Cytoplasmic NOTCH and membrane-derived β-catenin link cell fate choice to epithelial-mesenchymal transition during myogenesis

How cells in the embryo coordinate epithelial plasticity with cell fate decision in a fast changing cellular environment is largely unknown. In chick embryos, skeletal muscle formation is initiated by migrating Delta1-expressing neural crest cells that trigger NOTCH signaling and myogenesis in selected epithelial somite progenitor cells, which rapidly translocate into the nascent muscle to differentiate. Here, we uncovered at the heart of this response a signaling module encompassing NOTCH, GSK-3β, SNAI1 and β-catenin. Independent of its transcriptional function, NOTCH profoundly inhibits GSK-3β activity. As a result SNAI1 is stabilized, triggering an epithelial to mesenchymal transition. This allows the recruitment of β-catenin from the membrane, which acts as a transcriptional co-factor to activate myogenesis, independently of WNT ligand. Our results intimately associate the initiation of myogenesis to a change in cell adhesion and may reveal a general principle for coupling cell fate changes to EMT in many developmental and pathological processes. DOI: http://dx.doi.org/10.7554/eLife.14847.001


Introduction
During early embryogenesis, a succession of extensive tissue rearrangements and cell migration events, intimately associated with rapid cell fate changes, establish the tissues and organs of the future adult. A model where such complex issues are amenable to experimentation is the early formation of skeletal muscles in the chick embryo.
Over many days of development, the medial border of the dermomyotome (DML) generates the first skeletal muscle cells that assemble into a primary myotome ( Figure 1A-D). This arises from a crucial cell fate decision: epithelial cells in the DML either self-renew or undergo myogenic differentiation and this is accompanied by an EMT that allows their translocation into the primary myotome (Denetclaw et al., 1997;Gros et al., 2009;Ordahl et al., 2001;Rios et al., 2011).
Classical experiments demonstrated that signaling cues from tissues surrounding the somites act as inducers of muscle formation (Ordahl and Le Douarin, 1992). It was later shown that WNT1 and WNT3a are expressed in the dorsal neural tube and that, in vitro, both ligands enhance myogenesis in somites (Munsterberg et al., 1995;Stern et al., 1995;Tajbakhsh et al., 1998). A major consequence of the activation of the 'canonical' WNT pathway by WNT ligands is that b-catenin, normally degraded by the APC/Axin destruction complex, accumulates in the cytoplasm and enters the nucleus, where it engages DNA-bound TCF transcription factor to activate their targets (hence the name of WNT/b-catenin-dependent pathway to this cellular response). Since it was demonstrated that MYF5 promoter contains TCF binding sites required for its in vivo expression in the DML (Borello et al., 2006) and that dominant negative and constitutively active forms of TCF and b-catenin modulate MYF5 expression in somites (Abu-Elmagd et al., 2010;Gros et al., 2009), this has led to the largely accepted theory that myogenesis in somites is under the control of WNTs from the dorsal neural tube acting through a WNT/b-catenin-dependent pathway.
This raises a paradox: although presumably all cells in the DML are equally exposed to WNTs from axial structures and are fully competent to initiate myogenesis, only a small proportion of DML cells do so at any given time (Rios et al., 2011), suggesting that other mechanisms affect the cell fate decision in selected DML cells.
We recently introduced NOTCH signaling as a novel player in early myogenesis, by demonstrating that the activation of MYF5 and MYOD in the DML is dependent upon the transient activation of NOTCH signaling, triggered by Delta1-positive neural crest cells migrating from the dorsal neural tube (Rios et al., 2011). This finding is compatible with the well-documented role of the dorsal neural tube in myogenesis. Importantly, the mosaic expression of Delta1 in the migrating neural crest cell population ensures that NOTCH signaling is regularly triggered in selected DML cells, thus explaining the cell fate choice necessary to generate myotomal cells over an extended period of time, all while self-renewing the progenitor population. However, this finding was not in accordance with the established role of WNT in myogenesis, nor did it explain why the initiation of the myogenic program in selected progenitors within the DML is associated with their translocation into the primary myotome via an EMT.
