Wnts have been shown to provide a posteriorizing signal that has to be repressed in the anterior neuroectoderm for normal anteroposterior (AP) patterning. We have previously identified a zebrafish frizzled8a (fz8a) gene expressed in the presumptive anterior neuroectoderm as well as prechordal plate at the late gastrula stage. We have investigated the role of Fz8a-mediated Wnt8b signalling in anterior brain patterning in zebrafish. We show that in zebrafish embryos: (1) Wnt signalling has at least two different stage-specific posteriorizing activities in the anterior neuroectoderm, one before mid-gastrulation and the other at late gastrulation; (2) Fz8a plays an important role in mediating anterior brain patterning; (3) Wnt8b and Fz8a can functionally interact to transmit posteriorizing signals that determine the fate of the posterior diencephalon and midbrain in late gastrula embryos; and (4) Wnt8b can suppress fz8a expression in the anterior neuroectoderm and potentially affect the level and/or range of Wnt signalling. In conclusion, we suggest that a gradient of Fz8a-mediated Wnt8b signalling may play crucial role in patterning the posterior diencephalon and midbrain regions in the late gastrula.
In vertebrates, several models have been proposed to explain anteroposterior (AP) patterning of the nervous system. One of these, which has attracted considerable attention, is the ‘activation/transformation’ model proposed by Nieuwkoop (Nieuwkoop et al., 1952; Nieuwkoop and Nigtevecht, 1954; Saxen, 1989; Sasai and De Robertis, 1997; Sive et al., 1989), based on experiments conducted in amphibian embryos. The model proposes that the initial signals of neural induction (‘activation’) also simultaneously define a forebrain state at the early gastrula stage, and later, a ‘transforming’ signal from the chordamesoderm converts some of this initially specified anterior neuroectoderm to more posterior fates (reviewed by Saxen, 1989).
Also in amphibians, β-catenin-mediated Wnt signalling has been shown to be involved in body axial patterning during two distinct developmental stages: one before and one after the midblastrula transition (MBT) (reviewed by Niehrs, 1999). Wnt signalling before MBT is required for the initiation of dorsal axis formation (McMahon and Moon, 1989; Smith and Harland, 1991; Sokol et al., 1991), whereas a later phase of Wnt signalling after MBT appears to be involved in anteroposterior patterning of the neural axis (Christian and Moon, 1993; McGrew et al., 1995). Ectopic expression of Xenopus Wnts (Saint-Jeannet et al., 1997; Fredieu et al., 1997; McGrew et al., 1995; Chang and Hemmati-Brivanlou, 1998; Popperl et al., 1997), β-catenin (McGrew et al., 1995) or the artificial activation of the Wnt pathway by LiCl treatment in amphibian embryos (Fredieu et al., 1997) after MBT all interfere with head development. These observations suggest a negative function of Wnt signalling for anterior neural fate determination. The identification of Wnt inhibitors (Cerberus, Frzb1 and Dickkopf 1) (Bouwmeester et al., 1996; Leyns et al., 1997; Wang et al., 1997; Glinka et al., 1998) as head inducers also suggests that blocking Wnt signalling in the anterior neuroectoderm is a prerequisite for patterning the brain along the AP axis. More recently, it has been suggested that repressor activity of Tcf3 encoded by the headless (hdl) locus, which binds to Wnt target genes, is an essential factor for head formation in zebrafish (Kim et al., 2000). These observations support the proposal of the ‘two-inhibitor model’: that repression of Wnt signalling is required for vertebrate head induction (Glinka et al., 1997).
Despite the prevailing idea that anterior neuroectodermal fate is determined prior to posterior neuroectodermal identity, it is not known what genes help to subdivide the anterior neuroectodermal field into various brain regions, including telencephalon, diencephalon, and eye primordium, midbrain and mid-hindbrain boundary (MHB).
Previously, we have isolated a cDNA clone for zebrafish fz8a that is expressed in the neuroectoderm as well as in the organiser (Kim et al., 1998). Zebrafish fz8a is expressed in the entire neuroectoderm at early gastrula and by late gastrula, its expression domain is gradually reduced until it becomes restricted to the regions of the anterior neuroectoderm fated to contribute to the telencephalon, diencephalon and eye primordium (Woo and Fraser, 1995). Similar to Fz8a expression in the anterior neuroectoderm, Wnt1 and Wnt8b, members of Class I Wnt family (which signal via β-catenin), also start to be expressed at the presumptive midbrain, including MHB in the late gastrula stage of zebrafish embryos (Kelly and Moon, 1995; Kelly et al., 1995). These observations raise the possibility that a gradient of Wnt signalling established within the anterior neurectoderm by late gastrulation plays contributes to patterning in the neurectoderm. To explore this possibility, we have studied the roles of Fz8a-mediated Wnt signalling in fate determination of the anterior neuroectoderm.
In this report, we show that the rostral and caudal parts of the zebrafish anterior brain are negatively and positively regulated, respectively, by Wnt signals through the Fz8a receptor. We also suggest that different thresholds of Fz8a-mediated Wnt8b signalling in the anterior neuroectoderm during late gastrulation is crucial for the proper patterning of the posterior diencephalon and midbrain
MATERIALS AND METHODS
Fish maintenance and sampling
Zebrafish were raised, maintained and staged as described in The Zebrafish Book (Westerfield, 1995). Embryos were obtained by spontaneous spawning and appropriate stages of the embryos were fixed with 4% paraformaldehyde in PBS and dechorionated with watchmakers’ forceps.
