Multiple functions of a Zic-like gene in the differentiation of notochord, central nervous system and muscle in Ciona savignyi embryos

Unfertilized eggs of various groups of animals are maternally provided with a considerable amount of β-catenin protein and mRNA in the cytoplasm. During very early cleavages after fertilization, β-catenin is translocated from the cytoplasm into the nucleus of certain embryonic cells. The nuclear accumulation of β-catenin, together with transcription factor Tcf/Lef, activates many target genes that play pivotal roles in embryonic axis formation and/or embryonic cell specification (reviewed by Cadigan and Nusse, 1997; Moon and Kimelman, 1998; Sokol, 1999).

As in the case of vertebrate embryos, the endoderm of the ascidian embryo is specified autonomously and then induces formation of the notochord and mesenchyme (reviewed by Satoh, 1994; Satoh, 2001; Satou and Satoh, 1999; Nishida, 1997; Jeffery, 2001; Corbo et al., 2001). In a previous study, we showed that during cleavages of the ascidians Ciona intestinalis and C. savignyi, β-catenin accumulates in the nuclei of vegetal blastomeres by the 32-cell stage, that mis- and/or overexpression of β-catenin induces the ectopic development of endoderm cells, and that downregulation of nuclear β-catenin induced by the overexpression of cadherin results in the suppression of endodermal cell differentiation (Imai et al., 2000). Thus, the accumulation of β-catenin in the nuclei of endoderm progenitor cells is most likely the first step in the process of ascidian embryonic endoderm specification.

To understand the function of endoderm in ascidian embryos, it is necessary to identify the genes that act as direct targets and/or act downstream of β-catenin. We previously took advantage of the availability of β-catenin-overexpressing embryos and cadherin-overexpressing embryos to address this problem; in the former, β-catenin targets may be upregulated and in the latter, β-catenin targets may be downregulated, and subtractive hybridization screening between them performed. We found that a LIM-homeobox gene, Cs-lhx3, an otx homolog Cs-otx, and an NK-2 class homeobox gene Cs-ttf1 are downstream genes of β-catenin (Satou et al., 2001a). Inhibition of the function of these genes revealed that inhibition of the possible early embryonic function of Cs-lhx3 resulted in the suppression of endoderm differentiation. In addition, we found that the nuclear accumulation of β-catenin directly activates Cs-FoxD, which encodes a transcription factor with a forkhead domain. Cs-FoxD is transiently expressed in endoderm blastomeres at the 16- and 32-cell stages, and this gene function is not associated with endoderm but is necessary for notochord differentiation (K. S. I., unpublished). Here, we showed that another β-catenin downstream gene encodes a Zic-like protein. The gene is expressed in three different domains; namely, A-line notochord and central nervous system (CNS), and B-line muscle cells, and plays pivotal roles in the differentiation of the three different cell types.

Ciona savignyi adults were maintained under constant light to induce oocyte maturation. Eggs and sperm were obtained surgically from the gonoduct. After insemination, eggs were reared at about 18°C in Millipore-filtered seawater (MFSW) containing 50 μg/ml streptomycin sulfate.

A cDNA clone for a Zic-like gene (named Cs-ZicL) was isolated as one of the β-catenin downstream genes; the procedure for subtractive hybridization screening of cDNA clones for potent β-catenin target genes was described previously (Satou et al., 2001a). The cDNA obtained by subtraction was partial, and cDNA clones that contained the entire coding region were isolated from a C. savignyi gastrula cDNA library. Nucleotide sequences were determined for both strands using a Big-Dye Terminator Cycle Sequencing Ready Reaction kit and an ABI PRISM 377 DNA sequencer (Applied Biosystems).

To examine the spatiotemporal expression patterns of Cs-ZicL, RNA probes were prepared with a DIG RNA labeling kit (Roche, Tokyo, Japan). Whole-mount in situ hybridization was performed as described previously (Satou et al., 2001a). Control specimens hybridized with a sense probe did not show any signals above background.

