that macrophages are the crucial cell for directing the muscle fibrosis response.

Furthermore, we found close interplay between myeloid cells and the FAPs of skeletal muscle. Our data suggests that it is this cross-talk that is mediated by TGF-β1 to alter the fibrosis phenotype and markedly improve muscle healing. The LysM-Cre system offers the most effective model to delete Tgfb1 within the circulating myeloid lineage but is not specific to macrophages versus neutrophils or other myeloid lineage cells. LysM-Cre also does not target the resident macrophage population within the muscle thus implicating inflammatory monocytes/macrophages. Our adoptive transfer experiments also support the idea that nonresident macrophages are responsible for the fibrotic phenotype as our column purification did not transfer neutrophils or resident macrophages from the LysM-Cre/R26^mTmG animals, and this was sufficient to restore the fibrotic phenotype in deletion mice. We demonstrate the reduced efficiency of TGF-β1 blockade when introducing the ligand trap, which correlates with inflammatory infiltration to baseline levels after injury. Thus, with a copious source for myeloid TGF-β1, mice treated with TGFBRII-Fc likely experience a minor subpopulation of transitioning macrophages that can elicit a discrete, albeit small, proliferative response in nearby PDGFRα FAPs. This is indeed seen in Fig. 6. Notably, this proliferative response is muted enough to still permit rapid muscle degeneration even with similar numbers of inflammatory infiltrate as seen in WT injured animals.

A crucial element within the injured muscle appears to be the organization of the muscle architecture to direct the interaction between FAPs, satellite cells, and myeloid cells. Our control animals have wide areas of redundant PDGFRα staining within the areas of fibrosis, whereas LysM-Cre;Tgfb1^fl/fl and TGFBRII-Fc animals have very narrow distribution of staining confined to the endomysium. In the control animals, this disorganization may result from the persistence inflammatory infiltrate effacing the muscle fiber structure and disrupting the cell–cell contact between these three cell types. With limited infiltrate, the deletion and treated animals may have a shorter distance for the regenerating cells to travel. Alternatively, the disorganization could be the result of enhanced FAP proliferation widening the endomysial space. Further studies of these stromal interactions may elucidate this observation.

Finally, our TGFBRII-Fc treatment does recapitulate much of the observed abrogation of muscle injury and fibrosis seen in the deletion animals. This ligand trap effect of this therapy is not confined to the myeloid cells nor the site of injury. As discussed above, injured myofibers produce TGF-β1, as do infiltrating inflammatory cells as well as myriad other cells within the muscle milieu and periphery. It is unclear whether the effect of this ligand trap occurs in the periphery (before monocytes infiltrate the muscle) or at the level of cell recruitment. Further evaluation of chemokine production and signaling within the muscle or exogenous administration of chemokines into muscle may help to define this mechanism but is beyond the scope of this study.

These studies show a clear potential for clinical application of a therapeutic target preventing muscle fibrosis. Muscle fibrosis is a key limiting factor in rehabilitating the severely injured patient. In patients with significant extremity vascular compromise or in free-tissue transfer of muscle, a therapeutic agent that can improve return to function and prevent muscular fibrosis is of great value. The timing of our TGFBRII-Fc administration, beginning 30 min after unclamping of the femoral vessels, represents a clinically relevant administration schedule, modeling administration of the agent at the end of an operative procedure to restore blood flow. Further studies are necessary to evaluate this agent for its safety in polytrauma but it offers a promising insight into a pathway to improve muscle healing.

This study has several notable limitations. First, we use a LysM-Cre system to target myeloid cells. We have done extensive research on myeloid cell specific cre mouse lines. Vi et al. has shown that LysM-Cre effectively targeted macrophage lineage cells during the fracture repair process in vivo (45). Additionally, there was an excellent analysis of multiple myeloid “specific” cre lines published by Abram et al. in 2014 (42). In this study, authors crossed each of the lines to a ROSA-YFP reporter mouse and demonstrated across a variety of nonstimulated hematopoietic cells, focusing on the myeloid cells, among multiple tissues the efficiency and specificity of each strain. In the blood monocytes, they demonstrated that the LysM-Cre line had the highest amount of recombination signified by the expression of YFP. Further studies would need to be done to identify whether this effect is entirely confined to the interaction of infiltrating macrophages at the site of injury. Additionally, we confined our analysis to the early time points as we believe this is the crucial time where therapeutic intervention would make a difference.

Nonetheless, we have defined a central role of myeloid cell–derived TGF-β1 in the response muscle injury and subsequent development of fibrosis via extensive single-cell RNA sequencing of a unique and novel muscular polytrauma mouse model. We have also identified a crucial interaction between infiltrating macrophages and FAP cells to direct the severity of injury and formation of ECM. Finally, we identify a potential therapeutic modality for preventing the development of muscle fibrosis through ligand trap inhibition of TGF-β1 binding. These insights into the mechanisms of muscle repair may offer important viable therapies for human patients in the context of musculoskeletal polytrauma and morbidity.

The authors have no financial conflicts of interest.