SOXF factors regulate murine satellite cell self-renewal and function through inhibition of β-catenin activity

Muscle satellite cells are the primary source of stem cells for postnatal skeletal muscle growth and regeneration. Understanding genetic control of satellite cell formation, maintenance, and acquisition of their stem cell properties is on-going, and we have identified SOXF (SOX7, SOX17, SOX18) transcriptional factors as being induced during satellite cell specification. We demonstrate that SOXF factors regulate satellite cell quiescence, self-renewal and differentiation. Moreover, ablation of Sox17 in the muscle lineage impairs postnatal muscle growth and regeneration. We further determine that activities of SOX7, SOX17 and SOX18 overlap during muscle regeneration, with SOXF transcriptional activity requisite. Finally, we show that SOXF factors also control satellite cell expansion and renewal by directly inhibiting the output of β-catenin activity, including inhibition of Ccnd1 and Axin2. Together, our findings identify a key regulatory function of SoxF genes in muscle stem cells via direct transcriptional control and interaction with canonical Wnt/β-catenin signaling.


Introduction
Maintenance, repair, and regeneration of adult tissues rely on a small population of stem cells, which are maintained by self-renewal and generate tissue-specific differentiated cell types (Weissman, 2000). Most adult stem cells are quiescent within their niche, dividing infrequently to generate both a copy of the stem cell and a rapidly cycling cell (Barker et al., 2010). These features make adult stem cells essential for either normal tissue homeostasis or repair/regeneration following damage (Slack, 2000). Hence, identification and manipulation of stem cells, including understanding mechanisms of cell fate decision and self-renewal, are essential to develop stem cell-based therapeutic strategies (Relaix, 2006).
Skeletal muscle contains a population of resident stem cells -termed satellite cells (Katz, 1961;Mauro, 1961). Around birth, fetal muscle progenitor cells adopt a satellite cell position, becoming embedded within the basal lamina in close contact to the muscle fibers (Ontell and Kozeka, 1984;Relaix et al., 2005). Importantly, during postnatal growth, the emerging satellite cells progressively enter quiescence, a molecular state poorly characterized in vivo. However, in response to injury or disruption of the basal lamina, satellite cells are activated and proliferate to form myoblasts that either fuse to existing myofibers to repair, or fuse together to form multinucleated de novo myotubes for regeneration. Alternatively, a subset of satellite cells self-renews to maintain a residual pool of quiescent stem cells that has the capability of supporting additional rounds of growth and regeneration (Zammit et al., 2006). Satellite cells are indispensable for muscle recovery after injury, confirming their pivotal and non-redundant role as skeletal muscle stem cells (reviewed in Relaix and Zammit, 2012).
Many studies have demonstrated a balance between extrinsic cues and intracellular signaling pathways to preserve stem cell function, with Notch and Wnt signaling being of particular importance (Brack and Rando, 2012;Dumont et al., 2015). Wnt signaling has been extensively studied in satellite cells (Brack et al., 2008;Kuang et al., 2008). Whereas canonical Wnt signaling, implying bcatenin/TCF activation, is upregulated upon muscle regeneration and regulates satellite cell differentiation (Otto et al., 2008;von Maltzahn et al., 2012), non-canonical Wnt signaling (independent of b-catenin), mediates satellite cell self-renewal and muscle fiber growth (Le Grand et al., 2009;von Maltzahn et al., 2012). However, how Wnt signaling pathways interact with intrinsic transcriptional regulators remains unclear. Therefore, identifying the transcriptomic changes in muscle progenitors and satellite cells through development, growth and maturity is fundamental in order to build a comprehensive model of satellite cell formation and function (Alonso-Martin et al., 2016). Focusing on the important transition from developmental to postnatal myogenesis, we identified the SOXF family (SOX7, SOX17, SOX18) as potentially having a pivotal role in muscle stem cell function (Alonso-Martin et al., 2016).
