Main

Mitochondrial ETC dysfunction impacts the human population in multiple ways. Germline mutations in the mitochondrial and nuclear genome are estimated to affect as many as 1:5,000 individuals1, and accumulating ETC dysfunction is also observed as a secondary component of many common diseases, including aging itself2,3,4. Examining the mechanistic consequences of ETC dysfunction in various differentiated tissues has therefore been important to understanding disease physiology and progression. In contrast, our current knowledge of the consequences of ETC dysfunction in stem cell compartments has been understudied, despite the regenerative potential of these cells. In hematopoietic stem cells, depletion of complex III or mitochondrial DNA (mtDNA) results in impaired differentiation and anemia, as well as stem cell exhaustion5,6. However, despite numerous studies of metabolic changes in various stem cell types, a direct analysis of the consequences of ETC dysfunction is largely missing, including in MuSCs. MuSCs are largely quiescent and nonproliferative during adult life. In response to injury, quiescent MuSCs undergo rapid activation and proliferation, followed by cell fusion to regenerate new myofibers. Activated MuSCs exhibit altered metabolism to drive differentiation and have an high capacity to expand their mitochondrial population during formation of a new myofiber7,8. Whether there are unique mechanisms to maintain ETC functionality in MuSCs is currently unclear.

We report here that MuSCs are sensitive to ETC dysfunction associated with oxidative stress. We propose a model in which in response to elevated superoxide levels, MuSCs undergo a unique fate to be rapidly removed from muscular tissue, which is mediated by direct fusion of quiescent stem cells with preexisting myofibers. In this manner, ETC-dysfunctional MuSCs are deleted from the stem cell population and are thereby precluded from regenerating dysfunctional tissue. We identify that the MuSC–myofiber fusion event is downstream of ROS generation and dependent on the induced expression of an actin-network reorganizing protein, Scinderin. In the absence of Scinderin-mediated fusion, ETC-dysfunctional MuSCs remain and are competent to regenerate dysfunctional tissue. Thus, MuSCs retain a unique mechanism governing their response to ETC deficiency.

Results

Complex IV inhibition depletes MuSCs by MuSC–myofiber fusion

We examined the consequences of ETC dysfunction using genetic models in which complex IV activity was conditionally depleted in adult murine MuSCs. Using the Pax7-CreERT2(FAN) driver (which induces tamoxifen-dependent recombination in satellite cells, a subset of MuSCs critical for muscle regeneration9,10,11) and a floxed Cox10 (an accessory subunit of mitochondrial complex IV) allele (Cox10f)12,13, we depleted Cox10 transcript and protein levels in adult MuSCs using a 5-day tamoxifen administration (Fig. 1a and Extended Data Fig. 1a). To assess whether ETC function was impaired in our experimental model, we implemented an in vitro fluorescence-activated cell sorting (FACS)–based assay amenable to rare cell populations (Fig. 1b). Acutely isolated MuSCs were permeabilized and incubated with or without mitochondrial substrates in the presence of TMRE (tetramethylrhodamine ethyl ester; whose fluorescence is a measure of mitochondrial membrane potential (ΔΨm)). In MuSCs from tamoxifen-treated control mice (Cox10f/f; no Pax7-Cre driver; hereafter Cox10f/f), mitochondrial substrates stimulated TMRE fluorescence, which was blocked by ETC inhibitors (e.g., rotenone) or uncouplers (carbonyl cyanide 3-chlorophenylhydrazone (CCCP)), indicating the presence of a functional ETC (Fig. 1c and Extended Data Fig. 1b). In contrast, Cox10−/− MuSCs (from tamoxifen-treated Pax7-CreERT2(FAN); Cox10f/f mice; hereafter Cox10−/−) exhibited impaired stimulation of mitochondrial membrane potential, indicating significant ETC dysfunction in this stem cell compartment (Fig. 1c and Extended Data Fig. 1b). In addition, Cox10−/− MuSCs exhibited significant increases in mitochondrial mass, consistent with ETC dysfunction (Extended Data Fig. 1c).

Fig. 1: Complex IV dysfunction induces rapid MuSC depletion in vivo.
figure 1

a, Schematic for the tamoxifen (TMX) administration protocol to induce recombination in adult MuSCs. b, FACS-based assay for ETC function. Isolated MuSCs are permeabilized with perfringolysin O (PFO), followed by TMRE staining in the presence or absence of mitochondrial substrates and inhibitors. SC, stem cell. c, Mean TMRE fluorescence of wild-type (Cox10f/f) and Cox10−/− MuSCs in response to indicated substrates and inhibitors. ns, no substrate; pyr/mal, pyruvate/malate; rot, rotenone. d, MuSC frequency at indicated time points after the first dose of tamoxifen in mice with Cox10f/f or Cox10−/− MuSCs. P values reflect comparisons with t = 0 day data for each genotype. e, Pax7+ cell numbers (normalized to muscle area) at different times after the first dose of tamoxifen. P values reflect comparisons with t = 0 day data for each genotype. f, Representative images of endogenous Pax7+ cells (red), Laminin (green) and DAPI (blue) from tibialis anterior (TA) cross sections of indicated mice at different times after tamoxifen administration. Scale bar, 50 μm. g, Histology (H&E staining) and immunofluorescence images of TA cross sections 5 days after barium chloride (BaCl2) injury. Regenerative myofibers can be identified by their centrally localized nuclei and MYH3-positive staining at this timepoint. Scale bars, 50 μm. h, Quantitation of regenerative (central nuclei) myofibers (normalized to muscle area) in TA muscles 5 days after BaCl2 injury. i, TA weight (normalized to body weight) at various time points after BaCl2 injury. Statistical significance was assessed using two-way analysis of variance (ANOVA) (c–e,i) or two-tailed t test (h) with adjustments for multiple comparisons. Box plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates in each group and P values are indicated in the figure. Experiments were repeated five times (f) and six times (g), with similar results. d, days; DAPI, 4,6-diamidino-2-phenylindole; dpi, days postinjury.

Full size image

Deletion of Cox10 by tamoxifen administration induced a progressive decrease in MuSC numbers, as assed by the frequency of CD34+CD31−CD45−Sca1− cells by flow cytometry or staining of endogenous Pax7 in tissue sections (Fig. 1d–f and Extended Data Fig. 1d). As expected, the dramatic loss of MuSCs was associated with a severe regeneration defect in response to muscle cryoinjury (Extended Data Fig. 1e–h) or chemically induced muscular injury (Fig. 1g–i), as evidenced by a lack of activated (myogenin-positive (Myog+)) MuSCs and regenerative (myosin heavy chain 3–positive (MYH3+)) myofibers and impaired recovery of muscle mass. These results were confirmed using a second independent Pax7-CreERT2(KARDON) allele11, which also resulted in the loss of Pax7+ MuSCs cells upon tamoxifen-induced deletion of Cox10 (Extended Data Fig. 1i,j), accompanied by regenerative defects in response to injury (Extended Data Fig. 1k,l).

Cellular stress in MuSCs has been previously associated with premature activation or differentiation;14 however, we did not observe indicators of myogenic activation in Cox10−/− MuSCs (Extended Data Fig. 1m). To more directly assess the fate of Cox10-deleted MuSCs, we attempted to synchronize Cre-mediated recombination with a single high dose of tamoxifen15, which was sufficient to rapidly induce near-complete recombination, as well as loss of Cox10 protein (Extended Data Fig. 2a–d). This alternate protocol increased the kinetics of MuSC depletion, as detected by the CD34+ population (via FACS) or the Pax7+ population (via immunodetection of endogenous Pax7) (Extended Data Fig. 2e–g). Assessment of tissues at different time points after tamoxifen treatment revealed that the rapid depletion of MuSCs in this protocol was not associated with stem cell activation, apoptosis or senescence, as revealed by examination of myogenic regulatory factors (myogenic differentiation 1 (Myod), Myog and MYH3), TUNEL or β-galactosidase staining (Extended Data Figs. 3a–d and 4a–c). Although mitochondrial ETC impairment can be associated with apoptotic cell death in vitro16,17, Cox10 deficiency has not been typically associated with acute cell death in vivo13,18,19,20,21, and correspondingly, we did not observe any indication of apoptosis in Cox10−/− MuSCs based on TUNEL staining and assessment of cleaved caspase-3 levels (Extended Data Fig. 4d,e).