Here, we discovered that in DML cells, the activation of NOTCH signaling by Delta1-positive neural crest cells strongly decreases GSK-3b activity, independent of NOTCH transcriptional function. This leads to a dramatic stabilization of SNAI1, resulting in the initiation of the EMT program. As a consequence, b-catenin from the cell membrane pool is mobilized, allowing its entry into the nucleus where it activates MYF5 expression in a WNT ligand-independent manner. Therefore, our results suggest that non-canonical functions of NOTCH and b-catenin, associated in an efficient signaling circuitry, explain the coupling of myogenesis with changes in cell adhesion in the DML.

Co-activation of NOTCH and WNT reporters in early myogenesis
The first sign that myogenesis is initiated in the DML is the activation of MYF5 (or MYOD) expression, which serves as a read-out of a cell fate change in this structure (Rios et al., 2011). As WNT and Delta1 are reported to activate myogenesis (Munsterberg et al., 1995;Rios et al., 2011;Stern et al., 1995;Tajbakhsh et al., 1998), it was important to determine whether and how these pathways co-operate in this process. As a first step, we co-electroporated the DML of trunk-level somites (as shown in Figure 1A-D) with reporter constructs for the NOTCH and WNT pathways. The NOTCH reporter (Ohtsuka et al., 2006) contains the HES1 promoter region upstream of a destabilized red fluorescent protein. HES1 is a direct target of the NOTCH pathway, and we previously showed this construct serves as a faithful reporter of NOTCH activity in somites (Rios et al., 2011). The 'TOPflash' reporter (Korinek et al., 1997;Rios et al., 2010) contains twelve TCF-1 binding sites upstream of a destabilized EGFP (12Tf-d2EGFP). The activity of the TOPflash reporter is triggered by the binding of the transcription factor TCF/LEF together with its co-factor b-catenin. The destabilized fluorescent reporter proteins present in both constructs allow the visualization of only the cells that are actively engaged in NOTCH and WNT signaling, while the intensity of the signal indicates the magnitude of the response.
We found that, within the DML, almost half of the epithelial cells that activated the TOPflash reporter were also positive for the NOTCH reporter (Figure 2A,B). Remarkably, nearly all DML cells that were positive for both reporters were MYF5-positive ( Figure 2C), indicating a strong correlation between the activation of NOTCH and WNT reporters and the initiation of the myogenic program.
During this analysis, we made the surprising observation that a large majority of NOTCH reporter-positive cells displayed a high TOPflash fluorescence. On the contrary, most cells that expressed low levels of TOPflash fluorescence were NOTCH-reporter negative (Figure 2A,D). It was important to show that this difference was not due to electroporation, since DNA transfer with this technique is inherently variable between cells. To show that the variable levels of TOPflash activity were not the result of variable amounts of electroporated plasmids, we plotted the intensity of TOPflash response against that of the BFP as a control. BFP is driven by an ubiquitous promoter and its fluorescence should thus be directly correlated with the quantity of plasmid incorporated in each cell. We found no correlation between the intensity of the TOPflash reporter and that of BFP ( Figure 2E). This suggests that the low and high levels of TOPflash fluorescence observed in DML cells are due to genuine differences in TCF/b-catenin transcriptional activity. The graph also distinguishes cells that are positive (in red) for the NOTCH reporter from those that are negative (in blue, Figure 2E). It confirms that the vast majority of NOTCH reporter-positive (and MYF5-positive) cells display a high TCF/b-catenin transcriptional activity. This is the cell population that will be thoroughly analyzed in this study. The identity of the cell population that displays a low TCF/b-catenin transcriptional activity (about 60% of the entire TOPflash positive cell population) but is NOTCH reporterand MYF5-negative will become clear in the following chapters.

