Mutants in Sema6A were generated in a gene trap screen and have been previously described 1853. Macroscopic examination of the hindbrains of adult Sema6A mutants (n = 9) in comparison with wild-type (n = 3) and Sema6A heterozygous mice (n = 12) revealed dramatic defects in the size and trajectory of the CST. This tract was visible in wild-type and Sema6A^+/- mice as a highly myelinated tract running at the ventral surface of the medulla (Figure 2a). In contrast, in many Sema6A mutants (four out of nine) the CST was not visible as the medullar pyramids macroscopically (Figure 2b). In other mutants (five out of nine), the pyramids remained visible macroscopically, that is, they appeared as a tight bundle at the surface, but their size was reduced to various degrees between animals as well as between pyramids on either side of individuals (Figure 2c).

We used staining for the reporter gene placental alkaline phosphatase (PLAP), which is encoded on the gene trap cassette inserted into the Sema6A locus 18, to examine the CST of adult Sema6A^+/- and Sema6A^-/- mice at a microscopic level (Figure 2d–l). In adults, Sema6A, and thus PLAP, is strongly expressed in oligodendrocytes. Therefore, staining for PLAP labels all myelinated fibres in adults with one or both mutant alleles. The stainings confirmed a wide range in severity of CST size reduction and showed a reasonably smooth distribution in severity across the mutants analysed, including those that had no macroscopically visible CST. In addition, PLAP stainings revealed defects of the remaining CST axons. Instead of running at the surface of the medulla as a tightly fasciculated bundle as seen in Sema6A^+/- mice (Figure 2d, g, j), they were often split into several bundles that ran either normally at the surface (Figure 2f) or aberrantly deeper within the medulla (that is, at the rostrocaudal level of the inferior olive, they are embedded within the olive; Figure 2e, arrowhead). More strikingly, at the caudal olive a variable proportion of the remaining CST axons projected away from the midline towards the ventrolateral spinal cord (Figure 2i; but also detectable macroscopically when the CST remained visible at the surface, Figure 2c, arrowheads). Such axons failed to decussate and instead proceeded down the spinal cord in a ventrolateral position (Figure 2l, asterisk). The small portion of CST axons that remained close to the midline decussated correctly and entered the dorsal funiculus of the spinal cord (Figure 2h, k).

The hypoplasia of the CST might be the result of a guidance defect at the pyramidal decussation with many axons dying because they do not find their targets, while others manage to contact their targets even after misguidance and via an alternative route. Another possibility is that hypoplasia of the tract at the level of the medulla is due to fewer axons reaching that point in the first place; that is, that there is an additional choice point along the CST trajectory at which axons are misguided and fail to proceed down the spinal cord. To test these ideas we traced the CST during its development, that is, at postnatal day (P)4, when the majority of CST axons have just arrived in the spinal cord.

Placement of small crystals of the lipophilic dye DiI into the motor cortex of P4 post-mortem brains effectively labelled the descending CST (Figure 3). In Sema6A^+/- mice (n = 4; Figure 3, left column), traced fibres could be followed through the internal capsule and cerebral peduncle, revealing the normal trajectory of these fibres through the brain. At the MHB, traced fibres extending longitudinally on the ventrolateral surface of the midbrain turn medially and travel halfway towards the midline, then turn ventrally over the pontine nucleus to finally reach the ventral surface of the medulla to form the pyramids.

In Sema6A mutants (n = 5) we did not observe major trajectory defects of traced axons rostral to the MHB and the gross amount of traced fibres immediately rostral to the MHB was comparable to that seen in Sem6A heterozygous mice (data not shown; Figure 3a, b). However, at the MHB of Sema6A mutant mice (Figure 3, middle and right columns) we observed traced fibres taking several aberrant routes. At the caudal midbrain, most axons initially made a correct medial turn but some turned dorsally instead (three of five; Figure 3b, e, f, blue arrowheads). Of the axons that turned medially, a large fraction failed to turn ventrally over the pontine nuclei and instead projected all the way to the midline (five of five) and even crossed it (four of five), below the interpeduncular nucleus (four of five), or, as a minor fraction, turned dorsally (three of five; Figure 3b, yellow or green arrowheads, respectively) or ventrally (two of five). In the ventral projection over the pontine nuclei the tract appeared split into a couple of bundles. Some axons ran correctly close to the midline but others could be found to split from this main bundle to take an aberrant route within the pons or towards lateral areas (three of five; Figure 3e, f, h, i, pink arrowheads).

