Previously, we proposed that neural crest BC cells at the motor exit point (MEP) express repellent signals that help confine motor neuron somata within the chick embryo spinal cord as their axons first emerge into the periphery 18. This requires that candidate receptors for putative repellents should be expressed by motor neurons from embryonic day (E) 2 when the first motor axons exit the spinal cord at the MEP. Available expression data in mouse and chick embryos suggested that prominent amongst these were the multimeric surface receptors for secreted class 3 semaphorins, neuropilins and/or plexins 131415. We therefore analysed in detail expression of chick neuropilins and the three known members of the chick Plexin-A family, Plexin-A1, -A2 and -A4 (there is no chick Plexin-A3 orthologue 15) during the appropriate period. Our results confirm and extend those of Mauti et al. 15, showing dynamic patterns of expression in spinal motor neurons, with Npn-1, and Plexin-A1 and -A2 already present in progenitors at Hamburger Hamilton (HH) stage 18 (not shown). We found that Npn-2 is also expressed in motor neurons (or their progenitors) as early as HH stage 18, and in rostral parts of the embryo and throughout the embryo by HH stage19 (not shown). At HH stage 20 all Npn and Plexin-A receptors are expressed by motor neurons to some extent (Figure 1b–f), with Npn-1 most prominently and distinctly expressed. Plexin-A1 and -A2 are widely expressed throughout the ventral spinal cord at this stage. Within the spinal cord, Npn-2 and Plexin-A4 expression is restricted to laterally positioned motor neurons.

At the stage when we observed ectopic migration of motor neuron cell bodies following neural crest ablation (HH stage 23) 18, we found strong expression of npn-1 in most motor neurons (Figure 1h). Npn-2 is expressed in medial and lateral subsets of motor neurons but excluded from motor neurons located between these (Figure 1i). Plexin-A1 (Figure 1j) is more diffusely expressed throughout the ventral spinal cord, whereas Plexin-A2 (Figure 1k) and -A4 (Figure 1l) expression is restricted to lateral and medial motor pools, respectively. These expression studies show that all chick neuropilin and Plexin-A receptors examined could potentially regulate motor neuron cell body positioning.

To assess the involvement of these receptors we applied a loss-of-function strategy in the chick, with vectors co-expressing shRNA and enhanced green fluorescent protein (EGFP), which marks those cells targeted by RNAi in an otherwise wild-type background 19.

We first knocked down Npn-1 and Npn-2, using shRNA vectors described previously 1920. We used an electroporation protocol that targets the plasmid DNA towards the ventral neural tube (for details see Materials and methods and Additional file 1). By doing so, effects on migration of the neural crest derivatives, including BC cells, can be excluded, an important consideration since both Npn-1 and Npn-2 have been implicated in neural crest migration 2122.

We quantified the effect of shRNA-EGFP electroporation, compared to that of EGFP only, by counting ectopically positioned, EGFP-positive motor neurons. These were localised outside the spinal cord, along the ventral roots, and were characterised by their large soma linked to both trailing and leading EGFP-positive processes. Their identities were confirmed by dual labelling with antibodies to the transcription factors Islet-2 and motor neuron restricted protein (MNR)2. In our initial experiments with Npn-2 specific shRNA, we consistently observed ectopic EGFP/Islet-2/MNR-2 positive motor neurons in presumptive white matter and ventral roots (Figure 2a,b), but not with Npn-1 specific shRNA (Figure 2g). We subsequently generated four more vectors targeting Npn-2 and screened these for the potential to reduce Npn-2 expression in a co-expression assay 19. One of these constructs, Npn-2 E shRNA, was as efficient as Npn-2 B shRNA in reducing Npn-2 expression (Additional file 2) and also consistently induced ectopic motor neurons, when electroporated in the ventral neural tube (Additional file 3).

