An essential role of a FoxD gene in notochord induction in Ciona embryos

A key issue for understanding the early development of the chordate body plan is how the endoderm induces notochord formation (reviewed by Kimelman and Griffin, 2000; Rodaway and Patient, 2001). One possible scenario for notochord induction is that differentiation of the endoderm is tightly associated with the potential of the endoderm to induce the notochord, so that these two phenomena cannot be separated. Alternatively, the endoderm expresses a gene or produces a molecule that is not involved in the differentiation of the endoderm itself but is essential for notochord induction. We provide evidence for the latter scenario.

The endoderm of the ascidian tadpole larva is a simple tissue comprising about 500 cells, while the notochord is composed of exactly 40 cells (reviewed by Satoh, 1994; Satoh, 1999; Satoh, 2001). Cell lineage studies have documented that both the endodermal cells and notochord cells are derived from the vegetal A4.1 and B4.1 blastomeres of the eight-cell stage embryo (see Fig. 1). As early as the 32-cell stage, a pair of vegetal blastomeres (A6.1) becomes restricted to the generation of endoderm only (Fig. 1C), and at the 64-cell stage, five pairs of vegetal blastomeres (A7.1, A7.2, A7.5, B7.1 and B7.2) become endoderm restricted (Fig. 1D). Reflecting such an early fate restriction, presumptive endodermal blastomeres show strong potential for autonomous differentiation when they are isolated from early embryos (Whittaker, 1990; Nishida, 1993). It has been shown in Halocynthia roretzi embryos and may also be the case in Ciona embryos that, at the 32-cell stage, the endodermal cells (A6.1 and A6.3) induce the neighboring cells (A6.2 and A6.4) to differentiate into notochord cells (Fig. 1C), and at the 64-cell stage, A7.3 and A7.7 are destined to give rise to notochord cells (Fig. 1D) (Nakatani and Nishida, 1994). Later, B7.3 also receives induction signal(s) from B6.1, A7.5 or A7.6 that induce B7.3 to give rise to notochord cells (B8.6) (Darras and Nishida, 2001). Therefore, the 40 notochord cells are constituted from two different lineages: 32 cells are of A-line (primary), while eight cells are of B-line (secondary) lineage.

Recently, convincing evidence has been accumulated showing the involvement of β-catenin in axis determination and embryonic cell specification in a wide range of organisms from cnidarians to vertebrates (reviewed by Cadigan and Nusse, 1997; Moon and Kimelman, 1998; Sokol, 1999). β-catenin, together with TCF/LEF1, enters the nucleus and activates downstream genes, which include siamois (Brannon et al., 1997; Fan and Sokol, 1997), twins (Laurent et al., 1997), Nodal-related3 (McKendry et al., 1997) and fibronectin (Gradl et al., 1999) in Xenopus embryos; boz/dharma in zebrafish embryos (Fekany et al., 1999); and Brachyury (Yamaguchi et al., 1999; Arnold et al., 2000) in mouse embryos. In early Ciona embryos, β-catenin accumulates in the nuclei of endoderm precursor cells by the 32-cell stage, and this nuclear accumulation of β-catenin is the first step of endodermal cell specification (Imai et al., 2000). If β-catenin is mis- and/or overexpressed, the fate of presumptive notochord cells and epidermal cells changes so that they become endodermal cells. If β-catenin nuclear localization is downregulated by the overexpression of cadherin, which binds to cytoplasmic β-catenin, endodermal cell differentiation is suppressed and increased differentiation of epidermal cells occurs (Imai et al., 2000). We conducted subtractive hybridization screens of mRNAs between β-catenin-overexpressing embryos and cadherin-overexpressing embryos to identify potential β-catenin target genes in Ciona savignyi embryos (Satou et al., 2001a). We describe one β-catenin target gene belonging to the winged-helix/forkhead class of transcription factors (this class of transcription factors was recently renamed the Fox proteins, for Forkhead box) (Kaufmann and Knöchel, 1996; Kaestner et al., 2000). The transient expression of this gene is essential for notochord specification but not endodermal cell differentiation itself in the ascidian embryo.

Adults of Ciona savignyi were collected near the Otsuchi Marine Research Center, Ocean Research Institute of the University of Tokyo, Iwate and the Maizuru Fisheries Research Station of Kyoto University, Maizuru, Japan. Adults were maintained under constant light to induce oocyte maturation. Eggs and sperm were obtained surgically from the gonoduct. After insemination, embryos were reared at about 18°C in Millipore-filtered seawater (MFSW) containing 50 μg/ml streptomycin sulfate.

