Endothelial cell specification in the somite is compromised in Pax3-positive progenitors of Foxc1/2 conditional mutants, with loss of forelimb myogenesis

Foxc1 and Foxc2 transcription factors are important for tissue and organ development at different sites in the mouse embryo (Kume, 2009, 2010). They are expressed in paraxial mesoderm (Kume et al., 1998, 2000) and are required for somite maturation. In Foxc1 or Foxc2 single mutants, the ventral compartment of the somite, the sclerotome, is affected and its derivatives, the cartilage and bones of the ribs and vertebral column, are compromised. In the double mutant, somites fail to develop (Kume et al., 2001). The transcription factor Pax3, which plays an important role in myogenesis, is co-expressed with Foxc1/2 in paraxial mesoderm; however, it is subsequently confined to the dorsal compartment of the somite, the dermomyotome, whereas Foxc1/2 continue to be expressed at a high level in the sclerotome, as well as at a lower level in the dermomyotome where Foxc2 transcripts are notably detectable in the hypaxial domain (Lagha et al., 2009). The dermomyotome is the source of all skeletal muscle in the trunk and limbs and also gives rise to other mesodermal derivatives, including vascular endothelial and smooth muscle cells (Buckingham and Mayeuf, 2012). Genetic experiments have shown reciprocal inhibition between Pax3 and Foxc2 with consequences for cell fate choices in the multipotent cells of the dermomyotome when this equilibrium is perturbed (Lagha et al., 2009). Thus, higher expression of Foxc2, relative to Pax3, promotes vascular derivatives whereas upregulation of Pax3 promotes skeletal muscle at the expense of a vascular cell fate. The onset of skeletal myogenesis in the trunk results from delamination of Pax3-positive progenitors from the dermomyotome to form the underlying myotome (Buckingham and Mayeuf, 2012). At the limb level in the mouse embryo, bipotent progenitors (Kardon et al., 2002) marked by both Pax3 and Flk1 (Kdr ­– Mouse Genome Informatics; also known as VegfR2) (Mayeuf-Louchart et al., 2014) can give rise to endothelial (Pecam-1-positive) cells that migrate from the somites into the limb bud to form a subset of superficial blood vessels (Hutcheson et al., 2009), or to migrating myogenic progenitors that retain expression of Pax3 and contribute all the skeletal muscles of the limb (Buckingham and Mayeuf, 2012). Signalling pathways can potentially affect the balance between Pax3 and Foxc2, as shown for Notch, which promotes Foxc2 expression and the endothelial cell fate of dermomyotome progenitors that migrate into the forelimb (Mayeuf-Louchart et al., 2014).

Foxc1 has overlapping functions with Foxc2 (Kume et al., 1998, 2001) and because it is also expressed in the somite, we have investigated the phenotype of Foxc1/2 double mutants with respect to endothelial versus myogenic cells of the forelimb. Disruption of somitogenesis was avoided by targeting conditional mutations of both Foxc genes to Pax3-expressing cells. Whereas overproduction of myogenic cells was observed in the trunk, myogenic cells were absent from the limb resulting in no limb muscle formation. This striking result is discussed in the context of the loss of somite-derived endothelial cells observed in the double mutant.

Pax3^Cre/+ mice (Engleka et al., 2005) were crossed with conditional Foxc1^flox/flox;Foxc2^flox/flox mice (Sasman et al., 2012). In control and conditional mutant embryos, a Rosa26^flox-nLacZ or Rosa26^{tomato-floxGFP} reporter allele was also introduced into the crosses so that cells that express or had expressed Pax3 could be followed. We observed a delay in recombination with the Pax3^Cre allele, probably reflecting a delay in Cre recombinase accumulation, as indicated by GFP labelling of Pax3^Cre/+;Rosa26^{tomato-floxGFP/+} embryos (Fig. 1A) at embryonic day (E) 9.25, when recombination has occurred at forelimb level but is not detected in more posterior somites. By E10.5, recombination extends more posteriorly. To check the efficiency of recombination, the somites and forelimb region of E9.25 embryos were dissected to remove the neural tube and Pax3-positive neural crest cells (Fig. 1B). GFP-positive cells were isolated by flow cytometry. RT-qPCR analysis demonstrates that Foxc1 transcripts are strongly reduced in Pax3^Cre/+;Foxc1^flox/flox;Foxc2^flox/+ (Pax3^Cre/+C1^Δ/ΔC2^Δ/+) and Pax3^Cre/+;Foxc1^flox/flox;Foxc2^flox/flox (Pax3^Cre/+C1^Δ/ΔC2^Δ/Δ) embryos as well as transcripts of Foxc2 in Pax3^Cre/+;Foxc1^flox/+;Foxc2^flox/flox (Pax3^Cre/+C1^Δ/+C2^Δ/Δ) and Pax3^Cre/+;Foxc1^flox/flox;Foxc2^flox/flox (Pax3^Cre/+C1^Δ/Δ C2^Δ/Δ) embryos (Fig. 1C).

