The location and regulation of fusion events within skeletal muscles during development remain unknown. Using the fusion marker myomaker (Mymk), named TMEM8C in chicken, as a readout of fusion, we identified a co-segregation of TMEM8C-positive cells and MYOG-positive cells in single-cell RNA-sequencing datasets of limbs from chicken embryos. We found that TMEM8C transcripts, MYOG transcripts and the fusion-competent MYOG-positive cells were preferentially regionalized in central regions of foetal muscles. We also identified a similar regionalization for the gene encoding the NOTCH ligand JAG2 along with an absence of NOTCH activity in TMEM8C+ fusion-competent myocytes. NOTCH function in myoblast fusion had not been addressed so far. We analysed the consequences of NOTCH inhibition for TMEM8C expression and myoblast fusion during foetal myogenesis in chicken embryos. NOTCH inhibition increased myoblast fusion and TMEM8C expression and released the transcriptional repressor HEYL from the TMEM8C regulatory regions. These results identify a regionalization of TMEM8C-dependent fusion and a molecular mechanism underlying the fusion-inhibiting effect of NOTCH in foetal myogenesis. The modulation of NOTCH activity in the fusion zone could regulate the flux of fusion events.
Skeletal muscle development, homeostasis and regeneration rely on muscle progenitors and stem cells that undergo a multistep process to form multinucleated cells named myofibres. Myofibres are formed by myoblast fusion and constitute the main structural unit of skeletal muscles. The fusion of myoblasts has been poorly studied and remains challenging to study in vivo because it is difficult to dissociate the differentiation and fusion processes during myogenesis. Moreover, the location of fusion events within the developing muscles remains unknown.
The transcriptional control of the skeletal muscle programme is dependent on four bHLH transcription factors, MYF5, MRF4, MYOD and MYOG, named the myogenic regulatory factors (MRFs) (Buckingham and Rigby, 2014). MRFs have the ability to trigger the muscle programme, including the successive steps of specification, commitment, differentiation and fusion from muscle progenitors (Buckingham and Rigby, 2014; Comai and Tajbakhsh, 2014; Esteves de Lima and Relaix, 2021), but also from non-muscle cells in vitro and in vivo (Delfini and Duprez, 2004; Weintraub et al., 1991). Multinucleated myofibre formation is a multistep process involving cell cycle withdrawal of already-committed myoblasts, cell elongation, cell-cell contact and fusion (Biressi et al., 2007; Comai and Tajbakhsh, 2014). The three main steps that underlie myoblast fusion are: (1) cell recognition and adhesion; (2) enhancement of cell proximity via F-actin-propelled membrane protrusions from one fusion partner cell and myosin II-dependent cortical tension in the other fusion partner cell; (3) destabilization of the two apposed plasma membrane lipid bilayers and formation of a fusion pore (Hernández and Podbilewicz, 2017; Kim et al., 2015). Numerous transmembrane proteins and molecules of the intracellular actin machinery are involved in myoblast fusion in vertebrates. However, these proteins are not specific to myoblast fusion and are recruited for any cell-cell fusion processes, such as sperm-egg fusion in fertilization, cytotrophoblast fusion in placentation and axonal fusion during neuronal repair (Hernández and Podbilewicz, 2017). Recently, a muscle-specific gene, myomaker (named TMEM8C in chicken, Mymk in mice and MYMK in humans) has been identified as being essential for myoblast fusion in mice, chicken and zebrafish during development and muscle regeneration (Landemaine et al., 2014; Luo et al., 2015; Millay et al., 2013, 2014, 2016). Myomaker is a transmembrane protein of 221 amino acids with seven membrane-spanning regions. Autosomal recessive mutations in the MYMK gene, which reduce but do not eliminate MYMK function, cause a congenital myopathy, Carey–Fineman–Ziter syndrome, in humans (Di Gioia et al., 2017). Myomaker is sufficient to trigger fibroblast-myoblast fusion but not fibroblast-fibroblast fusion (Millay et al., 2014). However, when combined with the micropeptide myomixer (also named myomerger or minion), myomaker triggers fibroblast-fibroblast fusion (Bi et al., 2017; Quinn et al., 2017; Zhang et al., 2017). By itself, myomixer does not possess 10T1/2 fibroblast-C2C12 myoblast fusion activity (Bi et al., 2017). Two MRFs, MYOD and MYOG, have been shown to positively regulate the transcription of myomaker genes in mouse, zebrafish and chicken via E-boxes located in the myomaker promoter regions (Ganassi et al., 2018; Luo et al., 2015; Millay et al., 2014). The direct transcriptional regulation of the myomaker gene by MRFs couples the differentiation and fusion processes making it complex to study them independently.
