Sclerotome-derived Slit1 drives directional migration and differentiation of Robo2-expressing pioneer myoblasts

Vertebrate muscles develop from somites, transient epithelial structures which, in response to environmental signals, dissociate to form the dermomyotome (DM) and sclerotome (Scl) (Christ and Ordahl, 1995). Although the DM contributes to most myotomal fibers (Christ et al., 1978; Cinnamon et al., 2006; Cinnamon et al., 1999; Gros et al., 2004; Huang and Christ, 2000; Kahane et al., 1998b; Kahane et al., 2002) and to mitotic muscle progenitors (Ben-Yair and Kalcheim, 2005; Gros et al., 2005; Kahane et al., 2001; Kassar-Duchosoy et al., 2005; Relaix et al., 2005), it is not the source of the myotomal founder cells.

The earliest muscle progenitors, termed ‘myotomal pioneers’, arise along the medial portion of the epithelial somite (Kahane et al., 1998a), where many of them are already post-mitotic and express MyoD, Myf5 and desmin. Upon dissociation, these progenitors bend underneath the forming dorsomedial lip (DML) of the DM, become mesenchymal and engage in a typical directional pattern of migration towards the rostral pole of each somite. This process is then followed by a rostral-to-caudal (R-C) and medial-to-lateral (M-L) order of fiber differentiation (Kahane et al., 2002; Kalcheim et al., 1999). Whereas medial pioneers encompass the entire M-L extent of cervical and limb-level segments, at flank regions they are complemented laterally by a population of myoblasts emerging from the lateral epithelial somite. In addition to being the first skeletal muscle cells in the embryonic somite, a recent study showed that in the absence of medial pioneers, the whole myotome is mispatterned (Kahane et al., 2007). Hence, the significance of myotomal pioneers resides in their capacity to organize further somitic waves that contribute to the developing muscle. Despite our knowledge of their cellular dynamics, the molecular mechanisms directing their unique rostralward migration and differentiation are unknown.

The Roundabout (Robo) receptor and its ligand Slit were first identified in Drosophila as guidance molecules for CNS axons (Kidd et al., 1998) and for muscle progenitors (Kidd et al., 1999; Kramer et al., 2001). Their vertebrate homologs play distinct roles across multiple systems (Brose and Tessier-Lavigne, 2000; Legg et al., 2008; Shiau et al., 2008; Wang et al., 2003). Robo receptors are single-pass transmembrane receptors. Robos 1-3 are mainly expressed in the nervous system (Morlot et al., 2007; Wong et al., 2001), whereas Robo 4 is specific to endothelium (Shibata et al., 2008). Slits 1-3 are multidomain leucine-rich repeat (LRR) molecules. Slit-Robo interactions are mediated by the second LRR domain of Slits and the two N-terminal Ig domains of Robos (Morlot et al., 2007). Robo members regulate Rho GTPases to affect cell motility and axonal guidance (Patel and Van Vactor, 2002). Whereas attraction is primarily mediated by Cdc42 and Rac, repulsion is executed through Rho GTPases, in particular RhoA (Dickson, 2001; Kozma et al., 1997). RhoA signals through Rock1 to enhance actomyosin contractility. It has been implicated in myosin activation and retraction of the back of the cell (Ridley, 2004; Wong et al., 2006), and has also been found to be active at the leading edge during migration (Kardash et al., 2010; Kurokawa et al., 2005; Pertz et al., 2006).

The present study is the first to document that regulation of directional migration and differentiation of pioneer myoblasts is intrinsic to the somite. In search for somite-derived signals, we found that Robo2 is expressed by pioneer myoblasts already at the epithelial stage whereas Slit1 is present in the DM and is abundant in the caudal half of the dissociating Scl. Loss of Robo2 function impaired the caudorostral migration of pioneers as well as myofiber formation. Similar results were obtained when Scl-derived, but not DM-derived Slit1, was abrogated. Loss of RhoA activity phenocopied the above effects and activation of endogenous Rho partially rescued the normal phenotype in pioneers lacking Robo2 activity. In all cases, myogenic specification was unaffected. By contrast, expression of desmin, the earliest myotomal intermediate filament protein (Bar et al., 2004), but not of myosin, was dramatically affected suggesting that Robo and Slit, through RhoA, regulate specific cytoskeletal properties. Taken together, we propose that the caudal Scl, via Slit1, acts as a negative guidance cue to direct overlying Robo2-expressing pioneer myoblasts towards the rostral domain of each segment. Slit1-Robo2 signaling also mediates their subsequent differentiation into properly aligned muscle fibers.

