M-cadherin-mediated intercellular interactions activate satellite cell division

Muscle is one of the few tissues composed of cells with various nuclei. Based on transmitted electron microscope observations, initially it was postulated that endoreplication and amitotic divisions model syncytium formation during early muscle development (Boudjelida and Muntz, 1987). Later, several studies demonstrated the inability of differentiated muscle cells to proliferate and that post-mitotic mononuclear myocytes fuse with each other, giving rise to the syncitium (Hamilton, 1969; Hilfer et al., 1973; Muntz, 1975; Goldspink, 1979; Kielbówna and Daczewska, 2005). This primary musculature models the type, form and location of the fibers. Subsequently, satellite cells (Mauro, 1961) contribute to increase the diameter and length of the existing fibers, as they are the main source of myogenic cells after birth and during regeneration (Snow, 1978; Cusella-De Angelis et al., 1994; Edom-Vovard et al., 1999; Schmalbruch and Lewis, 2000; Seale et al., 2000; Chanoine and Hardy, 2003; Gargioli and Slack, 2004; Spalding et al., 2005; Chen et al., 2006; Mochii et al., 2007; Tseng and Levin, 2008; Cavaco Rodrigues et al., 2012). It was initially described (Moss and Leblond 1971) and widely assumed (Grounds et al., 2002; Chargé and Rudnicki, 2004) that satellite cells first become activated, divide and differentiate and only later do they fuse to the existing myofibers.

The physiology of satellite cells is tightly controlled by a complex network of transcription factors. Pax7 is the most frequently used marker to label satellite cells both in quiescent and activated states (Seale et al., 2000; Zammit et al., 2006). This transcription factor modulates the activation of satellite cells by controlling the induction of specific myogenic regulatory factors (MRFs) (Sassoon, 1993; Cornelison and Wold, 1997; Nicolas et al., 1998; McKinnell et al., 2008). Myf5 and MyoD are the first transcription factors expressed in committed myogenic progenitors (Davis et al., 1987; Hopwood et al., 1989; Kuang et al., 2007), whereas myogenin and MRF4 are the last genes to be induced during the differentiation of the satellite cells to myoblasts, just before specific muscle proteins, such as myosin, become expressed (Rhodes and Konieczny, 1989; Wright et al., 1989; Jennings, 1992; Cooper et al., 1999).

Although initially described as a homogenous population, nowadays it is widely accepted that satellite cells (Pax7+) are a heterogeneous population composed of adult stem cells (Pax7+/Myf5−) and committed myogenic progenitors (Pax7+/Myf5+) that coexist at different levels of differentiation. Pax7-positive cells can divide symmetrically, giving rise to identical daughter cells, or asymmetrically, giving rise to different daughter cells (Conboy and Rando, 2002; Shinin et al., 2006; Kuang et al., 2007). Moreover, satellite cell coexist in the same anatomical location together with satellite-cell-derived myoblasts (Pax7−/MyoD+/myogenin+) (Zammit et al., 2004; Kuang et al., 2007; Zammit, 2008; Day et al., 2009). Although myogenin is expressed before committed myoblasts fuse with myofibers, a minor percentage of this population still retains proliferative capacity (Andrés and Walsh, 1996; Zammit et al., 2004). In this report, both populations, satellite cells and derived myoblasts, are referred to as myogenic cells.

The transcription factor Mef2 (myocite enhancer factor-2) has been described as a cofactor of the MRFs, driving myoblasts differentiation together with MyoD and myogenin (Wong et al., 1994; Molkentin et al., 1995; Molkentin and Olson, 1996; Black and Olson, 1998). Desmin (a muscle-specific intermediate filament protein) and p21 (a regulator induced in post-mitotic myoblasts bringing about withdrawal from the cell cycle) are other markers of terminally differentiated myocytes (Lazarides and Hubbard, 1976; Guo et al., 1995; Halevy et al., 1995).

Another marker found in both quiescent and activated satellite cells is M-cadherin (Mcad), a calcium-dependent homophilic cell-to-cell adhesion molecule (Donalies et al., 1991; Irintchev et al., 1994; Rose et al., 1994; Cooper et al., 1999). The intercellular interaction between two sets of Mcad is necessary for the fusion between embryonic myoblasts (Zeschnigk et al., 1995; Kaufmann et al., 1999), as well as for the fusion of derived myoblasts with the existing myofiber (Wernig et al., 2004). However, mice with a null mutation in Mcad show no muscle defects, probably because of compensation by the other cadherins (Hollnagel et al., 2002). It has been described that the cytoplasmatic tails of Mcad interacts with beta-catenin (Bcat), and that both proteins are essential for the differentiation of the fusing myoblasts (Kuch et al., 1997; Wróbel et al., 2007). Moreover, different studies suggest a role for Bcat in controlling the transcription of target genes in proliferating satellite cells, in addition to its role in myoblast fusion. It has been described that in non-proliferating satellite cells, Bcat is located in a membrane-associated region, whereas in proliferating cells it translocates to the cytoplasm and the nucleus where it functions as a coactivator for the Wnt canonical signaling pathway (Otto et al., 2008; Zammit, 2008).

