Control of the migratory pathway of facial branchiomotor neurones

During the formation of the vertebrate central nervous system (CNS), many cell populations migrate over long distances and follow complex trajectories before reaching their final destination. These migrations usually involve multiple changes in the direction of the movement and/or in the substrate on which the cells progress. The corresponding transitions may depend on changes in the environment itself, on responses elicited in migrating cells by the environment or on programmed, cell-autonomous mechanisms. In the past few years, many studies have demonstrated the existence of environmental cues that can act as contact or diffusible attractants or repellents to provide directional information to migrating cells through interactions with cell-surface receptors (Serafini et al., 1996; Ackerman et al., 1997; Robinson et al., 1997; Birchmeier and Gherardi, 1998; Artigiani et al., 1999; Bloch-Gallego et al., 1999; Yee et al., 1999; Alcantara et al., 2000; Brose and Tessier-Lavigne, 2000). However, the changes elicited in the migrating cell by the environment to adapt to a novel context, i.e. modifications in the sensitivity to a guiding cue or in the expression of an adhesion molecule, are only beginning to be investigated.

Facial branchiomotor (fbm) neurones undergo a complex and stereotyped migration in the mouse embryonic hindbrain (Auclair et al., 1996; Studer et al., 1996; McKay et al., 1997; Schneider-Maunoury et al., 1997; see Fig. 1A). This migration takes place between E10 and E14 and its initiation occurs while the hindbrain is transiently segmented along the anterior-posterior (AP) axis into metameric units called rhombomeres (r) (Lumsden and Keynes, 1989; Lumsden, 1990; Schneider-Maunoury et al., 1998). Fbm neurones are born in the basal plate of r4 between E9 and E11, and extend their axons dorsally towards the exit point of the VIIth nerve (Altman and Bayer, 1982; Ashwell and Watson, 1983; Auclair et al., 1996; Goddard et al., 1996; Studer et al., 1996; O. Voiculescu, S. Schneider-Maunoury and P. Charnay, unpublished data). From E10, their cell bodies start migrating tangentially along the ventral midline, reaching first r5 and then r6. In r6, these neurones begin a dorsal migration between the neuroepithelium and the mantle. Subsequently, approximately at the level of the alar/basal plate boundary, fbm neurones adopt a radial pathway towards the pial surface, where they finally settle to form the facial motor nucleus. The first fbm neurones reaching this final destination can be detected at E12. At this stage, some fbm neurones have still not left the territory derived from r4, which we will still call r4, although rhombomere boundaries cannot be morphologically identified any more (Marin and Puelles, 1995; Wingate and Lumsden, 1996; Mathis et al., 1999). These late fbm neurones follow the same migratory pathway as the early ones and the facial motor nucleus migration is completed at around E14. Thus all fbm neurones execute a complex sequence of movements involving tangential progression through several rhombomeres followed by dorsal and then radial migrations. This migration pattern is atypical, since most motor neurones in the vertebrate hindbrain undergo a ventral to dorsal migration within a single rhombomere (Altman and Bayer, 1982; Lumsden, 1990). Furthermore, the caudal migration of fbm neurones is not observed in all vertebrate species. In the chick, these neurones perform their lateral and radial migrations within r4 (Lumsden and Keynes, 1989; Lumsden, 1990; Szekely and Matesz, 1993).

