Neuropilin 1 and 2 control cranial gangliogenesis and axon guidance through neural crest cells

The cranial ganglia of vertebrates contain sensory neurons, which are derived from neurogenic placodes and neural crest cells(D'Amico-Martel and Noden,1983). The neurogenic placodes are focal thickenings in the head epidermis, whereas neural crest cells are a transient population of multipotent stem cells that delaminates from the neural tube and disseminates throughout the body to give rise to several other cell types in addition to sensory neurons (reviewed by Graham,2000; Graham,2003; Trainor,2005). Cranial neural crest cells also contribute to cranial gangliogenesis by organising the placodally-derived sensory neurons and their axons in the chick (Begbie and Graham,2001). Thus, hindbrain extirpation prior to cranial neural crest cell emigration causes mispositioning of placodally-derived neurons.

The emigration of cranial neural crest cells from the hindbrain follows its segmental organisation into morphologically distinct compartments termed rhombomeres (r1-r7); specifically, neural crest cells emanate in distinct streams from r2, r4 and r6, which are separated by crest-free zones at the levels of r3 and r5 (Lumsden et al.,1991; Sechrist et al.,1993). Neural crest cells actively avoid the cranial mesenchyme at r3 and r5 level, which suggests that inhibitory guidance mechanisms are present in these areas (e.g. Farlie et al.,1999; Kulesa and Fraser,1998; Kulesa and Fraser,2000; Sechrist et al.,1993). Members of the class 3 semaphorin family are good candidates for mediating neural crest cell guidance, as they provide repulsive cues for a number of different cell types, most notably migrating neurons and their axons (reviewed by Huber et al.,2003).

Consistent with the idea that class 3 semaphorin/neuropilin (SEMA3/NRP)signalling guides cranial neural crest cells, SEMA3F and its receptor NRP2 are expressed in a complementary pattern during cranial neural crest migration,with SEMA3F being expressed in r3 and r5 and NRP2 being expressed by the r2-derived (trigeminal) and r4-derived (hyoid) neural crest cells(Gammill et al., 2007; Eickholt et al., 1999). Moreover, cranial neural crest cells travel through the normally crest-free zone at r3 level when semaphorin function is perturbed by ectopic NRP2 expression in the chick (Osborne et al.,2005) and when NRP2 or SEMA3F are ablated by targeted gene inactivation in the mouse (Gammill et al.,2007). Like SEMA3F, SEMA3A is expressed in r3 and r5 of the chick hindbrain (Eickholt et al.,1999), and the ectopic expression of a soluble form of its receptor, NRP1, in the chick hindbrain causes invasion of neural crest cells into the normally neural crest-free territory adjacent to r3(Osborne et al., 2005). However, it is not known whether NRP1 is a major factor controlling neural crest guidance in mammals, as conflicting data exist with respect to the role of NRP1 in neural crest guidance in the mouse.

On the one hand, it has been suggested that NRP1 is not expressed in mouse neural crest cells and that SEMA3A is not expressed in a pattern consistent with a role in neural crest cell migration(Kuan et al., 2004). On the other hand, it has been shown that antibodies specific for NRP1 recognise mouse neural crest cells in vitro, and that dorsal root ganglia are organised more loosely in Nrp1-null mutants, even though the same study reported that trunk neural crest cells migrate normally(Kawasaki et al., 2002).

To reconcile the conflicting information provided by previous studies, we have re-examined the expression pattern and functional requirement of NRP1 during cranial neural crest cell migration in the mouse. Contrary to previous reports, we have found that murine hyoid neural crest cells express Nrp1, and that SEMA3A/NRP1 signalling is essential to guide their migration. Furthermore, we have discovered that SEMA3A/NRP1 and SEMA3F/NRP2 cooperate to guide cranial neural crest cells, and that neuropilin-mediated neural crest cell patterning is essential for cranial ganglia segmentation. Based on our findings, we propose a novel and comprehensive model that explains the multiple roles of neuropilin signalling in the peripheral nervous system of the vertebrate head.

