SMPCs associate with human-only myofibres in vivo
We previously showed hPSC SMPCs can engraft and regenerate hundreds of dystrophin+ myofibres in immunocompromised mouse models of Duchenne muscular dystrophy (mdx-NSG) following cardiotoxin (CTX) injury20. We have continued to improve myogenic potential by supplementing survival/maturation factors during directed differentiation of hPSCs in vitro, as well as reducing culture time of ERBB3+NGFR+PAX7+ hPSC SMPCs before engraftment (Fig. 1a and Extended Data Fig. 1a–c). In vivo, we show that, while hPSC SMPCs fused with mouse skeletal muscle to form hundreds of multinucleated chimeric myofibres (maximum number 75–230 per cross sectional area (CSA), N = 7 mice), hPSC SMPCs also generated about twice as many human-only myofibres (maximum number 140–450 per CSA), which occupied large regions of engrafted skeletal muscle (Supplementary Video 1). Myofibre size as measured by Imaris software and exclusivity of human nuclei marked by human lamin AC were used to distinguish and quantify chimeric and human-only myofibres in transverse and longitudinal sections (Fig. 1a and Extended Data Fig. 2a).
An effective cell therapy must maintain the muscle stem cell pool. Yet so far retention of human PAX7 cells after transplantation of hPSC or foetal SMPCs has proven challenging. Thus, we were encouraged to find 20 or more human PAX7+ hPSC SMPCs within 0.15 mm2 (20×). A total of 127 human PAX7+ cells were counted and categorized on the basis of proximity to human-only myofibres, chimeric myofibres, interstitial space or regions with high numbers of human lamin AC cells but few myofibres (Fig. 1b). We found regions containing the greatest numbers of human-only myofibres also contained 98% of all engrafted PAX7 hPSC SMPCs. In contrast, regions containing only lamin AC, but few or no human-only myofibres, did not contain PAX7 cells. We also evaluated chimeric myofibre regions, and found these regions contained the other 2% of PAX7 cells (P < 0.05, Fig. 1b and Extended Data Fig. 2b). We also evaluated hPSC SMPC engraftment in another more severe model of muscular dystrophy, the mdx-DBA/2-NSG mice (Extended Data Fig. 2b). We similarly found that hPSC SMPCs form both human-only and chimeric myofibres, but PAX7+ cells predominately associated with the human-only myofibres (Fig. 1b). These data suggested that human-only myofibres were better at supporting PAX7 SMPC retention than chimeric myofibres or other regions within the skeletal muscle microenvironment.
In vivo niches from SMPCs resemble early foetal niches
Recent work from our lab has shown that directed differentiation of hPSCs generates embryonic or foetal-like skeletal muscle in vitro27 and we hypothesized that engrafted hPSC SMPCs would produce similar immature skeletal muscle in vivo. To understand the developmental trajectory of transplanted PAX7+ hPSC SMPCs, we compared PAX7 cells and their niches with human skeletal muscle tissues from embryonic weeks 9–11, foetal weeks 16–20 and adult years 25–40 (Fig. 2a). We found embryonic and foetal myofibres were ∼30-fold smaller than adult myofibres, and this was similar to the small human-only myofibres observed after hPSC SMPC transplantation. PAX7+ cells in human embryonic week 10 tissues closely associated with the sarcolemma of embryonic myofibres that clustered together but lacked basal lamina expression of laminin. As myofibres matured from the embryonic to foetal stage, PAX7+ cells associated with the sarcolemma became encompassed by laminin, demonstrating the emergence of human niches (Fig. 2a). Here the interior sarcolemma lacked a laminin-rich basal lamina, but formed spectrin+ cross bridges, which are a known hallmark of myogenic fusion36. Similarly, several engrafted hPSC myofibres could be found in regions undergoing potential fusion in myobundles marked by spectrin that were encompassed by laminin basal lamina. Both foetal week 20 tissues and engrafted hPSCs contained PAX7+ SMPCs within these myobundles (Fig. 2b). These data suggested hPSC SMPCs formed niches that more closely resembled a foetal week 20 phenotype upon engraftment in vivo. This was different than adult SC niches/myofibres, which are primarily in homeostatic state where all spectrin+ sarcolemma was closely juxtaposed to a laminin-rich basal lamina.
