The role of Pitx2 and Pitx3 in muscle stem cells gives new insights into P38α MAP kinase and redox regulation of muscle regeneration

Skeletal muscle regeneration depends on satellite cells. After injury these muscle stem cells exit quiescence, proliferate and differentiate to regenerate damaged fibres. We show that this progression is accompanied by metabolic changes leading to increased production of reactive oxygen species (ROS). Using Pitx2/3 single and double mutant mice that provide genetic models of deregulated redox states, we demonstrate that moderate overproduction of ROS results in premature differentiation of satellite cells while high levels lead to their senescence and regenerative failure. Using the ROS scavenger, N-Acetyl-Cysteine (NAC), in primary cultures we show that a physiological increase in ROS is required for satellite cells to exit the cell cycle and initiate differentiation through the redox activation of p38α MAP kinase. Subjecting cultured satellite cells to transient inhibition of P38α MAP kinase in conjunction with NAC treatment leads to their rapid expansion, with striking improvement of their regenerative potential in grafting experiments.


Introduction
In adult mammals, many tissues can undergo repair in response to injury due to the presence of stem cells. In contrast to embryonic stem cells, most adult stem cells are maintained in a quiescent state in specialized niches (Lander et al., 2012;Ema and Suda, 2012). In response to tissue damage or changes in their microenvironment, they exit quiescence, proliferate and then differentiate in order to regenerate the tissue, or self-renew to reconstitute the stem cell pool. Progression from quiescence to proliferation and differentiation requires metabolic flexibility to adapt to changing energy demands, as exemplified by hematopoietic stem cells which when quiescent mainly rely on glycolysis, but upon activation depend on active mitochondrial biogenesis and oxidative phosphorylation (Simsek et al., 2010;Takubo et al., 2013). Such a metabolic switch, which is critical for responding to bio-energetic needs, may also be implicated in the regulation of adult stem cell behaviour, as observed for reactive oxygen species (ROS) that are by-products of oxidative phosphorylation (Murphy, 2009). Although detrimental when in excess, ROS can serve as signalling molecules at physiological levels (D'Autréaux and Toledano, 2007;Sena and Chandel, 2012), and their regulation is now emerging as an important facet of stem cell biology (Bigarella et al., 2014;Khacho and Slack, 2017;Le Moal et al., 2017). Thus increased ROS production when quiescent hematopoietic stem cells switch from glycolysis to oxidative phosphorylation has been shown to be critical for priming their activation (Jung et al., 2013). In addition to enhancing stem cell proliferation, moderate levels of ROS can also affect cell fate decisions (Khacho and Slack, 2017). However, it is still unclear how ROS levels in adult stem cells can modulate the critical balance between proliferation and differentiation in an in vivo context and whether this impacts tissue homeostasis and regeneration.
To address this issue we have used the model of adult muscle stem cells, known as satellite cells, which progress from quiescence to proliferation and differentiation during skeletal muscle regeneration. While quiescent satellite cells have a low redox state, we observed a marked increase in ROS levels in activated cells, with a peak occurring prior to the onset of muscle differentiation. To investigate the effect of redox regulation on satellite cell behaviour, we employed conditional mouse mutants for Pitx2 and Pitx3 genes (Gage et al., 1999a), encoding homeodomain transcription factors that we previously identified as critical regulators of cell redox state during foetal myogenesis (L'honoré et al., 2014). We found that the intracellular ROS level is a critical regulator of satellite cell behaviour, acting through p38a MAP kinase activity. While the moderate overproduction of ROS observed in the single Pitx3 mutant results in the premature differentiation of satellite cells, excessive ROS levels seen in double Pitx2/3 mutants lead to impaired skeletal muscle regeneration due to accumulation of DNA damage and senescence of satellite cells. Reduction of ROS levels by the antioxidant N-Acetyl-Cysteine (Richards et al., 2011), together with inhibition of P38a MAP kinase signalling (Segalés et al., 2016), leads to robust expansion of satellite cells in culture. Satellite cells cultured under these conditions show high in vivo expansion and regenerative potential upon grafting, with implications for muscle cell therapy.