To examine the occurrence of differentiation markers in experimental embryos, in situ hybridization of whole-mount specimens was also carried out. The probes used were for a muscle actin gene, Cs-MA1 (Chiba et al., 1998), an epidermis-specific gene, Cs-Epi1 (Chiba et al., 1998), a mesenchyme-specific gene, Cs-mech1 (DDBJ/GenBank/EMBL database accession number, AB073374), and a CNS-specific gene, Cs-ETR (DDBJ/GenBank/EMBL database accession number, AB073375). Notochord differentiation was assessed with probes for the C. savignyi Brachury gene (Cs-Bra; Imai et al., 2000) and notochord-specific fibrinogen-like gene Cs-fibrn (DDBJ/GenBank/EMBL database accession number, AB073373).

Differentiation of endodermal cells was monitored by histochemical detection of alkaline phosphatase (AP) activity (Whittaker and Meedel, 1989).

To deduce the function of Cs-ZicL, we used morpholino antisense oligonucleotides (hereafter we simply refer to these as “morpholinos”), which have been shown to be very effective in ascidian embryos (Satou et al., 2001b). The 25-mer morpholino for Cs-ZicL was order-made (Gene Tools, LLC). The nucleotide sequence of the Cs-ZicL morpholino is shown in Fig. 1A. For rescue experiments, in vitro synthesized capped mRNA for Cs-ZicL was prepared from Cs-ZicL cDNA cloned into pBluescript RN3 vector (Lemaire et al., 1995) using a Megascript T3 kit (Ambion). To obtain a 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-ZicL mRNA was designed to lack the morpholino sequence, and therefore the Cs-ZicL morpholino does not recognize the synthetic mRNA. In the present study, we also used morpholinos for Cs-macho1 (Satou et al., 2002), and Cs-FoxD (K. S. I., unpublished).

After insemination, fertilized eggs were microinjected with 15 fmole of morpholinos and/or synthetic capped mRNAs. Each injection contained 30 pl of solution, and microinjection was carried out using a micromanipulator (Narishige Sci. Instru. Lab., Tokyo, Japan), as described previously (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 further cultured for about 12 hours, when control embryos reached the early tailbud embryo stage.

Subtractive hybridization screening of mRNAs from β-catenin-overexpressing embryos versus cadherin-overexpressing embryos yielded a cDNA fragment of a β-catenin-downstream gene which encoded a zinc finger protein. The gene was named Cs-ZicL (Ciona savignyi Zic-like). By screening of a gastrula cDNA library with this cDNA fragment as a probe, cDNA clones for Cs-ZicL were obtained and the longest was completely sequenced. As shown in Fig. 1, the insert of Cs-ZicL cDNA consisted of 1,217 nucleotides, which encoded a predicted polypeptide of 355 amino acids (Fig. 1A: DDBJ/GenBank/EMBL database accession number, AB057747). The predicted polypeptide contained five zinc finger domains (Figs 1 and 2).

Fig. 2A shows a comparison of the amino acid sequences of the five zinc finger domains of Cs-ZicL, Cs-macho1 of C. savignyi (Satou et al., 2002), macho-1 of Halocynthia roretzi (Nishida and Sawada, 2001) and mouse Zic3 (Aruga et al., 1996). Although there is variation in the first zinc-finger domain, these proteins share comparatively high level of amino-acid identity. Using these amino acid residues within the five zinc finger domains, we constructed a molecular phylogenetic tree using the neighbor-joining algorithm (Fig. 2B). The tree demonstrated that mouse and Xenopus Zic gene products form one clade, from which Ciona macho1, Halocynthia macho-1 and Ciona ZicL are distant. Within the three ascidian Zic-like gene products, Ciona macho1, Halocynthia macho-1 and Ciona ZicL did not form a discrete group, but there was a tendency for Halocynthia macho-1 and Ciona ZicL to form a group. These results suggest that the ascidian Zic-like genes evolved from a common ancestor gene of the Zic family independently of vertebrate Zic genes.

Analyses of whole-mount specimens by in situ hybridization revealed that Cs-ZicL is expressed in multiple embryonic regions. As shown in Fig. 3, the embryonic expression of Cs-ZicL is transient; the transcript was first detected at the 32-cell stage and was downregulated by the early tailbud stage. The expression was found in embryonic cells that give rise to notochord, CNS and muscle. The multiple expression domains and functions of Cs-ZicL in each lineage were further examined, with the following results.