To obtain mouse embryos of defined gestational ages, mice were mated in the evening, and the morning of vaginal plug formation was counted as 0.5 days post-coitum (dpc). To stage-match embryos within a litter, or between litters from different matings, we compared somite numbers. Mice carrying a Sema3a-null (Taniguchi et al.,1997) or Nrp1-null(Kitsukawa et al., 1997)allele, and mice carrying mutations that disrupt semaphorin signalling through both neuropilins (Nrp1^sema-/- Nrp2^-/-)(Gu et al., 2003) have been described. Conditional-null mutants for Nrp1(Nrp1^fl/fl) (Gu et al., 2003) were mated to mice expressing CRE recombinase under the control of the endothelial-specific Tie2 promoter(Kisanuki et al., 2001) or the neural crest-specific Wnt1 promoter(Jiang et al., 2000). The Erbb4-null allele has been described(Gassmann et al., 1995). Genotyping protocols are available on request.

In situ hybridisation was performed according to a previously published method (Riddle et al., 1993),using digoxigenin-labelled riboprobes transcribed from cDNA-containing plasmids. Plasmids encoding Nrp1 and Sox10 were provided by M. Fruttiger and N. Kessaris (University College, London, UK), respectively. Plasmids containing Sema3a or Sema3f were obtained from M. Tessier-Lavigne (Genentech). Plasmids containing Phox2b, Ngn1 and Brn3b were provided by C. Goridis (INSERM, Marseille, France), F. Guillemot (National Institute for Medical Research, London, UK) and E. Turner(University of California, La Jolla, CA), respectively.

Mouse embryos were fixed in 4% formaldehyde in PBS, washed in PBS and incubated for 2 hours in a blocking solution of PBS containing 0.1% Triton X-100 (PBT) and 10% goat serum. In some experiments, samples were subjected to in situ hybridisation prior to immunolabelling. The following primary antibodies were used: for blood vessels, rat anti-endomucin (Santa Cruz Biotechnology); for mouse neural crest cells, rabbit anti-p75 (a gift of Drs K. Deinhardt and G. Schiavo, Cancer Research UK, London); for neuronal cell bodies, mouse anti-HUC/D (Molecular Probes); for axons, rabbit anti-neurofilament (Chemicon). Samples were washed in PBT and incubated overnight at 4°C with secondary antibodies in blocking solution. Secondary antibodies used were Alexa 488-conjugated goat anti-rat, and Alexa 594-conjugated goat anti-rabbit IgGs (Molecular Probes). In some experiments,horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (DAKO) was used as a secondary antibody. Neurofilament-stained samples were dehydrated in methanol and cleared in a solution containing 2:1 benzyl benzoate:benzyl alcohol. HRP-labelled samples were visualised by conventional light microscopy, fluorescently labelled samples with a LSM510 laser-scanning confocal microscope (Zeiss).

In the chick, trigeminal and hyoid neural crest cells express Nrp1(Eickholt et al., 1999). To address whether NRP1 plays a role during cranial neural crest guidance in the mouse, we determined its expression pattern during the period of neural crest migration from the hindbrain into the pharyngeal arches using wholemount in situ hybridisation. We found that Nrp1 was not expressed in murine trigeminal neural crest cells; these were, however, clearly identified with the neural crest marker Sox10(Fig. 1, compare A with C and B with D). By contrast, the murine hyoid neural crest stream expressed both Nrp1 and Sox10 (Fig. 1, compare black arrowheads in A,B with C,D). In addition, Nrp1 was expressed in the second pharyngeal arch mesenchyme(Fig. 1C,D, open arrowheads). Double labelling of coronal sections from an 8.5 dpc embryo by in situ hybridisation with the Nrp1 probe and immunocytochemistry with antibodies specific for the p75 neurotrophin receptor (NGFR), a neural crest marker (Rao and Anderson,1997), confirmed Nrp1 expression in the p75-positive hyoid neural crest stream (Fig. 1J-L, black arrowheads) and in the p75-negative distal pharyngeal arch mesenchyme (Fig. 1J,L,open arrowheads).