Increased ROS and mitochondrial biogenesis mark the progression of satellite cells towards terminal differentiation
To investigate the regulation of mitochondrial metabolism in quiescent and committed myoblasts, we first performed a transcriptome analysis with Pax3(GFP)-positive satellite cells (Pallafacchina et al., 2010) purified by flow cytometry from adult Pax3 GFP/+ (Adult), postnatal day 7 Pax3 GFP/+ (P7) and adult dystrophic mdx/mdx:Pax3 GFP/+ (mdx) mice ( Figure 1A). Compared with quiescent satellite cells from adult muscle, activated satellite cells present during post-natal development, or in mdx muscles (Pallafacchina et al., 2010), which in the absence of Dystrophin undergo chronic regeneration, showed up-regulation of genes implicated in fatty acid metabolism and in oxidative phosphorylation, including regulators of mitochondrial biogenesis and function ( Figure 1A, , showing that activation is accompanied by metabolic changes involving increase in both mitochondrial activity and ROS production. We then investigated mitochondrial activity during the transition from proliferation to differentiation. Satellite cells were purified by flow cytometry from pectoralis, diaphragm and abdominal muscles of adult Pax3 GFP/+ mice and both respiration and glycolysis were measured by Seahorse analysis after different days of culture (D2-D4) (Figure 1figure supplement 1E-G). While proliferative satellite cells (D2) predominantly display glycolytic metabolism, the contribution of oxidative phosphorylation increases at day 3 (D3, Figure 1-figure supplement 1E-G), a time that precedes differentiation by one day, as evidenced by Myogenin expression (Figure 1-figure supplement 1J). To investigate the kinetics of mitochondrial function and of ROS production in vivo during muscle regeneration, satellite cells were then purified by flow cytometry from pectoralis muscles of adult Pax3 GFP/+ mice at different days (D0 to D7) after notexininduced injury ( Figure 1B). While quiescent satellite cells are characterized by a low mitochondrial mass (Latil et al., 2012;Tang and Rando, 2014), we observed an increase in mitochondrial mass occuring two days after injury and continuing thereafter ( Figure 1C (adult) and activated (P7 and mdx) satellite cells were purified respectively from adult Pax3 GFP/+ (Ad, 98% quiescent), postnatal day 7 Pax3 GFP/+ (P7, 80% activated) and adult dystrophic mdx:Pax3 GFP/+ (mdx, 30% activated) mice. In each case, pectoralis, abdominal and diaphragm muscles were used for isolation by flow cytometry of Pax3(GFP)-positive satellite cells. After purification, RNA was isolated from Pax3(GFP)-positive cells for transcriptome analysis Figure 1 continued on next page onset of differentiation, marked by Myogenin expression and down-regulation of Pax7 ( Figure 1E). Strikingly, quantification of total H 2 O 2 production and mitochondrial ROS production revealed a marked increase at two days after activation ( Figure 1F, Figure 1-figure supplement 1H), a time when satellite cells initiate differentiation ( Figure 1E). Analysis of cytoplasmic ROS levels showed an increase at day 2 after injury that reaches a peak at day 3 ( Figure 1G), when the level of GFP, reflecting Pax3 transcription, is low, as also seen for Pax7 ( Figure 1E). Subsequent increase in GFP fluorescence ( Figure 1G) corresponds to later self-renewal of the Pax3-positive satellite cell pool. As these results clearly indicate that intracellular ROS content reaches a peak before the onset of satellite cell differentiation, we then examined whether it is required to drive differentiation. To address this issue, and since we had previously identified Pitx2/3 genes as essential regulators of the redox state in foetal muscle precursor cells (L'honoré et al., 2014), we employed conditional Pitx2/3 mutant mice. Both Pitx2 and Pitx3 transcripts and their corresponding proteins are up-regulated in activated satellite cells, marked by MyoD expression and then by differentiation markers Myogenin and Troponin T, in vivo after injury ( Figure 1H Conditional ablation of the floxed-Pitx2 gene (Gage et al., 1999b) alone, or in combination with the constitutive Pitx3 mutant (L'honoré et al., 2014) was carried out by five intra-peritoneal injections of 4-hydroxy-tamoxifen (4-OHT). Pax3(GFP)-positive cells were then purified by flow cytometry from the pectoralis muscles of Pitx2 flox/+ :Pitx3 +/-:Pax3 GFP/+ :R26R Cre-ERT2/+ controls, Pitx3 -/-:Pax3 GFP/+ and Pitx2 flox/flox :R26R Cre-ERT2/+ :Pax3 GFP/+ single mutants and Pitx2 flox/flox :Pitx3 flox/-:R26R Cre-ERT2/+ : Pax3 GFP/+ double mutants, in the absence of injury (D0) or 2 and 3 days (D2, D3) after notexin injection ( Figure 1I). As previously shown ( Figure 1F,G), ROS levels increase markedly from day 2 after injury in control satellite cells isolated from Pitx2 flox/+ :Pitx3 +/-:Pax3 GFP/+ :R26R Cre-ERT2/+ mice ( Figure 1J). While Pitx2 mutant satellite cells display a similar redox state as control cells during regeneration, we observed abnormal moderate or high increases in ROS at days 2 and 3 after injury in satellite cells from single Pitx3 or double Pitx2/3 mutant mice, respectively ( Figure 1J). Comparison of ROS levels in Pax3(GFP) satellite cells isolated from double heterozygotes versus wild-type mice after notexin injury gave similar results (Figure 1-figure supplement 1M). We therefore systematically employed the double heterozygote as a control.  (Pallafacchina et al., 2010). (B) Experimental scheme used for the purification of quiescent and activated satellite cells during regeneration. (C-H) GFPpositive cells were isolated from the pectoralis muscles of adult Pax3 GFP/+ mice at different days (D) after notexin injection or from control muscles at D0. (C) GFP-positive cells isolated at D0 (quiescent) and D2 to D4 after notexin injury (activated) were immediately incubated with Mitotracker probe and analysed by flow cytometry to measure fluorescence intensity (f.i). (D) DNA samples prepared from Pax3(GFP)-positive cells were analysed by qPCR for the level of the mitochondrial COI gene relative to the level of the nuclear encoded NDUFV1 gene, expressed as a log ratio. (E) RNA samples prepared from Pax3(GFP)-positive cells were analysed by qPCR for transcripts of Pax7 and Myogenin relative to the level of Hprt transcripts, expressed as a log ratio. (F, G) Immediately after purification at D0-D5 (F) or D0-D7 (G), GFP-positive cells were incubated with Mitosox (F) or Cell Rox (G) probes and the relative fluorescence intensity (f.i) measured by flow cytometry. (H) RNA samples as in (E) were analysed by qPCR for Pitx2 and Pitx3 transcripts. (I) Double Pitx2/3 mutant Pitx2 flox/flox :Pitx3 flox/-:Pax3 GFP/+ :R26R Cre-ERT2/+ , single Pitx2 mutant Pitx2 flox/flox :Pax3 GFP/+ :R26R Cre-ERT2/+ , single Pitx3 mutant Pitx3 -/-:Pax3 GFP/+ , and control Pitx2 flox/+ :Pitx3 +/-:Pax3 GFP/+ :R26R Cre-ERT2/+ adult mice were obtained by 4-hydroxytamoxifen (4-OHT) intra-peritoneal (IP) injection on 5 consecutive days (D) and subjected to muscle injury by notexin injection into the pectoralis muscle (D0). (J) GFP-positive cells isolated as in (B) at D0, 2 and 3 after injury were incubated with Cell Rox probe and the relative fluorescence intensity (f.i.) measured by flow cytometry. Error bars (C-H, J) represent the mean ± s.d with n ! 6 independent animals (C, E-H, J) or 4 independent animals (D). Error bars represent the mean ± s.d, with *p<0.05, **p<0.01, ***p<0.001, ****p<0.001. Please see Increased ROS production is critical for satellite cell progression from proliferation to differentiation through redox activation of p38a MAP kinase To test the consequences of a moderate ROS increase on activated satellite cell behaviour, we used Pitx3 -/mice. At birth, Pitx3 mutants were indistinguishable from controls, but then displayed a progressive growth defect (Figure 2-figure supplement 1A), with reduced muscle mass and a decrease in the myofibre cross-sectional area (CSA) (Figure 2-figure supplement 1B,C), a phenotype that persists in adults (Figure 2A,B, Figure 2-figure supplement 1D). To investigate the behaviour of Pitx3 mutant satellite cells during muscle regeneration, we injured the Tibialis Anterior (TA) muscle of Pitx3 -/mutant and Pitx3 +/animals by cardiotoxin injection. Three weeks after injury, regenerated fibres of control mice have a similar size to that of uninjured fibres, whereas mutant muscles exhibit a decreased CSA (Figure 2A,B). This phenotype, which is correlated with a deficit in centrally located myonuclei in regenerating mutant fibres compared to controls ( Figure 2C), may reflect abnormal behaviour of Pitx3 mutant satellite cells. We therefore examined the kinetics of their differentiation during regeneration. While in control animals, most satellite cells are still proliferative 3 days after injury, Pitx3 mutant cells show a marked decrease in the number of Pax7/EdU double positive cells ( Figure 2D,E), together with a two fold increase in the number of differentiating cells marked by Myogenin and embryonic-Myosin Heavy Chain (Emb-MyHC) ( Figure 2D,F). These results indicate that Pitx3-deficient satellite cells undergo premature differentiation, a feature that could account for the muscle regeneration defect observed in Pitx3 -/mice.
To challenge their regenerative potential, Pitx3 mutant and control mice were subjected to three rounds of TA muscle injury by cardiotoxin injection ( Figure 2G). After the last round, muscle sections from control animals displayed a normal pattern of regenerating centro-nucleated fibres ( Figure 2H-J). In contrast, Pitx3 mutant muscles exhibited impaired regeneration marked by fibrosis ( Figure 2H To evaluate the contribution of altered redox state to the premature differentiation of Pitx3 mutant satellite cells, Pax3(GFP)-positive cells were purified by flow cytometry from muscles of control Pitx3 +/-:Pax3 GFP/+ and mutant Pitx3 -/-:Pax3 GFP/+ adult mice and their behaviour analysed in culture in the presence or absence of the ROS scavengers N-Acetyl-Cysteine (NAC) (Richards et al., 2011) or Trolox (Le Moal et al., 2017). As expected, while the absence of Pitx3 does not affect satellite cell activation marked by MyoD expression (data not shown), it leads to their premature differentiation. This phenotype is illustrated by a significant increase in the number of Myogenin-and Troponin T-positive cells at days 3 and 4 in the mutant population compared to controls ( Figure 3A, We then investigated the mechanism by which ROS regulate the onset of differentiation and focussed on the redox sensor p38a MAP kinase (Ito et al., 2006;Brien et al., 2013;Segalés et al., 2016). Measurement of p38a MAP kinase phosphorylation demonstrated its premature activationin Pitx3 mutant cells at day 3 of culture ( Figure 3D), concomitantly with Myogenin expression ( Figure 3A). In addition, treatment of Pitx3 mutant cells with either the p38a MAP kinase chemical inhibitor SB203580 (Bain et al., 2007) (Figure 3E), siRNA directed against p38a MAP kinase ( Figure 3F Figure 3J, addition of 0.5 nM TNFa leads to reduced proliferation together with premature differentiation. These effects are prevented by the presence of NAC, showing they result from ROS increase. Control Pax3(GFP)-positive satellite cells were transfected with siRNA directed against transcripts encoding the mitochondrial uncoupling factor, Ucp2 (Mailloux and Harper, 2011) and analysed after 3 days of culture for mitochondrial content, ROS levels, and for Myogenin expression. As expected, inhibition of mitochondrial uncoupling (  (Gleyzer et al., 2005) using siRNA led to a decrease in ROS levels together with delayed differentiation (Figure 3-figure supplement 2D and E). We then investigated whether NADPH oxidases, enzymes present at the cell membrane and in the endoplasmic reticulum, could also contribute to ROS production in activated satellite cells. qRT-PCR analyses indicated that Nox2 and Nox4 expression increases in the days (D2-D4) following injury (  The number of fibres with one or two and more centro-located nuclei was counted on transverse sections of control and mutant muscle, and expressed per section. (D) Three days after injury, control and mutant animals were treated with EdU for 5 hr and muscles were dissected. Muscle cryo-sections were stained with DAPI (blue) and analysed by immunofluorescence with Pax7 (red), Myogenin (green) and Embryonic Myosin Heavy Chain (Emb-MyHC, red) antibodies and by the EdU reaction (green). (E) Pax7/EdU double positive-cells expressed as a percentage of Pax7-positive cells and (F) Myogenin-positive and Emb-MyHC-positive cells were counted per field. (G-J) Mutant Pitx3 -/and control Pitx3 +/mice were subjected to three successive muscle injuries (D0, D21 and D42) by cardiotoxin injection into the TA muscle. Three weeks after the third injection (D63), injured and contralateral uninjured muscles were dissected. (H) Cryo-sections were analysed by hematoxylin-eosin (HE) and red sirius (RS) staining and by immunofluorescence with a Pax7 (red) antibody on DAPI stained sections. (I) CSA of control and mutant fibres was measured in regenerated and contralateral uninjured TA muscles, and is represented as a mean value. (J) Pax7 positive-cells were counted and quantified per field. The experiments were performed with n ! 3 animals, and a representative image is shown. (B, C, E, F, I, J), error bars represent the mean ± s.d, with **p<0.01, ***p<0.001. (A, H) Scale bar, 100 mm, (D) Scale bar, 50 mm. Please see Figure Figure 4A). Three weeks after injury, Pitx2/3 double mutant mice show severely impaired muscle regeneration with very small fibres, signs of fibrosis, and numerous cells undergoing senescence, in comparison to sections of controls, which display the usual pattern of regenerating centro-nucleated fibres ( Figure 4B). This phenotype was also seen in the absence of Pitx2/3 in the muscles of mdx mice, where chronic muscle regeneration takes place in the absence of Dystrophin ( Figure 5A). Consistent with what we observed during foetal myogenesis (L'honoré et al., 2014), mutation of both Pitx2 and Pitx3 in satellite cells leads to altered expression of genes for antioxidant proteins (Mgst2, SelX, SelM, Gstz, Hspb2) and for the regulator of mitochondrial function NRF1 (Gleyzer et al., 2005) (Figure 5-figure supplement 1A). We then assayed proliferation and differentiation properties of control and double mutant satellite cells. Similarly to what we observed with the Pitx3 mutant, Pitx2/3 deficient satellite cells exhibit an expansion defect ( Figure 5B,C). NAC addition to the culture medium alleviates this phenotype ( Figure 5B,C), and also exerts a positive effect on control cells by improving both their cloning efficiency, probably reflecting improved survival ( Figure 5-figure supplement 1B), and their expansion ( Figure 5B,C). In contrast to Pitx3-deficient satellite cells, which undergo differentiation following premature cell cycle exit, mutation of both Pitx2 and Pitx3 genes leads to a marked reduction in differentiation markers ( Figure 5D,E(i), Figure 5- (Puri et al., 2002). In the presence of both NAC and SB (NAC/SB), the effect is similar to that of NAC alone ( Figure 5E, Figure 5-figure supplement 1G). Altogether, these data demonstrate that the excessive ROS levels observed in Pitx2/3 double mutant satellite cells are responsible for their senescence, due to protein oxidation and DNA damage, a behaviour that probably accounts for the regeneration deficit observed in Pitx2/3 mutant mice in vivo after injury ( Figure 4B) or in dystrophic conditions (Figure 4-figure supplement 1).