Further caudally, in Sema6A^+/- mice, traced fibres travelled through the medulla close to the midline (Figure 3j). This could also be observed for the reduced amount (to varying degrees) of axons that were routed properly at the MHB of Sema6A mutants (Figure 3k, l). However, at the level of the inferior olive, which appeared to form normally in Sema6A mutants, some (two of four) or all (two of four) of these axons projected away from the midline and were some distance away from the spot where CST axons of wild-type mice would turn dorsally and cross the midline (Figure 3n, o). These axons proceeded ipsilaterally at a ventrolateral position.

Sema6A is known to mediate some effects through interactions with PlxnA2 4252 and/or PlxnA4 4045. Some members of this class of transmembrane semaphorins or their orthologues in invertebrates have also been shown to be capable of reverse signalling, presumably as receptors 4748495051. To assess whether these plexin proteins might be involved in mediating these functions of Sema6A and/or whether Sema6A might be acting cell-autonomously in CST axons, we examined the expression patterns of these genes in motor cortex and in the regions of the midbrain and hindbrain where defects are observed.

We used reporter gene staining, RNA in situ hybridization and immunohistochemistry to examine Sema6A expression, which we found to be highly dynamic. At embryonic day (E)17.5, CST axons have crossed the mid-hindbrain border and reached the pons. At this stage, CST axons stain positively for PLAP in Sema6A gene trap embryos (Figure 4a, a'). This is far prior to the onset of myelination, demonstrating that neurons that give rise to the CST normally express Sema6A in late embryogenesis. To examine the distribution of the Sema6A protein itself, we performed immunohistochemistry on sections from E17.5 and P1 brains, using a commercially available polyclonal antibody against murine Sema6A (Figure 4b–q), which we proved to be specific for Sema6A by comparing stainings on wild-type and Sema6A mutant brain sections (Additional file 1). At E17.5, we observed Sema6A-immunopositive corticofugal fibres at the level of the internal capsule and cerebral peduncle (Figure 4b), and positive CST axons at the level of the MHB and pons (Figure 2c). Sema6A was also expressed in the MHB region by the pontine nuclei (Figure 4c–h). By postnatal stage P1, localisation of Sema6A on CST axons was no longer evident, either at the MHB, where expression in the pons persists (Figure 4i–k), or in the medulla, where the inferior olive was strongly Sema6A-immunopositive (Figure 4l–q).

We used in situ RNA hybridization to examine expression of Sema6A, PlxnA2 and PlxnA4 in the neocortex at postnatal stage P4. This revealed laminar and areal-specific staining for each of the three genes (Figure 5). The pattern of layer-specificity is complementary between Sema6A and the Plxn genes in some areas (for example, primary somatosensory cortex (S1)). The laminar patterns also change in an areal-specific manner, especially between S1 and primary motor cortex (M1). Within the primary motor cortex all three genes are expressed to some extent in layer V, the source of CST axons. Sema6A mRNA at these stages is expressed highly in upper layer V, PlxnA2 mRNA has the highest expression throughout layer V, and PlxnA4 mRNA is expressed moderately, also throughout the layer. Corticospinal projection neurons are found predominantly in deep layer V 5455. Consistent with this, and with the lack of antibody staining at this age, staining for PLAP activity in Sema6A heterozygotes does not strongly label CST axons at postnatal stages (data not shown). As PlxnA1 and PlxnA3 have also been indirectly implicated in Sema6A functions 4356, though not shown to bind to it directly, we also examined the expression patterns of the genes encoding them in the motor cortex. While both genes are expressed in the cortex at this age, neither is particularly strongly expressed by deep layer V neurons in M1 (Additional file 2).