NOTCH signalling is known to be involved in skeletal muscle differentiation during development. NOTCH inhibition promotes muscle differentiation in both in vivo and in vitro systems (Kitzmann et al., 2006; Kopan et al., 1994; Schuster-Gossler et al., 2007; Vasyutina et al., 2007). Conversely, NOTCH is a potent inhibitor of muscle differentiation in chicken and mouse models during embryonic and foetal myogenesis (Bonnet et al., 2010; Delfini and Duprez, 2004; Esteves de Lima et al., 2016; Hirsinger et al., 2001; Mourikis et al., 2012a; Vasyutina et al., 2007; Zalc et al., 2014). Active NOTCH also inhibits muscle differentiation in C2C12 cells (Kopan et al., 1994; Kuroda et al., 1999). NOTCH displays additional functions in postnatal myogenesis, as NOTCH is required to maintain the quiescence and to regulate the migratory behaviour of satellite cells (Baghdadi et al., 2018; Bjornson et al., 2012; Conboy and Rando, 2002; Conboy et al., 2003, 2005; Mourikis et al., 2012b). The canonical NOTCH pathway mediates cell-to-cell communication involving a transmembrane NOTCH receptor and a transmembrane NOTCH ligand. Upon ligand activation, the NOTCH receptor undergoes a proteolytic cleavage to produce the NOTCH intracellular domain (NICD), which translocates to the nucleus and interacts with RBPJ to regulate gene transcription (Andersson et al., 2011; Kopan et al., 1994; Lubman et al., 2007). NICD immediate transcriptome and chromatin immunoprecipitation (ChIP)-sequencing analyses have identified key genes activated downstream of NOTCH, which include genes encoding the basic helix-loop-helix (bHLH) transcriptional repressors HES (hairy and enhancer of split) and HEY (hairy/enhancer-of-split related with YRPW motif) (Andersson et al., 2011). During mouse foetal myogenesis, Heyl is the main transcriptional response to NICD (Mourikis et al., 2012a). HES and HEY proteins repress the transcription of muscle differentiation genes to maintain muscle cells in a progenitor state in mice (Bröhl et al., 2012; Fukada et al., 2011; Lahmann et al., 2019; Mourikis et al., 2012a; Noguchi et al., 2019; Zalc et al., 2014). Because NOTCH is involved in muscle differentiation, NOTCH function in myoblast fusion is difficult to address. No link has been established between components of the NOTCH intracellular pathway and the fusion gene myomaker.
The identification of myomaker as a transmembrane protein required for myoblast fusion (Millay et al., 2013) has stimulated research on muscle fusion. However, the signalling pathways that regulate myomaker expression have not been fully identified. Moreover, the location of fusion events within muscles remains to be understood in vivo. Here, we identify a regionalized location within foetal muscles for the key molecular actors involved in myoblast fusion and the NOTCH ligand jagged 2 (JAG2). We also identified NOTCH as a regulator of TMEM8C expression and of the fusion process during foetal myogenesis using in vitro and in vivo chicken systems.
Single-cell RNA-sequencing analysis shows that TMEM8C+ fusion competent cells co-express MYOD and MYOG in chicken limb foetal muscles
Given the requirement and sufficiency of TMEM8C (myomaker) for myoblast fusion (Millay et al., 2013), we used TMEM8C gene as a readout for muscle fusion. We first analysed the distribution of TMEM8C+ fusion-competent cells in single-cell RNA-sequencing (scRNAseq) datasets of chicken whole-limb cells at the embryonic day (E) 6 and E10 foetal stages. At E6, the transition time point between the end of embryonic myogenesis and the initiation of foetal myogenesis (Duprez, 2002), limb muscles are not yet individualized and spatially organized, whereas at E10 the final limb muscle pattern is set. We performed a Seurat-based clustering analysis, described elsewhere (Esteves de Lima et al., 2021), leading to the identification of all the limb cell clusters. As the focus of our study is muscle fusion, we only considered the muscle clusters (Fig. S1A,B) for this analysis. At E6, myogenic cells were segregated into two muscle clusters, the PAX7 (progenitor) and the MYOD/MYOG (myogenic) clusters (Fig. 1A), as evidenced by the differential expression of the PAX7, MYOD and MYOG genes in these clusters (Fig. S1C). At E10, the myogenic cell segregation could be further refined as four clusters, the PAX7, MYOD, MYOD/MYOG and myosin clusters (Fig. 1D, Fig. S1D), illustrating the cell progression through the canonical myogenic maturation steps. At both E6 and E10 stages, TMEM8C+ cells were mainly located in the MYOD/MYOG cluster, rather than in PAX7 (progenitor) or myosin clusters (Fig. 1B,E for feature plots and 1C,F for violin plots). We further analysed the co-expression of TMEM8C and myogenic markers (PAX7, MYOD, MYOG) at the single-cell level (Fig. 1G-I). At E6 and E10, fusion-competent TMEM8C+ myogenic cells represent 56.5% and 33.2% of the MYOG+ cells, respectively, which is the highest proportion compared with that for the PAX7+ (4.8% at E6 and 0.7% at E10) and MYOD+ (17.6% at E6 and 12.5% at E10) populations (Fig. 1I, top). Conversely, the repartition of TMEM8C+ cells across the PAX7+, MYOD+ and MYOG+ populations shows that the majority of the TMEM8C+ cells are not in a PAX7+ progenitor state but rather in MYOD+ and MYOG+ advanced states of myogenic differentiation. For example, MYOG+ cells represent 72.2% and 60.5% of TMEM8C+ cells at E6 and E10, respectively (Fig. 1I, bottom). The frequency of the combinations that include MYOD and/or MYOG amounts to 75% and 84% of the TMEM8C+ cells at E6 and E10, respectively. These frequencies are higher than those of corresponding combinations within the total myogenic (PAX7+, MYOD+ or MYOG+) population (39.6% at E6 and 54.8% at E10) (Fig. S1E-G, right panels), suggesting an enrichment of the MYOD and/or MYOG combination in the TMEM8C+ population. Altogether, we conclude that TMEM8C expression essentially correlates with MYOD+/MYOG+ advanced states of myogenic differentiation.
