Biopsies of healthy human muscle (n = 12, Supplementary Table S1) were obtained with informed consent and approved by the University of Southern California IRB. Healthy and diseased human muscle biopsies (Supplementary Table S1) were fixed in 4% formalin (Thermo Fisher Scientific) for histological and immunohistochemical analyses. Muscles were dissected into 1–2 mm diameter pieces and fragments were digested with 0.5 mg/ml Collagenase IV (Millipore Sigma) at 37°C for 40 min on a microplate shaker (Fisherbrand). Cell-containing supernatant was filtered through a 70 μm cell strainer (Fisherbrand). After centrifugation (Allegra X-14R, Beckman Coulter), cell pellet was resuspended in Endothelial Cell Growth Medium-2 (EGM-2, Lonza) and seeded into uncoated culture dishes (Corning) or Matrigel Basement Membrane Matrix-coated culture dishes (0.09 mg/ml, Corning). Pre-plating cultures were prepared by seeding human muscle fragments into uncoated culture dishes, and outgrown cell clusters were cultured in EGM-2. Where indicated, confluent adherent cells were then removed (Trypsin-EDTA 0.25%, Corning) for sorting of either CD56⁺CD90⁺ or CD90⁺CD56⁻ cell subsets. For all experiments, primary muscle cells were used between passages 2-6.

Human muscle sections were stained with H&E for general tissue structure analysis or picrosirius RED for collagen expression according to manufacturer instructions (Abcam, Cambridge, UK). For fluorescence microscopy, stained muscle sections were labeled with primary antibodies overnight at 4°C and with secondary antibodies for 30 min at room temperature. The antibodies used for this study are listed in Supplementary Table S2. DAPI (Molecular Probes, 1:1000) was used for nuclei labeling. Images were acquired with a CKX53F3 Compact Cell Culture Microscope (Olympus, Oberkochen, Germany) and a Keyence BZ-X (Itasca, IL, United States).

Human muscle cells from primary cultures (passages 0–6) were removed (Trypsin-EDTA 0.25%, Corning), washed in PBS (Corning), centrifuged (Allegra X-14R, Beckman Coulter), and labeled according to the manufacturer instructions. The antibodies used for this study are listed in Supplementary Table S2. Cell sorting was performed using FACS via an Aria III Flow Cytometer (BD Biosciences). Debris and dead cells were excluded according to forward and side scatter data. Analyses were carried out using an Attune flow cytometer (Thermo Fisher Scientific) and FlowJo v10.8.1.

For adipogenic induction, cells were cultured in DMEM/10% FBS (Corning) supplemented with 1 μM dexamethasone, 1 μM insulin, 0.5 mM IBMX (all purchased from Millipore Sigma), and 1% pen/strep (Gibco). Adipogenic differentiation was assessed with Oil Red O staining of lipids according to manufacturer instructions (ScienCell Research Laboratories). For fibrogenic induction, cells were cultured in DMEM/10% FBS supplemented with 5 ng/ml human-TGFβ1 (R&D Systems). For myogenic induction, cells were cultured in DMEM supplemented with 20% Human Male AB Serum (Access Biologicals) and 1% pen/strep. For adipo-fibrogenic induction (AF), cells were cultured in DMEM/10% FBS supplemented with 5 ng/ml human-TGFβ1, μM dexamethasone, 1 μM insulin, 0.5 mM IBMX, and 1% pen/strep. For adipo-myogenic (AM) induction, cells were cultured in DMEM supplemented with 20% Human Male AB Serum, 1 μM dexamethasone, 1 μM insulin, 0.5 mM IBMX, and 1% pen/strep. For fibro-myogenic (FM) induction, cells were cultured in DMEM supplemented with 20% Human Male AB Serum and 5 ng/ml human-TGFβ1 and 1% pen. For adipo-fibro-myogenic (AFM) induction, cells were cultured in DMEM supplemented with 20% Human Male AB Serum, 1 μM dexamethasone, 1 μM insulin, 0.5 mM IBMX, 5 ng/ml human-TGFβ1, and 1% pen/strep. All differentiating cells were analyzed at 6 days post induction. Fibrogenic differentiation was quantified via picrosirius red staining of collagens I and III (Abcam). Collagen-bound picrosirius red dye was extracted by 0.1 M NaOH solution (Millipore Sigma), and collagen quantification was performed using spectrophotometry via a SpectraMax iD3 Microplate Reader (λ = 540 nm, Molecular Devices). Adipogenic differentiation was quantified by Oil Red O staining and red pixel intensity measurements by AdobePhotoshop version 24.4.2. For identification of myotubes, cultures were fixed with 4% PFA (Thermo Fisher Scientific) for 10 min at room temperature, washed in PBS, and labeled with nuclear DAPI and anti-human striated MyHC I. A MyHC I⁺ cell that contained >3 nuclei with typical morphology was considered a myotube. All differentiation conditions were also tested in the presence of 1 μM and 10 μM 423F. Cultures from Matrigel coated dishes were the source for CD56 subset for all differentiation experiments. Cultures from uncoated dishes (both cell suspension and pre-plating conditions) were the source for CD90⁺ and CD90⁻ subsets for all differentiation experiments. Cytotoxicity of 423F was measured by XTT assay according to manufacturer’s instructions (Biotium). Cells were cultured in the presence of 1, 10 and 30 μM 423F in DMEM/10% FBS and XTT mixture was added after 3 days in culture and incubated at 37°C for 2 h. All experiments were performed in triplicates.