The procedures for isolation of cDNA clones for strongly induced β-catenin target genes are described elsewhere (Satou et al., 2001a). Briefly, β-catenin-overexpressing embryos and cadherin-overexpressing embryos were prepared by microinjection of synthetic mRNAs as described previously (Imai et al., 2000). Total RNA was isolated from 120 of the former type of embryos and 129 of the latter type of embryos, both at the 110-cell stage, and cDNAs were synthesized from 0.3 μg of each type of total RNA using a SMART PCR cDNA Synthesis kit (Clontech). The subtraction procedure of Wang and Brown (Wang and Brown, 1991) was adopted with several modifications (Satou et al., 2001a). The cDNA fragments amplified by PCR after three subtraction cycles were inserted into pGEM-T vector (Promega). The average length of the cDNA fragments obtained by PCR was about 300 to 400 bp. First, 300-400 bp of either end of the cDNAs were completely sequenced, and the nucleotide sequence information was used to check the independence of clones and sequence similarity to reported genes.

cDNA clones containing entire coding regions were isolated from a gastrula stage cDNA library using probes derived from the cDNA fragments. Nucleotide sequences were determined for both strands using a Big-Dye Terminator Cycle Sequencing Ready Reaction kit and an ABI PRISM 377 DNA sequencer (Perkin Elmer).

Genomic DNA was isolated from the gonad of Ciona savignyi and was completely digested with EcoRI or MunI, and then cloned into lambdaZAPII (Stratagene). By screening the libraries with Cs-FoxD cDNA probe, we isolated a genomic clone that covers the 3.6 kb upstream region of Cs-FoxD.

The entire upstream region was amplified with KOD DNA polymerase (TOYOBO) and ligated to the lacZ gene cloned into pSP1.72 (Promega) in frame to produce the –3.6 kb construct. The amplified nucleotide sequence was confirmed by sequencing. The –2.3 kb, –2.0 kb, –1.6 kb and –0.66 kb constructs were made by cleavage of the –3.6 kb construct at internal restriction sites (SalI, PstI, XbaI and HindIII, respectively). The constructs harboring the 1138, 1119, 1051 and 983 bp upstream regions were made from the –1.6 kb construct by the standard deletion method using exonuclease. The mutation constructs were made from the –1138 bp construct using a Quick Change site-directed mutagenesis kit (Stratagene). The mutations introduced consisted of transversions (A↔C, T↔G). The nucleotide sequences of the resultant constructs were confirmed by sequencing.

Before injection, the constructs were linearized by digestion with ScaI. The constructs were microinjected into fertilized Ciona savignyi eggs, and the eggs were allowed to develop to early gastrula stage and then fixed for β-gal staining. The details of the methods for injection and histochemical detection of β-gal activity have been described before (Satou and Satoh, 1996)

The relative β-gal activity was calculated as follows. First, the number of blastomeres with lacZ expression was counted in control or experimental embryos (the number of embryos examined is shown in parentheses in Fig. 4), and the average number was calculated. Second, the average number in the control embryos was represented as 1.0, and the ratio of blastomeres with expression in experimental embryos/blastomeres with expression in control embryos was calculated.

We isolated a PCR fragment of Cs-Notch using the following degenerate primers: 5′-TGGGCIGCIGCIGTIAAYAA-3′ and 5′-TCCATRTGRTCIGTIAYITC-3′. Using the obtained fragment, a cDNA clone with a 2.7 kb insert was isolated from the fertilized egg cDNA library. The cDNA (5.4 kb), containing four EGF repeats, three DSL domains, a transmembrane domain and six ankyrin repeats, was amplified by two successive 5′ RACE reactions using a SMART RACE cDNA Amplification Kit (Clontech).

To make a synthetic mRNA for a constitutively active form of Cs-Notch protein, the region encoding the transmembrane domain and intracellular domain was cloned into pBS-RN3 vector. Because the cloned cDNA did not have the sequence encoding the N terminus of Cs-Notch, a fragment encoding the initiator methionine and the signal peptide of Ci-cadherin was inserted into 5′-end of the constitutively active Cs-Notch construct.

In situ hybridization was carried out using standard protocols (Satou et al., 1995). Differentiation of endodermal cells was monitored by two methods: histochemical detection of AP activity (Whittaker and Meedel, 1989) and in situ hybridization with a probe for an endoderm-specific thyroid hormone receptor gene, Cs-THR, of C. savignyi. The Cs-THR gene (GenBank Accession Number, AB057767) is expressed in endodermal cells of embryos at the neurula and later stages. Differentiation of epidermal cells was monitored with a probe for an epidermis-specific gene, Cs-Epi1 (Chiba et al., 1998), and that of muscle cells with a probe for a muscle actin gene, Cs-MA (Chiba et al., 1998). Differentiation of mesenchyme cells was monitored with a probe for a mesenchyme-specific gene, Cs-mech1, whose expression begins at the tailbud stage (Imai et al., 2002a; Imai et al., 2002b), and that of neuronal cells with a probe for a nervous system-specific gene, Cs-ETR (Imai et al., 2002a). Notochord differentiation was assessed with probes for C. savignyi Brachyury gene (Cs-Bra) (Imai et al., 2000; Imai et al., 2002b) and a notochord-specific fibrinogen-like gene, Cs-fibrn (Imai et al., 2002a). Control embryos that were hybridized with sense probes did not show signals above background.

