Muscle biopsies were obtained from patients with a genetic diagnosis of DMD and gender matched healthy controls. Clinical data of the patients are described in Table 1.

We obtained informed consent forms signed by parents or legal guardians as well as patient’s assent forms for donating these samples to research. Muscle samples from the biceps brachii of age- and gender-matched controls were obtained by the orthopaedic surgery department of Hospital de la Santa Creu I Sant Pau in Barcelona. Primary cells were obtained from the John Walton Muscular Dystrophy Research Centre Biobank (REC Ref: 19/NE/0028), which is supported by the NIHR Newcastle Biomedical Research Centre. The study was approved by the Ethics Committee of the Hospital de la Santa Creu I Sant Pau Hospital.

Frozen muscle explants from healthy controls and DMD patients were cultured to obtain FAPs as previously described (Suárez-Calvet et al., 2021). Briefly, muscle fragments were cultured in gelatine-plasma coated dishes with FAP medium (DMEM-GlutaMAX (ThermoFisher [Gibco], Waltham, MA) supplemented with 20% FBS (Gibco), 1%PS (Lonza, Basilea, Switzerland) and 2.5 ng/mL of basic fibroblast growth factor (Peprotech, Rocky Hill, NJ)). Sprouting cells from explants were trypsinized (Gibco), expanded and subcultured in Cell + culture flasks (Sarstedt, Nümbrecht, Germany).

Cultured human muscle derived cells were labelled using anti‐PDGFRα (BAF322, R&D Systems, Minneapolis, MN) followed by streptavidine‐PECy5 (405202, Biolegend, San Diego, CA) and anti‐CD56 (Milteny-Biotech). Stained cells were sorted on a FACSAria using FACSDiva software (Becton Dickinson, Ashland, OR). Doublet cells were excluded using forward scatter area and height and fluorescence minus one (FMO) controls were used to determine positivity. PDGFRα+/CD56− fraction was defined as FAPs. Human FAPs isolated from healthy human sample were transduced with a retroviral vector containing a sequence encoding the catalytic subunit of human telomerase reverse transcriptase (hTERT) alone or with both hTERT and cyclin-dependent kinase 4 (Cdk4) as described previously.

FAPs were seeded at 5,000 cells/cm² concentration in 96 well plates in FAP growth medium. To achieve adipogenic differentiation, media was changed into adipogenic differentiation medium (StemPro, Gibco) for 6 days. In order to induce fibrogenic differentiation, FAPs were cultured for 3 days with DMEM-GlutaMAX with 10% FBS with 5 ng/mL TGFβ (R&D Systems). Pharmacological compounds were obtained from the Wnt/Hedgehog/Notch Compound Library (186 compounds) and Tyrosin-kinase inhibitors (TKI) Library (322 compounds) from MedChemExpress, United States. Adipogenesis differentiation was initially assessed using the Wnt/Hedgehog/Notch Compound Library. Negative controls (non-differentiated FAPs) were maintained in DMEM+2% FBS and positive controls (adipogenic-differentiated FAPs) were treated with StemPro+0.01% DMSO (vehicle compound). Each compound was added to one well and incubated for 6 days to allow adipogenic differentiation process. Fibrogenic differentiation was initially assessed using the TKI library. Negative controls (non-differentiated FAPs) were maintained in DMEM+10%FBS while positive controls (fibrogenic-differentiated FAPs) were treated with fibrogenic medium+0.01%DMSO (vehicle compound). Compounds were incubated for 3 days to allow complete fibrogenic differentiation process.