Neuropilins do not signal independently but upon ligand binding associate with other molecules to confer an intracellular signal. Members of the Plexin-A family in particular have been implicated as obligate co-receptors 23 (for a recent review, see 24). Therefore, we generated shRNA vectors targeting the three known members of the chick Plexin-A family, Plexin-A1, -A2 and -A4 (for details on shRNA vectors see Materials and methods and Additional file 4). When we screened these Plexin shRNA constructs for the potential to induce ectopic migration of motor neurons, one shRNA vector, targeting plexin-A2 (Plexin-A2A shRNA), substantially and selectively knocked down expression in motor neurons and triggered ectopic migration of motor neurons (Figure 2e,f; Additional files 5 and 6). In contrast, those vectors targeting Plexin-A1 (Figure 2c,d) or -A4 (not shown) that efficiently reduced expression of their target genes (Additional file 5) had no effect on motor neuron cell body positioning (Figure 2g and Additional file 3).

We constructed three more shRNAs targeted at Plexin-A2 and one of these, Plexin-A2D, also generated many ectopic motor neurons (Additional file 3).

In conclusion, quantitative analysis of these loss-of-function studies (Figure 2g) clearly identified Npn-2 and Plexin-A2 as receptors for ligands regulating motor neuron cell body position.

In order to test whether this mechanism was conserved in a mammalian species, we examined Npn-2 null mice for evidence of ectopically positioned motor neurons. Npn-2 mutant mice have been analysed extensively in a number of studies of axon guidance phenotypes 13252627 and neuron migration 162228 but defects in spinal motor neuron migration have not been reported. Yet we identified large numbers of ectopically positioned HB9-positive motor neuron cell bodies in Npn-2 null mouse embryos (Figure 3b). Interestingly, this phenotype appeared to be confined to more caudal regions of the trunk, prompting an analysis of the incidence of ectopic motor neurons throughout the rostro-caudal axis. A typical result of such an analysis in serial cryosections taken throughout the trunk revealed ectopically positioned HB9-positive motor neurons were largely restricted to the hindlimb level (Figure 3c). At forelimb and thoracic levels there was no significant increase above the background level in wild-type littermate embryos. The pooled results of several E12.5 embryos analysed in this manner (n = 4 for each genotype) confirmed the presence of large numbers of ectopic motor neurons in the hindlimb region of Npn-2 null mice (Figure 3d).

Although the outcome of targeted Npn-2 RNAi in the chick spinal cord indicates that the ectopic motor neuron phenotype could be a cell-autonomous effect, it remained a possibility that a reduced prevalence of BC cells at the MEP might contribute to the phenotype in Npn-2 null mice since the absence of Npn-2 leads to aberrant neural crest migration 22. Therefore, using a specific marker for BC cells (Egr2/Krox20 29) we tested for the presence of these cells in Npn-2 null mouse embryos. The results (Additional file 7) show that at hindlimb level there was no obvious difference in egr2 expression at the MEP between Npn-2 null mice, and heterozygous and wild type littermates. Since in earlier grafting experiments in chick embryos we showed that small numbers of BC cells at the MEP were sufficient to confine motor neuron soma within the spinal cord 18, this supports the idea that the ectopic migration of motor neurons in Npn-2 null mice is a cell-autonomous effect.

When expression patterns of potential Npn-2 ligands in the chick were examined we found that Sema3B and Sema3G 30 are distinctly expressed by cad-7 expressing BC cells (Figure 4a) both at the dorsal root entry zone (DREZ; Figure 4b,b',e,e', red arrows) and the MEP (black arrows). In contrast, neither Sema 3C, prominently expressed in motor neurons (Figure 4c,c'), nor Sema3F, expressed broadly within somites (Figure 4d,d'), were localised to BC cells.

The finding that knockdown of Plexin-A2 compared to Npn-2 gave rise to greater numbers of ectopic motor neurons suggested that its role was more extensive than as just a Npn-2 co-receptor. Since the repellent activity of the transmembrane semaphorin Sema6A is mediated in sympathetic axons and hippocampal mossy fibres by direct interactions with Plexin-A2 and -A4 3132 we next looked for evidence of its expression. We found that Sema6A is expressed by both dorsal and ventral BC cells (Figure 4f,f').

Thus, we conclude that BC cells express at least three candidate repellent ligands for receptors on motor neurons identified in the RNAi screen, Sema3B, Sema3G and Sema6A.