Pax3 expression is not restricted to somites, but is also a feature of neural crest cells derived from the dorsal neural tube. Developmental defects have been described in Foxc1 and/or Foxc2 null mutants in derivatives of neural crest cells where these genes are also expressed (Kume et al., 1998; Seo and Kume, 2006). Similar defects are observed in the conditional mutants in which the cranial skeleton is affected (Fig. S1A). At the trunk level, formation of the axial skeleton is also affected (Fig. S1A), in keeping with the role of Foxc1/2 in the development of derivatives of the sclerotome (Kume et al., 2001). However, the axial skeleton and ribs begin to form and somitogenesis takes place in the double conditional mutants (Fig. 2), in contrast to the Foxc1/2 null mutant. At later stages, we observed truncation of the tail and loss of posterior somites (Fig. S1B), reflecting extensive Pax3-mediated recombination by E11.5.

As predicted from the analysis of single Foxc2 mutants (Lagha et al., 2009), excessive production of cells expressing the myogenic determination gene Myod1 is observed in the trunk region of double conditional (Pax3^Cre/+C1^Δ/ΔC2^Δ/Δ) mutants compared with control heterozygote (Pax3^Cre/+C1^Δ/+C2^Δ/+) embryos at E11.5 (Fig. 2A). Quantification of MyoD-positive cells in interlimb somites indicates an increase of myogenic cell number in double conditional (Pax3^Cre/+C1^Δ/ΔC2^Δ/Δ) mutants compared with control heterozygote (Pax3^Cre/+C1^Δ/+C2^Δ/+) embryos at E10.5 (Fig. 2B). The epithelial structure of the dermomyotome is maintained, as shown on sections of interlimb somites (Fig. 2B). This is also the case at the forelimb level at E9.5 (Fig. 3A).

However, there is a striking absence of myogenic cells expressing Myod1 in the forelimb of the double mutant (Fig. 2A; Fig. 4A). These cells are somewhat reduced in the absence of Foxc1 and are more severely reduced in the absence of Foxc2 (Fig. 4A). In the double conditional mutant, forelimb muscles are absent at E19.5 as shown by myosin staining on whole embryos and by Haematoxylin and Eosin staining on forelimb sections (Fig. 4B). Absence of skeletal muscle is due to a failure of migration of Pax3-expressing myogenic progenitor cells into the limb bud (Fig. 3B; Fig. 4C). In the genetic-tracing experiment shown in Fig. 4C, endothelial cells derived from Pax3-expressing progenitors in the somite would also normally be labelled in the limb bud (Hutcheson et al., 2009; Mayeuf-Louchart et al., 2014). However such labelling is not evident in the absence of Foxc1/2 and foci of residual β-galactosidase-positive cells correspond to neural crest cells that will contribute to the sympathetic nervous system, derived from dorsal root ganglia, marked by AP2α (Tfap2a – Mouse Genome Informatics), which are not affected in the Foxc1/2 conditional mutants (Fig. S2).

In the forelimb of the mouse embryo, only a small proportion of endothelial cells are derived from the somite (Hutcheson et al., 2009; Mayeuf-Louchart et al., 2014). In Pax3^Cre/+ embryos in the presence of a Rosa26^{tomato-floxGFP} allele, such Pecam-1-positive endothelial cells can be distinguished (Fig. 5A). These cells are not detectable on the double conditional Foxc1/2 mutant background and are reduced in the single mutants, notably in the absence of Foxc2. Quantification of GFP-positive cells that are positive for Pecam-1, after separation by flow cytometry, confirms the reduction in this cell population in the forelimb of Foxc1/2 conditional mutants at E10.5 (Fig. 5B).

