A temperature-sensitive mutation in the nodal-related gene cyclops reveals that the floor plate is induced during gastrulation in zebrafish

The floor plate is a specialized group of cells in the ventral neural tube of vertebrates and plays a crucial role in patterning the central nervous system. The specification of motoneurones, interneurones and differentiation of oligodendrocytes has been shown to require a functional floor plate, which secretes the glycoprotein, Sonic hedgehog (Shh)(Ericson et al., 1996; Orentas and Miller, 1996; Poncet et al., 1996; Pringle et al., 1996; Briscoe et al., 2001; Lewis and Eisen, 2001). The floor plate also provides guidance cues that are essential for the axonal outgrowth of many neurones (Colamarino and Tessier-Levigne, 1995; Matise et al., 1999).

Studies in several vertebrates (Placzek et al., 1990; Lawson and Pedersen, 1992; Le Douarin et al., 1998) that investigated the development of the floor plate led to several models for its induction. One model proposes that a signalling cascade mediated by the Shh protein secreted from the notochord induces the floor plate in the overlying neural tube(Placzek et al., 2000), and mutations in the mouse Shh gene indeed result in floor plate deficiencies (Chiang et al.,1996). Moreover, grafting experiments in the chick have indicated that floor-plate markers can be induced in ectopic locations of the neural tube by signals from the notochord, or the floor plate itself, or by the expression of SHH protein in ectopic locations of the neural tube(van Straaten et al., 1985; van Straaten et al., 1988; Placzek et al., 1990; Placzek et al., 1991; Yamada et al., 1991; Marti et al., 1995; Roelink et al., 1995; Ericson et al., 1996).

However, recent experiments in the chick, as well as analyses of zebrafish mutants suggest that the floor plate may be induced independent of notochord specification (Halpern et al.,1997; Le Douarin et al.,1998; Le Douarin and Halpern,2000; Charrier et al.,2002). For example, zygotic mutations in zebrafish cyclops (cyc), which encodes a Nodal-related secreted signalling factor (Hatta et al.,1991; Rebagliati et al,1998; Sampath et al.,1998), and one eyed pinhead (oep), which encodes an essential co-factor for Nodal signalling(Gritsman et al., 1999),result in the lack of a floor plate, in spite of the presence of a morphologically normal notochord expressing shh(Strahle et al.,1997; Schier et al.,1997). On the other hand, mutations in the flh and ntl genes, which are required for the formation of the notochord,result in embryos that exhibit a patchy or wider floor plate, respectively(Halpern et al., 1997). Furthermore, medial floor plate cells are not abolished by mutations in the zebrafish shh gene (Schauerte et al., 1998) or its receptor, smoothened(Chen et al., 2001; Varga et al., 2001), or by abrogation of Hedgehog signalling using antisense knockdown with morpholino-modified oligomers (Etheridge et al., 2001). Therefore, an alternate model proposes that the floor plate is induced in the organizer-derived midline precursor cells(Le Douarin and Halpern, 2000; Charrier et al., 2002). As the precursor cells give rise to both the notochord and the floor plate, this model predicts that floor-plate induction takes place early during gastrulation. A key unresolved issue in the models pertains to the timing of floor plate induction.

We have isolated a temperature-sensitive mutation in the zebrafish cyc locus. In contrast to null mutations in cyc, embryos homozygous for the cyc^sg1 mutation manifest variable cyc phenotypes at 22°C. Mutant embryos exhibit variably fused eyes, ventral curvature and patchy to complete floor plate, with motoneurones that may or may not be at their normal positions. At 28.5°C, cyc^sg1 mutant embryos are indistinguishable from cyc-null mutant embryos, with fused eyes, lack of medial floor plate cells and ventral curvature. Using this allele in temperature shift-up and shift-down experiments, we show that Cyc function is essential at gastrulation to induce the floor plate in zebrafish. By modulating Nodal signalling levels in mutants, and by overexpressing cyc in wild-type embryos, we show that high levels of Cyc signalling are required for induction of the floor plate. Furthermore, we show that continuous and high levels of Cyc signalling during gastrulation are essential for formation of a complete ventral neural tube. These results show that the floor plate inducing activity of Cyc is essential during gastrulation, and that it is required at multiple steps of the floor plate induction pathway for the development of a complete ventral neural tube.