When we targeted Sema3B expression in the chick embryo neural crest by RNAi, we found no increase in the number of ectopic motor neurons (data not shown). However, since only small numbers of transplanted BC cells at the MEP suffice to stabilise motor neuron cell body position in the crest-ablated chick spinal cord 18, the possibility remained that residual untransfected BC cells were sufficient to mask the effects of RNAi.

Therefore, we next analysed Sema3B mutant mice 27. We first confirmed that Sema3B is also strongly expressed in mouse BC cells (Figure 5a). Analysis of the incidence of ectopic motor neurons throughout the rostro-caudal axis, however, did not reveal a difference between Sema3B mutant mice and heterozygous and wild-type littermates (Figure 5b). Therefore, loss of Sema3B expression despite its distinct expression by BC cells has no effect on positioning of motor neuron cell bodies.

The finding of sema6A expression in chick motor exit point BC cells led us to examine expression in mouse. At E11.5 a distinct Sema6A signal is seen in motor exit point BC cells (Figure 6a, black arrowheads), in addition to prominent expression in the neural tube ventricular zone. Co-labelling with Egr2 (Krox-20) in an adjacent section (Figure 6a') confirmed expression was in BC cells. This prompted us to look for evidence of ectopic motor neuron positioning in Sema6A null mice 33. Like in Npn-2 null mice, here we found numerous ectopic motor neurons in ventral nerve roots (Figure 6b,c). Also, in common with Npn-2 null mice, HB9-positive ectopic motor neurons are most prevalent at the level of the hindlimb (Figure 6d,e). An effect on neural crest migration and on the formation of boundary caps could account for this phenotype. However, as judged by Egr2 expression, there was no difference between Sema6A mutants and wild-type littermates (Figure 6f,g) in the arrangement of BC cells at the MEP and DREZ.

Since at later stages in the mouse Sema6A is expressed not only in BC cells but also in spinal motor neurons 14 (data not shown), it could not be excluded that the effect on motor neuron positioning in Sema6A null mice was due to its absence from motor neurons. Therefore, we next tested in chick embryos the effects of targeted knock down of Sema6A, either ventrally in motor neurons, or dorsally in the neural crest (Additional file 1). We found a significant increase in ectopic motor neurons, but only when Sema6A shRNA was introduced in the neural crest (Figure 7a,b), not in the ventral neural tube (Figure 7c). Knockdown of Npn-2 in the neural crest or ventral neural tube had the converse effect, with ectopic motor neurons found only after expression of Npn-2 shRNA in the ventral neural tube (Figure 7c). Finally, we analysed the effect on boundary cap formation of Sema6A knockdown in the crest. The data show that Sema6A shRNA-expressing neural crest cells are normally positioned at the MEP and these correspond to cad7-positive BC cells (Figure 7d, white arrows). Together, these results support the idea that the effect of Sema6A loss of function on motor neuron positioning is explained by loss of its expression in BC cells, disrupting a putative interaction with motor neurons.

Having demonstrated that cell surface molecules involved in semaphorin signalling are implicated in regulating motor neuron cell body positioning, we next looked for possible common downstream cytoplasmic mediators. Members of the MICAL (molecule interacting with CasL) family of flavoprotein monooxygenases 3435 are among an increasing number of molecules implicated in downstream semaphorin signalling. To investigate the role of MICALs in motor neuron cell body positioning, we performed RNAi on its chick orthologues.

We identified three MICALs in the chicken genome, MICAL1 on chromosome 26, MICAL2 on chromosome 5 and MICAL3 on chromosome 1. In contrast to MICAL2 and 3, MICAL1 sequences are not present in a comprehensive chick expressed sequence tag (EST) expression database 36, suggesting it is not highly expressed. In situ hybridisation expression analysis for MICAL3 showed that this gene was distinctly expressed by motor neurons (Figure 8a; a more comprehensive study on chick MICALs is in progress; MV, unpublished results). We constructed shRNA expressing vectors targeting the predicted MICAL3 mRNA and electroporated these in the chick embryo ventral spinal cord. We found that MICAL3 shRNA causes substantial numbers of motor neurons to exit the spinal cord by translocating along their axons in the ventral roots (Figure 8b,c). The magnitude of the effect of MICAL3 shRNA (Figure 8d) is considerably larger than that found with Npn-2 or Plexin-A2 shRNA, approaching the numbers we observed after complete surgical neural crest ablation in chick 18. Ectopic migration following loss of MICAL3 in motor neurons occurs despite the presence of BC cells at the MEP (Figure 8d,e). Together, these results support the conclusion that the cytoskeletal re-organisation that uncouples somal migration from axon elongation in motor neurons involves multiple semaphorin-induced signals that converge on a common cytoplasmic mediator.