SMPCs and SCs differ in niche occupancy in vivo
We next set out to compare niche acquisition ability of human muscle stem and progenitor cells. We performed fluorescence-activated cell sorting enrichment on embryonic/foetal SMPCs (weeks 8–20, Lin-ERBB3+NGFR+) or adult SCs (years 25–60, Lin-CD82+NCAM+) dissociated from primary human tissues and immediately transplanted human cells into mdx-NSG mice. These data were used to determine whether niche acquisition resembled the endogenous niches from which the stem and progenitor cells were derived.
We first tested engraftment efficiency across a range of foetal SMPC numbers, from 10,000 to 500,000. Foetal SMPCs (weeks 17–20) produced >100 human myofibres per cross section when 500,000 foetal SMPCs were transplanted, but at lower densities, foetal SMPCs did not restore dystrophin efficiently and only 0–1 PAX7 cells remained at day 30 (Extended Data Fig. 3). Due to embryonic tissue size (weeks 8–10) we could only obtain 50–75,000 embryonic SMPCs, and these formed myofibres poorly upon engraftment and no PAX7 cells were found, which could have been related to survival or immaturity as seen in other stem cell systems37. In contrast, adult SCs fused to form dystrophin+ myofibres and retained 5–20 PAX7+ SCs per cross section at 30 days in vivo with as few as 5,000 transplanted SCs (Fig. 2c and Extended Data Fig. 3). Furthermore, more than 80% of PAX7+ adult SCs were located in human–mouse chimeric myofibre niches 30 days post engraftment. Thus, chimeric niche acquisition is not a species compatibility issue (Fig. 2c). Foetal SMPC ability to colonize niches was intermediate between hPSC and adult SCs as we found that ∼20% foetal SMPCs were associated with small human-only myofibres and ∼30% were found inside chimeric myofibres. In contrast, 80% of PAX7+ hPSC SMPCs were associated within clusters of human-only myofibres and less than 5% were found within chimeric niches (Fig. 2c). These data suggest that both foetal and hPSC SMPCs can associate with human-only myofibres and ability for niche entry differs across development and adult.
SC-ablated mice enable human PAX7 retention
As it was unclear why human PAX7+ hPSC and foetal SMPCs were less able to form niches with chimeric myofibres, we tested two hypotheses related to niche entry and regeneration illustrated in Fig. 3a. We generated a new inducible Pax7 ablation dystrophin model by crossing dystrophic mdx-NSG mice to Pax7-Cre/Ert2 mice and Rosa-DTX mice. When these mice are treated with tamoxifen (TMX), expression of Pax7-driven Cre recombinase removes LoxP stop codons to induce expression of diphtheria toxin (DTX) resulting in SC death38,39 (Fig. 3b). Thus, in this model, regeneration and niche formation would preferentially occur from transplanted human cells as competition with mouse SCs was removed. Pax7-DTX mice were evaluated for SC ablation following 7 days of TMX-infused chow, which we found was sufficient to ablate greater than 80% of endogenous Pax7 SCs relative to non-ablated controls (Fig. 3b). Following CTX- or BaCl2-induced injury of tibialis anterior muscles, Pax7 ablation resulted in decreased weight and cross-sectional area, demonstrating that Pax7 cells are required for muscle repair, but did not result in complete loss of all myofibres (37 mg versus 62 mg, P < 0.05) (Fig. 3b). We noted a subset of myofibres underwent hypertrophy after Pax7 ablation, growing up to four-fold larger than non-ablated controls, demonstrating myofibres are phenotypically influenced by loss of Pax7 SCs and their niches (P < 0.05, Fig. 3b and Extended Data Fig. 4a).