In the chick, Sema3a is expressed in r3 and r5 around the time of neural crest cell delamination (Eickholt et al., 1999). In the mouse, Sema3a was expressed only weakly within the hindbrain during the time of neural crest cell emigration(Fig. 1G). By contrast, it was expressed prominently in a stripe between the first and second pharyngeal arches (Fig. 1E, arrow); this expression domain originated just outside the hindbrain at the r3 level(Fig. 1G, arrows). Sema3a is therefore expressed in an appropriate position to prevent intermingling of migrating trigeminal and hyoid neural crest cells. In contrast to Sema3a, Sema3f was expressed strongly in the mouse hindbrain, with expression first in r3(Fig. 1F,H, arrows), and then in r3 and r5, consistent with previous reports(Gammill et al., 2007). In addition, both semaphorins were co-expressed in the distal arch mesenchyme. Taken together, the Sema3a and Sema3f expression patterns overlap only partially in the mouse embryo head at this stage. Importantly,the expression patterns of Sema3a and Nrp1 raise the possibility that SEMA3A signalling cooperates with SEMA3F to provide inhibitory cues for hyoid neural crest cells, preventing their invasion into the mesenchymal territory adjacent to r3(Fig. 1I).

To determine whether SEMA3A signalling through NRP1 is required for the guidance of hyoid neural crest cells in the mouse, we examined mouse embryos lacking either SEMA3A or NRP1 (Kitsukawa et al., 1997; Taniguchi et al., 1997) with the neural crest marker Sox10(Pusch et al., 1998). At 9.5 dpc, we observed ectopic neural crest cells in the normally neural crest-free area adjacent to r3 in 9/9 Nrp1-null mutants(Fig. 2B,E,K, arrowheads) and in 12/12 Sema3a-null mutants (Fig. 2C,F,L, arrowheads). Although all mutants examined showed such defects, in 4/9 Nrp1-null embryos and in 6/12 Sema3a-null embryos only the left or the right side was noticeably affected. In all cases,ectopically migrating neural crest cells travelled between the hyoid and trigeminal neural crest streams at the level of the dorsal neural tube(Fig. 2K,L and Fig. 3C, arrowheads). In 5/12 Sema3a-null and 4/9 Nrp1-null mutants, neural crest cells were additionally crossing between the trigeminal and hyoid neural crest streams in a more ventral region (Fig. 2B,F, lower arrowheads). In 1/9 Nrp1-null and 4/12 Sema3a-null mutants, the point of origin for the neural crest cells invading the territory adjacent to r3 was clearly identifiable as r4 (for example, Fig. 2C and Fig. 3B). In rare cases (one Sema3a-null and one Nrp1-null mutant), a small subset of trigeminal neural crest cells appeared to move into the territory adjacent to r3 (Fig. 2E, arrow).

NRP1 has been implicated in axon guidance as well as in blood vessel growth(Kawasaki et al., 1999; Kitsukawa et al., 1997). Accordingly, NRP1 has been suggested to play a role in the co-pattering of nerves and blood vessels, a subject of much recent interest(Carmeliet, 2003). We therefore considered the possibility that the neural crest defects were secondary to blood vessel defects in Nrp1-null mutants. Double labelling of 9.5 dpc embryos with the neural crest cell marker p75 and the blood vessel marker endomucin (Brachtendorf et al.,2001) revealed that cranial neural crest cells(Fig. 3A) normally migrate in close proximity to the anterior cardinal vein(Fig. 3D, arrows; Fig. 3G, overlay). Ectopic neural crest cells in Nrp1-null mutants also appeared to migrate in a rostral direction in close proximity to the anterior cardinal vein(Fig. 3H,I, arrowhead). We therefore used a mouse strain carrying a conditional Nrp1-null mutation (Gu et al., 2003) to directly address whether abnormal neural crest migration in the absence of NRP1 was secondary to defective blood vessel growth. We found that the tissue-specific ablation of NRP1 from blood vessel endothelium with CRE recombinase driven by the Tie2 (Tek - Mouse Genome Informatics) promoter (Kisanuki et al.,2001) did not impair cranial neural crest guidance in any of the four mutants examined (Fig. 3K), even though this mutation perturbs blood vessel patterning(Gu et al., 2003). Rather,ablation of NRP1 specifically from neural crest cells with CRE recombinase driven by the Wnt1 promoter(Jiang et al., 2000)phenocopied the defects seen in full Nrp1-null mutants (6/8 cases;compare arrowheads in Fig. 2B,Ewith that in Fig. 3L). These observations demonstrate that NRP1 plays a cell-autonomous role in the guidance of cranial neural crest cells.

