SDF1/CXCR4 signalling regulates two distinct processes of precerebellar neuronal migration and its depletion leads to abnormal pontine nuclei formation

To build a functional nervous system depends critically on the migration of newborn neurons from proliferative zones to their final destinations, where they assemble into nuclei, columns and layers, and integrate into neuronal circuits. Tangential migration is essential for the formation of neuronal assemblies that are tangentially displaced from their birthplaces. This is well documented for cortical interneurons(Marín and Rubenstein,2001) and hindbrain precerebellar neurons(Bloch-Gallego et al., 2005). Theoretically, there are multiple paths neurons can take to move from their birthplace to their final settlement. In reality, however, tangential migration takes place along stereotypic pathways, not only well-defined for their anteroposterior (AP) and dorsoventral (DV) directions, but also their depths within the ventriculopial (VP) span of the neuroepithelium(Marín and Rubenstein,2001; Sotelo,2004; Bloch-Gallego et al.,2005). The mechanisms underlying the specification of tangential pathways are only partially understood(Hatten, 2002; Park et al., 2002; Bloch-Gallego et al., 2005). Furthermore, the importance of taking a specific tangential pathway for the subsequent formation of nuclei, layers and columns remains elusive.

The migration of hindbrain precerebellar neurons presents an excellent model to address these questions, because the tangential migratory pathways and their resultant nuclei have been clearly mapped out and are genetically tractable (Altman and Bayer,1997; Wingate and Hatten,1999; Kawauchi et al.,2006; Taniguchi et al.,2006). Precerebellar neurons migrate over long distances, from their germinative zone at the lower rhombic lip (LRL), along distinct tangential pathways to form discrete precerebellar nuclei (see Fig. 1A). The mossy-fibre-projecting precerebellar neurons (PCN) migrate marginally beneath the pial surface of the hindbrain, forming two streams: the anteroventrally directed anterior extramural stream (AES); and the ventrally directed posterior extramural stream (PES). Although the ventral directions of the AES/PES have been attributed to the Netrin and SLIT/ROBO families(Serafini et al., 1996; Yee et al., 1999; Alcantara et al., 2000; Taniguchi et al., 2002; Marillat et al., 2004; Di Meglio et al., 2008), the mechanisms that control the marginal positioning of AES/PES cells, as well as the anterior direction of the AES, are poorly understood.

PCN in the AES and PES migrate abutting the pial meninges(Bourrat and Sotelo, 1990; Altman and Bayer, 1987a; Altman and Bayer, 1987b; Ono and Kawamura, 1990; Kawauchi et al., 2006), which are a source of the potent chemotactic molecule, chemokine stromal cell-derived factor 1 (SDF1; CXCL12 - Mouse Genome Informatics)(McGrath et al., 1999; Tissir et al., 2004). SDF1 and its receptor CXCR4, which were originally identified for their role in leukocyte trafficking (Nagasawa et al.,1996; Tachibana et al.,1998), are employed during the development of the nervous system for regulating the chemotaxis of neurons and axons(Tran and Miller, 2003; Lazarini et al., 2003; Stumm and Hollt, 2007; Li and Ransohoff, 2008). Thus,we hypothesised that the meningeal SDF1 might control PCN migration by confining them to the marginal pathways.

Here, we demonstrate that the meningeal SDF1 and its receptor CXCR4 are critically involved in two distinct aspects of PCN migration: confining the migrating cells to marginal streams, and promoting the anterior migration of AES cells. We also show that disruption of SDF1/CXCR4 signalling results in markedly abnormal formation of the AES-derived pontine nuclei.

The generation of Cxcr4 and Sdf1 knockout, floxed Cxcr4, and Wnt1-Cre mice have been described previously(Tachibana et al., 1998; Nagasawa et al., 1996; Tokoyoda et al., 2004; Danielian et al., 1998). Noon of the day on which a vaginal plug was detected was designated as embryonic day (E) 0.5. For expression studies, timed pregnant ICR mice (Nihon SLC,Shizuoka, Japan) were used. All animal maintenance and manipulations were carried out in accordance with the Guidelines for Animal Experiments at Osaka University.