Cells were cultured in 6-well tissue culture plates (Corning) with EGM-2. Following cell expansion, cells were induced with the appropriate differentiation media for each condition and cultured for 24 h. After 24 h, the cultures were rinsed with 4°C PBS (Corning) and then put on ice. RIPA Lysis and Extraction Buffer (Pierce) and Protease and Phosphatase Inhibitor (Pierce) were added to the cultures and incubated on ice for an hour, and then scraped with a cell scraper (VWR International). Protein lysates were then collected into 1.5 ml microcentrifuge tubes (Bioland Scientific) and sonicated 3 times with 5-s pulses at a 50% power via a Fisherbrand™ Model 50 Sonic Dismembrator (Fisher Scientific). Protein concentrations were determined using the BCA Protein Assay (Pierce) according to manufacturer instructions and then mixed with 4x Laemmli Sample Buffer (Bio-Rad) and beta-mercaptoethanol (Millipore Sigma). Samples were boiled at 100°C for 5 min with a Digital Heating Drybath (Thermo Scientific) and then loaded into Mini-PROTEAN TGC Precast Gels (Bio-Rad) for SDS-PAGE that were ran at 130 V until completion using a Mini-PROTEAN Tetra Cell (Bio-Rad). Gels were then transferred onto a 0.2-μm nitrocellulose Trans-Blot Turbo Transfer Pack (Bio-Rad). The running buffer used for SDS-PAGE was 10x Tris/Glycine/SDS (Bio-Rad). Following the transfer process, the nitrocellulose membranes were blocked with EveryBlot Blocking Buffer (Bio-Rad) and incubated overnight with primary antibodies anti-STAT3, anti-pSTAT3, anti-beta-actin, and anti-Histone H3 (all diluted 1:1000 with blocking buffer and purchased from Cell Signaling Technologies). Following primary antibody incubation, membranes were rinsed with TBST (1L double distilled H₂O, 29.25g NaCl, 20 ml 1 M Tris-HCl, 20 ml 10% Tween) and incubated with secondary antibody (31463 from Invitrogen) diluted 1:4000 with blocking buffer. Following secondary antibody incubation, membranes were washed with TBST and developed with Clarity Western ECL Max Blotting Substrate (Bio-Rad). All membranes were imaged using a ChemiDoc XRS+ (Bio-Rad) and quantified using Fiji ImageJ software.

Data are presented as mean ± SEM (unless otherwise indicated in figure legends). Statistical analysis was performed via GraphPad Prism software and Microsoft Excel using one-way ANOVA, Student’s t test, and Tukey’s post hoc multiple comparisons. p < 0.05 was considered statistically significant.

