Morpholino-based gene knockdown screen of novel genes with developmental function in Ciona intestinalis

In vertebrates, complete genome sequences are now available for human(International Human Genome Sequencing Consortium, 2001; Venter et al., 2001), mouse (Mouse Genome Sequencing Consortium, 2002) and Fugu rubripes(Aparicio et al., 2002). These genome sequences and extensive analyses of expressed sequence tags (ESTs),mostly derived from humans and mice, have demonstrated that there are many protein-coding genes, and biological function has been assigned to only a small proportion of the predicted genes. A considerable number of the remaining uncharacterized genes are thought to have special biological roles. One of the key goals of developmental genomics, therefore, is to determine the developmental function of such genes.

Urochordate ascidians are marine invertebrate chordates that have a common ancestor shared by the cephalochordate amphioxus and vertebrates. The organization of ascidian tadpole larvae shows basic features of the chordate body plan (Satoh, 1994; Corbo et al., 2001; Satoh, 2003). The ascidian genome has been proposed to contain a basic set of chordate-type genes corresponding to those present before the large-scale gene duplications occurred in the lineage leading to vertebrates(Holland et al., 1994), and thus ascidian genes have less functional redundancy than vertebrate genes. Recently, the draft genome sequence of the most-studied ascidian, Ciona intestinalis, has been reported (Dehal et al., 2002). Its 159 Mbp genome (17 times smaller than that of humans) contains 15,852 protein-coding genes, similar to the number in other invertebrates but only half that found in vertebrates. Vertebrate gene families are typically found in simplified form in the Ciona intestinalis genome, supporting the idea that ascidians contain the basic ancestral complement of genes involved in cell signaling and development(Dehal et al., 2002; Satou et al., 2003a; Wada et al., 2003; Yagi et al., 2003). In vertebrates, the function of a gene is often obscured by functional redundancy of the related genes. Therefore, Ciona embryos may provide an appropriate experimental system for exploring the functions of genes(Satoh et al., 2003).

The Ciona intestinalis cDNA project consortium has conducted comprehensive studies of gene expression profiles in fertilized eggs(Nishikata et al., 2001),cleavage stage embryos (Fujiwara et al.,2002), tailbud embryos (Satou et al., 2001b), larvae(Kusakabe et al., 2002) and young adults (Ogasawara et al.,2002). A total of more than 480,000 ESTs have been sequenced, and the spatial expression profiles of 5000 randomly selected genes have been determined (Satou et al.,2002a; Satoh et al.,2003). The genes have been classified into three major classes according to the function of the proteins they encode(Lee et al., 1999). Class A includes genes associated with functions in many cell types, class B contains genes associated with cell-cell communication and class C contains genes that function as transcription regulatory proteins. Besides genes that fall into these three classes, there are many genes for which not enough information is available to determine their biological function. These genes have been categorized into two classes, class DI and class DII. Class DI includes sequences that match ESTs or reported proteins with unknown function (mostly from humans and mice), and class DII consists of genes with no significant sequence similarity to known genes. Ciona intestinalis cDNA analysis suggests that at least 2500 genes are classified into class DI, and that nearly one quarter of them show specific spatiotemporal expression patterns. Most DI-class genes are assigned counterparts of vertebrate genes with unknown function. Therefore, a comprehensive functional analysis of Ciona intestinalis DI-class genes would provide significant information relevant to vertebrate genes with unknown function. We therefore decided to systematically investigate the loss-of-function phenotypes of DI-class genes,starting with 200 genes in the present pilot screen.

In addition to the small size of their genome and the small number of genes, ascidians provide a simple experimental system for investigating the molecular mechanisms underlying cell-fate specification during chordate development (Satoh, 1994; Satoh, 2001). The Ciona intestinalis fertilized egg develops within 18 hours into a tadpole larva, through invariant bilateral cleavage, gastrulation, neurulation and tailbud formation. The lineages of embryonic cells are invariant among individuals and have been well documented(Conklin, 1905; Nishida, 1987). The tadpole larva consists of ∼2600 cells that form distinct types of tissues and/or organs, including epidermis that covers the entire surface; a central nervous system with two sensory pigment cells; endoderm and mesenchyme in the trunk and notochord; dorsal nerve cord; ventral endodermal strand; and muscle in the tail (Fig. 1A). The Ciona intestinalis cDNA project has provided us with genes expressed in a tissue-specific manner for almost all types of tissue(Satou et al., 2001b; Kusakabe et al., 2002). These genes can be used as tissue-specific markers. Furthermore, over 5600 full-length sequences of cDNAs have already been determined(Satoh et al., 2003). These circumstances make Ciona intestinalis embryos a suitable system for a reverse genetic approach for identifying novel genes with developmental function.

