Nab proteins mediate a negative feedback loop controlling Krox-20 activity in the developing hindbrain

The development of the vertebrate hindbrain involves a transient segmentation process along the anterior-posterior (AP) axis, which leads to the formation of 7-8 morphological units termed rhombomeres (r) (Lumsden and Keynes, 1989; Lumsden and Krumlauf, 1996; Wingate and Lumsden, 1996; Schneider-Maunoury et al., 1998). The rhombomeres behave as units of lineage restriction (Fraser et al., 1990; Birgbauer and Fraser, 1994) and play an essential role in establishing the pattern of both hindbrain and craniofacial morphogenesis. In particular, this subdivision presages the periodic organisation of neurons (Lumsden and Keynes, 1989; Clarke and Lumsden, 1993) and correlates with the pathways of neural crest migration into the branchial arches (Lumsden and Guthrie, 1991; Lumsden et al., 1991; Serbedzija et al., 1992; Birgbauer et al., 1995; Kontges and Lumsden, 1996).

A number of putative regulatory genes have been shown to present spatially restricted patterns of expression along the AP axis in the hindbrain, with limits corresponding to prospective or established rhombomere boundaries (reviewed in Lumsden and Krumlauf, 1996). Mutational analyses indicate that several of these genes do indeed play important roles in the control of hindbrain development (reviewed in Schneider-Maunoury et al., 1998). Among them, Krox-20, which encodes a zinc finger transcription factor (Chavrier et al., 1988, 1990), appears to be essential for the formation and specification of r3 and r5. Krox-20 is successively activated in two transverse stripes, which prefigure and subsequently coincide with r3 and r5 (Wilkinson et al., 1989; Schneider-Maunoury et al., 1993). It is also expressed in neural crest cells emigrating from r5 toward the third branchial arch (Schneider-Maunoury et al., 1993; Nieto et al., 1995). Targeted inactivation of Krox-20 leads to progressive disappearance of the r3 and r5 territories (Schneider-Maunoury et al., 1993, 1997; Swiatek and Gridley, 1993) indicating that this gene participates in the control of the segmentation process. Furthermore, Krox-20 has been shown to regulate the transcription of several Hox genes in r3 and/or in r5, directly or indirectly (Sham et al., 1993; Nonchev et al., 1996a,b; Vesque et al., 1996; Seitanidou et al., 1997). Since the combinatorial expression of Hox genes is believed to determine the AP positional identity of the rhombomeres, this indicates that Krox-20 also plays an important role in the specification of positional information. Finally, Krox-20 has been shown to directly control the transcription of EphA4, which encodes a transmembrane tyrosine kinase receptor (Theil et al., 1998). Since Eph family members and their ligands have been implicated in the segregation of cells between odd and even rhombomeres (Xu et al., 1995, 1999; Mellitzer et al., 1999), this observation suggests that Krox-20 could participate in the establishment of lineage restrictions within the hindbrain. Taken together these various data indicate that Krox-20 is a key regulator of gene expression in r3 and r5 and co-ordinates various aspects of the segmentation and specification processes.

While several of the molecular targets of Krox-20 have been identified, very little is known about the regulation of Krox-20 expression and activity. Two potential modulators of Krox-20 transcriptional activity, Nab1 and Nab2, have been discovered (Russo et al., 1993, 1995; Svaren et al., 1996). These proteins were identified on the basis of their capacity to directly interact with Krox-24 (also known as NGFI-A, Egr1 or Zif268), a zinc finger transcription factor closely related to Krox-20 (Milbrandt, 1987; Christy et al., 1988; Lemaire et al., 1988; Sukhatme et al., 1988). It was shown that they can also bind to Krox-20 and that they inhibit transcriptional activation by both zinc finger proteins in cultured cells (Russo et al., 1995). This inhibition presumably involves the formation of a ternary complex between the DNA target sequence, Krox-20 or Krox-24, and the Nab protein, the latter carrying a transcription repression domain (Svaren et al., 1998; Swirnoff et al., 1998). The potential involvement of Nab1 and Nab2 in the modulation of Krox-20 function led us to investigate their expression in the developing hindbrain. We found that their expression patterns closely overlap with that of Krox-20 in r3 and r5. Furthermore, in Krox-20^−/− embryos, Nab1 and Nab2 are not expressed in the hindbrain, indicating that these genes are downstream of Krox-20. Finally, misexpression experiments performed in zebrafish embryos suggest that both Nab1 and Nab2 can antagonise Krox-20 action. In conclusion, these data indicate that Krox-20 controls the expression of its own antagonists and suggest that this feedback regulatory loop plays an important role in the development of the hindbrain.