In comparison to sections of healthy muscle biopsies (Figures 1A, D), diseased human RC muscles were characterized by loss of myofibers due to replacement with adipose tissue and fibrotic scars (Figures 1B–F). Small clusters of adipocytes were detected in proximity to large blood vessels (Figures 1B, E), and marked adipose tissue accumulation was observed next to massive vascular fibrosis (Figures 1C, F), implying perivascular cell origin of intramuscular fat and fibrosis. PDGFRα and CD90 were previously shown to demarcate FAP in diseased human muscle (Uezumi et al., 2014; Farup et al., 2021). In agreement, multiple other studies demonstrated that PDGFRβ and PDGFRα are strongly expressed by perivascular cells of various types of blood vessels, as well as interstitial cells in mouse and human muscles (Uezumi et al., 2014; Uezumi et al., 2016; Murray et al., 2017; Jensen et al., 2018). The combined expression of these markers in fibrotic muscle scars in early and late developmental stages is poorly defined, mainly due to the scarcity of biopsies from diseased human muscle. We therefore extended our analysis and mapped CD90, PDGFRβ, and PDGFRα co-expression in fibrotic lesions (Supplementary Table S1). As expected, PDGFRβ was highly expressed in healthy muscle by perivascular cells in all types of blood vessels as well as by perimysium and endomysium-residing interstitial cells (Figures 1G, I). In healthy muscle, CD90 expression was rarely detected in interstitial cells (Figure 1H) and was predominantly co-localized to PDGFRα⁺ perivascular cells of large veins (Figures 1I, J). All markers were dispersed between the three layers of the artery wall and were clearly co-expressed in the outermost layer, the tunica adventitia, as well as in the innermost layer, the tunica intima (Figure 1I). Next, we mapped the localization of PDGFRβ/PDGFRα/CD90 cells in fibro-adipogenic injured human RC muscles. The apparent development of massive fibrosis was accompanied by a robust increase in the number of PDGFRβ⁺PDGFRα⁺CD90⁺ cells, which populated the fibrotic scars (Figure 1K). Large fibrotic areas composed of PDGFRα⁺CD90⁺ cells were mostly located around large blood vessels that were identified by the presence of luminal red blood cells or von Willebrand factor⁺ endothelium (Figures 1L, M). Additionally, dense clusters of PDGFRβ⁺PDGFRα⁺ cells were seen emerging from the abluminal layer of perimysial blood vessels in areas of muscle undergoing fibro-adipogenic degeneration (Figures 1N–P). Taken together, these observations imply that PDGFRβ/PDGFRα/CD90 cell subsets are common drivers of fibro-adipogenic muscle disease with different etiologies.

To gain better understanding on how human muscle cells respond to fibro-adipogenic-myogenic stimulus in injury settings, we sought to develop a stable primary cell subset platform that could maintain the phenotype of FAP and myo-progenitors. To confirm the reproducibility of the protocol for cell subset generation, biopsies were taken from a variety of healthy skeletal muscles, including subscapularis, deltoid, quadriceps, and pectoralis. A wide range of donor ages (between 20–81 years, n = 12) was tested as well (Supplementary Table S1). Cell composition was analyzed by flow cytometry when cultures reached 80%–90% confluency at passage 0. In agreement with previous studies (Uezumi et al., 2016; Goloviznina et al., 2020), all cell subsets gained and maintained high expression of CD73 in culture (Figure 2A). Three CD73⁺ cell subsets were consistently identified regardless of muscle type or donor age: 1. CD56⁺CD90⁺, 2. CD90⁺CD56⁻, and 3. CD90⁻CD56⁻ cells (Figures 2A–C; Supplementary Figure S1). Sorted CD56⁻ cell populations (Supplementary Figure S1) co-expressed PDGFRβ and CD9 and lacked the expression of CD146, CD15 and CD34. Only the sorted CD90⁻CD56⁻ subset (Supplementary Figure S1), but not the sorted CD90⁺CD56⁻ subset (Supplementary Figure S1), expressed the fibro-adipogenic marker PDGFRα (Figure 2B). We have previously published studies regarding lineage-restricted (non-myo-fibro-adipogenic) pericyte-like cells (LRPC) generated from human embryonic stem cells (Mosich et al., 2019) that were used as controls for our subset differentiation studies. Interestingly, LRPC exhibited a mixed-marker profile, expressing characteristic progenitor cell markers such as PDGFRβ and CD146 but lacking the expression of PDGFRα, a common FAP marker (Figure 2B). Tremendous enrichment in the frequency of cell populations that expressed the myogenic marker CD56 was achieved when culture dishes were coated with Matrigel in comparison to seeding of cell suspensions onto uncoated dishes or pre-plating of muscle fragments (Figures 2D, F). CD56⁺ subset dominated the culture at the end of P0 (Figure 2F) and flow cytometry analysis revealed that CD56⁺ myoblast-like cells acquired a marker expression profile, that is, more characteristic of perivascular/mesodermal cells but lack the expression of the fibro-adipogenic marker PDGFRα (Figure 2E).