The analysis of several loss-of-function mutations have provided some information on the molecular mechanisms controlling fbm neuronal specification and migration. Hoxb1 expression in the otic/preotic hindbrain is restricted to r4 (Murphy et al., 1989) and it has been shown to be involved in specifying the identity of this rhombomere (Studer et al., 1996; Bell et al., 1999). In Hoxb1^–/– embryos, fbm neurones do not execute their normal caudal migration and directly progress laterally within r4, a behaviour similar to that of trigeminal branchiomotor neurones in r2 (Goddard et al., 1996; Studer et al., 1996). In addition, the inactivation of Gata2, Gata3, and Phox2b (Pmx2b – Mouse Genome Informatics), which are expressed in progenitors and precursors of fbm neurones, has been shown to impair fbm differentiation and to prevent their caudal migration out of r4 (Nardelli et al., 1999; Pata et al., 1999; Pattyn et al., 2000). Recent studies suggest that Hoxb1 controls Gata2, which in turn regulates Gata3 expression (Nardelli et al., 1999; Pata et al., 1999). These data suggest that the caudal migration of fbm neurones requires their correct AP specification and differentiation, which involves the Hoxb1, Gata2, Gata3 cascade. In contrast, in embryos lacking normal r5, such as homozygous mutants for the genes Krox20 (Egr2 – Mouse Genome Informatics; Schneider-Maunoury et al., 1993; Swiatek and Gridley, 1993) or kreisler (Mafb – Mouse Genome Informatics; Frohman et al., 1993; McKay et al., 1994), caudal migration appears to occur, suggesting that the presence of r5 is not required for its initiation (McKay et al., 1997; Schneider-Maunoury et al., 1997; Manzanares et al., 1999). Taken together, these data suggest that the initiation of caudal migration of fbm neurones is mainly dependent on intrinsic properties of these cells or on local cues in r4. However, very little is known about the regulatory mechanisms that control their further progression through r5 and into r6.

To address the latter issue, we have searched for molecular markers specifically expressed during the different phases of fbm neuronal migration. We show that fbm neurones regulate the expression of three genes, TAG-1 (Cntn2 – Mouse Genome Informatics), Ret and cadherin 8 (Cdh8), during the different phases of their migration in r4, r5 and r6. We then studied the expression of these genes and the migration of fbm neurones in the context of the kreisler, Krox20 and Ebf1 mutations. kreisler and Krox20 loss-of-function mutations lead to elimination of r5 and thus affect the environment through which cells migrate. We show that in both mutants, fbm neurones migrate dorsally into anteriorly positioned r6 and express a combination of genes characteristic of an r6 location. Ebf1 encodes an HLH transcription factor (Hagman et al., 1991, 1993; Wang and Reed, 1993) expressed in differentiating fbm neurones in r4 and during the entire migratory process (Garel et al., 1997; Pattyn et al., 2000). The Ebf1 loss-of-function mutation (Lin and Grosschedl, 1995) results in a subpopulation of fbm neurones adopting an r6-like pattern of migration within r5. Furthermore, these cells express a combination of markers normally characteristic of an r6 location. Altogether, these data show that the combination of r6-specific environmental cues and of cell-autonomous mechanisms involving Ebf1 elicit specific gene responses in fbm neurones and the subsequent selection of appropriate migratory pathway.

Krox20/lacZ heterozygous mice (Schneider-Maunoury et al., 1993) were maintained in a mixed C57Bl6/DBA2 background and crossed to produce homozygous embryos. PCR genotyping of embryo yolk sac was performed as previously described (Schneider-Maunoury et al., 1993). kreisler mice (kindly provided by Dr Julian Lewis) were maintained in 129 background. Homozygous or heterozygous males were crossed with heterozygous females to produce homozygous embryos. These were identified by examination of the position of the inner ear or by PCR genotyping (Frohman et al., 1993). Ebf1 heterozygous mice (Lin and Grosschedl, 1995) were maintained in C57Bl6 background and crossed to produce homozygous embryos. PCR genotyping was performed as described previously (Garel et al., 1999). For staging of embryos, midday of the vaginal plug was considered as embryonic day 0.5 (E0.5).