Adult fish were maintained and reared as described in Westerfield(Westerfield, 1994). The mutant lines used in this analysis are cyc^m294/+(Schier et al., 1996), cyc^tf219/+ (Brand et al., 1996), cyc^b16/+(Hatta et al., 1991), sqt^cz35/+(Heisenberg and Nusslein-Volhard,1997; Feldman et al.,1998) and cyc^sg1/+ (this work). Double mutants were generated by crossing heterozygous cyc^m294/+ to sqt^cz35/+ and cyc^sg1/+ to sqt^cz35/+. The zebrafish wild-type strain AB(Johnson et al., 1994) was used for out-crosses, and the polymorphic WIK strain was used for mapping(Rauch et al., 1997). Embryos were collected after natural matings and staged according to Kimmel et al.(Kimmel et al., 1995).

Adult zebrafish males of the AB strain were mutagenized with the chemical ethyl nitrosourea as described (Riley and Grunwald, 1995), and mosaic F1 progeny were screened for non-complementation with cyc^tf219/+ fish. Putative mutant fish were subsequently tested with other known cyc alleles as well as other mutants affecting the Nodal signalling pathway. Identified heterozygous fish were out-crossed to AB or to WIK fish. Embryos from identified heterozygous fish in the next generation were split into two groups at the one-cell stage, allowed to develop at 28.5°C or at 22°C until prim-5 stage, and analysed for cyc phenotypes of fused eyes, ventral curvature and the floor plate. Mapping was carried out using PCR on a AB/WIK mutant panel (n=80 haploid and 1200 diploid embryos) using primers flanking a CA repeat in the 3′-untranslated region of cyc(Sampath et al., 1998). For identifying the mutation, genomic DNA was isolated from single cyc^sg1 mutant embryos at 24 hours post-fertilization (hpf)and used as templates for sequencing. In addition, DNA and RNA were extracted from individual cyc^sg1 homozygous embryos at shield stages using TRIZOL reagent (Gibco, BRL), and single embryos were genotyped using the cyc CA repeat marker. RT-PCR was carried out using pooled RNA from identified cyc^sg1mutant embryos and the nucleotide sequence was determined.

The genotype of sqt^cz35 mutant embryos and cyc^m294 mutant embryos was determined as described(Feldman et al., 1998; Sampath et al., 1998). For determining the genotype of cyc^sg1 homozygous mutant embryos, genomic DNA was isolated from single embryos (live or after analysis of in situ hybridization patterns), and PCR was performed with the primers 5′-AACAGGAGCTACCGAGCAGGC-3′ and 5′-ACTGGCCCCGTCCTGCTGCT-3′. The PCR products were digested with the restriction enzyme PvuII (New England Biolabs), and analysed by agarose gel electrophoresis.

Embryos obtained from matings of cyc^sg1/+ fish were split into two groups at the one-cell stage, and raised at 22°C and 28.5°C, respectively. At regular intervals from 50% epiboly to 10-somite stages, embryos at 22°C were shifted to 28.5°C. Conversely, embryos raised at 28.5°C were shifted to 22°C at the same intervals. For temperature pulse experiments, embryos were incubated at either 22°C or 28.5°C with a brief shift-up or shift-down period during mid-gastrulation. Embryos were fixed at 100% epiboly or prim-5 stages for in situ hybridization with various markers.