Manipulation of satellite cell redox state has implications for muscle cell therapy
In view of our results showing the impact of ROS levels on proliferation, differentiation and survival, we reasoned that manipulation of ROS levels and/or signalling by exogenous treatments could be an efficient way to expand wild-type satellite cells in culture, while preventing them from undergoing differentiation, two crucial features for maintaining their regenerative potential. Pax3(GFP)-positive satellite cells were cultured in the presence or absence of NAC, SB203580 (SB), or both compounds (NAC/SB). Immunostaining experiments ( Figure 6A) show that while NAC or SB treatment alone leads to a reduction in p38a MAP kinase phosphorylation, the NAC/SB combination has a more pronounced effect on p38a MAP kinase phosphorylation and therefore activation of this signalling pathway. Accordingly, while NAC and SB treatments, under which proliferation is promoted ( Figure 6B C), both reduce differentiation ( Figure 6D), treatment with the dual NAC/SB combination leads to even stronger effects ( Figure 6D). Indeed, whereas after 3 days of culture in control conditions only 40% of cells show EdU incorporation, treatment with the NAC/SB combination leads to a major increase, with 80% of cells positive for EdU ( Figure 6E, Figure 6-figure supplement 1A(i)). Clonal analysis shows that this strong proliferative activity promotes a 10 fold increased expansion compared to controls, with a mean number of 250 cells per clone compared to 25 ( Figure 6B).  (ii)) were significantly affected under these conditions. It is notable that SB has little effect on ROS levels compared to NAC (Figure 6-figure supplement 1E), which is also reflected in cloning efficiency where NAC promotes cell viability and prevents senescence by reducing ROS levels ( Figure 6-figure supplement 1F). In view of the synergistic effect seen with NAC and SB together ( Figure 6E), we checked whether ERK, JUNK, p65 (NFkB) and ATM were affected and found no evidence that these signalling pathways were implicated ( Figure 6-figure supplement 1G,H, data not shown).
In view of these results, we anticipated that NAC/SB treatment, in addition to promoting expansion of satellite cells in culture, might improve their regenerative potential in vivo, by maintaining them in a highly proliferative and undifferentiated state. To test this hypothesis, 15.000 freshly isolated Pax3(GFP)-positive satellite cells, or 15.000 Pax3(GFP)-positive satellite cells cultured for three days (D3) in the absence or in the presence of NAC/SB, were grafted into the Tibialis Anterior muscle of Rag2 À/À :Il2rb -/-:dmd/dmd mice (Vallese et al., 2013) (Figure 6-figure supplement 1I). Three weeks after grafting, muscle sections were assayed for Dystrophin expression by immunofluorescence ( Figure 6F). While satellite cells cultured under standard conditions display a regenerative potential at least 10 times lower than freshly isolated satellite cells (Montarras et al., 2005) ( Figure 6F,G), NAC/SB treatment of cultured satellite cells leads to a very efficient engraftment, with a number of Dystrophin-positive cells similar to that obtained with freshly isolated satellite cells ( Figure 6F,G). To test whether engrafted cells also contributed to the satellite cell compartment in recipient muscles, we performed immunostaining with GFP and Pax7 antibodies. Both staining colocalized in a subset of nuclei in the engrafted muscle, indicating that the transplanted cells also successfully populated the satellite cell compartment (Figure 6-figure supplement 1J). To address the issue of the expansion and self-renewal of control and treated cells, 15.000 cells cultured for three days in control, NAC, SB or NAC/SB conditions were grafted into the Tibialis Anterior muscle of Rag2 À/À :Il2rb -/-:dmd/dmd mice. Injury regeneration assays performed 3 weeks after grafting ( Figure 6H-K) show that donor satellite cells treated with both NAC and SB203580 are much more abundant, (x20, Figure 6K) than when treated with NAC or SB alone, which also have a positive effect compared to controls (x2, Figure 6I). When put in culture, all the Pax3(GFP)-positive cells recovered from grafted muscles displayed a MyoD-positive phenotype (data not shown), confirming their myogenic identity. In addition, this grafting analysis ( Figure 6J,K) further indicates that while NAC or SB treatments both improve the regeneration potential of cultured satellite cells ( Figure 6J, K), the use of both compounds leads to a much higher number of Dystrophin-positive fibres ( Figure 6J,K). Taken together, these results indicate that cells treated with both compounds display (A) After 4 days in culture, cells were processed for phospho-p38a (pp38a) expression by immunofluorescence. pp38a-positive cells were counted, and Figure 6 continued on next page a higher regeneration potential, together with a much higher capacity to expand and self-renew upon grafting.