At the level of the MHB, RNA in situ hybridisation confirmed expression of Sema6A by the pontine nuclei and also by the pontine reticular nuclei (Figure 6). The CST runs between these structures at this point. Just caudally, Sema6A is also expressed by the superior olivary nuclei. PlxnA2 and PlxnA4 are also expressed by the pontine nuclei, in distinct patterns, and by other structures in this region (Figure 6). Just rostral to the level of the pyramidal decussation, Sema6A and PlxnA4 are strongly expressed in the inferior olives at P4, in a partly complementary fashion (Figures 6 and 7). PlxnA2 is weakly expressed in this structure at this age, in a pattern that is very similar to that of PlxnA4. At this point, CST axons are anatomically constrained near the midline by the inferior olives.

As the expression data suggested the possible involvement of the Plxn genes, we investigated further whether similar CST guidance defects can be found in mutants for PlxnA2 and PlxnA4.

Immunohistochemistry with antibodies against the cell adhesion molecule L1 stains CST fibres selectively at P10. Using this method, we analysed the CST projections in PlxnA2 (n = 3), PlxnA4 (n = 4), and PlxnA2;PlxnA4 double (n = 2) mutants and compared them to Sema6A mutants (n = 2) or normal control mice (Sema6A^+/-, n = 1; wild-type, n = 2; PlxnA2^+/-, n = 1; PlxnA4^+/-, n = 1) (Figure 8).

At the MHB of Sema6A^-/- mice at P10, the number of CST axons was already moderately to severely reduced. Only a few misguided axons could be found, indicating that the majority of such axons had already died between P4 and P10. At the decussation level, in some cases virtually no axons were visible. When present at this level, CST axons could be seen either forming a ventrolateral projection, decussating properly or split into two bundles to do both.

PlxnA2^-/- mice showed no apparent guidance defects or reduction in size of the CST. In PlxnA4^-/- mice, the size of the CST appeared normal from the MHB, where no misguided axons could be found, up to the level of the caudal medulla. Here, the CST split into two bundles that either formed a ventral or ventrolateral projection or decussated normally to join the contralateral dorsal funiculus. This defect, which is fully penetrant, thus strongly resembles that seen in Sema6A mutants.

Double mutants of PlxnA2 and PlxnA4 show very reduced viability and are thus extremely difficult to obtain. We have analysed two double mutants at P10, using anti-L1-immunohistochemistry. These showed no evidence of the defects observed at the level of the MHB and pons in Sema6A mutants (Figure 8). In particular, there was no apparent reduction in the number of fibres passing the pons. The defect in the caudal medulla was essentially the same as that observed in PlxnA4 single mutants. These data suggest that PlxnA2 is not involved in CST guidance in these regions and does not compensate for the lack of PlxnA4 at the MHB.

To investigate the possibility that misguided axons were not detected at the MHB at P10 in PlxnA4-deficient mice because they might have already died, we traced the CST in these mutants at P4 (Figure 9). The tracing studies confirmed the CST defect at the pyramidal decussation (two of five animals showed no decussating axons, but only ventral projection; three of five mutants showed splitting into decussating and ventrally projecting axons; Figure 9) as described for the anti-L1-staining at P10. We did not find any traced fibres that were misguided at the MHB in PlxnA4 mutants (n = 5 animals, versus PlxnA4^+/-, n = 5, and wild-type, n = 2; data not shown).

The phenotypes observed in Sema6A mutants identify the MHB as a major choice point, or series of choice points, for CST axons. In fact, CST axons must perform rather complicated manoeuvres to successfully navigate this region and reach the hindbrain. First, exiting the cerebral peduncle at the caudal end of the midbrain, the CST turns sharply from a longitudinal direction towards the midline. Before reaching the midline, it must subsequently turn from this transverse direction ventrally to course over the pons. In parallel, the CST must partially defasciculate to allow collateral branching into the pons while retaining directional projection to reach and refasciculate at the ventral surface of the medulla and form the pyramids. Defects in each of these events can be observed in Sema6A mutants, thus identifying Sema6A as the first guidance molecule controlling CST trajectory at the MHB. Mutations in Ctip2, which encodes a transcription factor expressed in CST neurons, also result in defects at the MHB, where the tract is disorganised and defasciculated, with many axons failing to turn ventrally at the pons and no axons extending past the pons by P0 7. While these phenotypes reinforce the notion of this region as a complicated point in the trajectory of CST axons, they are somewhat distinct from those observed in Sema6A mutants.