NOTCH signaling regulates TCF/b-catenin transcriptional activity during myogenesis
The results above raised the possibility that NOTCH and the TCF/b-catenin transcriptional activity may be interconnected during the initial phases of myogenesis. To address this, we first co-electroporated the DML with a constitutively active form of b-catenin (Rios et al., 2010) together with the NOTCH reporter. This did not lead to any change in activity of the NOTCH reporter, compared to controls ( Figure 3A-C). In contrast, the co-expression of a constitutively active form of NOTCH1 (NICD [Daudet and Lewis, 2005]) led to rapid and robust increase in the activity of the TOPflash reporter, compared to controls ( Figure 3D,E). To further investigate the interrelation between both pathways, we used a dominant-negative form of the NOTCH co-activator Mastermind (DN-MAML1 [Weng et al., 2003]). We previously showed that it strongly inhibits MYF5 expression in the DML (Rios et al., 2011). A constitutively active form of the most downstream effector of the WNT pathway, LEF1/TCF1 (named CA-LEF1 [Abu-Elmagd et al., 2010]), co-electroporated with DN-MAML1 not only rescued the inhibition of MYF5 expression by DN-MAML1, but also led to a strong activation of its expression that was indistinguishable from that observed after electroporation of CA-LEF1 alone ( Figure 3F,G) or NICD (Rios et al., 2011). These data support the premise that NOTCH regulates TCF/b-catenin transcriptional activity in the DML. Immunostaining of a HH 15 chick embryo with PAX7 (in green), HNK1 (in blue) and MYF5 (in red) to label the dermomyotome, the neural crest and the myotome, respectively. (C) Schematic illustrating regions of the somite that are represented in the confocal stacks of images shown throughout the study. It shows the medial portion of a somite 6-24 hr after its DML was electroporated. Typically, it leads to the mosaic expression (in green) of the electroporated construct(s) in the DML, the transition zone (in blue) and the nascent primary myotome (in purple). (D) is a maximum intensity projection of a confocal stack of a somite electroporated with GFP 6 hr prior to analysis. DOI: 10.7554/eLife.14847.002 The activation of TCF/b-catenin transcriptional activity by NOTCH is WNT ligand-independent The TOPflash reporter reflects TCF/b-catenin transcriptional activity and it is a known outcome of WNT ligand binding to its cognate receptor. However, the results above indicate that NOTCH somehow regulates TOPflash activity in the DML. To understand this apparent contradiction, we decided to test whether WNT ligands are at all required during this process. Dickkopf (Dkk) genes comprise an evolutionarily conserved family that encode secreted proteins that antagonize WNTs, by inhibiting in the extracellular space the WNT co-receptors Lrp5 and 6 (Niehrs, 2006). First, we tested whether DKK1 efficiently inhibits WNT signaling by co-electroporating DKK1 into the neural tube of developing embryos, together with the TOPflash reporter described above. As control, the TOPflash reporter was electroporated alone. Without DKK1, we observed a robust TOPflash reporter activity in the neural tube ( We then performed a similar experiment in somites. A striking difference from the neural tube control experiment was that, even in the presence of DKK1, the TOPflash reporter was still active ( Figure 4A,B). While this response showed a decrease of about half in the proportion of TOPflash reporter-positive cells compared to controls ( Figure 4C), remarkably nearly all cells that remained TOPflash-positive also expressed MYF5 ( Figure 4D). As a consequence, the overall proportion of MYF5-positive cells was unaffected ( Figure 4E). Of these cells that remained TOPflash-positive in the presence of DKK1, nearly all expressed the NOTCH reporter ( Figure 4F,G), a result coherent with our observation that MYF5 is tightly associated with the activation of this reporter ( Figure 2C).  These surprising results suggested that WNT ligands are dispensable for the NOTCH-dependent myogenic response in DML cells. To confirm this, we performed double electroporations ( Figure 4H). First, we electroporated WNT1 or a dominant-negative form of WNT1 into the dorsal neural tube and the neural crest. WNT1 is predicted to activate WNT1 and WNT3a cognate receptors, while DN WNT1 is believed to act as a competitive inhibitor for both WNT1 and WNT3a (Hoppler et al., 1996). Second, we electroporated the TOPflash reporter and an electroporation control plasmid in somites. This allowed us to quantify the activity of the TOPflash reporter together with the myogenic response (determined by immunostaining for MYF5). We observed that WNT1 and DN WNT1 significantly increased or decreased, respectively, the activity of the TOPflash reporter, compared to controls ( Figure 4I-L), but importantly with no visible change to their myogenic response ( Figure 4M).