A plasmid with nucleotide 162-1072 of Cxcr4 mRNA (Accession number D87747) was used to generate riboprobes for in situ hybridisation. The Mbh2 (Barhl1 - Mouse Genome Informatics) plasmid was a gift from Dr Tetsuichiro Saito (Chiba University, Japan).

Expression vectors pCAGGS-Egfp(Niwa et al., 1991; Hatanaka and Murakami, 2002)and pCAGGS-Sdf1 (Tanaka et al.,2009), expressing EGFP and SDF1α respectively, were described before. pCAGGS-Cxcr4-IRES-Egfp was constructed to co-express Cxcr4 and Egfp.

In vitro electroporation of DNA into the LRL of E12.5 mouse hindbrains was performed essentially as previously described(Taniguchi et al., 2002). Whole hindbrains were cultured with or without the meninges in open-book configurations on Millicell culture inserts (0.4 μm pore size,Millicell-CM, Millipore). The culture medium contained DMEM-F12(Sigma-Aldrich), 10% fetal bovine serum (GIBCO, Invitrogen) and 1× N2 supplement (GIBCO, Invitrogen). After 2 days in vitro, the cultured hindbrains were fixed for 2 hours in 4% paraformaldehyde (0.1 M PBS, pH 7.4), and frozen transverse sections (20 μm) were obtained for observation.

In utero electroporation was performed as previously described(Hatanaka et al., 2004) with the exception that 2 μl of expression vectors (between 0.5 mg/ml and 2.5 mg/ml) was injected into the fourth ventricle of E12.5 embryos.

LRL explants (approximately 200-300 μm) were microdissected from the E12.5 LRL. The meninges of either wild-type or Sdf1^-/-hindbrains were divided into pieces of about 400 μm in size. Either recombinant murine SDF1 (0.2 μM, PeproTech EC, London, UK) or PBS as a control was embedded into a collagen block(Zhu et al., 2002) by mixing with 10 μl rat tail collagen. The LRL explant was placed next to a meningeal explant or a collagen block 300-500 μm away and covered with 40μl matrigel (Matrigel Basement Membrane Matrix, BD Biosciences,Massachusetts, USA). The culture medium was the same as for hindbrain organotypic culture. After 2 days, the explants were stained with 0.03%4,6-diamidino-2-phenylindole (DAPI; Nacalai Tesque) to reveal the migrating cells.

In situ hybridisation on whole-mount mouse hindbrains was performed as previously described (Zhu et al.,2002). In situ hybridisation on 20 μm frozen sections was based on that of Cheng et al. (Cheng et al.,1995) with some minor modifications. The sections were fixed with methanol and the H₂O₂ treatment was omitted.

Immunohistochemistry was performed on 20 μm frozen sections, as previously described (Zhu et al.,2006). Primary antibodies used were: goat anti-CXCR4 polyclonal(1:300, Abcam), goat anti-SDF1 (c-19) polyclonal (1:50, Santa Cruz Biotechnology), rabbit anti-PAX6 polyclonal (1:400, Chemicon; or a gift from Dr Noriko Osumi), rabbit anti-laminin polyclonal (1:300, Sigma) and rat anti-GFP monoclonal (1:800, Nacalai Tesque). Secondary antibodies used were:Cy3-conjugated donkey anti-goat IgG (1:500, Jackson ImmunoResearch),Cy3-conjugated goat anti-rabbit IgG (1:250, Jackson ImmunoResearch), Alexa Fluor 488-conjugated donkey anti rabbit IgG (1:400, Invitrogen) and Alexa Fluor 488-conjugated goat anti-rat IgG (1:200, Invitrogen).

Anterograde DiI tracing of motor corticospinal axons was performed essentially as previously described(O'Leary and Terashima, 1988)in postnatal day (P) 3 pups. The injected pups were returned to their mothers until P5 before perfusion with 4% paraformaldehyde. Parasagittal vibratome sections (50 μm) of these hindbrains were obtained and immunostained with a PAX6 antibody under detergent-free conditions to preserve DiI signals.

Fluorescence and brightfield images were captured with a CCD camera(Axiocam, Zeiss) linked to an upright microscope (BX-60, Olympus). Whole hindbrains after in situ hybridisation were imaged with a Multi-Viewer System(Keyence, Osaka, Japan). In some cases, fluorescence images were obtained by a laser-scanning confocal microscope (TCS SP2 AOBS; Leica Microsystems).