Embryos were fixed overnight in 4% paraformaldehyde (PFA) at 4°C. Whole-mount immunohistochemistry and in situ hybridisation on dissected hindbrains were performed as previously described (Schneider-Maunoury et al., 1997). Hindbrains were subsequently embedded in paraffin and sectioned (8-12 μm), or flat-mounted (Schneider-Maunoury et al., 1997). The following probes were used for in situ hybridisation: Cdh8 (Korematsu and Redies, 1997); Ebf1, Ebf2 and Ebf3 (Garel et al., 1997); EphA4 (Gilardi-Hebenstreit et al., 1992); Hoxb1 (Wilkinson et al., 1989); islet 1 (Isl1; Nardelli et al., 1999); Phox2b (Pattyn et al., 1997); Ret (Pachnis et al., 1993); TAG-1 (a kind gift of D. Karagogeos). For flat mounts of embryos between E9 and E12.5, the ventricular surface was positioned upwards except for the Ret probe at E12.5. For E14.5 embryos, the pial surface was positioned upwards. DiI (Molecular Probes) injections in the root of the VIIth nerve were performed as previously described (Schneider-Maunoury et al., 1997; Helmbacher et al., 1998). Injected hindbrains were kept in the dark at room temperature for 3 to 7 days in 4% PFA to allow diffusion of the dye. They were then placed in a diaminobenzidine (DAB) solution (1.5 mg/ml of DAB, Tris pH 8.2 0.1 M) and illuminated on a fluorescence microscope for 30 minutes to 2 hours to obtain photoconversion of the DiI fluorescent signal, as previously described (Sandell and Masland, 1988; Marcus et al., 1995). After rinsing in PBS, the hindbrains were processed for whole-mount in situ hybridisation as above, cryoprotected in 30% sucrose and sectioned (20-40 μm) using a cryostat.

To perform a detailed study of the successive phases of migration of fbm neurones (Fig. 1A), we first searched for specific and informative markers. To identify fbm neurones unambiguously, we performed retrograde labelling experiments by injecting DiI into the root of the VIIth nerve at E11.5. This procedure labels fbm neurones migrating from r4 towards r6, and visceromotor (vm) neurones born in r5 and migrating dorsally within this rhombomere (Fig. 1A,B) (Auclair et al., 1996; Studer et al., 1996; McKay et al., 1997; Schneider-Maunoury et al., 1997). Photoconversion of the injected DiI combined with in situ hybridisation analyses allowed the characterisation of the patterns of expression of several genes in migrating fbm neurones (Fig. 1B-G and data not shown).

We confirmed that Phox2b, which encodes a homeodomain transcription factor, is expressed in progenitors and mature fbm neurones in r4 as well as during their migration through r5 and into r6 (Pattyn et al., 1997, 2000; Fig. 2A,B). Its expression was maintained in the forming facial motor nucleus at E14.5 (Pattyn et al., 1997; data not shown). In contrast to this expression pattern covering the entire migratory process, several genes were expressed by fbm neurones only during specific phases of the migration. TAG-1, which encodes a cell adhesion molecule of the immunoglobulin family (Dodd et al., 1988), was expressed in migrating fbm neurones in r4, r5 and r6 at E11.5 (Figs 1C,D, 2C). However, its expression was specifically downregulated in neurones migrating dorsally within r6 at E12.5 (compare Fig. 2C,D with 2A,B). Ret, which encodes a GDNF receptor subunit (Durbec et al., 1996a; Trupp et al., 1996), was expressed in r4 in a subset of ventral cells but not in fbm neurones at E11.5 (Fig. 2E and data not shown). However, Ret RNA was detected in fbm neurones located more caudally, in r5 and r6 (Fig. 2E). Double in situ hybridisation with an Hoxb1 probe, which specifically labels r4 (Murphy et al., 1989), indicated that fbm neurones switch on Ret precisely when they exit r4 to enter r5 (compare Fig. 2E and 2G). At E12.5, Ret expression was maintained in fbm neurones migrating in r5 and r6 where they form the facial motor nucleus (Fig. 2F). Finally, Cdh8, which encodes an adhesion molecule of the cadherin family (Tanihara et al., 1994) was first expressed at E12 in fbm neurones migrating away from the midline towards the dorsal region in r6 (Korematsu and Redies, 1997; data not shown). This expression was maintained from E12.5 to E14.5 in fbm neurones finishing their migration in r6 and forming the facial motor nucleus (Fig. 2H and data not shown). In conclusion, fbm neurones display dynamic patterns of gene expression during their migration. They switch on Ret when they exit r4 and enter r5, downregulate TAG-1 when they migrate dorsally in r6 and turn on cadherin-8 during the final phase of their migration within r6 (see Fig. 8). The conservation of this precise position-specific pattern of expression between E11.5 and E12.5 indicates that early and late migrating neurones display the same expression profile during their migration. Furthermore, the sharp transitions observed in TAG-1, Ret and Cdh8 expression suggest that fbm neurones may regulate the expression of these genes according to their position in the hindbrain, possibly in response to the rhombomeric environment. To test this hypothesis, we investigated how fbm neurones behave when this environment is affected.