The fusion marker TMEM8C and the myogenic differentiation gene MYOG display a regionalized transcript expression within foetal skeletal muscles
In order to compare the spatial location of TMEM8C transcripts with those of MYOD and MYOG within foetal skeletal muscles, we performed in situ hybridization experiments on adjacent sections of chicken forelimb at foetal stages. At E6, when the muscle pattern is not yet set, faint TMEM8C expression was observed in limb muscle masses, labelled with MYOD and MYOG transcripts and myosins (Fig. S2). We did not observe any obvious regionalization of MYOD and MYOG transcripts within limb muscle masses, labelled with myosins, at this stage (E6) (Fig. S2). At E10, when the musculoskeletal pattern is set, the tendon marker SCX, which encodes a bHLH transcription factor, was used to label tendons and allow the visualization of muscle tips (muscle extremities close to tendons) in both longitudinal and transverse limb sections (Fig. 2A,E,I, Fig. S3A-C,G). In situ hybridization experiments on longitudinal (Fig. 2A-H) and transverse (Fig. 2I-L, Fig. S3C-J) muscle sections of E10 chicken embryos showed that TMEM8C transcripts were preferentially located in the middle of muscle and less so at muscle tips (Fig. 2B,F,J, Fig. S3D,H). MYOG expression was also higher in the central muscle region compared with muscle tips where the expression was faint (Fig. 2C,G,K, Fig. S3E,I), although MYOG expression domain was broader compared with that of TMEM8C. In contrast, MYOD expression encompassed the entire muscle, including muscle tips close to tendons (Fig. 2D,H,L, Fig. S3F,J). These in situ hybridization experiments show that the TMEM8C and MYOG expression is regionalized, whereas MYOD is homogeneously expressed within foetal muscles.
To confirm the preferential central location of TMEM8C transcripts, we compared the expression of TMEM8C with that of FGF4, which is known to be expressed in myonuclei at muscle tips close to tendons in limbs (Edom-Vovard et al., 2001, 2002) of chicken embryos. Comparison of TMEM8C expression with that of FGF4 showed a complementary expression pattern, i.e. a central muscle location for TMEM8C and muscle extremities for FGF4 (Fig. 3A-D,G,H), whereas MYOD expression (Fig. 3E,F) encompassed the entire muscle. We conclude that the transcripts of TMEM8C fusion gene displayed a preferential central location and is excluded from muscle tips labelled with FGF4.
At a cellular level, scRNAseq datasets indicated an exclusion of TMEM8C expression from PAX7+ progenitors and a preferential expression of TMEM8C in the MYOD+/MYOG+ population (Fig. 1). Consistently, double-fluorescence in situ hybridization confirmed that TMEM8C expression was excluded from PAX7+ progenitors (Fig. S4A,B,B′,B″). We also observed numerous TMEM8C+/MYOD+ cells and TMEM8C+/MYOG+ cells within foetal limb muscles (Fig. S4C-F″, arrows). Lastly, in central muscle regions TMEM8C expression (Fig. S4G,G′) was preferentially observed outside myosin+ myotubes (Fig. S4H,H′, arrows), although we detected a rare myotube expressing TMEM8C (Fig. S4H,H′, arrowheads). We conclude that the transcripts of the TMEM8C fusion gene displayed a preferential central location in myogenic cells outside myosin+ myotubes.
MYOG protein is regionalized in foetal muscles
Because of the observed co-segregation of TMEM8C+ cells and MYOG+ cells in the scRNAseq datasets (Fig. 1) and the regionalized location of TMEM8C and MYOG transcripts within limb muscles during foetal myogenesis (Figs 2 and 3), we investigated the location of the MYOG protein within muscles by immunostaining of limb sections (Fig. 4, Fig. S5). At E6, MYOG+ cells follow a similar pattern of distribution as PAX7+ progenitor cells and myosin+ differentiated cells in dorsal and ventral muscle masses (Fig. S5), which is consistent with the homogeneous MYOG transcript distribution (Fig. S2B-B″). At E10, when the final muscle pattern is set, muscle tips were visualized with pSMAD1/5/9 antibody (Esteves de Lima et al., 2021; Wang et al., 2010) and tendons were labelled with collagen type XII. Immunohistochemistry on longitudinal muscle sections showed fewer MYOG+ nuclei at muscle tips close to tendons, compared with the central muscle regions where MYOG+ cells were more abundant (Fig. 4A-C), consistent with the regionalization of MYOG transcripts (Fig. 2). PAX7+ cells displayed a homogeneous distribution within myosin+ domains (Fig. 4A′,A″,B′). Immunohistochemistry on transverse sections of a ventro-posterior limb muscle, the flexor carpi ulnaris (FCU), further showed the higher density of MYOG+ cells in central muscle regions compared with muscle tips (Fig. 4D,D′), whereas PAX7+ muscle progenitors (Fig. 4D′,E,F′) and MYOD+ cells (Fig. 4E) displayed a homogeneous distribution within the muscle, the muscle tips being visualized with pSMAD1/5/9+ myonuclei (Fig. 4F,F′) close to tendons (Fig. 4G). Quantification of the MYOG+ cells showed a twofold increase in the MYOG+ cell percentage in central regions compared with muscle tips (Fig. 4H-J), with no change in overall DAPI+ density of nuclei (Fig. 4K).
The transcripts of the NOTCH ligand JAG2 are regionalized within limb foetal muscles
The regionalized expression of fusion-associated markers suggested a preferential location of myoblast fusion in central muscle regions. In order to identify signalling pathways that could regulate myoblast fusion, we looked for genes displaying similar regionalization. Our previous observations suggested a regionalized expression of JAG2 in limb foetal muscles (see figure 2A,B in Esteves de Lima et al., 2016). JAG2 is the main NOTCH ligand produced by myotubes during limb foetal myogenesis in chicken embryos (Delfini et al., 2000; Esteves de Lima et al., 2016). The classical action mode for NOTCH signalling is described as NOTCH ligand being produced by differentiated muscle cells and acting in muscle progenitors (Delfini et al., 2000; Esteves de Lima et al., 2016; Kassar-Duchossoy, 2005; Zalc et al., 2014). Analysis of scRNAseq data sets from limb cells confirmed the expression of JAG2 within differentiated muscle cell clusters at E6 and E10, whereas HEYL expression, a readout of NOTCH activity, was associated with the PAX7 progenitor clusters (Fig. 5A,B compared with Fig. 1A,D). As expected, cells expressing JAG2 and TMEM8C partially overlapped, in contrast to those that express HEYL and TMEM8C, which are mutually exclusive (Fig. S6A,B,F,G). We confirmed in vivo that JAG2 was regionalized in limb foetal muscles (Fig. 5). JAG2 transcripts were excluded from muscle tips (Fig. 5D,D′,G,G′) close to tendons, visualized with SCX expression (Fig. 5C,C′,F,F′), whereas MYOD transcripts were present homogeneously throughout the muscle (Fig. 5E,E′,H,H′). The regionalization of JAG2 within muscle fibres suggests a regionalization of NOTCH activity in the central region of the muscle, where TMEM8C+/MYOG+ fusion-competent cells are preferentially located, during foetal myogenesis. As expected from the role of NOTCH signalling as a gate-keeper of myogenesis entry, the fraction of HEYL+ cells in myogenic populations decreased with the progression of myogenic differentiation in the scRNAseq data analysis (Fig. 5I, Fig. S6D). Strikingly, within the MYOD1+ and/or MYOG+ myocyte population, the percentage of HEYL+ cells was higher in the TMEM8C− population versus the fusion-competent TMEM8C+ population (Fig. 5I,J, Fig. S6C-E,H), showing that NOTCH activity and TMEM8C expression are mutually exclusive in the differentiated myogenic population. This suggests that downregulation of NOTCH signalling in differentiated myogenic cells may be required for fusion to occur.