Before cells migrate from the somite to the forelimb, they co-express Pax3 and Flk1, which encodes vascular growth factor receptor 2 (Mayeuf-Louchart et al., 2014). Subsequently, maintenance of Pax3 expression and upregulation of the gene for the transcription factor Lbx1 characterises myogenic progenitor cells, whereas endothelial progenitors continue to express Flk1 and rapidly activate markers of the endothelial phenotype such as Pecam1. It is therefore possible to distinguish these two progenitor types in the somite. GFP-positive cells isolated from somites at the forelimb level of Pax3^Cre/+;Rosa26^{tomato-floxGFP/+} embryos at E9.25 were separated by flow cytometry and the ratio of Lbx1 to Flk1 transcripts quantified by RT-qPCR (Fig. 5C). On the double conditional Foxc1/2 mutant background this ratio is significantly higher. This demonstrates that migratory myogenic progenitors are specified and that they accumulate at the expense of endothelial progenitors in the Pax3-positive population of cells that give rise to both cell types in the somite.

Delamination and migration of Pax3-positive muscle progenitors into the limb bud depends on the Met tyrosine kinase receptor (Bladt et al., 1995). In the double conditional Foxc1/2 mutant, the Met gene (also known as c-Met) is expressed normally in the hypaxial domain of somites (Fig. 6A). Transcripts are not detected in the forelimbs at E10.5, as expected from the absence of Pax3-positive myogenic progenitors. Lbx1, which activates the gene for the cytokine receptor Cxcr4 required for the migration of a subset of myogenic progenitors, is also expressed in these Pax3-positive cells in the somites at forelimb level. It is still present in Foxc1/2 mutants at E9.25 (Fig. 5C) and at E10.5 (Fig. 6B), but not in the limbs where migration has failed to occur. By E10.5, Pax3- and Lbx1-positive cells have left the somite and accumulated in the adjacent region proximal to the forelimbs (Fig. 6B). Labelling with the epithelial marker Zo-1 (Tjp1 – Mouse Genome Informatics) reveals Pax3-positive cells that are Zo-1 positive in this region, where the hypaxial somite has broken down, but cells have failed to migrate into the limb. It also shows many cells that are Pax3 positive and Zo-1 negative, indicating that they had left the dermomyotome epithelium (Fig. 3B). At E9.5, when Pax3-positive cells delaminate and begin migrating into the forelimb, somites at the limb level in the double conditional mutant have a normal morphology (Fig. 3A) and we see no indication of cell death at this or later stages (data not shown).

At E10.5, a few myogenic progenitor cells in the forelimb begin to express the myogenic determination factor Myf5, as well as myogenic cells in the myotome of the somites where muscle is forming in control embryos (Fig. 6C). In the Foxc1/2 conditional mutant, excess Myf5-positive cells are dispersed around the myotome, as seen for MyoD-positive cells (Fig. 2B); however, additional labelled cells are observed outside this structure, in the region immediately adjacent to the proximal forelimb (Fig. 6C). By E11.5 in the mutant, myogenic cells in this position express the myogenic differentiation factor myogenin and are labelled with a myosin heavy chain antibody that marks muscle fibres (Fig. 6D). Differentiated cells are not normally observed in this position. We therefore conclude that myogenic progenitors, which do not migrate into the limb, differentiate prematurely.

In conclusion, we show that deletion of Foxc1 and Foxc2 specifically in Pax3-positive cells affects cell fate choices in the dermomyotome of somites at forelimb level, promoting the myogenic cell fate at the expense of endothelial cells that migrate to the limb. However, despite this increase in myogenic cell fate specification, no myogenic cells are found in the forelimb. Loss of endothelial cell specification compromises myogenic cell migration. Instead of migrating into the forelimb, these cells undergo premature myogenic differentiation. We conclude that the small percentage of endothelial cells specified in the somite plays a crucial role in ensuring correct migration of myogenic cells into the limb.