The migration of neurons to their final destinations is a key process in the development of the nervous system. This process must be coordinated with the establishment of connectivity and requires differential responses of distinct neuronal compartments to the multitude of guidance cues in the environment. This could be achieved in several ways: first by the sorting of compartment-specific receptors, such as at the axonal initial segment 37; the selective accumulation of kinesin motors in axons versus dendrites 38; by the localised intracellular content of modifiers, such as small cyclic mononucleotides 39; and lastly by the different extracellular environments in which each compartment is located. While this is obvious at the level of the growth cone, where localised signals lead to the re-organisation of the cytoskeleton and growth cone turning, an appealing idea is that the cell bodies of neurons are similarly responsive to cues in their local environment. The logistics of this problem is particularly challenging in the case of long distance projection neurons like somatic motor neurons. Their cell bodies begin to settle in position within the ventral spinal cord at an early phase of differentiation before their axons and dendrites make appropriate connections with their target cells.

In previous work we identified a key role for neural crest derived BC cells in the regulation of this process. In the absence of these cells, spinal motor neuron cell bodies follow their axons at the MEP and migrate into the periphery 18. Here we have begun to identify the molecules involved in the interaction between BC cells positioned at the MEP and motor neurons that prevent somal migration out of the spinal cord. Using loss-of-function studies in chick and mouse, we present evidence for the involvement of members of the semaphorin family in BC cells and their receptors on motor neurons. Loss of Npn-2, a receptor for class 3 semaphorins, in motor neurons leads to ectopic migration of these neurons, both in chick and mouse spinal cord, as does loss of the transmembrane semaphorin Sema6A in BC cells. RNAi studies in chick identified a role for Plexin-A2 in motor neurons, either as a transducing component in a signalling complex with the ligand binding receptor Npn-2, or as a direct receptor for Sema6A. The requirement for semaphorin-plexin signalling in cell body positioning is revealed by the potent effect of loss-of-function of MICAL3, a member of a conserved family of intracellular Plexin-A binding proteins that have been identified as key modulators of semaphorin-plexin signalling 3435.

We started our search for candidate molecules involved in the control of motor neuron migration at the MEP using an RNAi approach that we had developed in the chick, utilising a vector system co-expressing shRNA and EGFP 19. We combined this loss-of-function approach with protocols for targeted expression of transgenes in chick embryos 40 (Additional file 1), which allowed us to selectively knock down expression of candidates in either the ventral neural tube, where motor neurons develop, or in the neural crest, the progenitors of BC cells. Thus, we were able to conclude that Npn-2 and Plexin-A2 are required in motor neurons, whereas Sema6A functions as a BC cell expressed ligand in helping stabilise motor neuron cell body position in the ventral spinal cord.

Since the emergence of RNAi as a powerful experimental tool, several pitfalls associated with this technique have become apparent. First, when using randomly selected small interfering RNA (siRNAs) duplexes or shRNAs, at least half of the sequences will not induce any reduction of their targets and less than 5% will reduce expression by 90% or more 41. Therefore, we selected our RNAi targets using an algorithm for rational siRNA design (SFOLD 42) that predicts target accessibility and applies empirical rules for functional siRNA duplexes 41. Consequently, approximately one in three of our constructs proved effective, in that they significantly reduced expression of their intended targets as determined by an in vitro co-expression assay or by in situ hybridisation (Additional files 2, 5 and 6).

