Defined concentrations of a posteriorizing signal are critical for MafB/Kreisler segmental expression in the hindbrain

In the vertebrate brain, each of the eight hindbrain segments or rhombomeres (r) is endowed with a definite identity, characterized by a unique combination of gene expression. These genes encode proteins of various families: transcription factors, transmembrane, cytoplasmic and secreted proteins. Among them, particular attention has been devoted to Hox genes. The 39 vertebrate members of this family are distributed in four clusters. Their expression is segmentally restricted in the nervous system and somites and colinear with their genomic layout. Each rhombomere thus expresses a specific set of genes, which constitutes its ‘Hox code’. By homology with their role in Drosophila development, vertebrate Hox genes are thought to define the identity of the rhombomere phenotype. Accordingly, changes in the Hox code lead to alterations of the developmental fate of specific rhombomeres (Lufkin et al., 1991; Chisaka et al., 1992; Carpenter et al., 1993; Dollé et al., 1993; Mark et al., 1993; Zhang et al., 1994; Alexandre et al., 1996; Goddard et al., 1996; Barrow and Capecchi, 1996).

If the decisive role of Hox genes in rhombomere specification is now well established, the pathways that lead to their expression in specific rhombomeres are still elusive. It was, however, recently shown that positional signals arising from the environment are involved in Hox gene induction (Grapin-Botton et al., 1995). These signals, distributed in an anteroposterior increasing gradient, are transmitted by the somites and along the neural tube itself in a planar manner (Grapin-Botton et al., 1995, 1997; Itasaki et al., 1996) and may be likened to the ‘transformer’ signal proposed by Nieuwkoop (1952a,b,c) as responsible for the anteroposterior patterning of the developing nervous system.

Signals regulating Hox gene expression also influence other genes considered as potential Hox gene targets, like certain members of the Eph family of tyrosine kinase receptors (Taneja et al., 1996, Itasaki et al., 1996). The fact that posterior transpositions of rhombomeres lead in certain cases to the homeotic transformation of rhombomeres (Grapin-Botton et al., 1995) strongly suggests that expression of several other genes acting downstream from Hox genes may be perturbed as a consequence of this change of the environmental cues.

Less is known about the effect of the posterior environment on the expression of genes acting upstream of Hox genes. The cascade that leads to Hox gene activation is well characterized in Drosophila but its evolutionary conservation in vertebrates is not clear. Only two genes acting upstream of Hox genes have been identified in vertebrates: the zinc finger transcription factor, Krox-20, and the basic leucine zipper transcription factor, Kreisler (Kr) (Sham et al., 1993; Nonchev et al., 1996; Manzanares et al., 1997). The mouse Kr mutation was identified in an X-ray mutagenesis screen by behavioural defects, characterized by head tossing and running in circles (Hertwig, 1942a,b). Kr/Kr mutants are deaf, which together with the circling behaviour indicates a deficiency of the vestibulo-acoustic functions. Morphological analysis revealed that the otic vesicles of these animals, which normally lie adjacent to rhombomeres 5 and 6 (r5/6) of the developing hindbrain, are displaced laterally and develop into a cystic structure without an organized vestibular apparatus or cochlea (Hertwig, 1944, Deol, 1964).

The dysmorphogenesis of the otic vesicle is now thought to be secondary to a defect that occurs early in hindbrain patterning. In Kr/Kr mutants, the caudal hindbrain has a smooth unsegmented aspect (Deol, 1964) caudal to r4 (Frohman et al., 1993, McKay et al., 1994). The changes in the expression pattern of several genes expressed in this region were interpreted as a loss of r5 and at least of part of r6 cells in kr/kr mice (Frohman et al., 1993, McKay et al., 1994). In particular, FGF3 transcripts, encoding a secreted protein required for the development of the otic vesicle, are repressed in rhombomeres 4-6 adjacent to this structure (Wilkinson et al., 1988; Represa et al., 1991; Mahmood et al., 1995). The derivatives of the neural crest originating from r5/6, i.e. the glossopharyngeal ganglion and nerve and the abducens nerve, are missing in the mutant (McKay et al., 1994). The hyoid bone, which is partly derived from these rhombomeres (Couly et al., 1996; Köntges and Lumsden, 1996), is also abnormal (Frohman et al., 1993).

