Cdx-Hox code controls competence for responding to Fgfs and retinoic acid in zebrafish neural tissue

The anteroposterior patterning of neural tissue is established through various inductive signals. The neuroectoderm is initially induced by signals from the dorsal organizer and subsequently receives a diffusible signal(s)from the non-axial mesoderm and endoderm (mesendoderm in zebrafish) that is responsible for establishing the anteroposterior axis in the neuroectoderm. Subsequently, the neural tissue receives more defined positional information from the mesodermal tissues and the secondary organizing centers.

The hindbrain is a segmented neural structure that contains seven or eight compartments called rhombomeres (r). The formation and anteroposterior patterning of the posterior hindbrain and anterior spinal cord is regulated by fibroblast growth factor (Fgf) and retinoic acid (RA) signaling. r4 is the first-formed rhombomere, and it functions as a secondary signaling center that expresses fgf3 and fgf8, which are required for r5 and r6 to form (Maves et al., 2002; Walshe et al., 2002; Waskiewicz et al., 2002; Wiellette and Sive, 2004). The retinaldehyde dehydrogenase 2 gene (raldh2) [the aldehyde dehydrogenase 1 family, member A2 gene (aldh1a2) -Zebrafish Information Network] encodes an enzyme that synthesizes RA from retinaldehyde, the intermediate product of vitamin A oxidation(Niederreither et al., 2000); raldh2 is expressed in early mesendodermal cells and persists in the lateral/paraxial mesoderm in zebrafish(Begemann et al., 2001; Grandel et al., 2002). Mutations in the raldh2 gene in zebrafish lead to the loss of r7 and the anterior spinal cord (Begemann et al.,2001; Grandel et al.,2002). Inhibition experiments showed that the RA signal is required for the formation of the posterior hindbrain (r5-r7) and anterior spinal cord (Begemann et al.,2001; Grandel et al.,2002); high RA activity is required for the more posterior region(anterior spinal cord) and lower RA activity is required for the more anterior one, as reported for other vertebrate species(Dupe et al., 1999; Dupe and Lumsden, 2001; Gale et al., 1999; Niederreither et al., 2000; Wendling et al., 2001; White et al., 2000). The Fgf and RA signals not only control the formation and patterning of the posterior hindbrain and anterior spinal cord but also regulate other developmental processes. The counter gradients of Fgf and RA signals control neurogenesis in the posterior spinal cord and the segmentation of the paraxial mesoderm(Diez del Corral et al., 2003; Dubrulle et al., 2001; Sawada et al., 2001). It remains to be elucidated how the different tissue responses to Fgf and RA are controlled.

caudal-related homeobox (cdx) genes are members of the ParaHox cluster, a cluster of homeobox genes closely related to the Hox cluster that function in the formation of the posterior body in vertebrate and invertebrate species (Deschamps and van Nes, 2005; Lohnes,2003). Cdx proteins directly regulate the expression of the posterior hox genes through direct binding to the cis-regulatory elements of the hox genes(Charite et al., 1998; Gaunt et al., 2004; Isaacs et al., 1998; Pownall et al., 1996; Subramanian et al., 1995). Zebrafish cdx4/kugelig mutant embryos have a reduced posterior body and reduced expression of the posterior hox genes(Davidson et al., 2003; Hammerschmidt et al., 1996). Inhibition of both Cdx1a and Cdx4 leads to loss of the hoxb7a and hoxa9a expression at the early segmentation stage and causes a more severe posterior truncation than does the inhibition of Cdx4 alone(Davidson and Zon, 2006; Shimizu et al., 2005).

Here, we show that the inhibition of Cdx1a and Cdx4 induces ectopic expression of the posterior hindbrain and anterior spinal cord markers in the posteriormost neural tissue in zebrafish. Both Fgf and RA signals are required for this ectopic expression, which can be suppressed by expression of the posterior hox genes. Our results reveal an essential role for the Cdx-Hox code in modifying tissue responsiveness to Fgf and RA signaling.

Wild-type zebrafish (Danio rerio) embryos were obtained from natural crosses of fish with the AB/India genetic background. The islet1-GFP transgenic line Tg(isl1:GFP) was used for the analysis of cranial motoneurons (Higashijima et al.,2000). The embryos were incubated at 28.5°C. The morphant embryos and inhibitor-treated embryos showed aberrant development. Developmental stages were determined as hours post fertilization (hpf).