Fertile chick (Gallus gallus) and Japanese quail (Coturnix coturnix Japonica) eggs were from commercial sources (Moshav Orot and Moshav Mata).

Isotopic and isochronic grafts of inverted epithelial somites were performed both before and after the onset of MyoD or Myf5 expression in the medial domain; embryos were used at 16-18 or 25-27 somite stages (ss), respectively. A fragment containing the three most recently formed somites was surgically removed (Goldstein and Kalcheim, 1992). Then, somites from the left side of the donor embryo were inverted in the R-C direction and grafted into the right side of recipients from which the equivalent segments were removed in advance. Similarly, left stripes of somites were implanted into the right side of host embryos without rotation to achieve a M-L inversion.

DNA plasmids pCAGGS-GFP and YFP-C1-Lyn (from T. Meyer, Stanford University) served as controls. Rb2dN-GFP and Slit1-GFP were from S. Guthrie (Hammond et al., 2005). Dominant-negative RhoB and RhoA lacking GTPase activity (N19-RhoA and N19-RhoB) were from G. Prendergast (Adini et al., 2003). Dominant-negative MyoD consisted of the basic helix-loop-helix (bHLH) domain of mouse MyoD fused to the engrailed repressor domain [MyoD (bHLH)-enR]; E12 (bHLH)-enR served as control (Steinbach et al., 1998). MyoD and Myf5 expression vectors were from D. Duprez (Universite Pierre et Marie Curie, Paris). The following RFP-tagged RNAi constructs were employed: RNAi to cSlit1, RNAi to cRobo2 or RNAi to GFP (Shiau et al., 2008). DNA (3-5 mg/ml) was microinjected into the three or four recently formed somites. A square wave electroporator (BTX, San Diego, CA, USA) was used to deliver two pulses of current at 15 V for 10 milliseconds to the medial somite and one pulse at 9 V for 10 milliseconds to the ventral or dorsal somite (prospective Scl or DM, respectively).

Pluronic F-127 gel was prepared as described (Groysman et al., 2008) and mixed with 100 μg/ml lysophosphatidic acid (LPA). Small pieces of control or LPA-containing gels were placed dorsal to the embryo at the level of epithelial somites, and replaced every 4-5 hours, for a total of 14 hours.

Embryos were fixed with 4% formaldehyde and either wax embedded and sectioned at 10 μm, or observed as whole-mount preparations following evisceration and flattening.

Antibodies used were: anti-Desmin and anti-GFP (Molecular Probes), Myf5 (Yablonka-Reuveni and Paterson, 2001), N-cadherin (Shoval et al., 2007), p27 (Kahane et al., 2007), anti-caspase 3 (Cell Signaling) and anti-sarcomeric myosin [MF20 (Developmental Studies Hybridoma Bank)]. Secondary antibodies coupled either to Cy2, Cy3 or Cy5 were used (Jackson ImmunoResearch). In situ hybridization was performed as described (Cinnamon et al., 2001) using probes for chick MyoD and Myf5 (from D. Duprez); Robo1, Robo2, Slit1, Slit2 and Slit3 (from Ed Laufer, Columbia University, NY); and desmin, rhoA, rhoB and Wnt11 (from the BBSRC ChickEST Database) (Boardman et al., 2002). Nuclei were visualized with Hoechst.

The number of desmin-expressing mesenchymal progenitors, elongating cells or unit-length myofibers was analyzed on whole-mount preparations. Confocal sections that encompassed the entire thickness of the myotome were monitored using 40× or 60× objectives at 0.5-0.65 μm increments through the z-axis. In each optical section constituting part of a transfected cell, green and red channels were systematically compared to assess expression of desmin or myosin proteins. The dorsoventral analysis was complemented by performing transverse projections of the images. Results represent the mean percentage ± s.e.m. of total GFP-labeled cells in 10-20 embryos per treatment. Mesenchymal cells were defined by their typical fibroblastic shape, elongating cells were defined when their long axis was at least threefold the length of the short axis.

Whole-mounts and sections were photographed using a DP70 (Olympus) cooled CCD digital camera mounted on a BX51 microscope (Olympus) with Uplan FL-N 10×/0.30, 20×/0.5 and 40×/0.75 dry objectives (Olympus) at room temperature, using DP controller v1.2.1.108 acquisition software (Olympus). Confocal scanning was carried out using an Olympus Fluoview FV1000 with 10×/0.4 20×/0.75 40×/0.9 dry and 60×/1.35 oil objectives, software version 1.7c. For figure preparation, images were exported into Photoshop CS2 (Adobe). If necessary, the brightness and contrast were adjusted to the entire image and images were cropped without color correction adjustments or γ adjustments.