The Kreisler gene was recently cloned in mouse after a submicroscopic chromosomal inversion was identified in the mutant (Cordes and Barsh, 1994). The avian homologue of this gene, MafB (sic MafB/Kr) was cloned by virtue of its homology to other members of the Maf family of transcription factors (Kataoka et al., 1994) and of its interaction with Ets-1 (Sieweke et al., 1996). The expression of this gene in r5/6 in mouse and chick corresponds to the tissues affected by the mutation (Cordes and Barsh, 1994; Eichmann et al., 1997). MafB/Kr is later expressed in several vestibular and acoustic nuclei and might be necessary for their maturation. These early and late expression domains may account for the deafness and circling behaviour. Additionally, MafB/Kr is expressed in several other tissues in the developing embryo (Eichmann et al., 1997). In Kreisler mutant mice, expression of the Kreisler gene in these tissues is not affected and their development is normal (Eichmann et al., 1997) meaning that, in the mutant, the chromosomal inversion perturbs the regulation of the expression of this gene only in the r5/6 region.

The fact that MafB/Kr acts upstream of Hox genes is strongly suggested by two sets of data : (i) in Kreisler mutant mice the expression of several Hox genes is modified and (ii) a direct binding of Kreisler on Hoxb-3 promoter appears to activate Hoxb-3 expression in r5 (Manzanares et al., 1997). In order to detemine whether the environmental signal that regulates Hox gene expression in the hindbrain also regulates upstream genes, we used the grafting strategy previously applied to Hox gene analysis. We found that MafB/Kr is downregulated in rhombomeres 5/6 grafted into a more caudal position, consistent with the acquisition of an r8 identity. In contrast, like Hox genes, MafB/Kr expression is not altered when r5/6 are transposed anteriorly. We show that the effect of the posterior environment is mimicked by grafting either posterior somites or beads soaked in retinoic acid (RA), laterally to the rhombencephalon. Our experiments thus suggest that the expression of anteroposterior identity markers in a restricted level of the neural tube, can be triggered by a strictly defined concentration of a morphogen secreted by the somitic mesenchyme and that this effect can be obtained by defined concentrations of RA. In addition, we show that the signal provided by the somites can change neural crest cell identity together with that of neural tube cells.

Quail (Coturnix coturnix japonica) and chick (Gallus gallus) eggs from commercial sources were used throughout this study. Microsurgery was performed on embryos at the 5- to 8-somite stage (about 30 hours of incubation in a humidified atmosphere at 38°C). A window was cut in the shell and India ink diluted 1:1 in PBS was injected into the sub-blastodermic cavity in order to make the embryonic structures more visible without using any ‘vital’ stain. The vitelline membrane was then windowed to make the embryo accessible. For rhombomere transpositions, a gap corresponding to a unilateral pair of rhombomeres was made in the neural tube using a microscalpel, made by sharpening a steel needle on an Arkansas stone. At 5-somite stage, the limits of the grafted tissues and of the sites of implantation are defined according to the map of presumptive rhombomere territories established by Grapin-Botton et al. (1995). At later stages, constrictions between rhombomeres were used as landmarks. Hemineural plates corresponding to a pair of rhombomeres were dissected out from a quail donor and grafted heterotopically after removal of a size-matched piece of neuroepithelium in the chick host embryo. The neuroepithelium was not dissociated from its surrounding tissues by enzymatic digestion. The notochord was left intact in the host embryos and was never included in the graft. Carbon particles were positioned dorsally at the rostral and caudal extremities of the graft to indicate the limits of the transplant after healing. For somite grafting, the somites were dissected surgically from a quail donor and grafted in a slit performed lateral to r4-6 or r1-3.

The operated embryos were fixed in 4% formaldehyde : 2 Mm EGTA in PBS. When expression in the neural tube was in focus, expression was analyzed between stages 11 and 15 according to Hamburger and Hamilton (HH) (1951). For neural crest cell expression, embryos were fixed between stages HH 13 and 14 since neural crest cell expression fades away after this stage.