In the present study, we used a 25 mer morpholino oligo (Gene Tools LLC) for Cs-FoxD. The nucleotide sequence of the Cs-FoxD morpholino is shown in Fig. 2A. For rescue experiments, synthetic capped mRNA for Cs-FoxD was synthesized from Cs-FoxD cDNA cloned into pBluescript RN3 vector (Lemaire et al., 1995) using a Megascript T3 kit (Ambion, Austin, TX, USA). To obtain capped mRNA, the concentration of GTP was lowered to 1.5 mM and the cap analog 7mGpppG was added at a final concentration of 6 mM. The synthetic Cs-FoxD mRNA was designed to lack the morpholino sequence, and therefore the Cs-FoxD morpholino does not bind the synthetic mRNA.

The isolation and characterization of Cs-FGF4/6/9 (GenBank Accession Number, AB057767) were reported in Imai et al. (Imai et al., 2002a). Synthetic capped mRNA for Cs-FGF4/6/9 was synthesized as described above.

After insemination, fertilized eggs, usually with intact chorion, were microinjected with 15 fmole of morpholino and/or synthetic capped mRNA (Satou et al., 2001b). For each microinjection, 30 pl of solution was injected using a micromanipulator (Narishige Science Instruments Laboratories, Tokyo, Japan) as described (Imai et al., 2000). Injected eggs were reared at about 18°C in MFSW containing 50 μg/ml streptomycin sulfate. Cleavage of some embryos was arrested at the 110-cell stage with cytochalasin B and the embryos were then cultured for about 12 hours as control embryos that reached the early tailbud stage.

Semi-quantitative RT-PCR was carried out as described (Imai et al., 2000; Imai et al., 2002b). RNA extracted from 20 normal tailbud embryos and 20 embryos at the same stage developed from eggs injected with Cs-FoxD morpholino was used for cDNA synthesis. PCR was performed by using the following primers:

Cs-THR, 5′-GATGGACGTGAACCAATTG-3′ and 5′-GCGTATGTCGTGTATTCATA-3′; and

Cs-ETR, 5′-TTCAATGCACCGTGTACATA-3′ and 5′-GGACCTCGGGTTCTAGC-3′.

Each reaction (for Cs-THR or Cs-ETR) was performed for 23 cycles, with each cycle consisting of incubation using [α-³²P]dCTP at 94°C for 30 seconds, 55°C for 30 seconds and 72°C for 30 seconds. The reaction products were resolved by 5% PAGE and subjected to autoradiography.

Subtractive hybridization screening of mRNAs from β-catenin-overexpressing embryos versus cadherin-overexpressing embryos in C. savignyi yielded two β-catenin-downstream genes belonging to the winged-helix/forkhead class of transcription factors or Fox proteins (Fig. 2). The first cDNA clone consisted of 1949 nucleotides, which encoded a predicted polypeptide of 579 amino acids, and represented a gene identical to the previously characterized forkhead/HNF-3β (Cs-FoxA5) (Shimauchi et al., 2001a) (Fig. 2B,C). The second cDNA was 1915 base pairs in length and encoded a novel Fox protein of 506 amino acid residues (Fig. 2A) (DDBJ/GenBank/EMBL database Accession Number, AB057738). Comparative analyses of the amino acid sequences of the forkhead domain suggested that the second gene is a FoxD class member (Fig. 2B,C), and we therefore named it Cs-FoxD.

Expression of Cs-FoxD began at the 16-cell stage, and the in situ hybridization signal was evident in A5.1, A5.2 and B5.1, which have the developmental potential to give rise to endoderm, notochord and other tissues (Fig. 1H, also see 1B). During the next cleavage to form the 32-cell stage embryo, the expression was inherited by only one of each of the daughter cells, namely A6.1, A6.3 and B6.1 (Fig. 1I, also see 1C). A6.1 and B6.1 are destined to give rise to endoderm only, while A6.3 has the potential to form endoderm and trunk lateral cells (a type of mesenchyme cell). This early embryonic expression of Cs-FoxD was transient: the in situ signal was not detected by the 64-cell stage (Fig. 1J,K).

Later during embryogenesis, Cs-FoxD was expressed again in cells that included cells in the nervous system (Fig. 1L-N). This second phase of expression was detected in a pair of cells in the posterior neural plate (Fig. 1L). In the neurula, the expression was seen in several pairs of cells, and in the tailbud embryo, the transcript was detected in five domains (arrowheads in Fig. 1N), including anterior nerve cord cells (Fig. 1N).

If the early transient expression of Cs-FoxD is triggered by β-catenin nuclear localization, the expression of Cs-FoxD should be upregulated in embryos injected with in vitro synthesized β-catenin mRNA and downregulated in embryos injected with in vitro synthesized cadherin mRNA. As shown in Fig. 3A, injection of β-catenin mRNA into fertilized eggs resulted in an ectopic expression of Cs-FoxD in all of the blastomeres in 32-cell stage embryos (eight out of nine, the number of embryos with ectopic Cs-FoxD expression/the number of embryos examined). By contrast, injection of cadherin mRNA into fertilized eggs resulted in no expression of Cs-FoxD in 32-cell stage embryos (0/9; Fig. 3B). The embryos in which the expression of Cs-FoxD was completely suppressed (Fig. 3B) did not form notochords, presumably owing to the lack of endodermal cell differentiation (data not shown) (Imai et al., 2000).