The cyc^sg1 mutation was introduced into pCS2cyc⁺ (Sampath et al., 1998) by PCR-based mutagenesis. The Flag epitope-tagged pCS2cyc^+FLAG and pCS2cyc^sg1FLAGconstructs were generated by PCR-based methods and their nucleotide sequence was confirmed. In both constructs, the Flag epitope was fused in frame after the cleavage site, between Val 385 and Arg 386 in Cyc.

The plasmids pCS2cyc⁺,pCS2cyc^+FLAG, pCS2cyc^m294,pCS2cyc^sg1, and pCS2cyc^sg1FLAG were linearized with NotI, and sense strand capped mRNA was synthesized with SP6 RNA polymerase using the mMESSAGE mMACHINE system (Ambion). In vitro synthesised RNA was injected into one- to four-cell stage wild-type embryos. Animal caps were dissected at late blastula stages and cultured as described(Sagerstrom et al., 1996; Dheen et al., 1999) until sibling stage 80% epiboly at 22°C or 28.5°C. Animal cap explants and embryos at various stages were fixed for antibody staining or in situ hybridization with various markers.

Cos-7 cells were cultured at 37°C in DMEM (Gibco-BRL) containing 10%foetal bovine serum (Sigma), 10 U/ml penicillin and 10 mg/ml streptomycin sulphate (Sigma). The plamids pCS2cyc^+FLAG and pCS2cyc^sg1FLAG were transfected into Cos-7 cells using the Superfect transfection reagent (Qiagen). Cells were fixed after 24 hours in 4%paraformaldehyde and processed for detection of the Flag epitope.

Whole-mount in situ hybridization was performed as described(Sampath et al., 1998) on embryos fixed at 100% epiboly (10 hpf at 28.5°C or 20 hpf at 22°C),six-somite (12 hpf at 28.5°C or 24 hpf at 22°C) or prim-5 stages (24 hpf at 28.5°C or 48 hpf at 22°C). The following plasmids were linearized and antisense probes were synthesized by in vitro transcription:pBSshh (EcoRI, T7) (Krauss et al., 1993), pBStwhh (PstI, T7)(Ekker et al., 1995),pBSislet2 (EcoRI, T7) (Appel et al., 1995), pBShgg1 (XbaI, T7)(Thisse et al., 1994), pBSgsc(EcoRI, T7) (Stachel et al.,1993), pBSflh (EcoRI, T7)(Talbot et al., 1995). Single-or double-colour in situ hybridization was performed as described(Sampath et al., 1998). For digoxigenin-labelled probes, BM purple substrate (Roche) was used; for fluorescein-labelled probes, fast red (Roche) or 4-iodonitrotetrazolium violet(Molecular Probes) were used. For cryosections, whole-mount bicolour in situ hybridized embryos were embedded in 1.5% agarose:30% sucrose blocks. The blocks were frozen and sections were obtained on a Leica CM 1900 cryomicrotome at 16 μm intervals.

Embryos and animal caps were fixed in 4% paraformaldehyle at 4°C overnight. Embryos were incubated with a monoclonal antibody raised against the zn-5 epitope (Trevarrow et al.,1990), and colour was developed using the ABC kit (Pierce) with the substrate diaminobenzidine (Sigma). Animal caps and Cos-7 cells were incubated with an anti-Flag polyclonal antibody (Sigma), and detected with an anti-rabbit secondary antibody conjugated with Alexa 568 (Molecular Probes). Optical sections were obtained at 0.5 μm intervals on a Zeiss Axiovert 200M microscope, and images were processed using the Zeiss LSM image browser software.