In three embryos, a small DiA crystal was put in a slit at the level of r4 neural crest and somites 7-8 were grafted lateral to r4-6. One night after grafting, embryos were fixed in 4% paraformaldehyde in FISH buffer (4% sucrose, 0.12 mM CaCl₂, phosphate buffer 0.1 M pH 7.4), rinsed and included in gelatin 7.5%/sucrose 15% in phosphate buffer. 10 μm cryostat sections were cut, photographed under fluorescent stimulation and treated for in situ hybridization.

AG1-X2 beads (Biorad) were quickly rinsed in PBS. They were then incubated for at least 1 hour at room temperature in a retinoic acid solution in PBS. All-trans-RA (Sigma) stock solution was first prepared in ethanol at a concentration of 10 mg/ml and then diluted at the appropriate concentration in PBS. The beads were quickly rinsed in PBS and grafted in the embryo at the 5- to 8-somite stage, lateral to r4/5. In one experimental series, RA beads were grafted at stage HH19-20 in the anterior part of the limb bud, in a dorsal position opposite to the zone of polarizing activity across the proximodistal axis of the limb bud.

Heparin beads (Sigma) were soaked in a 1 mg/ml solution of bFGF (R and D systems) for at least 1 hour. They were rinsed and grafted lateral to r5 in a 5- to 8-somite-stage embryo.

A 920 bp MafB/Kr cDNA fragment spanning nt 84-936 of the coding region and 68 bp of the 3′ UTR (Sieweke et al., 1996), cloned into a pGEM3z vector (Promega), was used to make a digoxigenin-UTP (Boehringer Mannheim)-labelled RNA probe (Promega, Riboprobe Gemini II). Whole-mount in situ hybridizations were performed as described by Henrique et al. (1995). The washing time of embryos after E3 was multiplied by three when compared to Henrique et al. (1995), covering a total of about 9 hours.

In situ hybridization was performed using a protocol described by Strähle et al. (1994) with minor modifications.

MafB/Kr is expressed in the hindbrain in a dynamic pattern, as described by Eichmann et al. (1997). Its signal begins in r6 at the 5-somite stage. The level of expression increases quickly, reaching its highest levels at 6-somite stage, in r5 and r6. The presumptive territories of the rhombomeres were transplanted at the 5- to 8-somite stage, i.e. after the onset of MafB/Kr expression in r5/6 and before emigration of neural crest cells. MafB/Kr transcripts are present in the neural crest cells as they start migrating and thereafter decrease as the cells move away from the tube, to disappear toward stage HH14. Expression of the gene in the neural tube at r5/6 level disappears at stage HH16-17. In addition, MafB/Kr transcripts are present during stages HH13-15 in the roof plate, from rhombomere 1 down to the 10th somite.

In order to know if the environment exerts an effect on MafB/Kr expression, we transplanted r5/6 down to the r8 level (Fig. 1A) at the 5- to 8-somite stage, after the onset of MafB/Kr expression and localized its mRNA by whole-mount in situ hybridization, 10 hours (at stage HH12), and 15-18 hours (at stage HH13-15) after grafting. We found that, after 10 hours, MafB/Kr was totally repressed in 3 out of 4 cases and in the last one only a faint expression remained (Fig. 1B). 15-18 hours after grafting, the expression was lost in all 13 chimeras analyzed (Fig. 1C).

Two interpretations could be given to these experiments: (i) the expression in r5/6 could be maintained by a signal acting in the r5/6 position but absent in r8 (ii) alternatively, MafB/Kr expression in r5/6 could be actively downregulated by the posterior environment. To discriminate between these two possibilities, we withdrew r5/6 from its normal environment by putting it into the r3/4 position (Fig. 1A). In this position, MafB/Kr expression was maintained in the graft in the 10 chimeras fixed at stage HH14-15 (Fig. 1D). Moreover, grafting either r3/4 (n=3) or r8 (n=10) (Fig. 1E,F) into r5/6 position did not result in the onset of MafB/Kr expression at stage HH14-15.