These results led us to hypothesize that the overall TCF/b-catenin transcriptional activity observed in DML cells is the sum of two TCF/b-catenin transcriptional activities: the first is NOTCH-dependent, WNT ligand-independent and it is associated with myogenesis; the second is WNT ligand-dependent, but it is not associated with myogenesis. To show this, we inhibited both DLL1 in neural crest and the WNT ligands in the DML, while monitoring the activity of the TOPflash reporter and myogenesis in somites. Compared to controls (no inhibition of DLL1 in the neural crest), this led to a near complete absence of TOPflash reporter activity in DML cells ( Figure 4O,P), which was now accompanied by a robust decrease in their myogenic response ( Figure 4O,Q).
Together with the results presented in Figure 2, these data suggest that two distinct populations coexist in the DML: the first displays a strong TCF/b-catenin transcriptional activity triggered by DLL1 from migrating neural crest cells, which results in the initiation of MYF5 expression. The second shows a low TOPflash reporter activity, initiated by WNT ligand transported by neural crest cells but with no sign of myogenesis. These findings prompted a search for a mechanism that could explain how DLL1 can trigger the high TCF/b-catenin transcriptional activity that is required for the initiation of the myogenic program.

Membrane-derived b-catenin is required for the NOTCH-mediated activation of MYF5
In the DML, the initiation of myogenesis is accompanied by an EMT Rios et al., 2011). Because it integrates two distinct functions i) as a transcriptional co-factor together with TCF/ LEF and ii) as a structural adaptor protein linking cadherins to the actin cytoskeleton in cell-cell adhesion, b-catenin is a prime candidate to play a central role in these two processes. However, despite experimental evidence that the dissociation of adherens junction (AJ) can induce the release of b-catenin into the cytoplasm, a functional connection between the nuclear and membranal pools of b-catenin in vivo is still a matter of debate (Brembeck et al., 2006;Gavard and Mège, 2012;Nelson and Nusse, 2004).
To test whether b-catenin located at the AJ and at the plasma membrane of DML cells is required for the activation of MYF5, we electroporated DML cells with a mutant Y489F b-catenin that cannot be mobilized from the junctional complex (Rhee et al., 2007). As control, wild-type (WT) b-catenin was utilized. We verified that the overall levels of exogenously provided WT and mutant b-catenin were similar to those of the endogenous b-catenin ( Figure 5-figure supplement 1). In control embryos, we observed a normal proportion of MYF5-positive DML cells ( Figure 5A,C). In contrast, the electroporation of Y489F b-catenin resulted in a robust reduction of MYF5 expression ( Figure 5B,C), suggesting that junctional b-catenin is required for the early activation of myogenesis.
We then tested whether junctional b-catenin is required, downstream of NOTCH, to induce MYF5. We repeated the same experiment as above, this time in combination with NICD. Y489F or WT b-catenin were expressed for 10 hr in the DML, after which (using a Tet-on inducible system, see Materials and methods) NICD expression was induced for 6 hr ( Figure 5D). In control embryos, the induction of NICD expression resulted in a robust increase in overall (nuclear and cytoplasmic) b-catenin protein levels in electroporated cells (as determined by immunostaining for the Myc tag present on the exogenous WT b-catenin as well as by staining for endogenous b-catenin [ Figure 5E and Figure 5-figure supplement 2]). In these embryos, the vast majority of NICD+/WT b-catenin+ cells expressed MYF5 as expected ( Figure 5E,G). The myogenic response in embryos electroporated with Y489F b-catenin was strikingly different, and the presence of the mutant form of b-catenin not only prevented the activation of myogenesis by NICD, but also resulted in a dramatic reduction of MYF5 expression ( Figure 5F,G) that was similar to the decrease observed when Y489F b-catenin was electroporated alone ( Figure 5C). As reported (Rhee et al., 2007), Y489F b-catenin remained mainly distributed at the membrane and AJ of DML cells ( Figure 5-figure supplement 2), suggesting that it did not mobilize from the junctional complex.
Together with the observation that MYF5 expression is associated with high activity of the TOPflash reporter in DML cells (Figures 2 and 4), our data suggest that the pool of b-catenin at the AJ and along the plasma membrane of DML cells acts as a reservoir for the NOTCH-mediated, strong b-catenin-dependent response required for the activation of MYF5.