In the mouse, loss of the neuregulin receptor ERBB4, which is normally expressed in r3 and r5, causes invasion of hyoid neural crest cells into the normally neural crest-free mesenchyme adjacent to r3(Golding et al., 2002; Golding et al., 2000). It has therefore been hypothesised that ERBB4 controls the expression of a guidance cue that normally repels hyoid neural crest cells, perhaps by inducing its expression in the territory that separates the trigeminal and hyoid neural crest streams. To test the idea that ERBB4 is required for cranial neural crest cell patterning because it regulates Sema3a expression, we examined the expression pattern of Sema3a in the absence of ERBB4. We found that Sema3a was expressed normally in 5/5 Erbb4-null mutants (Fig. 4). This observation suggests that ERBB4 controls a signalling pathway that is distinct from, but cooperates with, SEMA3A/NRP1 signalling.

NRP2 signalling has previously been reported to exclude neural crest cells from the mesenchymal territory adjacent to r3(Eickholt et al., 1999; Gammill et al., 2007). In agreement, we found that 3/7 Nrp2-null mutants displayed unilateral invasion of neural crest cells into this region(Fig. 5B, arrowhead). However,the penetrance of this defect appeared lower in Nrp2-null than in Nrp1-null mutants (3/7 versus 12/12 cases, respectively). We next asked whether SEMA3F/NRP2 and SEMA3A/NRP1 signalling act synergistically during cranial neural crest cell guidance. Given that lack of both NRP2 and NRP1 is lethal prior to 9.5 dpc owing to severe vascular defects(Takashima et al., 2002), we took advantage of a mouse mutant that is deficient in semaphorin signalling,but not in vascular endothelial growth factor signalling through NRP1, and which therefore survives to birth (Nrp1^sema-/-Nrp2^-/-) (Gu et al.,2003). In 3/3 mutants lacking semaphorin signalling through both NRP1 and NRP2, the territory adjacent to r3 was invaded by ectopic neural crest cells much more heavily than in single-null mutants, and the trigeminal and hyoid crest streams appeared to fuse (compare Fig. 5C with Fig. 5B and Fig. 2B,E). SEMA3A/NRP1 and SEMA3F/NRP2 therefore act synergistically to exclude neural crest cells from the mesenchyme at the r3 level and to separate the trigeminal and hyoid neural crest cell streams.