We conclude from these experiments that, after the 5- to 8-somite stage, no signal from the environment is necessary to maintain MafB/Kr expression since, after anterior transplantation of r5/6, the gene continues to be normally expressed. In contrast, a negative signal present in the environment of r8 but not in that of r3/4 is able to repress this gene in r5/6 grafts.

In a second set of experiments, we looked for the source of the repressing signal. It was previously shown (Itasaki et al., 1996; Grapin-Botton et al., 1997) that the somites are able to induce Hoxb-4 expression when grafted adjacent to rhombomeres r4-The efficiency of the somites to induce Hoxb-4 increased rostrocaudally. We thus decided to graft somites 7/8 or 9/10 laterally to r4-6 (Fig. 2A) and assessed MafB/Kr expression at stage HH13-15. A repression of MafB/Kr expression was observed in r5/6 in all 13 chimeras, although in some cases only partially (Fig. 2B,C). The somites can therefore mimic the posterior environment. This effect was not reproduced by anterior somites. Somites 1/2 were not able to modify MafB/Kr expression (n=7). The effect on MafB/Kr expression in r4-6 required close contact between neural tube and somites since grafting somites 7-8 or 9-10 at the level of r6-8 only repressed MafB/Kr expression in r6 (n=4).

A signal unevenly distributed in the somites is thus able to repress Kreisler expression.

When somites 7-8 or 9-10 were grafted lateral to r4-6, in addition to the repression of Kreisler in r5/6, MafB/Kr was induced in r4 in 6 out of the 13 chimeras (Fig. 2C). Thus rhombomere 4 acquired a molecular characteristic of r5/6.

We then grafted the somites more rostrally to test whether MafB/Kr could also be induced in more anterior rhombomeres. In 4 chimeras, somites 7/8 or 9/10 were grafted lateral to r1-3. Under these conditions, MafB/Kr expression was not modified in the rhombencephalon. More posterior somites (12-14) were also unable to induce MafB/Kr when transplanted at the same anterior levels (n=4). Thus, only r4 is able to respond to the somitic signal by inducing MafB/Kr expression, meaning that the response of a given anteroposterior level to the somitic signal depends on the competence of the rhombomeres. Thus, the competence of r4 is higher than that of r1-3.

In view of the fact that RA administration to vertebrate embryos is able to modify the anteroposterior patterning of the hindbrain (see Mavilio, 1993 and Conlon, 1995 for reviews), we investigated whether the graft of a bead soaked in various concentrations of RA could affect MafB/Kr expression in rhombomeres. RA-soaked beads were inserted laterally to r4-6 and the experiments relied on the local delivery of RA to the neighbouring neural epithelium (Fig. 3A). When beads were soaked in a solution containing 10⁻⁴ M RA, the expression in r5/6 was always maintained and, in most cases, no other effect was observed (n=11/17). In a few cases, however, an extension of the expression domain to r4 (n=3) or r3/4 (n=3) took place (Fig. 3B). When RA concentration was raised to 5×10⁻⁴ M, the effect was more striking. MafB/Kr expression was repressed in r5/6 in most cases (n=10/11) (Fig. 3C). An anterior induction that reached various anteroposterior levels (r4, n=2; r3, n=4; r2, n=4) was also observed (Fig. 3C). In only one case, was no effect noted. Increasing the concentration to 10< sup>−3 </ sup >M resulted in the death of the embryos in about 50% of cases before the 15-somite stage. 30 embryos that were alive 18 hours after the operation were analyzed. Two of them had a normal MafB/Kr pattern. One showed a maintenance of r5/6 expression and a rostral extension into r3/4. Most of them lacked the r5/6 expression domain while expression was displaced anteriorly to r3 (n=8), r2 (n=3) or r1 (n=6). In the 10 remaining embryos, expression was completely abolished (Fig. 3D). It should be noted that the expression in the roof plate remained always unchanged despite the vicinity of the bead.

These observations show that (i) all the rhombomeres are competent to express MafB/Kr, provided that the concentration of RA reaches sufficient levels and (ii) the increase of threshold levels necessary to induce MafB/Kr indicates that the competence of rhombomeric levels to express this gene obeys a caudorostrally decreasing gradient. The fact that the limits of induction do not coincide with rhombomere boundaries indicates that the gradient is uniform.