SNAI1 is a necessary and sufficient step for the NOTCH-dependent activation of MYF5
Major players in cell adhesion are the zinc finger transcriptional repressors Snail (or SNAI) family members. An essential function of SNAI is to repress epithelial gene expression and thereby promote epithelial-mesenchymal transitions (EMT) during vertebrate and invertebrate development and metastatic progression of cancers (Barrallo-Gimeno and Nieto, 2005). In early chick embryos, SNAI1 is expressed in the epithelial dermomyotome, including the DML. Its activation leads to the down-regulation of N-cadherin expression present at the adherens junction of epithelial cells located in the central dermomyotome and to their subsequent EMT (Delfini et al., 2009). Coherent with this, the over-expression of SNAI1 in the DML induced a significant increase in the translocation of DML cells into the myotome ( Figure 6A,B), while the inhibition of its function by a SNAI1-specific siRNA (Delfini et al., 2009) inhibited their translocation ( Figure 6C,D). Since we showed above that the release of b-catenin from the AJ is a key event in the activation of MYF5, we tested whether SNAI1 could regulate the activation of MYF5 expression in the DML. The electroporation of SNAI1 in the DML induced a strong activation of MYF5 expression, compared to controls ( Figure 6F,G). Conversely, the electroporation of a dominant-negative form of SNAIL-1 or an siRNA directed against SNAI1 (Delfini et al., 2009) resulted in a significant decrease in MYF5 expression ( Figure 6C,E,F,G). Finally, we co-electroporated NICD together with the dominant-negative form of SNAI1 and observed that the massive increase in MYF5 expression observed after NICD expression (Rios et al., 2011) was profoundly reduced by DN-SNAI1 ( Figure 6F,G), thus demonstrating that SNAI1 activation is a necessary step, downstream of NOTCH, in the chain of events that leads to the activation of MYF5 in DML cells.

NOTCH regulates SNAI1 degradation through inhibition of GSK-3b activity
We then determined whether NOTCH regulates SNAI1. NICD regulates the transcription of many target genes (Fortini, 2009;Guruharsha et al., 2012), including SNAIL (Grego-Bessa et al., 2004). Therefore, we wondered whether this would be the case in the DML. However, the expression of NICD did not result in any significant increase in SNAI1 transcription (as judged by in situ hybridization, data not shown). These results led us to explore alternative mechanisms. SNAI1 mRNA is widely expressed; however, its activity is tightly regulated post-translationally through phosphorylation by GSK-3b, which leads to its b-Trcp-mediated ubiquitination and degradation. Due to this continuous degradation, SNAI1 protein has a short half-life, estimated to be about 25 min (Zhou et al., 2004). We first determined whether NOTCH signaling could modify the stability of SNAI1 in DML cells by electroporating SNAI1 fused to GFP (SNAI1-GFP), alone or in combination with NICD. SNAI1-GFP was used in low concentration, leading to no visible phenotype, e.g. EMT or activation of MYF5, data not shown). In control embryos, GFP was undetectable under UV examination, but faintly visible after immunostaining and confocal examination in electroporated cells ( Figure 7A). In sharp contrast, the co-electroporation of NICD with SNAI1-GFP led to a massive increase in GFP staining in most electroporated cells ( Figure 7A), suggesting that the activation of NOTCH signaling prevents the degradation of SNAI1 that normally occurs in DML cells.
Since GSK-3b is the main regulator of SNAI1 stability (Barrallo-Gimeno and Nieto, 2005;Weng et al., 2003), we tested whether NOTCH regulates the activity of GSK-3b. Taelman and colleagues (Taelman et al., 2010) cleverly designed a GSK-3b fluorescent 'biosensor' (see Materials and methods). When GSK-3b is active, the biosensor fluorescence is reduced; in contrast, when GSK-3b is inactive, its fluorescence is increased. We first determined whether the GSK-3b biosensor is active in DML cells, and observed that fluorescence of the biosensor was only detected in a small proportion of electroporated DML cells ( Figure 7B). This observation indicates that in the majority of DML cells under normal, unperturbed conditions, GSK-3b is active (i.e. low or no biosensor detectable). However, in a minority of cells, GSK-3b is inhibited (i.e. high level of biosensor). To find out the identity of the biosensor-positive cells in normal, unperturbed condition, we performed a series of co-electroporations followed by immunostaining. This showed that the cells in which the fluorescence of the biosensor is observed are those that i) activate the NOTCH reporter ( Figure 7C, D), ii) in which SNAI1 is active ( Figure 7F,G) and iii) are MYF5-positive ( Figure 7C,E). This reinforces the hypothesis of a mechanistic link between the initiation of myogenesis, NOTCH activation, the stability of SNAIL and the inhibition of GSK-3b activity in DML cells.