Discussion
Our analysis of Pitx2 and Pitx3 mutant mice demonstrates the importance of these transcription factors in regulating the behaviour of adult muscle stem cells and consequently the regeneration of skeletal muscle. In the double mutant, these satellite cells undergo DNA damage and become senescent. This is in contrast to the situation in the embryo where muscle progenitor cells respond to DNA damage by undergoing apoptosis in the absence of both factors (L'honoré et al., 2014). As during foetal myogenesis, we show that Pitx2/3 factors are involved in the redox control of the muscle satellite cell. The link between Pitx factors and the control of ROS has also been reported in cardiomyocytes where Pitx2, the only Pitx gene expressed in the heart, plays a critical role in the regulation of antioxidant scavenger genes (Tao et al., 2016).
In the absence of Pitx3 alone, the balance between proliferation and differentiation is perturbed so that satellite cells differentiate prematurely, a phenotype that impacts the efficiency of muscle regeneration. This mutant provides a valuable tool to investigate the regulation of this critical balance. Although historically viewed as harmful, ROS are now established as important regulators of signalling pathways (Bigarella et al., 2014;Holmströ m and Finkel, 2014;Khacho and Slack, 2017). Our analysis of skeletal muscle regeneration in the Pitx3 mutant reveals a critical role for the redox state of satellite cells in the regulation of their behaviour (Figure 7). While quiescent satellite cells exhibit low ROS levels, we observed a marked increase in activated cells at the onset of differentiation. Using chemical as well as genetic manipulation of the redox state of satellite cells during muscle regeneration, we show that the increase in ROS occurring in proliferating satellite cells leads to activation of p38a MAP kinase and link this redox activation to its function as a critical regulator of satellite cell differentiation (Segalés et al., 2016;Perdiguero et al., 2007;Brien et al., 2013;