Plasmid constructions and RNA injection
The zebrafish fz8a-coding region (Kim et al., 1998) was subcloned into the BamHI and XhoI site of plasmid pCS2(+) to generate plasmid pCS2(+)-Fz8a. To construct a dominant negative form of Fz8a (Fz8a-CRDTM), pCS2(+)-Fz8a plasmid DNA was digested with DraI/StuI to remove the DNA sequences encoding transmembrane domains (second to seventh) and intracellular domain of Fz8a. The resulting larger DNA fragment was self-ligated to generate Fz8a-CRDTM that encodes the first 267 amino acids including the CRD domain with the first transmembrane domain. To construct a secreted dominant negative form of Fz8a (Fz8a-CRD), DNA sequence encoding seven transmembrane domains and an intracellular domain of Fz8a was removed from pCS2(+)-Fz8a plasmid DNA by digestion with StuI/BssHII. To construct a Fz8a-CRD- EGFP, the resulting Fz8a-CRD that encodes the first 198 amino acids including the CRD domain was fused with EGFP cDNA (Clontech) and then placed at the downstream of zebrafish hsp70-4 heat-inducible promoter (Lele et al., 1997). A form of Tcf3 (dNTCF3), pCS3+MT-ΔTCF3, that lacks the first 47 amino acids of β-catenin-binding domain of Tcf3 (Kim et al., 2000), was used to repress Wnt target genes. Capped mRNAs for injections were synthesised from linearised plasmids pCS2(+)-Fz8a, fz8a-CRDTM, pT7Ts-Wnt8b (a gift from R. T. Moon) and pCS3+MT-ΔTCF3 using an in vitro transcription kit with m7G(5′)ppp(5′)G (Boehringer Mannheim). Capped mRNAs were diluted in 0.1 M KCl solution containing 0.5% Phenol Red, 0.5-5 pg of wnt8b, 20-300 pg of fz8a-CRDTM, 5-350 pg of fz8a and 12-50 pg of dNTCF3 were injected into one-cell stage embryos. Morpholino oligonucleotides for fz8a (fz8a-MO, 5′-AGTTCAGTTCAGATGTTGCAGAG-3′) and wnt8b (wnt8b-MO, 5′-ACTTTTCTTCACCTTTCAC-3′), and four base pair mismatched control morpholinos for fz8a (4misfz8a-MO, 5′-AGTTGAGTTCTGATGATGCACAG-3′) and for wnt8b (4miswnt8b-MO, 5′-ACTTATCTACACGTTTCTC-3′), which are designed to have no predicted internal hairpins and avoiding the presence of four consecutive G nucleotides, were purchased from Gene-Tools, LLC (Corvallis, OR). fz8a-MO (10 ng/embryo) and wnt8b-MO (10 ng/embryo) were solubilised in 0.1 M KCl solution prior to use and then injected into each of the one-cell stage zebrafish embryos. Mismatched bases are underlined.
Expression of secreted dominant negative form of Fz8a under the heat-inducible promoter
For stage-specific expression of a secreted dominant negative form of Fz8a at late gastrula stage, Fz8a-CRD-EGFP DNA (50 pg/embryo) was injected into one-cell stage embryos and cultured at 26°C. Embryos were cultured to 75% epiboly stage, subjected to heat-shock at 37°C for 1 hour and then further incubated at 28.5°C. Embryos that showed higher level of EGFP expression were selected under the fluorescence microscope, further cultured until seven-somite stage, fixed and subjected to in situ hybridisation.
Lithium chloride treatment
To collect the same staged-embryos, embryos were harvested within 10 minutes after mating. For LiCl treatment, various stages of embryos (0.3 g sea salt, 2 mg Methylene Blue in 1 l water) were exposed to 0.3 M LiCl in egg water for 15-20 minutes at 28°C, washed with egg water and then allowed to develop at 28.5°C.
Whole-mount in situ hybridisation
Antisense digoxigenin-labelled RNA probes for anf, eng2, fkd5, fz8a, opl, pax2, pax6, rx1, rx3, six3 and wnt8b were produced using a DIG-RNA labelling kit (Boehringer Mannheim) according to the manufacturer’s instructions. Hybridisation and detection with an anti-digoxigenin antibody coupled to alkaline phosphatase (Vector Laboratories) were performed as described by Jowett and Lettice (Jowett and Lettice, 1994). Double in situ hybridisation was performed as described by Hauptmann and Gerster (Hauptmann and Gerster, 1994). Embryos were photographed using a Nikon coolpix900 digital camera system attached to a Zeiss Axioskop microscope.
For transplantation, one-cell stage donor embryos were injected with 1 nl of RNA solution [4 ng/μl, wnt8b RNA; 5 μg/μl, lysinated tetramethylrhodamine-dextran (Mr 70,000, Molecular probes); 4 μg/μl, lysinated biotin-dextran (Mr 2,000,000, Molecular Probes)] and further cultured to the blastula stage. Transplantation of blastomeres at midblastula stage was performed into same-staged sibling hosts. A small group of 5-10 cells from a labelled donor embryo was gently drawn up into the transplantation pipette and transplanted into the blastoderm of the host embryos. The host embryos were allowed to develop at 28.5°C in Ringer’s solution. Host embryos at 100% epiboly stage were fixed in 4% paraformaldehyde and whole-mount in situ hybridisation was performed as described above. In order to detect donor-derived cells in the host after in situ hybridisation, the host embryos were re-fixed, and then processed for biotin detection using avidin and biotinylated peroxidase complex method (Vectastain ABC kit).