In a mutagenesis screen for new mutations in the cyc locus, we identified one mutant, sg1, which did not complement the cyc^m294, cyc^tf219 and cyc^b16 mutations(Hatta et al., 1991; Brand et al., 1996; Schier et al., 1996), and mapped to the cyc locus (Talbot et al., 1998; Sampath et al.,1998). In contrast to null mutations in the cyc locus,embryos homozygous for the cyc^sg1 mutation exhibit variably fused eyes (Fig. 1A-E)and variable degrees of ventral curvature(Fig. 1F-J) at 22°C, the permissive temperature for cyc^sg1 mutants. The mutation is incompletely penetrant at 22°C (Table 1), with only 2-3% of homozygous mutants exhibiting the `classic'cyc phenotype (Hatta et al.,1991). In addition, only 50% (n=25) of the homozygous mutant embryos at 22°C manifest the cyc phenotypes(Table 1). The mutation is fully penetrant at 28.5°C (Table 1), the restrictive temperature for cyc^sg1mutant embryos.

Nucleotide sequence analysis revealed an A to T transversion at position 853 of the cyc-coding sequence, which results in a premature stop codon (Fig. 2A,C). The mutation also introduces a site for the restriction enzyme PvuII in the cyc cDNA (Fig. 2B). To confirm if the A-T transversion causes the temperature-sensitive (ts)phenotype, synthetic mRNA encoding Cyc^sg1 was generated and microinjected into wild-type embryos. Microinjection of in vitro synthesized cyc^sg1 mutant mRNA into wild-type embryos resulted in cyc overexpression phenotypes(Rebagliati et al., 1998; Sampath et al., 1998) at 22°C, shown by the expansion (Table 2 and Fig. 2E) or duplication (Fig. 2F,G) of shh-expression domains. By contrast, at 28.5°C, similar to embryos injected with cyc^m294 mutant RNA, embryos injected with cyc^sg1 mRNA were indistinguishable from control embryos (Table 2 and Fig. 2D), confirming that the A-T transversion in cyc^sg1 is responsible for the temperature-sensitive phenotype.

To detect wild-type and mutant proteins, synthetic mRNA encoding Cyc^+FLAG or Cyc^sg1FLAG was injected into wild-type embryos, and animal cap explants dissected from the injected embryos were processed for detection of the Flag epitope. Animal caps loaded with either wild-type or mutant Flag-tagged cyc RNA show localization of the wild-type (not shown) and mutant protein at 22°C(Fig. 2K). By contrast, at 28.5°C, only the wild-type protein is detected(Fig. 2I), whereas Cyc^sg1FLAG is not detected(Fig.2L), similar to control explants (Fig. 2H). In situ hybridization with gsc in the animal cap explants showed expression in explants from cyc^sg1 injected embryos at 22°C but not at 28.5°C (data not shown). Cos-7 cells transfected with the plasmids pCS2cyc^+FLAG and pCS2cyc^sg1FLAG show localization of the Cyc^+FLAG protein at 37°C(Fig. 2J), but not the Cyc^sg1FLAG protein (Fig. 2M).

Expression of cyc transcripts in gastrula stage cyc^sg1 mutant embryos at 22°C is similar to that seen in wild type embryos (Fig. 3A,B) or reduced (Fig. 3D,E). At 28.5°C, similar to ENU-induced null mutations in the cyc locus (Rebagliati et al.,1998; Sampath et al.,1998), cyc transcripts in cyc^sg1mutants are reduced by mid-gastrula stages(Fig. 3C), and are not detected by the end of gastrulation (Fig. 3F). Analysis of expression of goosecoid (gsc),a marker of the prechordal plate mesendoderm(Thisse et al., 1994), which is reduced in cyc null mutants, reveals variably reduced(Fig. 3I,J) to normal(Fig. 3G) prechordal plate mesendoderm in cyc^sg1 mutants at 22°C, compared with those at 28.5°C (Fig. 3H).