NOTCH loss of function increases myoblast fusion in limb foetal muscles
In order to analyse the consequences of NOTCH inhibition on foetal muscle fusion, a dominant-negative form of DELTA1 (DELTA1-DN) under the control of the myosin light chain promoter was overexpressed specifically in differentiated myogenic cells. This prevented NOTCH ligand processing in ligand-producing cells and consequently inhibited NOTCH activation in signal-receiving cells (Chitnis, 2006; Esteves de Lima et al., 2016; Henrique et al., 1997). Limb somite electroporation with the pT2AL-MLC-Tomato-DELTA1-DN construct led to increased myoblast fusion (Fig. 6A-C,E-G, arrows indicate examples of Tomato+ fibres), as assessed by the increase in the percentage of myonuclei (MF20+ nuclei) with respect to all nuclei (Fig. 6I) and in the percentage of muscle fibres with a high number of myonuclei compared with control muscles (Fig. 6J). TMEM8C expression pattern became wider within the muscle in the NOTCH loss-of-function context compared with controls (Fig. 6D,H). We conclude that NOTCH activity regulates fusion in developing limb muscles.
Chicken embryo immobilization by decamethonium bromide (DMB) treatment decreases NOTCH activity and mimics a NOTCH loss-of-function phenotype, i.e. a reduction in the number of muscle progenitors and increased differentiation, this myogenic phenotype being rescued by NOTCH activation (Esteves de Lima et al., 2016). We used the same immobilization condition to address further the involvement of NOTCH in myoblast fusion in vivo. In contrast to the electroporation technique, which led to a mosaic expression of the electroporated gene within muscle cells, DMB-mediated immobilization affected all muscles. In immobilization conditions, following DMB treatment, we observed larger myotubes compared with controls (Fig. 6K,N). We also observed a wider expression pattern of the TMEM8C fusion gene within the muscles, associated with the loss of its regionalization in immobilized embryos compared with controls (Fig. 6L,M,O,P). We further observed the increase of TMEM8C mRNA levels by real-time quantitative PCR (RT-qPCR) experiments in immobilized foetal limbs (Fig. 6Q). The increase in the percentage of myonuclei versus all nuclei and in the percentage of muscle fibres with a high number of myonuclei (Fig. 6R,S) confirmed the increased fusion in immobilized embryos compared with controls. The phenotype observed with DMB-mediated immobilization is consistent with that of NOTCH loss of function, albeit stronger, which could be due to the wider exposure of all muscles to DMB compared with electroporation. To investigate whether the DMB-mediated increased fusion could be prevented by forced NOTCH activity, we performed grafts of DELTA1-expressing cells in one limb of immobilized embryos, using the contralateral limb as control. We observed that the percentage of myonuclei was decreased in DELTA1/DMB limbs compared with DMB-only limbs (Fig. S7I-N). This shows that active NOTCH prevents the increased fusion observed in DMB-treated muscles. Taken together, these experiments show that NOTCH loss-of-function conditions increase muscle fusion. Whether this increase in fusion results solely from increased differentiation (Esteves de Lima et al., 2016) or from a combined effect on differentiation and fusion cannot be resolved in vivo.