Cranial neural crest cells contribute only a small proportion of sensory neurons to the cranial ganglia(D'Amico-Martel and Noden,1983), but they are essential for neuronal development in the cranial ganglia by organising placodally-derived sensory neurons and their projections (Begbie and Graham,2001). We therefore asked whether ectopic cranial neural crest cell migration in neuropilin-null mutants affected sensory neuron positioning(Fig. 6). Labelling of neuronal cell bodies at 10.5 dpc with an antibody specific for HUC/D (ELAVL3/4)proteins (Wakamatsu and Weston,1997) identified ectopic neurons between the proximal parts of the facioacoustic and trigeminal ganglia in 5/5 Nrp1-null mutants; in 1/5 cases, ectopic neurons were also seen between the distal parts of the facioacoustic and trigeminal ganglia (Fig. 6H, arrowheads). The location of these ectopic neurons correlated with the position of ectopic neural crest cells at earlier stages (compare Fig. 6H with Fig. 2B,E). Similarly, ectopic neurons were found in positions prefigured by ectopic neural crest cells in Nrp2-null mutants (data not shown). The number of ectopic neurons increased dramatically when semaphorin signalling through NRP1 was reduced on a Nrp2-null background (Fig. 6, compare circled area in I with the equivalent area in H). Consequently, there was no clear separation between the trigeminal and facioacoustic ganglia in 3/3 compound mutants(Fig. 6C,F). This dramatic defect in gangliogenesis reflected the greater severity of the neural crest phenotype in compound mutants compared with single mutants at earlier stages(compare Fig. 6A with Fig. 2B, and Fig. 6C with Fig. 5C). Taken together, our observations suggest that abnormal cranial neural crest migration in neuropilin mutants directly affects sensory neuron localisation.

Loss of NRP1 or NRP2 results in ectopic projections and defasciculated nerve tracts in the peripheral nervous system(Giger et al., 2000; Kitsukawa et al., 1997). These axonal defects are generally attributed to the loss of semaphorin signalling at the axonal growth cone (e.g. Luo et al., 1993). We observed that the abnormal positioning of sensory neurons in neuropilin mutants also contributed to the disorganisation of sensory projections (Fig. 6). Thus, ectopic neurons in Nrp1-null mutants(Fig. 6H, arrowheads) extended axons between the trigeminal and facioacoustic ganglia(Fig. 6K, arrowheads). Similarly, ectopic neurons in Nrp2-null mutants also extended aberrant axons (data not shown). The axon tracts of compound mutants, with their greater number of ectopic neurons, were even more disorganised than those of single mutants: in areas where ectopic sensory neurons were situated in compound mutants (circled in Fig. 6I), a large number of axons extended through the normally axon-free space between the trigeminal and facioacoustic ganglia (circled in Fig. 6L). In addition, axon guidance defects within the pharyngeal arches were more severe in compound than in single mutants (Fig. 6, compare A with C).

The disorganisation of cranial sensory axons was also obvious in 11.5 dpc wholemount neurofilament stains (Fig. 7). Specifically, the trigeminal ganglion appeared to project posteriorly towards the geniculate ganglion in 4/5 mutants lacking NRP2(Fig. 7B), and the trigeminal and facioacoustic ganglion appeared fused in 2/2 mutants lacking semaphorin signalling through both NRP1 and NRP2 (Fig. 7C). Taken together, our findings are consistent with the idea that NRP1 and NRP2 play essential and non-redundant roles during growth cone guidance and axon fasciculation.

The role of neural crest cells in guiding placodal sensory neurons was previously demonstrated in the chick by surgical ablation of hindbrain segments prior to neural crest emigration(Begbie and Graham, 2001). Even though this experiment was consistent with the idea that cranial neural crest cells play a key role in the guidance of placodal neurons and their axons, it had technical limitations; specifically, the removal of whole hindbrain segments might have disturbed gangliogenesis indirectly by impairing communication between hindbrain tissue and head mesenchyme. Because neural crest cells are misrouted in Sema3a and Nrp1 mutants in the absence of gross alterations to other head structures, they present an ideal model system with which to test the hypothesis that neural crest cells organise placodal sensory neurons. Moreover, they allow us to determine the placodal origin of the ectopic neurons, as specific molecular markers exist for the different placodes in the mouse.

The ectopic neurons in Nrp1-null mutants are located between the trigeminal ganglion, which receives neurons from the trigeminal placode, and the facioacoustic ganglion complex, which comprises the vestibuloacoustic ganglion in its proximal part and the geniculate ganglion in its distal part. Whereas the vestibuloacoustic ganglion receives neurons from the otic placode,the geniculate ganglion receives neurons from the first epibranchial placode(Graham and Begbie, 2000). To identify the placodal origin of the ectopic neurons, we therefore used Ngn1 (Neurog1 - Mouse Genome Informatics) as a marker for trigeminal placode-derived neurons (Ma et al., 1998), Phox2b as a marker for geniculate neurons(Fode et al., 1998), and Brn3b (Pou4f2) as marker for otic placode-derived neurons(Eng et al., 2004; Huang et al., 2001)(Fig. 8C,E,G). In addition, we used the general neuronal marker Isl1(Begbie et al., 2002) to demonstrate that ectopic sensory neurons could be identified by in situ hybridisation (Fig. 8A,B).