The antisense morpholino oligonucleotides (MOs) for cdx1a, cdx4,raldh2, fgf3, fgf8, wnt3a and wnt8 and the preparation of morphant embryos were previously published(Begemann et al., 2001; Davidson et al., 2003; Shimizu et al., 2005; Wiellette and Sive, 2004). The expression patterns of krox20 (egr2b - Zebrafish Information Network), valentino, hoxb1a, hoxa2b, hoxb4a, hoxb5a, hoxb6a, hoxb7a, hoxa9a, fgf3, fgf8, raldh2 and cyp26a1 have been reported(Begemann et al., 2001; Davidson et al., 2003; Emoto et al., 2005; Furthauer et al., 1997; Grandel et al., 2002; Koshida et al., 2002; Prince et al., 1998a; Prince et al., 1998b; Reifers et al., 1998; Shimizu et al., 2005; Shinya et al., 2001; Walshe and Mason, 2003). BM Purple and FastRed (Roche) were used for whole-mount in situ hybridization. Images were taken using an AxioPlan2 microscope equipped with an AxioCam CCD camera (Zeiss).

Commissure neurons in the hindbrain were stained with the monoclonal antibody zn-5 (Trevarrow et al.,1990; provided by the Zebrafish International Resource Center) and Alexa 488-conjugated antibodies (Invitrogen/Molecular Probes). Transplantation was performed principally as described previously(Ho and Kane, 1990). Briefly,FITC-dextran (Invitrogen/Molecular Probes) was injected with cdx1aMO and cdx4MO into one-cell-stage embryos. Cells were harvested from the donor embryos and transplanted into the blastoderm of sibling recipient embryos at the sphere stage. After the embryos were fixed, the transcripts were detected by in situ hybridization using BM Purple, and FITC-dextran was detected by immunostaining with an alkaline phosphatase-conjugated anti-FITC antibody (Roche) and FastRed.

DEAB (Wako) and SU5402 (Calbiochem) were dissolved in DMSO at 100 mmol/l and 20 mmol/l, respectively. Recombinant mouse FGF8b proteins were purchased from R&D Systems. Embryos were treated with 50 μmol/l DEAB and/or 300μmol/l SU5402, or 100 ng/ml of mFGF8b in the presence of 1 μg/ml heparin in embryonic medium from the shield stage to 22 hpf.

To construct expression vectors for hoxb7a, hoxa9a, hoxb9a, hoxb1aand hoxb1b, the full open reading frames of these genes were amplified by PCR and inserted into pCS2+MT. Plasmids for VP16 fusion proteins VP-Hoxb9a and VP-Hoxb9b were constructed by inserting the PCR fragments containing hoxb9a (encoding amino acids 156-251) or hoxb9b(amino acids 162-256) into pCS2+NLS VP16AD, which contains the transcriptional activation domain of VP16 (amino acid 412-490)(Shimizu et al., 2002). Plasmids for the Engrailed fusion proteins En-Hoxb9a and En-Hoxb9b were constructed by inserting the hoxb9a or hoxb9b fragments into pCS2+En, which contains the repressor domain of Drosophila Engrailed(amino acids 1-226) (Shimizu et al.,2002). To make synthetic capped RNAs for these genes, the plasmids were linearized with NotI and transcribed with SP6 RNA polymerase.

We investigated the roles of Cdx1a and Cdx4 in the anteroposterior patterning of neural tissue. We first examined the expression of the hox genes hoxb5a, hoxb6a, hoxb7a and hoxa9a, which display region-specific expression in the spinal cord(Prince et al., 1998a), in cdx1a/cdx4MO-injected embryos (cdx1a/4 morphant embryos), at 22 hpf (corresponding to the 26-somite stage for the control embryos). In wild-type embryos, the expression domains of these hox genes extended anteriorly in the neural tube no further than the level of somite (s) 1 for hoxb5a, s2 for hoxb6a and s4 for both hoxb7a and hoxa9a (Fig. 1A,C,E,G), as reported previously(Prince et al., 1998a). In the cdx1a/4 morphant embryos, no hoxb7a or hoxa9a was detected (Fig. 1F,H); the expression domains of hoxb5a and hoxb6a extended noticeably less far posteriorly than they normally do; and a hoxb5a- and hoxb6a-negative region was observed in the posteriormost domain of the neural tube (Fig. 1B,D). Seen in this light, both Cdx1a and Cdx4 are required for the expression of the posterior hox genes, which are normally expressed in the poster spinal cord.