After pharmacological treatment, collagen-I expression was analysed in fibrogenic differentiation and perilipin-1 in adipogenic differentiation using quantitative in-cell western (ICW). Plates were rinsed with PBS and fixed with 4% PFA for 10 min. After washing steps and blocking with casein (ThermoFisher) solution for 30 min, goat anti-collagen I (1:500; Cat. 1310-01; Southern Biotech, Birmingham, AL) or rabbit anti-perilipin A/B (Cat. P1873–200UL; 1:500; Sigma-Aldrich, St. Louis, MO) was added overnight at 4C. After the washing steps, wells were incubated with either donkey anti-goat IRDye800 or donkey anti-mouse IRDye800 secondary antibody (1:1000; Li- COR, Lincoln, NE) and CellTag IRDye 700 (Li-COR) for 1 h. After washing with PBS, the fluorescent signal was measured by an Odyssey Imaging system (Li-COR) and normalized by cell number measured in the 700 nm channel. To analyse the % of inhibition by each drug, each condition was normalized to the positive control (fibrogenic or adipogenic differentiated FAPs). After that, we represented each drug into a map were the 0% represented the standard fibrogenic or adipogenic differentiation of FAPs and the effect of each compound was plotted in the negative side of the graph while drugs promoting the differentiation are represented in the positive side of the graph. At this stage, drugs were tested at n = 1.

FAPs were plated in 96-well plate at 5,000 cell/cm² density for 24 h. Then, cells were treated with the pre-selected compounds diluted in adipogenic or fibrogenic differentiation media in 12 descending concentrations (2,200–3,124 nM) and the vehicle. The differentiation performed at this step was according to the drug original plate (fibrogenic differentiation for TKI library or adipogenic differentiation for Wnt/Hedgehog/Notch Compound Library and compounds were tested in triplicates (n = 3). After differentiation process, FAPs were stained with PrestoBlue Cell Viability Reagent (Invitrogen) following manufacturer’s instructions. Fluorescence was measured at wave-lengths 570 nm excitation and 600 nm emission using a microplate reader (Varioskan LUX, ThermoFisher). After that, FAPs were rinsed with PBS and the ICW protocol was performed in the same plate to stain collagen-I and perilipin proteins.

The normalized gene or protein expression data obtained from single-cell RNA-seq (Fernández-Simón et al., 2024) (data available at https://singlecell.broadinstitute.org/single_cell/studies/66703832771a5b0208cf92e3) and proteomic dataset was used to predict which signalling circuits were differentially expressed in the DMD condition compared to the control condition. Quantitative proteomics was performed using a Lumos Orbitrap mass spectrometer (Thermo Fisher Scientific,San Jose, CA, United States) coupled to an EASY-nLC 1000 (ThermoFisher Scientific (Proxeon), Odense, Denmark). Samples were analysed with the MaxQuant software (version1.6.1.0) through the human Swissprot database. Single cell RNA-seq from FAPs were partitioned into Gel Bead-In-Emulsions (GEMs) by using the Chromium Controller system (10X Genomics). cDNA sequencing libraries were prepared using the Next GEM Single Cell 3′ Reagent Kits v3.1 (10X Genomics, PN-1000268), following manufacturer’s instructions. cDNA libraries were indexed by PCR using the PN-220103 Chromiumi7 Sample Index Plate. Size distribution and concentration of 3′cDNA libraries were verified on an Agilent Bioanalyser High Sensitivity chip (Agilent Technologies). Finally, sequencing of cDNA libraries was carried out on an Illumina NovaSeq 6,000 using the following sequencing conditions: 28 bp (Read 1) + 8 bp (i7 index) + 0 bp (i5 index) + 89 bp (Read 2), to obtain approximately 20–30.000 reads per cell. We checked the quality of the cells in each sample in R (version 4.2.1). For downstream analysis, cells with less than 1,000 reads or 500 genes and more than 5% mitochondrial reads were filtered out (Wolock et al., 2019). Using DoubletFinder (version 2.0.3), we identified and filtered out the doublets from each sample. 2000 highly variable genes were then identified using which we conducted principal component reduction and clustering of the data.

To reveal the precise biological properties of each conditions, we used Metascape (http://metascape.org) to perform enrichment analysis including KEGG Pathway, GO Biological Processes, Reactome Gene Sets, Canonical Pathways, CORUM, WikiPathways and PANTER Pathway. Genes with a log2FC > 0.5 were analysed for each condition.