We then tested whether NOTCH regulates GSK-3b by co-electroporating the biosensor together with NICD. This resulted in a massive increase in the fluorescence of the GSK-3b biosensor ( Figure 7B), suggesting that NICD inhibits GSK-3b kinase activity. To eliminate the possibility that the increase in fluorescence of the biosensor was due to a non-physiological effect of NICD expression in DML cells, we induced the activation of NOTCH signaling in the DML by electroporating DLL1 in the neural crest cell population (as described in Figure 4H), with the GSK-3b biosensor electroporated in adjacent somites. Again, we observed a strong increase in the fluorescence of the biosensor in somites when compared to controls ( Figure 7H,I). The electroporation of a dominantnegative form of GSK-3b (DN-GSK-3b [Taelman et al., 2010]) mimicked the effects of NOTCH activation on SNAI1 and MYF5, resulting in a significant increase in the proportion of DML cells undergoing EMT ( Figure 7J,K) as well as a robust increase in MYF5 expression in DML cells ( Figure 7L).
Altogether our data suggest that a direct consequence of the activation of NOTCH signaling by migrating DLL1-positive neural crest cells is an inhibition of GSK-3b activity that leads to a stabilization of the SNAI1 protein, which triggers an EMT of DML cells.

NOTCH controls myogenesis independently of its transcriptional role in the nucleus
The decrease of GSK-3b activity in DML cells could be a transcriptional response to the activation of NOTCH signaling. However, NOTCH can also act in a manner independent to RBPJ/CSL co-activation (Andersen et al., 2012;Ayaz and Osborne, 2014). To distinguish between the two possibilities, we electroporated a constitutively-active and a dominant-negative form of RBPJ (Kuroda et al., 1999) in DML cells. As expected, the DN-RBPJ and CA-RBPJ significantly inhibited or activated the NOTCH reporter, respectively, compared to controls ( Figure 8A-D). However, both constructs failed to modify the expression of MYF5 in this structure, compared to controls ( Figure 8E). These results indicate that NICD regulates myogenesis in the DML independently of its transcriptional activity with RBPJ.
We hypothesized that NICD could be mediating this activity in the cytosol. To address this, we constructed an artificial, membrane-tethered, HA-tagged NICD (CD4-NICD, see Materials and methods). Unlike wild-type NICD, which upon electroporation readily enters the nucleus and massively activates the NOTCH reporter (Rios et al., 2011), immunostaining for CD4-NICD was not observed in the nucleus and, coherent with this, it did not activate the NOTCH reporter ( Figure 9A,B). Despite this, CD4-NICD expression resulted in a massive stabilization of the  Figure 4H, only in the DML with the GSK-3b biosensor (in green) and H2B-BFP (in blue, left panels) or doubleelectroporated in the DML with the GSK-3b biosensor (in green) and H2B BFP (in blue) and in the neural tube with DLL1 under the control of a neural crest-specific promoter (right panels).  Figure 7 continued on next page GSK-3b biosensor that was indistinguishable from that obtained after electroporation of NICD ( Figure 9C,D). This suggests that the regulation of GSK-3b activity by CD4-NICD is not a transcriptional response to NOTCH signaling.
These results prompted us to test whether CD4-NICD expression would mimic the effects of NICD on SNAI1 and MYF5. Indeed, we observed that the electroporation of CD4-NICD led to i) a massive stabilization of SNAI1 protein comparable to NICD alone ( Figure 9E,F) and ii) to a robust increase of MYF5 expression in DML cells, not significantly different from the NICD alone ( Figure 9G,H). These data demonstrate that the activation of MYF5 in DML cells is the result of a cytoplasmic function of NICD that can be uncoupled from its role as a transcriptional co-activator in the nucleus.