Overlapped expression of fz8a and wnt8b genes in the anterior neuroectoderm of late gastrula embryos
In the anterior neuroectoderm of the 90% epiboly embryos, low levels of wnt8b transcripts are first detected at the putative midbrain region. The anterior boundary of the wnt8b expression domain initially overlaps with the posterior boundary of the fz8a expression domain, thus fz8a and wnt8b mRNA transcripts were commonly detected in the anterior region of putative midbrain territory at 90% epiboly (Fig. 1A,C). By the end of gastrulation (100% epiboly), however, fz8a expression gradually decreases at its posterior boundary, where its expression initially overlapped with wnt8b, and eventually the two adjacent expression domains of fz8a and wnt8b are separated by a gap (Fig. 1B,D). However, subsets of cells expressing fz8a are still detected in the anterior region of wnt8b expression domain (Fig. 1C,D). The proximity of fz8a and wnt8b expression domains in the future anterior brain region suggests a potential role for Fz8a-mediated Wnt8b signalling in anterior brain patterning. The fz8a-expressing domain was found to encompass the expression domains for several regional anterior neuroectoderm markers: opl for telencephalon (Grinblat et al., 1998) (Fig. 2A), anf (Kazanskaya et al., 1997) for telencephalon and diencephalon (Fig. 2B) and six3 (Kobayashi et al., 1998) for telencephalon, retina and the anterior tip of the prechordal plate (Fig. 2C). This suggested a specific role for Fz8a and Wnt8b in forebrain and midbrain patterning.
LiCl treatment at gastrula stage mimics the wnt8b misexpression
Lithium chloride (LiCl) treatment of embryos after MBT can stimulate the β-catenin-mediated class I Wnt signalling pathway (Klein and Melton, 1996; Hedgepeth et al., 1997) and mimic the loss of brain structures caused by overexpression of Xwnt8 (Fredieu et al., 1997). Therefore, we further examined whether wnt8b overexpression and LiCl treatment exert similar effects on anterior brain development. One-cell stage embryos were injected with a low dose of wnt8b mRNA (2-3 pg/embryo) and allowed to grow until 100% epiboly. Concurrently, uninjected shield stage embryos were exposed to LiCl, cultured until the completion of epiboly (9.5 hpf), at which the effects of wnt8b overexpression and LiCl treatment on the expression of anterior neuroectodermal markers (opl, anf, six3, fz8a) were analysed. A complete loss of all anterior neuroectoderm markers was seen in the anterior neuroectoderm of both wnt8b-injected (Fig. 2F-I) and LiCl-treated embryos (Fig. 2K-N). However, expression of six3 and fz8a in the anterior tip of the prechordal plate was not severely affected by either treatment (Fig. 2H,I,M,N). Nevertheless, both LiCl-treated and wnt8b-injected embryos at 25 hpf showed very similar defects in the anterior brain structure (Fig. 2J,O), suggesting that LiCl treatment at the onset of gastrulation induces defects in forebrain patterning similar to that caused by overexpression of wnt8b.
Stage-specific effect of ectopic Wnt signalling on neuroectoderm patterning
To investigate the potential role of Wnt signalling in patterning the anterior brain during late gastrulation, zebrafish embryos were exposed to LiCl at the various gastrula stages. When embryos were exposed to LiCl at stages earlier than mid-gastrula (8 hpf), a progressive loss of neural structures from rostral to caudal was observed; most embryos lost the forebrain and eye primordia (Fig. 3B,C). By contrast, LiCl treatment at later gastrula stages had a different effect and resulted in embryos with smaller eyes (Fig. 3D,E). Finally, embryos treated with LiCl at the 100% epiboly appeared almost normal in their eye and forebrain structure (Fig. 3F). These results suggest that Wnt signalling has two different stage-specific posteriorizing activities, one before mid-gastrulation and the other at late gastrulation: activation of Wnt activity in early gastrula can transform the whole forebrain territory into midbrain. By contrast, increased Wnt activity after late gastrula stage has limited posteriorizing activity, which can transform rostral fate into more caudal fate within anterior brain region, as indicated by the caudal defect of eye primordia as well as accompanying posterior diencephalon expansion (see Fig. 3M-X′). These stage-dependent effects of LiCl are analysed statistically in Table 1.
To confirm the defect of anterior structures caused by LiCl, the changes in rx1 (Chuang et al., 1999) and eng2 (Fjose et al., 1992) expression in eye primordia and midbrain, respectively, were further examined. In embryos exposed to LiCl at the early gastrula stage (shield stage), rx1 expression in the eye primordia was completely abolished, but the midbrain region marked by eng2 was remarkably expanded (Fig. 3H). By contrast, embryos treated with LiCl at the late gastrula stage (80% and 90% epiboly) showed small eyes with decreased rx1 expression, without any notable increase of eng2 expression in the midbrain (Fig. 3J,K). Finally, LiCl treatment at the end of gastrula stage (100% epiboly) did not induce any significant change of rx1 and eng2 expression, suggesting that activation of Wnt signalling at this stage has little effect on the size of the eye and midbrain (Fig. 3L). These results further support our observation that activation of Wnt signalling has stage-specific effects during gastrulation.