The HiPathia method was used to estimate signalling circuit activities within DMD condition from the corresponding normalized gene and protein expression values. HiPathia identifies signalling circuits using nodes that represent genes or protein and edges that represents the interaction. The software estimates the signal transduction along these pathways by considering both the structure of the pathway and gene expression levels. We initially excluded circuits that maintained consistent activation statuses across the studied conditions. Then, we selected the circuits that were represented by the pathways targeted by the drugs used. Results from the scRNAseq and proteomic studies were normalized using the log-fold change (logFC) of each protein comparing the expression of the DMD condition against the healthy control condition. The accuracy of the classification was assessed based on the log fold change (logFC) of each circuit and averaging the total circuits by pathway. The HiPathia method uses a Wilcoxon test was used to assess differences in pathway activity between controls and DMD samples (Fernández-Simón et al., 2024).

To reproduce the fibrogenic and adipogenic differentiation of FAPs, we differentiated immortalized FAPs into fibroblasts and adipocytes in vitro. As a readout, we quantified collagen-I and perilipin-I expression using in-cell Western blot (ICW) technique (Figure 1A). With that method, we assessed how this differentiation is influenced by two different drug library containing multiple compounds that could potentially inhibit fibrogenesis and/or adipogenesis. For adipogenic differentiation, we used a library of compounds interfering with the Wnt/Hedgehog/Notch pathways as these have been identified to be involved in adipogenesis (Marinkovic et al., 2019; Keats et al., 2014; Reggio et al., 2020). Regarding fibrogenesis, we used a library of 339 tyrosin-kinase inhibitors (TKI) since they been shown to influence fibrogenesis (Piñol-Jurado et al., 2018; Zhang et al., 2016). To analyse the effect of the libraries, drugs were added at a high (1uM) and low (0.1uM) concentration (Figure 1B) and differentiated into fibroblasts using the TKI library or adipocytes using the Wnt/Hedgehog/Notch Compound Library. Negative controls (containing the same vehicle as drugs) were also added to the plates and served as a reference for the standard differentiation rate enabling the calculation of the induction or inhibition of each compound used. Specifically, drugs were defined as a hit when the adipogenic or fibrogenic differentiation potential was decreased by more than 30% at both high and low doses compared to the negative control. We then plotted the inhibition of fibrogenesis by the TKI inhibitors and adipogenesis by the Wnt/Hedgehog/Notch interfering compounds at both doses on the same graph (Figure 1C) and identified a total of 29 drugs interfering with FAPs differentiation (Table 2). Based on the mechanism of action of each drug we constructed an inhibitory map by selecting the main target of each compound (Figure 1D).

To further characterize the selected drugs, we ensured that the inhibition of fibrogenesis with the TKI did not promote adipogenesis and similarly, that the inhibition of adipogenesis with the Wnt/Hedgehog/Notch compounds did not promote fibrogenesis. We used triplicates and put them in either an adipogenic or fibrogenic differentiation media as in our previous assay. At this stage, drugs showing an increase in fibrogenesis or adipogenesis were eliminated (Figure 2A) as well as those that reduced the total amount of cells (measured by CellTag). Out of the 29 compounds, 8 compounds reduced fibrogenesis and adipogenesis at ≥30% but 4 were eliminated since they showed reduced number of cells (Salinomycin, Adavivint, AD80 and Geldanamycin). Final compounds selected for the next step were: Bruceine D, GSK-3 inhibitor 1, RGF 286638-free base. To find the most effective and non-cytotoxic dose, the pre-selected drugs were analysed for fibrogenic or adipogenic differentiation together with a viability assay using a range of decreasing concentrations (from 22 μM to 3.11 nM) (Figures 2B–D). Adjustments to find the exact concentration that had an effect on FAPs without affecting viability were performed in primary cell line of FAPs isolated from a healthy subject (Supplemental Figure 1). Out of the initial 508 drugs screened, we narrowed down the list to a final 4 compounds.