Second, off-target and non-specific effects have been widely reported 434445 and are a serious concern when using RNAi. However, we are confident that the ectopic motor neuron phenotypes we report here are related to the loss-of-function of the intended targets of our constructs, as they are duplicated by constructs targeting the mRNAs on different sites (Npn-2, Plexin-A2 and MICAL3). During this study we tested a total of 38 shRNA hairpin constructs aimed at 10 targets for the induction of ectopic motor neurons after ventral electroporation. Even when considering our collection of shRNAs as random (yet they are all aimed at molecules involved in semaphorin signalling, and all of those are expressed in the ventral spinal cord to some extent), the probability of six hits appearing in three pairs is quite small (p = 0.0111; using an analysis derived from a hypergeometric test). Were we to expand our analysis with shRNAs aimed at truly random targets, the predicted outcome would be that the hit rate would decrease and, concomitantly, the significance of hits appearing in pairs would increase even further. This indicates that general non-specific or random off-target effects are a highly unlikely explanation for the set of data that we have obtained. Finally, the fact that we were able to confirm the roles for Npn-2 and Sema6A in the mouse reinforces the main conclusion from the chick RNAi work that multiple semaphorin-induced pathways are involved in maintaining the stability of motor neuron somal position.

Our loss-of-function studies in chick and mouse implicate two proteins in the control of motor neuron migration at the MEP. Npn-2 is required in motor neurons and Sema6A in BC cells. Sema6A and Npn-2 are not known to interact with each other. Therefore, we propose the involvement of at least two separate pathways mediating this mechanism, although, in contrast to Npn-1 3146, a contribution of Npn-2 to class 6 Sema signalling cannot formally be ruled out. Our more extensive studies in chick identified a role for Plexin-A2 in motor neurons. Consistent with its expression in motor neurons (Figure 1e,k), chick Plexin-A2 could function as a dual receptor for both class 6 and class 3 semaphorins. However, in the mouse we failed to detect significant expression of Plexin-A2 in motor neurons at any stage (results not shown). Considering there are three members of the Plexin-A family in chick, unlike in mouse, where there are four, differences in the individual contributions of the various members of the Plexin-A family to chick or mouse neural development are to be expected. The role of Plexin-A2 as a class 6 semaphorin receptor in chick motor neurons could be adopted by Plexin-A4 in the mouse, as this receptor is more strongly expressed by mouse spinal motor neurons (unpublished observations) and can function as a Sema6A receptor in mouse 32. Plexin-A3, which is expressed by motor neurons in the mouse 47, has already been implicated as the preferred co-receptor for Npn-2 in mouse sympathetic neurons 48. Clearly, our observations warrant further studies in Plexin-A mutant mice 31324749.

So far we have been unable to functionally implicate a class 3 semaphorin in the maintenance of motor neuron position, despite the strong evidence for involvement of Npn-2, and the distinct expression of the Npn-2 ligands Sema3B and Sema3G in BC cells. Since at least two known Npn-2 ligands are expressed by BC cells, a single hit loss-of-function approach may fail as a result of compensatory mechanisms. Experiments aimed at reducing the expression of Sema3B and Sema3G in chick neural crest simultaneously by RNAi have so far not demonstrated a significant increase in the numbers of ectopic motor neurons (not shown). However, considering we found that the ectopic motor neuron phenotype after surgical ablation is effectively rescued by small numbers of grafted BC cells 18, residual levels of repellent molecules at the MEP after partial knockdown are unlikely to produce a strong phenotype. Therefore, a thorough analysis of the role of Sema3B and -3G in motor neuron somal positioning at the MEP will probably require creating double mutants in mouse. Such an analysis will be feasible when Sema-3G mutant mice become available.

It has been proposed, based on their complementary expression patterns, that semaphorins and their receptors could play a role in motor pool sorting in the embryonic mouse ventral spinal cord 14. We cannot presently exclude the possibility that some of our loss of function experiments could directly perturb interactions between neighbouring motor neurons, leading to a possible disruption of pool cohesion and motor neuron positioning. However, this seems unlikely for two reasons. Firstly, no evidence was found in earlier studies of semaphorin and Npn-2 null mutants of any effects on motor pool organisation 13. Secondly, in our chick electroporation experiments where motor neurons were targeted, we rarely saw HB9-positive ectopic cells that were not GFP-positive (less than 1% of the ectopic motor neurons were GFP-negative; for example, see Figure 2a,b,e,f). Moreover, targeted disruptions by electroporation of transgenes in spinal motor neurons engineered to induce pool sorting errors lead to minor cell body displacements within the ventral spinal cord 10 rather than the large scale ectopic migration we found.