β-catenin acts together with Tcf/LEF as a transcription activator (reviewed by Miller and Moon, 1996; Willert and Nusse, 1998). To determine whether nuclear localization of β-catenin directly controls the initiation of Cs-FoxD expression, we carried out the four types of experiments, described below, and all of the findings strongly supported that the initiation of Cs-FoxD expression is directly controlled by nuclear accumulation of β-catenin via Tcf-binding elements. First, we examined the sequence of about 3.6 kb of the 5′ flanking region of Cs-FoxD, and found that there are three Tcf-binding elements, TBE1, TBE2 and TBE3 [(C/G)TTTG(A/T)(A/T)], between nucleotide positions –1202 and –963 upstream from the transcription start site (Fig. 4A).

Second, we injected various deletion constructs of p(–3.6kb)CsFoxD/lacZ into fertilized eggs and examined the activity of the reporter gene expression or lacZ protein expression at the early gastrula stage. Various deletion constructs between –3600 and –1138 showed the same level of lacZ expression (data not shown). Therefore, we used p(–1138)Cs-FoxD/lacZ, which contained all three TBEs, as a control, and compared its expression level with those of three further deleted and mutated constructs. The first deletion construct, p(–1119)Cs-FoxD/lacZ, which lacked TBE1, and the second deletion construct, p(–1051)Cs-FoxD/lacZ, which lacked TBE1 and TBE2, showed about 60% and 80% reduction in the number of cells with the reporter gene expression as compared with the control (Fig. 4B). In addition, the third deletion construct, p(–983)Cs-FoxD/lacZ, which lacked all three TBEs, did not show detectable reporter gene expression (Fig. 4B).

Third, when p(–1138)Cs-FoxD/lacZ was co-injected with β-catenin mRNA, the number of cells with lacZ expression increased more than nine times compared with that in embryos injected with p(–1138)Cs-FoxD/lacZ only (second lane in Fig. 4C). However, co-injection of p(–1138)Cs-FoxD/lacZ with cadherin mRNA abolished lacZ expression completely (third lane in Fig. 4C).

Fourth, mutation of control sequences (N region of Fig. 4A) resulted in the same level of reporter gene expression as intact p(–1138)Cs-FoxD/lacZ (Fig. 4C,D). Mutation of either TBE1, TBE2 or TBE3 also resulted in comparable reporter gene expression. However, mutations in all three TBEs resulted in a lack of detectable reporter gene expression (Fig. 4C,E). Co-injection of β-catenin mRNA along with the construct with mutations in all three TBEs did not rescue the reporter gene expression (Fig. 4C).

The transient expression of Cs-FoxD occurs primarily in endoderm precursor cells, suggesting a pivotal role of this gene in endodermal cell differentiation. We examined the role of Cs-FoxD in the differentiation of endoderm and of other tissues and organs by suppressing Cs-FoxD gene function with a specific morpholino oligo. Recently, we found that morpholinos specifically inhibit Ciona gene function and that their functional specificity can be ascertained by rescue experiments with in vitro synthesized mRNAs that lack the target sequences of the morpholinos (Satou et al., 2001a; Satou et al., 2001b). As a control, we injected morpholino against lacZ, and this injection had no effect on C. savignyi embryogenesis (data not shown) (Satou et al., 2001b).

Fertilized eggs injected with Cs-FoxD morpholino cleaved normally, and gastrulation took place as in normal embryos, but formation of tailbud embryos was usually affected, resulting in disordered tailbud embryos (Figs 5, 6). Unexpectedly, however, suppression of Cs-FoxD function did not affect endodermal cell differentiation (Table 1; Fig. 5). Cs-FoxD morpholino-injected embryos showed expression of endodermal markers, i.e. alkaline phosphatase (Cs-AP) (Fig. 5A′,B′) and a gene for thyroid hormone receptor (Cs-THR) (Fig. 5C′,D′), comparable with that of normal control embryos (Fig. 5A,B for Cs-AP; Fig. 5C,D for Cs-THR).

We also examined the differentiation of epidermis, muscle and mesenchyme in embryos developed from eggs injected with Cs-FoxD morpholino. Differentiation of these cell-types was assessed by monitoring the expression of an epidermis-specific gene, Cs-Epi1 (Fig. 6A), a muscle-specific actin gene, Cs-MA1 (Fig. 6B), and a mesenchyme-specific gene, Cs-mech1 (Fig. 6C). As shown in Fig. 6 and Table 1, the expression of Cs-Epi1 (Fig. 6A′), Cs-MA1 (Fig. 6B′) and Cs-mech1 (Fig. 6C′) was not affected by injection of Cs-FoxD morpholino. Therefore, Cs-FoxD is not likely to be involved in the differentiation of epidermis, muscle or mesenchyme.