To analyse the influence of the environment on the migration of fbm neurones, we studied mouse mutants that modify the territories through or into which fbm neurones progress but presumably do not directly affect the migrating cells themselves. The kreisler (kr) mutation leads to the disappearance of r5 and, although the interpretation of the phenotype has been controversial (Frohman et al., 1993; Cordes and Barsh, 1994; McKay et al., 1994, 1997), recent results indicate that an r6-like (r6*) territory is maintained in kr/kr embryos (Manzanares et al., 1999). Krox20^–/– embryos also lack r5, but have no apparent defects in r6 (Schneider-Maunoury et al., 1993; Swiatek and Gridley, 1993; Schneider-Maunoury et al., 1997; Seitanidou et al., 1997). It has been shown that in kr/kr embryos, vm facial neurones are not present and fbm neurones migrate dorsally within r6* (McKay et al., 1997; Manzanares et al., 1999). Using the Phox2b marker, we confirmed that fbm neurones migrated dorsally after exiting r4 (Fig. 3A,A’). However fbm neurones formed abnormal thick streams during this dorsal migration (Fig. 3A,A’), which may be due to the abnormality of r6*. Analysis of Ret expression at E14.5 showed that early dorsal migration resulted in anterior displacement of the facial motor nucleus (Fig. 3E,E’).

As in kreisler mutants, in Krox20^–/– embryos fbm neurones migrated dorsally within the r6 territory, immediately after exiting r4, as demonstrated by DiI injections into the root of the VIIth nerve at E12.5 (Fig. 4A,A’). This early dorsal migration led also to an anteriorly positioned facial motor nucleus (data not shown). In both Krox20^–/– and kr/kr embryos, Ret was activated in fbm neurones as soon as they exited r4 and migrated into r6/r6* (Figs 3C,C’, 4C,C’), TAG-1 was downregulated during the dorsal migration (Figs 3B,B’, 4B,B’) and Cdh8 was expressed in the final phase of the migration (Figs 3D,D’, 4D,D’).

In conclusion, although the two mutants lack r5, fbm neurones are able to migrate dorsally in the r6 territory after exiting r4 and express a combination of genes characteristic of r6. These results indicate that environmental cues in r6, rather than an internal clock, control the dorsal migration of fbm neurones. Furthermore, they show that these cells regulate their pattern of gene expression in response to the changing environment.

As indicated (see Introduction), Ebf1 is expressed in fbm neurones in r4 and during their entire migratory process (Fig. 1E and data not shown; Garel et al., 1997; Pattyn et al., 2000). In contrast, the two other members of the mouse Ebf family, Ebf2 and Ebf3, are expressed in the progenitors of fbm neurones, but not in mature neurones or during their migration (Fig. 1F,G and data not shown; Garel et al., 1997; Pattyn et al., 2000). In a previous study, we have shown that in the embryonic striatum, where Ebf1 is the only family member expressed, it controls early neuronal differentiation (Garel et al., 1999). These data led us to analyse Ebf1^–/– embryos (Lin and Grosschedl, 1995), to investigate whether Ebf1 controls aspects of fbm neurone migration. We first performed retrograde labelling experiments involving DiI injection into the root of the VIIth nerve. At E11.5, both bm and vm populations were normally positioned in Ebf1^–/– embryos (data not shown), suggesting that the migration of fbm neurones was not affected at this stage. In contrast, at E12.5, while the position of vm neurones was still normal, the migration of fbm neurones appeared severely affected (compare Fig. 5A and 5A’). We observed an accumulation of fbm neurones in the ventral region of r4 and an ectopic stream of cells progressing dorsally within r5 (Fig. 5A’), which was more clearly visible upon DiI injection into the facial nerve branch, which only labelled fbm projections (Auclair et al., 1996) (Fig. 5B,B’). Therefore the Ebf1 mutation affects the migration of part of the fbm population, leading to abnormal dorsal migration within r5.