In ascidians, exactly 40 notochord cells are formed in the larval tail. Of them, 32 are derived from A-line cells and eight from B-line cells. At the 64-cell stage, A7.3 and A7.7 pairs are primordial notochord cells (Fig. 3C′), and they divide three times to form 32 A-line notochord cells. Pair B8.6 in the 110-cell stage embryo are also primordial notochord cells (Fig. 3D′), and they divide twice to form eight B-line notochord cells.

Expression: In situ hybridization signals were first seen in the A6.2 and A6.4 pairs of the 32-cell stage embryo (Fig. 3B,B′). These cells have the developmental fate to form notochord. Signals were then seen in the A7.3 and A7.7 pairs at the 64-cell stage (Fig. 3C,C′), and A8.5, A8.6, A8.13 and A8.14 pairs at the early gastrula stage (Fig. 3D,D′), although the signals became faint at the early gastrula stage (Fig. 3D,D′).

Function: To deduce the function of Cs-ZicL, we suppressed its translation by injection of morpholino into fertilized eggs. Injected embryos developed similarly to normal embryos, but the resultant tailbud embryos looked slightly flat, and the tail was not fully elongated (Fig. 4A′,B′). However, Cs-ZicL-suppressed embryos showed differentiation of the endoderm as assessed by histochemical detection of AP (100%, n=12; Fig. 4A,A′), of epidermis as assessed by the expression of Cs-Epi1 (100%, n=15; Fig. 4B,B′), and of mesenchyme as assessed by the expression of Cs-mech1 (97%, n=28; Fig. 4C,C′,D,D′), suggesting that the differentiation of endoderm, epidermis and mesenchyme is not dependent on Cs-ZicL function.

We next examined Cs-ZicL function in notochord differentiation. When we examined the differentiation of the notochord in embryos developed from eggs that had been injected with Cs-ZicL morpholino, the manipulated embryos failed to express notochord-specific Cs-Bra in A-line notochord cells (100%, n=35, data not shown).

When we examined the effects of functional suppression of Cs-ZicL on notochord formation with a probe for a notochord-specific structural gene, Cs-fibrn (Cs-fibrinogen-like) (Fig. 5A), the Cs-fibrn expression was greatly reduced, and only a few cells on both sides of the tail showed Cs-fibrn expression in experimental tailbud embryos (83% of embryos with a few cells expressing Cs-fibrn, 17% of embryos with no expression, n=23; Fig. 5A′). In order to determine whether differentiation of B-line notochord cells was affected by Cs-ZicL morpholino, we took advantage of ‘cleavage-arrest’, in which division of blastomeres is arrested with cytochalasin B but the differentiation of embryonic cells is not disturbed. In control embryos arrested at the 110-cell stage, eight A-line notochord and two B-line notochord cells expressed Cs-fibrn (Fig. 5B). In contrast, in Cs-ZicL morpholino-injected and 110-cell-stage arrested embryos, only B-line notochord cells, pair B8.6, expressed Cs-fibrn (77% of embryos with expression only in B-line notochord cells, 23% of embryos had no expression, n=17; Fig. 5B′).

To confirm the specificity of Cs-ZicL morpholino, a rescue experiment was preformed. Synthetic mRNA for Cs-ZicL lacking the Cs-ZicL morpholino recognition sequences (see Fig. 1) was co-injected into fertilized eggs together with Cs-ZicL morpholino. This co-injection rescued the expression of Cs-fibrn in the experimental embryos (65%, n=40; Fig. 5A′′). These results clearly indicate that Cs-ZicL function is essential for the differentiation of A-line notochord cells but not B-line notochord cells.

Possible cascade of Cs-ZicL function in notochord formation: The cDNA clone for Cs-ZicL was originally isolated as a β-catenin downstream gene. This suggests that the expression of Cs-ZicL is controlled by nuclear accumulation of β-catenin. To examine this issue, we injected in vitro synthesized Cs-cadherin mRNA (Imai et al., 2000) into fertilized eggs. Cadherin binds to cytoplasmic β-catenin, and thus downregulates β-catenin nuclear accumulation. As shown in Fig. 6B (arrowheads), the expression of Cs-ZicL in two pairs of A-line cells (A6.2 and A6.4) was markedly inhibited (80%, n=10). However, the expression of Cs-ZicL in two pairs of B-line cells (B6.2 and B6.4) was not blocked by Cs-cadherin mRNA injection (arrows in Fig. 6B). This indicates that the expression of Cs-ZicL in the A-line cells is regulated by β-catenin nuclear accumulation. However, it should be determined in future studies whether nuclear β-catenin directly activates Cs-ZicL, or whether nuclear β-catenin indirectly activates Cs-ZicL via some other molecules.