To directly determine the fate of Cox10−/− MuSCs, we performed lineage tracing using a conditional (Cre-dependent) mito-Dendra2 expression allele (Df)22, which encodes a fluorescent protein (Dendra2) targeted to mitochondria. In control mice (Pax7-CreERT2(KARDON); Df/f; Cox10+/+; hereafter “wt”), the mito-Dendra2 signal mainly labeled mononuclear cells throughout skeletal muscle tissue (Fig. 2a), with occasional labeling of mitochondria within myofibers. Since MuSCs are known to occasionally fuse with neighboring myofibers in wild-type animals23,24,25, these MuSC–myofiber fusion events can be marked by segments or ‘domains’ of mito-Dendra2 signal within myofibers and typically represent fusion of a single MuSC25. Cox10 deletion in MuSCs (Cox10−/−: Pax7-CreERT2(KARDON); Df/f; Cox10f/f) dramatically increased the frequency of mito-Dendra2 domains within myofibers (Fig. 2a), corresponding to an elevated rate of MuSC fusion into myofibers. The contribution of MuSCs to neighboring myofibers accumulated with time after tamoxifen administration, occurred in multiple muscles throughout the body and was randomly distributed through muscle tissue, as revealed by deep imaging of cleared tissue (Fig. 2a,b and Supplementary Videos 1–3). Fiber-type analysis revealed that MuSCs readily fused to both slow- (type I) and fast-twitch (type II) fibers, with a statistical preference for slow-twitch fibers (Extended Data Fig. 5a–c). Single-fiber analysis indicated that mitochondrial content was not notably increased in myofiber segments where a fusion event had occurred (Extended Data Fig. 5d–f).

Fig. 2: Complex IV dysfunction induces MuSC–myofiber fusion in vivo and in vitro.
figure 2

a, Longitudinal imaging of mito-Dendra2 signal in various muscles 21 days after the first dose of tamoxifen. Scale bar, 200 μm. b, Mito-Dendra2 domain number (normalized to muscle area) in myofibers of TA and soleus muscles at different times after the first dose of tamoxifen. c, In vitro MuSC–myotube fusion assay. Dendra2+ (green) freshly isolated MuSCs of the indicated genotype were overlaid on DsRed+ (red) C2C12 myotubes and assessed for fusion with myosin-positive (magenta) myotubes after 4 days. Cells were stained with myosin (magenta) and DAPI (blue). Scale bar, 50 μm. d, Quantitated myotube fusion indices for MuSCs of the indicated genotype. e, mito-Dendra2 domain numbers in TA and soleus muscles 21 days after the first dose of tamoxifen in mice of the indicated genotype. Mymk, Myomaker. f, MuSC (CD34+) frequency of the indicated genotypes 21 days after the first dose of tamoxifen (same color scheme as d,e). Statistical significance was assessed using two-way ANOVA (b), two-tailed t test (d,e), Kruskal–Wallis (b), two-tailed Mann–Whitney (e) or one-way ANOVA (f) tests with adjustments for multiple comparisons. Box plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates in each group and P values are indicated in the figure. Experiments were repeated three times (a) and five or six times (c), with similar results.

Full size image

Quantitating the number of spatially distinct domains (Fig. 2b), or the number of fibers receiving a labeled MuSC (Extended Data Fig. 6a), revealed a time-dependent increase in MuSC–myofiber fusion following tamoxifen administration, which corresponds with the slower kinetics of MuSC depletion in this protocol (Fig. 1d,e). This result was replicated in the single high-dose tamoxifen protocol (Extended Data Fig. 2a), where the more rapid depletion of MuSCs (Extended Data Fig. 2e,f) was accompanied by a corresponding rapid appearance of mito-Dendra2 domains within myofibers (Extended Data Fig. 6b). Thus, in two tamoxifen administration protocols with distinct kinetics, the appearance of mito-Dendra2 domains correlated in time with the depletion of MuSCs. In addition, experiments with an alternative reporter allele (tdTomatof/f) replicated the induction of MuSC–myofiber fusion upon ETC dysfunction (Extended Data Fig. 6c,d).

As Cox10 deletion was restricted to MuSCs in the above experiments, the induction of MuSC–myofiber fusion was likely due to cell-autonomous properties of these mutant MuSCs. Indeed, freshly isolated Cox10−/− MuSCs fused at increased frequencies with preformed C2C12 myotubes in vitro (Fig. 2c,d), despite defects in in vitro differentiation (Extended Data Fig. 6e). In addition, conditional genetic removal of the fusion machinery (myomaker; Mymk)26 was sufficient to block fusion of Cox10−/− MuSCs in vivo (Fig. 2e and Extended Data Fig. 6f,g) and rescue the loss of MuSCs (Fig. 2f and Extended Data Fig. 6h), despite continued loss of ETC function (Extended Data Fig. 6i,j). Together, these results indicate that loss of Cox10 in adult Pax7+ MuSCs triggers their depletion by inducing MuSC fusion into existing myofibers.

MuSC–myofiber fusion is induced by ROS

Although subsets of quiescent MuSCs are known to occasionally fuse with myofibers in wild-type adult animals27,28,29, the underlying mechanisms that induce these events are unknown. Our results thus far position ETC dysfunction as a critical trigger of MuSC–myofiber fusion. To test whether MuSC–myofiber fusion is a general consequence of ETC dysfunction, we next examined MuSC fate after loss of complex I function (Ndufa9−/−), loss of complex III function (Qpc−/−) or loss of mtDNA (complex I, III, IV and V; Tfam−/−), using conditional alleles of these genes crossed with the Pax7-CreERT2(Kardon) driver and Df lineage tracer. Tamoxifen-induced recombination of these alleles was sufficient to deplete target proteins and reduce ETC function in adult MuSCs (Extended Data Fig. 7a,b). Like Cox10−/− MuSCs, Ndufa9−/− MuSCs also acutely fused with existing myofibers, as indicated by the presence of Dendra2 domains within myofibers after tamoxifen administration (Fig. 3a,b). However, Qpc−/− or Tfam−/− MuSCs did not fuse with myofibers (Fig. 3a,b), indicating that not all modes of ETC dysfunction trigger MuSC–myofiber fusion. Correspondingly, deletion of Ndufa9 (but not Qpc or Tfam) also resulted in acute depletion of MuSCs, similar to Cox10 deletion (Extended Data Fig. 7c).

Fig. 3: MuSC–myofiber fusion is regulated by ROS.
figure 3

a, Representative longitudinal images of mito-Dendra2 domains in TA and soleus muscles of mice of the indicated genotype (21 days after the first dose of tamoxifen). Scale bar, 200 μm. b, Quantitation of mito-Dendra2 domain numbers (normalized to muscle area) 21 days after the first dose of tamoxifen. P values reflect comparisons with the wt group. c, Relative superoxide levels in MuSCs of the indicated genotype 5 days after the first dose of tamoxifen. Same color scheme as b. P values reflect comparisons with the wt group. d, Mito-Dendra2 domain numbers (normalized to muscle area) in mice of the indicated genotype pretreated with vehicle (PBS) or NAC 5 days after the first dose of tamoxifen. e, Representative longitudinal images of mito-Dendra2 domains in TA and soleus muscles of mice treated with vehicle (PBS) or BSO for 21 days. Scale bar, 200 μm. f, Mito-Dendra2 domain numbers (normalized to muscle area) in mice treated with PBS or BSO for 21 days. g, Representative longitudinal images of mito-Dendra2 domains in TA and soleus muscles of sedentary (SED) mice, exercised mice treated with PBS (EXER+PBS) or exercised mice treated with NAC (EXER+NAC). Scale bar, 200 μm. h, Mito-Dendra2 domain numbers (normalized to muscle area) in mice of the indicated groups. Statistical significance was assessed using one-way ANOVA (b,c,h), two-way ANOVA (d), two-tailed t test (f) and two-tailed Mann–Whitney (f) tests with adjustments for multiple comparisons. Box plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates in each group and P values are indicated. Experiments were repeated five or six times (a), five times (e) and 10–11 times (g), with similar results.

Full size image

Different mechanisms of ETC dysfunction are associated with variable inductions in oxidative stress due to a differential ability to impact the redox-active prosthetic groups required for superoxide generation30. We observed elevated levels of cellular superoxide, as well as elevated mitochondrial superoxide and total ROS, in the Ndufa9−/− and Cox10−/− MuSCs, but not Qpc−/− or Tfam−/− MuSCs (Fig. 3c and Extended Data Fig. 7d). The lack of increased ROS in Qpc−/− and Tfam−/− MuSCs is consistent with previous experiments in other cellular systems31,32. The correlation between increased ROS levels (Fig. 3c) and induced MuSC–myofiber fusion (Fig. 3a,b) suggests that elevated ROS may be a critical signal that initiates the fusion event. To test this idea, we treated mice with the antioxidant N-acetylcysteine (NAC) before and during tamoxifen-induced conditional deletion of Cox10 or Ndufa9. NAC pretreatment lowered MuSC superoxide and mitoROS levels (Extended Data Fig. 7e) and inhibited MuSC–myofiber fusion in both genotypes (Fig. 3d), indicating that mitoROS elevation is associated with the induction of MuSC–myofiber fusion. Treatments with an alternative antioxidant (Trolox), as well as the mitochondrial targeted antioxidants MitoTempo and mitoquinone (MitoQ), were also sufficient to lower mitoROS levels in Cox10−/− and Ndufa9−/− MuSCs and block MuSC–myofiber fusion (Extended Data Fig. 7f,g). Thus, antioxidant treatment blocks the fusion of ETC-dysfunctional MuSCs with neighboring myofibers. Conversely, we induced elevated ROS in MuSCs of wild-type animals by systemic treatment with the pro-oxidant glutathione synthesis inhibitor (buthionine sulfoximine (BSO)) (Extended Data Fig. 7h). Daily BSO treatment (for 21 days) was sufficient to stimulate MuSC fusion to neighboring myofibers (Fig. 3e,f) and deplete MuSC levels (Extended Data Fig. 7i). Thus, induced pharmacologic elevations in systemic ROS correlate with enhanced MuSC–myofiber fusion.