Mouse cDNAs corresponding to the complete coding sequence of Nab2 (1575 nucleotides encoding the 525-amino-acid protein) and the near-complete coding sequence of Nab1 (1383 nucleotides encoding the 461 N-terminal amino acids of the 489-amino-acid protein) were isolated by reverse transcription-PCR amplification. Total RNA was prepared as described (Chomczynski and Sacchi, 1987), from newborn mouse brain (Nab1) or Ras-transformed NIH3T3 fibroblasts (Nab2). cDNA was prepared with MMLV Reverse Transcriptase using degenerated hexamers (pDN6, Pharmacia, Uppsala, Sweden), following the conditions of the manufacturer (Life Technologies, Bethesda, MD). The product of the reaction was diluted 20-fold for PCR amplification in the buffer recommended by the manufacturer of the Taq DNA Polymerase (Eurobio). The sequences of the primers were deduced from the published Nab1 and Nab2 nucleotide sequences (Russo et al., 1995; Svaren et al., 1996): Nab1, 5′-ATGGCCACAGCCTTACCTAGGAC-3′ and 5′-CCCAAGGCTC-TCTGAGGAGTGGG-3′; Nab2, 5′-ATGCACAGAGCTCCCTCTCC-CACAG-3′ and 5′-TCACTGCCGGCTGGCTTCTGCCTC-3′. 25 cycles of amplification were carried out at 94°C for 1 minute, 55°C for 1 minute, 30 seconds and 72°C for 1 minute, 30 seconds. The expected products of amplification were purified by electrophoresis on 1% agarose gel (Quiaex kit, Quiagen) and inserted into the PCRIITm plasmid by TA cloning (TA cloning kit, Invitrogen, La Jolla, CA). The fragments were subsequently recloned into Bluescript KS+, sequenced and shown to encode bona fide Nab1 and Nab2 proteins. For the preparation of sense RNA for injection in zebrafish embryos, the cDNAs were recloned into the pCS2+ vector (Turner and Weintraub, 1994). The concentrations of RNA preparations were estimated by gel electrophoresis analysis.

Postimplantation mouse embryos were recovered at the appropriate stage (8-10 days post coitum: dpc) and fixed at 4°C in PBS containing 4% paraformaldehyde for 30 minutes for X-gal staining or for 8 hours for direct in situ hybridisation. X-gal staining was performed as described (Sham et al., 1993). Embryos were refixed for 8 hours after X-gal staining, dehydrated in methanol and processed for whole-mount in situ hybridisation as described (Seitanidou et al., 1997). Digoxigenin antisense RNA probes were generated using a Stratagene RNA transcription kit.

Zebrafish eggs were obtained by natural spawning, collected and maintained at 28.5°C in E3 embryo medium supplemented by Methylene Blue (4 μg/ml) to inhibit fungal growth (Haffter, 1996). Embryos were staged according to Kimmel et al. (1995). They were injected at the 2-4-cell stage with 200 pl of a solution containing the RNA and 0.2% Phenol Red as described (Wittbrodt and Rosa, 1994). They were then allowed to develop until the 5-to 10-somite stage and were fixed for 30 minutes at room temperature in 4% paraformaldehyde in PBS containing 0.1% Tween 20 (PBT). After three washes in PBT, they were dissected from the vitellus and stained with X-gal for 15 minutes as described (Haddon et al., 1998). The embryos were then refixed overnight at 4°C, dehydrated in methanol and processed for in situ hybridisation (Oxtoby and Jowett, 1993). Single in situ hybridisations were performed with digoxygenin-labelled probes. Double in situ hybridisations also included a fluorescein-labelled Krox-20 probe and were performed as described (Hauptmann and Gerster, 1994; Prince et al., 1998), except that we used BM purple (Boehringer) as a substrate for the detection of digoxygenin instead of NBT and BCIP. After hybridisation, the embryos were stored in PBT at 4°C, flattened under a coverslip and photographed in 100% glycerol. Several concentrations of Nab1 or Nab2 RNAs were initially tested (5, 10, 20, 50 and 100 ng/μl) in these injection experiments. The highest concentrations (50 and 100 ng/μl) led to a high proportion of embryos with gastrulation defects and the lowest (5 ng/μl) did not induce any phenotype. We therefore selected the concentration of 20 ng/μl for further experiments and coinjected lacZ RNA with Nab RNA to reach a constant total RNA concentration of 100 ng/μl.