The muscle niche plays a crucial role in the physiology of satellite cells (Velleman, 2002; Collins et al., 2005; Kuang et al., 2008). The main components of the satellite cells niche are the associated microvasculature, extracellular matrix and muscle fibers. Proteins such as dystrophin, a key component of a multiprotein complex located in the plasma membrane of the myofiber, play a prominent role in the crosstalk between the extracellular matrix and the muscle fiber. In addition dystrophin provides tensile strength to the muscle fiber and acts as a scaffold for several signaling molecules (Hoffman et al., 1987; Peng and Chen, 1992; Rando, 2001).

As opposed to the prevailing model, our study shows that activated satellite cells fuse with the myofibers while they are still in mitosis. The analysis of different muscle samples by electron microscopy revealed the presence of satellite cells with condensed chromosomes and without a plasma membrane in the region of contact with the myofiber. Further histochemical analysis showed activated satellite cells with an intimate contact zone with the myofiber where dystrophin and the plasma membrane were absent. These observations prompted us to perform a thorough analysis of satellite cell division in neonatal mice musculature, because of the high ratio of cell divisions and the possibility of studying such mechanisms in vivo and in vitro. In addition to the electron and confocal microscopy study, we confirmed that Mcad-mediated intercellular interactions are crucial to stimulate satellite cell division in vitro. Furthermore, we show that the inhibition of cell fusion in vivo reduces the cycling of satellite cells. Our results suggest that Mcad-facilitated intercellular interactions play a fundamental role in the proliferation of muscle satellite cells both in vitro and in vivo.

We analyzed, by immunohistochemistry, muscle samples obtained during normal growth (Fig. 1A) and regeneration after tail amputation in Xenopus tadpoles (Fig. 1B), as well as normal growth (P7 mice, Fig. 1C; P30 mice, Fig. 1D) and regeneration after cardiotoxin-induced injury in mice (3-month-old mice, Fig. 1E). The presence of satellite cells was determined using the marker Pax7 and activated satellite cells using markers of cell division: PCNA (proliferating cell nuclear antigen, a marker for entrance into the S phase and maximally expressed in that phase of the cell cycle) (Johnson et allen, 1993) or PHH3 (phosphohistone H3, a marker for entrance into the M phase) (Gurley et al., 1978). In order to clearly identify the myofiber we performed colocalization with dystrophin, which is located at the muscle fiber membrane.

In all cases we observed that dividing satellite cells were always located in close physical proximity with the myofiber and over a contact zone in which dystrophin staining in the myofiber was almost undetectable. The absence of dystrophin in the area of the muscle fiber that was in contact with the dividing satellite cells suggested a potential disruption of the plasma membrane of the fiber and fusion with the satellite cell. Moreover, electron microscopy revealed signs of cell fusion between dividing satellite cells and the muscle fiber (Fig. 1F–H), without the presence of plasma membranes or cytoplasmatic continuity in the contact zone in the different musculatures studied.

In order to obtain details about the contact zone between the muscle fiber and activated satellite cell, we analyzed ultra-thin sections from the musculature of 1-week-old mice by electron microscopy. As expected, myogenic cells were shown to be located between the plasma membrane of the muscle fiber and its basal lamina, with clumps of highly condensed heterochromatin in their nuclei (Mauro, 1961; Hawke and Garry, 2001; Seale et al., 2000). When sections were analyzed at higher magnification, about 80% (see supplementary material Table S1) of the total myogenic cells had areas of fusion with the myofiber, as judged by the presence of cytoplasmatic continuity between both cells, disappearance of the plasma membrane, imprecise boundaries and amorphous material (Fig. 2A). These characteristics have been previously described during the last stages of the fusion process (Gonzélez et al., 1993). In some cases, we observed only one edge of the contact region fused, whereas at the other edge the two plasma membranes and the corresponding intercellular space could be clearly identified (Fig. 2B). However, in the areas where the two plasma membranes were clearly visible we could observe fusion intermediates as aligned vesicles, plaques (Fig. 2C) and pores (Fig. 2D) suggesting that the fusion process was in progress (Shimada, 1971; Kalderon and Gilula, 1979; Doberstein et al., 1997; Chen et al., 2007). We also noticed the presence of cytoplasmatic extensions of the myofiber that tended to surround the myogenic cell, fading at the point of contact, as previously described as cytoplasmatic flaps by Gonzélez et al. (Gonzélez et al., 1993).

Some of the cells that showed signs of fusion were undergoing cell division (16% of fusing myogenic cells). According to previous descriptions of the mitotic cell ultrastructure [appearance of the nuclear membrane, mitotic spindle and degree of chromatin condensation, or the presence of two cells with a mid body between them (Robbins and Gonatas, 1964; Wendell et al., 1993)] we observed two satellite cells in prophase (Fig. 2E), five in metaphase (Fig. 2F) and six in telophase. Interestingly, the degree of fusion in the contact zone between the satellite cell and myofiber increased as the cell cycle progressed, being complete in cells in metaphase. In three out of the six telophases both daughter cells were totally fused to the myofiber, indicating that probably both daughter cells would stay inside the myofiber after division (Fig. 2G). In contrast, in the other three telophases, only one of the two daughter cells was fused to the myofiber, while the other had a clear plasma membrane and intercellular space between itself and the myofiber, indicating that probably only one of the cells would stay inside the myofiber after division (Fig. 2H).