MetaMorph (Version 6.1, Universal Imaging Corporation) was used to quantify cell migration from images of DAPI-stained LRL explants. Each LRL explant was divided into four quadrants (see Fig. 3B) intersecting at the centroid of the explant. The explant itself was masked, and cell migration in the quadrant facing towards(proximal) and away from (distal) the co-culture was quantified as the total fluorescence intensity within each quadrant. The ratio of cell migration in the proximal over distal quadrant represented the bias of cell migration towards the co-culture. A ratio above 1 indicates attraction, whereas close to 1 indicates a neutral effect. Statistical significance was evaluated by Mann-Whitney U-test.

The precerebellar neurons destined to form the mossy-fibre-projecting nuclei begin migrating marginally along the PES and AES from E12.5 and E13.5,respectively (Fig. 1A)(Altman and Bayer, 1997; Taber-Pierce, 1966), whereas those destined to form the climbing-fibre-projecting inferior olivary (IO)nucleus migrate submarginally from E10(Taber-Pierce, 1973; De Diego et al., 2002). PES cells migrate circumferentially across the midline to form the contralateral external cuneate nucleus (ECN) and lateral reticular nucleus (LRN) in the medulla oblongata. The AES is a more tortuous route, with an initial ventrally directed migration, then an anteriorly directed path and a final ventral turn to the midline to form the predominantly ipsilateral pontine nuclei (PN) in the pons (Kawauchi et al.,2006; Okada et al.,2007).

PCN both in the PES and AES migrate beneath the pial surface of the hindbrain neuroepithelium, close to the overlying pial meninges(Altman and Bayer 1987a; Altman and Bayer 1987b; Bourrat and Sotelo, 1990; Ono and Kawamura, 1990). This tempted us to hypothesise that the meninges might serve to confine the migrating PCN to the pial surface. To test this hypothesis, we labelled PCN by electroporating EGFP into the LRL at E12.5 and organotypically cultured these hindbrains either with or without their meninges(Fig. 1B). After culturing for 2 days, many GFP-labelled cells emigrated from the LRL, reaching the floor plate both in the presence (Fig. 1C) and absence (data not shown) of the meninges. Inspection of transverse sections of these samples showed that with the meninges, migrating cells formed a highly confined stream beneath the pial surface of the hindbrain, like those in vivo (Fig. 1D, n=8). By contrast, in the absence of the meninges,many PCN migrated submarginally so that the migratory stream appeared dispersed (Fig. 1E, n=5). These results indicate the importance of the meninges in confining the migrating PCN to the marginal route.

The meninges may control the migrating PCN by secreting a chemoattractant that recruits and maintains these neurons close to the pial surface. One candidate is the chemokine SDF1, a chemoattractant expressed mainly in the meninges overlying the entire neural tube during development(McGrath et al., 1999; Tissir et al., 2004). Accumulating evidence indicates the importance of SDF1 in regulating the chemotaxis of migrating neurons and growth cones via its cognate receptor CXCR4 (Tran and Miller, 2003; Lazarini et al., 2003; Stumm and Hollt, 2007; Li and Ransohoff, 2008). We therefore investigated the expression of CXCR4 and SDF1 in the developing precerebellar system.

PCN were identified by their expression of a transcription factor Mbh2, also known as mammalian Barhl1(Bulfone et al., 2000; Li et al., 2004). On E13.5 transverse sections, Cxcr4 in situ hybridisation labelled a marginal stream of cells extending from the LRL to the floor plate(Fig. 2B). The profile of this stream of cells closely resembles the PES labelled by Mbh2 in situ hybridisation on an adjacent section (Fig. 2A,C), suggesting the expression of Cxcr4 in the PES(Fig. 2B,D). Likewise, Cxcr4 mRNA was also found in the AES at E13.5 (compare Fig. 2F with 2E). At E15.5, Cxcr4expression became undetectable in PES neurons undergoing LRN(Fig. 2G,H) and ECN formation(data not shown). By contrast, Cxcr4 was still expressed in migrating AES neurons at E15.5 (Fig. 2I,J), but was undetectable in PN at E17.5(Fig. 2K,L). Double immunohistochemistry with a CXCR4 antibody and an antibody against PAX6, a marker for PCN (Engelkamp et al.,1999), demonstrated that CXCR4 protein was expressed in the PES and AES at E14.5 (Fig. 2M-P and insets). SDF1 immunohistochemistry showed that SDF1 protein was predominantly present in the meninges and the pial surface of the hindbrain, adjacent to the PAX6-positive migratory streams (Fig. 2Q,R). Together these data demonstrate the expression of CXCR4 in the migrating PCN and SDF1 in the overlying meninges, suggesting this receptor/ligand pair might mediate the function of the meninges in controlling PCN migration.