We next examined the expression of hindbrain markers in the cdx1a/4 morphant embryos (Fig. 2A) (Prince et al.,1998b). Unexpectedly, we observed the ectopic expression of krox20 (a marker for r3 and r5, n=30/31), hoxb1a(r4, n=21/22), hoxa2b (r2-5, n=7/10) and valentino (r5, 6, n=19/19) in the posteriormost region of the neural tube, in addition to their normal expression domains in the hindbrain region (Fig. 2B-I). The ectopic expression of valentino, krox20 and hoxb1a was detected as early as 11, 13 and 15 hpf, respectively, in the cdx1a/4morphant embryos (Fig. 2R-W). We did not detect the ectopic expression of krox20 (n=0/17), hoxb1a (n=0/19), or valentino (n=0/17) in the embryos that received cdx1aMO alone (cdx1a morphant embryos, Fig. 2O-Q). Co-staining for krox20 and hoxb4a, which labels r7 and posterior (r7-), hoxb1a, or valentino revealed that the cdx1a/4 morphant embryos showed a hoxb1a-expressing r4 identity in the posteriormost region (Fig. 2E,L), and, from posterior to anterior, a krox20⁺valentino⁺ r5 identity(Fig. 2C,I,M; note that the expression domains of krox20 and valentino overlapped in the insets of Fig. 2M), a krox20^- valentino⁺ r6 identity(Fig. 2M; also hoxb4a^- in Fig. 2K), and a hoxb4a-expressing r7 identity(Fig. 2K). The cdx1a/4morphant embryos expressed the anterior spinal cord markers hoxb5aand hoxb6a in the region anterior to the krox20-expressing region in the posterior-most region (data not shown). These results suggest that the cdx1a/4 morphant embryos display ectopic formation of the posterior hindbrain and the anterior spinal cord, and the anteroposterior polarity of the ectopic tissue is opposite to that of the normal one(Fig. 7B).

In addition to its neural expression, hoxb1a is normally expressed in the cranial mesoderm (Fig. 2D), while its expression in cdx1a/4 morphant embryos was expanded posteriorly and reached the posterior end(Fig. 2E). By contrast to ectopic hoxb1a expression in the neural tissue, however, we were unable to find any gap between the anterior and posterior mesodermal expression, suggesting that the anterior mesoderm expands instead of posterior mesoderm, rather than being ectopically induced at the posterior end. This is consistent with the posterior expansion of hoxb5a expression in the mesoderm of Cdx1a/4-defective embryos(Davidson and Zon, 2006). The data suggest that the ectopic induction of anterior tissues only took place in the neural tissue in the cdx1a/4 morphant embryos.

In an attempt to reveal whether the cdx1a/4 morphant embryos contained ectopic hindbrain neurons, we performed immunohistochemistry with zn-5, which stains neurons such as hindbrain commissure neurons and secondary motoneurons (Trevarrow et al.,1990) (Fig. 3A-H). We were able to detect the hindbrain commissure neurons with their axons in the hindbrain of control embryos (Fig. 3A-C), and in the hindbrain and the posteriormost neural tissue of the cdx1a/4 morphant embryos (Fig. 3E-H). We further examined the formation of cranial motoneurons by injecting the cdx1a/4MOs into the islet1-GFP transgenic embryos(Higashijima et al., 2000). Here, we could detect GFP expression in the cranial motoneurons, including trigeminal, facial and vagal neurons in the control embryos (V, VII and X in Fig. 3I). In the cdx1a/4 morphant embryos, GFP expression was first detected in the hindbrain and the entire posterior neural tissue at the pharyngula period. We also observed a cluster of GFP-positive neurons with their axons in the posteriormost neural tissue at 48 hpf (Fig. 3K, marked by arrowhead). Taken together, the results suggest that inhibition of Cdx1a and Cdx4 leads to posterior, mirror-image duplication of posterior hindbrain and anterior spinal cord.

We also observed the ectopic expression of krox20, in the embryos that received cdx4MO alone (cdx4 morphant embryos, Fig. 2N). However, the ectopic transcripts were scattered in the middle trunk region and were not detected in the posteriormost neural tissues (Fig. 2N). These data suggest that: (1) Cdx4 is required for repressing the posterior hindbrain fate at least partly non-redundantly; and (2) the ectopic formation of the posterior hindbrain depends on inductive signals that are affected differently in the cdx1a/4 morphant and cdx4morphant embryos.