Since mitotic myogenic cells show clear regions of fusion with the myofibers, we conclude that cell fusion takes place while the cell is still dividing. The rest of the cells that show fusion with interphasic nuclei (84% of the fusing satellite cells) could correspond to post-mitotic myogenic cells undergoing differentiation or activated cells in a pre-mitotic phase.

To determine whether the main population of satellite cells with signs of cell fusion were differentiating or activated cells, samples were analyzed by immunohistochemistry and confocal microscopy. Paraffin sections of 1-week-old mouse musculature were stained for Pax7, PCNA and dystrophin. We observed a vast population of Pax7+/PCNA+ satellite cells (63% of total immunodetected cells; supplementary material Table S2). Interestingly, these cells had a contact zone with the myofiber with a fainter and discontinuous or totally absent dystrophin staining (undetectable dystrophin; Fig. 3A), whereas the myofiber perimeter just beside the contact zone presented a continuous and thick dystrophin staining (detectable dystrophin; supplementary material Movies 1, 2). The highest magnification and axial resolution of the confocal microscope (HCX PL APO lambda blue 63.0× 1.40 oil UV objective and pinhole aperture 1 Airy, z-sections of 0.5 µm) was needed to observe the delimited gap in the dystrophin organization, thus the z-planes over or below Pax7 cells presented a detectable dystrophin, as reported by Zhang and McLenan (Zhang and McLennan, 1994). No similar gaps were observed in the rest of the myofiber without satellite cells. In some cases myofibers showed extended irregular regions without dystrophin staining in all the z-planes observed, although this seems more likely to be an artifact of the sectioning.

Similarly, we identified a small fraction of Pax7−/PCNA+ cells (16% of total cells) comparable in terms of morphology and location to the Pax7+ population, which was present above a contact region with undetected dystrophin (Fig. 3A, red arrow). As expected, the percentage of total activated cells detected by immunohistochemistry in close physical proximity with the myofiber and showing a contact zone with undetected dystrophin, was similar to the percentage (80%) of myogenic cells initiating fusion with the myofiber identified by electron microscopy.

To further characterize cells undergoing cell division, we used an anti-pHH3 antibody, which labels mitotic cells (Fig. 3B). Pax7+/pHH3+ cells (supplementary material Table S3), as well as a few Pax7−/pHH3+ cells, were mostly located in the contact zone where dystrophin was absent. It is worth mentioning that Pax7+/PCNA+ satellite cells were located mainly in areas where dystrophin was partially undetected, whereas Pax7+/pHH3+ cells were exclusively found in areas where dystrophin was totally absent. This observation suggests that dystrophin expression decreases as satellite cells progress through the cell cycle, which is consistent with our previous observations using electron microscopy. Similar results were obtained by using a third proliferation marker, Ki67. About 66% of all Ki67+ cells detected had a contact zone with undetectable dystrophin (Fig. 3C; supplementary material Table S4) and 58% of Ki67+ cells were also positive for Pax7 (Fig. 3D; supplementary material Table S5).

These data indicate that in the neonatal mouse muscle most satellite cells are activated and have a contact zone with the myofiber with undetectable dystrophin.

We then decided to stain the plasma membranes with wheat germ agglutinin (WGA). WGA colocalized with dystrophin along the myofibers. However, we observed discordant patterns, but only in the contact zones between activated satellite cells and the myofiber. In a few cases (8% of total cells; supplementary material Table S6) dystrophin was detected, which colocalized with WGA staining (Fig. 3Ea), demonstrating that we were able to stain the plasma membrane in the narrow space between the two cells. In most cases we detected contact zones with undetectable dystrophin, with clear WGA staining (45% of total cells; Fig. 3F) or with weaker or totally absent WGA staining (Fig. 3Eb; 47% of total cells). We did not observe contact zones with clear dystrophin without WGA staining, demonstrating the reliability of the WGA as a plasma membrane stain. These data suggests that there is a progressive process just in the contact zone between the activated satellite cell and myofiber, in which there is dystrophin and the plasma membrane, then the dystrophin disorganizes and disappears and finally the plasma membrane disappears between the two cells.

As previously highlighted, in this case it was also necessary to use the highest magnification and axial resolution to observe the gap in WGA staining between the Pax7+ cell and myofiber, whereas the rest of the myofiber perimeter showed a continuous WGA staining (supplementary material Movies 3, 4).

In agreement with the electron microscopy results, we conclude that most of the satellite cells that are activated do not have a continuous plasma membrane in the contact zone with the myofiber.

We next used the marker laminin to confirm that myogenic cells undergoing cell division in intimate contact with the myofiber were in the internal face of the basal lamina. We identified mostly Pax7+/PCNA+ or pHH3+ satellite cells (Fig. 3G,H) and a smaller population of Pax7−/PCNA+ or pHH3+ cells surrounded by laminin (supplementary material Tables S7, S8).

In an effort to characterize the Pax7−/PCNA+ or pHH3+ myogenic sub-population, antibodies against myogenesis markers such as MyoD, Myf5 and myogenin were used. It is worth mentioning that when using different methodologies to detect these two markers (MyoD and Myf5), the gaps in the dystrophin staining just below the myogenic cells also appeared, as with the paraffin embedding.