A myoblast culture system that mimics myogenesis
With the ultimate goal to uncouple differentiation and fusion processes, we turned to an in vitro system of chicken foetal myoblast cultures. We first assessed whether cultures of chicken foetal myoblasts mimic in vivo myogenesis. Foetal myoblasts were isolated from limbs of E10 chicken embryos and plated at low density with high serum-containing medium (proliferation conditions). Muscle cell cultures in proliferation conditions contained PAX7+ progenitor cells and myosin+ differentiated cells (Fig. S8A). When switched to a low serum-containing medium (differentiation conditions), confluent myoblasts were pushed to differentiation and fusion and formed multinucleated myosin+ cells while maintaining a pool of PAX7+ reserve cells (Fig. S8B), as described for mouse myoblast cultures (Baghdadi et al., 2018; Kitzmann et al., 2006). Consistent with the in vivo situation (Bröhl et al., 2012; Mourikis et al., 2012a), HEYL expression was decreased in differentiated muscle cell cultures compared with proliferative myoblast cultures, but still detected in differentiated myoblasts (Fig. S8C). We then performed NOTCH loss- and gain-of-function experiments in proliferative myoblast cultures and compared the phenotypes with those obtained in vivo. We performed NOTCH loss-of-function experiments in myoblasts using the NOTCH inhibitor DAPT. NOTCH inhibition with DAPT increased myoblast differentiation by increasing the expression levels of the differentiation markers MYOD, MYOG and MYHC, while decreasing those of PAX7 and MYF5 in proliferative myoblast cultures (Fig. S8D,E,G), consistent with our previous observations (Esteves de Lima et al., 2016). Conversely, we forced NOTCH activity by overexpressing DELTA1 using the RCAS retroviral system in proliferative myoblast cultures and in chicken limbs. DELTA1-activated NOTCH led to the mirror phenotype for myogenesis to that of NOTCH loss of function. The expression levels of MYOD, MYOG and MYHC mRNAs were decreased, whereas those of PAX7 and MYF5 were increased in DELTA1-activated NOTCH myoblast cultures (Fig. S7A-H). This is consistent with the decreased expression of MYOD and MYOG transcripts and the myosin protein previously described in DELTA1-activated NOTCH limbs of chicken embryos (Bonnet et al., 2010; Delfini et al., 2000). In both NOTCH loss- and gain-of-function conditions, the expression of the NOTCH target gene HEYL was accordingly dysregulated in myoblasts, whereas that of BMP signalling target genes, such as ID1 and ID2, were not modified (Fig. S8G,H), showing that BMP signalling, previously shown to interact with NOTCH in other systems (Guo and Wang, 2009), is not involved in this process. DELTA1-activated NOTCH increased the percentage of PAX7+ cells in proliferative myoblast cultures (Fig. S8I) and in chicken limbs without an increase in muscle area after 2 days of DELTA1/RCAS exposure (Fig. S7A-H) compared with respective controls. These results, combined with literature data, show that cultures of chicken foetal myoblasts mimic in vivo myogenesis and that NOTCH loss and gain of function lead to similar outcomes for muscle differentiation in vivo and in vitro.
NOTCH loss of function in differentiating myoblast cultures increases terminal differentiation and fusion
In order to assess the effect of NOTCH inhibition on the later steps of myogenesis, DAPT-mediated NOTCH loss of function was performed on differentiating myoblast cultures (Fig. 7A-E). DAPT-mediated NOTCH inhibition favoured the appearance of myotubes and a decrease in the percentage of PAX7+ cells compared with controls, with no change in the nuclei number (Fig. 7A-C). Consistent with this, the expression of the muscle differentiation markers MYOG and MYHC was increased, whereas that of PAX7 was decreased, in DAPT-treated cultures compared with control cultures (Fig. 7D). TMEM8C expression was also increased in DAPT-treated cultures, whereas the NOTCH target gene HEYL was downregulated (Fig. 7D). Consistent with the increase of TMEM8C and MYHC mRNA levels (Fig. 7D), the myotubes were larger (Fig. 7A,B) and the fusion index was significantly increased in DAPT-treated differentiating myoblast cultures compared with controls (Fig. 7E). These results show that NOTCH inhibition increases the terminal differentiation and fusion processes in foetal myoblasts cultured in differentiation conditions.
NOTCH loss of function in differentiated MYOG+ myoblasts increases fusion
In the previous NOTCH inhibition experiments (Fig. 7A-E), we could not dissociate muscle cell differentiation from fusion. In order to uncouple differentiation from fusion, we used a two-step protocol whereby proliferating myoblasts were first pushed to differentiate without fusion and then allowed to fuse in a second step (Fig. 7F). Based on previously described protocols (Girardi et al., 2021; Latroche et al., 2017), we induced differentiation of myoblasts, seeded at a sub-confluent concentration to avoid cell-cell contact and thereby fusion, by culturing them in a low serum-containing medium (Fig. 7F). As expected, after this procedure, myoblasts lost PAX7 expression and became MYOG+ (Fig. 7G-I). MYOG+ differentiated myoblasts (myocytes) were then plated at high density and treated with either DAPT or DMSO for 48 h to assess the effect of NOTCH inhibition on the fusion process of already-differentiated myoblasts (Fig. 7F). DAPT-mediated NOTCH inhibition favoured myotube formation compared with controls (Fig. 7J,K). The fusion index was significantly increased in the context of NOTCH inhibition compared with control cultures (Fig. 7L). Consistent with this, the expression of TMEM8C increased with DAPT exposure, whereas the expression of MYOD and MYOG did not significantly increase compared with controls in this two-step-culture system (Fig. S9). Taken together, these results show that NOTCH inhibition in differentiated myoblasts promotes fusion independently of differentiation in vitro.
HEYL binds to the promoter regions of TMEM8C in chick limbs and myoblast cultures
In order to define the molecular mechanism through which NOTCH inhibition promotes myoblast fusion, we analysed the recruitment of the NOTCH target gene HEYL to the TMEM8C promoter regions previously described (Luo et al., 2015). Because the HEY/HES are bHLH transcriptional repressors known to inhibit myogenesis via direct binding to E-box-containing regions of the MYOD enhancer (Zalc et al., 2014), we tested whether HEYL could bind to the three E-box-containing regions in regulatory regions of the fusion gene TMEM8C (Fig. 8A). In order to test this hypothesis, we performed ChIP experiments in samples from DAPT-treated cultures of differentiated myoblasts (two-step protocol) and immobilized limbs (as the phenotype observed with DMB was similar but stronger than that following DELTA1-DN electroporation). We found that in the control for both the myoblast culture and the limb, the transcriptional repressor HEYL was recruited to the three E-box-containing regions located upstream of the TMEM8C gene (Fig. 8B,C). In addition, NOTCH inhibition led to decreased binding of HEYL to these regions compared with control conditions (Fig. 8B,C). This result is consistent with the increase in TMEM8C expression and fusion index in the DAPT-treated myoblast cultures (Fig. 7) and in NOTCH loss-of-function experiments on limb foetal muscles (Fig. 6). The decrease of HEYL recruitment to the TMEM8C promoter in the absence of NOTCH activity in vivo and in vitro provides a potential mechanism for the fusion-inhibiting effect of NOTCH signalling pathway (Fig. 8D).