We next tested the ability of human foetal SMPCs to engraft and reside in the niche of Pax7-ablated mdx-NSG mice. We found a two-fold increase in PAX7+ foetal SMPCs 30 days post engraftment (P < 0.05, Fig. 3c). We evaluated the location of foetal PAX7+ SMPCs and surprisingly found these were less in mouse SC niches, and more associated with small human fibres upon mouse SC ablation (P < 0.05, Fig. 3d). While in some cases engrafted foetal SMPCs fused with mouse to form chimeric myofibres in mdx-NSG Pax7-ablated mice, we found foetal SMPCs predominately formed small human-only myofibres. Interestingly, adult SC engraftment in Pax7-DTX mice also followed a foetal trajectory where adult SCs were more likely to be found near small human-only myofibres (Fig. 5a,b and Extended Data Figs. 4b and 6c). Together, these data point to the idea that, in the absence of mouse Pax7 SCs, human-only myofibres are left in a muscle microenvironment that promotes a developmental or regenerative-like paradigm that enables the support of engrafted PAX7 SMPCs.
Myofibres that support SC niches are lost upon SC ablation
While it is well known that Pax7 cells are essential for myofibre formation39,40,41, it is less clear how the SC niche and regeneration are affected by the absence of Pax7 cells. Therefore, we set out to identify which genes or populations were missing in the myofibres of Pax7-ablated mice. Mouse SC niches are re-established in parallel with new myofibre formation 7–10 days post-injury (dpi)10,42. Thus we selected this timepoint post injury to study niche formation, which we confirmed by embryonic myosin heavy chain (eMyHC) expression and increased mouse Pax7 SCs in mdx-NSG mice 8 dpi, and these cell types were both lost in mdx-NSG Pax7-DTX mice (Fig. 4a). We performed single-nucleus RNA sequencing (snRNA-seq) on Pax7-ablated muscle compared with non-ablated and regenerating muscle 8 dpi, and used recently published datasets as a starting point to identify new and known populations in our ablation models43,44,45. In total 14,300 nuclei were sequenced, enabling identification of several population shifts between regenerating, non-ablated and Pax7-ablated myofibres (Fig. 4b, deposited as GSE241368).
Seurat analysis identified a total of 21 populations including 10 myonuclear and 11 mononuclear populations (Supplementary Table 1). We found a two-fold influx in mononucleated cells in regenerating mdx-NSG mice 8 dpi, which included immune, stromal and myogenic populations (Fig. 4c). In contrast, Pax7 ablation had a blunted influx of mononucleated cells 8 dpi. Macrophages and Itga7+Vim+ myoblasts (MB.2, described by ref. 46) were the only mononuclear populations to increase in Pax7-ablated mice. Seurat analysis identified a Pax7+Myod1− SC population (SC.1) which as expected was eliminated in Pax7 ablated muscle (0.1% ablated versus 5.6% non-ablated). In non-ablated muscle, Myod1+Ki67+ myoblasts (MB.1, 0.6% versus 4.5%) and Myog+Myh3+ myocytes (MC, 1.3% versus 4.2%) increased 8 dpi compared with no injury, respectively, but upon Pax7 ablation the regenerative response of myoblasts (0.7% versus 2.6%) and myocytes (0.6% versus 1.4%) was diminished 8 dpi, validating that regeneration was absent following Pax7 ablation (Fig. 4c).
We also found several changes in myonuclear populations. As sequencing was performed using tibialis anterior, the majority of myonuclei were fast-twitch glycolytic type IIB myofibres (Myh4+, 59–70%), while a minority of myofibres were type IIX (Myh1, 24–33%) or type IIA (Myh2, 3–4%) (Supplementary Table 1). Type IIB myofibres were subdivided into four populations identified as follows: IIB.1, a homeostatic Myh4 subcluster; IIB.2, marked by the high-glycolytic-capacity genes (Car3, Pkbfb3 and Amy1); IIB.3, marked by sarcomere assembly genes (Actc1, Tmod1 and Myhbp3); and IIB.4, which expressed mitotic inhibition (Trp63 and Gadd45a) and myogenic growth genes (Fst1 and Igfn1). Upon Pax7 ablation, Car3+ IIB.2 myonuclei increased by 2.4-fold relative to non-ablated controls, suggesting these are longer lasting and less regenerative myofibres (Fig. 4c). In non-ablated mice, upon injury the percentage of all type IIB myofibre populations decreased, except for Actc1+ IIB.3 myofibres, which increased at 8 dpi and was two-fold greater than Pax7-ablated muscle, suggesting Actc1 marked a regenerative myonuclear population. Uniform Manifold Approximation and Projection (UMAP) reduction of myogenic regenerative response highlights that Actc1+ myofibres are intermediary between Myog+Myh3+ myocytes and more mature Myh4+ myofibres (Fig. 4c).