Posteriorizing activity of Wnt signalling at late gastrula regulates anterior brain patterning
To further characterise the stage specific effects of Wnt signalling on the patterning in the anterior neuroectoderm, we analysed the changes of pax6 and pax2 expression in embryos treated with LiCl; pax6 and pax2 have been reported to regulate forebrain and midbrain patterning (Krauss et al., 1991; Matsunaga et al., 2000; Schwarz et al., 1999; Urbanek et al., 1997; Warren and Price, 1997). At the one-somite stage, pax6 is expressed in a graded pattern in the presumptive anterior brain; it is highly expressed in the putative posterior diencephalon (arrowhead in Fig. 3M) and weakly expressed in the more rostral brain including the telencephalon, eye primordia and anterior diencephalon. This graded expression pattern of pax6 in the posterior diencephalon is also maintained at the seven-somite stage (Fig. 3S,S′). When embryos were exposed to LiCl at the early gastrula stage (shield stage), the putative MHB domain (marked by pax2 expression) was markedly expanded anteriorly, while rostral brain structures (marked by pax6) were completely lost in one-somite stage embryos (Fig. 3N). A similar expansion of the pax2 and loss of pax6 expression domains was seen at the seven-somite stage (Fig. 3T,T′). LiCl treatment at the 70% epiboly also caused the expansion of the pax2 domain with reduction of the pax6 domain (Fig. 3O,U,U′).
In contrast to the changes induced by LiCl treatment during early gastrulation, activation of Wnt signalling at the late gastrula stage (80 and 90% epiboly) induced a rostral expansion of the pax6 expression without causing much of a change in the size of the MHB domain marked by pax2 expression (Fig. 3P,Q). A similar change was seen at the seven-somite stage (Fig. 3V,W,V′,W′). However, pax2 and pax6 expressions in embryos treated with LiCl at the end of gastrulation (100% epiboly) were not changed (Fig. 3R,X,X′), thus almost comparable with those in untreated embryos (Fig. 3M,S,S′).
The anteriorly expanded pax6 expression in the diencephalon suggests that Wnt signalling activated at the late gastrula stage increases in the relative size of posterior diencephalic character within the diencephalon territory. Because the length from the anterior tip to the MHB in the same developmental stage of embryos is constant, the changes in the relative ratios for sizes of each territory of the anterior brain domains should occur reciprocally. In our study, the expansion of posterior fates (posterior diencephalon) at late gastrula resulted in a concomitant reduction of the relative size of the anterior fates (ventral diencephalon and eye primordia) (Fig. 3V,W). These results, taken together, indicate that Wnt signalling in the late gastrula particularly contributes to anterior brain patterning by imposing a posteriorizing Wnt signal on the posterior region of anterior neuroectoderm.
Ectopic expression of fz8a reduces eye size
To investigate whether fz8a in the late gastrula contributes to anterior brain patterning by posteriorizing anterior neuroectoderm from caudal to rostral, we conducted gain-of-function and loss-of-function experiments for Fz8a (Fig. 4; Table 2). Injection of fz8a mRNA into the one-cell stage embryos reduced the sizes of the eye primordia and anterior head structures (87%, n=108; Fig. 4B,F) similar to the results seen in embryos exposed to LiCl at late gastrula (Fig. 3D,E). In accordance with this, overexpression of the dominant-negative form of Fz8a (Fz8a-CRDTM, Fig. 4I) caused the expansion of eye primordia and anterior head structures (84.5%, n=116; Fig. 4C,G). These observations implicate Fz8a in contributing to Wnt signalling that is required for proper anterior brain patterning.
Recently, it was shown that hdl, a zebrafish homologue of Tcf3, is a repressor of Wnt target genes (Kim et al., 2000). To further demonstrate that repression of Wnt signals induces an expansion of anterior fates, a synthetic RNA encoding a form of hdl (dNTCF3) that lacks its β-catenin binding domain (Fig. 4I) and acts exclusively as a repressor, was injected into embryos. Of all the dNTCF3 mRNA-injected embryos, 42.6% embryos (n=54) showed enlarged eye primordia (Fig. 4D,H) very similar to those seen in embryos injected with Fz8a-CRDTM mRNA (Fig. 4C,G). This similarity supports the conclusion that the effects of the dominant negative Fz8a construct are due to reduced activation of Wnt target genes.
Fz8a synergises with Wnt8b but not with Wnt1
wnt8b and wnt1 are expressed in an overlapping pattern in the putative MHB starting from the late gastrula stage in zebrafish embryos (Kelly and Moon, 1995; Kelly et al., 1995). In order to ascertain whether Fz8a functionally interacts with Wnt8b or Wnt1 in vivo, we investigated the synergy between Fz8a and these two Wnts by examining their ability to dorsalise the embryo (ectopic dorsal axis induction or production of radialised embryos) and mimic phenotypes typically caused by much higher doses of wnt8b (Kelly et al., 1995). We injected an optimised amount of wnt8b (0.5-1 pg/embryo) and fz8a (5-20 pg/embryo) mRNAs; neither of these genes alone induces notochord expansion or secondary axis formation (Table 2). When wnt8b and fz8a mRNAs were co-injected into the one-cell embryos, a significant population of embryos exhibited dorsalised phenotypes (Table 2; Fig. 5A), including enlarged notochord (data not shown). In addition, a complete secondary axis formation was also observed in a small fraction of co-injected embryos (5.5%, n=200). The degree of dorsalisation was proportional to the amount of fz8a and wnt8b mRNAs injected, suggesting a functional interaction between Fz8a and Wnt8b (Table 2).
Ectopic expression of wnt1 (5 pg/embryo) resulted in the embryos with loss of the anterior structures (100%, n=48) similar to that observed in embryos overexpressing wnt8b. However, co-injection of wnt1 (5 pg/embryo) and fz8a (50 pg/embryo) mRNAs did not induce dorsalisation (0%, n=59, Fig. 5B).