In the present study, we identified a regionalized location for the TMEM8C fusion gene transcripts and for the fusion-competent MYOG+ cells in limb foetal muscles of chicken embryos and established a molecular link between the NOTCH pathway and TMEM8C expression that could underlie the fusion-inhibiting effect of NOTCH.
Whether the fusion process occurs at specific places within the skeletal muscles is not known. We identified a preferential central location for TMEM8C transcripts and fusion-competent MYOG+ cells in skeletal muscles during foetal myogenesis in chicken (Fig. 8E). The similar location of MYOG+ cells and TMEM8C transcripts is fully consistent with the transcriptional regulation of TMEM8C expression by MYOG (Ganassi et al., 2018; Luo et al., 2015; Millay et al., 2013). This regionalized expression suggests that fusion would preferentially occur in the central region of the muscle during chicken foetal development. Because myomaker/TMEM8C is required in both fusing cells (Millay et al., 2013) and TMEM8C is preferentially expressed in fusion-competent cells outside muscle fibres (Fig. S4), two scenarios for fusion are possible. In the first, mononucleated fusion-competent cells fuse to form de novo fibres. In the second, mononucleated fusion-competent cells fuse to existing fibres that rapidly downregulate TMEM8C expression.
In apparent contradiction, based on bromodeoxyuridine (BrdU) incorporation experiments, myoblast fusion has been suggested to occur preferentially at muscle tips in rat and mouse muscles during foetal and perinatal growth (Gu et al., 2016; Kitiyakara and Angevine, 1963; Zhang and McLennan, 1995). However, these BrdU incorporation experiments did not address specifically myoblast fusion, but rather cell fusion. One cannot exclude that these BrdU experiments revealed fibroblast fusion preferentially at muscle tips, as fibroblasts can be recruited to the myogenic fate at the muscle-tendon interface (Esteves de Lima et al., 2021; Yaseen et al., 2021). Furthermore, a preferential location of proliferating PAX7+ cells at muscle tips close to tendons has been shown, indicating that the muscle-tendon environment favours a proliferative state of the myoblasts rather than myoblast differentiation and fusion (Esteves de Lima et al., 2014).
Because NOTCH inhibition is a potent trigger of muscle differentiation, the role of NOTCH in muscle fusion has been neglected. We show now that expression of the gene encoding NOTCH ligand JAG2 is preferentially localized in the central muscle regions and excluded from the tips (Fig. 8E), similar to the regionalization observed for the MYOG+ cells and TMEM8C transcripts. We further demonstrate that NOTCH inhibition increased myoblast fusion in addition to differentiation in foetal myoblast cultures and limb muscles. These results suggest that the endogenous levels of JAG2 in the central muscle regions are potentially required to regulate the NOTCH ON (HEYL+) and NOTCH OFF (HEYL−) states in differentiated myoblasts (Fig. 5, Fig. S6) and, consequently, the flux of fusion events in this zone. This is consistent with the myotube hypertrophy observed in mouse C2.7 myoblasts upon NOTCH inhibition (Kitzmann et al., 2006). Many of the molecules identified as being involved in myoblast fusion can be linked to NOTCH signalling. The calcium-activated transcription factor NFATC2, recognized to control myoblast fusion after the initial formation of myotubes (Horsley et al., 2003), has the ability to suppress NOTCH transactivation and the expression of Hey genes (Zanotti et al., 2013). Shisa2, an endoplasmic reticulum (ER)-localized protein that regulates the fusion of mouse myoblasts via Rac1/Cdc42-mediated cytoskeletal F-actin remodelling, is repressed by NOTCH signalling (Liu et al., 2018). SRF, which was identified as a regulator of satellite cell fusion via the maintenance of actin cytoskeleton architecture (Randrianarison-Huetz et al., 2018), physically interacts with the NOTCH target gene Herp1, which interferes with SRF transcriptional activity (Doi et al., 2005). Recently, it has been reported that TGFβ inhibition promotes muscle cell fusion in chicken embryos and adult mouse muscles by modulating actin dynamics (Girardi et al., 2021; Melendez et al., 2021). Interestingly, crosstalk has been identified between TGFβ and NOTCH intracellular signalling pathways, leading to functional synergism for both pathways. TGFβ cooperates with NOTCH to induce Hes1, Hey1 and Jag1 expression in a Smad3-dependent manner through a Smad3–NICD interaction in different systems (Blokzijl, 2003; Zavadil et al., 2004). Given this positive interaction, we cannot exclude the possibility that TGFβ inhibition interferes with NOTCH signalling or vice versa during myoblast fusion. NOTCH decay has been also involved in the fusion of fusion-competent myoblasts into myotubes in adult Drosophila (Gildor et al., 2012), suggesting a conserved involvement of NOTCH inhibition in myoblast fusion in invertebrates and vertebrates.
In addition to showing a fusion-inhibiting effect of NOTCH in vertebrates, we established a molecular link between HEYL and the fusion gene TMEM8C (myomaker). In control conditions, the recruitment of the transcriptional repressor HEYL to the E-box-containing regions of the TMEM8C promoter could potentially be the basis for the absence of fusion ability of muscle progenitors displaying active NOTCH. In the context of NOTCH inhibition, the expression of HEYL is decreased and, consequently, HEYL recruitment to TMEM8C promoter regions is lost, which results in release of the inhibition of TMEM8C expression (Fig. 8D). The increase of TMEM8C expression is likely to be the molecular mechanism underlying the increased fusion observed in NOTCH inhibition conditions. This also indicates that during normal development, when NOTCH activity is decreased in muscle cells, muscle fusion is promoted in addition to differentiation. The myogenic differentiation factors MYOD and MYOG also bind to the E-box domain located close to the transcription start site of myomaker to positively regulate its expression in mouse (Mymk), chicken (TMEM8C) and zebrafish (mymk) (Ganassi et al., 2018; Luo et al., 2015; Millay et al., 2014). One attractive mechanism could be a competition between the transcriptional repressor HEYL and the transcriptional activator MYOG for occupancy of the TMEM8C promoter regions. In the context of active NOTCH, MYOG is not present and HEYL represses TMEM8C transcription, whereas in the context of NOTCH inhibition, HEYL is decreased and MYOG is present and activates TMEM8C transcription. This provides a possible new mechanism for tuning myoblast differentiation and fusion during development.