To study the consequences of this migratory defect, the morphology of the post-migratory facial motor nucleus was examined. At E14.5, in wild-type embryos, the facial motor nucleus forms a round structure, located caudally to the superior salivatory nucleus formed by the vm neurones that have migrated dorsally into r5 (Fig. 1A; Auclair et al., 1996; Studer et al., 1996; McKay et al., 1997; Schneider-Maunoury et al., 1997). Both nuclei are Ret positive at this stage (Fig. 5C). Ret in situ hybridisation in Ebf1^–/– embryos indicated that the facial motor nucleus presented an elongated shape, with a large abnormal anterior expansion in the r5 derived territory, overlapping with the superior salivatory nucleus (Fig. 5C’). Parasagittal sections processed for Nissl staining or Islet 1/2 immunohistochemistry, which labels motoneurones (Ericson et al.. 1992), confirmed this observation (Fig. 5D-E’). Moreover, Islet 1/2 immunohistochemistry performed earlier at E13.5 indicated that the elongated shape of the facial nucleus was due to the existence of two streams of fbm neurones, one located at the expected position and another one more rostrally (Fig. 5F’). Together these data indicate that in Ebf1^–/– embryos the facial motor nucleus expands anteriorly, owing to the abnormal dorsal migration within r5 of a significant proportion of fbm neurones.

In the hindbrain, the expression of Ebf1 is not restricted to migrating fbm neurones and the gene is also transiently expressed in numerous populations of differentiating cells, including motoneurones and commissural neurones (Fig. 1E; Garel et al., 1997; Wang et al., 1997). Since in Ebf1^–/– embryos a subpopulation of fbm neurones behave in r5 like wild-type fbm neurones in r6, we investigated whether this mutation would affect hindbrain patterning in the r4-r5 region or fbm cell differentiation. Expression of Hoxb1, a specific marker of r4 (Murphy et al., 1989), and of Epha4, which is specifically expressed in r2, r3 and r5 (Gilardi-Hebenstreit et al., 1992), were not modified in Ebf1 null embryos (Fig. 6A-B’), suggesting that general hindbrain patterning was not affected.

To investigate whether Ebf1 plays a role in fbm neuronal differentiation, we examined the expression of Islet1 (Isl1), a general motoneurone marker (Ericson et al., 1992), and Phox2b, which labels differentiating fbm neurones (Pattyn et al., 1997, 2000). From E10.75 to E11.5, Isl1 and Phox2b were not affected in Ebf1^–/– embryos (Fig. 6C-D’). In addition, the expression of Ebf2 and Ebf3 was also not modified in Ebf1^–/– embryos (data not shown). In conclusion, these data indicate that the Ebf1 mutation does not grossly affect hindbrain patterning and early neuronal differentiation of fbm neurones.

To test whether the r5 fbm dorsal migration in the Ebf1 mutant was related to abnormal gene expression, we examined the transcription of Phox2b, TAG-1, Ret and Cdh8. At E11.5, when fbm migration was not yet affected, Ret expression was normally induced in cells migrating in r5 (data not shown), while the level of TAG-1 expression was slightly reduced in these cells, as compared with the wild type (Fig. 7A,A’). At E12.5, examination of Phox2b expression confirmed the existence of migration defects: two streams were observed progressing dorsally in r5 and r6. The reinforcement of the labelling in r4 suggested a possible delayed migration of fbm cells from this rhombomere (Fig. 7B,B’). At this stage, drastic anomalies in TAG-1, Ret and Cdh8 expression were also observed in fbm neurones. TAG-1 was prematurely downregulated in cells that exited r4 (Fig. 7C,C’). Ret was normally expressed in cells migrating in r5 and r6, but it was prematurely activated in some fbm neurones in r4 (Fig. 7D,D’). Finally, Cdh8 was found to be expressed in cells located in ventral r4 and r5, in addition to cells migrating dorsally in r5 and r6 (Fig. 7E,E’).