In this regard, we have shown that nuclear β-catenin directly activates a forkhead domain transcription factor gene, Cs-FoxD (K. S. I., unpublished). Cs-FoxD is expressed very transiently in A5.1, A5.2 and B5.1 of the 16-cell stage embryo (Fig. 3A′) and A6.1, A6.3 and B6.1 of the 32-cell stage embryo (Fig. 3B′), and the expression is downregulated by the 64-cell stage. This transient expression of Cs-FoxD is not associated with endoderm differentiation but is essential for the differentiation of the notochord. Because Cs-FoxD is an essential factor for notochord formation, it is possible that the expression of Cs-ZicL in the presumptive A-line notochord cells is controlled by Cs-FoxD. As we expected, when Cs-FoxD function was blocked by the morpholino, the expression of Cs-ZicL was not found in A6.2 or A6.4 notochord/nerve cord lineage cells (100%, n=19; Fig. 6C), although the expression of Cs-ZicL in B6.2 and B6.4 muscle lineage cells was not affected (Fig. 6C). This result suggests that Cs-ZicL is downstream of Cs-FoxD. If so, ectopic expression of Cs-FoxD might induce Cs-ZicL expression in cells where it is not normally expressed. As shown in Fig. 6D, microinjection of in vitro synthesized Cs-FoxD mRNA into fertilized eggs resulted in ectopic expression of Cs-ZicL (90%, n=10). The ectopic expression was often seen in the a-line lineage of the anterior animal hemisphere in most of the embryos. These results indicate that in endodermal cells, nuclear accumulation of β-catenin directly triggers the expression of Cs-FoxD, which in turn induces the expression of Cs-ZicL in notochord/nerve cord lineage cells.

The CNS of ascidian tadpole larvae consists of about 330 cells that include sensory receptor cells and neuronal cells in the so-called sensory vesicle (or brain), motoneurons in the visceral ganglion, and ependymal cells in the nerve cord (reviewed by Wada and Satoh, 2001; Meinertzhagen and Okamura, 2001). Cells of the sensory vesicle are derived from a-line cells, while cells of the b- and A-lines contribute to the formation of the visceral ganglion and the nerve cord.

Expression: In situ hybridization signals for Cs-ZicL expression in CNS-forming cells were first seen in the A6.2 and A6.4 pairs of the 32-cell stage embryo (Fig. 3B,B′), cells that have the developmental fate to form the nerve cord. Signals were then found in the A7.4 and A7.8 pairs at the 64-cell stage (Fig. 3C,C′). At the early gastrula stage, further zygotic expression of Cs-ZicL became evident in the a8.17, a8.19, a8.25, b8.17 and b8.19 pairs (Fig. 3D,D′), in addition to faint expression in the A8.7, A8.8, A8.15 and A8.16 pairs of the nerve cord cells (Fig. 3D,D′). The expression of Cs-ZicL in the CNS is evident in the mid-gastrulae (Fig. 3E) and in the anterior tip cells of neurulae (Fig. 3F). No signal was detected at the early tailbud embryo stage (Fig. 3G).

Function:Cs-ZicL is expressed in cells of the CNS, suggesting its role in the differentiation of CNS cells. Therefore, we examined the differentiation of the nervous system with a probe for Cs-ETR, which is a pan-neural marker of C. savignyi embryos (Fig. 7A). As seen in Fig. 7A′, the expression of Cs-ETR was greatly reduced in Cs-ZicL morpholino-injected embryos examined at the tailbud stage (100%, n=12). Cs-ZicL, therefore, has an important role in the differentiation of the nervous system.