days in culture, cells were incubated with EdU and processed for EdU detection by fluorescence (C) or for Myogenin (Myog) detection by immunofluorescence (D). EdU-positive cells (C) and Myogenin-positive cells (D) were counted and quantified as the percentage (%) of total cells. (E) GFP-positive cells isolated as in (A) were cultured for 3 days (D3) in the absence (ctrl) or in the presence of both NAC and SB203580 (NAC/SB). Cells were incubated with
EdU and processed for EdU detection by fluorescence, and for Myogenin or P21 detection by immunofluorescence. Myogenin, EdU and p21 positive cells were counted, and quantified as the percentage (%) of total cells. (A-E) The experiments were performed with n ! 3 animals for each condition. Error bars represent the mean +/-s.d, with *p<0.05, **p<0.01, ****p<0.001. (F-G) Tibialis Anterior (TA) muscles of Rag2 -/-:Il2rb -/-:dmd/dmd mice were irradiated and grafted with 15.000 cells freshly isolated from Pax3 GFP/+ mice (Fresh) or cultured for 3 days in the absence (D3) or in the presence of both NAC and SB203580 (D3 NAC/SB). Three weeks after grafting, TA muscles were dissected and cryo-sections analysed by immunofluorescence. (F) Immunofluorescence for Dystrophin and (G) the number of Dystrophin-positive fibres was counted for each condition. (F-G) The experiments were performed with n ! 3 grafted TAs for each condition, and a representative image is shown, scale bar, 200 mm (F). (H-K) Tibialis Anterior (TA) muscles of Rag2 -/-:IlrRb -/-:dmd/dmd mice were irradiated and grafted with 15.000 cells isolated from Pax3 GFP/+ mice and cultured for 3 days in the absence (ctrl) or in the presence of NAC, SB203580 (SB), or both NAC and SB203580 (NAC/SB). Three weeks after grafting, for each mouse, one TA muscle was subjected to injury by cardiotoxin injection, and the contralateral muscle was used as a control. Six weeks after grafting, all TA muscles were dissected. Given the striking synergistic effect that we observe with NAC which lowers ROS levels and SB which inhibits p38a MAP kinase, it is probable that ROS levels also affect satellite cell behaviour independently of p38a MAP kinase. This is demonstrated by our analysis of the double Pitx2/3 mutant, where DNA damage is reversed by NAC treatment but not by SB, indicating that they reflect a p38aÀindependent effect of high ROS levels. In wild-type satellite cells too, cloning efficiency is improved by NAC but not SB treatment suggesting that lowering ROS levels in normal cells also reduces a background level of DNA damage. The striking effect of NAC and SB together may reflect reduced DNA damage due to ROS, which will repercuss on the cell cycle, as well as delayed differentiation (Shaltiel et al., 2015).

Redox regulation of muscle stem cells during muscle regeneration
ROS are mainly generated by the mitochondrial respiratory chain and by NADPH oxidases (Murphy, 2009;Katsuyama et al., 2012). While quiescent satellite cells, with low energetic requirements, are characterized by reduced mitochondrial mass and low ROS production, their activated progeny is marked by increased mitochondrial mass and activity and enhanced ROS levels. We show that a peak of mitochondrial ROS production occurs at day 2 of regeneration, when mitochondrial mass is only beginning to increase in activated satellite cells. Under conditions in which the cellular energy demand is still moderate, increased ROS production has been reported to occur from complex I (Murphy, 2009). In addition to mitochondrial ROS, by using Apocynin, an inhibitor of NADPH oxidases (Petrô nio et al., 2013), we demonstrate a contribution of these enzymes to the redox state of satellite cells and the regulation of satellite cell differentiation. In this study we have focused on the cell-autonomous role of the satellite cell redox state, but extracellular sources of ROS released into the environment by injured muscle fibres and inflammatory cells (Le Moal et al., 2017), will also contribute to the in vivo scenario.

Interplay between redox regulation, metabolic reprogramming and epigenetic modifications
The metabolic switch that accompanies the satellite cell transition from quiescence to ex vivo activation (Ryall et al., 2015) results in a decrease of intracellular NAD(+) levels and of the activity of the histone deacetylase, Sirt1, leading to elevated H4K16 acetylation and activation of muscle genes required for the entry of satellite cells into the myogenic program (Ryall et al., 2015). In the hematopoietic system, Sirt1 has been shown to be critical for the maintenance of hematopoietic stem cells (Matsui et al., 2012). The premature differentiation seen in Sirt1 mutant satellite cells (Ryall et al., 2015), resembles what we observe in the Pitx3 mutant (Figure 7-figure supplement 1). In addition, Sirt1 has been characterized as a direct redox sensor, which is inactivated upon oxidation of its cysteine residues (Zee et al., 2010). It is therefore possible that increased ROS levels during normal regeneration lead to Sirt1 inactivation in parallel with activation of p38a MAP kinase, and consequent onset of differentiation. This parallel action may contribute to the more striking effect that we observe with NAC, which reduces ROS levels with a consequent increase in Sirt1 activity, compared to the effect of inhibiting p38a MAP kinase with SB203580 alone, which does not correct the DNA damage and only partially corrects the differentiation defect seen in Pitx2/3 double mutant mice. It should be noted that the level of SB203580 that we employ (5 mM) is lower than that used by other authors (10 to 25 mM) who have reported beneficial effects of this drug on the regenerative properties of satellite cells (Bernet et al., 2014;Charville et al., 2015). We chose to use a lower dose because of off-targets effects of the drug on other signalling pathways (Bain et al., 2007). The effects that we observe on p38a phosphorylation in wild type cells are similar with SB203580 and with NAC, indicating that the dose of SB203580 is effective. SB203580 acts on threonine 106 to inhibit phosphorylation, whereas ROS, and therefore the ROS scavenger NAC, can affect the redox state of cysteine residues involved in substrate binding and activation of p38a MAP kinase (Bassi et al., 2017). Additional effects of NAC due to the lowering of ROS levels, through a reduction in DNA damage and a potential increase in Sirt1 activity, are independent of p38a MAPK kinase (Figure 7).