We then further established the identity of each cell population by comprehensive evaluation of their differentiation potential using a broad differentiation culture system. To better mimic in vivo conditions, where cells simultaneously receive various signals with diverse regulatory effects, cell subsets were exposed to mixed culture induction conditions for characterization of cell fate decisions when interactions between various pathways take place. FAP (CD73⁺CD90⁻CD56⁻), CD90 (CD73⁺CD90⁺CD56⁻), and CD56 (CD73⁺CD56⁺CD90⁺) cell subsets, as well as LRPC (CD73⁺CD90⁺CD56⁻), were induced in either single fibrogenic (F), adipogenic (A), and myogenic (M) culture conditions, or mixed culture conditions at various combinations, including adipo-fibrogenic (AF), adipo-myogenic (AM), fibro-myogenic (FM) and adipo-fibro-myogenic (AFM) conditions. Following 6 days of culture, their fibrogenic and adipogenic responses were measured. In single-condition-induced cultures, FAP subset exhibited the highest fibrogenic collagen production in all tested conditions in comparison to CD56, CD90, and LRPC (Figures 3A, D, G, 4D, G). Exposure of prospectively isolated CD90 subset to adipogenic conditions defined it as a novel non-adipogenic subset with low fibrogenic potential and thus similar to LRPC (Figures 3B, 4C, E). Only FAP exhibited extensive adipogenic differentiation in adipogenic cultures (Figures 3B, 4C, E). FAP adipogenesis was mild in myogenic cultures and was abrogated in fibrogenic TGFβ-1-induced cultures (Figures 3B, F4C, E, F, H). FAP maintained high fibrogenic potential in all combined culture conditions that were slightly reduced in the presence of AM combinations, including AM and AFM (Figures 5A, D, G, 6B, E). AF (Figures 5B, C, E) and AFM (Figures 5B, 6D, F) inductions failed to rescue adipogenic differentiation of FAP, demonstrating a dual role of TGFβ-1 as a positive regulator of fibrosis and a negative regulator of adipogenesis.

As expected, CD56 (Uezumi et al., 2016; Goloviznina et al., 2020) and LRPC cells (Mosich et al., 2019) lacked adipogenic potential in all tested conditions (Figures 3B, C, E, F, H, 4C, E, 5B, C, E, F, H, 6A, C, D, F) except for a minimal adipogenic response observed for CD56 under AM and AFM conditions (Figures 5B, F, H, 6D, F). Unpredictably, in our system, CD90 subset did not demonstrate any magnitude of adipogenic potential either in single or mixed conditions (Figures 3B, C, E, F, H, 4C, E, 5B, C, E, F, H, 6A, C, D, F). Only CD56⁺ cells readily formed multinucleated, myosin heavy chain I (MyHC I) expressing myotubes in control and myogenic cultures (Figures 4A, B; Supplementary Figure S3). Increased number and size of MyHC I⁺ CD56-derived myotubes were measured in the presence of human-serum-enriched DMEM medium that was defined in accordance as myogenic medium (data not shown). Of notice, replacement of FBS with human serum evoked for the first time a functional response of LRPC, inducing an increase in collagen production (Figure 3A). This finding indicates that classical TGFβ-1 fibrogenic induction is sometimes not enough to reveal the fibrogenic features of certain cell populations. Overall, our differentiation studies reveal that cell fate is mainly governed by intrinsic features characteristic of each muscle cell subset with a mutable degree of differentiation provided by extrinsic signals (Figure 6G).

