Neuropilin-mediated neural crest cell guidance is essential to organise sensory neurons into segmented dorsal root ganglia

The division of embryonic tissues into segments is a fundamental patterning process in early vertebrate development. It is most obvious in the trunk,where mesoderm on either side of the neural tube forms metameric epithelial spheres termed somites. Whereas the sclerotome layer of the somites gives rise to the spinal vertebrae, the dermomyotome layer differentiates into dermis and epaxial muscle (reviewed by Brent and Tabin, 2002). In higher vertebrates, the posterior half of each sclerotome is repulsive for motor axons and the neural crest cell (NCC)precursors of trunk sensory and sympathetic ganglia, forcing them to migrate preferentially through the anterior sclerotome (reviewed by Le Douarin and Kalcheim,1999). This pattern of NCC migration was previously thought to lay the foundation for the segmental organisation of their neuronal progeny (e.g. Kalcheim and Teillet, 1989; Kuan et al., 2004). However,loss of the repulsive guidance cue SEMA3F or its receptor NRP2 in the mouse causes NCC invasion into both anterior and posterior sclerotome without disrupting the metameric pattern of the sympathetic or dorsal root ganglia(DRG) (Gammill et al., 2006; Waimey et al., 2008). We have since shown that the related SEMA3A/NRP1 pathway provides an alternative mechanism for trunk NCC guidance (Schwarz et al., 2009). Specifically, loss of SEMA3A or NRP1 leads to excessive NCC migration along intersomitic and perisomitic routes and consequently to ectopic sensory and sympathetic neuron differentiation. However, this defect does not abolish metameric NCC migration and therefore perturbs DRG segmentation only mildly. The presence of either SEMA3A/NRP1 or SEMA3F/NRP2 signalling is therefore sufficient for the formation of segmented DRG. We now show that the simultaneous loss of both signalling pathways severely disrupts metameric NCC streaming and DRG segmentation. Thus, trunk NCCs in compound mutants lacking either the two ligands or the two receptors invade both intersomitic furrows and the posterior sclerotome halves, and the sensory NCC progeny consequently coalesce into a continuous neuronal band adjacent to the neural tube, rather than forming separate DRG. These findings are of great significance for our understanding of peripheral nervous system organisation and will help to elucidate the developmental mechanisms that coordinate the organisation of NCC-derived sensory neurons with the sclerotome-derived vertebrae.

Mice carrying a Sema3a-(Taniguchi et al., 1997) or Sema3f- (Sahay et al.,2003) null allele or mutations that disrupt semaphorin signalling through both neuropilins (Gu et al.,2003) have been described. In situ hybridisation was performed with digoxigenin-labelled riboprobes transcribed from plasmids containing cDNA for Isl1 (Pfaff et al.,1996), Nrp1 or Sox10(Schwarz et al., 2008a). For immunolabelling, we used goat anti-NRP1 and goat anti-NRP2 (R&D Systems),rabbit anti-p75, rabbit anti-neurofilament and rat anti-endomucin(Schwarz et al., 2008a).

Trunk NCCs form by delamination from the embryonic neuroectoderm and disseminate into the body along specific pathways. These pathways are characterised by their relationship to the emerging somites, which are added to the elongating embryo in a rostrocaudal fashion (reviewed by Le Douarin and Kalcheim,1999). In the mouse, a small population of early-born,NRP1-negative NCCs normally travels ventrally in the space between newly formed somites towards the dorsal aorta to seed the sympathetic ganglia (see Fig. S1 in the supplementary material)(Schwarz et al., 2009). By contrast, most other NCCs that give rise to neurons and glia in the trunk migrate through the anterior half of the somite, after it has segregated into sclerotome and dermomyotome layers. The onset of this migration pattern correlates with NRP1 expression by NCCs and SEMA3A expression in their environment (Schwarz et al.,2009). Moreover, NRP1 and its ligand SEMA3A are essential to prevent trunk NCC migration through the intersomitic furrow and to keep NCCs on a ventromedial path (Schwarz et al.,2009). Consistent with the phenotype of Sema3a- and Nrp1-null mutants, a mutation that reduces semaphorin signalling through NRP1 (Gu et al., 2003)leads to excessive intersomitic NCC migration, with entry of NCCs into the dorsolateral path (Fig. 1C,F)(Schwarz et al., 2009). By contrast, SEMA3F and its receptor NRP2 exclude NCCs from the posterior sclerotome (Fig. 1B,H)(Gammill et al., 2006). Significantly, we found that NRP2 signalling is not essential to confine NCC migration to the correct dorsoventral plane(Fig. 1E,H,K). Together, these observations raise the possibility that SEMA3A/NRP1 and SEMA3F/NRP2 synergise to control distinct aspects of trunk NCC guidance, perhaps by affecting different NCC subpopulations.