An unexpected finding is the restriction of an ectopic motor neuron phenotype to caudal-most regions of Npn-2 and Sema6A null mouse embryos. It seems unlikely that BC cells in the mouse confine motor neurons to the spinal cord only at caudal levels since we have previously described the occurrence of ectopic motor neurons at rostral levels in the Splotch (Pax3) mutant mouse 18, as well as in Sox10 null mouse embryos (Vermeren and Cohen, unpublished). In both mutants, embryos lack BC cells at rostral levels and their ventral nerve roots contain numerous ectopic motor neurons. One possible explanation is that this could reflect level-specific differences in expression of both Npn-2 and Sema6A in spinal motor neurons and BC cells. These differences clearly exist in the mouse 14 (unpublished observations) but not to such an overt extent that could explain the restricted localisation of this phenotype. Another explanation may lie in the temporal gradient of maturation along the rostro-caudal neuraxis, making the more immature hindlimb motor pools more susceptible to perturbations. This is perhaps hinted at by the higher background levels of ectopically positioned motor neurons we saw in caudal-most regions of wild-type mouse embryos (Figures 3c and 6d).

Further evidence for the involvement of semaphorin-plexin signalling in motor neuron somal settling is the potent effect of MICAL3 loss-of-function in the chick. MICALs are a family of conserved cytosolic signalling proteins known to interact directly with Plexin-As and cytoskeletal or cytoskeleton associated molecules (for a review, see 50). Drosophila MICAL was identified in a yeast-two-hybrid screen using D-Plexin-A cytoplasmic domain as bait and the D-MICAL loss of function (LOF) mutant has severe motor axon guidance defects, similar to those found in Sema-1a and PlexA mutants 34. In relation to axon guidance, MICALs have been implicated in Plexin-A signalling only. Furthermore, D-MICAL does not interact with PlexB 34. Thus, the evidence so far suggests that MICALs might be specific mediators of semaphorin-plexin-A signalling. Experiments with a pharmacological inhibitor of flavoprotein monooxygenases, epigallocatechin gallate (EGCG), suggested that axon repulsion by class 3 semaphorins might depend on the oxidoreductase activity of MICALS 35. EGCG did not inhibit the repulsive effects of a number of other ligands, including Sema6A. However, this does not rule out that MICALs are intermediaries in the signalling induced by Sema6A and other repellents, independent of their oxidoreductase activity.

Our loss-of-function studies in the chick ventral spinal cord suggest a ranking of targets, according to the severity of the phenotype, MICAL3 LOF > Plexin-A2 LOF > Npn-2 LOF (quantified in Additional file 8). This ranking is consistent with a signalling hierarchy, where a diverse set of semaphorins (Sema3B, 3G, 6A) interacts with a number of signalling receptors or receptor complexes (Plexin-A2 and Npn-2/Plexin-A). The signals emanating from these receptors then converge on a common intermediate (MICAL3). The effect of MICAL3 LOF approaches the severity of the ectopic motor neuron phenotype observed after surgical neural crest ablation 18. This suggests that, if they exist at all, the contributions of parallel, MICAL3-independent pathways will be necessarily minor. We are currently investigating the role of the two other chick MICAL orthologues, alongside other intermediary molecules known to be involved in semaphorin-plexin signalling pathways, such as CRMPs 51525354.

The saltatory motion of migrating immature neurons is dependent on a microtubule perinuclear cage that drags the nucleus and soma unidirectionally 5556. Several intracellular proteins are known to be involved in this process and their loss leads to pathologies characterised by ectopic migration, such as lissencephalies and epilepsy 4. We propose that the effect of the localised semaphorin-PlexinA2-MICAL3 signal that we have identified is to neutralise or provide an opposing force to the axon-driven perinuclear cage drag (Figure 9).