We found MyoD+ cells over undetectable dystrophin (Fig. 4A) but we did not observe MyoD−/PCNA+ cells. Even though we found a large number of Myf5+/PCNA+ cells located just over areas where the dystrophin was undetectable (Fig. 4B), most Myf5+ cells outside the myofiber were also positive for Pax7 (supplementary material Tables S9, S10 and S11). These results suggest that neither MyoD nor Myf5 exclusively mark the subpopulation of Pax7−/PCNA+ or pHH3+ cells.

Cells positive for myogenin were mainly located under the laminin (Fig. 4C; supplementary material Table S12). Myogenin+/PCNA+ cells (21% of total cells, Fig. 4D; supplementary material Table S13) or myogenin+/pHH3+ cells (4% of total cells, Fig. 4E; supplementary material Table S14) were always located over undetectable dystrophin and surrounded by a large population of PCNA+ or pHH3+ cells that were negative for myogenin (76% and 15% of total cells in each immunodetection, respectively). Triple staining for Pax7, myogenin and PCNA (Fig. 4F) showed that all PCNA+ cells in intimate contact with the myofiber were positive for either Pax7 or myogenin and that these two markers were never co-expressed (supplementary material Table S15).

Similar to the Pax7+ cells, myogenin+/PCNA+ cells showed a myofiber contact zone with undetectable dystrophin, whereas the rest of the myofiber showed a clear and continuous staining. Similarly, we observed some myogenin+ cells with WGA staining in the myofiber contact zone (Fig. 4G) and others with altered or absent WGA staining in that area (Fig. 4H), while the staining in the rest of the fiber was continuous.

Our results suggests that two activated sub-populations of myogenic cells, positive for Pax7 (satellite cells) or myogenin (derived myoblasts), present a myofiber contact zone with altered or absent dystrophin and some without WGA staining.

The described observations prompted us to investigate whether Mcad could be localized in the contact zone with undetectable dystrophin. First, we confirmed that other cadherins, such as E-cadherin or N-cadherin, were not expressed between the muscle fiber and satellite cells (data not shown). Next, we analyzed whether both sub-populations of myogenic cells, Pax7- or myogenin-positive cells, express Mcad in quiescent and activated states (Fig. 5) as already described for Pax7+ cells (Irintchev et al., 1994; Rose et al., 1994). A few Pax7+ cells were negative for the proliferation marker Ki67 and positive for Mcad (Fig. 5A). Most Pax7+ cells were positive for Ki67, some of them (60%) were also positive for Mcad (Fig. 5B; supplementary material Table S16), and the rest negative for Mcad. The same study using an anti-myogenin antibody gave rise to similar results, with a few myogenin+ cells negative for Ki67 and positive for Mcad (Fig. 5C), and a majority of cells positive for Ki67, some of them (60%) also positive for Mcad (Fig. 5D; supplementary material Table S17), and the rest negative for Mcad. This data suggests that satellite cells express Mcad before (Mcad+/Ki67−) and during cell proliferation (Mcad+/Ki67+) and that its expression goes down concomitantly with cell cycle progression (Mcad−/Ki67+). In agreement with this, we observed a clear reduction of Mcad signal in mitotic Mcad+/pHH3+ cells (Fig. 5E).

Staining with anti-Mcad and anti-dystrophin antibodies (supplementary material Table S18) showed a few cells with perfect colocalization of the two cell markers at the contact region (10%, Fig. 5F), whereas 76% showed Mcad staining over discontinuously detectable dystrophin (Fig. 5G) and 14% showed Mcad over undetectable dystrophin (Fig. 5H).

It has been described that Mcad interacts with Bcat (Kuch et al., 1997; Wróbel et al., 2007) so we decided to analyze potential correlations between the expression of Bcat, Mcad, Ki67 and dystrophin. We found that Bcat and Mcad always colocalize at the contact area between the myofiber and quiescent (Fig. 5I) and activated (Fig. 5J) satellite cells, and follows the same expression pattern, with most Bcat+ cells (50%) also positive for Ki67 and located over a contact zone with discontinuously detectable dystrophin (70%; Fig. 5K,L; supplementary material Tables S19, S20).

This data indicates that Mcad+/Bcat+ quiescent cells are located over a detectable dystrophin contact zone, whereas the main population of Mcad+/Bcat+ activated cells are situated over discontinuous or undetectable dystrophin, before the disappearance of both proteins at the contact region.

To further confirm that the main population of myogenic cells observed by electron microscopy with signs of fusion were not differentiated cells, we used antibodies against four proteins of fully differentiated myogenic cells: Mef2, desmin, p21 and macroH2A2 (mH2A2). The expression of the histone variant mH2A2 has been described to be high in terminally differentiated cells (Pehrson et al., 1997) and during adult stem cell differentiation (Barrero et al., 2013).

