Organoid culture promotes dedifferentiation of mouse myoblasts into stem cells capable of complete muscle regeneration

SkMOs contain satellite-like cells

Satellite cells express Pax7, an essential transcription factor (TF) for satellite cell maintenance and commitment^2,22. During skeletal muscle regeneration, myoblasts maintain Pax7 expression but activate MyoD, a TF involved in myogenic commitment and differentiation^23,24. MyoD expression, coupled with the downregulation of Pax7, commits cells to differentiation, resulting in the loss of stem cell potential¹². To investigate whether myoblasts (Pax7⁺MyoD⁺) can reduce the expression of MyoD and dedifferentiate back to a satellite cell-like state (Pax7⁺MyoD⁻), we seeded myoblasts into spinner flasks in proliferation medium¹⁷, where they self-assembled in three dimensions into SkMOs²⁵. After 20 d, we reduced the serum concentration and allowed the cells to differentiate. After an additional 10 d, the SkMOs were predominantly composed of cells that expressed myosin heavy chain (MyHC), a marker for terminally differentiated myotubes²⁶, and cells that expressed Pax7, many of which were present near myotubes (Fig. 1a,b). Although MyoD was expressed in the vast majority (98% ± 0.70%) of the Pax7⁺ myoblasts used to generate SkMOs, it was expressed in only a small percentage (9% ± 4.28%) of the Pax7⁺ cells present in late-stage SkMOs (Fig. 1c,d).

a, Schematic of the isolation and expansion of mouse myoblasts in vitro followed by the formation, growth and maturation of an SkMO. b, Immunofluorescent images of a late-stage SkMO stained with Pax7 (green), MyHC (red) and nuclear counterstain Hoechst (blue). Scale bars, 100 µm. c, Immunostaining of myoblasts and late-stage SkMOs for MyoD (red), Pax7 (green) and Hoechst (blue). Scale bars, 20 µm. d, Percentage of Pax7⁺MyoD⁺ cells in myoblasts and late-stage SkMOs. Box bounds represent s.e.m.; red line represents mean (n = 3, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P value: Myo versus SkMO 3.37 × 10⁻⁵; ***P < 0.001. e, Immunofluorescent images from EdU pulse chase experiments over 48 h comparing myoblasts, quiescent satellite cells and late-stage SkMO cells. Images show Pax7 (green), EdU (red) and Hoechst (blue). Scale bars, 10 µm. f, Percentage of Pax7⁺EdU⁺ cells in each cell condition. Box bounds represent s.e.m.; red line represents mean (satellite cell n = 4, myoblast n = 3, SkMO n = 3, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction between samples. P values: SC versus SkMO 0.205, SC versus Myo 2.0 × 10⁻⁶; Myo versus idSC 3.5 × 10⁻⁵; ***P < 0.001. g, Representative images of myoblasts, SkMO-derived GFP⁺ cells and quiescent satellite cells immediately after FACS isolation. Scale bars, 10 µm. h, Comparisons of cell diameter in proliferative myoblasts, SkMO-derived GFP⁺ cells and quiescent satellite cells. Box bounds represent s.e.m.; red line represents mean (n = 3, biological replicates with ≥100 cells per replicate). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction between samples. P values: Myo versus SC 2.22 × 10⁻⁴; Myo versus SkMO 8.98 × 10⁻⁴; SkMO versus SC 1.7 × 10⁻⁴; ***P < 0.001. NS, not significant; Myo, myoblast; SC, satellite cell.

To validate our findings, we used a transgenic Pax7^nGFP reporter mouse strain (Tg:Pax7^nGFP)²⁷ to FACS purify satellite cells (>95% GFP⁺) and derive, in standard two-dimensional (2D) growth conditions, several primary myoblast lines subsequently used to generate SkMOs (Supplementary Fig. 1a,b). After 30 d of organoid culture, GFP⁺ cells accounted for 18% ± 2% of the mononuclear cells within each organoid (Supplementary Fig. 1c). This provided approximately 3 ± 1.5 million GFP⁺ cells per spinner flask (Supplementary Fig. 1d). Finally, we compared the presence of canonical myogenic cell surface markers among myoblasts, SkMO GFP⁺ cells and freshly isolated satellite cells (Supplementary Fig. 1e) and found that SkMO GFP⁺ cells closely resembled the profile of satellite cells (Cxcr4⁺, CD104⁻ and CD200⁺)²⁸.