To investigate a role for Nab1 or Nab2 in the modulation of Krox-20 transcriptional activity in the hindbrain, we examined their expression patterns by whole-mount in situ hybridisation during the period of Krox-20 expression in r3 and r5, i.e. between 8 and 10 dpc. During this period, we found that, like Krox-20, the expression of both genes was restricted to r3 and r5 (Fig. 1). The localisation of the sites of expression was first estimated according to morphological landmarks (pre- and post-otic sulci, otic placode). Nab1 is activated in the hindbrain at around the 6- to 8-somite stage approximately simultaneously in prospective rhombomeres (pr) 3 and 5 (Fig. 1I). At this stage, the level of expression appears stronger in pr5 than in pr3 and close to the interface between these prospective rhombomeres and the adjacent ones. Subsequently, the expression becomes progressively more uniform within r3 and r5, while remaining higher in the latter rhombomere (Fig. 1J). Finally Nab1 expression fades away in r3 and in r5, starting from their ventral part (Fig. 1K,L), persisting in dorsal r5 until the 30-somite stage (data not shown). The pattern of Nab1 extinction in r3 and r5 is very similar to that of Krox-20 (Fig. 1D-F; Irving et al., 1996).

Nab2 activation in the hindbrain occurs earlier than Nab1, at around the 1-somite stage, and is first restricted to a narrow stripe likely corresponding to pr3 (Fig. 1N). Later on, this expression becomes more intense (Fig. 1O) and at around the 6-somite stage a second, more caudal stripe of expression is observed coinciding with pr5 (Fig. 1P). This sequential activation of Nab2 in pr3 and pr5 faithfully reflects the pattern of activation of Krox-20 in these prospective rhombomeres (Fig. 1A-C; Irving et al., 1996). At later stages Nab2 is expressed at similar levels in r3 and r5, with a slight reinforcement at the boundaries of these rhombomeres (Fig. 1Q-S). At around the 12-somite stage, Nab2 expression was also detected in the r5-derived neural crest migrating toward the third branchial arch (Fig. 1Q, arrowheads). From the 16-somite stage, Nab2 became progressively downregulated in r3 and r5 and the RNA was not detected beyond the 25-somite stage (data not shown).

To confirm the coexpression of Nab1 and Nab2 with Krox-20 in r3 and r5, we performed double labelling experiments. For this purpose, we made use of a mouse line carrying a ‘knock-in’ of lacZ into the Krox-20 locus (Schneider-Maunoury et al., 1993). Embryos heterozygous for this mutation develop normally and the Krox-20/lacZ gene recapitulates the normal expression of Krox-20. Nab1 and Nab2 mRNAs were detected by in situ hybridisation and Krox-20 expression by X-gal staining. The Nab1 and Nab2 expression domains were completely included within the X-gal positive regions (Fig. 1M,T). The Nab1 signal was only slightly visible (i.e. not completely masked by the X-gal) close to the transversal limits of r5 (Fig. 1M), confirming that the stripes of reinforced expression abut these boundaries.

In conclusion, both Nab1 and Nab2 present patterns of expression within the developing hindbrain which strikingly overlap with that of Krox-20. More specifically, the activation pattern of Nab2 and the downregulation profile of Nab1 very closely resemble those of Krox-20. An interesting feature of the Nab1 pattern is the reinforcement at the boundaries of r3 and r5 at around the 8-somite stage, which does not reflect the rather uniform expression of Krox-20 in these rhombomeres, as seen by in situ hybridisation, reporter tracing and protein immunochemistry (Fig. 1C and data not shown; Schneider-Maunoury et al., 1993; Irving et al., 1996). Together these data raise the possibility that the Nab proteins could indeed modulate the transcriptional activity of Krox-20 in the hindbrain.