The increase of myogenic cells in the absence of Foxc1 and Foxc2 is also observed in interlimb somites where MyoD-positive cells extend beyond the normal limits of the myotome. This observation is similar to the phenotype observed in the presence of the gain-of-function Pax3^Pax3-FKHR allele (Relaix et al., 2003) and is consistent with an upregulation of the Pax3-dependent myogenic pathway in the absence of Foxc1/2. In somites of double conditional mutants at forelimb level, analysis of the expression of Lbx1, which is an early marker of myogenic progenitors, shows that the proportion of Lbx1-positive cells is increased, compared with Flk1-positive cells (which include the endothelial population) (Mayeuf-Louchart et al., 2014). This therefore demonstrates promotion of the myogenic cell fate at the expense of the endothelial fate, which reflects the loss of Foxc1/2 in Pax3-expressing cells. At these early stages, mutant somites have a similar morphology to controls, so that delamination and migration of myogenic progenitors is not prevented by somite disorganisation. The delay in Cre recombinase activity from the Pax3^Cre allele avoids early loss of Foxc1/2, which affects somitogenesis, as seen in posterior somites of the double conditional mutant. The double mutant phenotype that we observed at forelimb level is more pronounced than that of the single Foxc2 conditional mutant indicating the role of Foxc1, and indeed the Foxc1 conditional mutant also shows a reduction of limb myogenesis although this is less pronounced than in the absence of Foxc2. We had previously observed a decrease in Pax3-positive cells in the forelimbs of Foxc2 mutant embryos (Lagha et al., 2009), similar to that reported here for the conditional Foxc2 mutant. We had proposed that it might be due to upregulation of Pax7, which perturbs myogenic progenitor cell proliferation in the limb (Relaix et al., 2004). However, the absence of all Pax3-positive cells in the forelimbs of the double conditional Foxc1/2 mutants precludes this explanation.

The failure of myogenic progenitors to migrate into the forelimb does not appear to reflect an obvious deficit in these cells and indeed they express Pax3, which ensures myogenic, migratory and survival functions (Buckingham and Rigby, 2014). Rather, it is the absence of endothelial cell migration from the somite that correlates with this phenotype. In the Foxc1/2 double mutant, we conclude that bipotent progenitor cells in the somite assume a myogenic, at the expense of an endothelial, cell fate, although remaining endothelial cells might also have a migratory defect (Hayashi and Kume, 2008; Hayashi et al., 2008) that prevents them entering the limb. In the chick embryo, it had been shown that endothelial and myogenic cells derived from the somite migrate independently and distribute to different locations (Huang et al., 2003). This is also indicated by genetic-tracing experiments in the mouse (Hutcheson et al., 2009; Mayeuf-Louchart et al., 2014). Furthermore, endothelial cells migrate from the mouse somite to the limb bud before myogenic progenitors (Tozer et al., 2007; Yvernogeau et al., 2012). Grafting of genetically marked mouse somites into the chick embryo at limb level demonstrated this clearly and showed that endothelial cell migration is independent of Pax3 whereas it depends on Flk1. Grafting of somites from Flk1 mutant mouse embryo resulted in the loss of endothelial cell migration and also of migrating myogenic cells, leading the authors to propose that prior migration of somite-derived endothelial cells is essential for myogenic cell migration into the limb (Yvernogeau et al., 2012). Interpretation of this experiment is complicated by the initial expression of Flk1 in Pax3-positive progenitors that can give rise to both myogenic and endothelial derivatives (Mayeuf-Louchart et al., 2014), with the possibility that this expression plays an early role in determining myogenic cell behaviour. The analysis of Foxc1/2 conditional mutants reported here provides another line of evidence supporting the hypothesis that somite-derived endothelial cells, which in the mouse represent only a small proportion of all the endothelial cells in the limb, are crucial for myogenic cell migration. In their absence, vascularisation of the limb takes place but no skeletal muscle forms. It remains to be seen how this effect is mediated; presumably, it occurs via signalling to myogenic progenitors by endothelial cells, necessary for their migration. Foxc1 in mesenchymal cells in the developing cerebellum has been shown to directly activate the gene that encodes Sdf1α (Cxcl12 – Mouse Genome Informatics) (Zarbalis et al., 2012), the ligand of Cxcr4, with major consequences for radial glial cell organisation and neuronal migration via Cxcr4 signalling in the absence of Foxc1 (Haldipur et al., 2014). Cxcr4 is expressed on myogenic progenitor cells that migrate to the limb and, in its absence, migration of a subset of progenitors is compromised and some limb muscles are missing (Vasyutina et al., 2005). A minor loss of limb muscle is seen in the Foxc1 mutant but an effect on Cxcr4 signalling cannot explain the loss of all migrating muscle progenitors in the double mutant. Furthermore, it is not clear whether somite-derived endothelial cells express Sdf1. As the trajectories appear to be different (Yvernogeau et al., 2012), it seems less likely that myogenic cells follow a trail laid by their endothelial sisters. It is more probable that the specification of endothelial cells that is Foxc dependent has an indirect effect on the environment of myogenic progenitors within the somite, which modifies their subsequent migratory behaviour.