For many years, mutant screens in the invertebrates C. elegans and Drosophila have provided us with functional information on many genes. However, reverse genetic approaches have become increasingly attractive, because they offer the possibility of taking interesting candidate genes from genome/EST databases and speedily identifying their roles. RNA-based interference (RNAi) is often useful for inhibiting gene function in C. elegans and Drosophila(Hannon, 2002). However, there are no reports yet showing that RNAi works well in ascidians. In ascidians,antisense phosphorothioate DNA oligonucleotides (S-DNAs) have been used to block embryonic gene function (Swalla and Jeffery, 1996; Nishida and Sawada, 2001). S-DNAs form RNA-DNA hybrids with the target mRNA and act as substrates for degradation of the mRNA by RNAse H(Cazenave et al., 1989). However, generally speaking, S-DNAs have nonspecific toxic side effects on embryos and therefore their specific effects are not always distinguishable. Recently, another strategy for inhibiting gene function has been shown to be extremely effective. Morpholino antisense oligonucleotides (hereafter referred as only `morpholinos') have been shown to block translation by preventing ribosomes from binding to target mRNA(Summerton, 1999). Morpholinos cause less toxicity and effectively enable the study of maternal and zygotic gene functions in a range of model organisms(Heasman, 2002). This is also the case in ascidian embryos (Satou et al., 2001a; Wada and Saiga,2002). Injection of a morpholino into eggs specifically targets a certain gene by suppressing translation and thereby gene function. In zebrafish embryos, it has been reported that the morpholino-induced phenotype is the same as that of a mutant of the targeted gene, although the penetrance of the phenotype is variable (Heasman,2002). Morpholinos therefore offer a high and rapid throughput assay system for chordate functional genomics applications.

In the present study, we conducted an extensive analysis of genes discovered in the Ciona intestinalis cDNA project. To identify genes with developmental function, translation of a total of 200 genes expressed in embryos was suppressed using specific morpholinos. Most of these 200 genes were classified into class DI, genes whose vertebrate counterparts have unknown function. The suppression of the translation of any of 40 of these genes (20% of the genes examined) caused specific embryonic defects. These results demonstrate that the Ciona embryo is a powerful tool for identifying developmental genes essential for formation of the chordate body plan. Furthermore, the present study might represent the first example of a screen using morpholinos as a reverse genetic tool to identify genes required for chordate development.

Ciona intestinalis were cultivated at the Maizuru Fisheries Research Station of the Kyoto University, Maizuru, and the Education and Research Center of Marine Bio-Resources of Tohoku University, Miyagi, Japan. Eggs and sperm were obtained surgically from the gonoduct. After insemination,eggs were dechorionated by treatment with actinase E and sodium thioglycolate as described previously (Satou et al.,2001b). Dechorionated fertilized eggs were microinjected with morpholinos as described below. After microinjection, they were maintained in agar-coated dishes with Millipore-filtered seawater containing 50 mg/ml streptomycin sulfate at 18°C. They developed into tailbud embryos and tadpole larvae by about 13 hours and 18 hours of development,respectively.

For the 200 genes examined in the present study, clones of the corresponding cDNA that encoded the full length of the putative protein were isolated and sequenced as described previously(Satou et al., 2002b). In all of these cDNA clones, there are in-frame stop codon(s) upstream of the putative initiation codon, which supports the validity of the prediction of the initiation codon. Motifs in the putative proteins were searched with SMART(http://smart.emblheidelberg.de/). Homology searches were performed against the DDBJ nucleotide database with BlastX, and gene sequences with P values less than 10^-15were selected as homologs. Temporal expression profiles of genes were determined based on the results of the EST analyses(Satou et al., 2002a). A list of the 200 genes including gene collection IDs, clone IDs (used as the name of each gene here), Accession Numbers of cDNA sequences and nucleotide sequences of the corresponding morpholino are provided in Table S1. EST information and the cDNA sequences of the 200 genes are available in our web site(http://ghost.zool.kyoto-u.ac.jp/indexr1.html).