When we combined Mef2, PCNA and dystrophin staining, we observed a large number of Mef2+ nuclei inside the myofiber (Fig. 6A; supplementary material Table S21), a few Mef2+ cells outside the myofiber that were likely to be smooth muscle cells from the vasculature (Black and Olson, 1998), and a small percentage of Mef2+/PCNA+ cells in areas where the dystrophin expression was altered (Fig. 6B). These Mef2+/PCNA+ cells are likely to be myogenin+ cells, since we could not detect any colocalization between Pax7 and Mef2 signals (Fig. 6C; supplementary material Table S22). On the contrary, we detected colocalization between Mef2 and myogenin+/PCNA+ cells (Fig. 6D; supplementary material Table S23). Importantly, we could not detect any significant pool of Mef2 cells surrounding the myofiber before fusion.

Desmin signal was observed only inside the myofiber (Fig. 6E), where there were no mononuclear derived myoblasts with desmin+ cytoplasm outside the area with dystrophin staining (Fig. 6F). mH2A2+ nuclei (Fig. 6G) and p21+ nuclei were also in intimate physical contact with the myofiber, never colocalized with PCNA+ nuclei (Fig. 6H), or Pax7+ cells (Fig. 6I) and were always located under the dystrophin signal (Fig. 6J).

These observations indicate that satellite cells and derived myoblasts that initiate their fusion with the myofiber are activated cells, and that final myogenic differentiation occurs once the myogenic nuclei are inside the myofiber.

Taking into account the in vivo results, we decided to study mouse myogenic cell division in culture. Following the prevailing model, we expected to see individual Pax7+ proliferating cells, and some groups of Pax7-negative-derived myoblasts, initiating their fusion through Mcad. Surprisingly, immunocytochemical analyses revealed a significant proportion of proliferating Pax7+ (Fig. 7A) or myogenin+ (Fig. 7B) cells that were in intimate physical contact expressing Mcad between them (80% of total Pax7+/Ki67+ cells; and 89% of total myogenin+/Ki67+ cells respectively; supplementary material Tables S24, S25), suggesting a potential role for Mcad in cell division, in addition to cell recognition and fusion. No E-cadherin or N-cadherin signal was detected (data not show), and Mcad signal colocalized with Bcat in the contact regions of Ki67+e cells (Fig. 7C).

To study live myogenic cell division, we transduced cells with a GFP–PCNA encoding vector and performed time-lapse experiments. Although constitutive expression of GFP–PCNA was observed as homogeneous nuclear staining, clear replication foci could be detected from the beginning and throughout the S phase (Leonhardt et al., 2000; Kisielewska et al., 2005). Therefore, we were able to detect the activation of myogenic cells through changes in the GFP signal, as well as mitotic cells by transmitted light imaging (supplementary material Movies 5, 6; Fig. 3B). Moreover, we were able to measure the duration between two clear changes in the GFP signal: from the appearance of the first replication foci, to the breakdown of the nuclear membrane, characterized by a decrease in the GFP signal (transition time, t_T).

Interestingly, we observed cells dividing in intimate physical contact with a bi-nuclear cell that was likely to be an incipient myotube (Fig. 7D; supplementary material Movie 7). Some other cells divided in a way that strongly reminded us of mouse embryo division (Fig. 7E; supplementary material Movie 8). These observations prompted us to think that cell-to-cell contact mediated by Mcad could activate myogenic cell proliferation, in a similar way to E-cadherin in mouse blastomeres (Shapiro et al., 1995; Stockinger et al., 2001). Experiments by González et al. showed that the adhesion of chimeric E-cadherin to single blastomeres recreates the signaling of the neighboring blastomere, increasing their division rate (González et al., 2011). In a similar way, and to confirm our hypothesis, we tested the effects of stimulating or blocking the Mcad response (by using Mcad recombinant protein or specific antibodies, respectively) on myogenic cell division in 15 hour time-lapse experiments.

In the absence of an Mcad recombinant protein or antibody (control conditions) we detected 14% of the cells activating S phase, 20% in mitosis and a t_T of 6 hours and 42 minutes (n = 336 cells; supplementary material Movie 9). Addition of Mcad recombinant protein to the culture medium (Mcad treatment) resulted in an increase in the number of activated myogenic cells (65% of total cells) and detected mitosis (64%) compared with the control. Interestingly, t_T was reduced to 4 hours and 38 minutes (n = 296 cells; supplementary material Movies 10–12). When an antibody against the extracellular domain of the Mcad was added to the medium (Ab 1 treatment) we observed an increase in the percentage of activated myogenic cells (28%), but a clear decrease in the number of mitotic cells (12%) compared with the control, and the t_T was extended to 7 hours 24 minutes (n = 448 cells). The delay in t_T was evident in some videos, where myogenic cells took more time to undergo a complete cell cycle in the presence of anti-Mcad (supplementary material Movies 13, 14), and this is probably why there were more activated cells than in the control.

This growth retardation is very similar to what has been described in cells overexpressing E-cadherin (Stockinger et al., 2001). In these cells, overexpressed E-cadherin prevents the translocation of Bcat to the nucleus, resulting in cell-cycle arrest. To test whether blocking Mcad with antibodies resulted in changes in Bcat-driven transcriptional activity, we performed qRT-PCR to evaluate the expression of Bcat target genes (supplementary material Fig. S1A). Interestingly, we found a statistically significant reduction in the Mcad transcripts in the Ab-1-treated cells, as well as a significant reduction in c-Myc expression levels, a Bcat target gene (He et al., 1998).