RA is well known to induce Hox gene expression in both the hindbrain and the limb (see Mavilio, 1993 and Conlon, 1995 for reviews; Gale et al., 1996). Since MafB/Kr is also expressed in the limb, we designed an experiment to investigate whether MafB/Kr expression can be modified by RA in the limb in the same manner as Hox gene expression (Oliver et al., 1990; Nohno et al., 1991; Izpisua-Belmonte et al., 1991). Fig. 4B shows that, in control limbs (left) at E4, the mesenchyme in anterior and posterior sides of the limb bud express MafB/Kr. When a bead soaked in 10⁻³ M RA was grafted at stage HH19-20 into the anterior part of the limb bud (Fig. 4A), in a position opposite to the zone of polarizing activity with respect to the proximodistal axis of the limb, the anterior expression of MafB/Kr at E4 was extinguished (n=4/6) (Fig. 4B).

Experiments carried out mostly in Xenopus embryo have shown that several molecules are also able to posteriorize the neural tube and, in particular, to induce Hox genes. eFGF and bFGF (Cox and Hemmati-Brivanlou, 1995; Lamb and Harland, 1995; Kengaku and Okamoto, 1995; see also Doniach, 1995 for a review) have this capacity. We thus tested, at 5- to 8-somite stage, the effect of bFGF-soaked beads grafted lateral to r5 on MafB/Kr expression. bFGF-soaked bead neither repressed MafB/Kr expression in r5/6 nor induced it in r4 (n=7) (Fig. 5).

The neural crest cells that migrate from r5/6 caudal to the otic vesicle express MafB/Kr during the early stages of their migration (Eichman et al., 1997). MafB/Kr is lost by the 20-somite stage, when crest cells move further away from the neural tube. The expression rostral to the otic vesicle is very low and detected only if the time of alkaline phosphatase reaction is significantly increased. We analyzed MafB/Kr expression in neural crest cells after the graft of somites 7/8 or 9/10 (Fig. 6A) or RA beads lateral to r4-6 (Fig. 3A), in embryos fixed before the normal downregulation of MafB/Kr, i.e. stage HH13-14. As mentioned before, in most cases, r4 neuroepithelial cells were induced to express MafB/Kr in this situation. Such is also the case for neural crest cells of r4 origin, at a level comparable to that observed normally in r6-derived neural crest cells (n=21/23) (Fig. 6B). In some cases, the somites grafted lateral to r4 modified the migration pathway of neural crest cells exiting from this rhombomere. However, even in these cases MafB/Kr-positive cells could be seen forming a stream in close contact to the neural tube at the r4 level (Fig. 6B). In some cases (n=3), r4 neural fold was labelled with DiA before the onset of migration. During migration, a subpopulation of DiA-labelled cells was found to be MafB/Kr-positive (Fig. 6C,D). In 19/23 cases, MafB/Kr-expressing cells were also found rostral to the otic vesicle opposite the graft. While MafB/Kr is repressed in the neural epithelium of r5/6, it is maintained in the associated neural crest cells.

When beads soaked in RA were inserted in the same position (Fig. 3A), the response varied according to RA concentration. At 10< ^>−4 M, the neural crest pattern was normal in 6/9 cases (Fig. 3B). MafB/Kr was found lateral to r4 in 3/9 cases. At 5×10⁻⁴ M (n=5), MafB/Kr expression was anteriorized to the level of r4 (n=2) or r3 (n=2) (Fig. 3C). At 10⁻³ M, the expression of MafB/Kr in the neural crest outflow lateral to r3 or 4 was noted in 8/8 cases (Fig. 3D).

The identity of the metameric units constituting the hindbrain is defined by a combinatorial activity of several developmental genes, among which those of the Hox clusters occupy a central place. The spatiotemporal expression pattern of these genes is regulated by external cues unevenly distributed along the anteroposterior axis in the neural plate and paraxial mesoderm (Grapin-Botton et al., 1995, 1997; Itasaki et al., 1996).