Muscle stem cell ageing and perturbation of redox homeostasis
In skeletal muscle, as in many tissues, ageing is accompanied by a decline in regenerative capacity (Sousa-Victor et al., 2014;Cosgrove et al., 2014). While the aged systemic environment significantly impacts the function of satellite cells (Gopinath and Rando, 2008), recent studies have shown that impaired regeneration is also due to a cell-autonomous functional decline in satellite cells (Sousa-Victor et al., 2014;Cosgrove et al., 2014;Bernet et al., 2014). Skeletal muscle is a remarkably stable tissue, and satellite cells are long-lived resident adult stem cells. Despite their low metabolic rate and their high antioxidant capacity (Pallafacchina et al., 2010), these long-lived cells are particularly exposed to ROS generated either by mitochondria during muscle fibre contraction (Jackson, 2011), or by NADPH oxidases shown to be over-expressed in sarcopenic fibres (Kerkweg et al., 2007). This exposure to ROS throughout life combined with a decline in their antioxidant capacity during ageing (Sullivan-Gunn and Lewandowski, 2013;Fulle et al., 2005) makes geriatric satellite cells particularly vulnerable to accumulation of genomic and mitochondrial DNA mutations, and as a consequence, to mitochondrial dysfunction and genomic instability. Geriatric satellite cells have been shown to have elevated p38a MAP kinase activity (Cosgrove et al., 2014;Bernet et al., 2014) and defective autophagy leading to mitochondrial dysfunction, excessive ROS levels (García-Prat et al., 2016), with increased DNA damage (García-Prat et al., 2016;Sinha et al., 2014) and senescence (Sousa-Victor et al., 2014;Latella et al., 2017). In view of the Pitx2/3-mutant phenotype (Figure 7-figure supplement 1), and in correlation with observations on hyper-homocysteinemic mice (Veeranki et al., 2015), we propose that Pitx2/3 mutants may represent a useful model of premature muscle ageing. Future experiments will address this issue.

Redox homeostasis and stem cell therapies
A major obstacle to the development of stem cell based therapies is the number of cells required for transplantation (Briggs and Morgan, 2013;Walasek et al., 2012). Ex vivo amplification of tissue specific stem cells is an option, but it generally leads to a reduction of their regenerative potential, due to their commitment to differentiation (Montarras et al., 2005;Walasek et al., 2012;Briggs and Morgan, 2013). In the muscle field, advances have been made by manipulating dystrophic muscle (Price et al., 2014;Tierney et al., 2014), by modifying culture conditions (Gilbert et al., 2010;Liu et al., 2012;Parker et al., 2012;Charville et al., 2015), or by grafting selected satellite cell sub-populations (Jean et al., 2011), however further improvements are mandatory. Treatment with SB203580, that inhibits p38a MAP kinase activation required for the onset of differentiation, promotes proliferation of human myoblasts in culture and their contribution to the reconstitution of the satellite cell pool upon grafting into injured skeletal muscles (Charville et al., 2015). NAC treatment of Pitx2 deficient mice restores the regenerative capacity of the newborn heart (Tao et al., 2016), a situation that is reminiscent of our finding for skeletal muscle in triple Pitx2/3:Dystrophin-deficient mice, where the dystrophic muscle phenotype is improved by NAC treatment. We now show that the use of the NAC/SB203580 combination in culture not only leads to fast and very efficient amplification of satellite cells, but also maintains their regenerative potential at levels similar to that of freshly isolated cells in grafting experiments. In vivo also, treatment with both compounds promotes expansion of the transplanted satellite cell pool after injury. These observations have important implications for stem cell therapy.

Materials and methods
Animal breeding, Cre recombination and genotyping All animal procedures were approved and conducted in accordance with the Institut Pasteur animal ethics committee (CEEA Institut Pasteur n˚2013-0017 and APAFIS #2455 2015 1122133311) following the regulations of the Ministry of Agriculture and the European Community guidelines.
Recombination of floxed alleles was obtained by five daily intra-peritoneal injections of 4hydroxy-tamoxifen (4-OHT, Sigma). 4-OHT was prepared by dissolving a freshly opened bottle of 50 mg in EtOH 100% at 80 mg/ml. The mixture was incubated at 55˚C with constant agitation until 4-OHT was completely dissolved (~1-2 hr). After complete solubilisation, 4-OHT was diluted 1:2 with Cremophor EL (Sigma) and kept at 4˚C. Just before injection, 4-OHT from this stock was diluted 1:4 in PBS and administered to 6 to 8 weeks-old control and mutant mice (1 mg final 4-OHT/mouse of 25 g). Recombination efficiency was verified by PCR using genomic DNA isolated from purified satellite cells three days after the last injection.
Description of mouse lines and their use are given in Supplementary file 1. Briefly, experiments that do not require satellite cell purification were performed on animals in which Pitx2/3 genes were conditionally recombined using the Pax7 Cre-ERT2/+ mice line, except for experiments using the single constitutive mutant Pitx3. As Pax3 GFP/+ :Pax7 Cre-ERT2/+ animals are not viable due to non-muscle defects, we used the inducible R26R CRE-ERT2/+ mouse line for all experiments requiring control and mutant satellite cell purification by flow cytometry, except for experiments using the single constitutive mutant Pitx3.
Unless indicated, all experiments were performed on 6 to 8 weeks-old adult animals.
N-Acetyl-Cysteine (NAC, Sigma) was diluted just prior to use at 500 mM in H 2 O in light-protected bottles (Richards et al., 2011) and given ad libitum to mice, with the solution being changed every three days.