In the ascidian embryo, the induction of notochord cells by signal(s) from endodermal cells takes place at the 32-cell stage (Nakatani and Nishida, 1994), and immediately after the induction or at the 64-cell stage, the primordial notochord cells begin to express the Brachyury gene (Yasuo and Satoh, 1993; Corbo et al., 1997), which encodes a T-box transcription factor. As mentioned above, Cs-FoxD is expressed transiently in endodermal cells, but suppression of its function does not block differentiation of endodermal cells themselves. However, it is possible that this gene functions in the induction and subsequent differentiation of notochord cells. The effects of Cs-FoxD morpholino on notochord induction were examined using two markers, expression of Brachyury (Cs-Bra) (Fig. 7A,B) and expression of a notochord-specific fibrinogen-like gene (Cs-fibrn) (Fig. 7G,H).

As shown in Fig. 7B and Table 1, injection of Cs-FoxD morpholino into fertilized eggs resulted in failure of the expression of Cs-Bra. In addition, suppression of the function of Cs-FoxD in the right AB2 cell of the two-cell stage embryo resulted in the failure of Cs-Bra expression in the right A7.3, A7.4, A7.7, A7.8 (white arrowheads in Fig. 7D) and B7.6 (white arrow in Fig. 7D). The suppression of the function of Cs-FoxD in one of the blasomeres of the four-cell stage embryo resulted in the failure of Cs-Bra expression in notochord precursors derived from the Cs-FoxD-suppressed blastomere (white arrow in Fig. 7E and white arrowheads in Fig. 7F). Expression of Cs-fibrn was also suppressed in embryos developed from eggs injected with Cs-FoxD morpholino (Fig. 7H; Table 1).

To examine whether the suppression is specifically caused by downregulation of Cs-FoxD function, we attempted rescue experiments. mRNA for Cs-FoxD lacking the Cs-FoxD morpholino recognition sequences was synthesized in vitro and injected into fertilized eggs together with Cs-FoxD morpholino. This co-injection rescued the expression of Cs-Bra (Fig. 7C; Table 1) and Cs-fibrn (Fig. 7I; Table 1) in the resultant embryos, although the morphology of the embryos were not recovered possibly because the effect of Cs-FoxD mRNA overexpression. All of these experimental results strongly suggest that Cs-FoxD is essential for the differentiation of notochord cells. It is also likely that the Cs-FoxD-downstream inductive signal does not act over a long range.

The presumptive notochord cells in Cs-FoxD-suppressed embryos were likely to have changed their fate into that of endodermal cells. As shown in Fig. 5B′,D′, in Cs-FoxD-suppressed embryos, the expression of endodermal markers Cs-AP and Cs-THR was detected not only in the endodermal cells but also in presumptive notochord cells (arrowheads); this was confirmed in 22 out of 30 embryos examined for Cs-AP and in all 22 embryos examined for Cs-THR. To confirm this result further, we performed semi-quantitative RT-PCR analysis. As is evident in Fig. 8, the quantity of Cs-THR transcripts in Cs-FoxD morpholino-injected embryos was higher than that in control embryos.

We also examined the effects of overexpression of Cs-FoxD by injection of its synthetic Cs-FoxD mRNA into fertilized eggs, and found that this injection did not affect notochord formation; that is, notochord cells were formed in the experimental embryos as in normal embryos (data not shown). However, we noticed that the epidermis failed to differentiate in Cs-FoxD mRNA-injected embryos (data not shown).

Notochord cells of the ascidian embryo are involved, either directly or indirectly, in pattern formation in the nervous system, and therefore inhibition of notochord formation by suppression of Cs-FoxD function may affect the nervous system formation. We examined this question with a marker gene, Cs-ETR (Fig. 6D). Injection of Cs-FoxD morpholino did not diminish the expression of Cs-ETR (Fig. 6D’), but the quantity of Cs-ETR transcripts in Cs-FoxD morpholino-injected embryos decreased when compared with that in the control (Fig. 8).

Two signal transduction pathways have been suggested for notochord induction in ascidian embryos: a Notch-related pathway in Ciona intestinalis embryos (Corbo et al., 1998) and a bFGF-related pathway in Halocynthia roretzi embryos (Nakatani et al., 1996; Shimauchi et al., 2001b). We examined the possible involvement of these signal pathways downstream of Cs-FoxD.

To examine the Notch signal pathway, we characterized a cDNA for Cs-Notch. As shown in Fig. 9A, although the cDNA lacks the 5′ region encoding the N-terminal half of Cs-Notch, it contains sequences encoding four EGF repeats, three DSL domains, the transmembrane domain and six ankyrin repeats. As in the case of Hr-Notch (Halocynthia homolog of Notch) (Hori et al., 1997), Cs-Notch is expressed maternally and zygotically. The maternal expression of Cs-Notch is evident from the fact that we isolated the Cs-Notch clone from a cDNA library of C. savignyi fertilized eggs. Whole-mount in situ hybridization with a Cs-Notch probe showed that the maternal transcript is distributed evenly in fertilized eggs and early embryos (Fig. 9B-D). Zygotic Cs-Notch expression begins at the neurula stage, and the Cs-Notch transcript is evident in the cells of the central nervous system (Fig. 9E,F).