A dominant-negative form of Fz8a can rescue the defects of anterior brain structure caused by the ectopic expression of Wnt8b
To provide additional evidence for an interaction between Fz8a and Wnt8b in generation of a posteriorizing signal during anterior brain patterning, we tested whether a dominant-negative form of Fz8a (Fz8a-CRDTM) could prevent the posteriorisation caused by wnt8b overexpression. When a low dose of wnt8b mRNA (0.5 pg/embryo) was injected, 61.1% of 25 hpf embryos (n=169) developed a small-eye phenotype, while a high dose of wnt8b mRNA (2 pg/embryo) induced an anterior brain defect together with a loss of the eye (90.6%, n=106; Table 2 and Fig. 5D). Co-injection of fz8a-CRDTM with wnt8b mRNA successfully rescued the loss of anterior brain structures caused by the overexpression of wnt8b mRNA (2 pg/embryo) in a dose-dependent manner (Table 2). As a result, 27.9% of the embryos (n=86) showed loss of eye (compare with 90.6% in wnt8b mRNA injection) when fz8a-CRDTM mRNA (20 pg/embryo) was co-injected with wnt8b mRNA; twice the amount of fz8a-CRDTM mRNA (40 pg/embryo) co-injected with wnt8b mRNA led to an almost complete recovery of the eye and forebrain (Fig. 5E,F). In accordance with these observations, fz8a-CRDTM mRNA, co-injected with wnt8b mRNA, almost restored anf (Fig. 5I), rx3 (Fig. 5M,N) and six3 (Fig. 5Q,R) expression, which disappeared in the anterior neuroectoderm of the 100% epiboly embryos overexpressing wnt8b (Fig. 5H,L,P). These results further support the suggestion that Fz8a and Wnt8b interact to provide a posteriorizing signal in the anterior neuroectoderm.
Although, as described in the previous section, no synergistic interaction was observed between Wnt1 and Fz8a, an interaction between this ligand and receptor cannot be ruled out because co-injection of Fz8a-CRDTM mRNA with wnt1 mRNA suppressed the posteriorisation caused by Wnt1 in a dose-dependent manner (data not shown), as shown in experiments with Wnt8b.
To demonstrate that Fz8a-CRDTM blocks posteriorisation defects caused by ectopic expression of Wnt1 and Wnt8b as direct consequence of its ability to interact with these ligands and not due to indirect effects on intracellular components of the Wnt signalling pathway, we showed that Fz8a-CRDTM cannot reverse posteriorisation caused by LiCl. One-cell stage embryos were injected with high dose of fz8a-CRDTM mRNA (300 pg/embryo), allowed to grow until the shield stage and exposed to LiCl for 15 minutes. In embryos injected with fz8a-CRDTM mRNA without LiCl treatment, an enlargement of anterior brain structures was observed. However, overexpression of Fz8a-CRDTM did not lead to a recovery of the eye and anterior brain structures lost after LiCl treatment (Fig. 5J). These results suggest that Fz8a-CRDTM is able to antagonise Wnt ligands, but it is unable to block the effects of LiCl, which leads to intracellular stabilisation of the Wnt signalling effector, β-catenin. In accordance with this observation, fz8a-CRDTM overexpression was unable to rescue the loss of anterior head structures in the hdl mutant (Kim et al., 2000), in which the repressor activity of Tcf3 for Wnt target genes was abolished by a null mutation (data not shown).
Knockdown of fz8a and wnt8b supports an important role for Fz8a-mediated Wnt8b signalling in anterior brain patterning
Recently, an antisense morpholino-based gene knockdown technique has enabled targeted loss-of-function studies in zebrafish embryos (Nasevicius and Ekker, 2000). To test directly whether Wnt8b and Fz8a are required for patterning the anterior brain, we generated gene-specific loss-of-function phenotypes for the two loci by injecting antisense morpholino oligonucleotides against fz8a (fz8a-MO) and wnt8b (wnt8b-MO) into the one-cell stage embryos (Fig. 6). The effects of fz8a-MO and wnt8b-MO on the anterior brain patterning were analysed at 100% epiboly with several anterior brain markers: opl for telencephalon and eye primordia, six3 for eye primordia, fkd5 for diencephalon (Varga et al., 1999), eng2 for midbrain and pax2 for MHB. The putative telencephalic and eye fate marked by opl was not affected by either gain- or loss-of-function of Wnt signal (Fig. 6A-D). By contrast, putative eye primordia, especially the posterior region of six3 expression domain, was either expanded by wnt8b-MO (Fig. 6G) or slightly reduced by LiCl treatment at the late gastrula stage (Fig. 6H). However, no changes were seen in the eye primordia of embryos injected with fz8a-MO (Fig. 6F). These results suggest that while Wnt signalling mediated by Wnt8b at late gastrulation contributes to determining the size of the of eye primordia, Fz8a is not essential to transduce its effects and other Fz receptors may be involved. While the telencephalon and eye primordia markers were not noticeably affected by knockdown of fz8a, the posterior diencephalon and midbrain, marked by lateroposterior domain of fkd5 and eng2, respectively, were significantly reduced in embryos injected with either fz8a-MO or wnt8b-MO (Fig. 6J,K,N,O). This observation indicates that Fz8a-mediated Wnt8b signalling is necessary for determining both posterior diencephalon and midbrain fates. By contrast, activation of Wnt signalling at the late gastrula by LiCl treatment expanded the posterior diencephalon (Fig. 6L), but the midbrain region was not affected by LiCl treatment (Fig. 6P). These results further suggest that putative midbrain region probably maintains relatively high activity of Fz8a-mediated Wnt8b signalling, while the posterior diencephalon region is likely exposed to the relatively low activity of Fz8a-mediated Wnt8b signalling during anterior brain patterning at late gastrula. Therefore, eng2 expression in the midbrain might not be sensitively increased by the further activation of Wnt signalling driven by LiCl treatment, although it could be sensitively decreased by the reduction of Wnt signalling caused by either fz8a-MO or wnt8b-MO injection. By contrast, fkd5 expression in the posterior diencephalon, where relatively low activity of fz8a-mediated wnt8b signalling probably persists, might be sensitive to both increased and decreased Wnt signalling. The expression level of pax2 in MHB was partially decreased by fz8a-MO (Fig. 6R), although no changes were seen in embryos either injected with wnt8b-MO or treated with LiCl at the late gastrula stage (Fig. 6S,T). This observation further suggests that pax2 expression in MHB fate is not regulated by Wnt8b signal, but is likely to be regulated by other earlier Wnt signal through Fz8a receptor during pre/early gastrulation.