SOX factors belong to the high mobility group (HMG) superfamily of transcription factors (Bernard and Harley, 2010), and act in the specification of stem cells in a number of tissues during development (Irie et al., 2015;Lizama et al., 2015). SOX17 plays important roles in development, particularly in embryonic stem cells (Sarkar and Hochedlinger, 2013;Séguin et al., 2008) and endoderm formation (Hudson et al., 1997;Kanai et al., 1996), and is critical for spermatogenesis (Kanai et al., 1996) and specification of human primordial germ cell fate (Irie et al., 2015). SOX17 is also implicated in stem cell homeostasis in adult hematopoietic tissues and in cancer (Corada et al., 2013;He et al., 2011;Lange et al., 2009;Ye et al., 2011). SOX7 shares a role in endoderm formation with SOX17, and interestingly, genetic interaction of Sox7 with Sox17 has been recently reported in developmental angiogenesis (Kim et al., 2016;Shiozawa et al., 1996;Takash et al., 2001). Finally, loss of SOX18 leads to cardiovascular and hair follicle defects (Pennisi et al., 2000). Moreover, SOX18 together with SOX7 and SOX17 regulates vascular development in the mouse retina (Zhou et al., 2015).
While SoxF genes play key functions in different stem cell systems, little is known of their role in myogenesis. Here, using a set of ex vivo and in vivo experiments including genetic ablation and regeneration studies, we demonstrate that these factors regulate skeletal muscle stem cell selfrenewal as well as satellite cell-driven postnatal growth and muscle regeneration. Moreover, we show that SOXF factors operate via interaction with b-catenin in myogenic cells to modulate the output of Wnt canonical signaling during postnatal myogenesis.

SoxF gene expression parallels satellite cell emergence and promotes satellite cell self-renewal
To characterize the formation, establishment and maintenance of satellite cells, we performed a chronological global transcriptomic profiling in embryonic, fetal, and postnatal muscle progenitors and satellite cells (Alonso-Martin et al., 2016). These cells were prospectively isolated from a Pax3 GFP/+ population, with minimal contamination of endothelial cells, as previously reported (Alonso-Martin et al., 2016) (Figure 1-figure supplement 1). Focusing on establishment of satellite cells, we identified the SOXF family (SOX7, SOX17, SOX18) of transcriptional regulators as likely key regulators of satellite cell function.
Strikingly, SoxF genes are barely detectable during embryonic and fetal stages ( Figure 1A-B) but are induced at onset of the emergence of satellite cells and robustly expressed in postnatal satellite cells at the transcript and protein level ( Figure 1A-C).
To examine whether SOXF factors were present specifically in quiescent satellite cells, we performed primary culture experiments in proliferation and differentiation conditions. We isolated freshly FACS-sorted quiescent satellite cells and compared their expression profile to those undergoing culture ( Figure 1D). Whereas activation (Myod), proliferation (Ki67), and differentiation (Myog, Myh1) transcripts were all induced in culture conditions, SoxF were predominately detectable in quiescent (Pax7) satellite cells ( Figure 1D).
To characterize the role of SOXF factors in satellite cell function, we used the myofiber culture model, which maintains a functional niche for skeletal muscle stem cells while allowing their  (Sox7,Sox17,Sox18) in FACS-isolated Pax3 GFP/+ cells from Affymetrix expression analysis (A) and RT-qPCR (B). E, Embryonic day; P, Postnatal day; MO, age in months. (C) Representative immunolabeling of a satellite cell (PAX7+) co-expressing SOX17 on a freshly isolated adult myofiber (T0). Scale bar, 10 mm. Nuclei are counterstained with DAPI. (D) Expression profile of fresh FACS-sorted and cultured satellite cells for quiescence (Pax7), activation/ commitment (Myod, Myog), proliferation (Ki67), terminal differentiation (Myh1), and for SoxF (Sox7, Sox17, Sox18) transcripts. Quiesc., quiescence; Prolif., proliferation; Diff., differentiation conditions. n = 3 mice (each quantified in triplicate) for all experiments. Data expressed as mean ± s.e.m. DOI: https://doi.org/10.7554/eLife.26039.002 The following figure supplement is available for figure 1: observation (Zammit et al., 2004). We generated retroviruses encoding a bi-cistronic expression for full-length SOX7FL, SOX17FL or SOX18FL, or transactivation defective SOX7DCt, SOX17DCt or SOX18DCt proteins (Figure 2-figure supplement 1A), together with GFP to identify transduced cells. As SOXF proteins share the same consensus DNA binding sequence, any SOXFDCt is expected to behave as a dominant negative for all three transcription factors (Hou et al., 2017). Retrovirus encoding IRES-GFP only was used as a control (CTRL). Overexpression of any of the SoxF genes (SOXF-FL) induced a similar phenotype in satellite cells, increasing the pool of self-renewing satellite cells (PAX7+/GFP+) ( This SOXF overexpression in satellite cells parallels the effects observed in other stem cell types, such as adult hematopoietic progenitors (He et al., 2011). Conversely, expression of transactivation defective SOXFDCt caused a decrease in self-renewal (PAX7+/GFP+) (