A construct that lacks the extracellular domain of Notch acts as constitutively active form (Kopan et al., 1996). We made a constitutively active form (ΔE-Notch) of Cs-Notch (Fig. 9A), and examined whether the overexpression of ΔE-Notch was able to rescue the notochord development in Cs-FoxD-suppressed embryos. As shown in Fig. 10C, injection of ΔE-Notch mRNA into fertilized eggs resulted in the recovery of Cs-fibrn gene expression in B-line notochord cells of Cs-FoxD-suppressed embryos, but none of the A-line notochord cells expressed Cs-fibrn. In addition, the experimental embryos showed ectopic expression of Cs-fibrn in B-line endodermal cells and mesenchyme cells (Fig. 10C).

We also examined the effects of injection of the ΔE-Notch construct into intact fertilized eggs, and found that all 40 experimental embryos showed ectopic expression of Cs-fibrn in B-line endodermal cells and mesenchyme cells (Fig. 10B). In addition, eleven out of the 40 experimental embryos showed ectopic Cs-fibrn expression in A-line endodermal cells as well (data not shown). Therefore, it is likely that there is a difference in the genetic cascade for the induction of notochord between the A- and B-lines, even though Cs-FoxD functions in the notochord induction in both lines.

Recently we found that Ciona contains at least four members of the FGF family, FGF3/7/10, FGF4/6/9/20, FGF8/17/18 and FGF11/12/13/14 (K. S. I., N. S. and Y. S., unpublished). Among these FGF family genes, Cs-FGF4/6/9 encodes an FGF protein with features of both FGF4/6 and FGF9/20 of vertebrates, and the gene is expressed in endodermal cells and notochord-lineage cells around the notochord induction stage (Imai et al., 2002a). When Cs-FGF4/6/9 function was suppressed with morpholino, the differentiation of mesenchyme cells was completely blocked (Imai et al., 2002a). However, the role of Cs-FGF4/6/9 in the induction of notochord cells is only partial; the initiation of notochord-specific gene expression was inhibited by Cs-FGF4/6/9 morpholino, but notochord-specific genes were expressed later, resulting in formation of a partial notochord.

The previous study showed that injection of Cs-FGF4/6/9 morpholino into fertilized eggs did not affect Cs-FoxD expression (Imai et al., 2002a). Vice versa, injection of Cs-FoxD morpholino did not affect Cs-FGF4/6/9 expression, suggesting that the expression and function of Cs-FGF4/6/9 and Cs-FoxD are not interdependent (Imai et al., 2002a). In the present study, we first confirmed these experimental results (data not shown). Then, we examined whether overexpression of Cs-FGF4/6/9 is able to rescue the notochord development in Cs-FoxD-suppressed embryos. As shown in Fig. 10D, injection of Cs-FGF4/6/9 mRNA into fertilized eggs failed to rescue the Cs-fibrn gene expression in Cs-FoxD-suppressed embryos (0/22, the number of embryos with Cs-fibrn expression/the number of embryos examined). Therefore, it is unlikely that Cs-FGF4/6/9 plays a role as a downstream signal of Cs-FoxD in notochord specification.

The results of the present study indicate that (1) during early Ciona embryogenesis, nuclear localization of β-catenin triggers the transient expression of Cs-FoxD in endodermal cells at the 16- and 32-cell stages, (2) the function of Cs-FoxD is not associated with the differentiation of the endoderm itself but is essential for the induction and subsequent differentiation of notochord cells, and the putative Cs-FoxD-downstream inductive signal does not act long-range, (3) the expression and function of Cs-FoxD and Cs-FGF4/6/9 are not interdependent, and (4) there is a difference between the induction processes in the A-line and B-line notochord cells, and the Notch pathway is likely to be involved in the B-line notochord induction downstream of Cs-FoxD.

Unfertilized eggs of various kinds of animals are provided maternally with a considerable amount of β-catenin mRNA and protein in the cytoplasm. During early cleavages after fertilization, β-catenin, together with TCF/LEF1, enters the nucleus and activates downstream genes which include siamois (Brannon et al., 1997; Fan and Sokol, 1997), twins (Laurent et al., 1997), Nodal-related3 (McKendry et al., 1997) and fibronectin (Gradl et al., 1999) in Xenopus embryos, and boz/dharma in zebrafish embryos (Fekany et al., 1999). Via the function of these induced genes, β-catenin plays pivotal roles in early developmental processes including axis determination and embryonic cell specification (reviewed by Cadigan and Nusse, 1997; Moon and Kimelman, 1998; Sokol, 1999). As in the case of vertebrate embryos, β-catenin accumulates in the nuclei of blastomeres in early Ciona embryos (Imai et al., 2000). However, Ciona β-catenin enters into the nuclei of endoderm precursor cells and triggers the activation of, for example, a LIM-homeodomain gene, Cs-lhx3, which is required for endodermal cell differentiation (Satou et al., 2001a).

As shown in Fig. 4A, about 1.2 kb of the 5′ flanking region of Cs-FoxD, which contains three Tcf-binding elements between nucleotide positions –1202 and –963 upstream from the transcription start site, is sufficient to drive specific expression of a reporter gene. Deletion of or mutation in the three Tcf-binding elements resulted in dramatic reduction of the reporter gene expression (Fig. 4B,C). Co-injection of p(–1138)Cs-FoxD/lacZ with β-catenin mRNA increased the number of cells with lacZ expression, while co-injection of cadherin mRNA with this construct abolished lacZ expression completely (Fig. 4C). In addition, co-injection of p(–1138)Cs-FoxD/lacZ without the three Tcf-binding elements along with β-catenin mRNA resulted in a failure to express lacZ. All of these data strongly suggest that the initiation of Cs-FoxD expression is directly controlled by nuclear accumulation of β-catenin, via Tcf-binding elements, although it remains possible that other binding element(s) are also involved in the control of Cs-FoxD expression.