To demonstrate directly that SDF1 expressed in the meninges chemoattracts migrating PCN, we performed co-culture experiments in matrigel(Fig. 3A,B). PCN from LRL explants migrated preferentially towards co-cultured meninges or a collagen block embedded with recombinant SDF1 (Fig. 3C,E,I), whereas they showed no preference towards a control collagen block (Fig. 3F,I). The chemoattraction elicited by the meninges was abolished by 20 μM AMD3100, a selective antagonist for CXCR4 (Donzella et al., 1998; Lazarini et al.,2000) (Fig. 3D,I),or when Sdf1^-/- meninges were used for co-culture (compare Fig. 3H with 3G; Fig. 3I), suggesting that the meningeal attraction is mediated by SDF1/CXCR4 signalling. Thus, the co-culture experiments demonstrate that the meningeal SDF1 can chemoattract migrating PCN in vitro.

To address the in vivo role of SDF1/CXCR4, we analysed the PCN migratory paths in E14.5 Cxcr4^-/- mice. Transverse sections across different axial levels of both wild-type(Fig. 4A, schematic) and Cxcr4^-/- hindbrains were subjected to Mbh2 in situ hybridisation. Anteriorly, where the AES transits from anteriorly to ventrally directed migration in the wild type(Fig. 4Aa, arrow), no Mbh2-positive cells were detected at the corresponding position in Cxcr4^-/- (Fig. 4Ab, asterisk). In posterior AES, newborn PCN emigrated marginally from the LRL in the wild type (Fig. 4Ac, arrow). In Cxcr4^-/-, most Mbh2-positive cells departed from the marginal streams and headed directly towards the ventral midline deep within the hindbrain parenchyma(Fig. 4Ad, arrow). PCN in the PES also derailed from marginal positions in Cxcr4^-/-(compare Fig. 4Af with Ae),although to a lesser extent with derailed neurons migrating at a short distance from the pial surface following the general curvature of the marginal zone (Fig. 4Af, arrow). These defects observed in Cxcr4^-/- (n=2) were replicated in Sdf1^-/- (n=2, data not shown),suggesting that SDF1/CXCR4 is the main ligand/receptor pair that mediates the marginal migration of PCN. The phenotypes at E14.5 are consistent with our in vitro data and demonstrate a crucial in vivo role of SDF1/CXCR4 signalling in guiding PCN to migrate marginally.

We then asked whether precerebellar nuclei could form properly in the absence of SDF1/CXCR4 signalling by Mbh2 in situ hybridisation on transverse sections across different axial levels(Fig. 4B, schematic) at E16.5. At the pontine flexure, we observed a prominent and symmetrical PN in the wild type (Fig. 4Ba, arrow). By contrast, much smaller and bilaterally asymmetrical PN were present on corresponding sections from Cxcr4^-/- mice(Fig. 4Bb, arrow, n=2), consistent with a previous study(Vilz et al., 2005). Interestingly, posterior to PN, where no Mbh2-positive cells were present around the midline in the wild type(Fig. 4Bc), we found a large cluster of Mbh2-positive cells spanning the midline in Cxcr4^-/- (Fig. 4Bd, arrow, n=2). Mbh2-positive cells could be traced from the LRL to this ectopic cluster(Fig. 4Bd, arrowheads) raising the possibility that the cluster derived from the derailed PCN. At posterior levels, the PES-derived LRN and ECN appeared comparable in size and location between the wild type and Cxcr4^-/- (compare Fig. 4Be,g with Bf,h).