Besides the Hox genes, other genes acting upstream from them are selectively expressed in different anteroposterior levels of the hindbrain. Krox-20, expressed in r3 and r5, was shown to be one of them (Sham et al., 1993; Nonchev et al., 1996). In this study, we have been interested in another example of this group of genes, the basic leucine zipper regulatory gene, MafB/Kr, which is expressed in r5/6 and is necessary for the development of the vestibulo-acoustic system (Hertwig, 1942a,b; Deol, 1964; Cordes and Barsh, 1994).

We have previously shown that transposition of r5/6 (which normally yield the auditory nuclei magnocellularis, laminaris and a small part of angularis) (Marin and Puelles, 1995) to an r8 position results in the acquisition by the transposed neuroepithelium of the Hox code characteristic for r8. This is followed by the expression of the corresponding posterior phenotype (Grapin-Botton et al., 1995). It was therefore pertinent to see if expression of MafB/Kr is affected when r5/6 are transferred to r8. We show that, if r5/6 are transposed to r8, MafB/Kr is repressed by the environment of r8. This demonstrates that the r8 environment induces not only the gain of posterior markers but also the loss of r5/6-specific markers. Therefore, as it is the case for Hox genes, anterior transposition of rhombomeres has no effect on MafB/Kr expression while posterior transposition induces changes that are in agreement with the novel position of the neuroepithelium along the neural axis. However, unlike Hox genes, the posterior environment can either repress or induce MafB/Kr, depending on the receptive field. This property was also observed by Itasaki et al. (1996) for an element of Hoxb-2 promoter.

Interestingly, in the absence of MafB/Kr expression, posteriorly transposed r5/6 express genes of the fourth paralogue group and produce a phenotype characteristic of the posterior brain stem including the motor nuclei of nerves X and XII, the inferior olivary nucleus and nucleus supraspinalis (Grapin-Botton et al., 1995). Whether MafB/Kr plays an active role (i.e. other than by repressing Hox genes) in the development of the auditory system is not known. However, MafB/Kr is not only expressed early in rhombomeres 5/6 but also later in some of their derivatives, the vestibular and acoustic nuclei, in particular nucleus laminaris and angularis (Eichmann et al., 1997). The expression of MafB/Kr in these nuclei, is not directly linked to the early expression in r5/6 since the nucleus angularis, which is derived mainly from r3/4 (Marin and Puelles, 1995), is completely labelled by MafB/Kr. The expression of this gene may therefore be required for these nuclei to differentiate. When r5/6 are transposed into r8, the nucleus laminaris does not differentiate (Grapin-Botton et al., 1995). Thus, the lack of acoustic nuclei in r5/6 transposed into r8 could be a direct consequence of MafB/Kr downregulation.

A direct involvement of Kreisler in Hox gene regulation was recently attested by Manzanares et al. (1997), who found two sites in the Hoxb-3 promoter able to bind the Kreisler protein and to direct Hoxb-3 expression in r5, and r6 under certain conditions. Although no studies have been carried out yet with respect to a MafB/Kr function on other Hox gene promoters, modifications in the gene expression pattern in Kr/Kr mutant mice suggest that it might also act upstream to Hox genes of the fourth paralogue group. In the Kreisler mutant mice, Hoxb-4 is extended rostrally to the anteroposterior level of r6 instead of r7 (Frohman et al., 1993) and Hoxd-4 to that of r5 (McKay et al., 1994). Hoxb-1, which is normally expressed in r4, r7 and r8 is extended to r5 (Frohman et al., 1993). These modifications can be interpreted either by a loss of r5/6 cells or by a change in r5/6 cell specification (Frohman et al., 1993; McKay et al., 1994). In the latter interpretation, MafB/Kr would downregulate Hoxb-4 and Hoxd-4 expression in r5/6 under normal conditions and the absence of Kreisler expression in Kr/Kr mutant would allow the anterior extension of paralogue group four Hox genes. Our transplantation experiments are in agreement with this view since the upregulation of paralogue group four genes correlates with the inhibition of MafB/Kr expression. Timing considerations are also consistent with the hypothesis that absence of the MafB/Kr gene product in vivo might derepress the expression of paralogue group four genes. By whole-mount in situ hybridization, MafB/Kr is faintly revealed 10 hours after grafting and completely absent after 18 hours. Even though Hoxb-4 induction is observed after 10 hours by the more sensitive radioactive in situ hybridization on sections, it is only seen after more than 24 hours by whole-mount techniques.