During Ciona embryogenesis, 36 unicellular and striated muscle cells are formed in the larval tail: 18 cells on each side of the tail. Of these 36, 28 are derived from B-line cells, four from A-line and four from b-line cells. Regarding the B-line (or primary lineage), B6.2, B6.3 and B6.4 pairs in the 32-cell stage embryo are presumptive muscle cells. At the 64-cell stage, the B7.4 (a daughter cell of B6.2) and B7.8 (a daughter cell of B6.4) pairs are primordial muscle cells, while the B7.5 pair forms larval muscle and adult muscle.

Expression: The first in situ signals for Cs-ZicL expression were detected in the B6.2 and B6.4 pairs of the 32-cell stage embryo (Fig. 3B,B′). As mentioned above, these cells have the developmental fate to form muscle. Signals were next seen in the B7.3, B7.4, B7.7 and A7.8 pairs at the 64-cell stage (Fig. 3C,C′). In situ signals in these two muscle lines became faint by the gastrula stage (Fig. 3D,D′). The 64-cell stage embryo also expressed Cs-ZicL in the B7.5 pair (Fig. 3C,C′). Signals in B7.5 were strong at the early gastrula (Fig. 3D,D′) but became undetectable by the mid gastrula (Fig. 3E).

Function: The role of Cs-ZicL in the differentiation of muscle cells was examined by monitoring the expression of muscle actin gene Cs-MA1 (Fig. 8A,A′′). Because Cs-ZicL is expressed in all the B-line muscle cells, first in B6.2 and B6.4 at the 32-cell stage (Fig. 3B), and then in B7.5 at the 64-cell stage (Fig. 3C), it is highly likely that Cs-ZicL has a role in muscle cell differentiation. When we examined the expression of Cs-MA1 in Cs-ZicL morpholino-injected embryos at the tailbud stage, a certain number of muscle cells in experimental embryos expressed Cs-MA1 (100%, n=10; Fig. 8B′′). Because we could not judge the exact number of muscle cells expressing Cs-MA1, we checked the expression of Cs-MA1 in early-gastrula stage (Fig. 8B) and mid-gastrula stage embryos (Fig. 8B′). In both early-gastrula stage and mid-gastrula stage embryos, blastomeres derived from B6.2 (arrowheads in Fig. 8A,A′), which is located at the anterior-most position of all the B-line muscle cells, did not express Cs-MA1 (96% of embryos at the early-gastrula stage, n=24, and 87% of embryos at the mid-gastrula stage, n=39; Fig. 8B,B′). However, muscle cells derived from B6.4 (arrows in Fig. 8A,A′) expressed Cs-MA1 normally in Cs-ZicL-function-suppressed embryos (Fig. 8B,B′).

To examine whether this suppression is specifically caused by downregulation of Cs-ZicL function, we performed another rescue experiment. As in the case of notochord cells, co-injection of Cs-ZicL morpholino and Cs-ZicL synthetic mRNA rescued the expression of Cs-MA1 in the B6.2-derived muscle cells (72%, n=50; Fig. 8C).

Possible cascade of Cs-ZicL function in muscle formation: Recently, Nishida and Sawada (Nishida and Sawada, 2001) isolated and characterized a Zic-like gene named macho-1 from Halocynthia roretzi (see Fig. 2). macho-1 is expressed only maternally and its mRNA shows a segregation pattern characteristic of the myoplasm. When macho-1 function is suppressed with antisense oligonucleotides, B-line muscle cell differentiation is blocked. In contrast, when the gene is overexpressed by injection of synthetic mRNA, muscle cells are formed ectopically. Because macho-1 is considered to be a muscle determinant, it is possible that Cs-ZicL is downstream of macho-1. Previous characterization of Cs-macho1 of Ciona savignyi showed that its maternal expression pattern coincides with that of macho-1, although Cs-macho1 is also expressed in the CNS zygotically (Satou et al., 2002).

When the expression of Cs-ZicL was examined in Cs-macho1 morpholino-injected 32-cell stage embryos, it was evident that Cs-ZicL expression in B-line muscle cells was not affected (100%, n=21; Fig. 9A). This indicates that the expression of Cs-ZicL is controlled independently of Cs-macho1.