Significance of results was determined using the Mann-Whitney non-parametric test or Fisher's exact test. Control and two experimental groups were also compared using the Kruskal-Wallis non-parametric test, followed by the Mann-Whitney test, for pairwise comparisons, with Bonferroni correction of the significance level. All tests applied were two-tailed, and a P-value of 0.05 or less was considered significant.

To investigate whether the caudal-to-rostral migration of pioneer myoblasts is an intrinsic property of the somite or is controlled by surrounding tissues, quail or chick epithelial somites were grafted into chick hosts so that either the R-C or the M-L axes were independently switched. This was performed both before as well as after initial expression of MyoD, Myf5 and desmin. Chimeras were fixed at a stage when an overall triangular shape of pioneer cells was apparent (Figs 1, 2). As previously shown, this pattern is accounted for by pioneers differentiating in both R-C and M-L directions, with medially located fibers and the lateral-most cells still mesenchymal and preferentially restricted to the rostral half of segments (Kahane et al., 2002).

A stripe of three somites that already expressed desmin was removed from the left side of 25ss embryos and grafted into the right side of equivalent hosts after R-C inversion (Fig. 1A-B′, asterisks). In all chimeras, the grafted somites showed an inverted pattern of fiber elongation, demonstrating that the pioneer myoblasts had migrated and were undergoing differentiation according to their original polarity (n=13, Fig. 1A-B′). When the same procedure was performed at 16-18ss, prior to the onset of MyoD, Myf5 or desmin expression (Fig. 1C-D′, asterisks), a similar, inverted pattern of pioneer development was observed (Fig. 1C,D″, asterisks) when compared with the contralateral side (n=10, arrows). These results suggest that the directional R-C behavior of pioneer myoblasts is determined early during somite formation.

A stripe of three desmin-positive epithelial somites was removed from the left side of 25ss embryos (Fig. 2A, asterisk in 2B) and grafted into the right side of equivalent hosts, thus achieving a M-L inversion without altering the R-C axis (Fig. 2A, asterisk in 2B′). In all chimeras, the grafted somites showed an inverted pattern of fiber elongation with long fibers located laterally and short fibers/mesenchymal cells pointing medially (n=19, Fig. 2A,B′); this pattern suggests that pioneers kept their original M-L polarity. When performed on young somites before expression of myogenic genes, a de novo induction of desmin+ progenitors was observed along the medial edge adjacent to the neural tube, consistent with the notion that axial structures have a role in early myogenesis (see Discussion). However, in 10/12 embryos, only few original medial pioneers appeared laterally (not shown), probably because they failed to upregulate myogenic markers in the absence of axial signals and/or were deprived of survival signals and died. Regardless of the exact mechanism, analysis of the migratory behavior of medial pioneers following M-L inversions at a young stage was not achievable. To control for possible effects of microsurgery, similar somite replacements that kept their original R-C (n=4) or M-L (n=3) orientations were performed, which resulted in a normal pattern of pioneer behavior (not shown).

Collectively, R-C and M-L somite inversions show that the rostralward migration of mesenchymal pioneers and subsequent M-L expansion of the nascent fibers are directed by cues intrinsic to the somite.

Next, we searched for somite-derived factors for which expression differs along the R-C domain. Genes differentially expressed in rostral (R) versus caudal (C) domains of the Scl were shown to determine the segmental patterning of migrating neural crest precursors, as well as of peripheral axons, via repellent mechanisms (Kalcheim, 2000; Krull, 2001; Le Douarin and Kalcheim, 1999; Roffers-Agarwal and Gammill, 2009; Vermeren et al., 2000).

Robo2 mRNA was already expressed in early pioneer cells that constitute the medial epithelial somite (Fig. 3A,D, asterisk in 3G), then in the developing primary myotome at a stage previously shown to correspond to caudal-to-rostral (C-R) migration (Fig. 3A,C,F) (see Kahane et al., 1998a; Kahane et al., 2002) and in fully differentiated fibers (Fig. 3A,B,E). By contrast, Robo1 and Robo3 were not transcribed in the medial somite (not shown).

Slit1 mRNA was preferentially transcribed in the C domain compared with the R domain of the ventral epithelial somite and nascent Scl; in addition, it was homogeneously expressed in the DM (Fig. 3H,J-L). Upon full dissociation into DM and Scl, Slit1 was downregulated in both primordia (Fig. 3H-I′). Only later (E3.5-4) did Slit1 appear in the DML and ventrolateral lip (VLL) after dissociation of the DM sheet (not shown). Slit2 and Slit3 were restricted to the dorsal neural tube (NT) and notochord, or notochord only, respectively (not shown). The specific expression of Robo2 to pioneer myoblasts and of Slit1 to the Scl and DM, motivated us to examine their function in development of the pioneer wave.