Although the above genetic manipulations and pharmacologic (BSO) treatment resulted in substantial ETC dysfunction and/or oxidative stress, the physiological stresses experienced by MuSCs in wild-type conditions are likely less severe. We examined the effects of physiologic elevations in ROS, making use of a daily treadmill exercise regimen. Exercise regimens are known to be associated with transient oxidative stress in skeletal muscle tissue33,34,35,36,37, although other pleiotropic effects are also present. Four weeks of daily treadmill running was sufficient to induce MuSC–myofiber fusion in hindlimb muscles of wild-type mice (Fig. 3g,h). Interestingly, this effect was blocked by daily treatment with NAC or other antioxidants (Fig. 3g,h and Extended Data Fig. 7j,k), indicating that the induced fusion was ROS dependent. Thus, physiologic elevations in ROS experienced by wild-type animals is associated with induction of MuSC–myofiber fusion.

Induction of Scinderin is required for MuSC–myofiber fusion

To identify mechanisms underlying the fusion of MuSCs with myofibers, we performed RNA sequencing (RNA-seq) analysis comparing wild-type and Cox10−/− MuSCs. We did not observe upregulation of Pax3 or other myogenic regulatory factors associated with MuSC activation in Cox10−/− MuSCs (Supplementary Table 1). Gene ontology analysis indicated that upregulated transcripts in Cox10−/− MuSCs were enriched for actin network binding and regulatory proteins (Supplementary Table 1 and Fig. 4a,b). Actin-dependent protrusions of the plasma membrane have been previously implicated in myoblast fusion38,39, and analysis of the highest upregulated genes identified Scinderin (Scin), a member of the gelsolin family with actin network reorganization activity40. Scinderin mRNA and protein was upregulated in both Cox10−/− and Ndufa9−/− MuSCs, but not Qpc−/− or Tfam−/− MuSCs (Fig. 4c,d). Antioxidant treatment was sufficient to block Scinderin induction in Cox10−/− and Ndufa9−/− MuSCs (Fig. 4e).

Fig. 4: Scinderin induces actin network changes and MuSC–myotube fusion.
figure 4

a, Volcano plot of gene expression changes in Cox10−/− versus wild-type MuSCs. Log2(fold change) is plotted against the log10(adjusted P value) for each gene. b, Gene ontology analysis of upregulated genes in Cox10−/− MuSCs. c, Normalized Scinderin transcript levels in isolated MuSCs of the indicated genotype 5 days after the first dose of tamoxifen. P values reflect comparisons with the wt group. d, Scinderin protein levels in isolated MuSCs of the indicated genotype 5 days after the first dose of tamoxifen. Histone 2B (H2B) levels are shown as a loading control. e, Scinderin protein levels in isolated MuSCs of the indicated genotype and treatment 5 days after the first dose of tamoxifen. H2B levels are shown as a loading control. f, Scinderin protein levels in C2C12 cells infected with empty or Scinderin-overexpressing virus. H2B levels are shown as a loading control. g, Violin plots of actin protrusion numbers (per cell) for control and Scinderin-overexpressing C2C12 cells. h, Representative images of actin networks (stained by Acti-stain555 phalloidin, red) and nuclei (DAPI, blue) in control and Scinderin-overexpressing C2C12 cells. Scale bar, 10 μm. i, Scinderin (α-HA− green) and actin (phalloidin; red) localization in cells overexpressing Scinderin. Scale bar, 10 μm. j, Representative images from an in vitro MuSC–myotube fusion assay; 10,000 Dendra2+ (green) freshly isolated control or Scin-overexpressing MuSCs were overlaid on DsRed+ (red) C2C12 myotubes and assessed for fusion with Myosin+ myotubes after 4 days. Cells were stained with myosin (magenta) and DAPI (blue). Scale bar, 50 μm. k, Quantitated myotube fusion indices for MuSCs of the indicated genotype. Statistical significance was assessed using one-way ANOVA (c), two-tailed Mann–Whitney test (g) or two-tailed t test (k). Box and violin plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates per group and P values are indicated in the figure. Experiments were repeated three times (h,i) and four times (j), with similar results.

Full size image

In vitro, overexpression of Scinderin stimulated actin-rich protrusions of the plasma membrane in C2C12 myoblasts (Fig. 4f–h). Scinderin protein accumulated at sites of actin protrusion and colocalized with actin accumulation (Fig. 4i), suggesting a model whereby Scinderin induces actin cap formation at the plasma membrane to induce cellular protrusions and invasion into neighboring myofibers. We therefore infected primary wild-type MuSCs with Scin-expressing (or control) retrovirus and then overlaid them with preformed C2C12 myotubes. Although control MuSCs rarely fused with C2C12 myotubes, Scin-expressing cells often fused with preformed myotubes (Fig. 4j,k). The data implicate a model whereby Scinderin induces fusion of primary MuSCs via local activation of actin polymerization at the plasma membrane and subsequent membrane protrusion.

We next used a conditional knockout allele of Scinderin41 to examine its in vivo role in MuSC fate and function. Deletion of Scinderin in Pax7+ MuSCs did not impair myofiber regeneration in response to injury (Extended Data Fig. 8a–c) or impair in vitro differentiation into myotubes (Extended Data Fig. 8d), indicating that Scinderin is not required for MuSC–MuSC fusion events necessary for new fiber formation. However, quiescent Cox10−/−Scin−/− MuSCs no longer fused with existing myofibers at elevated rates in vivo (Fig. 5a,b), despite continued ETC dysfunction and elevated superoxide (Extended Data Fig. 8e–g), indicating that Scinderin is required for MuSC–myofiber fusion events induced by ETC dysfunction. Correspondingly, removal of Scinderin rescued the depletion of MuSCs mediated by loss of Cox10 (Fig. 5c,d). In addition, Scinderin was required for MuSC–myofiber fusion induced by systemic elevations in oxidative stress (BSO treatment or treadmill running) in wild-type animals (Extended Data Fig. 8h,i). Together, these data indicate a requirement for Scinderin in MuSC–myofiber fusion in response to multiple forms of oxidative stress.

MuSC–myofiber fusion prevents regeneration of damaged tissue

As described above (Figs. 1,2), Cox10 deletion in MuSCs induces stem cell depletion, resulting in impaired myofiber regeneration. We therefore retested regenerative capacity in mice with Cox10−/−Scin−/− mice, which retain MuSCs (Fig. 5c,d). In contrast to Cox10−/− animals, codeletion of Scinderin was sufficient to rescue regeneration from Cox10−/− MuSCs, resulting in the formation of de novo MYH3+ myofibers (Fig. 5e,f), as well as recovery of muscle mass at 3 weeks after injury (Fig. 5h). As expected, deletion of Myomaker did not rescue regeneration from Cox10−/− MuSCs, as Myomaker is required for MuSC–MuSC fusion26,42 (Fig. 5e,f). Thus, the induction of MuSC–myofiber fusion and resulting depletion of MuSCs is responsible for the decline in regenerative capacity observed in mice with Cox10−/− MuSCs. However, de novo fibers formed from Cox10−/−Scin−/− MuSCs exhibited significant mitochondrial impairment, as indicated by their depleted mitochondrial complex IV activity (Fig. 5e,g). Consistent with this, functional analysis performed 21 days after injury revealed significant defects in twitch and tetanic force generation (Extended Data Fig. 8j–m). Thus, in the absence of Scinderin-mediated MuSC–myofiber fusion, ETC-dysfunctional stem cells are retained in this experimental model and available to regenerate dysfunctional tissue.

Fig. 5: Scinderin is required for MuSC–myofiber fusion.
figure 5

a, Representative longitudinal images of mito-Dendra2 domains in TA and soleus muscles of mice of the indicated genotype 21 days after the first dose of tamoxifen. Scale bar, 200 μm. b, Mito-Dendra2 domain numbers (normalized to muscle area) 21 days after the first dose of tamoxifen. c, Representative images of endogenous Pax7+ cells (red), laminin (green) and DAPI (blue) from TA cross sections from mice of the indicated genotype 21 days after the first dose of tamoxifen. Scale bar, 50 μm. d, MuSC (CD34+) and Pax7+ frequency in mice of the indicated genotype 21 days after the first dose of tamoxifen. e, Representative images of histology (H&E), MYH3+ fibers (red) and complex IV activity (COX) in cross sections of TA muscles from mice of the indicated genotypes. Muscles were removed and assessed 5 or 21 days after BaCl2 injury. For immunofluorescence staining, sections were stained with α-laminin (green), α-MYH3 (red) and DAPI (blue). Scale bar, 50 μm. f, Quantitation of regenerative fiber numbers (normalized to muscle area) in TA muscles 5 and 21 days after BaCl2 injury. g, Normalized COX activity from regenerative myofibers 5 and 21 days after BaCl2 injury. The number of analyzed fibers per group is indicated. h, TA weight (normalized to body weight) 21 days after BaCl2 injury. Statistical significance was assessed using one-way ANOVA (b,d,f,h) or two-tailed Mann–Whitney (g) tests, with adjustments for multiple comparisons. Box and violin plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates in each group and P values are indicated. Experiments were repeated five times (a), six times (c) and five to eight times (e), with similar results.