Our previous study indicated that suppression of Cs-macho1 function resulted in blockage of Cs-MA1 expression at the cleavage and early gastrula stages, but the blockage was not complete, and later transcripts of Cs-MA1 appeared in muscle cells at the mid-gastrula and later stages (Satou et al., 2002). Therefore, we first re-examined this issue. As shown in Fig. 8D, the expression of Cs-MA1 at the early-gastrula stage was almost completely inhibited in Cs-macho1 morpholino-injected embryos (93%, n=26). However, Cs-MA1 expression recovered in embryos by the mid-gastrula stage (80%, n=39; Fig. 8D′). This suggests that in Ciona embryos, pathway(s) independent of Cs-macho1 are also responsible for muscle differentiation, and Cs-ZicL may be involved in these pathways. To examine this issue, we injected both Cs-macho1 and Cs-ZicL morpholinos together into fertilized eggs. As shown in Fig. 9B,C, this injection completely suppressed the expression of Cs-MA1 at the mid-gastrula stage (100%, n=26) and tailbud stage (100%, n=44). These results demonstrate that both Cs-macho1 and Cs-ZicL are essential for muscle differentiation in Ciona embryos, and there are at least two independent genetic cascades for muscle differentiation, one that is Cs-macho1-dependent, and another that is Cs-ZicL dependent.

The present study demonstrated that Cs-ZicL of Ciona savignyi encodes a Zic-like zinc finger protein whose function is associated with the differentiation of three different types of embryonic cells: A-line notochord, CNS, and B-line muscle cells. During early embryogenesis, Cs-ZicL is expressed in blastomeres that give rise to these three types of tissue. The present study also demonstrated that the expression of Cs-ZicL in the A-line cells is downstream of β-catenin/Cs-FoxD, while the expression of Cs-ZicL in the B-line muscle cells is independent of another Zic-like gene, Cs-macho1.

In ascidian embryos in situ hybridization signals for zygotic gene expression are usually first detected in the nucleus, and the signals become distributed throughout the cytoplasm as development proceeds (e.g., Yasuo and Satoh, 1993; Satou et al., 1995). Therefore, we can judge the timing of the gene expression as well as cells exhibiting the gene expression. Based on such criteria, it can be said that the zygotic expression of Cs-ZicL takes place in at least three embryonic domains. The signals were first evident in the nuclei of A-line notochord/nerve cord cells at the 32-cell stage (Fig. 3). At the 64-cell and 110-cell stages, the signals were seen in the cytoplasm of this lineage, but the expression was downregulated by the mid-gastrula stage. The fact that injection of cadherin mRNA into fertilized eggs suppressed the Cs-ZicL expression in the A-line notochord/nerve cord cells suggests that Cs-ZicL expression there is controlled by the nuclear accumulation of β-catenin. There are two possibilities regarding the relationship between β-catenin and Cs-ZicL expression: either β-catenin together with Tcf/LEF directly activates the Cs-ZicL expression or β-catenin indirectly activates the Cs-ZicL expression via other molecules. Here we examined the latter possibility by asking whether Cs-FoxD, which is a direct target of the nuclear accumulation of β-catenin (K. S. I., unpublished), communicates between the two molecules. Injection of Cs-FoxD morpholino into fertilized eggs resulted in the failure of Cs-ZicL expression, and injection of Cs-FoxD mRNA into fertilized eggs resulted in ectopic expression of Cs-ZicL. Therefore, it is highly likely that Cs-ZicL is downstream of Cs-FoxD. However this does not exclude the first possibility. It is possible that a combinational regulation by β-catenin and Cs-FoxD (or its target gene) is required for the expression of Cs-ZicL.

At the 16-cell stage, Cs-FoxD is expressed in A5.1, A5.2 and B5.1 (Fig. 3A′), while Cs-ZicL is not expressed at this stage. At the 32-cell stage, Cs-FoxD is expressed in A6.1, A6.3 and B6.1, while Cs-ZicL is expressed in A6.2, A6.4, B6.2 and B6.4 (Fig. 3B′). Apparently, the expression of these two genes does not overlap in the 32-cell stage embryo. However, A6.2 and A6.4, which express Cs-ZicL, are daughter cells of A5.1 and A5.2, respectively, which express Cs-FoxD in the 16-cell stage embryo. Therefore, it is likely that Cs-FoxD expression in the 16-cell stage embryo may trigger the activation of Cs-ZicL in one daughter cell that gives rise to A6.2 or A6.4, although FoxD proteins are known to be as a transcriptional repressor in Xenopus (Pohl and Knöchel, 2001; Sullivan et al., 2001).