As the inhibition of Cdx1a/4 also affects the development of the mesoderm(Davidson et al., 2003; Davidson and Zon, 2006; Shimizu et al., 2005), it was not clear whether cdx1a and cdx4 repressed the formation of the posterior hindbrain and anterior spinal cord cell-autonomously or non-cell-autonomously. To address this issue, we transplanted wild-type or Cdx1a/4-deficient blastomeres into wild-type host embryos(Fig. 4). Although cells from the wild-type donor embryos never expressed krox20 (n=0/18),ectopic expression was occasionally detected in cells from the cdx1a/4 morphant embryos, when the transplanted cells were incorporated into the neural tissue (n=20, 20% of the embryos; Fig. 4), indicating that Cdx1a and Cdx4 suppress the posterior hindbrain fate cell-autonomously. However,this ectopic expression was detected only when the Cdx1a/4-deficient cells were located in the middle trunk region of the neural tissue, and not in the anterior or the posteriormost spinal cord(Fig. 4B,C), further supporting the idea that inductive signals for hindbrain gene expression were localized differently in the wild-type and cdx1a/4 morphant embryos.

Since the normal formation and anteroposterior patterning of the hindbrain and anterior spinal cord is regulated by Fgf and RA signals, we considered the possibility that the ectopic expression of the posterior hindbrain and anterior spinal cord markers might also depend on these signals. In an effort to investigate this possibility, we first examined the expression of the fgf genes raldh2 and cyp26a1, which codes for an RA-degrading enzyme, in wild-type, cdx1a morphant, cdx4morphant and cdx1a/4 morphant embryos, at the early segmentation stage (Fig. 5A-P), as the ectopic expression is initiated at the early segmentation stages(Fig. 2R-W). We found that the expression of these genes did not significantly differ between wild-type and cdx1a morphant embryos (Fig. 5). The fgf3 and fgf8 expression domains in the anterior neuroectoderm and r4 were not affected in the cdx4 and cdx1a/4 morphant embryos, and their expression in the posterior mesoderm was retained at reduced levels(Fig. 5A-H). The expression of fgf8 in the somitic mesoderm was relatively well maintained in the cdx4 morphant but was strongly reduced in the cdx1a/4morphant embryos (Fig. 5E,G,H). The expression domain of raldh2 in the trunk region of the paraxial/lateral mesoderm shifted posteriorly in the cdx4 morphant embryos. The raldh2 expression domain shifted more posteriorly and was located closer to the posterior end in the cdx1a/4 morphant embryos, than in the cdx4 morphant embryos(Fig. 5I-L). The expression of cyp26a1 in the posteriormost region was retained in the cdx4and cdx1a/4 morphant embryos but that in the anterior spinal cord was strongly increased and shifted posteriorly in the cdx1a/4 morphant embryos (Fig. 5M-P). The expression of fgf8, raldh2 and cyp26a1in the posterior region was initiated at the gastrula period and was affected in the cdx4 and cdx1a/4 morphant embryos in a similar way to that observed at the early segmentation stage (see Fig. S1 in the supplementary material). Given that Fgf signaling activity was high in the posteriormost region and RA signaling activity was high in the middle trunk region in wild-type animals, the results from the cdx4 and cdx1a/4morphant embryos indicate that the gradients of the Fgf and RA signals in the ectopic posterior neural tissue were opposite to those in the hindbrain and anterior spinal cord (Fig. 5R-T). Our results also show that the region of high activity for Fgf and RA signaling overlapped in the posteriormost region of the cdx1a/4 morphant embryos (Fig. 5T), but in the cdx4 morphant embryos these domains overlapped in the middle trunk region, where raldh2 and fgf8are coexpressed in the somitic mesoderm(Fig. 5S). Considering these findings, we hypothesized that the overlapping Fgf and RA signaling recapitulated the signaling conditions for development of the posterior hindbrain and anterior spinal cord, thereby inducing their ectopic development in the posteriormost neural tissue in the cdx1a/4 morphant embryos and in the posterior-trunk region in the cdx4 morphant embryos. Consistent with this, we detected ectopic krox20 transcripts in the vicinity of the fgf8 expression domain in the somitic mesoderm of the cdx4 morphant embryos (Fig. 5Q).