Membrane-tethered GFP (YFP-C1-Lyn) or Rb2dN-GFP, a truncated version of Robo2 lacking the cytoplasmic domain, were electroporated into medial pioneers at the epithelial somite stage (Fig. 4A). Sixteen hours later, control YFP-C1-Lyn+ cells had organized into a typical triangular shape composed medially of elongating fibers and laterally of mesenchymal cells primarily confined to the R domain (n=186 cells counted in 12 segments from a total of five embryos, Fig. 4B,D,J-L). In contrast with the controls, the majority of Rb2dN-GFP+ cells remained mesenchymal (Fig. 4E,F,J, P<0.001) and distributed homogeneously along the R-C extent of the transfected segments (Fig. 4C,K, P<0.001). Those which succeeded in elongating (13% versus 66% in controls, P<0.001) did so aberrantly from either the R lip, the C lip or from elsewhere along the segment compared with control cells that coherently elongated in a R-C direction (n=202 cells in 18 segments/ten embryos, Fig. 4F,L).

When fixed 24 hours after electroporation, 96% of control GFP+ cells had already differentiated into typical long and thin unit-length muscle fibers compared with only 37% in Rb2dN-GFP-transfected cells (see Fig. S1 in the supplementary material, P<0.006). The Rb2dN-GFP+ cells that elongated, did so aberrantly from different origins (n=68 cells in 19 segments/seven embryos), were abnormally wide in shape and exhibited irregular membrane protrusions that were not observed in wild-type myofibers (see Fig. S1B-D in the supplementary material).

The vast majority of control pioneers co-expressed desmin immunoreactivity (97%), a marker that first appears at the epithelial stage. In striking contrast, only 8% and 23% of Rb2dN-GFP-treated progenitors did so when monitored 16 and 24 hours post-transfection, respectively (P<0.001 for both stages, Fig. 4G-J and see Fig. S1 in the supplementary material). This dramatic reduction of desmin protein was not accounted for by cell death, as few, if any, caspase-3 or TUNEL-positive cells were apparent in control or treated myotomes (see Fig. S1E,F in the supplementary material; not shown). Furthermore, in spite of interfering with the onset of desmin protein expression, Rb2dN-GFP did not inhibit desmin transcription (n=12, see Fig. S2A-B′ in the supplementary material). Next, we examined whether Robo2 activity is required for steady desmin expression. To this end, Rb2dN-GFP was transfected into epithelial pioneers at a later stage (25ss) following initial appearance of desmin protein. Under these conditions, desmin immunoreactivity was preserved (not shown) suggesting that Robo2 is necessary for initial expression of desmin protein but not for its maintenance. Because desmin lies downstream of Myf5 and MyoD, we examined whether loss of Robo2 affects the latter. Early transcription of either MyoD (n=20) or Myf5 (n=6), or expression of Myf5 immunoreactive protein (n=2) were unaffected (see Fig.S2C-H′ in the supplementary material). As part of the myogenic program following MyoD and Myf5 expression, pioneer myoblasts become postmitotic (Kahane et al., 1998a). Loss of Robo2 activity did not prevent them from exiting the cell cycle normally (not shown) or from expressing p27 (n=5, see Fig. S2I-J′ in the supplementary material). Furthermore, epithelial pioneers were shown to express N-cadherin yet lose it upon dissociation and migration (Cinnamon et al., 2006); loss of Robo2 function did not alter the normal dynamics of N-cadherin (n=5, see Fig. S2K-L′ in the supplementary material and not shown) indicating that, in this context, Robo might act via cadherins other than N-cadherin (Emerson and Van Vactor, 2002; Rhee et al., 2002).

Next, we asked whether Robo2 specifically affects desmin protein. To this end, we monitored expression of myosin with the MF20 antibody that labels both fast and slow myosin heavy chains. Fig. S3A in the supplementary material shows that in normal embryos, although desmin immunoreactivity is already detected in the medial epithelial somites (see also Fig. 1B and Fig. 2B), myosin immunoreactivity becomes first apparent in somite 12 of 25ss embryos where it stains only a subset of differentiating myoblasts (see Fig. S3B,B3,B4 in the supplementary material). Hence, because of the relatively late onset of myosin expression (Sacks et al., 2003), the early behavior of pioneer myoblasts (epithelial stage and C-R migration) can only be monitored using desmin as a marker. Therefore, to monitor effects on myosin, embryos were electroporated and fixed at a differentiation stage at which full-length myofibers were present in control segments (see Fig. S3C-C‴ in the supplementary material). As described above, Rb2dN-GFP-transfected cells were abnormal in shape and failed to generate normally elongated fibers (see Fig. S1 and Fig. S3D-D″ in the supplementary material). Nevertheless, myosin expression was not affected (see Fig. S3C-F in the supplementary material). A faithful quantification of the proportion of myosin-expressing out of total transfected cells was, however, not possible as this protein was expressed by only a subset of myotomal cells under both experimental conditions. Hence, Robo2 signaling specifically influences the onset of desmin expression without affecting the later appearance of myosin.