The marked reduction of PN at their normal position concurrent with the emergence of posterior ectopic Mbh2-positive clusters raised the possibility that many pontine neurons are posteriorised in the absence of SDF1/CXCR4 signalling. The migratory pattern of AES at E14.5 in Cxcr4^-/- also supported this possibility(Fig. 4A). We sought direct evidence for pontine identity of the posterior ectopic Mbh2-positive cells. We found that Nuclear factor I (NFI) family members Nfix and Nfib are expressed at a high level in the PN but weakly in the LRN and ECN from E16.5 (data not shown) (see Fig. S1C,D in the supplementary material). In situ hybridisation on adjacent sections of an E16.5 Cxcr4^-/- hindbrain with Mbh2, Nfix and Nfib showed that the ectopic Mbh2-positive cells were strongly positive for Nfix and Nfib signals (see Fig. S1A,B in the supplementary material), suggesting that these cells were indeed pontine neurons.

Therefore, whereas formation of the LRN and ECN appears normal, PN formation is largely posteriorised in animals depleted of SDF1/CXCR4 signalling.

To better appreciate the spatial distribution of derailed pontine neurons in Cxcr4^-/-, we performed Mbh2 in situ hybridisation on whole-mount E16.5 hindbrains from wild type, Cxcr4^+/- and Cxcr4^-/- embryos. In wild type (n=2) and heterozygotes (n=6), Mbh2 labelled the bilaterally symmetrical PN anteriorly at the level of the trigeminal ganglion (gV) (Fig. 5A,B) and the LRN (Fig. 5A,B, asterisks)and ECN (data not shown) posteriorly. In Cxcr4^-/-, the LRN and ECN appeared normal in both position and size(Fig. 5C,D, asterisks; data not shown), consistent with our data on transverse sections(Fig. 4B). By contrast, the disruption in PN revealed on the whole-mount was dramatic(Fig. 5C,D; n=6). PN in their normal position (at or above the gV) were very small(Fig. 5C,D, notched arrowheads); instead multiple ectopic pontine clusters were found around the midline posterior to the gV. Closer inspections on whole-mount and transverse sections (Fig. 5E) enabled us to discern two types of ectopic pontine clusters: type I, defined by a single large cluster located deeply and spanning the midline (appears out of focus in Fig. 5C,D, arrowheads; Fig. 5E, arrowheads),corresponding to the cluster shown in Fig. 4Bd; and type II, defined by multiple superficial clusters situated adjacent to the midline (Fig. 5C-E, arrows). Most type II clusters were anterior to the type I cluster and appeared to derive from distinct marginal migratory streams, some of which were still visible (Fig. 5C,D). Another feature of type II clusters was their left-right asymmetrical distribution, the pattern of which varied among individual samples (Fig. 5C,D; n=6). In summary, whole-mount in situ hybridisation provided a three-dimensional perspective of the mutant phenotype that could not be easily appreciated on sections. Two types of defects were observed in pontine neurons(Fig. 5F): (1) those migrating deeply forming a deep cluster posteriorly; and (2) those migrating marginally,but departed from the anterior migratory path prematurely, ending up in multiple posterior locations.

We have shown so far that depletion of SDF1/CXCR4 signalling results in derailment of PCN from their marginal migration as well as posteriorised pontine migration. Although their expression pattern and the co-culture data imply that SDF1/CXCR4 signalling may function cell-autonomously, we cannot rule out the possibility that these mutant phenotypes are secondary to other defects in the hindbrain cytoarchitecture or a possible earlier defect in AP patterning. Indeed, CXCR4 is expressed in the ventricular zone and several other hindbrain nuclei (Fig. 2B,F; data not shown) (McGrath et al., 1999; Tissir et al,2004; Lieberam et al.,2005).