Previous studies carried out in traditional single input differentiation cultures demonstrated that the gp130/STAT3 pathway plays a key role in the regulation of these differentiation pathways and is associated with intrinsic features of muscle cell populations (Wang et al., 2008; Price et al., 2014; Tierney et al., 2014; Madaro et al., 2018). We therefore elaborated upon this in our studies and evaluated the potential therapeutic effects induced by modulation of the gp130 pathway. For that purpose, we used the compound 423F, a structural analog of an experimental small-molecule drug (RCGD 423) that has been shown to target IL-6/STAT3 signaling and promote cartilage repair (Shkhyan et al., 2018). We first extended findings in human chondrocytes to human muscle cells and validated that 423F is neither cytotoxic nor an inducer of muscle cell differentiation (Figures 3C–E; Supplementary Figure S2). We found that 423F not only failed to induce cell differentiation, but it also inhibited minor FAP adipogenic differentiation and fibrotic collagen production in non-induced control cultures (Figures 3C–E). 423F decreased FAP adipogenesis in a dose-dependent manner in all single induction conditions that mediated FAP adipogenic differentiation, including untreated control and adipogenic cultures. Myogenic cultures (Figures 3C, E, 4C, E, F, H) induced further decrease in FAP adipogenesis in AM, FM, and AFM combined conditions (Figures 5H, 6C, F), and further attenuated the minimal adipogenic differentiation of FAP measured in AF cultures (Figure 5E). Spectrophotometric quantification of picrosirius red staining for collagens I and III revealed that 423F is also a potent inhibitor of fibrosis, as indicated by dose-dependent reductions in collagen production in the highly fibrogenic FAP subset, which was significant in most of the tested conditions (Figures 3D, G, 4D, G, 5D, G, 6D, G). Similarly, the presence of 423F in the much less fibrogenic subsets of CD90 and CD56 also reduced collagen production in fibrogenic cultures (Figure 3G). Predictably, the 423F-treated non-fibro-adipogenic LRPC, which served as non-differentiating control cells in this study, maintained basal levels of collagen production in comparison to untreated matched induced cultures at all tested combinations of inductions, except for increased collagen expression in the presence of human serum (myogenic cultures) (Figure 3A). Still, myogenic cultures did not induce myogenic differentiation of LRPC as indicated by the lack of multi-nucleated myotube formation (Figure 4F). Collectively, these data demonstrate that capability of myogenic, fibrogenic, or adipogenic differentiation is largely dependent on the intrinsic features of cultured subsets, and that the degree of lineage differentiation varies when signals are multiplexed (Figure 6G).