In intact muscle, satellite cells are normally quiescent. Although the parental myoblasts used to form SkMOs incorporated the nucleoside analogue 5-ethynyl-2′-deoxyuridine (EdU) (93% ± 4.20% EdU⁺), indicative of cycling cells, very few Pax7⁺ cells (3% ± 1.5% EdU⁺) present in late-stage SkMOs were EdU⁺ (Fig. 1e,f). Additionally, a quiescent satellite cell can be distinguished from an activated satellite cell by cell size alone^29,30. Consistent with those findings, FACS-sorted late-stage SkMO-derived GFP⁺ cells were significantly smaller (9.15 ± 0.17 µm in diameter) than FACS-sorted GFP⁺ myoblasts (15.39 ± 0.68 µm), more closely resembling freshly isolated GFP⁺ primary satellite cells (6.76 ± 0.05 µm) (Fig. 1g,h). To further confirm that SkMO GFP⁺ cells have acquired properties of satellite cells, we set up a clonal growth assay modified from those used frequently to ascertain stem cell function in the neural and hematopoietic fields^31,32,33,34. Using FACS to sort one GFP⁺ cell per well of a 96-well spheroid plate, we quantified the ability of freshly isolated Pax7^nGFP satellite cells, myoblasts (derived from the satellite cells in two dimensions) and SkMO-derived GFP⁺ cells to form colonies in suspension (Supplementary Fig. 2a). After the confirmation that a single cell resided per well after 24 h in culture (Supplementary Fig. 2b), we found that freshly isolated satellite cells and SkMO-derived GFP⁺ cells, but not myoblasts, could form distinct clones (≥3 cells in a colony) (Supplementary Fig. 2c,d). Quantification of the approximate number of cells per clone revealed that satellite cells and SkMO-derived GFP⁺ cells gave rise to an average of 12.5 ± 0.9 cells per well and 6.9 ± 1.5 cells per well, respectively, whereas myoblasts rarely proliferated under these conditions (1.18 ± 0.16 cells per well) (Supplementary Fig. 2d). The regenerative properties of SkMO GFP⁺ cells were further illustrated by their ability to expand the number of SkMOs when passaged in spin culture (Supplementary Fig. 2e). Finally, we found that purified SkMO-derived GFP⁺ cells, as would be expected, retained their capacity to give rise to proliferative myoblasts capable of fusing to form multinucleated myotubes when cultured under standard differentiation conditions in vitro (Supplementary Fig. 2f).

Thus, using multiple criteria, including loss of MyoD expression, quiescence, cell size and clonal growth, combined with the ability to form differentiated muscle, it appears as if a population of cells resembling satellite cells arises when myoblasts are cultured in organoid-like conditions. For the remainder of this manuscript, we refer to SkMO-derived GFP⁺ cells as idSCs.

Satellite cells and idSCs share transcriptional similarities

To study transcriptional changes that define idSCs as they arise in myoblast-seeded organoids, we conducted multiplexed RNA sequencing (RNA-seq) analyses (3′ DGE) of freshly isolated satellite cells, myoblasts and idSCs, all derived from the Tg:Pax7^nGFP mouse. We further obtained Pax7⁺ reserve cells, which are present in small numbers after traditional 2D differentiation in vitro of Tg:Pax7^nGFP myoblasts into myotubes, as they represent the closest in vitro analogue to satellite cells^35,36 (Supplementary Fig. 3a). A focused analysis of genes known to be involved in satellite cell maintenance/activation¹¹ clustered idSCs with endogenous satellite cells while clearly distinguishing them from both myoblasts and reserve cells (Fig. 2a). Among the genes separating idSCs and satellite cells from myoblasts were genes associated with the Notch signaling pathway (Notch3, HeyL and Jag1), early activation/stress response (Fos, Egr1 and Socs3), quiescence (Spry1, Gas1 and Timp3), cell surface (Cdh15, Cxcr4 and Fgfr4), extracellular matrix (ECM) (Dcn, Dag1 and Col15a1) and secreted growth factors (Igf1, Bmp4 and Wnt4). Quantitative polymerase chain reaction (qPCR) measurements and transcripts per million (TPM) analysis from our bulk RNA-seq dataset confirmed many of these findings (Supplementary Fig. 3b,c).

a, Heatmap depicting genes key to satellite cells, early activation and stress response, quiescence, cell surface receptors, ECM, secreted growth factors and Notch pathway genes colored based on z-score from high (red) to low (blue) (n > 4). b, Venn diagram depicting significant genes (FDR ≤ 0.1) among myoblasts, idSCs and satellite cells. c, Volcano plots depicting log₂ fold change in gene expression between satellite cells versus myoblasts (bottom left), idSCs versus myoblasts (top right), and idSCs versus satellite cells (bottom right). Genes that are significant (FDR ≤ 0.1) are shown in red. d, KEGG pathways identified based on comparing differentially expressed genes among myoblasts, idSCs and satellite cells and visualized based on significance using a hypergeometric test (−log(P value)). Highlighted pathways are directly referenced in the text. Myo, myoblast; SC, satellite cell.

We then filtered this RNA-seq dataset based on differentially expressed genes (false discovery rate (FDR) < 0.1) between myoblasts and satellite cells (Supplementary Table 1), myoblasts and idSCs (Supplementary Table 2) or idSCs and satellite cells (Supplementary Table 3). Next, we conducted unbiased hierarchical clustering to analyze genes within a given clade according to KEGG pathway and curated gene groups. idSCs and satellite cells shared transcriptional patterns observed in clades (7, 4, 1, 3, 9, 8) broadly associated with a reduction in transcription, translation, cell cycle progression and metabolism, along with increases in quiescence, stress response and HDAC activity (Supplementary Fig. 3d and Supplementary Table 4). idSCs and myoblasts differed from satellite cells in clades (10, 11, 12, 13), which represent KEGG pathways involved in oxidative phosphorylation, proteosome activation and carbon metabolism, suggesting that idSCs have a shared, yet distinct, transcriptional profile with respect to myoblasts and satellite cells.