To test our hypothesis, we conducted experiments inhibiting Fgf and/or RA signaling in the cdx1a/4 morphant embryos (Figs 6, 7; see Fig. S4 in the supplementary material). We inhibited Fgf signaling by treating the embryos with 300 μmol/l SU5402, a specific inhibitor of the FGF receptor(Mohammadi et al., 1997) or by co-injecting fgf3MO and fgf8MO(Wiellette and Sive, 2004). We inhibited RA signaling by treating the embryos with 50 μmol/l 4-(Diethylamino)-benzaldehyde (DEAB), a potent retinaldehyde dehydrogenase inhibitor (Russo, 1997) or by injecting raldh2MO (Begemann et al., 2001). Marker expression in embryos treated only with an inhibitor of Fgf or RA signaling is shown in Fig. S2 in the supplementary material (marker expression in wild-type untreated embryos is shown in Fig. 2). Inhibition of either the Fgf or RA signal did not perturb the other signaling gradient in the cdx1a/4 morphant embryos (Fig. S3 in the supplementary material). Inhibition of the Fgf signal in the cdx1a/4 morphant embryos by the fgf3/8MOs abolished the ectopic expression of hoxb1a (r4, n=26/34), krox20 (r5, n=36/36) and valentino (r5, 6, n=29/34) and reduced the normal expression of krox20 in r5, but did not inhibit the expression of hoxb4a (r7-) and hoxb5a (s1-)(Fig. 6A,E,I,M,Q; Fig. 7C). Inhibition of the Fgf signal by SU5402 abolished the ectopic expression of hoxb1a(n=14/14) and the ectopic and normal expression of krox20(n=14/14) and valentino (n=14/14) but retained the hoxb4a and hoxb5a expression(Fig. 6B,F,J,N,R; Fig. 7D). The weaker phenotypes with the fgf3/8MOs are probably due to incomplete inhibition of the Fgf3 and Fgf8 function under our experimental conditions, as the MOs did not disrupt the normal formation of r5 and r6(Maves et al., 2002; Walshe et al., 2002; Wiellette and Sive, 2004). The results suggest that the Fgf signal is required for the ectopic expression of the hindbrain r4-r6 markers, but dispensable for both normal and ectopic expression of the r7 and the anterior spinal cord markers. The inhibition of Raldh2 by its MO in the cdx1a/4 morphant embryos repressed the expression of hoxb5a (s1-, n=26/26), but did not suppress the normal or ectopic expression of krox20 (r5), hoxb1a(r4), valentino (r5, 6) or hoxb4a (r7-)(Fig. 7E; see Fig. S4 in the supplementary material). Strong inhibition of RA signaling by DEAB in the cdx1a/4 morphant embryos completely abolished the expression of hoxb4a (r7-; n=15/15) and hoxb5a (s1-, n=11/11) and strongly inhibited the expression of valentino(r5, 6), but did not significantly inhibit the expression of hoxb1a(r4) (Fig. 5C,G,K,O,S; Fig. 7F). These data indicate that high RA signaling activity is required for the normal and ectopic expression of the anterior spinal cord markers, and lower RA signaling activity is required for both normal and ectopic expression of the posteriormost hindbrain markers (r7 and probably r6).

Inhibition of both the Fgf and RA signals by SU5402 and DEAB in the cdx1a/4 morphant embryos suppressed the ectopic hoxb1aexpression (r4, n=15/15) and the normal and ectopic expression of krox20 (r5, n=18/18), valentino (r5, 6, n=9/9), hoxb4a (r7-, n=10/10) and hoxb5a(s1-, n=10/10) (Fig. 6D,H,L,P,T; Fig. 7G). This inhibition also elicited a slight expansion of the hoxb1a-expressing domain (Fig. 6H; Fig. 7G),suggesting that these embryos lost the duplicated posterior hindbrain and anterior spinal cord, but instead showed an expanded r4 domain. Although Wnt signaling is reported to be involved in patterning and neurogenesis in the hindbrain (Amoyel et al., 2005; Riley et al., 2004),inhibition of Wnt3a and Wnt8, which disrupt the posterior body formation in wild-type embryos (Shimizu et al.,2005; Thorpe et al.,2005), did not inhibit the formation of the ectopic neural tissue(Fig. 6U). Our results indicate that the ectopic formation of r4-6 requires high Fgf signaling, whereas that of anterior spinal cord and the posteriormost hindbrain requires high RA signaling, in a similar manner to the normal formation of these tissues.