In situ hybridization with the early marker of the floor plate, tiggy-winkle hedgehog (twhh), shows patterns that are either comparable with wild-type embryos or reduced in cyc^sg1mutants at 22°C (Fig. 4A-C). At 28.5°C, similar to null mutations in cyc,twhh expression is not detected in the midline(Fig. 4D) of cyc^sg1 mutants. Strikingly, expression of shh in prim-5 stage cyc^sg1 mutants reveals a range of patchy to complete floor plates at 22°C (Fig. 4E-G), whereas at 28.5°C, similar to null mutations in cyc (Hatta et al.,1991), there is a complete lack of medial floor plate cells(Fig. 4H). The cyc^sg1 embryos with patchy shh expression reveal the intermittent presence of floor plate cells throughout the length of the embryo (Fig. 4E), with gaps in the anterior of the embryo but a fairly complete floor plate in the trunk(Fig. 4F), or with normal shh expression in the anterior and gaps in the trunk(Fig. 4G).

Primary motoneurones revealed by islet2 expression also show a range of phenotypes in cyc^sg1 mutants at 22°C(Fig. 5B-D) compared with wild-type siblings (Fig. 5A). The position of primary motoneurones may be similar to that seen in wild-type embryos, either with a shh-positive medial floor plate(Fig. 5B), or even in the absence of a shh-positive floor plate(Fig. 5C). Alternatively, in the absence of a medial floor plate, similar to null mutations in cyc(Beattie et al., 1997), the primary motoneurones collapse in the midline(Fig. 5D). Immunostaining using antibodies raised against the zn5 epitope to detect retinal ganglion cell axons in the anterior of the embryo (Fig. 5E-G), and secondary motoneurones in the trunk(Fig. 5H-J) show patterns comparable with wild-type embryos (Fig. 5E,F,H,I) in cyc^sg1 mutants raised at 22°C, in contrast to those raised at 28.5°C(Fig. 5G,J).

A fundamental question regarding the induction of the floor plate is when this event takes place (Dodd et al.,1998; Le Douarin and Halpern,2000; Placzek et al.,2000). Because embryos homozygous for the temperature-sensitive mutation, cyc^sg1, have medial floor plate cells at the permissive temperature and lack them at the restrictive temperature, we used the cyc^sg1 allele to address this question. Embryos collected from matings of cyc^sg1/+ heterozygous fish were incubated at 22°C and shifted to 28.5°C (22-28 shift) to abrogate Cyclops function at various stages of gastrulation and segmentation. In these temperature shifts, the floor plate should develop until the point in embryogenesis when Cyclops is essential for this event. Interestingly, cyc^sg1 mutant embryos shifted to 28.5°C at early gastrula stages did not show any medial floor plate cells (n=42) as assessed by expression of the early markers of the floor plate, twhhand shh (Fig. 6A), as well as the markers of differentiated floor plate cells, f-spondin2(spon1b – Zebrafish Information Network) and col2a1(data not shown). Furthermore, 22-28 shifts performed after midgastrula stages(75% to 80% epiboly) resulted in the presence of medial floor plate cells in∼75% of mutant embryos (n=110; genotype confirmed by PCR)(Fig. 6A,C,D).

Conversely, embryos were incubated at 28.5°C, and shifted down to 22°C (28-22 shift) at various stages to determine if medial floor plate cells could be rescued in these embryos. When 28-22 shifts were performed at early to mid-gastrula stages, more than 80% of mutant embryos exhibited medial floor plate cells (n=94; genotype confirmed by PCR)(Fig. 6B). The proportion of mutant embryos with shh-positive floor-plate cells decreases as shifts were performed later in gastrulation(Fig. 6B). A 28-22 shift-down after 80% epiboly failed to induce any medial floor plate cells(n=188), similar to embryos incubated at 28.5°C alone until 24 hpf (n=93). These results indicate that Cyclops is required for inducing the floor plate between 70 and 80% epiboly.