We next conducted differential gene expression analysis (FDR < 0.1) comparing satellite cells, idSCs and myoblasts to further characterize changes that occur at the transcriptional level. We identified 1,347 differentially expressed genes between idSCs and myoblasts, 1,272 genes differentially expressed between satellite cells and myoblasts and only 774 genes differentially expressed between idSCs and satellite cells (Fig. 2b). Volcano plots depicting differentially expressed genes among idSCs, satellite cells and myoblasts confirmed these findings and further demonstrated conserved trends in gene expression between satellite cells and idSCs (Fig. 2c). Functional enrichment analysis identified KEGG pathways for focal adhesion, ECM–receptor interactions, proteoglycans in cancer and the Notch signaling pathway as being conserved and upregulated in idSC and satellite cell samples relative to myoblasts (Fig. 2d). Conversely, myoblasts and idSCs are more similar with respect to their increased activation of KEGG pathways broadly involved in cell cycle and a variety of metabolic pathways (glycolysis, OxPhos, carbon and tricarboxylic acid (TCA) cycle) when compared to satellite cells (Fig. 2d). These data suggest that idSCs are intermediate with respect to their transcriptional profile when compared to both myoblasts and satellite cells.

Epigenomic changes regulate/underlie idSC formation

The organization of the epigenome with respect to chromatin structure is a major regulator of stem cell identity^37,38. We reasoned that a temporal analysis of chromatin dynamics would provide insight into the process by which idSCs are generated when myoblasts are grown in an organoid format. We conducted Omni-ATAC-seq³⁹ to measure changes in the epigenome during the generation of idSCs within SkMOs at early (day 10), mid (day 20) and late (day 30) timepoints (Supplementary Fig. 4a). An initial correlation heatmap distinguished early-stage, mid-stage and late-stage idSC samples from satellite cells and from their parental myoblast lines (Supplementary Fig. 4b). We observed dynamic patterns of change in chromatin accessibility among proliferative myoblasts, idSCs present in SkMOs at the three different timepoints and freshly isolated satellite cells. These regions (or peaks) existed across the spectrum of open and closed chromatin, and we identified approximately 40,000 myogenic responsive chromatin regions (adjusted P < 0.0001, Wald test)⁴⁰ that differed after pairwise comparison among the five cellular states. Unbiased hierarchical clustering identified five clades (1.1, 1.2, 2.1, 2.2 and 3) that distinguished the myogenic responsive chromatin regions. Association of chromatin peaks present in each clade to candidate gene and Gene Ontology (GO) terms was conducted using the Genomic Regions Enrichment of Annotations Tool (GREAT)⁴¹. Clades 1.1 and 1.2 included peaks that were primarily open in myoblasts (myoblast state) and closed in satellite cells and in idSCs starting at day 10 onward (Fig. 3a, dashed box). Peaks found in these clades were enriched for GO Biological Processes (GO BPs) involved in striated muscle development (Actin, Atg5/7, Lmod3, MyoG, Prox1 and Pitx2), contraction (α and β adrenergic receptors, Calm1, Prok2, Ttn and Na/K channel genes), myotube differentiation (MyoD, MyoG, Myocd, Cyp26b1 and Cxcl12) and histone citrullination (Padi gene family) (Fig. 3a,b and Supplementary Table 5). Clade 2.1 is shared between myoblast and mid- to late-stage idSC samples (transition state) and includes GO BPs involved in vascular development (Tcf4, Dcn, Fgf1, Ets1 and Btg1) and cytokine production (Runx1, Fn1, Pik3r1, Thbs1, Csf1r, Trib2, Foxp1, Smad7 and Smad3). Clade 2.2, which is unique to mid-stage and late-stage idSC samples, represents GO BPs involved in actin cytoskeleton organization (Nrp1, Actn1, Foxp1 and Tpm1) and ECM organization (Col5a1, Dpt, Grem1 and Has1). Finally, clade 3, which was predominantly open in satellite cells (satellite cell state) and closed in other samples, is associated with genes involved in regulation of carbohydrate metabolism (Foxo1, Irs2, Prkce and Rora), cellular response to insulin stimulus (Errfi1, Foxo1, Pik3r1 and Zfp36l1) and negative regulation of cell cycle phase transition (Zfp36l1, Zfp36l2, Foxn3 and Cradd) (Fig. 3a,b and Supplementary Table 5). We observed a progressive closure of chromatin accessibility for myoblast state peaks in a temporal fashion across idSC timepoints (early, mid and late stage), ultimately resulting in closed chromatin in satellite cells (Fig. 3a, dashed box). Unbiased pathway analysis involving differential binding (DiffBind) comparisons among myoblasts, day 30 idSCs and satellite cells identified candidate signaling pathways for genes found in open chromatin unique to satellite cells (VEGF signaling, vascular smooth muscle contraction and Tgfβ signaling), shared between myoblasts and day 30 idSCs (Jak/Stat, insulin and mTOR signaling pathways) and candidate pathways shared by both day 30 idSCs and satellite cells (regulation of actin cytoskeleton, MAPK signaling, leukocyte trans-endothelial migration and focal adhesion) (Supplementary Fig. 4c). These broad genome-wide changes suggest that idSCs transition over time to silence the myoblast epigenomic state, in line with satellite cells, yet maintain a unique chromatin profile linked to ECM and actin remodeling.