We see a number of possible scenarios to explain the phenotypes observed. First, Sema6A could act directly in this region in canonical fashion as a repulsive ligand to constrain CST projections to an appropriate channel. This fits generally with the observation that Sema6A expression in this region seems to surround the CST. However, some of the phenotypes, such as midline crossing, are not obviously consistent with this model, given the absence of prominent midline expression of Sema6A. The apparent lack of involvement of the known Plxn receptors might also argue against it. Second, Sema6A could act as an attractive or permissive cue at this level. CST axons project in apposition to a number of pontine nuclei that are strongly Sema6A-positive. Changing responsiveness to guidance cues at different points of an axon's trajectory is a common mechanism 57, which has been observed directly with semaphorins 5859606162. The importance of Sema6A expression in pontine nuclei is, however, called into question by analyses of mutants in Netrin-1 or its receptor, deleted in colorectal cancer (DCC). In these mice, the pontine nuclei and pontine reticular nuclei fail to condense due to defects in cell migration 63. While these mice show defects in CST guidance in the hindbrain (see below) they do not show any defects in CST guidance at the MHB. These findings suggest that the phenotype at the MHB in Sema6A mutants may instead be due to a lack of Sema6A expression on CST axons themselves during late embryogenesis. Sema6A could act in CST axons as a receptor for some unknown ligand in surrounding cells or, acting either as a ligand or a receptor (or both), could mediate axon-axon interactions that are important for proper guidance at this level. The orthologous Drosophila gene Sema-1a appears to function in such a fashion in motor axon guidance by setting an appropriate level of fasciculation between motor axons, allowing them to respond to guidance cues at specific choice points 64.

Our analyses at early postnatal stages indicate that the defects at this point are the most likely cause of the CST hypoplasia observed in adult Sema6A mutants. This is based on several observations: first, in none of the traced mutants did we see a gross difference in the number of CST axons arriving to the MHB at early postnatal stages. Although smaller defects within the forebrain might have been left undetected, we consider this possibility unlikely to be able to account for the significant hypoplasia observed. Second, many CST axons take widely divergent routes at this point and fail to enter the hindbrain. The failure to connect with appropriate targets is likely to result in the degeneration of these axons, which are not detected at later stages, and consequently the death of layer V cells themselves. Indeed, we do not observe any obvious derailed PLAP-stained CST fibres in the MHB region of adult Sema6A mutant brains (data not shown). Third, PlxnA4 mutant mice show similar defects as Sema6A mutants in the caudal medulla, but no defects at the MHB and no hypoplasia of the CST. The later defect within the caudal medulla, which is shared between Sema6A and PlxnA4 mutants, is thus unlikely to contribute to the observed hypoplasia in Sema6A mutants. In fact, many axons that adopt a ventrolateral route within the hindbrain and spinal cord in Sema6A and PlxnA4 mutants persist into adulthood, and have in Sema6A mutants been independently shown to contact appropriate targets in the spinal cord 56.