To gain better insight into this issue, we investigated whether Fgf and RA were sufficient to induce the posterior hindbrain in the absence of Cdx1a and Cdx4. We treated the cdx1a/4 morphant embryos with mouse FGF8. Compared with untreated cdx1a/4 morphant embryos(Fig. 2), the FGF8-treated cdx1a/4 morphant embryos showed anterior expansion of the ectopic expression domains of krox20 and valentino, but not of hoxb1a (Fig. 6V,W,X). As the posterior mesoderm expresses raldh2 in the cdx1a/4morphant embryos (Fig. 5L), our results suggest that Fgf and RA signals induced the posterior hindbrain fate in posterior neural tissue lacking Cdx1a/4.

Cdx proteins are known to regulate the expression of the posterior hox genes (Charite et al.,1998; Gaunt et al.,2004; Isaacs et al.,1998; Pownall et al.,1996; Subramanian et al.,1995), and the expression of posterior hox genes, such as hoxb7a and hoxa9a, is absent from the neural tissues of cdx1a/4 morphant embryos (Shimizu et al., 2005) (Fig. 1F,H), suggesting that the posterior Hox proteins function downstream of Cdx1a/4 to repress the fate of the posterior hindbrain. To address this issue, we injected RNAs for hoxb7a, hoxa9a or another posterior hox gene, hoxb9a, with or without the cdx1a/cdx4MOs. The misexpression of these posterior hoxgenes in wild-type embryos suppressed the expression of krox20(n=15/15 for hoxa9a, n=19/19 for hoxb9a and n=11/15 for hoxb7a) (Fig. 8A-C, and data not shown for hoxb7a). The ectopic expression of krox20, which was observed in the cdx1a/4morphant embryos (Fig. 2C), was abolished in these embryos (n=14/15 for hoxa9a, n=15/17 for hoxb9a and n=15/17 for hoxb7a)(Fig. 8D,E, and data not shown for hoxb7a), indicating that these posterior hox genes compensated for the loss of Cdx1a and Cdx4 in repressing the posterior hindbrain fate. Misexpression of the anterior hox genes hoxb1a or hoxb1b did not suppress the ectopic krox20 expression in the cdx1a/4 morphant embryos (data not shown). These findings suggest that the posterior hox genes function downstream of Cdx1a and Cdx4 to repress the posterior hindbrain fate.

Hox proteins are reported to function as transcriptional activators or repressors (Asahara et al.,1999; Lu et al.,2003; Saleh et al.,2000; Tour et al.,2005). We investigated whether the posterior Hox proteins function as transcriptional repressors of the posterior hindbrain genes or as transcriptional activators that indirectly repress the transcription of these genes. To approach this issue, we constructed activator and repressor versions of Hoxa9a and Hoxb9a, by constructing fusion proteins with either the transcriptional repressor domain of Drosophila Engrailed (En-Hoxa9a and En-Hoxb9a) or the activation domain of the Herpes Simplex Virus transcriptional activator VP16 (VP-Hoxa9a and VP-Hoxb9a), respectively. Although the expression of En-hoxa9a or En-hoxb9a led to a reduction of posterior structures and an expansion of the anterior structure in wild-type (data not shown) and cdx1a/4 morphant embryos(Fig. 8F,G), neither of these fusion proteins inhibited the ectopic expression of krox20(n=17/17 for En-hoxa9a and n=25/27 for En-hoxb9a) (Fig. 8F,G). By contrast, the expression of VP-hoxa9a or VP-hoxb9a suppressed the expression of krox20 in its normal and ectopic positions (n=10/15 for VP-hoxa9a and n=11/15 for VP-hoxb9a)(Fig. 8H,I), as did the wild-type RNA. Our data show that the posterior Hox proteins function as transcriptional activators that indirectly repress the formation of the posterior hindbrain.

In this study, we unexpectedly found that the loss of Cdx1a and Cdx4 functions led to ectopic expression of the hindbrain genes and ectopic formation of the hindbrain neurons (Figs 2, 3). Mutations in, or the inhibition of, caudal-related genes is reported to affect the development of the posterior body structure in invertebrate and vertebrate species (Chawengsaksophak et al.,2004; Macdonald and Struhl,1986; Shinmyo et al.,2005; van den Akker et al.,2002; van Nes et al.,2006), but no role has been reported for them in inhibiting the ectopic formation of an anterior structure in the posterior body. Our data reveal a previously unrecognized role for the caudal-related genes in body patterning.