Since Actc1 was the highest-enriched gene in IIB.3 myonuclei, we confirmed with immunostaining that Actc1 protein expression overlaps with timing of SC niche establishment during mouse muscle regeneration (Extended Data Fig. 6a). In contrast, while embryonic myosin protein is co-expressed on Actc1+ myofibres, the Myh3 RNA expression is gone at 8 dpi. Thus, Actc1 expression may be longer lasting than Myh3 and may mark a myofibre important for supporting new SC niche formation. Our snRNA-seq analysis identified 134 upregulated genes expressed by Actc1+ (IIB.3) myofibres, which included genes important for sarcomere formation, nucleotide and lipid metabolism, mitochondrial energetics and calcium regulation (for example, Tnnt3, Eya4, Prkag3, Ampd1 and Casq1). These myofibres also expressed membrane-associated gene signatures related to Notch, Integrin and Jak/Stat signalling (Fig. 4d, Extended Data Fig. 5 and Supplementary Table 2). We used CellChat analysis47 to analyse ligand–receptor niche interactions, which identified increased laminin signalling between Actc1+ myofibres and mouse Pax7+ SCs compared with other myofibre populations (Fig. 4d).
ACTC1+ myofibres enable repopulation in SC-ablated mice
As the Actc1+ gene was the highest-expressed candidate found in our snRNA-seq data in regenerating mouse satellite cell niches, we evaluated if ACTC1 shares common functions with the human-only myofibres supporting PAX7+ SMPC engraftment. ACTC1 is also known to be expressed during foetal skeletal muscle development48, which further supported the likelihood that ACTC1 would be important in the emerging niches formed by hPSC SMPCs. Indeed, we found that human-only myofibres from hPSC and foetal SMPCs strongly expressed ACTC1 following engraftment, but in myofibres that we identified as chimeric, ACTC1 was not expressed at 30 days (Fig. 5a and Extended Data Fig. 6b). To further demonstrate that ACTC1 was expressed by human-only myofibres, but not chimeric myofibres, we took advantage of mouse-specific myosin expression Myh4 in tibialis anterior (TA), which human skeletal muscle does not express. Staining confirmed the presence of chimeric myofibres (lamin AC+spectrin+Myh4+) and human-only myofibres (ACTC1+Myh4−), in which the smallest human myofibres had the greatest ACTC1 expression (Fig. 5b). ACTC1+ human myofibres were more prominent when engrafted in mdx-NSG Pax7-DTX mice, and moreover adult SCs now predominately formed ACTC1+ human-only myofibres in Pax7-ablated mice, suggesting that, when mouse regeneration is impeded, human SC niche formation initially occurs with regenerating human-only myofibres (Extended Data Fig. 6c).