The overlapping patterns of expression of the three genes in the developing hindbrain raise the possibility that Nab1 and Nab2 may lie downstream of Krox-20 in a regulatory pathway.

To test this possibility, we examined Nab1 and Nab2 expression in Krox-20^−/− embryos. Although Krox-20 inactivation leads to the disappearance of r3 and r5, this phenomenon is progressive and the r3 and r5 territories can be observed until the 6-8 and 12- to 14-somite stages, respectively (Schneider-Maunoury et al., 1993, 1997; Seitanidou et al., 1997). In the case of Nab1, the late and weak activation in r3, at around the 8-somite stage, precludes the investigation of its possible regulation by Krox-20 in this rhombomere. In contrast, we could show that the stripe of Nab1 mRNA corresponding to r5 is not observed in Krox-20 homozygous mutant embryos, while it is present in both wild type and heterozygous embryos at the 8- to 10-somite stage (Fig. 2A-C). The persistence of the Nab1 signal in the heart region in Krox-20^−/− embryos attests that hybridisation occurred normally (Fig. 2C). In the case of Nab2, the highest level of expression is observed in wild-type embryos in r3 at around the 4- to 6-somite stage (Fig. 1N-S). We therefore first performed an analysis at this stage and observed that in Krox-20^−/− embryos no activation of Nab2 occurs in the prospective r3 territory (Fig. 2D-F). Examination of older embryos (10-to 12-somite) demonstrated that in Krox-20^−/− embryos Nab2 expression is also lost in r5 and in the r5-derived neural crest cells (Fig. 2G-I). These latter cells, which express Krox-20 in the wild-type situation, have been shown previously not to be affected at this stage in their survival by the mutation (Schneider-Maunoury et al., 1993; Seitanidou et al., 1997). In contrast Nab2 expression is maintained in the developing heart (data not shown).

In conclusion, these data indicate that Krox-20 function is required for expression of Nab1 at least in r5 and of Nab2 in r3, r5 and the r5-derived neural crest. Therefore, in these territories, the transcription factor Krox-20 appears to positively regulate the transcription of genes encoding putative antagonists of its own transcriptional activity.

The analysis of Nab1 and Nab2 mRNA patterns indicated that these genes are essentially coexpressed with Krox-20 during hindbrain development, raising the possibility that they may antagonise its activity. Therefore we decided to investigate whether overexpression or premature activation of these genes might affect Krox-20 transcriptional activity. Ectopic expression can be very efficient in the zebrafish (Brachydanio rerio), following injection of RNA into early embryos. On the basis of the strong conservation of the Krox-20 expression pattern and of the amino acid sequence between mouse and fish (72% identity; Oxtoby and Jowett, 1993), it is likely that the protein plays very similar or an identical function during hindbrain development in the two species. Furthermore, since the region of the Krox-20 protein known to interact with the Nab proteins (Russo et al., 1995) shows an even higher degree of similarity between the two species (94% between residues 244-277 and 226-259 in the mouse and zebrafish sequences, respectively), it is also likely that mouse Nab1 and Nab2 can bind to the fish and mouse Krox-20 proteins with similar affinities.

Therefore we decided to investigate the capacity of Nab protein to antagonise Krox-20 function during hindbrain development by injecting mouse Nab1 or Nab2 RNA into zebrafish embryos at the 2-to 4 cell stages. To evaluate the effect of this ectopic expression of Nab genes, we studied its consequences on the expression of a Krox-20 target gene, EphA4 (previously known as Sek1 in the mouse and rtk1 in zebrafish), by whole-mount in situ hybridisation at the 6-to 8-somite stage. In the zebrafish hindbrain, EphA4 was shown to be expressed in r3 and r5, a pattern very similar, although not identical, to that of the mouse (MacDonald et al., 1994; Xu et al., 1994, 1995). In the mouse, EphA4 was demonstrated to be under the direct transcriptional control of Krox-20 in r3 and r5 (Seitanidou et al., 1997; Theil et al., 1998).