Interestingly, we found that embryos that were shifted down to the permissive temperature at 50% or 60% epiboly showed a complete floor plate and ventral neural tube, throughout the entire length of the neuraxis(Fig. 6E), whereas shift-down at later stages resulted in rescue of patches of floor-plate cells(Fig. 6F). The patches were distributed throughout the length of the neuraxis, regardless of the stage at which the embryos were shifted. In addition, embryos from 28-22 shift-down experiments at mid-gastrula stages typically showed longer stretches of cells expressing floor plate markers than embryos from 22-28 shifts(Fig. 6C,F). These results suggest that Cyclops function may be required first for induction of the floor plate from its precursors early during gastrulation, and, subsequently, for the complete development of the floor plate along the entire length of the neuraxis.

To confirm the above observations that the floor-plate inducing activity of Cyclops is essential during mid-gastrulation, we addressed whether a transient pulse at the permissive temperature during mid-gastrulation was sufficient for induction of the floor plate, or, conversely, if a brief incubation at the restrictive temperature could abrogate floor-plate fates in cyc^sg1 mutant embryos. When embryos were incubated at 28.5°C throughout gastrulation and segmentation, with a brief shift-down period at 22°C during mid-gastrulation, 27/29 mutant embryos (93%) that were incubated at 22°C between 70 and 80% epiboly showed rescue of medial floor plate cells, determined by the presence of shh- or twhh-expressing cells (Fig. 7A,D,H). However, the extent of rescue as determined by shh expression at prim-5 stage was usually patches of three or four cells, throughout the length of the neuraxis(Fig. 7H). If the pulse of 22°C was given between 60 and 90% epiboly, all mutant embryos(n=17) showed floor-plate cells(Fig. 7A,B,F). Furthermore, the embryos showed longer stretches of cells expressing shh(Fig. 7F), with a complete floor plate in the trunk in all mutant embryos. Conversely, only 9/30 (30%) of the mutant embryos that were raised at 22°C and incubated at 28.5°C between 70 and 80% epiboly showed floor plate cells(Fig. 7A,E,I). The number of cells expressing twhh or shh were also fewer, with large gaps between patches of floor plate cells(Fig. 7E,I). If the pulse of 28.5°C was given between 60 and 90% epiboly, all mutant embryos(n=24) lacked floor plate cells(Fig. 7C,G). Thus, although a transient pulse of Cyclops signalling between 70-80% epiboly is sufficient to initiate floor plate fates, continuous Cyclops signalling is required between 60 and 90% epiboly for a complete floor plate.

Ectopic overexpression experiments in zebrafish and Xenopus have suggested that different levels of Nodal signalling pattern the organizer,with high levels required for specification of the anterior organizer fates(prechordal plate mesendoderm), and lower levels for posterior (notochord)fates (Jones et al., 1995; Gritsman et al., 2000). To determine if the level of Nodal signalling is important for induction of the floor plate, we overexpressed wild-type cyc mRNA in wild-type embryos. Although low doses of cyc RNA [or squint(sqt) RNA, data not shown] are sufficient to induce ectopic domains of the marker of the notochord, flh(Fig. 8B)(Gritsman et al., 2000)(n=147), expansion or duplication of the gsc and twhh expression domains (n=124 and 273, respectively)requires higher doses of cyc mRNA(Fig. 8G,H,K,L). Thus, high levels of Cyclops signalling are required for specification of floor plate fates.

This was supported by analysis of compound mutants of cyc^sg1 generated with a null allele of cyc,cyc^m294 and with the other zebrafish nodal-related mutant, sqt^cz35. Although all mutant embryos (38/186 cyc; genotype confirmed by PCR) from cyc^sg1/+matings at 22°C had a complete floor plate(Fig. 9A), 4/79 embryos from matings of cyc^m294/+ with cyc^sg1/+ at 22°C had no shh expression in the floor plate(Fig. 9B), and 18/79 embryos were shh+, but showed several gaps in shh expression in the floor-plate domain (genotype cyc; confirmed by PCR).