To assess further the specificity of the Robo2 phenotype, we used RNAi-mediated knockdown (Shiau et al., 2008). Robo2-RNAi phenocopied the effects of truncated Robo2; it perturbed the normal generation of unit-length fibers and reduced expression of desmin, but had no effect on initial myoblast specification as monitored by transcription of Myf5 (n=4 and n=7 for control-RNAi and Robo2-RNAi, respectively, see Fig. S4 in the supplementary material).

Together, these data suggest that Robo2 activity is necessary for the normal directional migration of pioneer myoblasts and for the formation of fibers with a normal morphology.

Next, we examined whether myogenic genes upregulate cRobo2 transcription. To this end, MyoD and Myf5 were misexpressed in the prospective DM, a domain that is normally negative for Robo2 except for its medial-most domain. Sixteen hours later, Robo2 mRNA was upregulated in DM cells overexpressing MyoD and Myf5, compared with GFP controls (n=9, 5, 5, respectively; see Fig. S5A-C″ in the supplementary material).

Reciprocally, to examine whether myogenic genes are necessary for cRobo2 expression in pioneers, we transfected MyoD (bHLH)-enR into medial epithelial somites. As control, we electroporated E12 (bHLH)-enR, in which the basic region was replaced by the corresponding sequence of the non-myogenic dimerization partner of MyoD. Whereas MyoD (bHLH)-enR inhibited both MyoD as well as Myf5-induced myogenesis in chick somites, E12 (bHLH)-enR had no effect, confirming their specificity (not shown). Consistently, a dramatic decrease in cRobo2 expression was detected in pioneer progenitors expressing MyoD (bHLH)-enR whereas no effect was observed in controls (n=7 and n=5, see Fig. S5D-E″ in the supplementary material). Together, these data suggest that Robo2 acts downstream of myogenic genes to affect the assembly of the desmin-containing cytoskeleton.

The preceding grafting experiments (Figs 1 and 2) suggested that the cues for directional behavior of pioneer myoblasts are somite-intrinsic. Slit1 is transiently expressed in DM and caudal Scl. Consequently, we inhibited Slit1 function with plasmid-based RNAi previously used in the nervous system (Shiau et al., 2008). Concomitantly, to identify the origin of active Slit1, we separately abrogated its activity in either the Scl or the DM by electroporating labeled RNAi to Slit1 or control RNAi-RFP either to the ventral or dorsal somite.

Confocal analysis revealed normal desmin+, Myf5+ fibers when control RNAi-RFP (GFP-RNAi) was transfected into the Scl (n=53 somites/15 embryos, Fig. 5A,A′,C,H). Furthermore, electroporation of the DM with GFP-RNAi or Slit1-RNAi did not interfere either with normal pioneer development (n=13 somites/four embryos, Fig. 5G,G′ and not shown). By contrast, most myoblasts overlying the Scls that received Slit1-RNAi were devoid of, or exhibited reduced, desmin immunolabeling (n=95 somites/23 embryos, Fig. 5B,B′, P<0.001). However, they expressed desmin mRNA (Fig. 5D,E) and Myf5 (Fig. 5I) (n=10 and n=9, respectively). Expression of desmin mRNA revealed that cells were disorganized and, instead of generating fibers, remained mesenchymal (Fig. 5D,E), similar to pioneers depleted of Robo2 function. In an attempt to rescue the above phenotypes, we co-electroporated Slit1-RNAi and human Slit1-GFP, which is not blocked by chick RNAi. In nine of 11 such embryos (36/41 segments), we observed the presence of elongated, desmin+ fibers that resembled the control situation (Fig. 5F,F′, P<0.002), further confirming the specificity of the observed effects. Hence, cross talk between Scl-derived Slit1 and Robo2-carrying pioneers might mediate the early patterning of the avian myotome.

Slit- and Robo-mediated axonal repulsion was shown to stimulate RhoA activation (Patel and Van Vactor, 2002; Wong et al., 2001). Because Slit1- and Robo2-dependent migration of pioneers also consists of an initial phase of myoblast ‘repulsion’ from the C domain of each segment towards the R domain, we examined whether Rho GTPases are involved.