The ectopic neurons did not express the geniculate neuron marker Phox2b (Fig. 8D),suggesting that they were not derived from the first epibranchial placode. Moreover, they were not labelled with the Ngn1 probe(Fig. 8F), suggesting that they were not derived from the trigeminal placode either. As Ngn1additionally labels neurons derived from trigeminal neural crest cells(Ma et al., 1998), this experiment also ruled out the possibility that ectopic neurons in Nrp1-null mutants had differentiated from trigeminal neural crest cells, which had moved caudally. Subsequent neurofilament staining confirmed that ectopic bridges were present between the trigeminal and facioacoustic ganglia in all mutants examined with Phox2b and Ngn1 (data not shown). Thus, ectopic neurons were likely to be derived from the otic placode; consistent with this idea, the Brn3b marker was expressed by the ectopic neurons (Fig. 8H). Even though this marker also labels trigeminal neurons, the lack of Ngn1 expression implied that the ectopic neurons must have been derived from the otic placode. Taken together, the data presented here demonstrate that NRP1-dependent neural crest cells control the positioning of placodal sensory neurons.

SEMA3A and NRP1 have previously been implicated in neural crest patterning in the chick. However, it was generally assumed that this role was not evolutionary conserved (Gammill et al.,2006; Kawasaki et al.,2002; Kuan et al.,2004). The data presented here demonstrate that this assumption was incorrect, as SEMA3A/NRP1 signalling plays a key role in cranial neural crest patterning in the mouse. Specifically, SEMA3A/NRP1 signalling excludes hyoid neural crest cells from the cranial mesenchyme at the level of hindbrain rhombomere r3 and thereby prevents the intermingling of the trigeminal and hyoid neural crest streams (Fig. 2). SEMA3A is likely to act as a repulsive cue for hyoid neural crest cells because: (1) Sema3a is expressed in an area between the trigeminal and hyoid streams (Fig. 1); (2) hyoid neural crest cells express Nrp1(Fig. 1); and (3) hyoid neural crest cells require Nrp1 in a cell-autonomous fashion(Fig. 3).

The cranial neural crest defect of Nrp1-null mouse mutants resembles the defect seen in chick embryos when a vector expressing soluble NRP1 protein was electroporated into the hindbrain prior to neural crest cell emigration (Osborne et al.,2005). There are, however, several important differences between our previous chick and current mouse analyses. Firstly, the chick experiments did not attempt to demonstrate a specific role for SEMA3A, whereas the analysis of Sema3a mouse mutants established that it is indeed the repulsive signal detected by the NRP1-expressing cranial neural crest cells. Secondly, the Sema3a expression pattern during cranial neural crest cell migration differs in chick and mouse: whereas Sema3a and Sema3f are expressed in an overlapping pattern in r3 and r5 during neural crest delamination in the chick, Sema3a is expressed only weakly in the mouse hindbrain, but prominently in the periphery at this stage. These differences in the Sema3a and Sema3f expression patterns in the mouse might reflect an evolutionary divergence in the mechanisms that control cranial neural crest streaming in these two organisms. Importantly, Sema3a is expressed in the mouse in the correct spatiotemporal fashion to directly affect migrating neural crest cells at a short distance. This observation therefore eliminates the need to postulate a long-range SEMA3A gradient emanating from the hindbrain beyond the perineural membranes into the head mesenchyme. In addition, the overlapping expression patterns of Sema3a and Sema3f in the distal pharyngeal arch mesenchyme might present a barrier to neural crest migration as they invade the arches.