Full size image

Scinderin regulates MuSC functionality during aging

During aging, MuSCs are known to rapidly deplete, which is accompanied by elevations in ROS and metabolic signatures of ETC dysfunction15,43. Consistent with this literature, we assessed MuSC properties in a cohort of wild-type mice from 2 to 27 months of age, which revealed significant declines in MuSC numbers in early adulthood, as well as increased mitochondrial ROS in late adulthood (Fig. 6a–d). As mtDNA mutations have been observed to accumulate with age in various tissues, we purified MuSCs from wild-type mice of varying ages and performed deep sequencing of mtDNA genomes at high coverage (Extended Data Fig. 9a,b and Supplementary Table 2). Deep sequencing of MuSC mtDNA genomes revealed an age-dependent increase in the abundance of mtDNA mutations, including both substitutions and indels (insertions/deletions) (Fig. 6e,f and Extended Data Fig. 9c). Age-accumulating mutations occurred randomly throughout the mitochondrial genome, were predominantly of low allelic frequency and displayed mutational signatures consistent with mtDNA replicative errors44,45 (Fig. 6g and Extended Data Fig. 9d,e). Thus, the murine MuSC population accumulates random mtDNA errors in late adulthood, which is correlated with the overall observed increases in superoxide levels. However, our pooled sequencing analysis cannot distinguish allelic frequencies within individual cells, and thus, we cannot directly associate the presence of mtDNA mutations with elevated superoxide levels at the single-cell level. We noted that the observed frequencies of protein-coding indels, as well as nonsynonymous substitutions, were significantly lower than expected by random chance (Fig. 6h,i and Extended Data Fig. 9f), indicating that MuSCs select against dysfunctional mtDNA mutations during aging.

Fig. 6: MuSCs accumulate mtDNA mutations with age.
figure 6

a, Quantitation of MuSC number (frequency of CD34+ cells) from wild-type mice of the indicated ages. b, Representative superoxide FACS profiles of MuSCs from mice of the indicated age. NC, unstained negative control. Gating for the SOXHI (SuperoxideHI) population is indicated. c, Quantitation of mean superoxide levels in MuSCs of mice of the indicated age. d, Quantitation of SOXHI population frequency from MuSCs of mice of the indicated age. e, Number of identified mtDNA substitutions from deep mtDNA sequencing of isolated MuSCs from mice of the indicated age. f, Number of identified mtDNA indels from isolated MuSCs of mice of the indicated age. g, Mapped positions of identified somatic mtDNA mutations (indels, noncoding, synonymous and nonsynonymous substitutions) in MuSCs from 27-month-old mice. h, Abundance of mtDNA protein-coding indels (as a percentage of total indels) observed in isolated MuSCs from mice of the indicated age. The expected frequency is indicated by the red line. i, Abundance of mtDNA nonsynonymous substitutions (as a percentage of total protein-coding substitutions) observed in isolated MuSCs from mice of the indicated age. The expected frequency is indicated by the red line. Statistical significance was assessed using one-way ANOVA (a,c), Kruskal–Wallis (d–f) and chi-squared (h,i) tests with adjustments for multiple comparisons. Box plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates per group and P values are indicated.

Full size image

We therefore tested the role of MuSC–myofiber fusion in regulating MuSC properties during aging, making use of mice with conditional removal of Scin or Mymk in MuSCs starting at 6 weeks of age (Fig. 7a). As compared with wild-type animals, animals with Scin−/− MuSCs or Mymk−/− MuSCs exhibit increased MuSC numbers at multiple ages (12–30 months) (Fig. 7b and Extended Data Fig. 10a), indicating that the ongoing MuSC–myofiber fusion in wild-type animals significantly contributes to the depletion of MuSCs during physiologic aging. Assessment of mitochondrial ROS in 12-, 24- and 30-month-old MuSCs revealed that loss of Scinderin or Myomaker results in a significant accumulation of stem cells with elevated superoxide levels (Fig. 7c,d). Thus, in the absence of MuSC–myofiber fusion, MuSCs deplete more slowly with age and accumulate stem cells with elevated mitochondrial ROS.

Fig. 7: Scinderin is required for regenerating healthy tissue in aged animals.
figure 7

a, Schematic of aging experiments in wild-type, Mymk−/− and Scin−/− mice. Recombination is induced in MuSCs of animals at 6 weeks of age via tamoxifen (TMX) administration, followed by aging and the indicated analysis at various time points (2, 12, 24 and 30 months of age). At each time point, MuSCs were analyzed in situ by FACS; in addition, TA muscles were injured by BaCl2 and assessed for regeneration 21 days after injury. b, Pax7+ cell numbers (normalized to muscle area) in animals of the indicated genotype and age. P values represent comparison with wild type for each age group. c, Representative FACS profiles of superoxide (SOX) levels in MuSCs isolated from animals of the indicated genotype and age. Gating for SOXHI population is indicated. d, Quantitation of SOXHI population frequency from MuSCs of mice of the indicated age and genotype. P values represent comparison with wild type for each age group. e, Representative images of histology (H&E) in cross sections of TA muscles from mice of the indicated genotypes and ages. Muscles were removed and assessed 21 days after BaCl2 injury. Scale bar, 50 μm. f, Quantitation of fiber size from regenerative fibers (21 days after injury) in TA muscles of mice of the indicated genotypes and ages. g, Representative images of COX activity in cross sections of TA muscles from mice of the indicated genotypes and ages 21 days after BaCl2 injury. h, Quantitation of relative COX activity in regenerative fibers (21 days after injury) from TA muscles of mice of the indicated genotypes and ages. i, Representative images of SDH activity in cross sections of TA muscles from mice of the indicated genotypes and ages 21 days after BaCl2 injury. j, Quantitation of relative SDH activity in regenerative fibers (21 days after injury) from TA muscles of mice of the indicated genotype and ages. Statistical significance was assessed using two-way ANOVA (b,d) or two-tailed Mann–Whitney (f,h,j) tests, with adjustments for multiple comparisons. Box and violin plots indicate median values and interquartile ranges; whiskers are plotted using the Tukey method. The number of biological replicates in each group and P values are indicated. Experiments were repeated five to eight times (e,g,i), with similar results.

Full size image

Retention of dysfunctional MuSCs during aging potentially has consequences for the regeneration of healthy tissue. We inhibited MuSC–myofiber fusion via conditional deletion of Scinderin in MuSCs at 6 weeks of age and tested regeneration in young (2 months), middle-aged (12 months), older (24 months) and geriatric (30 months) animals. Loss of Scinderin had no functional consequences on regeneration in 2-month-old animals (Fig. 7e,f and Extended Data Fig. 8a-c). At 12 months of age, animals with Scin−/− MuSCs displayed intact regenerative capacity (Fig. 7e,f and Extended Data Fig. 10b), based on their ability to form new fibers. However, the regenerated tissue exhibited substantial defects, including reduced fiber size (Fig. 7f) and reduced complex IV activity, evident of mitochondrial dysfunction (Fig. 7g,h). Concomitant with this, regenerative fibers displayed significant increases in SDH activity (Fig. 7i,j), indicative of mitochondrial proliferation and accumulation commonly observed in the settings of muscular ETC dysfunction. These defects were retained in older (24-month-old) and geriatric (30-month-old) animals but were not observed in young (2-month-old) animals (Fig. 7e-j). Thus, deletion of Scinderin in MuSCs results in an age-dependent regenerative defect, likely due to the retention of MuSCs with elevated mitoROS that are unable to regenerate healthy tissue. In particular, the accumulating mitochondrial ETC dysfunction in aged MuSCs can be propagated to new myofibers if these cells are not removed by Scinderin-mediated MuSC–myofiber fusion.

Discussion

Our results indicate the consequences of mitochondrial damage in MuSCs. ETC dysfunction associated with elevated ROS triggers compromised MuSCs to directly fuse with existing myofibers. This mechanism is cell autonomous, as ETC-dysfunctional MuSCs are able to fuse to otherwise healthy myofibers; however, we cannot rule out potential contributions from signaling events within the receiving myofiber or the extracellular matrix, and this constitutes a limitation of our study. This self-removal of ETC-dysfunctional stem cells by MuSC–myofiber fusion contributes to the striking decline in MuSC numbers observed during aging (Fig. 8a,b) and limits the appearance of ETC-dysfunctional MuSCs in aged animals. Thus, our findings provide insight into mechanisms to attenuate loss and regulate MuSC health in aged individuals.