The fact that MafB/Kr is not expressed in r3/4 from the beginning supports the contention that the induction of paralogue group four Hox genes in these rhombomeres requires an inductive signal independent from MafB/Kr. The absence of MafB/Kr is thus a requirement for paralogue group four Hox gene expression only in rhombomeres that receive a sufficient amount of posteriorizing signal to express these genes. Their mutual exclusion in vivo in might therefore reflect the fact that MafB/Kr gene product is a downregulator of paralogue group four genes needed to establish a sharp limit between r5/6 and r7/8 identities.

As previously demonstrated for Hox gene induction (Itasaki et al., 1996; Grapin-Botton et al., 1997; Gale et al., 1996), the posteriorizing signal able to switch off MafB/Kr expression in r5/6 can be transmitted to the neuroepithelium by direct contact with somites and by the direct application of RA.

The experiments involving either anterior transplantations of somites or exposure to various concentrations of RA led to two important results. First, rhombomeres anterior to r5 have the competence to express MafB/Kr but the threshold at which the gene is activated increases anteriorly from r4 to r1. Secondly, activation of MafB/Kr gene takes place only over a defined range of concentration of the inducer. For example, the concentration reached by apposition of somites 7-10 to r4-6 switches off MafB/Kr expression in r5/6 and switches it on in r4. Moreover, activation of MafB/Kr in r3 to r1 requires increasing concentrations to trigger a response. In these experiments, this was only reached by RA beads and not by somites. These experiments show that a single molecule originating from the environment might account for restricted gene expression domains in the hindbrain. The model of a gradient of a single morphogen distributed along the anteroposterior axis represented in Fig. 7 summarizes this interpretation of our observations. Our results are in agreement with the importance of RA concentration in Hox gene activation already observed by Simeone et al. (1990) and stress the fact that RA can regulate the expression of other segmentally expressed genes. Although we have no proof that RA is the actual endogenous signal molecule in posteriorization by the somite, these experiments show that it can mimic such a signal.

As in the case of Hox gene induction, how the signal is set up in normal development remains a question mark. Our experiments, however, indicate that at least at the stages immediately following the transposition of r5/6 to r8, the signal must travel via the plane of the neuroepithelium, since the anterior somites at the level of r8 are themselves unable to induce Hox genes and inhibit MafB/Kr. Other molecules than RA could be the endogenous morphogen released by the somites. Experiments carried out mostly in Xenopus embryo have shown that several molecules are also able to posteriorize the neural tube and, in particular, to induce Hox genes. Some members of the FGF family (Cox and Hemmati-Brivanlou, 1995; Lamb and Harland, 1995; Kengaku and Okamoto, 1995; see also Doniach, 1995 for a review) as well as of the Wnt family and downstream molecules (McGrew et al., 1995; Itoh and Sokol, 1997) have this capacity. We show here that in these conditions bFGF is not able to regulate MafB/Kr expression. It does not exclude however that FGFs may be involved in anteroposterior patterning at an earlier step or in more caudal regions.

It is difficult to determine precisely the exact thresholds of RA concentrations able to induce or repress MafB/Kr from our experiments involving RA-soaked beads. RA has been shown to be progressively released over about 20 hours (Eichele et al., 1984), but we can determine with precision neither the amount of RA released from the beads nor the kinetics at which it is released. Nevertheless these studies clearly support the general concept of the gradient hypothesis in the regulation of gene expression along the anteroposterior axis in the hindbrain.

The expression of MafB/Kr in its various sites is differentially controlled. We report here two examples that respectively show that MafB/Kr is (i) differentially regulated in the rhombomeres and the hindbrain roof plate and (ii) can be regulated by the same molecule in the rhombomeres and the limb.