A low level of RhoA transcripts was first detected in pioneer myoblasts residing in the medial epithelial somite. RhoA signal intensified upon pioneer delamination and migration and persisted throughout differentiation (see Fig. S6A-D in the supplementary material). By contrast, RhoB mRNA was absent during early phases of pioneer development and began to be transcribed in differentiated fibers with transcripts preferentially localized around the centrally located nuclei (see Fig. S6E-I in the supplementary material).

Next, we monitored the effects of inhibiting either RhoA or RhoB, the latter used as control because of its late onset compared with RhoA. Inhibition was achieved by overexpression of N19-RhoA or N19-RhoB, which lack GTPase activity (Adini et al., 2003). These constructs were shown to have differential effects on delamination of neural crest cells (Groysman et al., 2008). Control GFP or N19-RhoB-electroporated segments exhibited a characteristic pattern composed of desmin+ medial fibers and lateral elongating cells (Fig. 6A-B‴,D-F; n=330 cells/19 segments/eight embryos and n=42 cells/four segments/three embryos in control and N19-RhoB, respectively). By contrast, most N19-RhoA-transfected cells remained mesenchymal and many of them failed to migrate rostralward (P<0.001) and exhibited reduced desmin expression (Fig. 6C-F, n=387 cells/18 segments/seven embryos, P<0.001 for all parameters compared in Fig. 6D). Transfected cells that succeeded to elongate did so from different origins (R lip, C lip, elsewhere) when compared with control or N19-RhoB that exhibited, as expected, directional fiber differentiation from R to C (Fig. 6F). In summary, loss of RhoA function phenocopied the effects of loss of Robo2 function.

To assess whether RhoA operates downstream of Robo in myotomal pioneers, we examined whether Rho activation can rescue the phenotype caused by loss of Robo2. First, we electroporated full-length and constitutively active forms of RhoA or RhoB to medial pioneers. Both GTPases caused death of the transfected cells as previously observed in the neural tube (Groysman et al., 2008), precluding further analysis. To circumvent this limitation, endogenous Rho activity was stimulated by LPA. LPA is a bioactive phospholipid that signals through G-protein-coupled receptors and promotes cytoskeletal reorganization through activation of the Rho pathway (Li et al., 2003; Ren et al., 1999; Ridley and Hall, 1992; Weiner et al., 2001). Small pieces of control gel or LPA-containing pluronic gel were placed dorsal to the embryo in ovo, embryos were fixed 14 hours later and were analyzed for desmin expression, cell migration and differentiation.

Seventy-three percent of labeled pioneer cells were present in the R domain of somites that received control or LPA-containing gels (Fig. 7A-B′,F,G; n=199 cells/15 segments/seven embryos and 157 cells/16 segments/nine embryos, respectively). By contrast, only 42% of Rb2dN+/control gel-treated pioneers were present in the R domain (Fig. 7C,C′,F,G; n=74 cells/16 segments/seven embryos, P<0.001). Treatment with LPA of pioneers that received Rb2dN enhanced the proportion of rostrally located pioneers to 63% (Fig. 7D,D′,F; n=159 cells/20 segments/ten embryos, P<0.001). In addition, treatment with LPA stimulated expression of desmin immunoreactivity (50% of Rb2dN+ pioneers) compared with 15% in Rb2dN+/control gel and 95-98% in control GFP/control gel (Fig. 7A″-D″,F, P<0.001 for all comparisons). Nevertheless, LPA did not significantly enhance the percentage of elongating cells in Rb2dN-transfected cells (Fig. 7F). Hence, activation of endogenous Rho partially rescued the effects of loss of Robo2 function both on desmin expression and pioneer migration towards the R domain, but could not completely reverse the phenotype. Altogether, our results show that RhoA mediates the repulsive effect of Slit1 and Robo2 on selected aspects of pioneer myoblast development.

Lineage tracing studies identified the existence and dynamic behavior of distinct pioneer myoblasts in the avian embryo (Kahane et al., 1998a), and inhibition of their specification demonstrated their pivotal role in myotomal patterning (Kahane et al., 2007). However, the molecular mechanisms underlying pioneer development remained unknown. Here, we show that their unique directional migration and differentiation is regulated by somitic cues, and identify Robo2 and Scl-derived Slit1 as mediators of the these processes through RhoA signaling and desmin protein expression.