a, Heatmap depicting dynamic patterns of change in chromatin accessibility that occur between proliferative myoblasts and idSCs at early-stage (day 10), mid-stage (day 20) or late-stage (day 30) culture timepoints and freshly isolated satellite cells. Clustering of peaks into three broad states: myoblast specific, transition state specific and satellite cell specific. Boxed region highlights myoblast specific peaks in an open configuration remodeled into closed chromatin during idSC formation. b, Clade analysis with associated GO BP terms generated by GREAT based on an unbiased dendogram depicting similarities in chromatin accessibility across sample types. c, Heatmap depicting dynamic patterns of change in chromatin accessibility conserved between late-stage idSCs and satellite cells relative to myoblasts. d, Associated GO Cellular Component terms for genes with cis-regulatory peaks conserved between late-stage idSCs and satellite cells using a binomial test. e, Associated GO BP terms for genes with cis-regulatory bound E-box sites specific to myoblasts using a binomial test. f, Footprinting of E-box sites bound in myoblasts across idSC samples and satellite cells. Example shows progressive loss in accessibility over time in cultured idSCs, similar to satellite cell footprinting. SC, satellite cell.

To further study whether epigenetic similarities exist between idSCs and satellite cells, we next conducted a pairwise comparison to identify peaks that are open in both late-stage idSCs and satellite cells and closed in myoblasts (Fig. 3c). GREAT analysis of this refined list of peaks identified GO Cellular Components involved in ECM and cytoskeletal signaling (Dcn, Tgfbr3, Dag1 and Timp3)^42,43 and collagen deposition, most prominently collagen V trimers (Col5a1, a2, a3, Col6a1 and Col6a2), known to promote satellite cell function⁴⁴, along with other collagens present in the satellite cell niche (Col15a1 and Col27a1)^43,45 (Fig. 3d). Further examination of candidate genes known to regulate satellite cell fate identified chromatin closure at the MyoD and MyoG loci (Supplementary Fig. 4d) conserved between late-stage idSCs and satellite cells⁸. Conversely, we observed open chromatin surrounding Notch pathway members, as well as TFs associated with the satellite cell state, in idSC timepoints and satellite cells (Supplementary Fig. 4d). These data highlight the involvement of key ECM molecules, found in the satellite cell niche, as well as Notch signaling, in orchestrating the process by which idSCs arise when myoblasts are cultured in organoid format.

In accordance with other cell types undergoing dedifferentiation, initial transcriptional and epigenomic downregulation of the myogenic program might be expected to occur early in the process, followed by activation of genes highly expressed in satellite cells. We examined whether specific TF binding motifs differed between each distinct cellular state, therefore allowing us to predict TF regulatory networks that characterize the transition from myoblast to idSC. We used chromVAR to calculate a genome-wide chromatin variability score and found that the most dynamically regulated TF binding sites across all samples were predominantly members of the basic leucine zipper (bZip) or basic helix-loop-helix (bHLH) TF families (Supplementary Fig. 4e). These dynamic binding sites include the canonical myogenic bHLH TFs (MyoD, Myf5, Myog and Myf6), along with bZIP TFs, including several members of the activator protein (AP-1) complex. Given that TF binding sites present in open chromatin suggest active transcriptional regulation by the cognate TF, we conducted transcription factor occupancy prediction by investigation of ATAC-seq signal (TOBIAS)⁴⁶ to establish whether a given TF is bound at a specific site. As MyoD is a master TF that regulates myogenic identity, we analyzed myoblast-specific peaks that contained consensus MyoD binding sites that were bound (footprint) using TOBIAS, and then we associated these bound peaks with their nearest gene using GREAT and identified GO BP terms enriched for skeletal muscle fusion/contraction (Fig. 3e). We subsequently asked whether these peaks remain bound in satellite cells and in idSCs. We observed a notable temporal reduction in TF occupancy on MyoD sites normally bound in myoblasts, such that late-stage idSCs and satellite cells were essentially unbound (Fig. 3f), further supporting our findings that idSCs have reduced expression of MyoD protein and RNA (Figs. 1c,d and 2a,c).