Growth curves comparing control and Mcad or Ab 1 treatments further confirmed that recombinant Mcad addition stimulates the proliferation of myogenic cells. Moreover, treatment with an anti-Mcad antibody raised against the cytoplasmatic domain of the protein (Ab 2 treatment) had no significant effects on growth rates (supplementary material Fig. S1B).

To further prove the differences in the three experimental conditions, we performed a EdU (5-ethynyl-2′-deoxyuridine) incorporation test (pulse of 45 minutes) 5 hours after starting the treatment. Dot plots and histograms (Fig. 7F) show an increase in the levels of incorporated EdU in Mcad-treated cells and a decrease in anti-Mcad Ab-1-treated cells, compared with the control. Accordingly, we detected (supplementary material Fig. S1C–E) clear differences in the mean intensity of fluorescence (MIF) suggesting that the Mcad-treated cells were incorporating EdU faster, whereas the anti-Mcad (Ab-1)-treated cells were replicating their DNA slower than the control. Although we did not detect significant differences in the percentage of EdU+ cells (S phase) between the three groups, an increase in the total number of cells was observed after Mcad treatment, as well as an increase in the number of cells in G2/M after the Ab 1 treatment.

When EdU pulses were longer (15 hours as in the time-lapse studies) no significant differences in the cell cycle profile between control and Mcad-treated cells were observed (Fig. 7G). However, in the presence of Mcad we detected a significant increase in the intensity of the EdU signal. Moreover, we detected (supplementary material Fig. S1F–H) double the number of cells after Mcad treatment, which is consistent with our observation in in vitro cultured cells that were able to undergo two rounds of cell division during the course of the time lapse study (15 hours; see supplementary material Movie 15). After anti-Mcad (Ab-1) treatment the differences were notable, with an increase in the number of cells that did not undergo replication (EdU−) and a decrease in the EdU+ cell fraction and MIF value, compared with the control. These effects were not observed after the treatments of mouse embryonic fibroblasts (MEFs).

These results suggest that Mcad-mediated cell interactions work as a trigger to boost cell cycle progression in cultured myogenic cells.

In order to investigate whether the cell-to-cell interaction of myogenic cells with the myofiber mediated by Mcad affects their proliferation capacity in vivo, we analyzed the effect of blocking Mcad interactions by two strategies: injecting antibodies against Mcad (using the same antibodies as in the in vitro study), and injecting lentiviruses encoding shRNAs against Mcad directly into the muscle of neonatal mice.

Fourty-eight hours after injecting Ab 1 in neonatal mice legs (supplementary material Fig. S2A) the number of total and activated Pax7+ cells was dramatically reduced compared with Ab-2-injected and control legs. Accordingly, we detected a clear reduction in Mef2+ nuclei inside the myofiber in the Ab-1-injected legs compared with Ab-2-injected and control legs.

We next tested the effects of knocking down the expression of Mcad in vivo using lentiviruses encoding shRNAs against Mcad. In order to confirm the efficiency of the shRNAs, we infected myogenic cells in culture with four different shRNAs against Mcad. Growth curves (supplementary material Fig. S2B) showed a clear reduction in the ratio of cell division in cells infected with the anit-Mcad shRNAs (LV-ShMcad) compared with those infected with the empty viruses (LV-ShCtrol). From the four tested shRNAs, shRNA1 produced more conspicuous effects and efficiently reduced the expression of Mcad, as judged by immunodetection (supplementary material Fig. S2C) and by qPCR (supplementary material Fig. S2D). Therefore, we used this lentivirus to inject neonatal mouse muscles. By 72 hours after the injection (supplementary material Fig. S2E) a clear reduction of the total and activated Pax7+ cells, as well as Mef2+ nuclei inside the myofiber, was observed in muscles injected with LV-shMcad compared to LV-shCtrol and non-injected (control) legs.

Owing to methodological difficulties in using antibodies or shRNAs in adult Xenopus, we performed a cell fusion inhibition experiment in tadpoles using tunicamycin (supplementary material Fig. S2F). Previous findings revealed that tunicamycin inhibits myoblast fusion by blocking the synthesis of UDP-N-acetyl-glucosamine, reducing the number of nuclei inside the muscle fibers (Gilfix and Sanwal, 1980). After 48 hours of tunicamycin treatment, we observed a clear reduction in the number of total and activated Pax7+ cells in the body musculature. We also detected a decrease in Mef2+ nuclei inside the myofiber.

We thus concluded that inhibition of satellite cells fusion in vivo blocks their proliferation and potentially their differentiation.

We have explored the mechanisms of myogenesis in neonatal mouse and Xenopus tadpoles. We found that mouse myogenic cells interact through Mcad with the myofiber in a contact zone where dystrophin expression is absent. Our observations in cultured mouse myogenic cells demonstrate that such cell-to-cell interaction activates and accelerates their cell division. Although more studies should be performed in order to reveal the mechanistic nature of these effects, our data suggest that the Mcad–Mcad interactions could be regulating cell proliferation through Bcat transcriptional activity, and thus probably through the canonical Wnt signaling pathway. The fact that we could not detect nuclear Bcat accumulation in our samples is in agreement with other reports that describe changes in Bcat transcriptional activity in the absence of detectable changes in nuclear localization, since low levels of nuclear Bcat were found to be sufficient to stimulate the transcription of its target genes (Stockinger et al., 2001).