Studies on avian mesoderm patterning showed that specification of individual somites along their R-C extent is a somite-intrinsic property arising already prior to segmentation; by contrast, dorsoventral somite patterning, as well as the commitment of the medial somite to a myogenic fate, seem to be induced by signals from adjacent structures (Borman and Yorde, 1994; Buffinger and Stockdale, 1994; Dietrich et al., 1997; Kenny Mobbs and Thorogood, 1987; Münsterberg et al., 1995; Münsterberg and Lassar, 1995; Pownall et al., 1996; Rong et al., 1992; Stern et al., 1995; Vivarelli and Cossu, 1986) (for reviews, see Bothe et al., 2007; Christ et al., 2007; Christ and Scaal, 2008). Our results stemming from M-L somite inversions, performed prior to the onset of MyoD or Myf5 expression, are consistent with the axial structures inducing the medial somite to generate pioneer myoblasts. In addition, results of R-C somite inversions show that actual patterning of the first wave of muscle progenitors, represented by an initial C-R migration of myoblasts, is subjected to somite-intrinsic polarity cues, previously described to control the R-C patterning of the somite-derived Scl. Likewise, the segmental organization of adjacent neural crest-derived peripheral ganglia, Schwann cells and nerves is also subservient to intrinsic R-C differences within the Scl that are mediated by several families of repellent molecules (see Results). These intra-Scl differences are also reflected in the generation of components of the vertebrae (Goldstein and Kalcheim, 1992). Thus, collectively, intrinsic differences between early R-C somite domains affect the patterning of virtually all local derivatives including neural as well as skeletal (muscle and vertebrae) components.

Once differentiated, pioneer myofibers provide a scaffold for further myotomal organization (Kahane et al., 2007). The latter is accounted for by a second wave of muscle colonization stemming from all four lips of the DM once it forms (Cinnamon et al., 2001; Cinnamon et al., 1999; Gros et al., 2004; Huang and Christ, 2000; Kahane et al., 1998b; Venters et al., 1999), including the DML (Denetclaw et al., 2001; Denetclaw et al., 1997; Ordahl et al., 2001) and VLL (Denetclaw and Ordahl, 2000). Nevertheless, we suggest that the mechanisms by which DM-derived progenitors develop differ significantly from that of pioneers. Only pioneers are likely to respond to caudal Scl-derived Slit1, which drives their rostralward migration. Subsequent to Scl maturation, at a stage corresponding to the onset of DM lip contribution to the myotome, Slit1 is no longer expressed (see Fig. 3). This downregulation of Slit1 might enable second wave cells to elongate in both R and C directions, contrary to the transient directional behavior of earlier pioneers. Furthermore, Wnt11 from the DML, shown to orientate elongation of myocytes derived from the DML of the DM (Gros et al., 2009), is unlikely to direct migration and/or differentiation of pioneers as detectable Wnt11 mRNA became apparent only after pioneer myoblasts had accomplished migration, and its expression was homogeneous along the R-C extent of the lip (see Fig. S7 in the supplementary material and O.H.-B., unpublished). Furthermore, pioneer myoblasts can be distinguished from DM-derived fibers by differential electroporation to the medial versus dorsal somite domains, respectively. Labeling of the medial somite generates a triangular pattern of elongating fibers and lateral mesenchymal cells 14 hours later and fully elongated fibers by 24 hours. By this time, the DM, including the DML, has not yet contributed to myotome colonization (see Fig. S8 in the supplementary material). Consistently, pioneer myoblasts are mostly a post-mitotic (Brdu−/p27+) and finite cell population, whereas progenitors of the DML of the DM behave as a stem-like cell population. Hence, the present results add to previous lineage and molecular evidence to further support the notion that the establishment of pioneer fibers and that of subsequent DML-derived fibers are distinct and separable events (Cinnamon et al., 2006; Kahane et al., 2002).

We report that Slit1 mediates the repellent activity of the caudal Scl by acting upon Robo2-expressing pioneer myoblasts to drive rostralward migration. Similarly, Drosophila midline Slit acts as a long-range cue to initially repel muscle precursors away from the midline (Kidd et al., 1999). The present report is the first to document vertebrate muscle patterning through a repellent mechanism and, particularly, to demonstrate the activity of specific Slit-Robo members. In addition to an effect on R-C redistribution, we also report for the first time an effect of Robo/Slit on the generation of normal fiber morphology. It is likely that defects in migration provoked further abnormalities in myofiber development, as normal pioneers elongate from the R lip of each segment towards the opposite pole. Mesenchymal cells that fail to anchor to the R lip might be unfit to differentiate properly. It remains unclear whether, in addition to repellent signals underlying R-C polarity, there exist also chemoattractant cues that direct migration towards the R domain and/or signals that control attachment at the R edges. Surprisingly, no defects were observed when abrogating DM-derived Slit1. Possibly, pioneer cells are particularly sensitive to differences in ligand concentrations present along the R-C extent of the Scl but absent in the DM.