SOX17 is required for satellite cell quiescence and myofiber maturation
Considering the important role of SOX17 in cell stemness and cell fate decisions (Chhabra and Mikkola, 2011;Irie et al., 2015;McDonald et al., 2014), we chose to investigate its function in postnatal skeletal muscle satellite cells in vivo. Since Sox17 mutant mice die during development (Kim et al., 2007), we combined a null Sox17 reporter allele (Sox17 GFP ) with a conditional Sox17 fl allele to perform tissue-specific genetic ablation of Sox17: intercrossing with Pax3 Cre/+ mice to achieve lineage-specific Sox17 deletion during development and consequently postnatally, or Pax7-CreERT2/+ mice for an inducible adult satellite-cell-specific deletion. Pax3 Cre/+ ;Sox17 GFP/fl mutant mice had no obvious differences in body or muscle weight during postnatal growth or in adulthood (Figure 3-figure supplement 1A-C). Yet, Sox17-knockout Soleus muscle in adult Pax3 Cre/+ ;Sox17 GFP/fl mice contained more myofibers, but with reduced cross-sectional area ( Figure 3A-D). Myofibers from Pax3 Cre/+ ;Sox17 GFP/fl Soleus also had a lower myonuclei density ( Figure 3E), suggesting that Sox17-deficient muscles have less satellite cells contributing to postnatal muscle growth (White et al., 2010;Yin et al., 2013;Zammit, 2008). Indeed, direct quantification using PAX7 or MCAD immunolabeling, including reduction of Pax7 transcripts, revealed that there were fewer satellite cells in Pax3 Cre/+ ;Sox17 GFP/fl muscles ( Figure 4A,B,D and Figure 4-figure supplement 1). Interestingly, this reduction was already evident by two weeks of postnatal growth ( Figure 4B), a time when a significant proportion of satellite cells are becoming quiescent, forming the pool of adult muscle stem cells. Finally, consistent with our myofiber culture experiments (Figure 2), we found that the decrease in muscle stem cells in Sox17-knockout mice was associated with a striking decrease of quiescent cells ( Figure 4C). Instead, an increased proportion of satellite cells expressed PAX7 and MYOD (18.3% vs. 3.4% in controls) in Sox17-knockout mutants, and thus were activated, and 16.8% even expressed just MYOD (compared to 2.4% in controls), indicating that they were potentially entering the differentiation program ( Figure 4C).
Conditional knockout of Sox17 specifically in adult satellite cells caused a similar loss of satellite cells as soon as three weeks after tamoxifen injection in Pax7 CreERT2/+ ;Sox17 fl/fl mutant mice ( Figure 4E-G). Myofiber content and morphology was not affected in satellite-cell-specific Sox17conditional knockout (Pax7 CreERT2/+ ;Sox17 fl/fl ) adult mutant mice though ( Figure 3-figure supplement 2), suggesting that the phenotype in Pax3 Cre/+ ;Sox17 GFP/fl mice was linked to impaired early postnatal growth and satellite cell-derived myonuclear accretion (White et al., 2010). These results demonstrate that SOX17 plays an important role in induction and maintenance of satellite cell quiescence.     Myogenic stem cell function is impaired during muscle regeneration in Sox17-deficient mice To evaluate the role of SOX17 during satellite cell activation, renewal and differentiation in vivo, we carried out skeletal muscle regeneration assays. Following cardiotoxin (CTX)-induced regeneration in Tibialis anterior (TA) muscle of wild type mice, we first assessed the dynamics of SoxF gene expression by RT-qPCR in total injured muscle. We observed progressive up-regulation of SoxF genes, with distinct peaks at days (d) 4, 6, and 15 following injury (  (Rocheteau et al., 2012) through muscle regeneration depicts an identical behavior of all SoxF transcripts, being downregulated upon injury, and induced as regeneration proceeds ( Figure 5A). SoxF genes and Pax7 display a similar profile, contrary to commitment and differentiation markers (Myod and Myog, Figure 5-figure supplement 1D), inferring that SOXF have stem cell specific activity during regenerative myogenesis ( Figure 5A).
Regenerating TA muscles in Pax3 Cre/+ ;Sox17 GFP/fl mice were strikingly smaller than controls and expressed lower levels of myogenic genes ( Figure 5B-C). Furthermore, we observed a loss of quiescence in Sox17-knockout satellite cells after muscle regeneration, likely preventing cells from reestablishing the pool of quiescent satellite cells ( Figure 5D-E) so that when regeneration was over, the satellite cell pool was smaller in Sox17-knockout mutants ( Figure 5F). Interestingly, when plating fresh FACS-sorted isolated satellite cells in vitro, Sox17-knockout cells proliferated more than control cells, yielding bigger colonies (Figure 5-figure supplement 2A-B). This result mimicked the effect obtained in satellite cells transduced with SOXFDCt, with increased satellite cell proliferation at the expense of self-renewal ( Figure 2). Histological analysis of TA muscles in Pax3 Cre/+ ;Sox17 GFP/fl mice at d7 after CTX-induced regeneration revealed cell infiltration, fat accumulation and fibrosis, that were absent in regenerating muscles of control Sox17 GFP/fl mice ( Figure 5G), suggesting abnormal regeneration and impaired satellite cell function (Mann et al., 2011;Sambasivan et al., 2011). Moreover, this delay in regeneration was still observed at d28, with signs of cell infiltration still evident ( Figure  To confirm that muscle regeneration defect in Pax3 Cre/+ ;Sox17 GFP/fl mice was due to satellite cell function compromised by loss of SOX17, we also examined regeneration in TA muscles of Pax7-CreERT2/+ ;Sox17 fl/fl mice ( Figure 6A). Analysis of regeneration at d7 in Sox17-conditional knockout mutants revealed that satellite cell numbers were reduced, with fewer in quiescence ( Figure 6B-D). At d28, diminution of the satellite cell pool was confirmed in regenerating muscle of adult conditional Pax7 CreERT2/+ ;Sox17 fl/fl mutant mice ( Figure 6E-G) as observed with Pax3 Cre/+ ;Sox17 GFP/fl mice. Again, consistent with the phenotype in Pax3 Cre/+ ;Sox17 GFP/fl mice, histological analysis of regenerated Sox17-conditional knockout TA muscles revealed cell infiltration, fat and fibrosis deposition, that were absent in regenerating muscles of control Sox17 fl/fl mice ( Figure 6H-L), confirming abnormal regeneration and impaired satellite cell function in the absence of SOX17.

Impaired SOXF function leads to severe muscle regeneration defects
Both myofiber culture and in vivo experiments suggested that SOXF factors are involved in satellite cell self-renewal. Alterations of SoxF gene function in myofiber culture experiments yielded stronger phenotypes than in vivo genetic ablation of just Sox17, suggesting a compensatory mechanism between SOX17 and other SOXF proteins. To study such a possible compensatory effect between SOXF members, we performed myofiber culture experiments in control Sox17 GFP/fl and Pax3 Cre/+ ; Sox17 GFP/fl mutant mice, and analyzed the effect of expressing each of the SOXF factors  Figure 7C, CTRL vs. KO). Interestingly, transduction with retrovirus encoding either SOX7 or SOX17 rescued this defect in self-renewal, whereas expression of SOX18 was unable to revert this effect ( Figure 7A). Moreover, overexpression of SOX7 or SOX17 strongly decreased the number of activated satellite cells, to even lower levels compared to control animals ( Figure 7B). Expression of SOX18, however, did not modify the activation status of the cells. Finally, overexpression of each SOXF proteins induced a strong decrease in differentiation ( Figure 7C), as previously observed in wild type cells (Figure 2-figure supplement 1C-F). These results demonstrate that overexpression of SOX7 or SOX17, but not SOX18, rescues the quiescence and activation phenotype of Sox17knockout satellite cells.
To further characterize the redundant activity of SoxF genes in vivo, we took advantage of the dominant negative effect of SOX17DCt ( Figure 2) to carry out electroporation into regenerating muscle ( Figure 7D-F). Two days after CTX injection of wild type TA muscles, we electroporated a bi-cistronic construct co-expressing SOX17DCt and GFP ( Figure 7D and Figure 7-figure supplement 1), together with a TdTomato reporter that revealed efficient electroporation along the regenerating muscle (Figure 7-figure supplement 1). Post-electroporation, we observed many areas of regenerating muscle devoid of fibers, with accumulation of fat and fibrosis, compared to control, indicating a general failure of muscles to regenerate ( Figure 7E). A dramatic reduction in Pax7 expression was associated with the exacerbated phenotype of SOX17DCt electroporated into muscle, compared to regeneration in Sox17-knockouts (Figures 5C,G and and 7E-F). These results are consistent with SOXF activity being required for skeletal muscle regeneration and confirm the overlapping role of SOXF members, as previously reported in other tissues (Matsui et al., 2006;Sakamoto et al., 2007;Sarkar and Hochedlinger, 2013).
Inhibition of b-catenin activity by SOXF factors in muscle stem cells SOXF and b-catenin (CTNNB1) interact through a site located in the C-terminus of SOXF proteins (Figure 2-figure supplement 1A) and that deletion of this region is sufficient to ablate SOXF -bcatenin interaction (Guo et al., 2008;Sinner et al., 2007;Sinner et al., 2004;Zhang et al., 2005). Moreover, expression of constitutively active b-catenin in satellite cells in vivo leads to reduced myofiber size (Hutcheson et al., 2009;Kuroda et al., 2013), a phenotype similar to that we observe with the ablation of SOX17 in these cells (Figure 3). This suggests that SOXF inhibition of b-catenin activity could be required for muscle homeostasis. Upon activation of Wnt signaling, non-phosphorylated b-catenin is stabilized and translocates to the nucleus where it associates with TCF/LEF transcription factors to regulate target gene expression (MacDonald et al., 2009).
We designed two transcriptional reporter assays in C2C12 myoblasts to further characterize the SOXF -b-catenin interaction following b-catenin canonical signaling activation by LiCl ( Figure 8A-B). All SOXF proteins individually, strongly activated our novel SoxF reporter, SoxF-B-TKnLacZ (containing five multimerized SOXF consensus binding motifs), demonstrating binding to the same consensus sequence ( Figure 8A). Upon b-catenin co-expression with SOXF proteins, SoxF-B-TKnLacZ transactivation was further increased ( Figure 8A). Conversely, we explored the role of SOXF proteins on LEF/TCF-b-catenin transcriptional activity ( Figure 8B). In this system, b-catenin expression led to a four-fold increase in b-catenin reporter pTOP-TKnLacZ activity, while co-expression of SOXF impaired b-catenin-mediated induction of this reporter ( Figure 8B). These functional assays indicate that while b-catenin enhances the transactivation activity of SOXF members, SOXF proteins hinder bcatenin-mediated activation of a TCF/LEF reporter in myogenic cells. Hence, our results imply that  SoxF genes modulate b-catenin signaling during myogenesis. Strikingly, expression levels of known target genes of the canonical b-catenin pathway appear modified in Sox17-knockout muscles ( Figure 8C). Indeed, Jun, Ccnd1, and Axin2 expression were all increased two-to ten-fold in Sox17 mutant muscles ( Figure 8C).
In agreement with previous reports (Otto et al., 2008;Rudolf et al., 2016), we observed nuclear b-catenin expression in activated, but not quiescent, satellite cells indicating that induction of canonical signaling is synchronous with the activation of satellite cells ( Figure 8D). To assess the functional significance of b-catenin binding to SOXF proteins, retroviral constructs of SOXF lacking b-catenin binding domain (SOXFDBCAT) were generated (Figure 2-figure supplement 1A). Expression of SOXFDBCAT in wild type satellite cells ex vivo caused a significant decrease in self-renewal capacity and increased activation ( Figure 8E

SOXF factors modulate b-catenin transcriptional activity in satellite cells
To further demonstrate the functional interplay between SOX17 and b-catenin transcriptional activity in myogenic stem cells, single myofiber-associated satellite cells were treated with LiCl. This induction of b-catenin signaling yielded an expansion of the activated satellite cell pool (CTRL, Figure 9A). Overexpression of Sox17 (SOX17FL) abolished the expansion of satellite cells ( Figure 9A), while SOX17DCt did not affect the enhanced LiCl-driven expansion. Similar results were obtained when using CHIR9902, a specific inhibitor of the Glycogen synthase kinase-3 (GSK3B), which targets b-catenin for degradation (data not shown) (Ying et al., 2008). Our findings point to modulation of cell cycle by SOXF activity: satellite cells fail to acquire quiescence when SOXF function is impaired in vivo and ex vivo. In accord with these observations, the cell cycle regulator Ccnd1 (Cyclin-D1) was up-regulated in Sox17-knockout satellite cells but absent in wild type cells ( Figure 8C and Figure 9B). We next investigated how SOXF proteins affect the b-catenin transcriptional regulation of two target genes found increased in Sox17-knockouts, Ccnd1 [also a SOX17 target (Lange et al., 2009)] and Axin2. We designed a cell-based transcriptional reporter assay using either 1 kb of the 5'UTR of Ccnd1 (Ccnd1-nLacZ), encompassing binding motifs for TCF/LEF and SOXF proteins, or 5.6 kb of the proximal Axin2 promoter (Axin2-nLacZ) ( Figure 9C-D). b-catenin expression increased activity of both Ccnd1-nLacZ and Axin2-nLacZ reporters following LiCl treatment, while co-expression of SOX17 impaired b-catenin-mediated induction of these two reporters in a dose-dependent manner ( Figure 9C-D). SOX7DBCAT, lacking the b-catenin binding site, however, was unable to influence activation of either the Ccnd1-nLacZ or Axin2-nLacZ reporters. Accordingly, Axin2 expression levels appeared to be progressively down-regulated at the onset of satellite cells emergence, thus displaying general inverse dynamics to SoxF genes ( Figure 9E) (Alonso-Martin et al., 2016).
Together, our data demonstrate that SOXF factors control expansion and self-renewal of adult muscle stem cells, associated with an inhibition of TCF/LEF-b-catenin target genes.

Discussion
We previously performed a global transcriptomic analysis of the changes in gene expression in murine muscle stem cells throughout life (Alonso-Martin et al., 2016). Focusing on the signature associated with establishment and maintenance of satellite cells from their developmental progenitors, we identified SoxF genes, Sox7, Sox17, and Sox18 as of interest. SoxF transcripts become expressed at the time of satellite cell emergence, with a maximum expression in the quiescent adult state, highlighting their role in establishment, maintenance and function of muscle stem cells. Of relevance, SOX17 is involved in cell fate decisions in human primordial germ cells and embryo-derived stem cells (Irie et al., 2015;McDonald et al., 2014).
Absence of SOX17 leads to impaired postnatal muscle development, with an increase of smaller fibers. Postnatal muscle fiber hypertrophy depends on the total number of muscle fibers within a muscle; thus, the postnatal growth rate of the individual muscle fiber would be lower when there are more myofibers (Rehfeldt et al., 2000). In addition, the reduction of myonuclei per myofiber suggests that myofiber growth impairment may be due to a reduced contribution of satellite cell fusion (White et al., 2010). Consistent with these findings, we observed fewer satellite cells in Sox17knockout mice, associated with a loss of quiescence and a reduced stem cell pool in postnatal muscles. Moreover, when SOXF function is impaired in satellite cells, self-renewal capacity is reduced and both activation and proliferation are increased. Satellite cell self-renewal is critical to maintain the pool of the satellite cells, so impairment of this process translates into reduced cell numbers, resulting in defective muscle regeneration in both Pax3 Cre/+ ;Sox17 GFP/fl and Pax7 CreERT2/+ ;Sox17 fl/fl mutant mice, highlighting the specific relevance of SoxF genes postnatally and specifically in adult satellite cells. Moreover, we show that SOXF overexpression in satellite cells inhibits proliferation and differentiation and promotes self-renewal, with SOX17 promoting self-renewal in other stem cell types, such as adult hematopoietic progenitors (Chhabra and Mikkola, 2011;He et al., 2011).
Specific genetic ablation of Sox17 leads to milder phenotypes than when dominant negative constructs are used, which suppress transcriptional activation through all SOXF proteins, in myofiber cultures (ex vivo) or injured muscle electroporation (in vivo). Yet, despite apparently normal expression of Sox7 and Sox18 in Sox17 mutant mice ( Figure 4D), there is a general loss of quiescence in satellite cells. SoxF genes have been reported to act with redundant functions, as versatile regulators of embryonic development and determination of different stem and progenitor cell fate (Matsui et al., 2006;Sakamoto et al., 2007;Sarkar and Hochedlinger, 2013). However, our data suggest that in muscle stem cells, redundancy between SoxF genes is more complex. For instance, overexpression of SOX7 or SOX17 but not SOX18 is sufficient to rescue the phenotype in Sox17 mutant mice. Recently, a Sox7 fl mutant mouse has been reported, revealing the genetic interaction of SOX7 with SOX17 in developmental angiogenesis (Kim et al., 2016). Furthermore, during revisions for this study, a muscle-specific ablation of Sox7 (Pax3 Cre/+ ;Sox7 fl/fl ) was reported, showing upregulation of Sox17 and Sox18 in the absence of Sox7 (Rajgara et al., 2017). Nevertheless, Sox7-deficient muscles demonstrated severe phenotypes in homeostatic and regeneration conditions (Rajgara et al., 2017), similar to Sox17 ablation in myogenic cells (Figures 3-6). Future studies analyzing the impact of ablating both SOX7 and SOX17 for muscle stem cell function will be of interest.
Finally, our data link SOXF regulation of satellite cell self-renewal with control of b-catenin activity in satellite cells. Interaction between SOXF and b-catenin has been reported in other cell types, i.e. repression of b-catenin-stimulated expression of dorsal genes (Zorn et al., 1999), regulation of endodermal genes (Sinner et al., 2004), or acting as tumor suppressors antagonizing Wnt/b-catenin signaling (Liu et al., 2016;Sinner et al., 2007;Takash et al., 2001), as well as regulators of this pathway in oligodendrocyte progenitor cells (Chew et al., 2011;Ming et al., 2013). More importantly, our data provide a molecular mechanism for previous reports which demonstrate that a tight regulation of the Wnt/b-catenin canonical signaling output is required to ensure skeletal muscle regeneration (Brack et al., 2008;Brack et al., 2007;Figeac and Zammit, 2015;Murphy et al., 2014;Otto et al., 2008;Parisi et al., 2015;Rudolf et al., 2016;Seale et al., 2003;von Maltzahn et al., 2012). Hence, SOXF factors display a dual activity as both intrinsic regulators of muscle stem cell quiescence and interacting with extrinsic signaling pathways to regulate the expansion of activated muscle stem cells. Moreover, recent findings demonstrate that old satellite cells are incapable of maintaining their normal quiescent state in muscle homeostatic conditions, by switching to an irreversible pre-senescence state (Sousa-Victor et al., 2014). Satellite cells fail to regulate their quiescence with aging, leading to depletion of the pool of stem cells (Blau et al., 2015). Interestingly, satellite cell functional impairment is associated with up-regulation of canonical Wnt/b-catenin (Brack et al., 2008;Brack et al., 2007). Our data therefore points to a potential role of SOXF-b-catenin interaction in this context.
In conclusion, we demonstrate that SOXF transcription factors play a key role in stem cell quiescence and myogenesis through both direct transcriptional control and by modulation of the output of b-catenin activity to affect canonical Wnt signaling.

Cell sorting and culture
For FACS, muscle samples were isolated from adult mice (forelimb, hindlimb, and trunk muscles). isolated from either Pax3 GFP/+ or Tg:Pax7-nGFP were obtained using the FITC channel to recover the GFP+ population.

RNA preparation and quantitative PCR
Total RNA from FACS-sorted satellite cells was extracted from independent experiments according to the RNasy Micro Kit (QIAGEN, Courtaboeuf, France) RNA extraction protocol. For whole muscle total RNA, RNeasy Fibrous Tissue Midi Kit (QIAGEN, Courtaboeuf, France) was used. cDNA synthesis was performed using Transcriptor First Strand cDNA Synthesis Kit (Roche-Sigma-Aldrich, St. Quentin Fallavier, France). RNA quality was assessed by spectrophotometry (Nanodrop ND-1000).
qPCR reactions were carried out in triplicate using LightCycler 480 SYBR Green I Master (Roche-Sigma-Aldrich, St. Quentin Fallavier, France). Expression of each gene was normalized to that of Hypoxanthine Phosphoribosyltransferase 1 (Hprt1) for total muscle, or TATA Box Protein (TBP) for cultured cells. Results are given as mean ± standard error. The single (*), double (**), triple (***), and quadruple (****) asterisks represent p-values p<0.05, p<0.01 and p<0.001, respectively, for Student's unpaired t-test. The oligonucleotides used in this study are listed in table 1.

Immunolabeling, microscopy and image treatment
Muscles were dissected and snap-frozen in liquid nitrogen-cooled isopentane. Eight mm cryosections were fixed in 4% paraformaldehyde (PFA) and immunofluorescence was carried out as previously described (Mitchell et al., 2010). Analysis was carried out using either a Leica TCS SPE confocal microscope or a Zeiss AxioImager. Z1 ApoTome (for scanning of whole Soleus cryosections). Images were processed with either Adobe Photoshop CS5 software (Adobe Systems) or MetaMorph 7.5 Software (Molecular Devices). Counting was performed using ImageJ (version 1.47 v; National Institutes of Health, USA, https://imagej. nih.gov/ij/). Transduced satellite cells in myofiber cultures were directly counted under a Leica fluorescent microscope at 40x magnification. Mean ± standard error (s.e.m.) was given. The single (*), double (**), and triple (***) asterisks represent p-values p<0.05, p<0.01, and p<0.001 respectively by Student's unpaired t-test. All experiments have been performed on at least three independent experiments for each condition. For the characterization of Sox17 mutant mice, 2-5 whole scanned cryosections in at least three different animals (controls and mutants) were analyzed.

Single myofiber isolation, culture and transduction
Single myofiber procedure was performed as previously described (Moyle and Zammit, 2014). Briefly, both Extensor digitorum longus (EDL) muscles were dissected and digested in Collagenase type I (Sigma-Aldrich, St. Quentin Fallavier, France) solution for 1.5 hr. Flushing medium against the digested muscle, myofibers detached from whole muscle and were placed into another culture dish. Fibers were taken at different time points, freshly isolated (T0), and 24 (T24), 48 (T48), and 72 (T72) hours after culture in activation medium [DMEM High Glucose (Life Technologies, Saint-Aubin, France), 10% horse serum (Life Technologies, Saint-Aubin, France) and 0.5% chicken embryo extract (MP-Biomedical, Illkirch-Graffenstaden, France)] at 37˚C in 5% CO 2 . Retroviral expression vectors and transduction were carried out as previously reported (Zammit et al., 2006). To transduce myofiber-associated satellite cells, 1:10 dilution of the retroviral supernatant was used 24 hr after fiber isolation. Satellite cells were transduced for 48 hr and then recovered for fixation and immunostaining. EdU (2 mM; C10340, Thermo Fisher Scientific, Montigny-le-Bretonneux, France) chase was performed for 72 hr (last 48 hr together with retroviral transduction). EdU-incorporating cells were detected according to the manufacturer's protocol.

Retroviral cloning
Sox7 and Sox18 cDNAs were amplified by PCR from IMAGE clones 40131228 and 3967084 respectively; Sox17 cDNA was cloned by PCR from mouse kidney cDNA (gift of Dr. J. Hadchouel). All were subcloned in pCig mammalian bi-cistronic expression vector and pMSCV-IRES-eGFP (MIG) retroviral packaging vector using XhoI and EcoRI added to cloning primers (Megason and McMahon, 2002;Pear et al., 1998).

Genes Sequences
Pax7 5' -AGGCCTTCGAGAGGACCCAC -3' 5' -CTGAACCAGACCTGGACGCG -3' profile, all days from day 0 up to day 7, and then days 10, 15, 21, and 28. Second injury was performed as above, 28 days after first injury. Muscle electroporation was performed using an Electro Square-Porator ECM 830 (BTX Ò , Genetronics Inc., Holliston, MA). According to (Sousa-Victor et al., 2014), 40 mg of DNA solutions were injected and TA muscles were electroporated using external plate electrodes two days after CTX injection. TAs were examined five, seven, or ten days later. Seven and 28 days after injury, TA muscles were processed for histology analysis by Hematoxylin and eosin , Oil red O, and Sirius red staining as previously described (Sambasivan et al., 2011).