Since we found that SC niche formation is predominantly occurring near human-only fibres in mdx-NSG Pax7-DTX mice we set out to evaluate whether these niches were functional. The hallmark in vivo assay for evaluating true muscle stem cell potential is the re-injury assay, in which PAX7 cells must regenerate new myofibres as well as maintain the stem cell pool12. After 30 days, engrafted muscles are re-injured using CTX, which damages myofibres but preserves PAX7 cell regeneration ability49. As controls for re-injury, day 30 and uninjured day 60 engrafted tissues are used. We found human adult SC and foetal SMPCs were able to form dystrophin+ myofibres and retain some PAX7+ cells at all timepoints, but engraftment dynamics were very different across mouse models (Fig. 5c and Extended Data Fig. 7a). In mdx-NSG mice, adult SCs where able to sustain PAX7+ SC numbers in chimeric niches following injury, but foetal PAX7+ SMPCs decreased by >2-fold, and the ratio of PAX7 foetal SMPCs in chimeric niches decreased by 43% (Fig. 5c). After re-injury, 1 in 3 transplanted adult SCs expressed PAX7, whereas only 1 in 15 foetal SMPCs expressed PAX7 (Extended Data Fig. 7b). In contrast, in the mdx-NSG Pax7-DTX mice, we found after re-injury there was a notable increase in the ability of foetal SMPCs to repopulate both human-only and chimeric niches (Fig. 5c). We also found several instances where two or more PAX7+ foetal SMPCs were within 5 µm of each other, which is suggestive of daughter foetal SMPCs derived from a common parent following re-injury (Fig. 5c). These data show that the increase in ACTC1+ human-only myofibres in the mdx-NSG Pax7-DTX mice can support SC niche formation and repopulation after re-injury.
Spatial analysis identifies key emerging niche pathways
To better define the ACTC1+ niche supporting human PAX7 cells, we optimized spatial RNA sequencing (RNA-seq) and evaluated the niche in both long-term engraftment and after re-injury (Fig. 6a). To do this, we used the Nanostring GeoMx platform, which enables cell type and location specific analysis of gene expression at single antibody resolution. GeoMx allows for region of interest (ROI) selection and segmentation on PAX7 cells and ACTC1+ and ACTC1− myofibres (Supplementary Tables 3 and 4). In brief, we engrafted foetal SMPCs for up to 4 months and induced an injury in a subset of engrafted mice midway at 2 months. We then mounted flash-frozen muscle sections on microscope slides, hybridized a whole transcriptome human RNA oligo library onto the muscle, and stained with morphology markers ACTC1, PAX7, lamin AC and spectrin to identify human muscle engrafted in mouse. After ROI selection and segmentation, the RNA oligos are photocleaved and collected for sequencing (N = 44 ROIs, Extended Data Fig. 10, deposited as GSE243875). Principal component analysis (PCA) found distinct clustering of cell types including ACTC1+ and ACTC1− myofibres (Fig. 6b), and our data confirmed a robust repopulation ability by foetal SMPCs in mdx-NSG Pax7-DTX mice that was sustained through 4 months (Fig. 6c). At the time of re-injury, we identified unique gene signatures in the ACTC1+ myofibres, which included increased embryonic and foetal myosins, ECM development and cell adhesion. In contrast the ACTC1− myofibres had higher expression of metabolism genes and maturation genes. We additionally performed DEG analysis of PAX7 foetal SMPCs after repopulation versus PAX7 foetal SMPCs that remained over a 4-month time period (Supplementary Table 5). Foetal PAX7 SMPCs that had undergone repopulation expressed many genes that could be unique candidates associated with new niche formation including NOTCH4, ITGB1, FGFR1 and MEGF10 as well as stem cell maintenance, chromatin organization and negative regulation of cell proliferation (Fig. 6c,d). We then used CellChat analysis to predict receptor–ligand interactions between the ACTC1+ myofibres and the PAX7 foetal SMPCs identified potential candidates associated with emerging niches including robust association of DAG1 on PAX7 cells and LAMA5 as well as LAMC1 on ACTC1-positive myofibres. Interestingly, THBS4 was a top candidate on ACTC1+ myofibres interacting strongly with receptors on PAX7 cells including integrins and with other myofibres through CD36, a regulator of lipid metabolism and fatty acid uptake (Fig. 6d). Thus, spatial analysis revealed several potential interacting pathways supporting emerging niche formation in vivo in addition to ACTC1.