An alternative explanation may be that Cs-FoxD expression in A6.1 and A6.3 at the 32-cell stage induces Cs-ZicL expression in A6.2 and A6.4 through cell-cell communication. This possibility is discussed below. In Halocynthia embryos, it has been shown that interaction between the A6.1/A6.3 (primordial endoderm cells) and A6.2/A6.4 (cells with developmental fate to give rise to notochord) takes place in the latter half of the 32-cell stage (Nakatani and Nishida, 1994), and this interaction activates Brachyury expression in the primordial notochord cells at the 64-cell stage (Nakatani et al., 1996). In ascidians, Brachyury is a key regulatory gene for notochord formation (Yasuo and Satoh, 1993; Yasuo and Satoh, 1998). Therefore, molecular events that take place at the 16-cell or 32-cell stages and to eventually activate Brachyury expression are key to understanding the mechanisms of notochord induction. In a previous study, we characterized Cs-FGF4/6/9 cDNA as a possible notochord inducer, and showed that the corresponding gene is expressed in the endoderm and notochord cells at the 32-cell stage (Imai et al., 2002). Cs-FGF4/6/9 is essential for the induction of mesenchyme. Although the expression of Cs-fibrn was reduced in Cs-FGF4/6/9 morpholino-injected embryos, the function of Cs-FGF4/6/9 in the notochord induction was partial. In addition, Cs-FGF4/6/9 is not downstream of Cs-FoxD nor upstream of Cs-ZicL, suggesting that the function of Cs-FGF4/6/9 in the formation of the notochord is independent of the Cs-FoxD/Cs-ZicL cascade. Therefore the notochord inducer downstream of Cs-FoxD is still unknown. Whether Cs-FoxD activates the expression of Cs-ZicL directly or via an unknown inducer should be determined in future studies. It should be noted here that Cs-ZicL is necessary for A-line notochord formation, but is not involved in B-line notochord formation. Cs-FoxD is expressed in both A-line and B-line endoderm cells and its function is essential for the differentiation of both A-line and B-line notochord cells. Therefore, in the formation of the B-line notochord, a different genetic cascade, which does not include Cs-ZicL, exists between the expression of Cs-FoxD and Cs-Bra.

Cs-ZicL is expressed in two anterior B-line muscle cells, B6.2 and B6.4, at the 32-cell stage, and later at the 64-cell stage the gene is expressed in posterior B-line muscle cells, B7.5 (Fig. 3). Injection of Cs-ZicL morpholio resulted in the failure of Cs-MA1 expression in B6.2-line muscle cells (Fig. 8), suggesting that the initiation of the muscle-specific structural gene in B6.2 is controlled by Cs-ZicL.

Cs-macho1 is a C. savignyi homolog of Halocynthia macho-1, which is a maternal muscle determinant gene (Nishida and Sawada, 2001). The maternal transcript of Cs-macho1 is distributed like a posterior end mark (Satou et al., 2002); it is localized to the posterior-most blastomeres throughout early embryogenesis (Yoshida et al., 1996; Satou and Satoh, 1997). However, Cs-macho1 protein is expected to be distributed in all the B-line muscle cells, because injection of Cs-macho1 morpholino blocks the initiation of Cs-MA1 expression in all of the B-line muscle cells (Fig. 8D). Therefore, the distribution of the products of the Cs-macho1 and Cs-ZicL genes may be overlapping in B-line muscle cells. As mentioned above, embryos injected with Cs-macho1 morpholino cannot initiate the expression of Cs-MA1 in any of the B-line muscle cells, while embryos injected with Cs-ZicL morpholino failed to initiate Cs-MA1 expression only in B6.2-line muscle cells. This suggests that the activity of Cs-macho1 is required for the initiation of Cs-MA1 expression in B6.4 and B7.5, and the combined activity of Cs-macho1 and Cs-ZicL is sufficient for the initiation of Cs-MA1 expression in B6.2.