MATERIALS AND METHODS
Fertilized chicken (Gallus gallus) eggs from commercial sources (White Leghorn strain: HAAS, Strasbourg, France; JA 57 strain: Morizeau, Dangers, France) were incubated at 38.5°C in a humidified incubator until appropriate stages. Embryos were staged according to the number of days in ovo (E). All experiments on chicken embryos were performed before E14 and consequently are not submitted to a licensing committee, in accordance with European guidelines and regulations.
scRNAseq analysis of whole-limb cells
The scRNAseq protocol and sample collection was performed as described by Esteves de Lima et al. (2021). Briefly, scRNAseq datasets were generated from whole forelimbs from three E6 embryos and three E10 embryos using the 10X Chromium Chip (10X Genomics) followed by sequencing with a HighOutput flowcel using an Illumina Nextseq 500 and by sequence analysis with Cell Ranger Single Cell Software Suite 3.0.2 (10X Genomics). Only mononucleated muscle cells are included in the datasets, as plurinucleated myotubes are excluded by the single-cell isolation protocol. The Seurat package (v3.0) (Stuart et al., 2019) under R (v3.6.1) (R Core Team, 2019) was used to perform downstream clustering analysis on scRNAseq data (Macosko et al., 2015). We then extracted the clusters identified as muscle clusters by the differential expression of classical myogenic markers (PAX7, MYOD, MYOG, MYHC) and performed the rest of the analysis on these muscle clusters only. Gene expression is defined by ‘gene log-normalized count>0’. The scRNAseq datasets were analysed using Seurat tools such as Feature plots and Violin plots. The R package ggplot2 v3.3.3 (Wickham, 2016) was used to generate custom feature plots highlighting gene co-expression. Population intersection plots were generated with the R package UpSetR v1.4.0 (Conway et al., 2017).
Chemical inhibitor administration
DMB (Sigma-Aldrich) solution and control Hank's solution were prepared as previously described (Esteves de Lima et al., 2016). Briefly, 100 µl of DMB or control solutions were administered in chicken embryos at E7.5 and E8.5 and embryos were collected at E9.5.
Grafts of DELTA1/RCAS-expressing cells
Chicken embryonic fibroblasts obtained from E10 chicken embryos were transfected with DELTA1/RCAS (Delfini et al., 2000) using the Calcium Phosphate Transfection Kit (Invitrogen). Cell pellets of approximately 50-100 µm in diameter were grafted into limb buds of E4.5 embryos as previously described (Bonnet et al., 2010; Delfini et al., 2000). DELTA1/RCAS-grafted embryos were harvested at E6.5 or E9.5. Embryos treated with DMB or control solutions at E7.5 and E8.5 were collected at E9.5.
Forelimb somite electroporation was performed as previously described (Wang et al., 2011). The DNA solution was composed of the pT2AL-MLC-Tomato-DELTA1/DN or the pT2AL-MLC-Tomato (control) vectors and a transient transposase-containing vector CMV/βactin-T2TP (Bourgeois et al., 2015; Esteves de Lima et al., 2016). This vector set allows the stable integration of the MLC-Tomato-DELTA1-DN or the MLC-Tomato cassettes into the chick genome. The concentration of each vector was 2 μg/μl and 600 ng/μl for the CMV/βactin-T2TP. DNA was prepared in solution containing carboxymethyl cellulose 0.17% (Sigma-Aldrich), Fast Green 1% (Sigma-Aldrich), MgCl2 1 mM (VWR Chemicals) and 1× PBS in water.
Primary myoblasts were obtained from hindlimbs of E10 chicken embryos and cultured in proliferating high-serum-containing medium (10%) or in differentiation low-serum-containing medium (2%), as previously described (Havis et al., 2012). For NOTCH gain-of-function experiments, myoblasts were transfected with DELTA1/RCAS plasmid or Empty/RCAS (control). We performed NOTCH loss-of-function experiments in myoblasts using the γ-secretase inhibitor DAPT (Sigma-Aldrich). For NOTCH loss-of-function experiments, myoblasts were treated with DAPT (Sigma-Aldrich) at a concentration of 5 µM for 24 h in low- and high-serum conditions (proliferation and differentiation assays, respectively) or DMSO (Fluka) (controls). For the fusion assay, MYOG+ myoblasts plated at a high density in low-serum conditions (Girardi et al., 2021; Latroche et al., 2017) were treated with 5 µM DAPT (Sigma-Aldrich) at 0 h and 24 h of culture and collected at 48 h.
Forelimbs of control and manipulated (DMB, DELTA1/RCAS) chicken embryos were fixed in 4% paraformaldehyde (Sigma-Aldrich) overnight at 4°C and then processed in 7.5%/15% gelatin/sucrose (Sigma-Aldrich) for 12 µm cryostat sections as previously described (Esteves de Lima et al., 2014). Monoclonal antibodies for PAX7 and MF20 developed by D. A. Fischman and A. Kawakami, respectively, were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology Iowa City, IA 52242, USA. The Collagen XII and MYOG antibodies were kindly provided by Manuel Koch (Germany) (Koch et al., 1992) and Christophe Marcelle (France) (Manceau et al., 2008), respectively. The MYOD (BD Biosciences) and pSMAD1/5/9 (Cell Signaling Technology) antibodies were obtained from commercial sources. Secondary antibodies were conjugated with Alexa 488, Alexa 555 or Alexa 647 (Invitrogen). Nuclei were visualized with DAPI (Sigma-Aldrich) staining. Detailed information for antibodies is provided in Table S1.