To further confirm the role of Fz8a and Wnt8b in the anterior brain patterning, the effects of fz8a-MO and wnt8b-MO were also analysed at the seven-somite stage embryos (Fig. 6U-Z′). Injection of fz8a-MO reduced the posterior diencephalon domain expressing pax6, the MHB domain expressing pax2 (Fig. 6V) and midbrain domain expressing eng2 (Fig. 6Z), similar to the results seen in the 100% epiboly embryos (Fig. 6J,N,R). These results further support a role for Fz8a receptor in mediating Wnt signal during anterior brain patterning. Injection of wnt8b-MO induced slightly enlarged eye primordia, but reduced the posterior diencephalon domain expressing pax6 without any changes of pax2 expression domain in MHB of the seven-somite stage embryos (Fig. 6W). In addition, wnt8b-MO also significantly reduced eng2 expression in the midbrain (Fig. 6Z′) as seen in the 100% epiboly stage (Fig. 6O).
As expression of fz8a starts from the MBT (Kim et al., 1988), changes in anterior brain patterning observed in embryos injected with fz8a-MO might be due to loss of its early function rather than its role at the late gastrula stage. Therefore, we analysed the effect of specifically inhibiting Fz8a function at the late gastrula stage on pax2 and pax6 expression in the anterior brain. A secreted dominant-negative form of Fz8a DNA with a heat-inducible promoter (hsp-fz8a-CRD-EFGP) was injected into embryos and its expression was induced from 75% epiboly by heat-shock at 37°C for 1 hour, and then embryos were returned to 28.5°C. Embryos with higher EGFP expression were selected (data not shown), further cultured to the seven-somite stage, and changes in pax2 and pax6 expression in the anterior brain were analysed. As a result, hsp-fz8a-CRD-EFGP induced a similar phenotype to that seen in embryos injected with wnt8b-MO (Fig. 6W,X).
This result further supports our data that Fz8a-mediated Wnt8b signalling during late gastrulation plays an important role in regulation of the anterior brain patterning.
Specificity of fz8a-MO and wnt8b-MO
Previously, we have shown that co-injection of fz8a and wnt8b can synergistically produce a dorsalised phenotype at doses where neither fz8a nor wnt8b alone would have this effect (Table 2). Based on this observation, we predicted that inhibiting the function of either fz8a or wnt8b with morpholinos would prevent the dorsalised phenotype. To reveal the specificity in the effects of fz8a-MO and wnt8b-MO, either morpholino was co-injected with a mixture of fz8a and wnt8b mRNAs into yolk of the one-cell stage embryos (Fig. 7). As before, embryos co-injected with a mixture of fz8a and wnt8b mRNAs showed a dorsalised phenotype; the axial mesoderm marked by ntl expression was laterally expanded and pax2 expression domain was radialised (Fig. 7B,H). By contrast, co-injection of fz8a-MO with the mixture of fz8a and wnt8b mRNAs prevented or reduced the dorsalised phenotype (Fig. 7C,I). Similarly, the wnt8b-MO reduced the dorsalised phenotype of embryos injected with fz8a and wnt8b mRNAs (Fig. 7D,J). To further confirm the specificity of fz8a-MO and wnt8b-MO, we designed four base pair mismatched fz8a-MO (4misfz8a-MO) and wnt8b-MO (4miswnt8b-MO). Neither 4misfz8a-MO nor 4miswnt8b-MO could rescue dorsalised phenotype induced by co-injection of fz8a and wnt8b (Fig. 7E,F,K,L). In addition, we co-injected wnt8b-MO (9 ng/embryo) with wnt8b mRNA (2 pg/embryo) into the one-cell stage embryos. The co-injection successfully rescued the eye defects (eyeless or small eye phenotype) caused by wnt8b mRNA (92%, n=36; data not shown). These results, taken together, suggest fz8a-MO and wnt8b-MO specifically knockdown function of fz8a and wnt8b, respectively.
Wnt8b can suppress fz8a expression in the anterior neuroectoderm
By the 100% epiboly stage, the expression domains for fz8a and wnt8b in the anterior neuroectoderm were separated by a gap of two to three cell diameters, though expression domains of fz8a and wnt8b partially overlapped at 90% epiboly stage (Fig. 1). Furthermore, a decreasing gradient of fz8a expression from rostral to caudal was noticed in the anterior neuroectoderm by the bud stage embryos (data not shown). This observation raised the possibility that Wnt8b might suppress fz8a expression in the anterior neuroectoderm just like wingless (wg) suppresses Dfz2 expression in Drosophila (Cadigan et al., 1998). To test this hypothesis, we transplanted cells from a donor (injected with wnt8b RNA at the one cells stage) into wild-type embryos at the mid-blastula stage (Fig. 8A). In the late gastrula of wild-type host embryos, fz8a mRNA transcripts almost disappeared from the anterior neuroectoderm near the transplanted cells, indicating that Wnt8b signalling suppresses fz8a expression in the anterior neuroectoderm (Fig. 8B). This suppression of fz8a by Wnt signalling was further confirmed in hdl mutant that are characterised by exaggerated Wnt signalling. As expected, fz8a expression in the blastoderm of the 30% epiboly stage as well as the anterior neuroectoderm of bud stage hdl mutant embryos was almost completely lost (Fig. 8C,D). Consistent with this suppression of fz8a by Wnt signalling, injection of fz8a-CRDTM or dNTCF3 mRNAs resulted in an increase of fz8a expression in the anterior neuroectoderm (data not shown). Taken together, these results indicate that fz8a expression in the anterior neuroectoderm is negatively regulated by Wnt8b signalling.