ACTC1+ myofibres are essential for PAX7 in emerging niches
To definitively prove that ACTC1 myofibres are critical for supporting PAX7 cells, we followed recent approaches that have introduced an inducible suicide gene in hPSCs50,51. HPSCs are uniquely able to be genetically manipulated while retaining pluripotency and ability to differentiate into downstream lineages. We engineered an inducible apoptosis system that included a FKBP12 linked to a caspase 9 (iCasp9) and a self-cleaving T2A peptide immediately following a homology arm of the last 3′ base pair in the ACTC1 gene (Extended Data Fig. 8a). Treatment with the bio-inert small molecule AP1903 induces cross-linking of the drug-binding domains of this chimeric protein, which in turn dimerizes caspase 9 and activates the downstream executioner caspase 3 molecule, resulting in gene-specific cellular apoptosis52 (Fig. 7a). To determine the optimal conditions for inserting iCasp9 by homology-directed repair, we tested 17 conditions in H9 and H1 hPSC lines including 3 guide RNA (gRNA) targets, lipo-STEM and nucleofection, and compared homology-directed repair protocols described by Skarnes et al., to the protocol provided by Lonza (ref. 53 and Extended Data Fig. 8b). We confirmed the presence of inserted iCasp9 in both H9 and H1 lines following nucleofection (Fig. 7a) and proceeded to single-cell cloning in 384 wells. Of these, 108 clones survived, and we found 2 clones that contained a heterozygous insertion for iCasp9. We performed sequencing to confirm an exact match for our engineered iCasp9 vector within the ACTC1 locus (Extended Data Fig. 8c).
We next performed directed differentiation of ACTC1-iCasp9 hPSCs to generate SMPCs for engraftment. We used Pax7-ablated mdx-D2 NSG Pax7-DTX mice as these mice readily enable human-only myofibre formation. Two weeks following engraftment, mice were treated with AP1903 (10 μM in 100 μl saline) or control (0.1% dimethyl sulfoxide in saline) every two weeks for 90 days. Similar to previous results, in non-treated mice we found hundreds of hPSC myofibres that were able to engraft over approximately 6 mm3 of mouse TA, and we counted up to 90 PAX7+ hPSC SMPCs associated with hPSC myofibres within a 1.5-mm2 region of tissue (Fig. 7b and Extended Data Fig. 9a). AP1903 treatment induced a ∼90% reduction in hPSC myofibres, as we found fewer than 20 myofibres per cross section in treated mice. Remarkably the apoptosis of ACTC1+ myofibres resulted in an almost complete elimination of PAX7 cells in the AP1903-treated mice. We counted a total of 269 human PAX7 cells in five regions of control mice and only 5 PAX7 cells within the same engrafted regions of AP1903-treated mice (Fig. 7b,c). However, in the AP1903-treated groups there were hundreds of remaining lamin AC+ cells that were neither myofibres nor PAX7+.
We further validated the iCasp9 system using GeoMx spatial sequencing on engrafted hPSC muscle cells with and without AP1903 treatment (ROI, N = 59). Spatial analysis showed an abundance of cell death-related pathways including several upregulated caspase genes in the AP1903-treated muscles but not in control muscles (Fig. 7c,d and Supplementary Table 6). We compared the remaining hPSC lamin AC+ cells (called hPSC PAX7−) with hPSC PAX7+ cells, and these expressed genes enriched in non-stem cell or non-myogenic related cell fate change including POSTN, CDH11 and PDGFRA, suggesting that in the absence of human myofibres these cells change fate to other mesoderm lineages, whereas the PAX7+ cells expressed key stem cell genes such as ITGA7, NOTCH3 and CDH15 (Extended Data Fig. 9 and Supplementary Table 7). To better define how the hPSC SMPCs are interacting with the emerging fibres we compared the PAX7 cells with the hPSC myofibres in vivo again using spatial RNA-seq. We also discovered that key stem cell signalling and adhesion pathways are turned on demonstrating maintenance of stem cell fate in this model when ACTC1-positive fibres are present, indicating support of human PAX7 cells by these emerging niches (Fig. 7d). To verify this, we performed CellChat analysis and found that hPSC myofibres send several signals through embryonic laminins and delta ligands (LAMA1 and DLK1) that are predicted to interact with receptors (NOTCH3, FGFR4, DAG1, ITGA7 and CD44) on the hPSC PAX7+ SMPCs (Fig. 7e and Supplementary Table 8). These identified key stemness pathways are turned on, which could be key to supporting hPSC PAX7 cells in emerging niches.