Of the axons that make it to the pyramidal decussation point in Sema6A mutants, some decussate normally and join the dorsal funiculus, while many take up an abnormal ventrolateral position and continue to project caudally on the ipislateral side. The relative proportions of axons that show this phenotype is somewhat variable, in both Sema6A and PlxnA4 mutants. Whether this reflects stochastic variation or some molecular distinction between subtypes of axons is not clear. The CST is thought to be pioneered by a small number of axons, with the majority of axons following these, possibly responding more to fasciculation cues on the surface of these axons than to guidance cues in the environment. Small variations in the number of pioneer axons adopting either of the alternative routes might thus be passively transmitted to the follower axons 65. While the phenotype is most apparent at the decussation itself, in fact the CST axons begin to spread laterally before this point, that is, at the level of the inferior olive. Sema6A is no longer detected on CST axons at this stage but is strongly expressed in the inferior olives, while PlxnA4 is expressed by the CST neurons. This suggests that Sema6A may normally act as a repulsive cue in the olives to constrain CST axons near the midline (Figure 10). Failure to turn dorsally and cross the midline at the normal decussation point may thus be indirect and simply reflect the fact that these axons are no longer situated in a position to respond to the normal medial cues that promote decussation.

A very similar phenotype has been reported in a number of mutants affecting Netrin signalling 20. In Unc5c (formerly Unc5h3) mutants, a reduced decussation is observed, with a substantial proportion of axons projecting ventrolaterally. This is preceded in the hindbrain by a lateral broadening of the descending CST, rostral to the level of the decussation. The CST eventually splits into two distinct domains, one close to the midline, which decussates, and one more lateral, which continues to project ipsilaterally. Mutants in netrin-1 and a viable mutant of DCC show similar phenotypes. These were interpreted as evidence for a direct role of Netrin-1, which is expressed at the midline in the region of the decussation, as a guidance factor for CST axons at this choice point 20. However, the involvement of Unc5c, which normally mediates repulsion from sources of Netrin-1, complicates this interpretation. Our results suggest an alternative model, which is that the decussation defect in CST guidance is secondary to the profound defect in the migration of inferior olivary neurons observed in these mutants 66. This would be consistent with the lateral broadening of the CST prior to the decussation point in these mutants. Our data suggest that the CST is normally actively hemmed in by the inferior olive, molecularly mediated by Sema6A (Figure 10). If this structure is not formed properly, or if it lacks Sema6A expression, then CST axons will spread laterally and will not decussate properly. (Notably, the olive forms normally in Sema6A mutants.) Interestingly, defects in the formation of the inferior olivary nuclei have also been noted in Joubert syndrome, where pyramidal decussation of the CST also fails to occur 2729.

A similar phenotype is observed in NCAM mutants, where a proportion of CST axons adopts a ventrolateral position, while the rest project dorsally and either cross the midline normally or join the ipsilateral dorsal funiculus 19. In this case, the laterally misrouted axons are detectable at early postnatal stages but do not persist in adults. The cellular basis of this phenotype is not entirely clear; it could involve a disruption of the homophilic fasciculation of follower axons on pioneers or the involvement of NCAM in mediating responses to other guidance cues 67. In addition, the integrity of the olive in NCAM mutants has not been reported.

Sema6A has two known binding partners, PlxnA2 and PlxnA4, which have been shown to act as receptors for Sema6A or as antagonists to Sema6A in different contexts 40424552. Our expression analyses show that both plexins are expressed in deep layer V of primary motor cortex, the origin of CST axons. Sema6A is expressed on CST axons at early stages but is not detected at perinatal stages, when CST axons are navigating through the hindbrain. At these stages it is expressed more selectively in upper layer V of M1. Analyses of PlxnA2 and PlxnA4 mutants suggest that PlxnA4 is the important player in the guidance of CST axons, while PlxnA2 has no essential function. PlxnA4 mutants show a qualitatively identical phenotype to Sema6A mutants at the level of the pyramidal decussation, although more axons reach that point in PlxnA4 mutants as they do not show any defects at the MHB. An abolished interaction of PlxnA4 and Sema6A is thus sufficient to explain the defect at the decussation choice point. The expression patterns of these molecules favour a classic mode of action in this context; that is, Sema6A acts as a guidance ligand for the receptor PlxnA4. Interestingly, an overlapping role for PlxnA3 in this process has also been demonstrated 56. These authors observed the presence of ventrolaterally projecting CST axons at low penetrance in either PlxnA3 or PlxnA4 single mutants but at very high penetrance in double mutants. However, PlxnA3 has not been shown to bind Sema6A directly.