Taking into account our observations, we propose a revised model for myogenic cell division during neonatal myogenesis (Fig. 8). Quiescent Pax7+ or myogenin+ myogenic cells express Mcad and Bcat in their contact zone with the myofiber, which still shows detectable dystrophin. Once the muscle growth process starts, the dystrophin in the contact zone starts to disorganize and myogenic cell proliferation becomes activated, followed by initiation of its fusion with the myofiber. Subsequently, the dystrophin completely disappears from the contact zone where Mcad, Bcat and plasma membranes are still present. Finally, the plasma membrane disappears between both cells while the myogenic cell is still in mitosis. After telophase, one daughter cell stays outside the myofiber to replace the Pax7-positive cell pool and the other one stays inside the myofiber and differentiates, or both daughter cells differentiate.

We conclude that the alteration of dystrophin, which gives resistance and rigidity to the myofiber in an organized state, is related to the activation of the myogenic cells and its fusion with the myofiber. This alteration seems to confer plasticity to the plasma membrane of the myofiber, as judged by the presence of cytoplasmatic extensions or flaps that surround part of the myogenic cell during cell fusion. It is also likely that the disorganization of the dystrophin is necessary for the cell-to-cell interaction of the satellite cells in humans, as it has been described that dystrophin becomes narrow in areas where activated satellite cells reside after intense exercise (Lindström et al., 2010).

We propose a model in which proliferation and differentiation of satellite cells take place while in physical contact with the myofiber. Therefore, the muscle fiber probably plays a critical role in satellite cell self-renewal and differentiation regulation. It has been described that changes in the muscle niche, rather than modifications of the satellite cells themselves, appear to be the main factor responsible for the declining regenerative response of old muscle (Carlson and Faulkner, 1989). In fact, aged satellite cells transferred into a younger muscle can divide at the same ratio as young satellite cells (Shefer et al., 2006; Carlson and Conboy, 2007).

Another aspect of our model that differs from the prevailing model is the final differentiation of the adult stem cells or the myogenic progenitors. Our results agree with the study done by Schulze (Schulze et al., 2005), in which mesenchymal stem cells only differentiate to myocytes once they fuse with them. In fact, it has been postulated that the direct interaction between differentiated muscle cells and stem cells is a prerequisite for their differentiation (Nunes et al., 2007; Boonen and Post, 2008).

Previous studies have already linked cell proliferation and fusion, suggesting that embryonic myoblasts can fuse among themselves or with the myofiber only just after mitosis, during the G1 phase (Bischoff, 1990; Bischoff and Holtzer, 1969). According to our model, the fusion of myogenic cells starts at the interphase of the mother cell and finishes just after a cell division in the G1 phase of the daughter cell. The data here reported are thus in agreement with those previously described by Bischoff and colleagues (Bischoff, 1990; Bischoff and Holtzer, 1969; Ishikawa et al., 1968), and in providing new insights into vertebrate myogenesis, they may serve as the basis for the development of new strategies to induce muscle regeneration in human muscle dystrophies.

Body musculature of stage 54 Xenopus and thigh muscle of B6 mice at post-natal day (P)7 were used in this study. All animal experiments were done following experimental protocols previously approved by the Institutional Ethics Committee on Experimental Animals, in full compliance with Spanish and European laws and regulations.

Samples were fixed with 2.5% glutaraldehyde for 2 hours at 4°C and post-fixed in 1% osmium tetroxide (2 hours at 4°C), dehydrated with ethanol and embedded in epoxy resin. Samples were examined with a JEOL 1011 transmission electron microscope (Tokyo, Japan).

Samples were fixed with 4% paraformaldehyde for 2 hours at 4°C for tadpoles and 24 hours at 4°C for mice samples. Cultured satellite cells were fixed with 2% paraformaldehyde for 20 minutes at room temperature. Paraffin samples were processed with antigen retrieval in the Dako Pascal system (Glostrup, Denmark), and treated with 0.5% Triton X-100 and 3% donkey serum for 30 minutes. Primary antibodies (see supplementary material Table S26) were incubated overnight at 4°C, and secondary antibodies (see supplementary material Table S27) for 2 hours at 37°C. To use the Myf5 antibody, samples were processed with OCT. For MyoD antibody, samples were frozen with isopentane, sectioned with a cryostat, and fixed with paraformaldehyde for 10 minutes at room temperature. mH2A2 antibody was previously described (Buschbeck et al., 2009). Images were taken using a Leica TCS SP5 microscope (Wetzlar, Germany).

Satellite cells were isolated from P7 mice as previously described (Perdiguero et al., 2007). Briefly, thighs were mechanically dissected, digested using 1% pronase, and satellite cells were recovered by centrifugation on a Percoll density gradient. Cells were cultured in collagen-coated dishes in the presence of Ham's F-10 medium containing 20% FBS, 2 mM Glutamax, and 0.1 µl/ml FGF. Phenotyping of the cell culture was performed prior to each experiment (see supplementary material Table S26 for antibodies used) by flow cytometry (Beckman Coulter, Indianapolis, IN), giving similar results to those published for satellite cells in culture (Ieronimakis et al., 2010).