The optimal amount of Nab RNA to inject was determined in a first series of experiments (see Materials and Methods). In the following experiments, the E. coli lacZ RNA was coinjected as a lineage tracer and β-galactosidase activity was detected by X-gal histochemistry. Embryos injected with lacZ RNA only or a Nab2 RNA carrying a frame-shift mutation immediately 3′ to the initiation codon (Nab2mut) were used as controls. In uninjected and control embryos, at the 6-to 8-somite stage, EphA4 transcripts were detected in the forebrain, the otic and olfactory placodes, the notochord, r3 and r5 (Fig. 3A,B and data not shown). In the hindbrain, the r3 and r5 territories of EphA4 expression appear as even transversal stripes, with straight interfaces orthogonal to the AP axis. In contrast, more than 40% of embryos injected with Nab1 (Fig. 3C-H) or Nab2 (Fig. 3I-N) present obvious alterations in the segmental EphA4 expression pattern (Table 1A), while the general organisation of the embryos did not seem to be affected except for a slight, general widening of the neural tube. The r3/r5 EphA4 stripes were irregular in shape and in some cases their extension was reduced as if parts of the expression domains had been eliminated. We also observed situations in which the level of EphA4 mRNA was reduced in a few cells, possibly directly reflecting a downregulation of EphA4 expression (Fig. 3H). The missing parts of r3 or r5 did not always coincide with areas of lacZ expression. Alterations of EphA4 expression were observed at the levels of r3 or r5 with comparable frequencies (Table 1A). Alterations affecting both r3 and r5 were less frequent (Table 1A) and were generally associated with a higher number and a more widespread distribution of X-gal-positive cells.

EphA4 expression in the forebrain or notochord was never affected by Nab1 or Nab2 injection (Fig. 3C-E, I-K and data not shown). Furthermore, cohybridisation with a Paxb probe, which labels a transversal territory at the level of the mid/hindbrain junction (Krauss et al., 1991), indicated that the Paxb-positive stripe was not affected like the r3/r5 EphA4 positive domains by the injection (Fig. 3E,H,K,N). Therefore, effects of Nab misexpression on EphA4 expression appeared to be specific and were only observed in territories where Krox-20 is known to be involved in the activation of EphA4 transcription, suggesting that the Nab proteins are indeed acting by interfering with the normal Krox-20 function and that this might include repression of transcriptional activation. This latter possibility is reinforced by the observation of cells where the level of expression of EphA4 is simply reduced (Fig. 3H, arrow).

To investigate directly whether the Nab proteins were acting solely by antagonising Krox-20, we attempted to rescue the Nab-mediated phenotype by coinjection of Nab1 or Nab2 with Krox-20 RNA. Preliminary experiments indicated that injection of Krox-20 mRNA at concentrations higher than 0.2 ng/μl resulted in a very high incidence of major developmental defects, with more than 70% of dead or grossly abnormal embryos at the stage of examination (data not shown). In addition, coinjection of 0.2 ng/μl of Krox-20 mRNA with Nab mRNAs was not effective in rescuing the EphA4 hindbrain phenotype (data not shown). Therefore, the induction of early defects during gastrulation by Krox-20 misexpression precluded our analysis of its capacity to rescue the Nab-mediated phenotype at the 6-to 8-somite stage.

The observation that loss of EphA4 expression is not always associated with lacZ expression in the Nab RNA injection experiments suggests that Nab misexpression might modify the fate of the Nab-expressing cells or even of adjacent Nab-negative cells. To investigate this possibility, we repeated the Nab RNA injection experiments and analysed the effects on the Krox-20-positive territories by whole-mount in situ hybridisation. Uninjected and control embryos (Fig. 4A,B and data not shown) presented a typical Krox-20 (Krx-20) expression pattern with two even transversal stripes corresponding to r3 and r5 and the r5-derived neural crest. In contrast, in embryos injected with Nab1 (Fig. 4C-E) or Nab2 (Fig. 4F-H), the stripes were distorted, with regions severely narrowed or even eliminated. The phenotype appeared to affect more frequently and, in a more dramatic manner, r5 than r3 (Table 1B, Fig. 4C,F), although r3 was also grossly abnormal in some cases (Fig. 4H). The affected regions were often X-gal-negative. The phenotypes and frequencies appeared similar in the cases of injection of Nab1 or Nab2 (Table 1B). In addition, the percentage of affected embryos was about twofold lower than in the case of the analysis of the consequences on EphA4 expression (Table 1B). As indicated above, the Paxb-positive stripe did not appear to be affected by the misexpression of the Nab genes (Fig. 4E,H), suggesting that the phenotype is restricted to the r3-r5 region. Finally, we also noted occasional alterations in the pattern of migration of r5-derived Krox-20-positive neural crest cells (data not shown).