Plexins are also well known co-receptors for secreted semaphorins, which have been indirectly implicated in CST axon guidance. As mentioned above, ablation of L1 in mice results in a failure of CST axons to decussate at the caudal medulla and, after correct contralateral or aberrant ipsilateral entry into the dorsal funiculus, to extend down the spinal cord beyond cervical levels 12. In cortical slice cultures, wild-type but not L1-deficient cortical axons are repelled by Sema3A secreted from the ventral spinal cord and this repulsion requires formation of a co-receptor of L1 and neuropillin-1 (in cis) 6168. In line with this idea, mice expressing L1 with a deleted homophilic binding site that leaves binding to neuropilin-1 intact show a normal CST 69. However, direct involvement of Sema3A in CST guidance has not been demonstrated. In fact, recent analyses of Sema3A mutant mice did not reveal any CST guidance defects 70, indicating the involvement of other molecules in L1-mediated CST guidance at the pyramidal decussation. A lack of defects in neuropilin-1 mutant mice 56 also suggests that transmembrane, rather than secreted semaphorins, may be the major players in vivo in CST guidance and that the phenotypes of Plxn mutants at this choice point reflect their role in direct signalling from transmembrane semaphorins.

The fact that neither the PlxnA2 nor the PlxnA4 single mutants have a CST defect at the MHB suggests either that Sema6A interacts with some other molecule at this choice point or that PlxnA4 and PlxnA2 compensate for each other. Analyses of double mutants of PlxnA2 and PlxnA4 did not reveal any additional defects at the MHB, however. PlxnA3;PlxnA4 double mutants also do not show any defect at the MHB 56. Though we can not rule out the possible involvement of PlxnA1 at this point 43, or of additional redundancy between plexins, it seems most likely that the function of Sema6A at this choice point, whether due to expression on the CST axons and/or in the surrounding region, is either homophilic or involves an as yet unknown interacting partner.

Sema6A mutant mice were described previously 1853. Genotyping of Sema6A mice was carried out as described 44. PlxnA2 and PlxnA4 mutant mice were described in 45 and 40, respectively. Staining for expression of the reporter gene PLAP was carried out on 100 μm vibratome sections of adult brains as described 18.

To identify guidance defect of CST axons during development, the tract was traced with the lipophilic fluorescent dye 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate (DiI). After sacrificing mice at P4, the skullcap was removed, leaving the brain in the base of the skull, and three DiI crystals were placed unilaterally into the motor cortex (Figure 3c'). The brains were left in phosphate-buffered saline for 3 h and subsequently immersion fixed and kept in phosphate-buffered saline containing 4% paraformaldehyde. After 5 to 7 months, the brains were removed and frontal vibratome sections, 50 μm in thickness, were prepared.

To visualize the CST at P10, immunostaining was carried out on 40 μm vibratome sections using monoclonal rat anti-L1 (1:200; Chemicon, now Millipore, Billerica, MA, USA) or mouse anti L1 (1:500; Abcam, Cambridge, UK) antibodies as described previously 13. The blocking solution contained 1% bovine serum albumin, 5% normal serum, and 0.5% Triton-100. Primary antibody was visualized with Cy2- or Cy3-conjugated anti-rat or -mouse IgG (Jackson ImmunoResearch, West Grove, PA, USA). For protein expression analysis at E17.5 and P1, we used goat anti mouse Sema6A (1:100; R&D Systems, Abingdon, UK) together with mouse anti-neurofilament-M (clone 2H3, 1:50; DSHB, University of Iowa, Iowa City, USA) as a reference staining of fibres using the same protocol except for an initial 30 minute incubation of sections in 20 mM sodium citrate, pH8.5 at 80°C for antigen retrieval.

In situ hybridization was performed on free-floating 50 μm vibratome sections of P1 and P4 brains, as previously described 72. The following digoxigenin-labeled RNA probes were used: Sema6A (kind gift of W Snider), PlxnA2 and PlxnA4 (from the lab of M Tessier-Lavigne).