In conclusion, in Ebf1^–/– embryos, a subpopulation of fbm neurones adopt an r5- or r6-specific expression pattern, while still located in r4 or r5 respectively. There is nevertheless a perfect correlation between gene expression pattern and migratory behaviour, since these cells perform an r6-like dorsal migration in r5 (Fig. 8). Furthermore, the molecular defects are likely to precede the modification in migration, since their first manifestations are observed in r4. Taken together, these data suggest that some fbm neurones in Ebf1^–/–embryos are unable to interpret their local environment correctly and raise the possibility that the inappropriate expression of molecules like TAG-1, Ret and cadherin 8 may play a role in their abnormal migration.

In this paper, we have investigated the migration of facial branchiomotor neurones and followed the expression of several migration markers in different mouse mutants. Our work demonstrates that these cells adapt their pattern of gene expression according to their local environment and supports the idea that these elicited responses of the migrating cells are required for the selection of the appropriate migratory pathways.

It has been proposed that cells undergoing complex migrations might respond to successive environmental cues by modifying some of their intrinsic properties, including their gene expression pattern, to follow the appropriate pathway. However, such an adaptation by migrating cells has not been investigated in detail in the vertebrate CNS so far. The migration of fbm neurones constitutes an attractive model to study this issue, since these cells progress through successive rhombomeres and adopt different trajectories within these well-defined territories. In the present study, we have shown that fbm cells express specific combinations of markers such as Ret, TAG-1 and Cdh8 during each phase of their migration (Fig. 8). Analysis of segmentation mutants lacking r5 (Krox20, kreisler) indicated that fbm neurones migrate dorsally after exiting r4 and express a combination of genes characteristic of r6 (Fig. 8). These data indicate that r6-specific cues control the dorsal migration and that no conditioning or change in gene expression that occurs in r5 is essential for further migration. Furthermore, they demonstrate that the r6-specific gene expression pattern is dictated by the location of migrating cells and not simply activated according to a prefixed chronological program. Overall, we show that fbm neurones adapt to their changing environment and respond to it by regulating their gene expression pattern.

In Ebf1^–/– embryos, a subpopulation of fbm neurones prematurely activate Ret expression in r4, which occurs normally only in r5 in wild-type embryos (Fig. 8). Similarly, these cells downregulate TAG-1 and activate Cdh8 in r5, while these subsequent modifications in gene expression normally occur only after the cells have reached r6 (Fig. 8). These data indicate that in Ebf1 null embryos there is a shift between the pattern of gene expression of fbm neurones and their rhombomeric location as compared with wild-type embryos. In contrast, since this subpopulation of fbm neurones migrate dorsally in r5 as if they were in r6, the association between the pattern of gene expression of fbm neurones and their migratory behaviour is maintained. Interestingly, the shift between location and gene expression is first detected in r4, before selection of an abnormal migratory pathway in r5. This suggests that modifications in gene expression in fbm cells reflects their interpretation (erroneous in Ebf1^–/– embryos) of the local environment and influences their subsequent migratory pathway.

Since Ebf1 expression is not restricted to fbm neurones in the hindbrain, the misinterpretation of their rhombomeric environment by fbm neurones in Ebf1 null embryos raises the question of the cell autonomy of the phenotype: are fbm neurones affected in their capacity to appropriately evaluate their environment or is the environment modified, providing them with misleading cues? Although we cannot discard the second hypothesis, we favour the first one for the following reasons: (1) no modification in expression of regional markers indicative of a change of identity in r4 or r5 was observed; (2) we did not detect any other migratory defect in the hindbrain among vm, somatomotor (including the abducens nucleus) or other branchiomotor neurone populations, suggesting that only fbm neurones are affected; (3) the existence of an inappropriate response of fbm neurones is reminiscent of another phenotype associated with the Ebf1 mutation in the embryonic striatum, where differentiating neurones do not downregulate the expression of early genes, leading to a discrepancy between the location of the cells and their pattern of gene expression (Garel et al., 1999).