The effects of knocking-down Robo2 or Slit1 were phenocopied by loss of RhoA, but not of RhoB, consistent with only the former being expressed in the early phases of pioneer ontogeny and similar effects were observed when transfecting C3-transferase (O.H.-B., unpublished). Furthermore, RhoA acts downstream of Robo/Slit signaling as activation of endogenous RhoA in Robo2dN-transfected cells significantly rescued cell migration and desmin expression. The observation that RhoA did not rescue fiber formation in the absence of Robo2 activity indicates that additional factors are required downstream of Robo-Slit. In other contexts, Slit-Robo-mediated axonal repulsion was shown to activate RhoA through specific Slit-Robo GTPase activating factors (srGAPs) (Patel and Van Vactor, 2002; Wong et al., 2001). We suggest that in myotomal pioneers, the cytoplasmic domain of Robo2 might similarly interact with srGAPs to recruit RhoA to the membrane and promote its activity. Furthermore, RhoA, through Rho kinase and type 1 phosphatase might be necessary for the assembly and dynamics of desmin filaments (Inada et al., 1999). RhoA, through Rho kinase and type 1 phosphatase was shown to affect the balance between intermediate filament protein phosphorylation and dephosphorylation, thereby affecting the continuous exchange of filament subunits between a soluble and a filamentous state (Inada et al., 1999). Hence, Slit1-Robo2-activated RhoA could be necessary for the assembly and dynamics of desmin filaments in pioneer myoblasts, similar to the effect of RhoA on the polymerization and retrograde flow of actin filaments in other cellular contexts (Edwards et al., 1999; Kardash et al., 2010; Sanders et al., 1999).

Loss of either Robo2, Slit1 or RhoA consistently affected the onset of desmin protein expression in pioneer myoblasts but did not inhibit its transcription, or that of Myf5 or MyoD. In addition, both myogenic genes are also able to upregulate Robo2 transcription. Taken together, these results suggest that Slit1-Robo2/RhoA act downstream of cell specification to affect the assembly of the desmin cytoskeleton. The precise mechanism by which this is accomplished remains to be investigated. This will be of particular interest as, in contrast to desmin, the onset of myosin expression is not affected by loss of Robo2. Desmin belongs to the family of intermediate filament (IF) proteins (Cooke, 1983; Lazarides and Hubbard, 1976; Small and Sobieszek, 1977). It is the main IF protein of all muscle cells. In avians and mammals, desmin is one of the earliest muscle markers, detected already in epithelial somites and myoblasts (Kaehn et al., 1988; Kahane et al., 1998a; Kaufman and Foster, 1988), where it forms longitudinal filaments associated with the Z-disc (Schroder et al., 2000). Desmin participates in the early establishment of the sarcomere (Furst et al., 1989; Schaart et al., 1989) and its expression precedes that of other proteins of the contractile apparatus, including myosins [this study and Sacks et al. (Sacks et al., 2003)]. Consistent with its roles in connecting the contractile apparatus with mitochondria, nuclei and costamers, knockout embryos exhibit abnormal alignment of myofibrils, disrupted anchorage of myofibrils to the sarcolemma and mitochondria with disturbed function (Bar et al., 2004). We observed that pioneers with either reduced or apparent lack of desmin failed to properly migrate, remained mesenchymal and did not generate fibers. The few cells that partially elongated were abnormally wide and exhibited prominent membrane protrusions, all reflecting abnormal cytoskeletal assembly. Whereas early electroporations abrogated desmin and generated a severe phenotype, late transfections performed after initial desmin expression had no apparent effect on protein immunoreactivity (although it was difficult to assess its relative intensity), and cellular abnormalities were correspondingly less dramatic. It is possible that after initial assembly, desmin filaments are stable and late treatment only impaired polymerization of newly synthesized protein. It is also possible that Robo2-dependent fiber shape and differentiation involves the regulation of additional genes. Along this line, further studies should directly address whether inhibiting desmin recapitulates all or only some of the phenotypes induced by loss of Robo2 function.

We thank Yuval Cinnamon for initial expression data of Robo and Slit and for providing Fig. S8 in the supplementary material. We thank S. Guthrie, M. Bronner-Fraser, Z. Yablonka-Reuveni, T. Meyer, G. Prendergast and E. Laufer for reagents. We are indebted to Tallie Bdolach for help with statistics. This study was supported by grants from the Israel Science Foundation (ISF), the EEU 6th Framework program Network of Excellence MYORES, the Association Francaise contre les Myopathies (AFM), and the DFG (SFB 488) to C.K.