To address this question, we asked if the defects in PCN migration could be rescued by expressing Cxcr4 in Cxcr4^-/- PCN progenitors in the LRL. Either pCAGGS-Cxcr4-IRES-Egfp expressing full-length Cxcr4 or pCAGGS-Egfp as a control was electroporated into the LRL of wild-type and Cxcr4^-/-embryos at E12.5 by in utero electroporation. Egfp electroporation in the wild type labelled the PES and AES unilaterally(Fig. 6A,E)(Kawauchi et al., 2006). The GFP-positive PES was confined marginally abutting the laminin-positive meninges, as shown at E14.5 on a transverse section(Fig. 6A; n=5). The labelling of AES and PN, better appreciated on E16.5 whole-mount hindbrains(n=12), showed the characteristic anterior migratory path laterally(Fig. 6E, arrow) and its ventral turn just below the rootlet of the gV(Fig. 6E, asterisk). Forced expression of Cxcr4 in the wild-type LRL did not cause notable changes in migratory profiles of the PES(Fig. 6B) or AES(Fig. 6F). Electroporation of Egfp alone into the LRL of Cxcr4^-/- embryos labelled the aberrant PES and AES similar to those revealed by Mbh2in situ hybridisation: broadened PES (Fig. 6C; n=2), many AES cells migrating deeply anterior to the PES (see Fig. S2 in the supplementary material), and the posteriorised type I and II clusters (Fig. 6G,arrowhead and open arrowhead, respectively; n=5). Of particular note,the characteristic anterior path of the AES was largely missing in the mutant. Inspection on transverse sections from these samples confirmed that many GFP-positive cells were located within the type I(Fig. 6I; n=2) and type II ectopic clusters (data not shown), which were double positive for PAX6(Fig. 6I, inset).

We then tested if these phenotypes could be rescued by expressing Cxcr4 in the Cxcr4^-/- LRL. Cxcr4expression restored a PES confined to the marginal position at E14.5(n=2; compare Fig. 6Dwith 6C), and resulted in few labelled cells migrating deeply anterior to the PES (data not shown, n=2). The anterior migration of pontine neurons was also restored(n=6), as evidenced by the re-emergence of the characteristic anterior migratory path (Fig. 6H, arrow) and prominent PN in the normal anterior position. This occurred concomitantly with few labelled cells in type I and II ectopic pontine clusters (compare Fig. 6H with 6G; a transverse section is shown in Fig. 6J), suggesting that expression of CXCR4 directed the strayed cells to their normal anterior-migratory pathway. Taken together, these results suggest that CXCR4 functions cell-autonomously in controlling the marginal migration of PCN and anterior migration of pontine neurons.

What are the consequences of ectopic PN on the later event of axon innervations? PN are the main hindbrain targets of the corticospinal tract(CST). Corticopontine innervation is formed by interstitial collateral branching from the trunk of CST axons(O'Leary and Terashima, 1988). Although collagen co-culture experiments implied that cues from PN control the formation of collateral branching (Heffner et al., 1990), direct in vivo evidence supporting this possibility has been missing. The ectopic pontine clusters resulted from depleted Sdf1/Cxcr4 signalling provided us with an opportunity to seek the missing evidence.

The mature pattern of motor corticopontine collaterals emerges at around P2 in mice (Bastmeyer and O'Leary,1996), but neonates of Cxcr4^-/- and Sdf1^-/- die at birth(Tachibana et al., 1998; Nagasawa et al., 1996). To circumvent this obstacle, we conditionally knocked out Cxcr4 in Wnt1-expressing LRL using a Wnt1-Cre driver and a Cxcr4^fl/fl responder line. The Wnt1-Cre driver was used because PCN precursors were shown to locate within the Wnt1-expressing domain of LRL(Rodriguez and Dymecki, 2000). Wnt1-Cre;Cxcr4^fl/fl mice survived into adulthood. First we confirmed that PN formation was indeed disrupted in Wnt1-Cre;Cxcr4^fl/fl mice similarly to Cxcr4^-/- (data not shown) and ectopic pontine clusters persisted until at least P5 (see Fig. S3 in the supplementary material). The CST was then traced anterogradely by DiI from the motor cortex between P3 and P5 (Fig. 7A). In control Cxcr4^fl/fl hindbrains, labelled motor CST showed a stereotypic pattern of collateral branches directed exclusively into PN(Fig. 7B-D; n=3 brains). Interestingly, the motor CST in Wnt1-Cre;Cxcr4^fl/fl hindbrains showed extensive collateral branching at multiple ectopic sites between the pontine flexure and the pyramidal decussation (Fig. 7E; n=4 brains). Notably, each ectopic cluster of collateral branches extended towards a PAX6-positive ectopic pontine cluster(Fig. 7F-H). The formation of ectopic branching could not be due to changes in corticospinal neurons in the motor cortex because Wnt1 expression was excluded from the developing forebrain (Parr et al., 1993). Although we could not formally exclude the possibility that the ectopic branching could be caused by disruption of hindbrain nuclei other than PN, the high correlation between the sites of ectopic collateral formation and ectopic pontine clusters suggest that pontine neurons themselves control the collateral branch formation from the corticospinal tract.