The drug 423F is a gp130 modulator that acts via promoting gp130 homodimerization and preventing IL-6 and Oncostatin M-induced heterodimerization with IL-6R and OSMR respectively. In chondrocytes, 423F was shown to induce a marked increase in the levels of activated receptor-gp-130 (phosphorylated gp130) and in the levels of activated phosphorylated STAT3 (pSTAT3), the downstream effector. Given that STAT3 has been shown to regulate myogenic differentiation in a context-dependent manner (Wang et al., 2008; Price et al., 2014; Tierney et al., 2014) we further assessed its expression and activation in human muscle subsets in single and multiplexed myogenic, fibrogenic, and adipogenic culture conditions. The lineage-restricted hES-derived cell line was used to validate lack of activation of STAT3 in mesodermal cell counterparts. 423F mediated a slight dose-dependent decrease in STAT3 activation after 24 h in untreated cultures (Figures 7A, E). A stronger dose-dependent reduction in pSTAT3 was observed following induction of FAP differentiation in fibrogenic and adipogenic cultures (Figures 7A, F, G). This demonstrates that 423F/gp130 signaling antagonizes TGF-β1-mediated expression of collagen I and III, as well as insulin-induced adipogenesis, suggesting that 423F/gp130 signaling promotes downstream activation of SOCS3 that has been shown to negatively regulate gp130 (Carow and Rottenberg, 2014), TGF-β1 (Ogata et al., 2006), and insulin (Fan et al., 2014) signaling. Conversely, the myogenic CD56 subset exhibited higher levels of pSTAT3 in comparison to all other subsets in untreated culture (Figures 7C, E). There was a significantly greater dose-dependent increase in STAT3 activation in myogenic conditions (Figures 7C, H). In accordance, myogenesis took place only in untreated and myogenic cultures as indicated by detection of MyHC I⁺ multinucleated myotubes (Figures 4A, B). Like differentiating FAP, 423F inhibited STAT3 activation in differentiating CD56 in a dose-dependent manner (Figure 7H). Dose-dependent 423F/gp130-induced reduction in FAP-pSTAT3 could be still detected in AF and FM cultures (Figures 7A, I, K) but was abolished in AM and AFM cultures (Figures 7A, J, L). The non-adipogenic CD90 subset in turn exhibited opposite response to that of FAP and CD56, exhibiting a 423F/gp130-mediated dose-dependent increase in pSTAT3 in fibrogenic cultures (Figures 7B, F). The levels of pSTAT3 remained unchanged in adipogenic, myogenic, AM, and AFM conditions that did not elicit CD90 adipogenic differentiation nor significant changes in their 423F-promoted fibrogenic response (Figures 7B, G, H, J, L, 5G, H, 6E, F). FAP exhibited the highest sensitivity to 423F-mediated inhibition of STAT3 activation in comparison to CD56 and CD90 in matched culture conditions (Figure 7). As expected, the levels of pSTAT3 were not affected by 423F/gp130 signaling at all tested conditions and 423F concentrations (Figure 7D). To confirm that gp130 mediated 423F effects, FAP were co-treated with 423F and anti-gp130 in non-induced, fibrogenic, and adipogenic cultures. In all tested conditions, a greater reduction in pSTAT3 levels was detected in the presence of both 423F and anti-gp130 than that seen in the presence of 423F alone, validating previous data (Shkhyan et al., 2018) demonstrating that 423F acts through gp130.

Considering that administration of 423F can concurrently modulate the differentiation and cell composition of the progressively degenerating skeletal muscle, we further evaluated the effect of 423F/gp130 signaling on muscle progenitor differentiated progeny. To evaluate myotube growth (Hindi et al., 2013), we measured myotube diameter and quantified the number of nuclei per myotube in cultures of CD56⁺CD90⁺ myogenic subset in the presence or absence of 423F. Progenitor derived myotubes were significantly larger in 423F treated cultures than myotubes in untreated cultures 423F. Greater increase in myotube size was measured in the presence of higher 423F concentration than that of untreated control cultures or cultures treated with lower concentrations of 423F (Figures 8A, B). A maximal increase in myoblast fusion, which was indicated by the number of nuclei per myotube, was already achieved at lower concentrations of 423F (Figure 8B).

It was previously shown that IL-6 and STAT3 are required for adipocyte lipolysis (Guo et al., 2021; Gandhi et al., 2022). We therefore hypothesized that mature adipocytes may be more susceptible to 423F-induced modulation of gp130/STAT3 signaling rather than FAP. FAP were first induced to differentiate into adipocytes in adipogenic medium for 6 days. At 6 days post adipogenic induction, FAP-derived cultures were composed of mature Oil red O⁺ adipocytes and a fraction of non-differentiated FAP (Figure 8C). The effects of 423F on mature adipocytes were further tested in these mixed cultures. Mature adipocytes derived from human muscle FAP following 6 days of adipogenic induction were eliminated from 423F-treated cultures in a dose-dependent manner (Figures 8C, D). Conversely, 423F treatment did not affect the growth of non-differentiated muscle FAP in these mixed adipocytes-FAP cultures (Figures 8C, D). Taken together, these data demonstrate that 423F can simultaneously promote myo-regeneration and diminish chronic muscle degeneration by inhibition of fibrosis progression as well as selective elimination of developing intramuscular adipose tissue.