Little is known about exactly what MICALs interact with except an affinity for the cytoskeleton 57. However, MICALs have clearly been implicated in class3 semaphorin signalling, which is best known for its role in axon guidance and cytoskeletal remodelling following growth cone turning (for reviews, see 2458). This process relies notably on type II myosin 59, the over-activation of which leads in the fly to the ectopic migration of photoreceptor cell bodies from the single layered eye disk along the optic nerve as far as the optic lobe 60. In growth cones, type II myosin generates a force that underlies the retrograde flow of actin. This force is antagonised by dynein 61, part of the dynein/dynactin negative end microtubule motor transport complex. Interestingly, null mutation of p150 glued, a subunit of dynactin, in Drosophila photoreceptors also leads to their ectopic migration into the optic lobe 6263. There are clear parallels between these observations and our results: in both scenarios normally unipolar neurons become bipolar through the perturbation of a signal (or its effectors) that antagonises perinuclear cage drag. Recently, it was shown that knock-down of both cytoplasmic dynein and myosin II stalls glial-guided somal migration of neuronal precursors 64. These findings resonate with our work but leave open the question how these changes are regulated at the cell surface of neurons. We suggest that for motor neurons, semaphorin signalling from BC cells triggers cytoskeletal changes that lead to migration arrest. When cell body migration stops, the stabilisation of neuronal cell body position requires that uncoupling of the perinuclear cage drag be sustained throughout the lifetime of the neuron. Since BC cells are a transient population 65, what constraints prevent motor neuron somata from emigrating at later stages? The most likely explanation is that as motor neurons mature, their polarised structure becomes stabilised by the onset of cadherin-mediated interactions with neighbouring motor neuron somata 10, under the influence of a variety of environmental factors, most importantly innervation from muscle targets, primary sensory afferents and the corticospinal tract.

All animal procedures were performed in accordance with institutional guidelines. Npn-2 66, Sema3B 27 and Sema6A-deficient mice 1767 were bred and genotyped as described.

Targeted electroporation towards the ventral neural tube was essentially performed as described by 40, using an Intracept TSS10 pulser (IntraCell Shepreth Herts, UK) or a Grass-Telefactor SD9 stimulator and 0.25 mm silver electrodes. Plasmid DNA, resuspended in TE buffer at 1 mg/ml with 0.2% Fast Green FCF, was injected in the neural tube of embryonic day 2 (HH stage 10–15) embryos and 3–5 square pulses of 10–12 V were applied with the positive electrode piercing the egg shell and positioned underneath the trunk and the bent, negative electrode above the embryo, parallel to the positive electrode.

For dorsal horn/neural crest targeting, the polarities of the electrodes were reversed and 2–3 square pulses of 7–10 V applied. To limit the size of the targeted area, the positive electrode was positioned perpendicularly to the negative electrode. As a consequence, expression in the dorsal neural tube is restricted to the equivalent of one or two somites, whereas EGFP expressing neural crest derivatives can be observed several somites away from the electroporation site (Additional file 1). The embryos were left to develop for a further two days post-electroporation (HH 22–25) before being fixed in 4% paraformaldehyde (PFA) for antibody staining and analysis. To facilitate the rapid generation of RNAi constructs, co-expressing shRNA and EGFP, we created the pCAβ-EGFPm5-mU6 vector. The ApaI site at position 864 in pCAβ-EGFPm5 1968 was removed by digestion with ApaI and filling in using T4 polymerase. Subsequently, a 376 bp cassette containing the mouse U6 promoter was cut from pBS/U6 (aka pSilencer1.0) 69 using BamHI and inserted into the SalI site at position 1 of the modified pCAβ-EGFPm5 vector, after blunt ending of the BamHI and SalI overhangs with T4 polymerase.

RNAi target sites of 21 nucleotides length were selected on the respective target mRNAs using the SFOLD algorithm 4270 at. The specificity of the selected targets was checked by BLAST. Synthetic complimentary oligonucleotides containing hairpin sequences were designed using the Ambion pSilencer Insert Designer 71, using the default loop sequence of TTCAAGAGA. This oligonucleotide design attaches overhangs matching ApaI (GGCC) to the 3' end and EcoRI (AATT) to the 5' end of the bottom strand. The EcoRI overhang was omitted from the design in order to allow the annealed oligonucleotides to match the blunt end in the vector generated by EcoRV. Oligonucleotides were purchased from MWG Biotech AG (Ebersberg, Germany) or OPERON, (Cologne, Germany). Annealed oligonucleotides were cloned into the unique ApaI and EcoRV sites of the pCAβ-EGFPm5-mU6 vector. Details of the shRNA constructs used in this paper can be found in Additional file 4.