These data underscore the closure of chromatin regulating the broad myogenic program, with specific evidence demonstrating that bound MyoD consensus sites are remodeled into closed chromatin during the appearance of idSCs in organoids, making them more similar to what is observed in satellite cells. For these reasons, we decided to compare the in vivo capacity of late-stage (day 30) idSCs to both myoblasts and freshly isolated satellite cells in vivo.

idSCs regenerate muscle on par with satellite cells

Cell functionality is traditionally determined by cell transplantation. To test the in vivo behavior of late-stage idSCs, we infected primary myoblast lines from Tg:Pax7^nGFP reporter mice with a lentivirus expressing luciferase and tdTomato (LT). This dual reporter facilitated FACS isolation of nGFP⁺tdTomato⁺ cells and the detection of donor-derived cells in histological sections, along with the ability to carry out longitudinal in vivo monitoring of transplanted cells using bioluminescent imaging (BLI)^47,48. We induced bilateral cardiotoxin (CTX) injuries in the left and right tibialis anterior (TA) muscles of immunodeficient NSG mice. This allowed us to test in parallel two different types of cells in the same animal, thereby reducing experimental variability⁴⁹. Forty-eight hours after injury, we transplanted 10,000 nGFP⁺tdTomato⁺ idSCs or nGFP⁺tdTomato⁺ parental myoblasts into regenerating TA muscles (Fig. 4a and Supplementary Fig. 5a). We subsequently conducted BLI measurements at 7 d, 14 d and 21 d after transplant to determine the persistence of engrafted cells (Fig. 4b and Supplementary Fig. 5b). Both myoblasts and idSCs displayed detectable radiance 48 h after transplant, confirming the success of the procedure (Supplementary Fig. 5c). However, at later timepoints, the radiance associated with transplanted idSCs progressively increased, whereas myoblast radiance levels steadily decreased (Fig. 4b). After 21 d, we observed a significant increase in radiance from TA muscles transplanted with idSCs (3.2 ± 0.7 million photons cm⁻² s⁻¹) compared to myoblasts (0.007 ± 0.001 million photons cm⁻² s⁻¹). This was observed even when a 50-fold excess of myoblasts was transplanted (Supplementary Fig. 5d). These data show that idSCs efficiently persist after transplantation, whereas myoblasts, in line with previous findings, do not^14,50.

a, Experimental schematic outlining the transplant and analysis of myoblasts or idSCs into regenerating TA muscle of NSG mice. b, Quantification of the average radiance (photons cm⁻² s⁻¹) across a standardized region of interest (ROI) superimposed over NSG hindlimbs over a 21-d timecourse after transplant of 10,000 myoblasts (yellow) or idSCs (green). Data represent mean ± s.e.m (n = 97 day 7; n = 102 day 14; n = 92 day 21, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P values: day 7 0.011; day 14 5.23 × 10⁻⁶; day 21 5.40 × 10⁻⁵; *P < 0.05, ***P < 0.001. c, Representative immunofluorescence micrograph depicting engraftment after the transplant of nGFP⁺tdTomato⁺ myoblasts or idSCs into newly regenerated muscle of NSG mice. Images were stained with laminin (green), tdTomato (red) and Hoechst (blue). Scale bar, 100 µm. d, Quantification of the percentage of tdTomato⁺ fibers per muscle cross-section after transplant of either myoblasts or idSCs. Box bounds represent s.e.m.; bar represents mean (n = 5, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P value: idSC versus Myo 6.70 × 10⁻⁵; ***P < 0.001. e, Representative immunofluorescent micrograph depicting the ability of donor-derived cells to repopulate the satellite cell niche and express Pax7 (green), tdTomato (red) and Hoechst (blue). Arrowheads represent sublaminar Pax7⁺tdTomato⁻ cells within NSG mice. Arrows indicate sublaminar Pax7⁺tdTomato⁺ cells. Scale bar, 100 µm. f, Quantification of the percent of sublaminar Pax7⁺tdTomato⁺ cells relative to the total Pax7⁺ cells per section. Percent double-positive cells was derived from an average total of Pax7⁺tdTomato⁺ to Pax7⁺ cells quantified per replicate (shown in panel). Box bounds represent s.e.m.; bar represents mean (n = 5, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P value: idSC versus Myo 7.24 × 10⁻³; **P < 0.01. g, 63-d longitudinal BLI study examining engraftment of 10,000 idSCs or myoblasts, followed by re-injury of muscle at 21 d and 42 d after transplant. Data represent mean ± s.e.m. (n = 18, days 21, 28, 35 and 42; n = 15, days 44, 49, 56 and 63; n = 3, day 23, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P values: day 21, 0.005; day 23, 0.077; day 28, 0.001; day 35, 0.001; day 42, 0.015; day 44, 0.019; day 49, 0.024; day 56, 0.044; day 63, 0.030; *P < 0.05, **P < 0.01, ***P < 0.001. h, Immunofluorescence micrograph of a TA cross-section before the second round of re-injury (day 42). tdTomato⁺ (red) and laminin (green). Scale bars, 500 µm. i, Percentage of tdTomato⁺ fibers per muscle section (n = 3, biological replicates). Box bounds represent s.e.m.; red line represents mean. Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction. P value: Myo versus idSC 0.012; *P < 0.05. Myo, myoblast.