Conditional mutants for Foxc1 (Foxc1^flox/flox) and Foxc2 (Foxc2^flox/flox) (Sasman et al., 2012) were crossed with Pax3^Cre/+ mice (Engleka et al., 2005) to generate Foxc1 (C1^Δ/Δ; C2^Δ/+), Foxc2 (C1^Δ/+; C2^Δ/Δ) and double (C1^ΔΔ; C2^Δ/Δ) conditional mutant embryos. Rosa26^flox-nLacZ/+ (gift from J. F. Nicolas, Institut Pasteur, Paris, France) or Rosa26^{tomato-floxGFP/+} (Muzumdar et al., 2007) mice were crossed onto this genetic background to introduce a Cre reporter allele. Embryos were dated taking E0.5 as the day after the vaginal plug and somite number was used for precise comparison. Double conditional heterozygotes Pax3^Cre/+;Foxc1^Δ/+;Foxc2^Δ/+ were used as controls in a number of experiments because their phenotypes were indistinguishable from Pax3^Cre/+;Foxc1^+/+;Foxc2^+/+ embryos. All animal procedures were carried out in the Institut Pasteur animal facility with approval by the French Ministry of Agriculture.

X-Gal staining and whole-mount in situ hybridisation with digoxigenin-labelled probes were performed as described by Lagha et al. (2009). The Met probe was kindly provided by C. Birchmeier (Max Delbrück Centrum für Molekulare Medizin, Berlin, Germany), the Myod1 probe was as described by Sassoon et al. (1989). Immunostaining of myosin heavy chain on a whole-mount embryo was carried out as previously described by Ouimette et al. (2010). Haematoxylin and Eosin histological staining and immunostainings were assessed on frozen sections (16 µm), after overnight incubation and 2 h of embryo fixation in PBS or 4% paraformaldehyde, respectively. Pax3, Pecam-1, GFP, Myf5, Lbx1 (gift of C. Birchmeier), MF20, ZO-1 and AP2α antibodies were used as previously described by Lagha et al. (2009) (see Table S1 for details). A confocal Zeiss LSM 700 laser scanning microscope was used for fluorescence image acquisition. Each experiment was performed on a minimum of three embryos per condition.

Skeletal preparations were performed as described by Bensoussan-Trigano et al. (2011).

Embryos were microdissected in F-12 medium (Gibco). For E9.25 embryos, the forelimbs and their five adjacent somites were separated from the neural tube and delaminating neural crest cells, viewed by GFP fluorescence under a fluorescent binocular microscope. For E10.5 embryos, the forelimbs were recovered without somites. Tissues were conserved in F-12 medium during the time of the dissection and were then mechanically dissociated before enzymatic digestion in F-12, 0.1% collagenase D (Roche) at 37°C. Cells were re-suspended in F-12, 1% horse serum after 18 min of centrifugation (1800 rpm; 600 g). After a 1-h incubation with Alexa Fluor 647 anti-mouse CD31 (Pecam-1) (BioLegend), followed by two rinses in PBS with 1% horse serum and two 18-min centrifugations (1800 rpm; 600 g), cells were filtered with a 40-µm Cell Stainer Cap. A MoFlo Cytomation or Astrios (Beckman Coulter) apparatus was used for cell sorting, on the basis of size, granularity and GFP or Alexa 647 fluorescence. The number of cells was determined for each embryo.

Sorted cells were recovered in 75 µl of RLT buffer and RNA was purified using the RNeasy Micro Kit (Qiagen). RT-qPCR analysis was performed as described by Lagha et al. (2009). See Table S2 for primer sequences.

PRISM software was used for statistical analysis with the Student's t-test, preceded by the Fisher test. Data are presented as the mean and standard error of the mean (s.e.m.).

We thank S. Coqueran, C. Cimper and P. Dardenne for technical assistance; P. H. Commere for cell-sorting analysis; P. Roux for imaging and Y. Lallemand; and B. Robert for advice.