To address whether NRP1 and NRP2 preferentially guide different trunk NCC subpopulations, we compared their expression pattern to that of the nerve growth factor receptor p75 (NGFR), a good marker for NCCs with neuronal or glial potential (Fig. 1M-T; the plane of sectioning is illustrated in Fig. S1A in the supplementary material). We observed that sympathetic progenitors preferentially expressed NRP1 over NRP2 (compare Fig. 1M with 1Q). By contrast, both neuropilins appeared to be expressed in NCCs in the anterior sclerotome(Fig. 1N-P,R-T). However, the high level of NRP2 in the anterior sclerotome(Fig. 1Q; Fig. 2B) precluded the unequivocal assignment of NRP2 expression to trunk NCCs in situ, as previously discussed (Gammill et al.,2006). We therefore examined whether NRP2 colocalised with p75 in Sema3f-null embryos, which contain ectopic NCCs in the NRP2-negative posterior sclerotome (Fig. 2B,H). Using this genetic manipulation, we could clearly identify NRP2 in NCCs (Fig. 2I-L). Interestingly, NCCs within the sclerotome appeared to be heterogeneous with respect to NRP2 levels, as cells were stained with variable intensity in the anterior sclerotome of wild types and in the posterior sclerotome of Sema3f-null mutants (Fig. 2F,L). It is possible that the intermixing of high- and low-level NRP2-expressing NCC populations is due to a community effect, in which subtypes of NCCs remain in physical contact while they migrate within the sclerotome (Kulesa and Fraser,1998).

We next examined NRP2 expression in Sema3a-null mutants, which display excessive NCC migration on the intersomitic route that is normally taken only by the early wave of sympathetic NCC precursors(Schwarz et al., 2009). We found that the mutant intersomitic stream predominantly contained NCCs with undetectable NRP2 expression (Fig. 2M-R). By contrast, the ectopic intersomitic stream prominently expressed NRP1 (Fig. 2S-X),suggesting that NRP1, but not NRP2, is a marker for intermediate wave NCC precursors with sympathetic potential (see Fig. S1B in the supplementary material). In support of this idea, Nrp1-null, but not Nrp2-null mutants contain ectopic sympathetic neurons(Kawasaki et al., 2002; Waimey et al., 2008).

Taking advantage of a mouse strain that carries the Nrp1^Sema mutation on a Nrp2-null background and preserves VEGF signalling through NRP1 (Gu et al., 2003) and hence does not disrupt vascular patterning(Vieira et al., 2007), we next asked how the simultaneous loss of both neuropilins affected trunk NCC migration. In these compound Nrp1^Sema Nrp2 null mutants,neither intersomitic migration nor posterior sclerotome invasion was aggravated relative to single mutants (Fig. 3). Rather, p75 labelling demonstrated an excessive migration of NCCs alongside blood vessels in the intersomitic furrow and the perisomitic space, as seen in single Sema3a- and Nrp1-null mutants, but the remaining NCCs entered the sclerotome with no preference for the anterior or posterior half, as in single Sema3f- and Nrp2-null mutants (Fig. 3A,B).

The analysis of compound Nrp1^Sema Nrp2 null mutants by wholemount Sox10 in situ hybridisation further revealed a consistent heterogeneity of NCC defects along the rostrocaudal axis(Fig. 3C-F). Thus, excessive NCC migration in the intersomitic furrow and alongside perisomitic vessels,typical of Nrp1-null mutants, was prominent in the caudal trunk(arrows in Fig. 3C). By contrast, the indiscriminate migration through both anterior and posterior sclerotome, characteristic of Nrp2-null mutants, was obvious in the rostral trunk (curved line in Fig. 3C). Transverse sections at forelimb level, where both migratory patterns overlapped at 9.5 dpc, established that the ectopic NCC streams followed the same dorsoventral trajectory as in single mutants; thus, the intersomitic stream travelled on both dorsolateral and ventromedial paths, as in single Nrp1- and Sema3a-null mutants(Fig. 3D), whereas NCCs in the anterior and posterior sclerotome were confined to a ventromedial path, as in single Nrp2- and Sema3f-null mutants (compare Fig. 3I,J with Fig. 3E,F). Consistent with previous studies on ligand-receptor pairings in axon guidance(Schwarz et al., 2008b; Yaron et al., 2005), compound Sema3a Sema3f null mutants precisely phenocopied the defects of compound Nrp1^Sema Nrp2 null mutants (compare Fig. 3C-F with Fig. 3G-J). Because the caudal trunk is younger than the rostral trunk, the variation of NCC phenotype across the rostrocaudal axis in compound mutants implies that SEMA3A/NRP1 signalling acts first to direct NCCs from the intersomitic path into the sclerotome, and that SEMA3F/NRP2 signalling subsequently restricts NCC migration to the anterior sclerotome. We therefore conclude that NRP1 and NRP2 do not act redundantly at the level of individual NCCs, but cooperate to pattern NCC migration because of their functional requirement in distinct NCC subpopulations (see Fig. S1B in the supplementary material).