To explore the capacity of transplanted cells to incorporate into regenerating muscle fibers, we sectioned and stained TA muscles for the presence of tdTomato within myofibers from mice previously imaged by BLI. Under these conditions, donor cells and endogenous host satellite cells are competing to generate new myofibers. Manual injection of cells as well as differences in the area of damage produced by CTX in individual muscles contribute to variability. Notwithstanding, we observed significant engraftment of idSCs compared to myoblasts 21 d after transplant of 10,000 cells into CTX-damaged TA muscles (Fig. 4c). We further quantified this engraftment across entire muscle sections, finding that idSCs gave rise, on average, to 5.8% ± 0.8% tdTomato⁺ fibers per section (193.60 ± 27.01 tdTomato⁺ fibers per section), whereas myoblasts only rarely gave rise to engrafted fibers (0.02% ± 0.02% or 0.80 ± 0.80 tdTomato⁺ fibers per section) (Fig. 4d).

To confirm the ability of idSCs to repopulate the satellite cell niche, we stained transplanted muscle sections to identify donor-derived cells defined by the expression of both Pax7 and tdTomato (Fig. 4e). We found that idSCs gave rise to 44.8 ± 10.76 Pax7⁺tdTomato⁺ cells per section. When compared to the total number of Pax7⁺ cells in muscle sections, idSCs accounted for 14% ± 4% of Pax7⁺tdTomato⁺ cells. In contrast, myoblasts produced a negligible number of Pax7⁺tdTomato⁺ cells, expressed either as a total (0.4 ± 0.24) or as a percentage (0.2% ± 0.1%) (Fig. 4f). Taken together, these data suggest that idSCs have acquired the capacity to repopulate the stem cell niche and strongly resemble satellite cells after transplant into damaged muscle.

To determine whether idSCs can self-renew after damage, we subjected NSG mice, that had previously been transplanted bilaterally with 10,000 myoblasts or idSCs, to two additional rounds of re-injury conducted at days 21 and 42 after transplant (Fig. 4a). Successful self-renewal of these transplanted cells would allow them to participate in muscle repair after multiple rounds of injury. After each episode of CTX-induced injury, we observed enhanced radiance from TA muscles transplanted with idSCs (228 ± 84 million photons cm⁻² s⁻¹ at 42 d and 3.09 ± 1.2 billion photons cm⁻² s⁻¹ at 63 d) (Fig. 4g and Supplementary Fig. 5e). This significant increase in radiance levels after re-injury was observed only in the idSC samples and not in TAs transplanted with parental myoblasts. To further confirm that idSCs were able to support muscle regeneration after injury, we evaluated sectioned muscle for the presence of donor-derived fibers (tdTomato⁺) after one round of re-injury (day 42). Donor-derived fibers were readily observed in idSC-transplanted muscle, whereas they were completely absent in myoblast-transplanted samples (Fig. 4h). After one round of re-injury, idSCs contributed to the formation of new myofibers (43% ± 10% tdTomato⁺ fibers) across the entire muscle section, whereas no discernable tdTomato⁺ fibers were observed after re-injury of TA muscles transplanted with 10,000 myoblasts (Fig. 4i). These findings differ substantially from transplanted muscle that is not re-injured (Supplementary Fig. 5f), underscoring the capacity of idSCs to self-renew, giving rise to cells able to engraft into muscle and repopulate the satellite cell niche.

To rigorously compare unmodified freshly isolated satellite cells, idSCs and myoblasts, we assessed engraftment, repopulation and self-renewal of transplanted cells in a mouse model of muscular dystrophy (mdx^5cv) in which muscle recovery from damage is impaired and revertant myofiber frequency is low (Fig. 5a)⁵¹. We found that the ability of idSCs to engraft (9.6% ± 1.3% dystrophin⁺ fibers per section) was as good as, if not superior to, that of freshly isolated satellite cells (7.3% ± 1.07% dystrophin⁺ fibers per section) (Fig. 5b,c). After re-injury, idSC engraftment appeared better than that achieved with freshly isolated satellite cells (21.5% ± 3.78% dystrophin⁺ fibers per section and 7.4% ± 4.1% dystrophin⁺ fibers per section, respectively) (Fig. 5c). Furthermore, the number of cells repopulating the satellite cell niche was similar at day 21 (9.9% ± 2.1% of Pax7⁺nGFP⁺ cells for idSCs and 10.9% ± 2.5% of Pax7⁺nGFP⁺ cells for satellite cells), whereas day 42 displayed a greater degree of repopulation for idSCs relative to satellite cells (28.2% ± 3.5% of Pax7⁺nGFP⁺ cells and 12.9% ± 3.2% of Pax7⁺nGFP⁺ cells, respectively) (Fig. 5d,e). At day 42, after re-injury, most of both populations (70.9% ± 8.9% of idSCs and 51.8% ± 7.7% of satellite cells) have withdrawn from the cell cycle and are Pax7⁺nGFP⁺Ki67⁻ (Fig. 5f,g). As expected, donor myoblasts were unable to engraft or repopulate the satellite cell niche in damaged mdx^5cv muscle (0.31% ± 0.1% dystrophin⁺ fibers per section and 0.7% ± 0.3% of Pax7⁺nGFP⁺ cells) and, when proliferated in vitro, were rarely Pax7⁺nGFP⁺Ki67⁻ (13.5% ± 4.1%) (Supplementary Fig. 6a,b). Together, these data indicate that, despite being somewhat different from satellite cells based on transcriptional and epigenomic signatures, the in vivo behavior of transplanted idSCs is nearly equivalent to that of bona fide satellite cells.