The non-overlapping expression patterns of Sema3a and Sema3f in conjunction with the expression pattern of their neuropilin receptors suggested that SEMA3A/NRP1 and SEMA3F/NRP2 act in a synergistic fashion to prevent the mixing of the trigeminal and hyoid neural crest cell streams. Consistent with this idea, compound mutants lacking semaphorin signalling through both NRP1 and NRP2 had a more severe defect than single mutants, with extensive intermingling of the trigeminal and hyoid neural crest streams in the mesenchyme at r3 level (Fig. 5).

Both cranial neural crest cells and neurogenic placodes give rise to the sensory neurons in the cranial ganglia (e.g. D'Amico-Martel and Noden,1983), and their development must therefore be coordinated. Previous experiments suggested that cranial neural crest cells play a key role in the guidance of placodal neurons and their axons in the chick(Begbie and Graham, 2001). We have now substantiated the concept that cranial neural crest cells orchestrate cranial gangliogenesis and have provided a molecular mechanism for neural crest cell positioning upstream of cranial gangliogenesis. Thus, in the absence of either NRP1 or NRP2, ectopic neurons were found in locations that were prefigured by the position of ectopic neural crest cells (compare Figs 2 and 5 with Fig. 6). Using a collection of different molecular markers for placodal neurons, we then established that the ectopic neurons of Nrp1-null mutants had placodal identity(Fig. 8).

Importantly, mutants lacking semaphorin signalling through both NRP1 and NRP2 (Figs 6, 7) displayed an even more extensive disorganisation of the cranial sensory nervous system, with a lack of separation between the trigeminal and facioacoustic ganglia and extensive misprojections between both ganglia; these defects occurred in addition to the extensive axon guidance defects within the pharyngeal arches that we had anticipated based on the reported phenotypes of the single mutants(Giger et al., 2000; Kitsukawa et al., 1997). Owing to the abnormal positioning of sensory neurons, the organisation of the trigeminal and facioacoustic ganglia no longer reflected the segmental nature of the hindbrain and pharyngeal arches. Based on the severity of these defects, we predict that more than one type of placodal neuron will be affected in double mutants. Even though the paucity of double mutants(frequency of 1:16 in an average litter size of <8 embryos) has precluded us from performing a placodal marker analysis similar to that which we carried out for single Nrp1 mutants, the observation that Brn3b is also expressed by ectopic neurons in 2/4 Nrp2-null mutants (data not shown) provides further support for the idea that NRP1 and NRP2 pathways cooperate to organise the position of placodal sensory neurons.

In conclusion, our study confirms that the neuropilin-mediated guidance of cranial neural crest cells is an essential prerequisite for the ordered positioning of cranial sensory neurons in the head and for the spatial separation of the cranial ganglia. We therefore propose a new working model to explain the role of class 3 semaphorin/neuropilin signalling in the sensory nervous system (Fig. 8I): by acting on neuropilin/plexin receptors, SEMA3A and SEMA3F synergise to repel axonal growth cones (Huber et al.,2003) and cranial neural crest cells (this study; indicated by the green box in Fig. 8I). As cranial neural crest cells in turn determine the position of sensory neuron cell bodies (Begbie and Graham,2001), semaphorin/neuropilin signalling plays a novel role in patterning both cell bodies and axons in the peripheral nervous system.

We thank Drs Hajime Fujisawa, Masahiko Taniguchi, Andrew McMahon, Masahi Yanagisawa, Dorothy Reimert, David D. Ginty and Alex L. Kolodkin for mouse strains; Drs Marcus Fruttiger and Nicoletta Kessaris for advice and reagents;Drs Anthony Graham and Matthew Golding for invaluable discussions; the staff of the Biological Resources Unit for help with mouse husbandry; and Dr John Greenwood for support. This research was funded by a project grant of the Medical Research Council to C.R. and a PhD studentship from the Fundação para a Ciência e Tecnologia (SFRH/BD/17812/2004)to J.M.V.