In conclusion, Nab1 or Nab2 injection appears to affect not only the expression of a Krox-20 target gene, but also the shape and extension of the domains of expression of Krox-20 itself, suggesting a role for the Nab proteins either in the control of Krox-20 transcription or on the fate of Krox-20-positive cells.

Since Nab misexpression often led to modifications of the domains of Krox-20 expression, we wondered whether this was accompanied by an alteration of the r4 territory. The Hoxb1 gene is specifically expressed in this rhombomere at the 6-to 8-somite stage (Prince et al., 1998). As anticipated, whole-mount Hoxb1 in situ hybridisation on uninjected or control embryos revealed a single transversal stripe of expression corresponding to r4 (Fig. 5A,B and data not shown). In contrast, about 20% of the embryos injected with Nab1 or Nab2 RNA presented visible alterations of the Hoxb1-positive territory (Table 1C). Frequently, these alterations concerned the posterior part of the territory, with caudal extensions; however, these did not cover the entire medio-lateral width of the rhombomere, the rest of it appearing sometimes reduced in width (Fig. 5D,E,G,H). Finally, by performing double in situ hybridisations, we examined whether the variations in shapes and extent of the Krox-20- and Hoxb1-positive territories maintained the complementarity between the respective expression domains. In embryos injected with Nab1 or Nab2 RNA, Krox-20- and Hoxb1-positive domains were found perfectly adjacent, with no cell in between expressing either marker (Fig. 5F,I).

Krox-20 has previously been shown to be a key regulator of gene expression in the hindbrain. It has been implicated in the determination of segmental identity in r3 and r5 by controling the expression of several Hox genes (Sham et al., 1993; Nonchev et al., 1996a,b; Vesque et al., 1996; Seitanidou et al., 1997) and in the segregation between even and odd numbered rhombomeres through regulation of a member of the Eph family, EphA4 (Seitanidou et al., 1997; Theil et al., 1998). Furthermore, Krox-20 has been shown to be required for the maintenance of the r3 and r5 territories, which might involve a control on cell proliferation or survival (Schneider-Maunoury et al., 1993, 1997; Swiatek and Gridley, 1993). Therefore, Krox-20 appears as a central co-ordinator of several essential cellular functions in these rhombomeres. The present work demonstrates that two additional genes, Nab1 and Nab2, are downstream to Krox-20, and introduces additional complexity in the regulatory cascade. It suggests the existence of a negative feedback regulatory loop, since Nab1 and Nab2 encode proteins able to antagonise Krox-20 transcriptional activity in the hindbrain. Therefore, the Nab genes are likely to play important roles in the development of the hindbrain by controlling Krox-20 activity.

We have shown that Nab1 and Nab2 present spatial-temporal patterns of expression within the hindbrain and the neural crest cells, that closely overlap with that of Krox-20. Furthermore, targeted inactivation of Krox-20 prevents the expression of these two genes in these territories of overlap (i.e. r3, r5 and the r5-derived neural crest), but does not affect their expression in the developing heart where Krox-20 is not expressed. These data indicate that the Nab genes are downstream of Krox-20 in a regulatory pathway acting in r3, r5 and the r5-derived neural crest, and that they can also present other modes of regulation, independent of Krox-20. Since Krox-20 can act as a positive transcriptional activator in the hindbrain, these data raise the issue of whether its involvement in Nab1 and Nab2 transcription control is direct or indirect. Analysis of the cis-acting elements governing the expression of the Nab genes will be required to answer this question.