In the developing brain, a directed tangential migratory pathway is specified for AP and DV directions, as well as depth within the VP span of the neuroepithelium. By focusing on the migrating precerebellar neurons, we have demonstrated that the chemokine SDF1 in the pial meninges and its receptor CXCR4 in PCN are crucial for controlling two distinct processes of PCN migration: (1) positioning the migratory streams to the appropriate VP level;and (2) regulating the anterior migration of pontine neurons.

Evidence suggests that SDF1/CXCR4 signalling controls the VP position of migrating PCN in both the PES and AES, most likely via chemoattraction by the meningeal SDF1: (1) PCN emigrating from the LRL were attracted by the meningeal SDF1 in vitro; (2) the highest SDF1 protein concentration was found in the pial meninges and the pial surface of the hindbrain(Fig. 2), suitably located to attract PCN to migrate along the marginal path. Thus, the scenario presented here resonates with recent findings in the developing cortex, where the meningeal SDF1 confines the tangential migration of hem-derived Cajal-Retzius cells and a subset of cortical interneurons to the marginal zone via chemoattraction (Borrel and Marín, 2006; Li et al., 2008; Lopez-Bendito et al., 2008; Tanaka et al., 2009). As SDF1 is expressed in the pial meninges surrounding the entire central nervous system (CNS) (McGrath et al.,1999; Tissir et al.,2004), these findings imply that chemoattraction by the meningeal SDF1 could be a general mechanism for controlling pathfinding events taking place in the CNS marginal zone.

The second role of SDF1/CXCR4 signalling appears to be regulating the anterior migration of pontine neurons. Two types of defective anterior migration were observed in Cxcr4 knockout mice. The first comprises pontine neurons that derail from the marginal stream and migrate towards the midline deep within the neuroepithelium. Two lines of evidence indicate that these neurons may have completely failed to turn anteriorly. First, their leading processes appear to orient circumferentially towards the midline in Egfp-electroporated Cxcr4^-/- samples (see Fig. S2 in the supplementary material). Second, the deep ectopic cluster they formed was located in rhombomeres (r) 6 and 7 (Y.Z. and F.M., unpublished), the axial levels where the pontine progenitors were shown to originate by genetic fate mapping (Farago et al., 2006). The second type of defect comprises pontine neurons that appear to migrate marginally. Many of them are likely to have migrated for some distances anteriorly, as most superficial type II clusters they formed were located more anterior than the deep type I cluster. Nevertheless, they failed to complete the full course of the anterior pathway, turning towards the ventral midline prematurely in a rather unpredictable manner. As both defects were rescued by expressing Cxcr4 in the Cxcr4^-/- LRL, it is unlikely that they were secondary to disruptions in other hindbrain architectures.

What is the mechanism(s) for the anterior migration of pontine neurons? We think the two anterior migration defects reflect the requirement of SDF1 in two distinct aspects.

The first type of defect in which cells went straight to the ventral midline without apparent anterior turning may be secondary to their failure in migrating marginally. This is because most pontine neurons that migrated marginally travelled anteriorly for some distances, as discussed above. If the guidance cue(s) instructing the anterior migration is spatially confined to the pial surface, the pontine neurons derailed from the marginal stream would miss the cue(s) and escape its influence.

The second type of defect may reflect a direct involvement of SDF1/CXCR4 signalling in the anterior migration of pontine neurons. A most straightforward explanation would be that an anterior-high posterior-low SDF1 gradient instructs the anterior migration of pontine neurons by way of chemoattraction. Indeed, we detected such a graded distribution of SDF1 protein underlying the anterior migrating pontine neurons (see Fig. S4 in the supplementary material). An instructive role of SDF1 is further supported by the gain-of-function experiment that showed that ectopic expression of SDF1 in the LRL prevented pontine neuron precursors from leaving the LRL (see Fig. S5 in the supplementary material). However, the fact that some marginally migrating pontine neurons can move anteriorly for some distances suggests the existence of other instructive cues. Alternative to, or in addition to, the instructive role of SDF1, SDF1/CXCR4 signalling may serve to modulate the responsiveness of pontine neurons to other anterior guidance cue(s), as has been demonstrated in developing retinal ganglion cell axons and sensory axons(Chalasani et al., 2003; Chalasani et al., 2007).