The vector pEGFP-PCNA-IRES-puro2b from Addgene (26461) was digested with XhoI and BamHI and the insert was subcloned into pMSCVpuro (Clontech). Retroviral particles were obtained by transfection into Phoenix Amphotropic cells (ATCC). Viral supernatants were harvested on two consecutive days every 24 hours and filtered through a 0.45 µmm PVDF filter (Millipore). 100,000 satellite cells were spinfected (at 750 g for 45 minutes) twice at 24 hours intervals with fresh retroviral preparations in the presence of polybrene. Transduced cells were selected with puromycin.

Cells were filmed for 15 hours using a Leica SP5 inverted microscope with an in vivo system, and a 63× objective; one image was obtained every 6 minutes, in Mat-tec plates. Ten fields were detected in the same recording using a motorized stage. At least two recordings were made for each condition.

Recombinant Mcad (12 µmg/ml; R&D, 4096-MC-050; Minneapolis, MN), antibody 1 against the extracellular domain of the Mcad (1∶50, Santa Cruz; sc-81471; Santa Cruz, CA) or antibody 2 against the cytoplasmic domain (1∶50, Santa Cruz; sc-374093) were added to satellite cell cultures for 5 hours or 15 hours. At least three replicates of each condition were analyzed, and the experiment was repeated at least twice.

Cells were incubated in 6-well plates with 1 µmM EdU for 45 minutes or 15 hours and analyzed using the Click-iT EdU cell proliferation assay kit from Invitrogen (Carlsbad, CA), detected with a Moflo flow cytometer (Beckman Coulter, Indianapolis, IN).

Isolation of total RNA from was performed using Trizol reagent (Invitrogen, Carlsbad, CA). All samples were treated with TURBO DNase inhibitor (Ambion) to remove any residual genomic DNA and 2 µg of RNA was used to synthesize cDNA using the Invitrogen SuperScript II Reverse Transcriptase kit. 25 ng of cDNA were used to quantify gene expression by quantitative RT-PCR using primers described in supplementary material Table S28.

The recombinant lentiviruses were produced by transient transfection of HEK293T cells cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen, Carlsbad, CA). At 60–70% confluency the cells were co-transfected with 4 µg of pCMV VSV-G, 8 µg psPAX2, carrying lentiviral envelope protein or lentiviral packaging proteins, respectively, and 12 µg of lentiviral pLKO.1-puro shRNA plasmids carrying interferential sequences to knock down the M-cadherin gene (MISSION shRNA NM_007662, TCRN0000094574-77; Sigma-Aldrich, St. Louis, MO). The transfection was performed using Fugene 6 transfection reagent (Promega Corporation, Madison WI, USA) following the manufacturer's instructions. 24 hours and 48 hours post-transfection, medium containing viruses was collected, pooled, centrifuged at 3500 r.p.m. at room temperature for 5 minutes and filtered through a 0.45 µm PVDF filter. The viruses were concentrated by centrifugation at 19,600 r.p.m. at 22°C for 140 minutes. The pellet was resuspended in phosphate-buffered saline (PBS) and aliquoted viral stocks were stored at −80°C.

2.5 µg of Ab 1 or Ab 2 against Mcad diluted in 25 µl of NaCl were injected in the thighs of P7 mice with a Hamilton syringe. 2.5 µl of LV-shCtrol or LV-shMcad-1 (TCRN0000094574) preparations (10⁹ IU) were injected in the thighs of P7 mice with a Hamilton syringe. Tunicamycin (Sigma; Ref: T7765; St. Louis, MO. USA) stock solution (1 mg/ml) was prepared in DMSO, and a 1/10 dilution in mQ H₂O was injected into the body musculature of stage 54 tadpoles. Animals were sacrificed at 24 hours and 48 hours after injection for antibody and tunicamycin treatments, 48 hours and 72 hours after injection for lentivirus experiments.

In the muscle sections we used images that had been obtained with a 63× objective, taking images of at least eight fields of each individual. For the study of dystrophin organization and Mcad distribution, an over zoom was needed. In the cultured cells we took 10 random fields in each slide with the 20× objective. The numbers of cells in the different satellite cell subpopulations were expressed as the percentages of the total number of cells immunodetected with each antibody combination. For the cell culture experiments, the results were expressed as the percentage of the total number of fields. In the flow cytometer assay, evaluation of cycling cells was based on the mean intensity of fluorescence (MIF), measured as signal to noise ratio (MIF from the EdU-positive cell fraction/MIF from the EdU-negative cell). In the in vivo experiments 10 fields from three different sections were taken randomly, and the percentage of the total myofiber number was calculated. We processed the images using the Metamorph software (Molecular Devices, Sunnyvale, California, USA). The statistical analysis was carried out using the Student's t-test.

We thank C. Fabregat, T. Ventura C. Gómez and M. Carrió for technical assistance, C. Mann, M. Zamora and P. Muñoz for sharing concepts, C. Jopling and B. Christen for discussing ideas and M. Buschbeck for mH2A2 antibody. MANDRA1 clone 7A10 (G. E. Morris), and F5D (W. E. Wright) were obtained from The Developmental Studies Hybridoma Bank, The University of Iowa.