a, Schematic summarizing isolation and transplant timelines. b, Representative immunofluorescence micrograph depicting engraftment after the transplant of 10,000 FACS-sorted nGFP⁺ myoblasts, idSCs or freshly isolated satellite cells into irradiated (IRR) (13 Gy) TA muscles of mdx^5cv mice. Images were stained with dystrophin (green), laminin (red) and Hoechst (blue). Scale bar, 100 µm. c, Quantification of the percentage of dystrophin⁺ fibers relative to the total laminin⁺ fibers per muscle cross-section. Box bounds represent s.e.m.; bar represents mean (day 21: n = 8 Myo, n = 10 idSC, n = 7 SC; day 42: n = 9 idSC, n = 4 SC, biological replicates) One-way ANOVA with Tukey HSD analysis conducted at day 21 and day 42 between samples. Day 21 ANOVA P value 6.82 × 10⁻⁶, Tukey HSD adjusted P values: day 21 idSC versus Myo 5.48 × 10⁻⁶, day 21 idSC versus SC 0.292, day 21 Myo versus SC 6.26 × 10⁻⁴; day 42 ANOVA P value 0.049. *P < 0.05, **P < 0.01, ***P < 0.001. d, Representative immunofluorescent micrograph depicting the ability of donor-derived cells to repopulate the satellite cell niche and express Pax7 (green), nGFP (red), laminin (white) and Hoechst (blue). Arrowheads represent sublaminar Pax7⁺nGFP⁻ (endogenous) satellite cells within mdx^5cv mice. Arrows indicate sublaminar Pax7⁺nGFP⁺ cells (donor derived). Scale bars, 100 µm. e, Quantification of the percent of sublaminar Pax7⁺nGFP⁺ cells per section. Box bounds repreent s.e.m.; bar represents mean (day 21 n = 8 Myo, n = 6 idSC, n = 7 SC; day 42 n = 8 idSC, n = 3 SC, biological replicates). One-way ANOVA with Tukey HSD analysis conducted at day 21 and day 42 between samples. Day 21 ANOVA P value 7.31 × 10⁻⁴, Tukey HSD adjusted P values: idSC versus Myo 4.73 × 10⁻³, idSC versus SC 0.91, Myo versus SC 1.26 × 10⁻³. Day 42 ANOVA P value 0.035. *P < 0.05, **P < 0.01, ***P < 0.001. Raw counts for the number of Pax7⁺nGFP⁺/total Pax7⁺ are displayed at the top of the panel. f, Representative immunofluorescent micrograph depicting Pax7, nGFP, Ki67 and Hoechst. Arrows represent Pax7⁺nGFP⁺ cells (donor derived). Scale bars, 25 µm. g, Quantification of the percent of sublaminar Pax7⁺nGFP⁺Ki67⁻ cells per section. Data represent mean ± s.e.m. (n = 8 idSC; n = 5 SC, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction between samples. P value: SC versus idSC 0.166. Raw counts for the number of Pax7⁺nGFP⁺Ki67⁻/total Pax7⁺nGFP⁺ are displayed at the top of the panel. HSD, honestly significant difference; NS, not significant; SC, satellite cell.

idSCs restore muscle function on par with satellite cells

To determine whether transplanted idSCs could improve the function of damaged muscle, another test of their similarity to satellite cells, we irradiated TA muscles of NSG mice (13 Gy), induced damage with CTX and subsequently transplanted 10,000 myoblasts, idSCs or satellite cells (Fig. 6a). Irradiation of hindlimb muscle prevents endogenous satellite cells from entering the cell cycle after muscle damage⁵², meaning that the only cells capable of repairing muscle would be those provided by transplantation. We observed a substantial difference in specific force production between transplanted idSCs and myoblasts. idSCs gave rise to 138 ± 20.49 kN × m⁻², similar to that produced by satellite cells (178 ± 16.43 kN × m⁻²), whereas myoblasts gave rise to negligible specific force readings (3 ± 0.88 kN × m⁻²) (Fig. 6b). These findings were further supported when assessing maximum tetanic and twitch forces (Supplementary Fig. 7a–d). Muscle force data were strongly supported by histological evaluation of the treated muscle at the conclusion of the study (Fig. 6c) and by quantifying the total number of myofibers per section obtained after the transplant of either satellite cells or idSCs (1,535 ± 145 fibers and 1,965 ± 313 fibers, respectively) (Fig. 6d). In contrast, sham (saline) and myoblast samples had much fewer fibers after regeneration (14 ± 8 fibers and 84 ± 12 fibers, respectively). These findings correlated with the percentage of centrally located nuclei in transplanted samples, a hallmark of newly regenerated muscle (Supplementary Fig. 7e), and with TA muscle weights (Supplementary Fig. 7f). Irradiated undamaged muscle was devoid of centrally nucleated myofibers consistent with the low levels of satellite cell turnover under normal conditions (Fig. 1f and Supplementary Fig. 7e). Finally, both idSCs and satellite cells reoccupied the satellite cell niche to a similar extent after regeneration (Supplementary Fig. 7g,h), further supporting the ability of late-stage idSCs to repopulate the satellite cell niche.