Once we established the in vitro system, we aimed to further validate the signalling pathways involved in the degenerative process of FAPs. To validate this system, we analysed whether the upregulated pathways observed in FAPs isolated from pathological conditions were decreased after drug treatment in our in vitro assay.

We decided to examine deregulated pathways in FAPs isolated from Duchenne muscular dystrophy patients, a genetic disorder characterized by muscle degeneration. DMD is one of the most severe forms of muscular dystrophies leading to rapid degeneration of the skeletal muscle accompanied by an increase in fibrotic and adipogenic tissue. For this reason, analysing deregulated pathways in FAPs isolated from DMD patients could be a valuable strategy for studying the molecular mechanisms involved in the process of muscle degeneration. By combining results from cell transcriptomic and proteomic studies, we unravelled the underlying mechanisms of cell signalling in FAPs, providing a functional link between gene and protein-level data and disease-related cellular mechanism of interest. On the one hand, we performed single-cell RNA sequencing (scRNAseq) study using 10X Genomics and Illumina technology to analyse the gene signatures of FAPs isolated from healthy samples (n = 2) and DMD patients (n = 3) (Xu et al., 2021) (doi: 10.3389/fcell.2024.1399319). In parallel, we performed mass spectrometry to analyse the protein expression of FAPs isolated from healthy samples (n = 3) and DMD patients (n = 3) (Figure 3A). To obtain a comprehensive understanding of the variations in gene and protein expression between DMD and control FAPs, we first analysed the differences at the genomic and proteomic level comparing the healthy samples and the DMD samples. Differences in proteins and genes are represented in a Volcano plot in Figure 3B. To reveal the precise biological properties of each condition, we used Metascape. Genes and proteins with a log₂FC > 0.5 were analysed for each condition. Top genes upregulated in the healthy condition (blue dots) were involved with the actin cytoskeleton organization and ECM interaction as well as skeletal system development while top genes in the DMD condition (red dots) were involved with cell migration, regulation of insulin-like growth factor or the complement system. Top proteins upregulated in the healthy condition were involved in cell homeostasis such as protein localization between organelles in the cell, membrane organization or regulation of muscle adaptation. However, upregulated proteins in the DMD condition were involved with mRNA metabolic process, regulation of angiogenesis or adherent’s junctions.

Then, we analysed potential changes in signalling pathways using high throughput pathway interpretation and analysis (HiPathia) tool. This tool transforms uninformative gene expression data into signalling pathway circuit activities (Hidalgo et al., 2016; Esteban-Medina et al., 2019). The signalling circuits selected are those that exhibit the most significant change in activity either upregulated or downregulated pathways (Figure 3C). The increased cell signalling circuits identified in both datasets allowed us to estimate the signalling activity profiles upregulated in pathogenic conditions. Upon comparing the two studies, we observed that signalling circuits related to Forkhead box O (FOXO), Nf-kB or ErbB pathways were upregulated in DMD FAPs, while circuits related to VEGF and Notch were decreased in DMD FAPs.

This orthogonal integrative analysis enabled us to predict the mechanism of action of drugs and to validate whether pathways increased in the DMD condition in vitro were in accordance with the highest inhibited targets. For instance, the ErbB signalling, which is associated with the epidermal growth factor receptor (EGFR), was found to be one of the molecular pathways most upregulated in DMD FAPs. When we analysed how FAPs respond to EGFR inhibitors, we observed that most of these compounds decreased FAP differentiation, although the pattern of inhibition was not completely homogeneous. Another example of the most representative pathways is the Nf-kB signalling pathway. When we analysed which of our compounds targeted that pathway, we observed that three compounds inhibited the IκB kinase, which is part of the upstream Nf-KB signal transduction cascade. All these three compounds showed an inhibitory pattern in FAPs differentiation. Finally, we analysed the effect of inhibiting insulin or FOXO pathway. Since FOXO is a transcription factor, we did not identify any drug targeting the factor directly. However, HiPathia tool classified-FOXO signalling by upregulation of cyclin kinase proteins. Once we analysed the inhibition of different CDK isoforms, we observed that most of them presented a negative regulation of FAPs differentiation (Figure 3D). That observation shows that hallmarks playing some type of upstream regulator role in pathogenic conditions are the ones most responsive to inhibitory treatments.