Similarly, Although 9/190 embryos (4.7%) at 22°C from cyc^sg1/+;sqt^cz35/+ matings did not express both shh and hgg1, a marker of the prechordal plate mesendoderm(genotype cyc/cyc;sqt/sqt double mutants, confirmed by PCR), clutches obtained from matings of cyc^sg1/+;sqt^cz35/+ and cyc^m294/+;sqt^cz35/+ fish showed a higher proportion (21/341; 6.2%) of shh-;hgg1- embryos(Fig. 9C,D) at 22°C. In addition, 68/254 embryos at 22°C from matings of cyc^sg1/+;sqt^cz35/+ and cyc^m294/+;sqt^cz35/+ were shh+hgg1+, but showed several gaps in the expression of shh in the floor-plate domain (genotype cyc; confirmed by PCR) (Fig. 9B,D). This is in comparison with no gaps in the floor-plate domain of hgg1+;shh+ embryos kept at 22°C from cyc^sg1/+;sqt^cz35/+ matings (36/141 cyc; genotype confirmed by PCR)(Fig. 9A,C). The floor plate in the trunk was also complete in all hgg1-;shh+ homozygous sqt mutant embryos (n=188). Similar results were obtained using the early markers, twhh and gsc, at 100% epiboly (data not shown). Therefore, although one copy of cyc^sg1 is sufficient for the initial development of the floor plate from its precursors,it is not sufficient for a complete ventral neural tube along the length of the neural axis. Furthermore, deficiencies in the prechordal plate mesendoderm did not affect induction of the floor plate by Cyc^sg1.

Nodal signalling has been shown to be required for several patterning processes in early vertebrate embryos, ranging from the specification of mesoderm, endoderm and the ventral neural tube, to establishment of left-right asymmetry (Schier and Shen,2000; Whitman,2001). Many of these events occur early in development, and may be temporally or spatially overlapping, making it difficult to assess the precise requirements for Nodal signalling in each of these germ layers and processes. Evidence for the functions of Nodal signalling in specific tissues/germ layers has been obtained primarily from the generation of chimeras in the mouse(Varlet et al., 1997; Brennan et al., 2001; Brennan et al., 2002). A hypomorphic allele has also been described in the mouse nodal gene(Lowe et al., 2001). However,the nodal^fl/∂ mutant embryos die before gestation and manifest more severe phenotypes than our cyc^sg1 homozygous mutant embryos at 22°C. Thus, the zebrafish temperature-sensitive cyc^sg1mutant provides a powerful tool with which to assess the precise requirements for Cyclops/Nodal signalling in all tissues and stages in which it functions.

Interestingly, the molecular lesion in cyc^sg1 is a transversion which results in a premature stop codon in the pro domain. The receptor-binding functional moiety of TGFβ family proteins is thought to lie within the C terminus mature domain(Kingsley, 1994), and this region should be lacking in the Cyc^sg1 mutant protein. However, cyc^sg1 is functional at 22°C. Accordingly, we find that Flag-epitope tagged Cyc^sg1 mutant protein is detected in zebrafish animal cap explants at 22°C, but not at 28.5°C. In addition,in Cos-7 cells, wild-type Cyc⁺ protein shows subcellular localization in a pattern reminiscent of the Golgi complex, whereas the mutant protein is not detected. Because an alternate start site (met 336) is present within the pro region after the stop codon, it is possible that the N terminus-truncated Cyc protein generated from this site is stable and functional in zebrafish embryos at 22°C but not at 28.5°C. Alternatively, a translational read-though mechanism similar to that described in mammalian cells (Laski et al.,1982; Hryniewicz and Vonder Haar, 1983; Phillips-Jones et al., 1995) may function at 22°C, the permissive temperature for cyc^sg1, allowing Cyc function at this temperature. It is also possible that the pro domain of Cyc has some activities that have not been previously identified. Analysis of Cyc^sg1 can therefore provide valuable insights into the functions of various domains of Cyc/Nodal proteins.