Fig. 8: Proposed model: MuSC–myofiber fusion regulates removal of ETC-dysfunctional MuSCs during aging.
figure 8

a, In young wild-type animals, healthy MuSCs predominate and are competent to regenerate healthy myofibers in response to injury. b, In aged, wild-type animals, ETC-dysfunctional MuSCs (depicted with red mitochondria) are sensitive to elevated ROS and Scinderin levels and thereby induced to fuse into neighboring myofibers. In this manner, the remaining MuSC population contains healthy mitochondria and generates healthy myofibers in response to injury. c, In the absence of Scinderin during aging, MuSC–myofiber fusion is not available to remove ETC-dysfunctional MuSCs. In this setting, injury can trigger the activation, proliferation and fusion of damaged MuSCs, resulting in de novo myofibers with mitochondrial dysfunction.

Full size image

MuSC–myofiber fusion has been previously observed in wild-type mice. During early postnatal growth (up to postnatal day 21), MuSCs directly contribute to existing myofibers, resulting in an approximately fivefold increase in myonuclear numbers and an approximately sevenfold increase in cross-sectional area46. After this initial period, the contribution of MuSCs to myofibers is slow in sedentary animals but can be readily observed with lineage tracing studies23,25. In our experimental models of induced MuSC–myofiber fusion in adult animals, we did not observe changes in cross-sectional area or myonuclear number, consistent with the low number of MuSCs (~2–3% of total nuclei) present in adult tissue. Our results instead suggest that MuSC–myofiber fusion in adult muscle is typically reserved for stem cells accumulating high levels of oxidative stress. Our data do not indicate evidence of premature activation or differentiation in response to oxidative stress; however, it is possible that myogenic activation occurs very close in time to MuSC–myofiber fusion such that the induction of myogenic regulatory factors is difficult to detect, and this represents a limitation of our study.

We observe that although aged MuSCs accumulate mtDNA mutations, these cells display signatures of selection against dysfunctional mtDNA variants. Thus, in the absence of MuSC–myofiber fusion, ETC-dysfunctional MuSCs are retained and available to regenerate de novo tissue embodied by mitochondrial dysfunction (Fig. 8c). Thus, MuSC–myofiber fusion appears to be largely aimed towards preservation of a functional stem cell population, primed to regenerate healthy tissue, despite the decline in overall stem cell numbers. We speculate that the loss of stem cells does not substantially impact an organism’s fitness until it is past reproductive age and suggest that this mechanism is optimized to promote healthy regeneration in response to injury in younger animals. However, future work will be needed to assess the consequences of contribution of ETC-damaged MuSCs into existing myofibers over long periods of time, which may directly impact the aging process.

In addition, we observe that exercise (running) is sufficient to induce MuSC–myofiber fusion in wild-type mice, and this process can be regulated by systemic antioxidant treatment. ROS (induced by exercise) regulate the activation of a number of signaling pathways associated with improved insulin sensitivity and aerobic capacity, as well as muscle hypertrophy47. Indeed, a number of studies have suggested that antioxidant treatment can sometimes inhibit the beneficial effects of exercise training48,49,50,51,52,53. Our findings suggest that induction of MuSC–myofiber fusion may constitute part of the redox-dependent adaptation to exercise training, and future work will address the role of this phenomenon in contributing to muscle hypertrophy and activation of exercise-associated signaling pathways.

Our findings add to the accumulating literature on the relevance of mitochondrial dysfunction in stem cells. Maintenance of quiescence is well known to prevent stem cell exhaustion in numerous tissue systems, and quiescent stem cells typically require protective mechanisms to protect them from damage as they age. In response to nuclear DNA damage, stem cells are dependent on DNA repair pathways to maintain functionality54,55. In contrast, mtDNA damage is not associated with a robust repair mechanism. We found that MuSCs retain a unique mechanism to cope with ETC deficiency, and it will be intriguing to investigate whether other stem cell compartments respond to mtDNA damage in a distinctive manner and how this mechanism interplays with other quality-control organelle mechanisms (e.g. mitophagy). A recent report from in vitro cultured mammary stem cells suggests that young mitochondria are preferentially sequestered into daughter cells to promote stemness, whereas retention of old mitochondria promotes differentiation56, and our findings add to the general model that damaged mitochondria hamper maintenance of a quiescent stem cell population.

The identification of Scinderin as a specific regulator of MuSC–myofiber fusion allows the dissection of MuSC–myofiber fusion separately from MuSC–MuSC fusion events (associated with injury-induced regeneration). Indeed, our results suggest that MuSC–myofiber fusion requires the low levels of Myomaker found in quiescent MuSCs, although it is possible that Myomaker expression is induced immediately before fusion. Our data thereby indicate that there are mechanistic differences between these two modes of cell fusion, and it will be important in the future to evaluate the biophysical differences in membrane dynamics therein. More generally, the observation that dysfunctional stem cells are induced to fuse with existing tissue suggests opportunities to target age-associated pathology via regulation of stem cell–tissue fusion events.

Methods

Mice

Throughout this study, all genotypes refer to animals with conditional (floxed (f)) alleles, targeted to the MuSC population. Cox10f/f (strain 024697), Pax7-CreERT2(FAN) (strain 012476), Pax7-CreERT2(KARDON) (strain 017763), C57BL/6 (strain 000664), tdTomatof/f (strain 007914) and mito-Dendra2f/f (strain 018385) mice were purchased from The Jackson Laboratory. For lineage tracing experiments, we exclusively used the Pax7-CreERT2(Kardon) allele, as the Pax7-CreERT2(Fan) allele displayed leaky recombination in the absence of tamoxifen. The generation of Qpcf/f, Tfamf/f, Mymkf/f and Scinf/f conditional knockout mice was described previously31,41,42,57.

The Ndufa9f/f conditional knockout mouse was generated at the Children’s Research Institute Mouse Genome Engineering Core using the Easi-CRISPR workflow58. sgRNAs surrounding exon 4 of Ndufa9 were selected by cross-referencing CHOPCHOP, the MIT CRISPR Design Website, and the CRISPR-Cas9 guide checker tool (Integrated DNA Technologies) (Supplementary Fig. 1a). CRISPR-Cas9 crRNAs and the Megamer single-stranded DNA fragment homology-directed repair template were purchased from Integrated DNA Technologies. Animals were screened for insertion of the conditional allele by PCR, and correct targeting was confirmed by Sanger sequencing. Genotyping primers for Ndufa9f/f mice are provided in Supplementary Table 3.

All mice were maintained on C57BL6 backgrounds, except Ndufa9f/f mice, which were on a mixed background. Both male and female mice were used in all experiments; sex-specific differences were not present, and male and female mice were analyzed together. Six- to 8-week-old mice were used for all experiments, except in Figs. 6 and 7, where ages are indicated. All mice were housed in the Animal Resource Center at the University of Texas Southwestern Medical Center under a 12-h light/dark cycle and fed ad libitum. All animal protocols were approved by the University of Texas Southwestern Institutional Animal Care and Use Committee (protocol 101323), and all relevant guidelines were adhered to while carrying out this study.

Generation and characterization of Ndufa9 f/f mouse embryonic fibroblasts (MEFs)

Embryonic day 13.5 embryos were collected from Ndufa9f/f or wild-type pregnant females after carbon dioxide asphyxiation and cervical dislocation. Embryos were dissected to remove the head and red organs, minced with a sterile scalpel and digested with trypsin for 45 min at 37 °C. Digested embryos were subsequently collected in DMEM (Sigma-Aldrich, D6429) with 10% FBS, 2 mM L-glutamine and 1% penicillin/streptomycin and plated to tissue culture flasks that were pretreated with gelatin (10 ml 0.1% gelatin/H2O solution per 75-cm2 surface area for 1 h at 37 °C). Ndufa9+/+ or Ndufa9f/f MEFs were immortalized by transduction with lentiCRISPRv2 expressing an sgRNA targeting Trp53 at a multiplicity of infection of ~0.5 (visual estimation) and subsequently selected with 5 μg ml−1 puromycin for 4 days. Approximately 5 × 106 immortalized MEFs were incubated in 1 ml media with 25 μl Ad5-CMV-EGFP or Ad5-CMV-CreEGFP adenovirus (University of Iowa Viral Vector Core; VVC-U of Iowa-4 and VVC-U of Iowa-1174, respectively) for 10 min at room temperature before being plated to tissue culture dishes. After transduction with EGFP or CreEGFP adenovirus, MEFs were continuously cultured in media containing 1 mM sodium pyruvate and 100 μg ml−1 uridine; 48 h later, transduced MEFs were sorted at the UT Southwestern Children’s Research Institute Moody Foundation Flow Cytometry Facility, and GFP+ cells were selected. Transduced cells were verified for genotype by PCR analysis (Supplementary Fig. 1b) and western blot (Supplementary Fig. 1c).