Whole-mount immunostaining of electroporated chick embryos was performed as described before 18. Mouse monoclonal 3A10, anti-Islet-1/2 (39. 5D5), anti-Islet-2 (51. 4H9), and MNR2 (81. 5C10) were developed by Dr Thomas Jessell. These antibodies were obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA, USA. Rabbit anti-neurofilament (NF1991) was purchased from Millipore (Watford, UK); rabbit anti-HB9 from Abcam (Cambridge, UK); rabbit anti-GFP from Invitrogen (Paisley, UK). Mouse anti-V5 was obtained from Sigma-Aldrich (Dorset, UK).

Post-staining, embryos were cleared in a glycerol series, embedded in 20% gelatin/PBS and sectioned on a Leica VT1000S vibratome. Sections (75 μm) were collected on glass slides and mounted with MOWIOL or Vectashield. For quantification of ectopic motor neurons, sections were viewed on a Zeiss Axioskop fluorescence microscope.

The effect of each shRNA-EGFP construct was quantified by counting the number of sections containing ectopic Islet-2/MNR2 positive motor neurons at the EGFP-positive electroporation site (for quantification, see Additional file 8). When the ventral spinal cord was electroporated, the ectopic neurons were EGFP-positive, and when the neural crest was electroporated, peripheral glia and BC cells were EGFP-positive. Since electroporation results in a mosaic embryo, there were instances where the region of the ventral spinal cord that was EGFP-positive contained few labelled motor neurons. As results were expressed as the number of sections containing EGFP-Islet2/MNR2 positive neurons over the overall number of sections containing EGFP, the proportions obtained represent an underestimate of the incidence of ectopic motor neurons. A minimum of six embryos were analyzed for each construct. Representative images were scanned on an Olympus Fluoview confocal laser scanning microscope connected to an Olympus AX70 research microscope. Images were processed using the software supplied by the manufacturer.

For analysis of ectopic motor neurons in mice, embryos were fixed for 2 h in 4% PFA on ice, followed by 2 × 30 minute washes in PBS. Subsequently, embryos were equilibrated in 30% sucrose in PBS, decapitated, the bodies cut in half at mid-trunk level and frozen in Optimal Cutting Temperature (OCT) compound. Serial sections of 20 μm thickness (approximately 20 sections per slide) were cut in the transverse plane on a Bright cryostat. Sections were subsequently dual-labelled with rabbit anti-HB9 and mouse monoclonal 3A10 antibodies and antibody binding visualised by incubating with appropriate fluorescent second antibodies. Slides were coded before the incidence of ectopic motor neurons (that is, HB9 positive neurons in the presumptive white matter or within ventral roots) was quantified by an investigator who was unaware of the code. Images were collected using a Zeiss Axioskop fluorescence microscope connected to a Spot CCD camera.

Digoxigenin-labelled antisense copy RNA was transcribed from linearised cDNA templates and hybridised with 20 μm cryostat sections, as described previously 7273. A cDNA probe for chick Plexin-A1 74 was a gift from Dr Britta Eickholt; chick Npn-1 from Dr Jon Raper; chick Npn-2 from Dr Gera Neufeld; chick Sema3F from Dr Yuji Watanabe; chick Plexin-A2 (ChEST887n12), Plexin-A4 (ChEST202o14), chick Sema3B (ChEST771a21), chick Sema3C (ChEST997l4), chick Sema6A (ChEST667n18), chick MICAL2 (ChEST58n1), and chick MICAL3 (ChEST59i4) were obtained from the BBSRC chick EST database 36 via the MRC Geneservice. Chick Sema3G cDNA (clone pgf1n.pk012.d10) was obtained from the chick EST database from the University of Delaware. Mouse Sema3B cDNA was a gift from Dr Andreas Püschel; mouse Sema6A cDNA from Dr Alain Chedotal. Mouse Egr-2 (Krox20) cDNA was a gift from Dr Piotr Topilko.

Images were viewed using Nomarski optics on a Zeiss Axioskop and images stored on a Spot 2 DCC camera.