Discussion
The findings presented here uncover a unique signaling module, triggered by a signal presented by migrating neural crest cells to selected progenitors in the somites. This signal is transduced into two distinct, but interrelated outcomes i) a cell fate decision in the receiving cells that leads to their entry into the myogenic program and ii) an EMT that allows their translocation into the myotome (Figure 10). While the key players in this module are well characterized, the demonstration that they cooperate to integrate cell fate determination and epithelial plasticity is novel.
EMT is intimately associated with essential changes in cell specification in many developmental processes. For instance, the massive and rapid cell rearrangement that takes place during neural crest formation or gastrulation in vertebrates vertebrate are associated with the acquisition of novel features that are characteristic of neural crest cells or of the forming mesoderm and endoderm, respectively (Barrallo-Gimeno and Nieto, 2005;Joubin and Stern, 1999). Inhibiting EMT in these models profoundly interferes with the associated cell fate changes. A large spectrum of secreted factors belonging to many signaling pathways have been shown to activate EMT. However, these mechanisms are unlikely to trigger at the same time the cell fate changes associated with the cell adhesion changes (Barrallo-Gimeno and Nieto, 2005;Lamouille et al., 2014). A remarkable aspect of the signaling module described here is that it targets GSK-3b, the main regulator of SNAI1 stability. In addition, aiming at GSK-3b may protect the b-catenin released during the EMT process from degradation, allowing it to accumulate sufficiently to act as a transcriptional co-factor, thereby resulting in a quasi-automatic coupling of EMT and b-catenin-dependent cell fate changes. Given the commonality of the key players described here in this paper, it is plausible that the same signaling circuitry could couple EMT to cell fate decisions in a wide variety of epithelia in normal and pathological conditions.
At both ends of this signaling module, NICD and b-catenin display unexpected functions. The 'canonical' role of NICD is to act as a co-transcriptional activator together with RBPJ. However, cytosolic interactions of NICD with various molecular partners that result in cellular responses independent of RBPJ, have been described in in vitro experimental settings (reviewed in [Andersen et al., 2014;Ayaz and Osborne, 2014]), but also in Drosophila (Le Gall et al., 2008). Our data uncover a novel in vivo function of NOTCH1 that takes place in the cytosol and leads to GSK-3b inhibition. It was shown that the transcriptional activity of NICD is modified by its phosphorylation in vitro, probably by GSK-3b (Espinosa et al., 2003;Foltz et al., 2002;Schweisguth, 2004). It is therefore not completely surprising that GSK-3b and NICD could (directly or not) interact in the DML. What is remarkable is the fact that this leads to a decrease in the overall activity of GSK-3b that has important consequences for the cell. The mechanisms for this are unknown. A hypothesis is that the response we observed results from a titration of GSK-3b kinase activity by NICD that is amplified by the stabilization of SNAI, itself a substrate for this kinase.
While it is largely accepted that b-catenin participates in adhesion and signaling functions in a mutually exclusive manner, changes in the transcriptional activity of b-catenin have also been associated with the disassembly of the junctional complex at the plasma membrane in vitro (Brembeck et al., 2006;Gavard and Mège, 2012;Nelson and Nusse, 2004). The results reported here provide direct evidence that this is also the case in vivo and demonstrate that the pool of b-catenin accumulated at the membrane and/or at the AJ has important signaling functions in addition to its established role in cell-cell adhesion.