An Agilent Seahorse XFe96 Analyzer was used for cellular oxygen consumption measurements of Ndufa9+/+ and Ndufa9f/f MEFs (Supplementary Fig. 1d). MEFs were plated at 10,000 cells per well in 80 μl media and allowed to adhere overnight. The following day, cells were washed twice with 200 μl per well assay medium (DMEM (Sigma-Aldrich, D5030) with 10 mM glucose, 2 mM L-glutamine, 1 mM sodium pyruvate and 1% penicillin/streptomycin), and 150 μl assay medium was added to each well after the second wash. Cells were transferred to a 37 °C, CO2-free incubator for 1 h. Standard calibration and baseline oxygen consumption measurements were performed using a 3-min ‘mix’/3-min ‘measure’ cycle with three measurements recorded at baseline and after injection of each compound. The following inhibitors were used: 2 μM oligomycin, 3 μM CCCP and 3 μM antimycin A. Data collection was performed with WAVE (v.2.4.1.1) software.

MuSCs isolation via FACS

Murine MuSC isolation was adapted from a published protocol59. Following carbon dioxide asphyxiation and cervical dislocation, skeletal muscle was rapidly dissected and sequential digested with Collagenase II (1 h) and Dispase (30 min) at 37 °C. Mononucleated cells were collected through a 70-μm cell strainer and suspended in HBSS with 2% horse serum (Gibco, 16050114). Cells were then incubated with the following antibodies on a rotator at 4 °C for 30 min: APC-conjugated anti-mouse CD31 (BioLegend, clone MEC13.3, 102510, 1:100), APC-conjugated anti-mouse CD45 (BioLegend, clone 30-F11, 103112; 1:100), PerCP-Cy5.5–conjugated anti-mouse Sca-1 (Invitrogen, clone D7, 45598182, 1:100) and biotin-conjugated anti-mouse CD34 (Invitrogen, clone RAM34, 13034181, 1:100). After incubation, cells were washed twice and then incubated with phycoerythrin (PE)/Cy7–conjugated streptavidin (BioLegend, 1:100, 405206) on a rotator at 4 °C for 20 min. Cells were washed twice and then suspended with 2% horse serum in HBSS with DAPI. Quiescent MuSCs (Extended Data Fig. 9a) were identified as the CD34+, CD31−, CD45−, DAPI− and Sca-1−. For isolation of mito-Dendra2+ cells, mononucleated cells were resuspended with 2% horse serum in HBSS with DAPI, and GFP+, DAPI− cells were isolated. Purity was confirmed by immunofluorescence staining for Pax7. All sorting was performed at the Moody Foundation Flow Cytometry Facility on a FACS Aria flow cytometer (BD Biosciences). Data were collected by FACSDiva (v.8.0.2, BD Biosciences) software and analyzed using FlowJo (v10.6.1) software.

ROS and mitochondrial membrane potential (ΔΨm) analysis

For ROS analysis, MuSCs were resuspended in HBSS with 5 μM final concentration Superoxide Detection reagent (total superoxide levels; Enzo Life Sciences, 51010), 5 μM final concentration mitoSOX (mitochondrial superoxide levels, Thermo Fisher, M36008), or 1 μM final concentration mitoROS (mitochondrial total ROS levels; Cayman Chemical, 701600) and then incubated at 37 °C for 30 min. After incubation, cells were washed and resuspended in HBSS with DAPI, followed by immediate FACS analysis.

For mitochondrial membrane potential analysis, MuSCs were isolated at 5 days after the first dose of tamoxifen and resuspended in mitochondrial assay buffer with the indicated mitochondrial substrates and inhibitors (see below for recipes). Cell suspensions were incubated at 37 °C for 30 min. After incubation, cells were washed and then suspended in HBSS with DAPI.

The mitochondrial assay buffer contained 220 mM mannitol, 70 mM sucrose, 10 mM KH2PO4, 5 mM MgCl2, 2 mM HEPES and 1 mM EGTA, pH 7.4, supplemented with fresh TMRE (Invitrogen, T669; final concentration, 150 nM) and fresh PMP reagent (Agilent, 102504-100; final concentration, 3 nM).

Pyruvate/malate buffer contained mitochondrial assay buffer supplemented with 10 mM pyruvate and 5 mM malate, pH 7.4.

Inhibitor concentrations were 5 μM for CCCP and 5 μM for rotenone.

RNA isolation and sequencing

Total RNA was purified from FACS-isolated MuSCs using the RNeasy Micro Kit (Qiagen, 74004) according to manufacturer’s instructions. Library preparation was performed using the SMARTer stranded pico input total RNA-seq kit (Takara Bio, 634411) following the manufacturer’s instructions. Next-generation sequencing was performed using an Illumina NextSeq 500 by the Children’s Research Institute’s Sequencing Facility at UT Southwestern Medical Center. RNA-seq analysis was performed on BICF RNASeq Analysis Workflow (ssh://git@git.biohpc.swmed.edu/BICF/Astrocyte/rnaseq.git) provided by the UTSouthwestern Bioinformatics Core Facility. Gene ontology analysis was performed using DAVID (https://david-d.ncifcrf.gov/home.jsp). Summary statistics are provided in Supplementary Table 1, and raw RNA-seq data have been deposited to the NCBI GEO under accession number GSE180867.

mtDNA sequencing and analysis

Total DNA was purified from FACS-isolated MuSCs using the QIAamp DNA Micro kit (Qiagen, 56304). mtDNA sequences were enriched by rolling circle amplification using the Repli-G Mitochondrial DNA kit (Qiagen, 151023), following manufacturer protocols60. Library preparation was performed using the Nextera XT DNA Library Preparation Kit (Illumina, FC-131-1096) following manufacturer instructions. Paired-end sequencing (2 × 150 bp) was performed using an Illumina NextSeq 500 by the Children’s Research Institute’s Sequencing Facility at UT Southwestern Medical Center. Sequencing analysis was performed using an in-house pipeline. Briefly, sequencing reads were checked and trimmed for quality control using FastQC 0.11.8 and TrimGalore 0.6.4. Trimmed reads were mapped to the Mus musculus mitochondrial genome GRCm38 using MapSplice2 2.2.1, and then paired-end reads were deduplicated (Picard 2.23.1) and quality filtered (SAMtools 1.9.0) to include high-confidence mappings. Per-base sequencing depth and point mutations were assessed using bam-readcount 0.8.0 for the quality-filtered mapped reads. Mean coverage was 7,676× (range = 5,088–9,601×) (Extended Data Fig. 9b and Supplementary Table 2). To avoid false positives and mtDNA-derived nuclear psueodgenes, variants (substitutions and indels) were called when detected on both heavy and light strands with greater than 50 supporting reads and greater than 0.5% allelic frequency. Substitutions displayed a bias towards C>T and T>C transitions (Extended Data Fig. 9e), consistent with previous mtDNA mutational signatures observed in human populations44,61, and were predominantly of low allelic frequency (Extended Data Fig. 9d). Large deletions were determined using the noncanonical junction detection mode of MapSplice2. The deletion frequency was calculated as:

$$\frac{{\mathrm{Supports}}}{{\mathrm{Supports} + {\mathrm{MeanCoverage}}\_{\mathrm{by}}\_{\mathrm{FullAlignments}}}}.$$

Here, supports is the number of read pairs supporting a large deletion, and MeanCoverage_by_FullAlignments is the average read depth using all reads that mapped to the region with no detected deletion. Junction coordinates for the deletion were extracted by regtools 0.5.1, and bedtools 2.29.2 was used to calculate MeanCoverage_by_FullAlignments from the nonjunction reads. A single large-scale deletion was noted (7,193–8,150, encompassing parts of the Cox2, ATP8, ATP6 and Cox3 genes) in MuSCs from older animals (Extended Data Fig. 9b,c (red arc)).

Statistical analysis of indel and nonsynonymous substitution frequency was performed based on a published method62. Briefly, expected frequencies of indels were calculated based on the relative size of protein-coding (69.88%) and non–protein-coding (30.12%) regions in the mouse mitochondrial genome. For expected frequencies of nonsynonymous substitutions, we took into account the observed mutational signatures at 18 and 27 months of age.

Age (months)

A>C

A>G

A>T

C>A

C>G

C>T

18

0.0058

0.18

0.072

0.011

0.00072

0.28

27

0.0044

0.23

0.063

0.0056

0.0015

0.22

Age (months)

G>A

G>C

G>T

T>A

T>C

T>G

18

0.097

0.00

0.0058

0.014

0.33

0.00

27

0.071

0.00037

0.0030

0.012

0.38

0.00

For each of the 13 protein-coding genes, we simulated 300,000 substitutions of the wild-type mouse sequence (extracted from the mouse reference mitochondrial genome: NC_005089) based on the above mutational signature. For each simulation, we calculated the frequency of nonsynonymous and synonymous substitutions. Statistical significance of the observed number of protein-coding indels and nonsynonymous substitutions was then calculated using a chi-squared test (Fig. 6h,i and Extended Data Fig. 9f ). Summary statistics are provided in Supplementary Table 2, and raw mtDNA-seq data are available at the NCBI GEO website under accession GSE180953.