Muscle injury
Mice were anesthetised with 0.5% Ketamine (Imalgene1000, 100 mg/kg, Merial) and 0.5% Xylazine (Rompun 2%, 20 mg/kg, Bayer). Pectoralis and Tibialis Anterior muscles were injected with 10 ml of notexin (12.5 mg/ml, Latoxan) and 20 ml of cardiotoxin (20 mg/ml, Latoxan) respectively. After injection, mice were kept on a warm plate until recovery. Mice were sacrificed by cervical dislocation at different days after injury and the muscles used for analysis by histology, immunofluorescence or for satellite cell purification.
Regeneration experiments followed by histology and immunofluorescence were performed on the Tibialis Anterior (TA) muscle injected with cardiotoxin. As Pax3(GFP) expressing satellite cells are much less abundant in TA muscle (Relaix et al., 2006), purification of activated satellite cells during regeneration was performed at different days after intramuscular injection of the pectoralis muscle with notexin. Notexin was used in this case in order to obtain a more widespread injury than with cardiotoxin (Hardy et al., 2016).

Primary cell purification and culture
Satellite cells were prepared and cultured as previously described (Montarras et al., 2005) with minor modifications. For primary cell culture, dissected diaphragm, abdominal and pectoralis muscles of independent adults (6-8 weeks old mice) were incubated in digestion buffer (0.1% Trypsin, 0.1% Collagenase V [Roche]) at 37˚C with frequent agitation. After successive digestions, pooled supernatants were centrifuged at 1800 rpm for 20 min and cells were purified by flow cytometry via MoFlow Astrios (Beckman Coulter) on the basis of size, granularity and GFP fluorescence. A mean of 70.000 to 100.000 Pax3(GFP)-positive cells were obtained per adult mouse. 5.000 Pax3(GFP)-positive cells were seeded on 35 mm plastic dishes pre-coated with gelatin. Cells were cultured in growth medium (40% F12, 40% DMEM, 20% SVF, 2% Ultroser [Biosepra]) and left to differentiate spontaneously. When cultured under 3% O 2 , cells were seeded immediately after purification in a Hypoxia Chamber Workstation (SCI-tive, Baker Ruskinn) at 37˚C, 5% CO 2 and 70% humidity. Each experiment corresponded to a minimum of three individual mice that were treated and analysed separately. Clonal analysis was performed as previously described (Pallafacchina et al., 2010). For each genotype, two 96 plates were analysed for each condition (in the absence or presence of N-Acetyl-Cysteine, SB20380 or both) for three individual mice.
For purification of satellite cells during regeneration, a similar protocol was used with minor modifications: dissected pectoralis muscles from different adult mice were incubated in digestion buffer (0.1% Trypsin, 0.1% Collagenase V [Roche]) at 37˚C with low agitation. After successive digestions, pooled supernatants were centrifuged at 1800 rpm for 20 min and cells were purified by flow cytometry via MoFlow Astrios (Beckman Coulter) on the basis of size, granularity and GFP fluorescence. A mean of 10.000 to 30.000 Pax3(GFP)-positive cells were obtained per muscle. When the number of purified Pax3(GFP)-positive cells was too low to get significant results, several animals were pooled in order to obtain each biological sample.
N-Acetyl-Cysteine (Sigma), Trolox (Abcam) and SB203580 (InVivoGen) were used for cell seeding at 10 mM, 10 mM, and 5 mM respectively in growth media. Apocynin (Santa Cruz Biotechnology), Buthionine Sulfoximine (Enzo Life Sciences) and TNFa (R and D systems) were added to the cell culture medium at 10 mM, 10 mM and 0.5 nM respectively.

Satellite cell transplantation
Three months-old immune-compromised Rag2 À/À :Il2rb -/-:dmd/dmd mice (Vallese et al., 2013) were subjected to leg-X ray irradiation using a Faxitron equipment 4 days before cell engraftment. For each injection, 15.000 satellite cells from Pax3 GFP/+ mice, collected by flow cytometry (freshly isolated cells) or obtained by trypsination after 3 days of culture (cultured cells), were resuspended in 15 ml of PBS and injected into the Tibialis Anterior muscle with a 25 ml Hamilton syringe. Three or six weeks after transplantation, mice were sacrificed by cervical dislocation and the muscle analysed.
For EdU staining, the Click-iT EdU Cell Assay kit (Invitrogen) was used following the manufacturer's instructions with minor modifications: EdU was incubated at 10 mM for 15 min in cell culture medium. For detection, immunostaining for primary (Myogenin on cell culture and Pax7 on muscle sections) and secondary antibodies was performed, followed by the click chemical reaction using Alexa488 as a reactive fluorophore for detection, followed by DAPI staining.
Senescence associated bÀgalactosidase assays were performed as previously described (Dimri et al., 1995) on cultured cells, and on frozen muscle sections.

RNA extraction and RT-qPCR analysis
Cells were collected after culture or after flow cytometry in RLT buffer (Qiagen). Total RNA was extracted and purified after DNase treatment (Promega) using the RNAeasy Micro kit (Qiagen) and was reverse-transcribed into cDNA, following instructions in the Transcriptor cDNA synthesis kit (Roche). Quantitative real time PCR was performed on a StepOnePlus PCR machine (Applied Biosystems), using the FastStart Universal SYBR Green Master (Roche). Relative levels of expression in each assay were obtained through the DDCt method, and are expressed on a log scale. Hprt and RPLO transcripts levels were used for the normalizations of each target. As similar results were obtained with both, normalization has been done with Hprt. At least three biological replicates were used for each condition. Custom primers were designed using the Idt online software. Sequences are provided in Supplementary file 2. For each primer set, dissociation experiments were performed to ensure that no primer dimers or false amplicons would interfere with the results, and amplification efficiency was calculated using serial dilutions of total cDNA.