In ascidians, Brachyury is essential for notochord cell differentiation. Hr-Bra, Ci-Bra and Cs-Bra are expressed exclusively in notochord cells (Yasuo and Satoh, 1993; Corbo et al., 1997; Imai et al., 2000). Ectopic expression of Hr-Bra (Yasuo and Satoh, 1998) or Ci-Bra (Takahashi et al., 1999a) promotes notochord cell differentiation of non-notochord lineage cells. Inhibition of expression of Hr-Bra (H. Takahashi, K. Hotta. and N. S., unpublished) or Cs-Bra (Satou et al., 2001b) results in failure of notochord cell differentiation. In addition, Ci-Bra activates more than 20 specific genes that may be associated with the structural construction and subsequent function of the notochord in Ciona embryos (Takahashi et al., 1999a; Di Gregorio and Levine, 1999; Hotta et al., 2000). Therefore, a key issue to explore in order to understand notochord induction is the pathway that lies between Cs-FoxD and Brachyury: what molecules are activated or repressed by transcription factor Cs-FoxD, and how do they trigger the activation of Brachyury?

In the above-mentioned context, the 5′ flanking sequences required for notochord-specific expression of Ci-Bra have been studied in detail (Corbo et al., 1997). The basal promoter of Ci-Bra is present in about a 600 bp 5′ flanking region of Ci-Bra, and consists of at least three regions: i.e. from distal to proximal to the transcription start site; a region responsible for suppression of the gene in mesodermal regions (mesenchyme and muscle) other than the notochord, a region responsible for activation of the gene in the notochord, and a region responsible for activation in other mesodermal regions (Corbo et al., 1997). The region responsible for the notochord-specific activation contains binding sites for suppressor of hairless [Su(H)], and these sites are essential for the notochord-specific expression of Ci-Bra, leading to the suggestion that Notch is a signal molecule (Corbo et al., 1998).

As shown in the present study, the overexpression of ΔE-Notch in normal embryos changes the fate of B-line endodermal cells into that of notochord cells (Fig. 10B). In addition, overexpression of ΔE-Notch in Cs-FoxD morpholino-injected embryos could rescue the B-line notochord differentiation (Fig. 10C). Therefore, it is highly likely that Notch protein is produced by translation of its maternal message, and its signal transduction pathway via binding sites for Su(H) is involved in B-line notochord induction. These results also suggest that both B-line presumptive notochord cells and endodermal cells are competent to respond to Notch signal transduction molecules, and some mechanisms may block the competence of the endodermal cells.

However, overexpression of ΔE-Notch in Cs-FoxD morpholino-injected embryos could not rescue the A-line notochord differentiation. This suggests that there are different induction mechanisms in the A-line and B-line notochord cells. The identity of the molecules that act downstream of Cs-FoxD and trigger the activation of Cs-Bra in A-line notochord cells will be discussed below in relation to Cs-ZicL.

Regarding the role of Cs-FGF4/6/9 in Ciona notochord induction, the previous study showed that its role in the notochord induction is partial, because the suppression of Cs-FGF4/6/9 function cannot completely inhibit the notochord formation (Imai et al., 2002a). As shown in the previous and present studies, the expression of Cs-FGF4/6/9 and Cs-FoxD are independent. Injection of Cs-FGF4/6/9 mRNA into fertilized eggs did not rescue the Cs-fibrn gene expression in Cs-FoxD-suppressed embryos. Therefore, some signals other than Cs-FGF4/6/9 may be involved in the notochord induction in Ciona embryos. However, it remains possible that Cs-FoxD is required for the competence for Cs-FGF4/6/9 or other signals, because Cs-FoxD is expressed in notochord/endoderm precursors at 16-cell stage.

Recently, we isolated cDNA clone for a Zic-like zinc finger transcription factor gene Cs-ZicL, as one of β-catenin downstream genes (Imai et al., 2002b). Cs-ZicL is transiently expressed in the A-line notochord/nerve cord lineage and in B-line muscle lineage from the 32-cell stage, and later in a-line CNS lineage from the 110-cell stage. Suppression of Cs-ZicL function with morpholino indicated that Cs-ZicL is essential for the formation of A-line notochord cells but not of B-line notochord cells (Imai et al., 2002b). The expression of Cs-ZicL in the A-line cells is downstream of Cs-FoxD. Therefore, it is highly likely that in the A-line notochord cells acts Cs-ZicL as downstream of Cs-FoxD, while in the B-line notochord cells acts Cs-Notch as downstream of Cs-FoxD.