Although the patterns of expression of the Nab genes overlap with that of Krox-20 in the hindbrain, they are not identical. Three features appear different: (1) Nab1 is activated much later in r3 than Krox-20; (2) Nab2 downregulation in r5 precedes that in r3, in contrast to Krox-20; (3) the levels of expression of the Nab genes are not uniform along the AP axis in r3 and r5, showing reinforcements close to the rhombomere boundaries. This character is more pronounced in the case of Nab1 in r5 and contrasts the uniform pattern of expression of Krox-20 at the same stages in r3 and r5. These differences between Nab1, Nab2 and Krox-20 expression patterns suggest either that other factors, in addition to Krox-20, participate in the transcriptional regulation of the Nab genes, or that differential post-transcriptional mechanisms are involved. In this respect, it is interesting to note that an r3/r5-specific, Krox-20-dependent transcriptional enhancer from the chicken Hoxb2 gene also gives rise to reinforced expression close to the rhombomere boundaries in r3 (Vesque et al., 1996), supporting the view that in the case of the Nab genes this effect might also have its origin at the transcriptional level.

Nab1 or Nab2 RNA injection in zebrafish embryos results in specific modifications of the expression patterns of EphA4, an established transcriptional target of Krox-20 in r3 and r5 in the mouse, of Krox-20 itself and of Hoxb1, a marker of r4. In the case of EphA4, these modifications appear as distortions and restrictions of the expression territories and affect r3 and r5 with similar frequencies (Table 1A). This effect is specific for this region of the central nervous system since the stripe of Paxb-positive cells at the mid/hindbrain junction is not altered. Since the Nab proteins are able to antagonise Krox-20 transcriptional activity in vitro, these observations could be interpreted as a simple inhibition of EphA4 activation due to much too high levels of Nab proteins in the affected cells. This possibility is supported by the reduced level of EphA4 expression observed in some cells (Fig. 3E,H). However, this interpretation does not directly account for the fact that the cells within the territories where EphA4 expression appears lost are not always all positive for the coinjected lacZ marker. Furthermore, it does not explain why the Krox-20-positive territories are also affected. Finally, another interesting observation to take into account is the clear segregation between EphA4-positive and -negative cells in injected embryos: no ‘salt and pepper’ distribution was observed. These additional data lead us to propose two other interpretations, which are not exclusive with the first one.

1. By antagonising Krox-20 activity and possibly other Nab targets, the overexpression of Nab genes may lead to a change of identity of odd-numbered rhombomere cells into even-numbered. This change of identity would be accompanied by a loss of Krox-20 expression, an activation of even-numbered rhombomere markers (e.g. Hoxb1), and the acquisition of adhesion properties of even-numbered cells. This latter manifestation may be the sole consequence of the loss of EphA4 expression, due to the co-ordinated cessation of expression of several Eph family receptors that are normally present in odd-numbered rhombomeres (Xu et al., 1994; the different Eph receptor genes might all be under the transcriptional control of Krox-20), or involve additional genes. Expression of dominant-negative forms of these receptors has previously been shown to affect segmental restriction in the hindbrain (Xu et al., 1995). This interpretation is obviously consistent with the role of Krox-20 in controlling the expression of several Hox genes, which have themselves been implicated in the specification of cell identity along the AP axis.

2. A second interpretation, which we favour, is that inhibition of Krox-20 transcriptional activity by overexpression of Nab genes prevents not only EphA4 expression, but also cell proliferation or survival, a possibility consistent with the phenotype associated with the Krox-20 null mutation (Schneider-Maunoury et al., 1993; Swiatek and Gridley, 1993) and supported by our recent analysis of the fate of r3 and r5 cells in Krox-20^−/− embryos (O. Voiculescu, P. Charnay and S. Schneider-Maunoury, unpublished results). This interpretation would explain why the territories of expression of Krox-20 itself are also reduced in some injected embryos. The caudal expansion of the Hoxb1 territory observed in some injected embryos may result from the invasion by Hoxb1-positive cells of the adjacent territory left over by the missing r3 or r5 cells, consistent with our double in situ analysis. We attempted to study modifications of the apoptotic pattern in injected embryos, but such an analysis was precluded by the high variability of the endogenous pattern of cell death at the stages analysed (data not shown).

