Scinderin promotes fusion of electron transport chain dysfunctional muscle stem cells with myofibers

Muscle stem cells (MuSCs) experience age-associated declines in number and function, accompanied by mitochondrial electron transport chain (ETC) dysfunction and increased reactive oxygen species (ROS). The source of these changes, and how MuSCs respond to mitochondrial dysfunction, are unknown. We report here that in response to mitochondrial ROS, murine MuSCs directly fuse with neighboring myofibers; this phenomenon removes ETC-dysfunctional MuSCs from the stem cell compartment. MuSC–myofiber fusion is dependent on the induction of Scinderin, which promotes formation of actin-dependent protrusions required for membrane fusion. During aging, we find that the declining MuSC population accumulates mutations in the mitochondrial genome but selects against dysfunctional variants. In the absence of clearance by Scinderin, the decline in MuSC numbers during aging is repressed; however, ETC-dysfunctional MuSCs are retained and can regenerate dysfunctional myofibers. We propose a model in which ETC-dysfunctional MuSCs are removed from the stem cell compartment by fusing with differentiated tissue. During aging, the ability of skeletal muscle to repair itself declines, in part due to a decrease in muscle stem cells. Here, the authors report that muscle stem cells that accumulate mitochondrial damage fuse with existing muscle fibers in a manner that depends on the induction of Scinderin.

M itochondrial 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 individuals 1 , and accumulating ETC dysfunction is also observed as a secondary component of many common diseases, including aging itself [2][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 exhaustion 5,6 . 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 myofiber 7,8 . 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.

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-Cre ERT2(FAN) driver (which induces tamoxifen-dependent recombination in satellite cells, a subset of MuSCs critical for muscle regeneration [9][10][11] ) and a floxed Cox10 (an accessory subunit of mitochondrial complex IV) allele (Cox10 f ) 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 (ΔΨ m )). In MuSCs from tamoxifen-treated control mice (Cox10 f/f ; no Pax7-Cre driver; hereafter Cox10 f/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-Cre ERT2(FAN) ; Cox10 f/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).
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 (myogeninpositive (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-Cre ERT2(KARDON) allele 11 , 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 tamoxifen 15 , 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 vitro 16,17 , Cox10 deficiency has not been typically associated with acute cell death in vivo 13,[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 (D f ) 22 , which encodes a fluorescent protein (Dendra2) targeted to mitochondria. In control mice (Pax7-Cre ERT2(KARDON) ; D f/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 animals [23][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 MuSC 25 . Cox10 deletion in MuSCs (Cox10 −/− : Pax7-Cre ERT2(KARDON) ; D f/f ; Cox10 f/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).
Quantitating the number of spatially distinct domains (Fig. 2b) 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 (tdTomato f/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. Cox10

MuSC-myofiber fusion is induced by ROS.
Although subsets of quiescent MuSCs are known to occasionally fuse with myofibers in wild-type adult animals [27][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-Cre ERT2(Kardon) driver and D f 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). 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 generation 30 . We observed elevated levels of cellular superoxide, as well as elevated mitochondrial superoxide and total ROS, in the 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 prooxidant 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 tissue [33][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 fusion 38,39 , and analysis of the highest upregulated genes identified Scinderin (Scin), a member of the gelsolin family with actin network reorganization activity 40 . 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).
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 Scinderin 41 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 fusion 26,42 (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.
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 dysfunction 15,43 . 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 agedependent 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 errors 44,45 (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.
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.
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.
MuSC-myofiber fusion has been previously observed in wildtype 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 area 46 . After this initial period, the contribution of MuSCs to myofibers is slow in sedentary animals but can be readily observed with lineage tracing studies 23,25 . In our experimental models of induced MuSCmyofiber 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. 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 BaCl 2 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 SOX HI population is indicated. d, Quantitation of SOX HI 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 BaCl 2 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 BaCl 2 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 BaCl 2 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.
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 hypertrophy 47 . Indeed, a number of studies have suggested that antioxidant treatment can sometimes inhibit the beneficial effects of exercise training [48][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 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 functionality 54,55 . 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 qualitycontrol 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 differentiation 56 , 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 MuSCmyofiber fusion allows the dissection of MuSC-myofiber fusion separately from MuSC-MuSC fusion events (associated with injuryinduced 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.
The Ndufa9 f/f conditional knockout mouse was generated at the Children's Research Institute Mouse Genome Engineering Core using the Easi-CRISPR workflow 58 . sgRNAs surrounding exon 4 of Ndufa9 were selected by crossreferencing 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 Ndufa9 f/f mice are provided in Supplementary Table 3.
All mice were maintained on C57BL6 backgrounds, except Ndufa9 f/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.   (Supplementary Fig. 1b) and western blot (Supplementary Fig. 1c). An Agilent Seahorse XFe96 Analyzer was used for cellular oxygen consumption measurements of Ndufa9 +/+ and Ndufa9 f/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, CO 2free 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.

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.
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 protocols 60 . 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 mtDNAderived 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 populations 44,61 , 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: 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 largescale deletion was noted (7,150, 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 method 62 . 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. 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.

Age
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 method 63 . For genomic PCR reactions, genomic DNA was purified from isolated MuSCs using phenol-chloroform extraction. Oligonucleotide sequences are provided in Supplementary Table 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.

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 BaCl 2 (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 BaCl 2 injury was administered. Mice were anesthetized with isoflurane. TA muscles were directly injected with 50 μl of 1.2 % BaCl 2 (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 BaCl 2 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).

Mito-Dendra2 and tdTomato myofiber domain imaging.
Myofiber domains were imaged and quantitated similar to previous studies 25 . 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.
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 protocol 65 . 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.
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 methods 26 .
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.

Data availability
RNA-seq and mtDNA sequencing data have been deposited in the NCBI GEO (accession numbers GSE180953 and GSE180867). mtDNA sequencing mapping used GRCm38 (M. musculus genome; https://www.ncbi.nlm.nih.gov/assembly/ GCF_000001635.20). Other data are provided within the article, source data and supplementary materials or available upon reasonable request.