Even though NCC migration is severely affected in Nrp1- and Nrp2-null mutants, and Nrp1-null mutants have ectopic sensory neurons, DRG segmentation is preserved in both types of single mutants(Gammill et al., 2006; Schwarz et al., 2009). These findings suggest that the maintenance of a NCC-free region either in the intersomitic furrow or the posterior sclerotome is sufficient to ensure the segmentation of the sensory NCC progeny. To test whether DRG separation is compromised if both neuropilin-mediated mechanisms of NCC guidance are lost in tandem, we examined compound Nrp1^Sema Nrp2 null mutants by in situ hybridisation with probes for the sensory neuron markers Isl1(Ericson et al., 1992) and Brn3a (Pou4f1 - Mouse Genome Informatics)(Eng et al., 2004). Whereas DRG neurons were organised into metameric ganglia in single mutants(Fig. 4B,C), they were arranged in a continuous band in compound mutants, and DRG therefore appeared fused(Fig. 4D,H and data not shown). There were a few sensory-neuron-free spaces along the rostrocaudal axis in compound mutants, and these correlated with the position of blood vessels;however, their presence was not sufficient to enforce proper segmentation of the DRG (Fig. 4D,H). Wholemount neurofilament labelling further confirmed that the DRG were not segmented (see Fig. S2 in the supplementary material).

In contrast to the DRG, the spinal nerves of compound mutants did not lose their segmentation. Thus, the spinal nerves between the fore- and hindlimbs maintained a metameric pattern, even though they were severely defasciculated(Fig. 4L), as previously shown[fig. 3S in Huber et al.(Huber et al., 2005)]. Similarly, sensory axons projecting from the ventral and dorsal rami of the spinal nerves towards the skin emerged in a segmental pattern, despite prominent defasciculation and precocious growth (see Fig. S2 in the supplementary material). The segmentation of spinal nerve axons is thought to be achieved through a combination of contact repulsion by posterior somite cells and chemoattractive cues from anterior somite cells (reviewed by Kuan et al., 2004). The preservation of spinal nerve segmentation is therefore consistent with the finding that rostrocaudal somite polarity is not lost in semaphorin or neuropilin mutants (Fig. 2N)(Gammill et al., 2006).

Importantly, the compound Nrp1^Sema Nrp2 null mouse is the first model in which it has been shown directly that loss of DRG segmentation is due to defective NCC patterning. Previously, loss of DRG segmentation has been observed in mouse mutants with defects in somite polarity. For example, mice lacking the transcription factor UNCX (previously known as UNC4.1) show defects in specific aspects of rostrocaudal somite polarity that are crucial for sclerotome patterning(Leitges et al., 2000; Mansouri et al., 2000). The observation that somitic Sema3a expression is perturbed in these mutants (Mansouri et al.,2000) raises the possibility that UNCX controls semaphorin expression to coordinate the position of NCCs with sclerotome derivatives. It would therefore be interesting to examine whether Sema3f expression is also affected in Uncx-null mutants, because loss of SEMA3A/NRP1 signalling disrupts DRG segmentation only when SEMA3F/NRP2 is lost in tandem(Fig. 3).

Loss of DRG segmentation has also been observed in chick embryos, in which caudal half-somites were replaced with rostral half-somites(Kalcheim and Teillet, 1989). Because this experiment disrupted the rostrocaudal distribution of NCCs within the somites (Stern and Keynes,1987), it led to the hypothesis that the rostrocaudal polarity of NCC migration lays the foundation for DRG segmentation. However, the intrasomitic patterning of NCC migration is not essential for DRG segmentation(Gammill et al., 2006). Because the juxtaposition of two rostral or caudal half-somites also perturbs intersomitic boundary formation (Stern and Keynes, 1987), we propose that the DRG phenotype exhibited upon transplantation of multiple rostral half-somites is similar to that of compound Nrp1^Sema Nrp2 null mutants, as intra- and intersomitic NCC guidance pathways are simultaneously disrupted in both experimental models.

Supplementary material available online at http://dev.biologists.org/cgi/content/full/136/11/1785/DC1