Quantification of mitochondrial DNA
Cells were collected after flow cytometry in extraction buffer (0.2 mg/ml proteinase K, 0.2% SDS and 5 mM EDTA) and incubated at 50˚C for 3 hr. DNA was precipitated with 0.3 M sodium acetate (pH 5.2) and isopropanol for 20 min on ice before centrifugation at 12.000 rpm at 4˚C. Quantitative real time PCR was performed on a StepOnePlus PCR machine (Applied Biosystems) using the FastStart Universal SYBR Green Master (Roche) and primers that amplify the mitochondrial COI gene and the nuclear encoded NDUFV1 gene as an endogenous reference (see Supplementary file 2).

Metabolic and bio-energetic analysis
H 2 O 2 production was quantified by the Amplex red assay (Life technology), which detects the accumulation of a fluorescent oxidized product.
Cytoplasmic reactive oxygen species were quantified by flow cytometry using the Deep Red Cell Rox probe (Invitrogen). For analysis of cells in culture, Cell Rox probe (0.5 mM) was directly added to the culture medium at the indicated time for 30 min in the cell culture incubator (at 37˚C and in the dark). After three rinses with PBS at 37˚C, cells were then rapidly trypsinised, fixed for 5 min in PFA, and rinsed three times in PBS. Cells were finally re-suspended in PBS and immediately analysed on a LSR Fortessa (BD Biosciences) apparatus, using FlowJo (TreeStar) software. For analysis of cells after flow cytometry, freshly sorted cells were re-suspended in 2 ml of F12 medium with 1% FCS at 37˚C and incubated for 30 min at 37˚C in a cell culture incubator in the presence of the Deep Red Cell Rox probe (0.5 mM). Cells were then rinsed in PBS at 37˚C, fixed for 5 min in PFA for the Cell Rox probe, rinsed 3 times in PBS, and re-suspended in PBS for analysis on a LSR Fortessa (BD Biosciences).
Mitochondrial reactive oxygen species were quantified using Mitosox probe (Invitrogen): freshly sorted cells were re-suspended in 2 ml of HBSS medium with 1% FCS at 37˚C and incubated for 20 min at 37˚C in a cell culture incubator in the presence of the Mitosox probe (0.5 mM). Cells were then rinsed in HBSS at 37˚C, and immediately analysed on a LSR Fortessa (BD Biosciences).
Mitochondrial mass was quantified using Mitotracker RedSox probe (Invitrogen): freshly sorted cells were re-suspended in 2 ml of F12 medium with 1% FCS at 37˚C and incubated for 30 min at 37˚C in a cell culture incubator in the presence of the Mitotracker RedSox probe (250 nM). Cells were then rinsed in PBS at 37˚C, and immediately analysed on a LSR Fortessa (BD Biosciences).
Cellular bioenergetics was analysed on a Seahorse extracellular flux analyser (XFe24) according to the manufacturer's instructions. Briefly, Pax3(GFP)-positive cells purified by flow cytometry were seeded at a density of 100.000 cells per well in a Seahorse XF24 plate coated with the cell immobilizer BD Cell Tak coating (BD Biosciences) and cultured for 2 to 4 days. The seeding density was based on initial assays that optimized the oxygen consumption rate (OCR) and the extracellular-acidification rate (ECAR). One hour before the assay, cultured cells were washed three times in minimal assay media (Seahorse Biosciences, 37˚C, pH 7.40) supplemented with glucose 2.25 g/L, 1 mM sodium pyruvate and 2 mM L-glutamine, and equilibrated for 1 hr at 37˚C in a non-CO 2 incubator. The basal and maximal rate of OCR and ECAR were measured successively in the control condition (basal rates) and after simultaneous injection of oligomycin (1 mM) that inhibits ATP synthase and of FCCP (0.5 mM) that uncouples mitochondrial OXPHOS (maximal rates). This dual injection allows measurement of the ability of cells to respond to increased energy demand using both oxidative phosphorylation (OCR) and glycolysis (ECAR). At the end of the experiment, proteins were extracted using 10 mL of lysis buffer and quantified using the Bradford assay. OCR and ECAR values were then normalized to protein concentration in order to compare different conditions.

Oxyblots
Irreversible protein oxidations were quantified by determining levels of protein carbonylation using the Protein OxyBlot kit (Millipore). Briefly, after protein extraction using lysis buffer, 10 mg of total proteins (in 5 mL) were denaturated by adding 5 mL of 12% SDS and boiling at 95˚C for 15 min. Derivatization was achieved by the addition of 5 mL DNPH (10 mM) and incubation at room temperature for 20 min. After neutralization by addition of 7.5 mL of neutralization solution, samples were loaded on 4-10% polyacrylamide gels. For controls, the same samples were treated as described except that a control solution without DNPH was used instead of the DNPH-containing derivatization solution. Proteins were blotted to a nitrocellulose membrane. After staining with red ponceau to quantifiy total proteins, carbonylated proteins were detected using the rabbit polyclonal anti-DNP primary antibody (1:100; Protein Oxyblot Kit) and the anti-rabbit peroxidase antibody. Detection was performed with ECL.

Statistical analysis
Statistical analysis was performed using GraphPad Prism software using the two-tailed unpaired t-test test and a minimum of 95% confidence interval for significance; P-values indicated on figures are <0.05 (*),<0.01(**) and <0.001 (***). Figures display mean values of all animals tested ±standard deviation (s.d), with all experiments being carried out on a minimum of 3 animals in 3 independent experiments.

Acknowledgments
This work was supported by the Institut Pasteur, the CNRS (UMR 3738 and UMR 8256) and Sorbonne Université s, with grants to MB, DM and BF from the AFM, grants to MB and DM from the ANR (REGSAT), the Laboratoire d'Excellence REVIVE (Investissement d'Avenir ANR-10-LABX-73) and the EU Optistem project (Health, FP7-2007, 223098) and grant to AL from the DIM Stem Pô le. A L'honoré was supported by a postdoctoral fellowship from the Fondation pour la Recherche Mé dicale (FRM) and a Marie Curie IRG grant (MC-IRG248496/SATELLITE CELL). G Pallafacchina was supported by a postdoctoral fellowship from the Fondation pour la Recherche Mé dicale. The authors thank C Bodin, C Cimper, N Benanteur and the laboratory of histology of the Institut Pasteur for technical help, A Lombes and F Devaux for reagents and advice on mitochondrial function, R Blaise for reagents and B Robert for critical reading of the manuscript and discussion. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Data availability All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for figures and supplement figures.

Author contributions
The following previously published datasets were used: