Overexpression of facioscapulohumeral muscular dystrophy region gene 1 causes primary defects in myogenic stem cells

Facioscapulohumeral muscular dystrophy (FSHD, OMIM 158900) is the third most common muscular dystrophy, exhibits autosomal dominant inheritance and has no cure (Cabianca and Gabellini, 2010). FSHD typically arises with a reduction of facial and shoulder girdle muscle mass. The disease may extend to abdominal and pelvic girdle muscles impairing the ability to walk. Although FSHD is primarily a disease of skeletal muscle, up to 75% of FSHD patients also present vascular defects (Fitzsimons et al., 1987; Osborne et al., 2007; Padberg et al., 1995).

FSHD differs from classical muscular dystrophies for several aspects. While in many myopathies sarcolemmal disruption is the primary pathogenetic mechanism (Dalkilic and Kunkel, 2003; Durbeej and Campbell, 2002), in FSHD patients there is no evidence for alteration of sarcolemmal integrity (Orrell et al., 1999) or mitochondrial involvement (Kilmer et al., 1995) and the mechanism responsible for the disease is currently unclear.

FSHD is associated with reduction in the copy number of a macrosatellite repeat, called D4Z4, located at the subtelomeric region of chromosome 4 long arm, in 4q35 (Wijmenga et al., 1992). The number of repeats varies between 11 and 100 in healthy individuals, while FSHD patients carry 1 to 10 repeats (van Deutekom et al., 1993). The contraction of the D4Z4 repeat array causes a Polycomb/Trithorax epigenetic switch leading to the overexpression of several genes within the FSHD region (Cabianca et al., 2012; Gabellini et al., 2002; Lemmers et al., 2010; Snider et al., 2009; Snider et al., 2010). The peculiar nature of the mutation at the basis of FSHD and its complex effect on chromatin surrounding the 4q35 region makes it highly unlikely that the root cause of the disease can be attributed to a single gene. Since expression of multiple genes is affected, the molecular pathogenesis of FSHD has been challenging to untangle, and as yet no therapy is available for FSHD patients. The two most important FSHD candidate genes are the D4Z4 repeat gene double homeobox 4 (DUX4) (Lemmers et al., 2010; Snider et al., 2009; Snider et al., 2010) and the proximal gene FSHD region gene 1 (FRG1) (Gabellini et al., 2002). Nevertheless, for both DUX4 (Jones et al., 2012; Tsumagari et al., 2011) and FRG1 (Klooster et al., 2009; Masny et al., 2010) significant controversy exists regarding their actual overexpression and their role in the disease. For this reason, the potential role of 4q35 gene overexpression in the disease has been investigated at the functional level. DUX4 transgenic mice do not show any obvious muscle phenotype, despite displaying a DUX4 expression pattern and an alteration of DUX4 target genes similar to FSHD patients (Krom et al., 2013). On the contrary, mice overexpressing FRG1 (FRG1 mice), selectively in the skeletal muscle, display reduced muscle size and develop a muscular dystrophy resembling FSHD (Gabellini et al., 2006). Moreover, studies conducted in Xenopus laevis and Caenorhabditis elegans demonstrated that frg1 is required for normal muscle development and its overexpression causes muscular defects and vascular abnormalities correlated with the clinical findings from FSHD patients (Hanel et al., 2009; Liu et al., 2010; Wuebbles et al., 2009). Collectively, these results suggest that FRG1 is important for muscle function and its aberrant expression could contribute to the FSHD pathogenesis.

FRG1 is a dynamic nuclear and cytoplasmic shuttling protein that, in skeletal muscle, is also localized to the sarcomere (Hanel et al., 2011). Interestingly, overexpressed FRG1 is almost completely nuclear and is localized in nucleoli, Cajal bodies, and actively transcribed chromatin (Sun et al., 2011; van Koningsbruggen et al., 2004). Although, it has been associated with RNA biology (Gabellini et al., 2006; Sun et al., 2011; van Koningsbruggen et al., 2004; van Koningsbruggen et al., 2007), the molecular and cellular mechanism that follows FRG1 overexpression leading to muscular dystrophy is currently unknown.

To address this point, we monitored the muscle pathology of the FRG1 mouse during its life from 3 weeks (no sign of disease) to 14 weeks of age (full development of the dystrophic phenotype). We found that the onset of the phenotype is at 4 weeks of age when FRG1 mice display reduced myofiber size. Muscle stem cell ex vivo and in vivo experiments and muscle regeneration assays indicated that myogenic stem/progenitor cells from FRG1 mice exhibit defects in activation, proliferative, clonogenic, differentiation and regenerative potential, suggesting that these defects contribute to FRG1 mouse pathology. Next, we performed a gene expression profiling that identified networks and genes affected by FRG1 overexpression that could explain the abovementioned tissue and muscle stem cell defects.

In order to investigate the progressive course of the disease, we sacrificed FRG1 and wild-type (WT) mice from 3 weeks (no phenotype) to 14 weeks of age (clear dystrophic signs) and cryosectioned their vastus muscles. Morphometric analysis of Gomori-trichrome stained vastus muscles revealed that the first alteration in FRG1 mice appears at 4 weeks of age with a reduction in the myofiber cross-sectional area (CSA) compared to WT controls (Fig. 1A; unpaired t-test, P = 0.0199, n = 3). Dystrophic symptoms, like infiltration of inflammatory cells and regeneration became markedly visible between 6 and 8 weeks of age. Flow cytometric quantification revealed an increase in the percentage of CD45⁺ cells in FRG1 muscle-cell preparations, though the total number of muscle cells isolated was not altered (Fig. 1B; unpaired t-test for 7 weeks, P = 0.0105, n = 3 and supplementary material Table S1). Regeneration was assayed by quantification of centrally nucleated myofibers in transversally cryosectioned vastus muscles (Fig. 1C,D; paired t-test for 8 weeks, P = 0.0129, n = 6) and regenerating myofibers were also confirmed using immunofluorescence for developmental myosin heavy chain (MHCd) (see supplementary material Fig. S1; unpaired t-test, P = 0.0453, n = 3). Other muscular dystrophy signs like necrosis, fibrosis and fat deposition were only evident at 14 weeks of age (supplementary material Fig. S1).

As shown in Fig. 1A, the myofiber CSA of WT mice rapidly increased between 3 and 8 weeks of age as expected. Surprisingly, the myofiber CSA of FRG1 mice remained almost the same (Fig. 1A). To further study this feature, longitudinal cryosectioning of vastus from 4-week-old mice and quantification of the number of myonuclei per myofiber was performed. FRG1 mice displayed a 40% reduction in the number of myonuclei compared to control mice (Fig. 1E; unpaired t-test, P = 0.0122, n = 3), while the number of myofibers was not altered (Fig. 1F). Hence, FRG1 mice suffer from a post-natal muscle growth defect.

Although the FRG1 transgene is under the human skeletal actin (HSA) promoter and thus it should be expressed solely in mature muscle fibers, Chen and colleagues recently reported that FRG1 overexpression also occurs in primary myoblasts, causing a proliferative defect (Chen et al., 2011). Prompted by this observation, we considered that a defect in satellite cells, the myogenic stem cells responsible for postnatal growth and muscle repair (Bentzinger et al., 2012), could contribute to the muscle growth defect of FRG1 mice. Accordingly, we found that the FRG1 transgene was expressed at similar levels in vastus muscle and in quiescent satellite cells, freshly isolated by flow-cytometry as SM/C-2.6⁺ cells (Fukada et al., 2004) (see supplementary material Fig. S2). Thus, in FRG1 mice the transgene is aberrantly overexpressed in satellite cells opening the possibility to investigate the FRG1 role in this important muscle compartment. We analyzed 2-week-old animals, in which the proliferation of satellite cells contributes to post-natal muscle growth by increasing the number of myonuclei (Bentzinger et al., 2012; White et al., 2010). Immunofluorescence for the paired box gene 7 (Pax7), the most important marker of quiescent satellite cells (Seale et al., 2000), and the Ki67 proliferation marker on transverse cryosections showed that the percentage of Ki67⁺ proliferating cells within the Pax7⁺ cell population was lower in FRG1 (see supplementary material Fig. S3) than WT muscles (paired t-test, P = 0.0271, n = 3). Hence, a satellite cell impairment is present in FRG1 mice well before the development of any histological sign of dystrophy suggesting a role for satellite cells in the muscle growth defect of FRG1 mice.

Next, we decided to repeat this analysis ex vivo. To exclude that our observations were a secondary effect of muscle degeneration, we analyzed mice at 4 weeks of age, when no dystrophic symptoms other than a reduced myofiber size are manifested. Immunofluorescence on muscle cells cultured for 24 hours, revealed a reduced percentage of Ki67⁺ within the Pax7⁺ cell population in FRG1 cultures compared to WT controls (Fig. 2A,B; unpaired t-test, P = 0.0427, n = 4). Upon isolation, one of the first steps in satellite-cell activation and exit from quiescence is the expression of MyoD (myogenic differentiation-1), which controls myogenic commitment of proliferating cells (Chargé and Rudnicki, 2004). Exit from quiescence is also associated with downregulation of the quiescence marker caveolin-1 (Cav1) (Gnocchi et al., 2009; Volonte et al., 2005). Therefore, we monitored the temporal expression of Pax7, MyoD and Cav1. Interestingly, 24 hours after isolation the percentage of MyoD⁺ cells within the Pax7⁺ cell population was lower in FRG1 than WT cultures (Fig. 2C,D; unpaired t-test, P = 0.0029, n = 3); while, the percentage of Cav1⁺ quiescent cells within the Pax7⁺ cell population was higher in FRG1 cultures compared to WT controls (Fig. 2E,F; unpaired t-test, P = 0.0003, n = 4). Notably, the percentage of Pax7⁺ cells in FRG1 cell cultures was not decreased compared to WT (data not shown) and we found no evidence for increased senescence or apoptosis in FRG1 satellite cells compared to WT (Dimri et al., 1995; Kudryashova et al., 2012) (see supplementary material Fig. S4). Overall, these data suggest that FRG1 satellite cells display an activation, and cell-cycle entry defect associated with an impairment in quiescence exit.

To further support our findings, we decided to assay the clonogenic potential of FRG1-derived myogenic stem/progenitor cells. Muscle cells were isolated, cultured in low-density and the number of myogenic clones was counted. In accordance with our previous results, FRG1 cultures contained less myogenic clones than WT (Fig. 3A; two-way Anova test, P = 0.0001, n = 10) resulting in a strikingly reduced number of total MyoD⁺ cells (Fig. 3B; unpaired t-test, P = 0.0001, n = 4), as assayed by immunofluorescence. On the other hand, the clones emerging in FRG1 cultures contained a number of cells similar to WT clones (Fig. 3C). Since the proliferative rate of FRG1 cells was only slightly lower than WT cells (Fig. 3D; unpaired t-test, P = 0.0176, n = 3), we concluded that the greatly reduced number of total myogenic cells (Fig. 3B) is primarily caused by an activation defect.

Next, we analyzed the differentiation ability of primary myoblasts from FRG1 mice. To this aim, we plated an equal number of muscle mononuclear cells before induction of differentiation, thus avoiding a possible bias due to the reduced cell number in FRG1 initial cultures. Interestingly, we found that FRG1 myoblasts displayed significantly reduced terminal differentiation compared to WT (Fig. 3E,F; two-way Anova test, P = 0.005, n = 3), though the percentage of MHC⁺ cells was not reduced in FRG1 cultures compared to WT (Fig. 3G). While this result suggests that a fusion defect could be involved in the terminal differentiation defect of FRG1 myoblasts, further work is required to determine the molecular mechanism underlying it.

To assure that the abovementioned defects are not caused by a reduction of the initial number of myogenic cells, several independent approaches were employed to measure the number of quiescent satellite cells in vivo. First, the expression level of Pax7 (Seale et al., 2000), was similar between FRG1 and WT mice (Fig. 4A). Second, the number of satellite cells defined anatomically as Pax7⁺ mononuclear cells located underneath the basal lamina of adult myofibers was equal between FRG1 and WT animals (Fig. 4B,C). Thirdly, muscle mononuclear cell preparations from FRG1 mice and WT littermates contained a similar percentage of satellite cells identified by flow cytometry and cytospins, as CD34⁺/Integrin-α7⁺/CD31⁻/CD45⁻/Sca1⁻ and immunofluorescent Pax7⁺ cells, respectively (Fig. 4D,E). Collectively, our results indicate that long before the development of muscular dystrophy, FRG1 mice display myogenic stem-cell activation, proliferative, clonogenic and differentiation defects.

Upon injury, satellite cells are activated, proliferate and fuse to repair the muscle (Chargé and Rudnicki, 2004). To evaluate the regeneration ability of satellite cells, muscle injury was induced by cardiotoxin (CTX) injection in vastus of FRG1 and WT mice at 4 weeks of age. Ten days after CTX injection, mice were sacrificed and their vasti were isolated and transversally cryosectioned. Morphological analysis revealed that muscles from FRG1 mice contained extremely small centrally nucleated (regenerating) myofibers (CSA: 1009 µm² ± s.e.m. = 119 µm²) compared to WT controls (CSA: 2178 µm² ± s.e.m. = 69 µm²; Fig. 5A; unpaired test, P = 0.0009, n = 3). While developmental MHC (MHCd) immunofluorescence staining was absent in WT muscles, as expected (Murphy et al., 2011), MHCd⁺ regenerating myofibers were still evident in FRG1 muscles confirming a delay in muscle regeneration (Fig. 5B). Four weeks after cardiotoxin injection, WT injured muscles, although not completely regenerated, contained a significantly lower percentage of centrally nucleated myofibers than FRG1 injured muscles (Fig. 5C,D; unpaired t-test, P = 0.004, n = 3). qRT-PCR for the muscle regeneration marker MyHC-emb (myosin, heavy polypeptide 3, skeletal muscle, embryonic) confirmed the delayed regeneration in FRG1 muscles (Fig. 5E; paired t-test, P = 0.0208, n = 6). To further evaluate the extent of this defect, we performed repeated injury experiments, where muscles were injected four times with an interval time of 1 week. Contrary to WT animals, FRG1 mice were unable to repair the damaged muscle and a part of the tissue was replaced by fat, as shown in vastus transversely cryosectioned, 4 weeks after the last damage (Fig. 5F).

An in vivo immunofluorescence analysis for Pax7 and Ki67 on muscle transverse cryosections 2 days after the damage, revealed that the number of Pax7⁺/Ki67⁺ proliferating cells was lower in FRG1 than WT muscles (Fig. 5G: paired t-test, P = 0.0096, n = 3), suggesting a role for satellite cells in the muscle regeneration impairment of FRG1 mice. Nonetheless, the FRG1 transgene is expressed in both satellite cells and adult muscle, thus a reduced regeneration could be due to defects in any of these compartments. To investigate if a cell-autonomous defect contributes to the muscle regeneration impairment of FRG1 mice, we performed satellite cell transplantation assays. To this aim, we generated WT and FRG1 EGFP⁺-transgenic mice. Freshly isolated satellite cells were transplanted in CTX-treated tibialis anterior of WT animals to assess the ability of purified muscle stem cells to engraft and proliferate upon transplantation into WT muscles. After 3 weeks, we sacrificed the mice and analyzed cryosectioned muscles by immunofluorescence with an anti-EGFP antibody (Fig. 6A). Although, the number of EGFP⁺ myofibers was similar in mouse muscles transplanted with WT and FRG1 satellite cells (Fig. 6B), the CSA frequency distribution was different for the FRG1 donor-derived myofibers compared to WT controls (Fig. 6C; Kolmogorov–Smirnov test, P<0.001, n = 6) and the FRG1 muscle stem cells gave rise to smaller myofibers compared to WT (Fig. 6D), indicating a reduced ability of FRG1 satellite cells to contribute to muscle upon transplantation.

Although we cannot completely rule out the possibility of defects in other cell types or a direct effect on the myofiber itself as potential contributing factors, these results suggest that pre-dystrophic FRG1 mice display a cell-autonomous defect in the muscle stem cell compartment that could contribute to the development of muscular dystrophy.

To explore the molecular mechanism underlying the disease onset, we performed a gene expression profiling of vastus muscles from FRG1 and WT mice at 4 weeks of age (n = 3), when no dystrophic symptoms other than a reduced myofiber size are present, using Illumina BeadChip expression arrays. 390 genes were differentially expressed (upregulated: 310; downregulated: 80) between FRG1 and WT samples (Fig. 7A; P<0.01; twofold difference). The microarrays results were validated by qRT-PCR, using a selection of differentially expressed genes (see supplementary material Table S2). Although our analysis was conducted on animals that did not present any sign of muscular dystrophy at histological level (Fig. 1), an Ingenuity Pathway Core Analysis (IPA) of all differentially expressed genes recognized an over-representation in several gene categories and networks associated with muscular dystrophy, like muscle disorders, cell death, muscle development and function and gene expression (see supplementary material Table S3). Intriguingly, performing a gene-set enrichment analysis using FSHD muscle datasets available in the Gene Expression Omnibus (GEO) database we found that FRG1 mice show a statistically significant, concordant expression profile with FSHD patients (Fig. 7B). Furthermore, to examine the lists of genes differentially expressed in these earlier studies, as a whole, we used rotation gene set testing (ROAST). We investigated if these genes display a higher upregulation or downregulation than average universal gene expression differences in our dataset. These tests were performed separately for each of the FSHD muscle datasets and significant correlation for both upregulated and downregulated genes between our study and the previous studies was observed (see supplementary material Table S4). Hence, our results further validate the FRG1 mouse as a useful animal model of FSHD and identify pathways relevant for its phenotype.

Despite its extensive study, FSHD pathogenesis remains unclear and controversial. All current models predict that deletion of D4Z4 repeats results in the de-regulation of a candidate gene(s), located in the FSHD region, leading to disease (Cabianca and Gabellini, 2010; van der Maarel et al., 2011). While the two most accepted FSHD candidate genes are DUX4 and FRG1, the molecular and cellular mechanism following their de-regulation and finally causing the disease remains elusive. Furthermore, FSHD is characterized by an extreme variability in disease onset, progression and severity. This heterogeneity in disease manifestation could reflect heterogeneity in gene expression of FSHD candidate gene(s). An interesting possibility, therefore, is that the complexity of FSHD could be explained envisaging that the epigenetic alteration of DUX4, FRG1 and other potential genes could collaborate to determine the final phenotype.

D4Z4 is a primate-specific repeat (Clark et al., 1996), consequently genetic mouse models of FSHD (displaying for example a different number of subtelomeric D4Z4 repeats) cannot be generated. Therefore, the functional consequence of D4Z4 deletion, like overexpression of an FSHD candidate gene, is the only disease aspect that can be modeled in mice. FRG1 overexpression in mouse, frogs and worms causes an FSHD-like phenotype (Gabellini et al., 2006; Hanel et al., 2009; Liu et al., 2010; Wuebbles et al., 2009). Importantly, we also found a remarkable similarity of the expression profile of FRG1 mice to one of the FSHD patients. Hence, the FRG1 mouse is to date the only mouse model that displays features of the human disease.

The detailed morphological analysis of vastus muscles from FRG1 mice demonstrated that they suffer from a postnatal muscle growth impairment, likely caused by defects of the satellite cells. In particular, we found that long before the development of any dystrophic sign FRG1 mice display myogenic stem cells activation, proliferative, differentiation and clonogenic defects. Satellite cells are necessary both for the post-natal muscle growth and regeneration and indeed, cardiotoxin injection revealed that muscle regeneration is also impaired, strengthening the importance of the abovementioned defects.

In the classical type of muscular dystrophies, the lack of a functional dystrophin–glycoprotein complex, causes mechanical fragility, contraction-induced damage and death of the muscle fibers. This leads to activation of the myogenic stem cells that take part in muscle regeneration. As a result, dystrophic muscles undergo repeated cycles of muscle fiber degeneration and regeneration. These vicious cycles continue until, in the late stages of the disease, the endogenous stem cell pool becomes exhausted and muscle fibers are replaced by fibrotic and adipose tissues, compromising normal muscle function (Mann et al., 2011). Hence, in classical muscular dystrophies, stem cells are not a primary target of the disease; but stem-cell defects are an indirect effect of the primary mutation, which causes muscular dystrophy, and they only contribute to disease progression.

Similarly, a proliferative defect of FRG1 primary myoblasts isolated from 18-week-old mice has been recently reported (Chen et al., 2011). FRG1 mice at 18 weeks of age show a pronounced dystrophic phenotype, therefore the myoblast defect described at this age could be secondary to muscle wasting and simply caused by the exhaustion of the muscle stem cell proliferative capacity in the severely dystrophic muscle. On the contrary, all our myogenic-cell and muscle regeneration experiments were performed in young animals, well before the appearance of dystrophic symptoms. This feature distinguishes our study, since we can exclude the possibility that the observed deficits are secondary effects of the dystrophic phenotype, strongly suggesting that stem-cell defects can be a primary component of muscular dystrophy in FRG1 mice.

Our data show that the FRG1 transgene is already expressed in satellite cells and in conjunction with our transplantation experiments in WT recipient muscles, where the donor myofibers arising from FRG1 satellite cells were smaller than the WT-derived ones, indicate a novel role for FRG1 in the regulation of stem/progenitor cell function. Nevertheless, FRG1 satellite cells gave rise to a similar number of donor myofibers as WT cells, suggesting that a WT myofiber and/or satellite-cell compartment can attenuate the FRG1 muscle stem cell defects. Thus, we do not exclude the possibility that a mature myofiber component can play an important role in the muscle abnormalities of FRG1 mice. Indeed, the muscle microenvironment in mice can influence the behavior/abilities of myogenic cells (Carlson and Conboy, 2007; Carlson et al., 2008; Conboy et al., 2003).

Our comparative gene expression profiling and the Ingenuity Pathway Core Analysis (IPA) of all differentially expressed genes recognized a high level of over-representation in several gene categories associated with muscular dystrophy, like inflammation, cell death, muscular disorders and genetic disorders. In addition, we have identified several genes that could explain the emerging tissue and satellite-cell defects. Interestingly, nitric oxide synthase 1 (Nos1) was found to be downregulated in FRG1 muscles. Nos1 activity inhibition by pharmacological treatment or a knockout approach reduces the activation of satellite cells and leads to muscle regeneration deficit (Anderson, 2000; Tatsumi et al., 2002; Tatsumi et al., 2006). Therefore, it is tempting to speculate that the FRG1 satellite cell defects could be associated with this pathway.

Overall, this study provides a significant insight in the mechanisms that contribute to the pathophysiology of muscular dystrophy in FRG1 mice and suggests that muscle stem cell defects could contribute to the disease.

FRG1 mice (Gabellini et al., 2006), control C57BL/6J littermates and C57BL/6-Tg (ACTbEGFP)1Osb/J mice (a gift from Dr Giuliana Ferrari) were maintained at Charles River (Calco, Italy). To generate EGFP transgenic mice that overexpress FRG1, C57BL/6-Tg(ACTbEGFP)1Osb/J heterozygotes were bred with FRG1-high mice. Mice at 3–14 weeks of age were sacrificed for this study. To induce muscle injury, 30 µl of CTX (0.1 mM) (Sigma) were injected into vastus lateralis muscles of 4-week-old males using a 29 G syringe. For repeated injury experiments, muscles were injected four times with an interval time of 1 week, and analyzed 4 weeks after the last damage. For transplantation experiments, WT males at 10–11 weeks of age were used as recipients and were injured 48 hours before transplantation into tibialis anterior muscles. Muscle SM/C-2.6⁺/GFP⁺ cells were sorted by flow cytometry as described below and 5000 cells were transplanted in each injured muscle. Mice were sacrificed 10 days, 3 or 4 weeks after injury.

Cell preparations were obtained by vastus lateralis muscles of 4-week-old males as previously described (Xynos et al., 2010) and were plated on collagen-coated dishes after pre-plating for 1 hour in uncoated dishes. Primary myoblasts were grown in nutrient mixture F-10 Ham (Sigma) supplemented with 20% FBS (Hyclone) and 5 ng/ml bFGF (Peprotech) for 1–5 days and differentiated in Dulbecco's modified Eagle medium (DMEM; EuroClone) supplemented with 5% donor horse serum (EuroClone). To calculate the fusion index, after the initial expansion, cells were trypsinized and re-plated in equal number before induction of differentiation, thus avoiding a possible bias due to the reduced cell number in FRG1 initial cultures. Next, cells differentiated for 1–2 days and were stained with anti-MHC antibodies and Hoechst. The fusion index was calculated by counting the total number of nuclei (Hoechst stain) and the number of nuclei that are present inside a mature myotube (defined as a MHC-positive syncytium with 3 or more myonuclei), and expressed as percentage of myonuclei. For clonogenic assays, muscle cells were plated at 200 cells/cm² and let grow (for 4–5 days) until visible, well-isolated colonies were formed. Next, the total number of clones and the number of cells per clone in the plate were counted. For proliferative assays, following an initial expansion of 3 days, cells were trypsinized, re-plated in equal numbers and counted after 14, 24, 38 and 48 hours. For cytospins, freshly isolated mononuclear cells were spotted in glass slides (20,000 cells/slide) and centrifuged for 5 minutes at 800 rpm. Cells were fixed in 4% paraformaldehyde (Electron Microscopy Science) for 15 minutes at room temperature and stained as described below. For flow cytometry, mononuclear cells freshly isolated from muscles were stained with antibodies for 45 minutes at 4°C and resuspended in DMEM supplemented with 10% FBS (EuroClone) at a density of 10×10⁶ cells/ml. The following antibodies were used: CD45-FITC (#553080), Sca1-FITC (#553335), CD31-FITC (#553335), CD34-Alexa-Fluor-647 (#560233) (BD PharMingen), Integrin α7-PE (#K0046-5, MBL) and SM/C-2.6-biotin (dilution 1:200) (Fukada et al., 2004). Flow cytometric analysis and sorting of muscle SM/C-2.6⁺ cells were performed with FACSCanto (Beckton Dickinson) and MoFlo Cell Sorter (Beckman Coulter) equipped with argon laser 488 nm and He-Ne laser 635 nm, respectively.

Vastus lateralis and tibialis anterior muscles were dissected, frozen in isopentane cooled in liquid nitrogen and cryosectioned (8-µm thick). Immunofluorescence and Gomori-trichrome and X-gal staining were performed as previously described (Dubowitz, 1985; Xynos et al., 2010). Characterization of fibrotic and fat tissue in the muscle was performed after Sirius Red and Oil Red O staining, respectively (Bortolanza et al., 2011). The following primary antibodies were used: mouse anti-MyoD1 clone 5.8A (#M3512, Dako; dilution: 1/50), rabbit anti-MyoD clone M-318 AC (sc-760AC, SantaCruz; dilution: 1/50), rabbit anti-Cav1 (sc-894, SantaCruz; dilution: 1/50), rabbit anti-Ki67 (NCL-Ki67p, Novocastra Laboratories; dilution: 1:200), mouse anti-MHCd (NCL-MHCd, Novocastra; dilution: 1/40), mouse MF20 antibody (Developmental Studies Hybridoma Bank; dilution: 1:2), rabbit anti-GFP (A11122, Molecular Probes; dilution: 1/200), chicken anti-laminin (ab14055, Abcam; dilution: 1/2000), rabbit anti-laminin (L9393, Sigma; dilution 1/300), rabbit anti-Casp3-activated (LS-C12476, LCBio; dilution: 1/100) and mouse anti-Pax7 (hybridoma bank, dilution: 1:2). Alexa Fluor 488 goat anti-rabbit, Alexa Fluor 488 goat anti-mouse, Alexa Fluor 555 goat anti-rabbit, Alexa Fluor 555 goat anti-mouse and Alexa Fluor 555 goat anti-chicken (Molecular Probes, 1:500) were used for secondary detection. Samples were mounted in aqueous medium and visualized at room temperature, using Imager.M2 (N-Achroplan 10× /0.25 NA and 20× /0.45 NA) and Observer.Z1 (N-Achroplan 10× /0.25 NA Ph1 and 20× /0.4 NA Ph2 Korr) microscopes (Zeiss). Pictures were acquired with AxionCamMRc5 and AxioCam MRm cameras respectively using its AxioVision Rel. 4.8.2 software by Nikon. All images were analyzed with ImageJ. Adobe Photoshop C5 was used to compose the final pictures.

Total RNA from tissues was extracted and treated with DNase 1, using the RNeasy Fibrous Tissue Midi or Mini Kit (Qiagen). Aliquots of RNA (500 ng) samples were checked for integrity quality number (RIN) above 8 using Agilent 2100 Byoanalyzer and amplified according to the specifications of the Illumina TotalPrep RNA Amplification Kit (Ambion, Austin, TX, USA). The cRNA samples were applied to the arrays of Sentrix MouseWG-6_V2 BeadChip (Illumina, San Diego, CA, USA) and hybridized according to the manufacturer's specifications. The Sentrix BeadChips were scanned with the Illumina's Beadarray system 500 G Scanner (Illumina). The hybridization-image signal intensity has been extracted, background subtracted and normalized using Illumina Inc. BeadStudio software version 3.3.7. The produced data has been checked to respect the Illumina internal quality control and loaded into Bioconductor software. All the raw data are available under the GEO record GSE28575. cDNA was synthesized using Invitrogen's SuperScript III First-Strand Synthesis Super-Mix. qPCRs (for primers see supplementary material Table S5) were performed with SYBR GreenER qPCR SuperMix Universal (Invitrogen) using Biorad's CFX96 Real-time System. Relative quantification was calculated with CFX Manager Software V.1.6. Validation of the differential expression of genes identified by DNA microarray was performed using TaqMan gene expression assays with custom-made TaqMan array microfluidic cards (Applied Biosystems). Relative quantification was calculated with qBasePLUS V.1.5 using Gapdh, Ppia and 18S rRNA as reference genes.

All the microarrays raw data has been background subtracted, log2 transformed and Quantile normalized using Bioconductor software package Lumi (Du et al., 2008). To identify differentially expressed genes based on moderate t-test, the bioconductor Limma package (Smythe, 2005) has been used. Genes have been selected using a P-value cut-off (after BH₁₁ adjustment) set to the minimal widely used P<0.01 (Shi et al., 2006) to control the false discovery rate and a log₂ fold-difference detection limit of 1. To test the association of selected differentially expressed genes with signaling pathways, cellular processes and molecular networks, the information provided in the Ingenuity Pathway Knowledge Base has been used with a score cut-off of 3.3 (i.e. P<0.0005). Comparison with public data has been performed using parametric gene set enrichment analysis (PGSEA) (Subramanian et al., 2005) and the derived GAGE and PAGE (Luo et al., 2009) and the Roast function from Bioconductor packages. All the Public data used has been downloaded from Gene Expression Omnibus (GSE15090, GSE3307, GSE10760 and GSE36398) (Arashiro et al., 2009; Bakay et al., 2006; Osborne et al., 2007; Rahimov et al., 2012) using an in-house adapted GEOquery function and converted in a single Bioconductor object, RMA normalized and processed to identify differentiated genes in the same way as done for our microarray data. To perform the PGSEA analysis on the public data, we used the PGSEA Simple Molecular Concepts class for holding the information and the HomoloGene database to map the identifier from Mus musculus to Homo sapiens.

Statistical analyses were two-tailed tests and performed using GraphPad Prism version 5.0a (GraphPad Software, San Diego).

We thank Giulio Cossu and Jose Teodoro for helpful discussions. Maria V. Neguembor conducted this study as partial fulfilment of her PhD in Molecular Medicine, Program in Cellular and Molecular Biology, San Raffaele University, Milan, Italy.