In the limb, Hox genes are initially expressed in the posterior side. When grafted anteriorly, RA beads are able to induce Hox genes in the anterior part of the wing mimicking the influence of the zone of polarizing activity (Oliver et al., 1990; Nohno et al., 1991; Izpisua-Belmonte et al., 1991). We show here that RA is also able to repress MafB/Kr expression anteriorly. In the hindbrain, Hox genes are induced anteriorly by somites and by appropriate concentrations of RA. As in the limb, Hox gene induction in the hindbrain is correlated with the downregulation of MafB/Kr expression. The hindbrain and the limb thus share the expression of certain gene families and of their regulatory pathways.

MafB/Kr expression in the rhombencephalon is not restricted to r5/6. It is also expressed in the roof plate and rhombic lip (Eichmann et al., 1997). While regulating MafB/Kr expression in the rhombomeres, the somites have no effect on its expression in the roof plate. Only the expression domains that are segment-specific are affected by changing RA concentration. The dorsal expression is probably dependent on other regulatory pathways: the repression by the posterior environment could be masked by another upregulating element active in the roof plate only. Alternatively, the dorsal cells could be unable to receive the signal or lack cofactors to respond to it. Consistent with this idea, in Kreisler mutant mice, MafB/Kr expression is lost only in the rhombomeres, whereas various MafB/Kr expression sites other than r5/6 such as the intestine, liver, spleen and bone marrow, are not affected (Eichmann et al., 1997).

The graft of somites 7-10 or of RA-soaked beads also affected the neural crest expression of MafB/Kr. The stream of neural crest cells (NCCs) that migrate rostrally to the otic vesicle derives from rhombomeres 3, 4 and 5 (Sechrist et al., 1993; Birgbauer et al., 1995). These crest cells normally express MafB/Kr at a very low level compared to the cells that migrate from r5/6 caudal to the otic vesicle (Eichman et al., 1997). Somites grafted lateral to r4-6 were able to induce MafB/Kr in the neural tube rostral to the otic vesicle. Positive cells immediately adjacent to r4 showed that MafB/Kr was also induced in NCCs issued from this rhombomere. It has been shown in two previous publications that Hox gene expression is independently regulated in neural tube and NCCs. Prince and Lumsden (1994) showed that Hoxa-2 expression in r2 is limited to the neural tube. Similarly, Hoxa-3 is expressed in r5 but only in the subset of its NCCs which migrate into branchial arch 3. Saldivar et al. (1996) showed that Hoxa-3 expression is maintained in r5 (an odd rhombomere) ectopically transposed to r4, while it is downregulated in the corresponding NCCs when they reach the distal part of branchial arch 2. In contrast, when r6 is transposed to r4, Hoxa-3 expression remains unchanged in the NCCs. These experiments show therefore that Hox gene expression can take place independently in the neural tube and in the neural crest. The case described here shows the unprecedented simultaneous upregulation of a positional marker in the neural tube and NCCs. Taken together, these results suggest that several steps can be distinguished in the specification of anteroposterior identity of NCCs. At an early stage, the cells that are still in the neural fold can be influenced by a posterior signal such as somites or RA beads showing that their anteroposterior identity is not yet determined. Later on, the same signals have no effect on gene expression in NCCs. This may be due to the fact that they are no longer competent to respond to these cues or that the signals cannot reach them during their migration. In a third step, signals from the branchial arch environment might refine Hox gene expression. This is suggested by the fact that cells derived from odd rhombomeres can be led to express a different Hox code from that of their rhombomere of origin (Saldivar et al., 1996).

In conclusion, we show here that expression of the leucine-zipper transcription factor MafB/Kr, which is normally expressed in a restricted hindbrain level, whose fate is related to fulfill auditory and vestibular functions, is controlled by defined concentration of a posteriorizing signal. The problem as to whether this signal is of the same nature as the one that regulates Hox gene expression is now raised.

The authors want to acknowledge Françoise Viala and Francis Beaujean for photographic assistance. They are also indebted to Heather Corbett for suggestions during writing. This work was supported by the CNRS, Collège de France, Association pour la Recherche contre le Cancer (ARC), Ligue contre le Cancer, and Fondation pour la Recherche Medicale (FRM).