SLIT2 expressed by the facial motor nucleus appears to regulate the anterior migration of pontine neurons by preventing the ROBO1/ROBO2-expressing pontine neurons from leaving their anterior pathways prematurely(Geisen et al., 2008). It is unlikely that SDF1/CXCR4 signalling promotes the anterior migration of pontine neurons chiefly via modulating the SLIT-ROBO signalling pathway, because the anterior migration defects reported here in Cxcr4 or Sdf1knockout mice were more severe than those in mice depleted of SLIT-ROBO signalling: in the former, a majority of pontine neurons failed to reach their normal anterior positions, whereas in the latter only a minority failed.

Precerebellar nuclei formation takes place as a consequence of proper migration of PCN. The derailment from marginal migratory streams in Cxcr4 knockout mice appears to be manifested differently in the PES-and AES-derived nuclei. Whereas the derailed cells in the PES can form the LRN and ECN in their normal positions, those in the AES fail to migrate anteriorly and form a deep ectopic pontine cluster at a posterior position. This difference could be due to two reasons. First, the displacement from the marginal position in the PES is smaller than that in the AES, meaning that the derailed PES cells may be able to encounter instructive cues for nuclei formation. The smaller derailment in PES might be due to the presence of additional guidance cues for PES neurons in the marginal/submarginal region in the caudal hindbrain. Second, whereas the PES is a straightforward ventral migration, the AES is a tortuous trajectory comprising ventral-anterior-ventral switches. Therefore, the consequence of failing to migrate marginally earlier on was amplified by their failure to encounter the anterior migratory cues.

Those pontine neurons that migrated marginally but fell short of anterior migration formed multiple ectopic clusters that exhibited left-right asymmetry. The fact that the pattern of this asymmetry varies among individuals raises the possibility that it might arise as a result of intrinsic and environmental noises that the developing pontine neurons experience during their production and migration. The fluctuations in phenotypes owing to noises might be normally buffered and equalised between the left and right by an intact and robust anterior guidance mechanism, but become unmasked phenotypically when the robustness of that mechanism is weakened, as might be the case in the Cxcr4 knockout. Interestingly,similar left-right variations were reported in a recent study in which anterior migration of pontine neurons was affected in Hoxa2-, Hoxb2-, Phox2b-, Robo1-Robo2- and Slit1-Slit2-deficient mice(Geisen et al., 2008). These observations taken together suggest that several molecular mechanisms may contribute additively to the optimal robustness of the anterior migration of pontine neurons.

The presence of multiple ectopic pontine clusters in Cxcr4knockout mice tempted us to test in vivo whether the pontine neurons themselves trigger the formation of collateral branches from the CST, a possibility that was only inferred from in vitro evidence(Heffner et al., 1990). By using a conditional knockout strategy to circumvent lethality at birth, we could obtain postnatal pups that had multiple ectopic pontine clusters. The motor CST in these samples extended few collateral branches at the pontine flexure, as PN were largely missing at their normal positions. Instead, we observed multiple ectopic clusters of collateral branches posteriorly along the CST, each of them correlated with an ectopic pontine cluster, suggesting a causal relationship between these two events. As the ectopic pontine clusters exist before the growth of motor CST, it is reasonable to conclude that the ectopic pontine neurons trigger the ectopic collateral formation.

It should be noted that not every ectopic pontine cluster was accompanied by a collateral cluster. This could be because our focal DiI injection labelled only a small fraction of motor CST axons. However, an alternative is that only subpopulations of pontine neurons can induce collaterals from the motor CST. In support of this alternative is the mature pattern of motor corticopontine projection in wild-type mice, in which collaterals arise at a specific rostral and a caudal site from CST segments overlying PN. The motor CST axons, however, seem to be capable of extending collateral branches from diverse rostrocaudal positions within the hindbrain in response to the induction of ectopic pontine clusters, suggesting that these axons themselves do not have an intrinsic mechanism that governs the sites of collateral extension. These observations, taken together, provide the first in vivo evidence that the formation of collateral branches from CST is controlled by cues derived from pontine neurons.

Supplementary material available online at http://dev.biologists.org/cgi/content/full/136/11/1919/DC1