Our data show that expression of the posterior hox genes inhibits the ectopic formation of the posterior hindbrain in cdx1a/4 morphant embryos, suggesting that the posterior hox genes mediate the activity of Cdx. The possibility remains that Cdx proteins by themselves or downstream genes other than the posterior hox genes suppress the formation of the posterior hindbrain and anterior spinal cord. To address this issue,loss-of-function experiments for all the posterior Hox proteins must be performed. This is not possible currently, because there are too many posterior hox genes to be knocked down by MOs. Our findings, however,show that the posterior Hox proteins play at least some part in the Cdx-mediated inhibition of posterior hindbrain formation.

The Hox proteins interact directly - or indirectly through Pbx proteins -with transcriptional co-repressors and histone deacetylases, and they function as transcriptional repressors (Asahara et al., 1999; Saleh et al.,2000), which implies that the posterior Hox proteins might directly repress the expression of the posterior hindbrain genes. Our data show, however, that the posterior Hox proteins function as transcriptional activators to repress the posterior hindbrain genes. Therefore, the Hox proteins may activate the expression of transcriptional repressors that inhibit the expression of the posterior hindbrain genes. Alternatively,posterior Hox proteins might repress the function of transcriptional activators that induce the expression of the posterior hindbrain genes,through a protein-protein interaction or competition for the binding sites. In any case, the posterior Hox proteins indirectly repress the expression of the posterior hindbrain genes.

It is well known that the same signaling molecules are often used for different developmental processes. fgf3 and fgf8 control the anteroposterior patterning of the hindbrain and spinal cord(Maves et al., 2002; Walshe et al., 2002; Waskiewicz et al., 2002; Wiellette and Sive, 2004), the morphogenesis of the posterior body(Dubrulle et al., 2001; Sawada et al., 2001) and the development of telencephalon (Shinya et al., 2001; Walshe and Mason,2003). The RA signal controls the anteroposterior patterning of the hindbrain/spinal cord (Begemann et al.,2001; Grandel et al.,2002; Maves et al.,2002; Waskiewicz et al.,2002) in the anterior region and regulates neurogenesis and segmentation in the posterior region (Diez del Corral et al., 2003). The biological activities of the Fgf and RA signals must therefore be controlled differently in the anterior and posterior regions, by other factors. Here, we were able to demonstrate that Cdx1a and Cdx4 function to control the responsiveness of Fgf and RA signals. First, inhibition of Cdx1a/4 led to ectopic expression of the posterior hindbrain and anterior spinal cord genes, and the ectopic expression was suppressed by inhibition of Fgf and/or RA signals (Figs 6, 7). The Fgf signal is known to be required for the formation of posterior body(Griffin et al., 1998), and it is possible that the inhibition of the Fgf signal might secondarily affect ectopic formation through the repression of inductive signals from the posterior body. The incubation of the cdx1a/4 morphant embryos with the FGF8 protein, however, led to expansion of the ectopic expression but did not significantly affect the posterior body structures(Fig. 6), suggesting that the Fgf signal acts directly on the neural tissue to induce the hindbrain genes,and Cdx1a and Cdx4 repress the Fgf-dependent ectopic expression. This is consistent with the proposed direct role of Fgf and RA signals in the normal formation of the posterior hindbrain and anterior spinal cord. Our transplantation experiment also showed that Cdx1a and Cdx4 function in repressing the hindbrain genes cell-autonomously(Fig. 4). Viewed as a whole,Cdx1a and Cdx4 can be seen as controlling the formation of the posterior neural tissue by modifying the competence of these tissues to respond to the Fgf and RA signals.

It is unlikely, however, that Cdx1a and Cdx4 repress the ectopic formation of the hindbrain and anterior spinal cord through inhibiting the Fgf and RA signaling pathways. Cdx1a requires Fgf signaling to induce the expression of the posterior hox genes (Shimizu et al., 2005), and RA signaling is known to be involved in neurogenesis of the spinal cord, where Cdx genes are expressed(Diez del Corral et al.,2003). Rather than inhibiting the Fgf and RA signals then, Cdx1a and Cdx4 actually control the responsiveness to the Fgf and RA signaling. Although the molecular mechanism by which the Cdx proteins control the Fgf and RA responsiveness remains unclear, the posterior hox genes are suitable candidates to be involved in this mechanism.