Quantitative real-time PCR

Total RNA was extracted from sorted MuSCs using the RNeasy Micro Kit (Qiagen, 74004) following the manufacturer’s instructions. After RNA isolation, real-time RT-PCR were performed with Luna Universal One-Step RT-qPCR Kit (New England Biolabs, E3005) following the manufacturer’s protocol, on a CFX384 Real-Time System (SN027118, Bio-Rad). Transcript levels were normalized to β-2 microglobulin using the 2−ΔΔCT method63. For genomic PCR reactions, genomic DNA was purified from isolated MuSCs using phenol-chloroform extraction. Oligonucleotide sequences are provided in Supplementary Table 3.

Western blot analysis

For C2C12 myoblasts and MEFs, trypsinized cells were spun down, washed with PBS and resuspended in RIPA buffer (Thermo Scientific, 89900) supplemented with protease inhibitor cocktail (Roche, 11873580001), and put on ice for 30 min. Lysates were spun down at 12,000 g at 4 °C for 10 min. Protein concentrations were quantitated with the DC protein assay (Bio-Rad, 5000112). For MuSCs, 50,000 FACS-isolated cells were collected in PBS, spun down and resuspended in 50 μl RIPA buffer and processed as above. The following antibodies were used: Cox10 (Abcam, ab84053, 1:1,000), Qpc (Proteintech, 14975-1-AP, 1:1000), TFAM (Santa Cruz Biotechnology, sc-166965; 1:1,000), Scin (Santa Cruz Biotechnology, sc-376136; 1:1,000), Ndufa9 (Thermo Fisher, 459100; 1:2,000), cleaved caspase-3 (Cell Signaling Technology, 9664; 1:1,000), β-actin (Cell Signaling, 4970; 1:5,000) and histone H2B (Santa Cruz Biotechnology, sc-515808; 1:1,000). Uncropped western blot images are provided in Supplementary Figs. 2,3.

Mouse injections

Tamoxifen (Cayman Chemical, 132585) was dissolved in corn oil (Sigma-Aldrich, C8267). For most experiments, 75 mg kg−1 body mass was administered by intraperitoneal injection to 6- to 8-week-old mice once per day for five consecutive days. For experiments shown in Extended Data Fig. 2, a single dose of 200 mg kg−1 body mass of tamoxifen was administered by intraperitoneal injection to 6- to 8-week-old mice. Mice were euthanized and tissue removed at various time points after tamoxifen administration (3 h to 9 months) as indicated.

For BSO (Cayman Chemical, 14484500) administration experiments, mice were first treated with tamoxifen for 5 days (as above). Starting the day after the fifth tamoxifen administration, BSO was administered at 4 mmol per kilogram body mass once per day by intraperitoneal injection for 21 days. Mice were then euthanized, and tissue was removed for analysis.

The following antioxidants were administered: NAC (Sigma-Aldrich, A7250; 200 mg kg−1 per day), Trolox (Sigma-Aldrich, 238813; 50 mg kg−1 per day), MitoQ (Biovision, B1309; 10 mg kg−1 per day) or mitoTEMPO (Cayman Chemical, 1662125; 1 mg kg−1 per day). Mice were treated with antioxidants once per day by intraperitoneal injection for 7 days. On days 3–7, tamoxifen (75 mg kg−1) was administered by intraperitoneal injection once per day. Mice were euthanized and tissue was removed 1 day after the fifth dose of tamoxifen (7th dose of antioxidant).

Muscle injury experiments

For cryoinjury experiments, 6- to 8-week-old mice were pretreated with tamoxifen (75 mg kg−1 intraperitoneal injection once per day for 5 days) as above. Two days after the last tamoxifen injection, a cryoinjury protocol was performed. Mice were anesthetized with isoflurane. A single incision was made in the skin overlying the TA, and a metal probe (0.5 mm diameter) was cooled in liquid nitrogen and then applied directly onto the exposed TA for 10 s. Following the cryoinjury, the wound was closed with size 7–0 polyamide threads (Ethicon, 1647G). Postoperative analgesia with meloxicam (Sigma-Aldrich, M3935, 2 mg kg−1 per 24 h) was administered subcutaneously once per day for 2 days. Mice were euthanized and tissue removed between 2 and 14 days after injury as indicated.

For BaCl2 (Alfa Aesar, 0361-37-2) injury experiments, 6- to 8 week-old mice were pretreated with tamoxifen (75 mg kg−1 intraperitoneal injection once per day for 5 days) as above. Two days following the last tamoxifen injection, a muscular BaCl2 injury was administered. Mice were anesthetized with isoflurane. TA muscles were directly injected with 50 μl of 1.2 % BaCl2 (in sterile saline) using a sterile 29G needle. Postoperative analgesia with meloxicam (Sigma-Aldrich, 2 mg kg−1 per 24 h) was administered subcutaneously once per day for 2 days. Mice were euthanized and tissue removed between 2 and 21 days after injury as indicated.

Treadmill running experiments

For long-distance running experiments, mice were randomized to different treatment groups (sedentary, exercise + PBS or exercise + antioxidant). Mice were first administered tamoxifen for 5 days (as described above). On the day after the fifth tamoxifen administration, mice were acclimated to the treadmill (Columbus Instruments) by running 30 min per day for 3 days at slow speeds (up to 10 m min−1). After the acclimation period, mice were run daily on a treadmill with mild electrical stimulus and 0° inclination. The treadmill speed was set at 15 m min−1, and running was conducted 30 min per day for 6 days per week for 4 weeks. Exercised mice were treated with subcutaneous injection of NAC (200 mg kg−1 per day), Trolox (50 mg kg−1 per day), MitoQ (10 mg kg−1 per day), mitoTEMPO (1 mg kg−1 per day) or PBS 6 h before each treadmill run.

TA muscle force measurement in situ

TA muscle force was measured in animals at 21 days after BaCl2 injury using a protocol (http://www.treat-nmd.eu/downloads/file/sops/dmd/MDX/DMD_M.2.2.005.pdf). Briefly, mice were anesthetized with isoflurane, and the hindlimb was secured. The sciatic nerve was exposed in the posterolateral thigh and clamped to a custom electrode. The TA muscle was exposed, and the distal tendon was severed and attached to a force transducer via a suture (Grass Instruments, FT03-E). During measurement, the TA muscle and nerve were kept moist with 37 °C saline. The TA muscle was stimulated via the sciatic nerve using a pulse generator (Siglent Technologies, SDG2042X), and the resulting force output from the force transducer was recorded via a digital acquisition board (Dataq Instruments, DI-1110) using WinDaq software (v.3.0.7). The stretched muscle length was optimized to achieve maximal force, followed by optimization of supramaximal stimulation voltage (typically 3–4 V) and pulse duration (typically 0.3–0.4 ms). Maximal twitch force was collected for 8–10 trials and normalized by muscle cross-sectional area. Following measurement of twitch force, the force–frequency relationship was measured over stimulation frequencies from 5 to 125 Hz, at supramaximal stimulation. Data were analyzed and plotted in MATLAB (MathWorks).

Skeletal muscle clearing and deep tissue imaging

The PEGASOS tissue clearing method64 was used for whole-mount muscle clearing before deep tissue imaging. Mice were anesthetized, transcardially perfused with 50 ml ice-cold heparin PBS and then 50 ml 4% PFA. Muscles were dissected and immersed in 4% PFA at room temperature for 24 h. Subsequently, samples were incubated with decolorization solution (25% (v/v in H2O) Quadrol (Sigma‐Aldrich, 122262) for 2 days at 37 °C with gentle shaking. Next, delipidation was performed at 37 °C with gentle shaking using the following gradient of tert-butanol (tB) solutions over 2 days: (1) 30% tB (Sigma‐Aldrich, 471712) solution (70% v/v H2O, 27% v/v tB and 3% w/v Quadrol), (2) 50% tB solution (50% v/v H2O, 47% v/v tB and 3% w/v Quadrol) and (3) 70% tB solution (30% v/v H2O, 67% v/v tB and 3% w/v Quadrol).

Next, samples were immersed into dehydration solution consisting of 70% tB, 27% (v/v) poly(ethylene glycol) methyl ether methacrylate average Mn500 (PEG MMA500) (Sigma‐Aldrich, 447943) and 3% (w/v) Quadrol at 37 °C for 2 days. Samples were then incubated in clearing medium (75% (v/v) benzyl benzoate (Sigma‐Aldrich, B6630), 22% (v/v) PEG MMA500 and 3% (w/v) Quadrol) at 37 °C with gentle shaking until they achieved fully transparency. Muscles were imaged on a Zeiss LSM780 Inverted confocal microscope, and data were collected with Zeiss Zen (v.14.0.0.201) software. Imaris software (Bitplane) was used to create Supplementary Videos 1–3.