Remarkably, we observed distinct thresholds of TCF/b-catenin transcriptional activity associated with discrete cellular outputs. It is known that WNT(s) expressed by the dorsal neural tube and transported by neural crest cells regulate the expression of WNT11 in the DML through canonical WNT signaling. WNT11 itself plays an essential role in the oriented elongation of early muscle fibers through planar cell polarity signaling Marcelle et al., 1997;Serralbo and Marcelle, 2014). It is therefore not surprising to observe a TOPflash activity in DML cells that results from WNT ligand binding. The data presented here suggest that the WNTs carried by neural crest cells induce a low TOPflash response that regulates myotome's organization, but has no influence on MYF5 expression. On the contrary, DLL1, via the signaling module uncovered here, provokes the release of junctional b-catenin that leads to a high TOPflash response sufficient to trigger myogenesis. The reason for such different cellular responses is only speculative at present. It could be due to varying quantities of b-catenin entering the nucleus, low when WNT is presented by neural crest cells, high when it is (presumably massively) mobilized from the cell membrane. WNT11 and MYF5 would then be morphogen-like responses of DML cells to varying quantities of b-catenin. An alternative hypothesis is that of a qualitative difference in the transcriptional activity of b-catenin when it results from bona fide WNT signaling or when it is mobilized from the cell membrane. Rhee and colleagues have shown that b-catenin phosphorylated at the AJ on Y489 is transported into the nucleus and is transcriptionally active in vitro (Rhee et al., 2007). It is thus possible that post-translational phosphorylation of b-catenin potentiates or modifies its transcriptional activity, allowing a sharp separation between the WNT-dependent and the NOTCH-dependent cellular responses. This hypothesis is attractive, as it would explain why we did not observe any change in MYF5 expression even when WNT1 was over-expressed in neural crest cells and would also explain why in the presence of the mutant Y489F b-catenin, endogenous b-catenin did not rescue MYF5 expression when NICD was over-expressed.
Finally, our study also provides a model that reconciles apparently divergent observations on the respective role of NOTCH and WNT signaling in the initial phases of myogenesis. The necessary function of NOTCH in this process (Rios et al., 2011) is confirmed, but the present work demonstrates that the NOTCH signal is a permissive one, while the instructive role is executed by b-catenin and TCF/LEF. This is in accordance with the role assigned to WNT signaling in somite myogenesis years ago (Borello et al., 2006;Gros et al., 2009;Munsterberg et al., 1995;Stern et al., 1995;Tajbakhsh et al., 1998); however, the results presented here shed an entirely novel light on the origin of the signal and the pathway that mediates it. More generally, the increasing complexity of the cellular responses triggered by migrating neural crest cells in somites is puzzling as it raises the question of how two distinct tissues have become so perfectly coordinated during evolution to generate such sophisticated interactions.

Materials and methods Electroporation
The somite electroporation technique that was used throughout this study has been described elsewhere 2009;Rios et al., 2010;. Briefly, we targeted the expression of various constructs to the dorso-medial portion of newly formed interlimb somites of Hamburger-Hamilton (HH [Hamburger and Hamilton, 1992]) stage 15-16 chick embryos (24-28 somite). We have previously shown that this technique allows the specific expression of cDNA constructs in the DML, and that fluorescent reporters (e.g. GFP) are detected 3 hr after electroporation in this structure . Since we have analyzed most embryos 6 hr after electroporation, this implies that the molecules under study here have been acting during a narrow timeframe (about Figure 10. A signaling module in the DML that combines EMT and cell fate change. The physical contact with DLL1-positive neural crest cells triggers the cleavage of NOTCH receptor and the release of NICD in DML cells. NICD represses the activity of GSK-3b, independent of its transcriptional activity. The accumulation of SNAI1 (blue arrows) allows its translocation in the nucleus (not shown) where it activates EMT. This releases b-catenin from the adherens junction and/or membrane, leading it to enter into the nucleus where, together with TCF/LEF coactivator, it activates MYF5 expression. Abbreviations: DLL1: Delta-like 1; NECD: NOTCH extracellular domain; NICD: NOTCH intracellular domain; EMT: Epithelial to Mesenchymal Transition. DOI: 10.7554/eLife.14847.014

Quantifications and statistical analyses
Electroporation results in the transfection of a portion of the targeted cell population, which is variable from embryo to embryo. To precisely evaluate the phenotypes obtained after electroporation of cell-autonomously acting cDNA constructs, the number of positive cells was compared to the total number of electroporated cells, recognized by an internal fluorescent reporter construct. On average, more than 700 cells were counted per point and the corresponding quantifications are shown in all figures.
To determine the fluorescence intensity of electroporated cells ( Figure 2D,E), the surface of electroporated cells were rendered manually with Imaris software. The mean intensity for each cell and each channel in three-dimensions was collected for statistical analyses.
Statistical analyses were performed using the GraphPad Prism software. Mann-Whitney nonparametric two-tail testing was applied to populations to determine the P values indicated in the figures. In each graph, columns correspond to the mean and error bars correspond to the standard deviation. ***p value < 0.001, extremely significant; **p value 0.001 to 0.01, very significant.