By abrogation of Cyclops signalling in cyc^sg1temperature-sensitive mutant embryos at various stages of early development,we have determined that the crucial window for Cyc function in inducing the zebrafish floor plate is during gastrulation. Disruption of Cyc function during gastrulation by temperature shift experiments and modulation of the level of Cyclops signalling results in patchy or no medial floor plate marker gene expression. Interestingly, the cyc^sg1 mutant embryos with patchy floor plate exhibit groups of floor-plate cells that are distributed throughout the length of the neuraxis. The presence of patches of floor plate cells throughout the length of the neuraxis suggests that the entire ventral neural tube arises from a group of precursors that are distributed throughout the length of the embryo, and that their differentiation into floor plate cells requires continuous Cyclops signalling during gastrulation. Previous observations by Hatta et al.(Hatta et al., 1991) where transplanted wild-type cells adopted floor-plate fates in cyc mutant hosts, and were able to recruit adjacent mutant host cells into floor plate fates, indicated that once specified, mutant cells had the ability to differentiate into floor plate cells. Our data indicates that sustained and high levels of Cyclops signalling are essential for the complete specification of floor-plate cells. Thus, in addition to being required for induction of cells of the floor plate and ventral neural tube, Cyclops signalling is also required for the development of a complete ventral neural tube throughout the entire length of the embryo.

Signalling in the anterior organizer cells, which give rise to the prechordal plate mesendoderm, has been implicated in induction of the floor plate (Sampath et al., 1998; Amacher et al., 2002). We find that deficiencies of the prechordal plate did not affect induction of the floor plate by Cyclops in sqt^cz35/sqt^cz35, cyc^sg1/cyc^sg1;sqt^cz35/sqt^cz35,or cyc^m294/cyc^sg1;sqt^cz35/sqt^cz35 mutant embryos. It is possible that the precursors of the prechordal plate cells or the remaining prechordal plate cells in these mutants are able to mediate floor plate induction via Cyclops signalling. Similar to the requirements for specification of prechordal plate mesendoderm (Gritsman et al.,2000), we find that uninterrupted and high levels of Cyclops signalling during gastrulation are crucial for induction and complete development of the floor plate in zebrafish. Sustained and high levels of Cyc/Nodal signalling during gastrulation can specify both floor plate and prechordal plate mesoderm fates. Therefore, we cannot rule out the possibility that it is the high level of Cyclops signalling, rather than signalling in the prechordal plate mesendoderm, which is responsible for floor plate induction.

Using the temperature-sensitive cyc^sg1 allele, we have conclusively provided evidence that the medial floor plate is induced during gastrulation in zebrafish. It will be important to identify the cells in which Cyc/Nodal signalling is required for inducing the floor plate, the molecules that function downstream of Cyclops signalling to mediate this process in fish, and its similarities and differences with floor plate induction in amniotes. Given that several aspects of Nodal signalling are highly conserved(Schier and Shen, 2000), and given that axial/FoxA2/HNF3β is a common downstream effector of the floor-plate induction pathways in zebrafish as well as mice(Rastegar et al., 2002),similar mechanisms and timing of floor-plate induction are also likely in other vertebrates.

We thank Mohan Balasubramanian, Suresh Jesuthasan, Yun-Jin Jiang, Vladimir Korzh, Snezhana Oliferenko, Srividya Rajagopalan and Srinivas Ramasamy for discussions and comments on the manuscript; A. Klar, V. Korzh, B. Thisse and C. Thisse for probes; D. Balasundaram and V. Korzh for antibodies; Bin Wei Jiao, Nur Nazihah, Soh Kun Peh and Mindy Tan for screening mutagenized fish;Bin Wei Jiao, Qin Yao and Tin Lay for technical assistance; and Chin Heng Goh and Amy Tan for zebrafish maintenance. K.S. is indebted to Laszlo Orban and Venkatesan Sundaresan for their support during early stages of this work. This work was supported by the Institute of Molecular Agrobiology,A^*STAR and Temasek Life Sciences Laboratory, Singapore.