Mito-Dendra2 and tdTomato myofiber domain imaging

Myofiber domains were imaged and quantitated similar to previous studies25. Briefly, acutely dissected skeletal muscle was immediately fixed in formalin for 3–4 h at room temperature, followed by overnight at 4 °C. Muscles were rinsed with PBS and then mounted in a glass-bottom dish (MatTek, P35G-1.5–14-C). Muscles were imaged using a Zeiss LSM780 inverted confocal microscope and analyzed with ImageJ software.

Histology analysis of skeletal muscle

Following CO2 asphyxiation and cervical dislocation, the indicated skeletal muscles were rapidly dissected and then freshly frozen in liquid nitrogen–cooled 2-methylbutane, before embedding in O.C.T compound (Thermo Fisher Scientific, 23-730-571). Sections (10 μM) were cut on a cryostat (Leica, CM3050S). For H&E staining, slides were prepared following the protocol from the TREAT-NMD website (http://www.treat-nmd.eu/downloads/file/sops/cmd/MDC1A_M.1.2.004.pdf). For COX/SDH staining, slides were stained for COX activity and SDH activity based on standard operating protocols (https://neuromuscular.wustl.edu/pathol/histol/cox.htm; https://neuromuscular.wustl.edu/pathol/histol/SDH.pdf). β-Galactosidase staining was performed using a β-galactosidase staining kit (Cell Signaling Technology, 9860) following the manufacturer’s instructions. Sections were imaged using an Olympus IX83 microscope and analyzed with Image J software (National Institutes of Health) to calculating myofiber numbers and COX/SDH intensity.

Immunofluorescence protocols

For muscle sections, freshly frozen 10-μm sections (prepared as above) were fixed in formalin at room temperature for 5 min and then blocked with blocking buffer (0.25% Triton X-100 (Sigma-Aldrich, X100)) and 10% goat serum (Gibco, 16210064) in PBS at room temperature for 1 h. Sections were incubated with primary antibodies diluted in blocking buffer at 4 °C overnight. On the second day, sections were washed with PBS and then incubated with secondary antibodies diluted in blocking buffer at room temperature for 1 h. Sections were stained with DAPI diluted in PBS and then washed with PBS and mounted with fluoro-gel mounting medium (Electron Microscopy Sciences, 1798510).

For C2C12 staining, cells were fixed in formalin at room temperature for 5 min and then blocked with blocking buffer at room temperature for 1 h. Cells were incubated with primary antibodies diluted in blocking buffer at 4 °C overnight. On the second day, cells were washed with PBS and then incubated with secondary antibodies diluted in blocking buffer at room temperature for 1 h. Cells were stained with DAPI, diluted in PBS and then washed with PBS and mounted with fluoro-gel mounting medium.

Single-myofiber isolation was performed based on a published protocol65. Briefly, extensor digitorum longus or soleus muscles were digested with 2 ml Collagenase I solution (2 mg ml−1 Collagenase type I (Worthington, #CLS-1) in F-10 medium) for 1 h at 37 °C and then triturated by hand with a wide-bore pipette to release single fibers. Fibers were fixed with formalin for 10 min at room temperature, followed by blocking buffer (PBS, 10% goat serum and 0.25% Triton X-100) for 1 h at room temperature and primary antibody (diluted in blocking buffer) overnight at 4 °C. On the second day, fibers were washed with PBS, incubated with secondary antibodies (diluted in blocking buffer) for 1 h at room temperature and mounted with fluoro-gel mounting medium on glass slides.

Muscle sections, single myofibers and C2C12 cells were imaged using a Zeiss LSM780 inverted confocal microscope and analyzed using ImageJ software. For single-myofiber analysis, longitudinal mito-Dendra2 profiles were calculated in ImageJ and aligned in MATLAB (MathWorks). The following antibodies were used: Pax7 (AB528428, 2 μg ml−1), MYH3 (BF-45, 2 μg ml−1), Myogenin (F5D, 2 μg ml−1), MYH (MF20, 2 μg ml−1) (all from Developmental Studies Hybridoma Bank), MyoD (Santa Cruz Biotechnology, sc-377460, 1:500), laminin (Sigma-Aldrich, L9393, 1:500), Ki67 (Abcam, ab15580, 1:500), Tomm20 (Proteintech, 11802-1-AP; 1:500), myosin heavy chain type I (Sigma-Aldrich, M8421; 1:500), myosin heavy chain type II (Sigma-Aldrich, M4276, 1:500), cleaved caspase-3 (Cell Signaling Technology, 9664, 1:500), HA (Abcam, ab9110; 1:500), Alexa Fluor 594 goat anti-mouse IgG2b (A21145, 1:500), Alexa Fluor 594 goat anti-mouse IgG1 (A21125, 1:500), Alexa Fluor 488 goat anti-rabbit IgG(H+L) (A11034, 1:500), Alexa Fluor 594 goat anti-rabbit IgG(H+L) (A11012, 1:500) and Alexa Fluor 647 goat anti-mouse IgG2b (A21242, 1:500) (all from Invitrogen). TUNEL staining was performed using the Click-iT Plus TUNEL Assay kit (Invitrogen, C10618).

Retrovirus generation and infection

Mouse Scin cDNA (obtained from transOMIC technologies) or Cox4-DsRed (Addgene, 23215) were cloned into the retroviral vector pQCXIP (Clontech, 631516), and 3 μg pQCXIP-empty vector or pQCXIP-Scin-HA or pQCXIP-Cox4-DsRed plasmids with 1 μg pCL-Eco (Addgene, 12371) plasmids were transfected using PEI (Polysciences, 24765-1) into HEK293T (ATCC, CRL-11268) at 80% confluence in a six-well plate; 48 h after transfection, viral medium was collected and filtered through a 0.45-μm filter.

For C2C12 myoblasts, cells were plated the day before infection at 30% confluence and infected with viral mixture containing polybrene (Sigma-Aldrich, H9268, 6 μg ml−1) for 24 h. Infected cells were selected with 2 μg ml−1 puromycin. After C2C12 cell line infection with Cox4-DsRed, DsRed-positive cells were sorted using FACS Aria flow cytometer (BD Biosciences).

For primary MuSCs infection, 10,000 freshly isolated MuSCs were plated in 96-well plate with growth medium (Ham’s F-10 (HyClone, SH30025.01) with 10% horse serum, 1% penicillin/streptomycin and 2.5 ng ml−1 basic fibroblast growth factor (Peprotech, 100-18B)) overnight. MuSCs were overlaid with viral supernatant for 24 h and then transferred to wells containing 4-day C2C12 myotubes (see in vitro fusion assay below).

In vitro MuSC fusion assay

For the MuSC–myotube fusion assay, DsRed-positive C2C12 myoblasts were first cultured in growth medium (DMEM with 10% FBS and 1% penicillin/streptomycin) in an eight-well chamber slide (Thermo Fisher Scientific, 177445). When the confluence reached 90%, the culture medium was replaced with differentiation medium (DMEM with 2% horse serum and 1% penicillin/streptomycin) for 4 days to induce myotube formation. Then, 10,000 Dendra-positive MuSCs (isolated at t = 5 days after the first dose of tamoxifen) were then overlaid on myotube-containing wells. After incubation for an additional 4 days, cells were fixed and stained for myosin and DAPI, and the myotube fusion index was calculated as the fraction of myotube (myosin-positive) nuclei that also contained Dendra2 signal.

For the in vitro differentiation (MuSC-MuSC fusion) assay, 100,000 freshly isolated MuSCs were plated in an eight-well chamber slide (Thermo Fisher Scientific, 177445) with growth medium (Ham’s F-10 (HyClone, SH30025.01) with 10% horse serum, 1% penicillin/streptomycin and 2.5 ng ml−1 basic fibroblast growth factor (Preprotech, 100-18B) for 24 h. The media was then switched to differentiation medium (DMEM with 2% horse serum and 1% penicillin/streptomycin) for 4 days. Cells were fixed and stained for myosin and DAPI, and the fusion index was calculated as the fraction of total nuclei in myosin-positive myotubes, based on published methods26.

Statistical analysis

All data are represented as median and interquartile range box plots; whiskers are plotted using the Tukey method. All data are from biological replicates. No statistical tests were used to predetermine sample size. Data sets for each group of measurement were tested for normality using the Shapiro–Wilk test. If the data were not normally distributed, the data were log-transformed and retested for normality. For normally distributed data, groups were compared using the two-tailed Student’s t test (for two groups) or one- or two-way ANOVA (for more than two groups), followed by Sidak’s, Tukey’s or Dunnett’s test for multiple comparisons. For data that were not normally distributed, we used nonparametric testing (Mann–Whitney or Kolmogorov–Smirnov tests for two groups and Kruskal–Wallis test for multiple groups), followed by Dunn’s multiple comparisons adjustment. For analyzing observed frequencies of protein-coding indels and nonsynonymous substitutions and fiber-type–specific analysis, a chi-squared test was used. Multiple independent experiments with biological replicates were performed for all reported data, and the number of biological replicates is indicated in the figures.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.