However, even if the initiation of Cs-MA1 expression is blocked by Cs-macho1 morpholino, transcripts of Cs-MA1 are later detected in muscle cells, suggesting that in Ciona embryos, Cs-macho1 is required but not sufficient for Cs-MA1 expression. In other words, the initiation of Cs-MA1 expression is governed by Cs-macho1 but the activity of other genes such as Cs-ZicL is required for the Cs-MA1 expression. Actually, muscle cells are not formed when the function of both genes is suppressed with morpholinos, indicating that Cs-macho1 is not the only muscle determinant and there are other muscle determinants govering the expression of Cs-ZicL in muscle cells in Ciona embryogenesis. In Halocynthia, macho-1 is essential and sufficient for the muscle cell differentiation (Nishida and Sawada, 2001), suggesting that determination mechanism of the muscle cells are slightly different in these two species. It will be interesting to study how these two different mechanisms were evolved.

Cs-macho1 and Cs-ZicL share highly similar zinc finger domains. It is also of interest to ask what are their own specific functions, and how do they cooperate with each other. One possibility is that these two factors recognize the same binding sequece and control the same genes. In this case, Cs-ZicL seems to work as a backup factor of Cs-macho1. Another possibility is that Cs-ZicL and Cs-macho1 do not share their recognition sequences. Our preliminary results showing these two factors have different recognition sequences (K. Yagi, N. S. and Y. S., unpublished data) support this hypothesis. In this latter case, two different pathways using Zic-like factors are working in muscle differentiation.

Cs-ZicL is expressed in embryonic domains that give rise to three different cell types. The first is A-line notochord/nerve cord cells at the 32-cell stage, the second is B-line muscle cells at the 32-cell stage, and the third is CNS cells at the 110-cell stage. In addition, Cs-ZicL expression in the two lineages at the 32-cell stage is regulated by independent mechanisms. The expression of Cs-ZicL in the A-line notochord/nerve cord cells at the 32-cell stage is downstream of β-catenin/Cs-FoxD but independent of Cs-macho1. The expression of Cs-ZicL in the B-line muscle cells at the 32-cell stage is not regulated by β-catenin/Cs-FoxD, and is also independent of Cs-macho1. It is very important to determine the cis-regulatory elements of Cs-ZicL, and this analysis is now being conducted. Cs-ZicL has at least three different regulatory functions in embryogenesis: the differentiation of the notochord, the central nervous system and muscle cells. This means that the same factor can work in different ways depending on the context of the cells. Therefore, Cs-ZicL provides a good experimental system for studying how the same factor recognizes different targets in different cell lineages.

The molecular phylogenetic analysis based on the comparison of amino acid sequences of the zinc finger domains demonstrated that mouse and Xenopus Zic-related genes form one clade, from which Cs-macho1, macho-1 and Cs-ZicL are distant (Fig. 2B). Blast search indicated that the zinc finger domains of Cs-ZicL and Cs-macho1 show the highest similarity to those of vertebrate Zic family gene products. In addition, a Ciona intestinalis EST project has characterized nearly 150 zinc finger transcription factor genes (L. Yamada, N. S. and Y. S., unpublished data), and only Ci-ZicL and Ci-macho1 have high similarity to vertebrate Zic family genes. This indicates that Cs-ZicL and Cs-macho1 and vertebrate Zic might have originated from a common ancestral gene. In ascidians, the expression and function of this gene diverged as macho-1 for a maternal muscle determinant gene, while ZicL diverged as a zygotic gene with multiple functions.

The expression pattern of Cs-ZicL resembles that of vertebrate Zic. It has been reported that the expression and function of vertebrate Zic genes are mainly associated with the nervous system. However, some vertebrate Zic genes are also expressed in embryonic mesoderm. For example, Xenopus Zic3 is expressed in involuting mesoderm at the early gastrula stage (Kitaguchi et al., 2000), and mouse Zic1, Zic2, and Zic3 are expressed in embryonic mesoderm at the early primitive streak stage (Nagai et al., 1997). Because the present study demonstrated the function of Cs-ZicL in mesoderm formation, the function of vertebrate Zic genes in mesoderm formation should be addressed in future studies.

We are thankful to Dr Yasuaki Yakagi, 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 form 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)