a, Experimental schematic outlining the transplant of 10,000 nGFP⁺ cells (myoblasts, idSCs or freshly isolated satellite cells) into irradiated (IRR) and subsequently damaged TA muscles of NSG mice. b, Specific maximum tetanic force assessed on TA muscles 21 d after CTX damage. Undamaged muscle and sham transplants gave rise to specific force values of 278 ± 20.24 kN × m⁻² and 8 ± 2.47 kN × m⁻², respectively. Box bounds represent s.e.m.; red line represents mean (n = 6 undamaged, n = 8 sham, n = 15 Myo, n = 12 SC, n = 12 idSC, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction between samples. P values: undamaged versus idSC 2.56 × 10⁻⁷; undamaged versus SC 9.02 × 10⁻⁵; Myo versus idSC 1.37 × 10⁻⁹; Myo versus SC 6.88 × 10⁻¹³; SC versus idSC 0.0405; Myo versus sham 0.818; *P < 0.05, ***P < 0.001. c, Stitched hematoxylin and eosin micrographs of cross-sectioned TA muscle 21 d after cell transplant. Scale bars, 100 µm. d, Total muscle fibers per section across samples. Box bounds represent s.e.m.; red line represents mean (n = 3, biological replicates). Student’s two-tailed t-test, unpaired with equal variance assumed and no multiple testing correction between samples. P values: undamaged versus idSC 0.358; undamaged versus SC 0.0175; Myo versus idSC 9.22 × 10⁻⁶; SC versus idSC 0.0897; Myo versus SC 8.39 × 10⁻⁵; Myo versus sham 0.804; *P < 0.05, **P < 0.01, ***P < 0.001. NS, not significant; SC, satellite cell.

To purify idSCs without relying on a reporter, we considered the different expression of cell surface markers on idSCs and myoblasts (Supplementary Fig. 1e) and identified a combination of CD9⁺ and CD104⁻ that effectively isolates idSCs (Supplementary Fig. 8a). Validation against nGFP⁺ idSCs resulted in 98.2% ± 2.2% CD9⁺CD104⁻, whereas the converse CD9⁺CD104⁻ population contained 87.6% ± 6.6% nGFP⁺ cells (Supplementary Fig. 8a). We subsequently transplanted 10,000 CD9⁺CD104⁻ sorted cells into TA muscles previously irradiated and damaged with CTX (Supplementary Fig. 8b). CD9⁺CD104⁻ idSCs, when compared to nGFP⁺ sorted idSCs, resulted in similar levels of muscle formation (Supplementary Fig. 8c), the presence of donor cells repopulating the satellite cell niche (Supplementary Fig. 8d) and similar mean specific force values (163 ± 22.7 kN × m⁻² relative to nGFP⁺ sorted 138 ± 20.49 kN × m⁻²) (Supplementary Fig. 8e). In addition, mean maximum force values (Supplementary Fig. 8f), TA muscle weights (Supplementary Fig. 8g) and percentage of centrally located nuclei (Supplementary Fig. 8h) in nGFP⁺ and CD9⁺CD104⁻ transplants were comparable.

Recapitulation of idSC formation in human SkMOs

To make our findings more clinically relevant, we obtained human myoblast lines from patient biopsies or purchased them from commercial sources. We subsequently generated human SkMOs in an enhanced serum-free medium containing well-characterized components and limited animal products using a shortened 15-d protocol (Supplementary Fig. 9a)⁵³. We observed subtle differences between the expression of human myogenic cell surface markers assessed at the myoblast stage and after 15 d of SkMO culture (Supplementary Fig. 9b). Although variable between replicates, %PAX7⁺ cells are rare across human myoblast lines (3.5% ± 3%) (Supplementary Fig. 9c,d). After the formation of human SkMOs, the percentage of PAX7⁺ cells decreases in variability across replicates and accounts for 5% ± 1% of total cells (Supplementary Fig. 9c,d). Notably, whereas PAX7⁺ cells in myoblasts are predominantly MYOD⁺ (84% ± 2%), in human SkMOs a minority are MYOD⁺ (5% ± 2%) (Supplementary Fig. 9c,e). Consistent with these findings, we observe a global decrease in the percentage of EdU⁺ cells when transitioning from proliferating myoblasts to human SkMOs (Supplementary Fig. 9f,g), resulting in a reduction in the percentage of PAX7⁺EdU⁺ cells between myoblasts (30% ± 22%) and human SkMOs (0.8% ± 1.7%) (Supplementary Fig. 9h). These findings suggest that SkMO formation drives myoblast dedifferentiation across species to a quiescent state whereby Pax7 is maintained while MyoD expression is absent.