During vertebrate development, head induction requires the blockage of posteriorizing Wnt activity. The three potent Wnt antagonists, Cerberus, Frzb1 and Dickkopf 1 (Dkk1), produced in endoderm and mesoderm have been reported to play critical roles in head induction (Bouwmeester et al., 1996; Leyns et al., 1997; Wang et al., 1997; Glinka et al., 1998). However, the molecular mechanism responsible for defining anterior neuroectodermal fates into more subdivided brain territories, such as telencephalon, retina, diencephalon, midbrain and MHB, during gastrulation has not been well established, although important roles of eng2/pax2 and pax6 in the determination of forebrain/midbrain boundary during the segmentation stage have been reported in mice and zebrafish (Araki and Nakamura, 1999; Matsunaga et al., 2000).
While this paper was in revision, two papers provided evidences that rostrocaudal brain characters can be specified by Wnt3A signal in a dose-dependent manner (Kieker and Niehrs, 2001; Nordstrom et al., 2002); an artificially generated Wnt gradient activity helps subdivision of the neuroectodermal field into various brain regions including the telencephalon, diencephalon, midbrain, MHB and hindbrain. However, our data show that the posteriorizing activity of Fz8a-mediated Wnt8b signalling in vivo at the late gastrula stage defines domains for the posterior diencephalon and midbrain. Our study thus provides the first evidence that Fz8a-mediated Wnt8b signalling at the late gastrula stage is necessary for proper patterning of the anterior brain.
A Wnt antagonist for head formation is required before mid-gastrula, but anterior brain patterning requires Wnt signalling at late gastrula
In previous studies, increased activity of a Wnts signal after MBT resulted in the loss of forebrain structures: increase of Wnt signal activity by injection of Wnt mRNA, LiCl or CMV promoter-driven wnt8 or wnt8b plasmids resulted in the loss of anterior neural fates, including the diencephalon (Saint-Jeannet et al., 1997; Fredieu et al., 1997; McGrew et al., 1995; Kelly et al., 1995; Chang and Hemmati-Brivanlou, 1998; Popperl et al., 1997). However, our LiCl treatment studies demonstrate that stimulation of Wnt signalling from the late gastrula onwards no longer has a loss of anterior brain structures, but instead result in transformation of anterior fate into more posterior one within anterior brain region. These observations suggest that Wnt antagonists expressed in mesendoderm (Niehrs, 1999) required for head induction might act at a relatively early stage, before cell fates are determined in the anterior neuroectoderm (which is likely to occur around the late gastrula stage).
Wnt8b-mediated repression of fz8a expression might control short- and long-range Wnt8b gradient in the anterior neuroectoderm
Our results demonstrate that Fz8a-mediated Wnt8b signalling is needed for the maintenance of proper forebrain territory, as well as specification of the midbrain. However, the presumptive forebrain, including posterior diencephalon, is far from the presumptive midbrain, where wnt8b is expressed. Therefore, additional explanation is required for the mechanism of long-range transfer of unstable Wnt8b to the forebrain region from midbrain.
In Drosophila wing development, the activity of highly unstable Wg (the Drosophila homologue of Wnt) protein alone rapidly decreases by diffusing a few cell diameters from the site of Wg expression (Couso et al., 1994; Neumann and Cohen, 1997a). However, the Wg morphogen is stabilised by binding to its receptor Dfz2, thus enabling Wg protein to migrate longer distances (at least 50 μm) from the region expressing Wg (Zecca et al., 1996; Neumann and Cohen, 1997b; Cadigan et al., 1998). In addition, Wg is defined as a localised factor that controls its own range of action by repressing the expression of a Dfz2 receptor that stabilises it during wing development (Cadigan et al., 1998). By these mechanisms, Wg can differentially regulate its short-range (higher activity of Wnt signal) and long-range (lower activity of Wnt signal) target genes during wing development.
In the anterior neuroectoderm of the zebrafish late gastrula, fz8a and wnt8b are expressed adjacent to each other. Therefore, it is expected that Wnt8b also can travel a longer distance from its site of synthesis by binding with Fz8a as seen in Drosophila. Furthermore, in our study, the transplanted cells that overexpress wnt8b suppressed fz8a expression in the anterior neuroectoderm of the recipient embryos. These observations suggest that the long-range and short-range transfer of Wnt signalling evidenced in Drosophila wing development might also be conserved in zebrafish anterior brain patterning, although the in vivo data that reveal the diffusion of Wnt8b through binding with Fz8a in zebrafish is not available yet. We therefore suggest as a possible mechanism that the short-range and long-range transfer of Fz8a-mediated Wnt signalling can define the midbrain and forebrain territory, respectively.