In situ hybridization
Chicken forelimbs of control and manipulated (DMB, DELTA1/RCAS, DELTA1-DN) embryos were fixed in 4% paraformaldehyde (Sigma-Aldrich) overnight at 4°C and processed for in situ hybridization on wax tissue or cryostat sections, as previously described (Esteves de Lima et al., 2016). Digoxigenin-labelled mRNA probes were prepared as previously described: MYOD, JAG2 and DLL1 (Delfini et al., 2000), FGF4 and SCX (Edom-Vovard et al., 2001), MYOG (Bonnet et al., 2010). The TMEM8C probe was obtained by PCR from E9.5 limb tissues using the primers 5′-ACCCTCAGCACTTTGGTCTTT-3′ (forward) and 5′-ACAGGGCACACCCCATACA-3′ (reverse), cloned into the pCRII-TOPO vector (Invitrogen), linearized with NotI (New England Biolabs) and synthesised with SP6 (Promega). Fluorescence in situ hybridization was performed according to Wilmerding et al. (2021).
RNA extraction and RT-qPCR
Total RNAs were extracted from control limbs, experimental limbs or primary foetal myoblast cultures and 500 ng to 1 µg of RNA was reverse transcribed using the High-Capacity Retrotranscription kit (Applied Biosystems). RT-qPCR was performed using SYBR Green PCR Master Mix (Applied Biosystems). Primer sequences used for RT-qPCR are listed in Table S2. The relative mRNA levels were calculated using the 2−ΔΔCt method (Livak and Schmittgen, 2001; Schmittgen and Livak, 2008). The ΔCts were obtained from Ct normalized with GAPDH and RPS17 levels in each sample. Each RNA sample was analysed in duplicate.
In vitro (Harada et al., 2018) and in vivo (Havis et al., 2006) ChIP assays were performed as previously described. Limbs from E9.5 chicken embryos were homogenized using a mechanical disruption device (Lysing Matrix A, Fast Prep MP1, 2×40 s at 6 m/s, Machine MP Bio) and 5 µg of the HEYL antibody, kindly provided by So-Ichiro Fukada (Osaka University, Japan) (Fukada et al., 2007), were used to immunoprecipitate 10 µg of sonicated chromatin. ChIP products were analysed by RT-qPCR to amplify the target regions. Primers are listed in Table S2.
After immunohistochemistry or in situ hybridization experiments, images were obtained using a Zeiss apotome epifluorescence microscope, a Leica DMI600B fluorescence microscope or a Leica SP5 confocal system.
Image analyses and quantification
For the distribution of MYOG+ and PAX7+ cells in the middle versus the tips of muscle, the MYOG/DAPI or PAX7/DAPI percentage is the number of MYOG+ or PAX7+ nuclei in each region divided by the total number of DAPI+ nuclei in that region. The number of PAX7+ and MYOG+ cells in control and DAPT foetal myoblast cultures was normalized to the number of total DAPI+ nuclei. The fusion index is the number of nuclei within myotubes (MF20+) (at least two nuclei) divided by the total number of DAPI+ nuclei in the field. For the distribution of the number of myonuclei per fibre, the fibre percentage is the number of fibres with a given number of myonuclei divided by the total number of fibres. All the described quantifications were performed using the Cell counter plug-in of the free ImageJ or Fiji software (Schneider et al., 2012).
Data were analysed using the non-parametric, two-tailed, Mann–Whitney test using GraphPad Prism V6 software, except for the ChIP-RT-qPCR samples, which were analysed by two-tailed, paired t-test. Results are shown as mean±s.d. No data were excluded. No randomization was allocated between the samples because we compared control and treated animals/samples. Blinding was used to perform image acquirement and quantification.
We thank lab members Matthew Borok, Philippos Mourikis, Sonya Nassari and Valentina Taglietti for critical reading of the manuscript. We thank So-Ichiro Fukada, Manuel Koch and Christophe Marcelle for reagents. We thank Sophie Gournet for illustrations.
Conceptualization: J.E.d.L., D.D.; Methodology: J.E.d.L., C.B., M.-A.B., E. Hirsinger, E. Havis, D.D.; Software: E. Hirsinger; Validation: J.E.d.L., C.B., M.-A.B., E. Hirsinger, E. Havis, D.D.; Formal analysis: J.E.d.L., E. Hirsinger, E. Havis, D.D.; Investigation: J.E.d.L., C.B., M.-A.B., E. Havis, D.D.; Resources: J.E.d.L., C.B., M.-A.B., E. Hirsinger, E. Havis, F.R., D.D.; Data curation: J.E.d.L., E. Hirsinger, D.D.; Writing - original draft: J.E.d.L., D.D.; Writing - review & editing: J.E.d.L., E. Hirsinger, E. Havis, F.R., D.D.; Visualization: J.E.d.L., C.B., M.-A.B., E. Hirsinger, E. Havis, D.D.; Supervision: F.R., D.D.; Project administration: J.E.d.L., D.D.; Funding acquisition: D.D.
The work was supported by the Association Française contre les Myopathies (MyoFibro 22234), the Agence Nationale de la Recherche (ANR-20-CE13-0020-01), the Fondation pour la Recherche Médicale (DEQ20140329500), the Centre National de la Recherche Scientifique, Sorbonne University and the Institut National de la Santé et de la Recherche Médicale. J.E.d.L. was part of the MyoGrad International Research Training Group for Myology and was supported by the Deutsche Forschungsgemeinschaft and the Association Française contre les Myopathies. J.E.d.L. was then supported by a Labex REVIVE post-doctoral fellowship.
scRNAseq data have been deposited in the NCBI Gene Expression Omnibus database under accession number GSE166981.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199928.
The authors declare no competing or financial interests.