Besides the genes discussed above, some other genes may be involved in this induction process. For example, FoxA5 (forkhead/HNF3β) gene is expressed in the blastomeres of both endoderm and notochord lineages (Shimauchi et al., 1997). Suppression of the function of Cs-FoxA5 results in a deficiency of endoderm development, which in turn leads to the failure of notochord induction (K. S. I., unpublished). However, it has been shown that FoxA and Brachyury act synergistically in the notochord formation in Halocynthia embryos (Shimauchi et al., 2001b), as previously demonstrated in Xenopus (O’Reilly et al., 1995). In addition, other genes such as BMP2/4 are involved in notochord formation (Darras and Nishida, 2001). The genetic cascade of notochord induction in ascidian embryos appears to be not so simple.

Ascidians are subdivided into two major orders: the Enterogona and the Pleurogona (see Satoh, 1994). Ciona species belong to the former order, while Halocynthia roretzi belong to the latter. Recent molecular phylogenetic studies suggest that the two orders diverged comparatively early in the history of evolution of urochordates (Wada et al., 1992; Cameron et al., 2000). Therefore, it is possible that the two ascidian species adopt different molecular mechanisms in notochord induction. Indeed, previous analyses of notochord-specific enhancer of the 5′ flanking region of ascidian Brachyury indicated that the enhancer sequences of Ciona Ci-Bra (Corbo et al., 1997) appear to be different from those of Halocynthia Hr-Bra (Takahashi et al., 1999b).

Using H. roretzi embryos, Nishida and colleagues conducted a series of blastomere isolation and recombination experiments, and studied the effects of the treatment of the blastomeres with exogenous human bFGF (Nakatani and Nishida, 1994; Nakatani and Nishida, 1997; Nakatani et al., 1996; Kim et al., 2000; Kim and Nishida, 2001; Darras and Nishida, 2001). They suggested that FGF signal transduction pathways are involved in notochord induction in this species. Because a dominant-negative form of FGF receptor (HrFGFR) of H. roretzi causes suppression of Hr-Bra (As-T) expression, it is likely that FGF-like molecules are involved in notochord formation (Shimauchi et al., 2001b). Therefore, it is important to isolate FGF-related genes from H. roretzi to determine their roles in the notochord induction.

In Ciona embryos, Cs-FGF4/6/9 is partially involved in the notochord induction (Imai et al., 2002a), and therefore molecules other than FGFs appear to be responsible for this process. As shown in the present study, Cs-FoxD play a pivotal role in the specification of notochord cells in Ciona embryos. In addition, Cs-Notch acts downstream of Cs-FoxD in the B-line notochord induction while Cs-ZicL acts downstream of Cs-FoxD in the A-line notochord induction. Therefore, it should be determined whether these genes are involved in the notochord formation in Halocynthia embryos. Such efforts may disclose evolutionary changes in the mechanism involved in the formation of ascidian larval notochord, the most prominent characteristic feature of chordates (Satoh and Jeffery, 1995).

Recent studies have shown that vertebrate FoxD genes are expressed in the central nervous system and are involved in neural crest formation and function (Pohl and Knöchel, 2001; Sasai et al., 2001; Kos et al., 2001). One important question to be answered is whether vertebrate FoxD genes are also expressed in early embryos and are involved in the specification of notochord cells. As shown in the present study, Cs-FoxD expression is very transient, as it is seen at just the 16-cell and 32-cell stages. Therefore, it is possible that vertebrate FoxD genes are also expressed at early stages of embryogenesis, as is Cs-FoxD. Although the functions reported for vertebrate FoxD genes are mainly associated with the neural lineage, some FoxD genes are also expressed in the mesoderm at early embryonic stages. For example, zebrafish fkh6, fkh8 and fkh9 are expressed in paraxial mesoderm (Odenthal and Nüsslein-Volhard, 1998). Xenopus foxD5a is expressed in paraxial mesoderm during primary gastrulation, and ventral injection of Xenopus foxD5a mRNA induces partial secondary axes composed of expanded mesoderm and ectoderm (Sullivan et al., 2001). Xenopus foxD5a is strongly induced by siamois (Sullivan et al., 2001). Mouse Fkh2 is expressed in notochord and anterior endoderm at the headfold stage (Kaestner et al., 1995). These results suggest that vertebrate FoxD family genes might have functions in the patterning of the mesoderm.

Conversely, it is also interesting that there is later expression of Cs-FoxD in epidermal cells and nerve cord cells in the early tailbud embryo. Is there any evolutionary link between neural crest cells and this later expression in ascidian embryos, which do not develop neural crests?

In conclusion, an ascidian FoxD gene is expressed transiently and primarily in endodermal cells, and its expression is essential for notochord differentiation. Further characterization of Cs-FoxD target genes is important for understanding the molecular mechanisms involved in the notochord induction of ascidian embryos.

We thank Dr Yasuaki Takagi, Mr Kouichi Morita and all of the members in the Otsuchi Marine Research Center, Ocean Research Institute of the University of Tokyo, Iwate, Japan for collecting Ciona savignyi. This research was supported by Grants-in-Aid to from the MEXT, Japan to Y. S., K. S. I. and N. S., and by a grant from the Human Frontier Science Program to N. S. (RG0290/2000-MR).