In addition to the cell-autonomous role of Cdxs in repressing hindbrain gene expression in the neural tissue, Cdxs also control the sources of Fgf and RA signals in the mesodermal tissues. In our study we observed the raldh2 expression domain in the paraxial/lateral mesoderm to shift posteriorly in the cdx1a/4 morphant embryos, resulting in overlapping regions of high Fgf and RA signaling in the posteriormost neural tissue(Fig. 5). This is involved in the ectopic formation of the posterior hindbrain and anterior spinal cord. In a previous study we had reported that cdx1a and cdx4 are required for the formation of the posterior mesoderm(Shimizu et al., 2005),suggesting that Cdx1a and Cdx4 function to separate the regions of high Fgf and RA signaling though regulating the posterior body formation, thereby preventing ectopic formation in the posteriormost neural tissue. We also detected upregulation of cyp26a1 expression in the anterior spinal cord of the cdx1a/4 morphant embryos(Fig. 5). This is probably due to high RA signaling activity, as cyp26a1 is strongly responsive to RA signaling (Emoto et al.,2005). It is not yet clear, however, exactly how the RA signaling gradient is generated in the posterior hindbrain in the presence of high Cyp26a1 activity in the cdx1a/4 morphant embryos. As the RA signal is high in these embryos, it could be the case that some part of the RA may escape from the Cyp26a1-meidated degradation and be sufficient for the formation and patterning of the posterior hindbrain and anterior spinal cord. In the posteriormost neural tissue, the high RA activity probably contributes to the mirror image duplication in the cdx1a/4 morphant embryos.

How are our present findings integrated with the current model for the anteroposterior patterning of neural tissue? The anteroposterior patterning of neural tissues is initially regulated by inductive signals from the dorsal organizer and the non-axial mesendoderm, in which Wnt and Fgf signaling are believed to be involved. The subsequent regional specification is controlled by inductive signals from the secondary organizing centers and the adjacent mesoderm tissues, in which Fgf and RA signaling are involved(Moens and Prince, 2002). cdx1a and cdx4 are regulated by Wnt and Fgf signals(Shimizu et al., 2005), and they confer on the neural tissues different competences for responding to the local Fgf and RA signals. The region in which cdx1a and cdx4are not expressed takes on the posterior hindbrain/anterior spinal cord fate in response to the counter gradients of Fgf and RA signaling. In the posterior neural tissue, where cdx1a and cdx4 are expressed, cdx1a and cdx4 not only suppress the posterior hindbrain/anterior spinal cord fate, but also are required for the formation of normal posterior neural tissue (Shimizu et al., 2005).

A previous study (Bel-Vialar et al.,2002) has reported that, in chick embryos, 3′ HoxBgenes - which correspond to anterior hox genes in this study - are responsive to RA signaling, while 5′ Hox genes (posterior hox genes) are responsive to Fgf signaling. The CDX activity makes the posterior hox genes competent to respond to Fgf signaling. We previously reported that Fgf signaling is also required for the Cdx-mediated expression of hoxa9a and hoxb7a(Shimizu et al., 2005). These reports suggest that Cdx proteins cooperate with Fgf signaling in controlling the patterning and formation of the posterior spinal cord. Consistent with this, our preliminary data show that misexpression of cdx1a activates ectopic expression of hoxb9a in the hindbrain region in an Fgf-dependent manner (data not shown). From this perspective, cdx1aand cdx4 are key genes for switching the tissue competence to respond to Fgf signaling from the anterior to the posterior mode. As in chick embryos,the anterior hox gene hoxb1b is shown to be responsive to the RA signaling at the gastrula period in zebrafish(Kudoh et al., 2002). As paralog group1 (PG1) of the anterior hox genes have been shown to be involved in the formation of posterior hindbrain (r4-r6) in various species(Carpenter et al., 1993; Chisaka et al., 1992; Dolle et al., 1993; Gavalas et al., 1998; Lufkin et al., 1991; Mark et al., 1993; McClintock et al., 2001; Rossel and Capecchi, 1999; Studer et al., 1998), then this suggests that the anterior hox genes also cooperate with Fgf and RA signaling in the formation of the posterior hindbrain.

Our findings provide compelling evidence that a Cdx-Hox code controls the tissue competence to respond to the inductive signals that control the anteroposterior patterning of neural tissues. The roles of the Cdx-Hox code in neural patterning illuminate at least one mechanism by which a given inductive signal can control different processes during embryogenesis.

We thank H. Okamoto for Tg(isl1:GFP) line; A. Yamamoto for comments on the manuscript; M. Royle for editing the manuscript; H. Akiyama,Y. Wataoka, K. Bando, and A. Katsuyama for fish care and technical assistance;and the members of the Hibi laboratory for helpful discussions. This work was supported by a grant from the Ministry of Education, Science, Sports and Technology (17570185 to T.S.) and RIKEN (to M.H.).