Conversely, we observed that VEGF and Notch signalling pathways were downregulated in the DMD FAPs in vitro. Targets of these pathways did not show an inhibitory effect, suggesting that molecular pathways that are not physiologically upregulated in DMD may not possess any regulatory activity on FAPs differentiation. We observed that targeting Notch pathway resulted in an increase of both fibro/adipogenic differentiation, although some of these drugs reduced the differentiation potential of FAPs. This heterogeneous response to the same pathway might be associated with the suppressive mechanism of adipogenesis related to Notch signalling. It has been previously studied that Notch reduces adipogenesis (Marinkovic et al., 2019), and therefore Notch inhibitors could lead to an increased differentiation of FAPs into adipocytes (Figure 3E).

Since the inhibitory pattern according to each pathway was not homogeneous due to some drugs targeting the same pathway could inhibit but also had no effect on FAP differentiation, we aimed to further assess the specificity of targeting single proteins from a determined pathway. We decided to focus on the EGFR pathway as it was the highest upregulated pathway in both the transcriptomic and proteomic study. The EGFR pathway has been previously linked to fibrosis (Sakuma, 2017; Zhuang and Liu, 2014) and all the compounds used in our study were from the TKI library, so we performed the first inhibition assay in fibrogenic conditions. The epidermal growth factor RTK family consists of four members: EGFR (ErbB1, HER1), ErbB2 (HER2), ErbB3 (HER3) and ErbB4 (HER4). By evaluating the circuits upregulated or downregulated in our scRNA-seq dataset, we observed that HER4 and HER3 signalling pathways were increased, while EGFR pathway was decreased. When circuits belonging to all 4 receptors were analysed, there was an increase of activity detected (Figure 4A). By analysing each target of the pathway in our library response map (Figure 4B), we observed that while EGFR and HER2 inhibitors did not influence FAP differentiation (green and dark blue dots, respectively), drugs targeting HER3 specifically inhibited FAP fibrogenic differentiation (purple dots). We could not analyse the effect of only inhibiting HER4 because none of the drugs of the drugs used in our library specifically target that receptor. However, we identified one compound interfering with HER4 and EGFR that inhibited FAPs differentiation (violet dot). Most of the compounds targeting HER4, HER2 and EGFR did also reduce FAPs differentiation (red colour) suggesting that HER3 and HER4 signalling may be more involved in fibrofatty differentiation than HER2 and EGFR (Figure 4B). This is another example of how our in vitro platform can help in the analysis of the molecular mechanisms that are altered in pathogenic conditions and which type of drugs can be developed to inhibit fibrofatty expansion in muscle tissue.

Finally, we wanted to assess whether the selected drugs identified in our assay had a similar effect on primary cells obtained from healthy individuals and from patients. We isolated FAPs from muscle tissue from three DMD patients and three healthy aged-matched individuals. By using the differentiation assay as previously described, we treated FAPs with the four top candidates of our libraries. Despite the high variability observed, which is probably related to the inherent variability of using primary cells from different individuals, we observed that the four candidate drugs reduced fibrogenesis and/or adipogenesis. In general, the inhibition was higher when samples from DMD patients were used, suggesting that FAPs isolated from a perturbed microenvironment could respond more to antifibrotic and antiadipogenic treatment and supported our previous results showing that the molecular pathways previously identified could be also upregulated in these cells (Figures 4C, D).