The precise function of Ebf1 is still the object of speculations. The r6-like migratory behaviour of fbm neurones in r5 in Ebf1^–/– embryos suggest that similar cues involved in guiding the neurones are present in r5 and r6. In wild-type embryos, these cues may not be recognised by migrating fbm neurones because these cells do not express the appropriate combination of genes. We propose that a function of Ebf1 is to prevent the establishment of the r6-specific pattern of expression in fbm cells until a signal provided in r6 releases this block. Therefore, in Ebf1^–/– embryos, fbm cells express the r6 pattern prematurely, while still in r4 or r5, and engage into a lateral migration in r5. Since only a part of fbm cells migrate laterally in r5, the Ebf1-controlled blocking mechanism is likely to be partially redundant with another, unknown system. An alternative hypothesis is that Ebf1 plays a differentiation. Although we cannot exclude this possibility, the absence of modification of Phox2b, Isl1, Ebf2 and Ebf3 expression suggests that at least the initial steps of fbm cell differentiation are not affected.

The correlation between the patterns of expression of TAG-1, Ret and Cdh8 in fbm neurones and their migratory behaviour is conserved in the mutant backgrounds that we have studied. This raises the possibility that the corresponding gene products may actually participate in establishing pathways of migration appropriate for each rhombomeric environment. Indeed, TAG-1 and Cdh8 encode cell-surface molecules that have been implicated in axonal outgrowth and cell adhesion (Furley et al., 1990; Redies and Takeichi, 1996). Ret, which encodes a GDNF-receptor subunit, is expressed by several populations of migrating cells (Pachnis et al., 1993; Durbec et al., 1996b) and may also be involved in migratory processes (Tang et al., 1998; Natarajan et al., 1999). During their migration in r5 and r6, fbm neurones are likely to interact successively with different substrates: they may first follow longitudinal fibre bundles present on both sides of the floor plate in r5, then progress through the mantle in r6 and finally interact with radial glia in r6. Such changes of migratory substrates are likely to require appropriate modifications in the adhesion properties of fbm neurones. TAG-1 might be involved in the tangential caudal migration of fbm cells within r5. It is expressed at the right location and it has been shown to interact with another cell adhesion molecule, L1, which is present on the longitudinal fibres mentioned above (Dodd et al., 1988; Kuhn et al., 1991) and has recently been shown to modulate the repulsion/attraction response to Sema3A, a specific guidance cue (Castellani et al., 2000). According to its site of expression, Cdh8 might be involved in the radial more general role in the control of neuronal differentiation of fbm neurones and that its mutation leads to premature migration within r6. Finally, in Ebf1 mutants, the abnormal lateral migration of fbm neurones in a number of separate streams may reflect defective adhesion between the fbm neurones themselves. This could in turn affect the response of fbm neurones to guidance cues.

On the basis of the present work, we propose a model for the control of fbm neurone migration involving a constant dialogue between the migrating cells and their environment. Within each rhombomere, fbm neurones encounter specific cues, that they interpret according to a pre-established program. Their responses involve modifications in gene expression, including changes in adhesion properties, which in turn participate in the selection of the local migratory pathway. We provide evidence that Ebf1 may contribute to the pre-established interpretation program, since its mutation affects the response of fbm neurones to the rhombomeric environment, and we propose that modifications in TAG-1, Ret, and Cdh8 expression, which are part of the elicited responses, are actually involved in the elaboration of the migratory pathways. The existence of such a dialogue between local cues and elicited responses of the migrating cells is likely to be of general significance and to explain complex migratory patterns by successive adaptation steps.

We are grateful to R. Grosschedl for the kind gift of the Ebf1 mutant mouse line. We thank members of the Charnay and Rubenstein labs for helpful comments on this work; and R. Grosschedl, F. Helmbacher, R. Hevner, F. Mechta, J. Nardelli, M. Studer, M. Wassef and K. Yun for stimulating discussions and critical reading of the manuscript. We are grateful to J.-F. Brunet, C. Goridis, D. Karagogeos, V. Pachnis, C. Redies, F. Sedel, M. Wassef and D. Wilkinson for the kind gift of probes, and to J. Lewis for providing the kr mouse line. S. G. was supported by fellowships from the ARC and the SSAS (Société de Secours des Amis des Sciences) and M. G.-D. was supported by a FEBS fellowship. This work was supported by grants from INSERM, MENRT, EC, ARC and AFM.