Expression of pax6 in the posterior diencephalon might be a consequence of anf repression caused by Fz8a-mediated Wnt8b signalling
In Xenopus, pax6 transcripts distributed at the anterior neuroectoderm are known to be under the inhibitory control of anf. The expression of anf is initially detected in the entire presumptive diencephalon and then eventually disappears except in the future ventral diencephalic cells of the midline at the late gastrula stage (Ermakova et al., 1999; Kazanskaya et al., 1997). Our data show that activation of Wnt signalling by LiCl treatment or wnt8b overexpression completely abolishes anf expression in the anterior neuroectoderm, but expands the posterior diencephalon region marked by pax6 expression. Therefore, this observation suggests that pax6 expression in the posterior diencephalon might be indirectly activated by the long-range action of Fz8a-mediated Wnt8b signalling. Instead, the suppression of anf expression accomplished by long-range action of Fz8a-mediated Wnt8b signalling probably contributes to the upregulation of pax6 expression in the posterior diencephalon.
Possible involvement of multiple Fz and Wnt interactions in anterior brain patterning
In morpholino experiment, eng2 expression in the midbrain was significantly reduced by either fz8a-MO or wnt8b-MO injection, thus indicating that eng2 expression requires both Fz8a and Wnt8b. By contrast, putative eye primordia, marked by six3, were expanded by wnt8b-MO, but no changes were induced in six3 expression domain by fz8a-MO. These results suggest that other Fz receptors probably mediate Wnt8b signal during fate determination of the eye primordia, although other fz genes expressed in the presumptive eye primordia at late gastrulation have not been identified.
However, pax2 expression in MHB, which was not reduced by wnt8b-MO but downregulated by fz8a-MO, raised a possibility that other earlier Wnt signals occur at pre/early gastrulation might regulate pax2 expression in the putative MHB through Fz8a receptor. Recently, two Wnt8 isoforms produced by a bicistronic transcript were reported to be required for proper initiation of pax2 expression in the presumptive MHB (Lekven et al., 2001). Thus, a possibility of fz8a and Wnt8 interaction in regulating pax2 expression in MHB still remains to be elucidated. Alternatively, we can not rule out the possible contribution of Fz7 and Wnt8 interaction for regulating pax2 expression in MHB, as fz7 expression was also reported in the presumptive MHB of zebrafish late gastrula embryos (El-Messaoudi and Renucci, 2001). Therefore, further studies are needed to address the detailed functions of multiple Fz and Wnt interactions in regulating fate determination of the eye primordia and MHB.
A gradient of Fz8a-mediated Wnt8b signal activity patterns the posterior diencephalon and midbrain
Our data suggest that LiCl treatment at the late gastrula stage (90% epiboly) acts as an artificial Wnt signal activator, thus significantly increasing fkd5 and pax6 expression in the posterior diencephalon. However, eng2 expression was not dramatically increased, although Wnt signalling was highly activated by LiCl treatment at the late gastrula stage. Nevertheless, injections of wnt8b-MO and fz8a-MO morpholinos, which might cause partial reductions of Wnt8b and Fz8a, reduced eng2 expression in the midbrain more sharply compared with decreased expressions of fkd5 and pax6 in the posterior diencephalon. These results indicate that eng2 in the midbrain is highly sensitive to a decrease of Wnt8b signal activity but less sensitive to an excess of Wnt signal, whereas fkd5 and pax6 in the posterior diencephalon is highly sensitive to an excess of Wnt signal but less sensitive to a decrease of Wnt8b signal. These observations indicate that patterning of the midbrain needs a higher threshold of Wnt8b activity, while that of the posterior diencephalon may require relatively lower Wnt8b thresholds.
To explain a gradient of Fz8a-mediated Wnt8b signal activity required for the proper patterning of the anterior neuroectoderm (posterior diencephalon and midbrain), we propose a model that can generate a sharp gradient of Fz8a-mediated Wnt8b signalling activity, with a peak at the midbrain (Fig. 9). First, at the 90% epiboly stage, two adjust expression domains for fz8a and wnt8b are partially overlapped in the putative midbrain. At the same time, a small amount of Wnt8b, possibly stabilised by binding to Fz8a, might further diffuse towards the presumptive posterior diencephalon from midbrain. Therefore, low Wnt8b signal activity and high Wnt8b signal activity might be imposed on the posterior diencephalon and midbrain region, respectively. Subsequently, at late gastrula stage, overlapping two expression domains are separated by the repression of fz8a expression caused by Wnt8b with generating a decreasing gradient of Fz8a receptor towards the caudal anterior neuroectoderm, thus a gradient of Fz8a-mediated Wnt8b signal activity becomes sharper at late gastrula stage. Consequently, a gradient of pax6 expression in the diencephalon from posterior to anterior can be established by low level of Wnt8b activity, while eng2 expression in the midbrain can be regulated by high level of Wnt8b activity. This hypothesis that pax6 and eng2 expression requires lower and higher level of Wnt signalling, respectively, was recently evidenced in chick gastrula (Nordstrom et al., 2002) during revision of this paper.
We conclude that the Wnt signal activity accomplished by the functional interaction between Wnt8b and Fz8a within the anterior neuroectoderm is required for zebrafish anterior brain patterning. In addition, we also propose that a gradient of Fz8a-mediated Wnt8b signalling with a peak in the midbrain might play an important role in the fate determination of the anterior brain.
We thank Drs Randal T. Moon, Makoto Kobayashi and Monte Westerfield for wnt8b, six3 and fkd5 plasmid constructs; Drs Claudio Stern, Lilianna Solnica-Krezel and Bruce Appel for helpful comments on the manuscript. We are also grateful to the members of our laboratory for their technical support. This work was supported by Korea Research Foundation Grant (KRF-2000-015-DS0036).