In conclusion, our data are consistent with the idea that an abnormally high level of Nab proteins in r3 or r5 cells may lead to the inhibition of at least three manifestations of the Krox-20 activity: transcriptional activation of specific targets, like EphA4, segregation from cells with even-numbered identity and control of cell proliferation/survival. As indicated above, the different interpretations are not exclusive and the phenotype might result from a combination of different behaviours at the cellular level. In particular, the different targets of Krox-20 might not be sensitive to the inhibition by Nab proteins at the same level. This possibility is consistent with the higher incidence of alterations of the Krox-20-positive domain corresponding to r5 following Nab misexpression, whereas EphA4-positive territories are equally affected in r3 and r5 (compare Table 1A and B).

While the specific effect of Nab misexpression in r3 and r5 argues in favor of an interference with the Krox-20 function, we cannot exclude that these two proteins might have additional targets. Indeed, we noticed that injection of Nab RNA into zebrafish embryos, besides its specific effect in the hindbrain, also leads to a general widening of the neural tube. This might reflect an interference with the convergence-extension process occurring during gastrulation. In this respect, it is interesting to note that Xegr-1, the Xenopus laevis ortholog of Krox-24/Egr-1 is expressed in the Spemann organiser region and during gastrulation in the involuting mesoderm with a pattern very similar to that of Xbra (Panitz et al., 1998). Therefore, the observed phenotype in injected embryos might reflect an interference of Nab ectopic expression with an early function of Krox-24/Egr-1.

Our injection experiments suggest that the mouse Nab proteins can interfere with the zebrafish Krox-20 function and that, at least in this assay, Nab1 and Nab2 are functionally interchangeable since no significant differences were observed in the phenotypes (Table 1). In this respect, it is interesting to note that the Nab1 protein used in these experiments was deleted of its 38 C-terminal amino acids. Therefore, this part of the protein is not necessary for the tested activity. The functional similarity of Nab1 and Nab2 is not surprising considering the homology between the two proteins, which is particularly high in the regions involved in the interactions with members of the Krox-20 family and in transcriptional repression (90% and 72% amino acid identity, respectively). Further studies will be required to determine whether the activities of the two proteins differ in other aspects.

What are the physiological functions of Nab1 and Nab2? In the hindbrain, our data are consistent with the existence of a negative feedback loop regulating Krox-20 function. Such a loop might be involved in ensuring a precisely defined level of Krox-20 activity in r3 and r5, which itself might be required to obtain an equilibrated expression of its multiple target genes. These latter genes may display different sensitivities to Krox-20 concentration, and deviation from the normal Krox-20 level may affect this sensitive balance. Since Krox-20 controls multiple aspects of cell behaviour, any unbalance may favour one function of Krox-20 versus the others (e.g. proliferation versus specification) and result in dramatic alterations in cell fate. More specifically, the reinforced expression of Nab1 close to the rhombomere boundaries could be responsible, in r3 and r5, for the reduced cell proliferation rates observed in these domains and involved in the narrowing of the neural tube width at this level (Guthrie et al., 1991). In this respect, the most evident effect of Nab misexpression in zebrafish embryos, distortion of the boundaries of the r3 and r5 territories, might reflect interference with this particular function.

Finally, Krox-20 has also been shown to be required for normal myelination by Schwann cells in the peripheral nervous system (Topilko et al., 1994). Recently, a mutation affecting a precise Krox-20 amino acid required for the Krox-24/Krox-20 interaction with the Nab proteins (Russo et al., 1995) has been observed in a recessive form of human peripheral myelanopathy (Warner et al., 1998). This suggest that the Nab proteins play an important role in the control of myelination in the peripheral nervous system and provides additional strong evidence for their involvement in regulating Krox-20 function. It will of course be of great interest to investigate the effects of this mutation on the development of r3 and r5. This should provide a direct assessment of the functions of the Nab genes in the developing hindbrain.

We are grateful to Drs D. Wilkinson and V. Prince for kindly providing the zebrafish EphA4 (rtk1) and Hoxb1 cDNAs, respectively. We thank L. Saint-Etienne for very helpful technical assistance in zebrafish injections, Dr F. Rosa for kind accommodation in his facilities and helpful discussions and members of the Charnay and Rosa laboratories for valuable discussions and critical comments on the manuscript. This work was supported by grants from the Institut National de la Santé et de la Recherche Médicale, the Ministère de l’Education Nationale, de la Recherche et de la Technologie, the European Community, the Association pour la Recherche sur le Cancer and the Ligue Nationale Française Contre le Cancer.