Morpholinos were made to order (Gene Tools, LLC). For most of the 200 genes, a complementary morpholino was designed to target a sequence containing the putative initiation codon. For the other genes, a morpholino was designed to target a sequence of the 5′ untranslated region (Table S1). Microinjection was carried out as described previously(Satou et al., 2001a). The amount of morpholino injected was 10 fmoles, an amount chosen based on the results of previous reports (Satou et al.,2001a) and our preliminary experiments. Ten fmoles of morpholinos dissolved in distilled water were injected into fertilized eggs and the effects on morphology were determined using stereomicroscopy about 18 hours later at the tadpole larva stage. In cases of the use of a second form of citb018a19 morpholino, 30 fmoles were injected into fertilized eggs. As controls, 10 or 30 fmoles of a morpholino against lacZ were injected,but these had no effect on embryogenesis, provided excess amounts were not injected.

Two researchers simultaneously, but independently, injected morpholinos into 25 or more fertilized eggs. When they obtained the same result, it was scored. When the results of the two researchers were inconsistent, a third researcher performed another injection to reach a final conclusion.

For rescue experiments, the protein-coding region of Ci-Bra,cieg003h01 or citb018a19 that lacked a target sequence for morpholino was amplified by PCR and cloned into pBluescriptRN3 vector. Capped mRNA was synthesized in vitro using a Megascript T3 kit (Ambion) together with the cap analog 7mGpppG. Because the synthesized mRNA lacks the target sequence for the corresponding morpholino, translation from it is not inhibited by the morpholino. The fertilized eggs were injected with 5 pg (Ci-Bra) or 25 pg (cieg003h01 and citb018a19) of the synthesized mRNA together with 10 fmoles of the corresponding morpholino, allowed to develop up to the tailbud embryo stage, and examined for marker gene expression or morphology.

Whole-mount in situ hybridization was carried out as described previously(Satou et al., 2001b). The probes used were Ci-Epi1 for epidermal cell differentiation(Chiba et al., 1998), Ci-ETR for nervous system development(Satou et al., 2001b), Ci-talin for notochord differentiation(Satou et al., 2001b; Sasakura et al., 2003) and ciad005j06 for mesenchyme cell differentiation(Satou et al., 2001b). Probes for Ci-arr (Nakagawa et al.,2002), Ci-opsin1(Kusakabe et al., 2001), Ci-opsin3 (Nakashima et al.,2003) and Ci-Gαi1a(Yoshida et al., 2002) were also used for experiments with the citb018a19 gene.

Histochemistry for endoderm alkaline phosphatase (ALP) and muscle acetylcholine esterase (AchE) was performed as described(Imai et al., 2000).

In the present study, the function of 200 Ciona intestinalis genes was examined using morpholinos. The genes were chosen from a collection identified through the Ciona intestinalis cDNA project(Satou et al., 2002b). Of them, 144 genes were selected because they belong to class DI (see Table S1). The rest of the genes were selected because they exhibit interesting features in structure and/or expression pattern. Although the 200 genes encode a wide spectrum of proteins with various structures, about one-third of them consist of genes for Zn-finger proteins, which is concurrent with the fact that a number of this type of proteins have been shown to play important roles in development (see Table S1). Most of the 200 genes are both maternally and zygotically transcribed, as shown by EST evidence(http://ghost.zool.kyoto-u.ac.jp/indexr1.html).

Translational suppression of each gene was achieved by microinjection of a corresponding morpholino into fertilized eggs, and the resultant embryos were examined for any defect in larval morphology to identify genes required for normal development. When morpholino against lacZ mRNA was injected as a negative control, all of the injected embryos developed normally (data not shown). However, injection of morpholino complementary to Ci-Bra, an essential regulator of notochord development in Ciona embryos(Corbo et al., 1997; Takahashi et al., 1999), at the same dose led to a shortened tail phenotype in almost all injected embryos(Fig. 1B). Expression of Ci-talin, a notochord marker (Fig. 1C) was lost in the morpholino-injected embryos(Fig. 1D). These findings are consistent with the proposed role of Ci-Bra in notochord formation. The specificity of the action of Ci-Bra morpholino was assessed by a rescue experiment with synthetic Ci-Bra mRNA. Co-injection of Ci-Bra morpholino and Ci-Bra mRNA abrogated the loss of Ci-talin expression caused by injection of Ci-Bra morpholino(Fig. 1E).

A gene was considered to be positive when half or more of the specimens within the group of injected embryos exhibited developmental deficiency. Out of the 200 genes examined, 40 genes were judged to be positive by this criterion (Table 1; information on the 40 positive genes is summarized in Table 2). The remaining 160 genes were judged to be tentatively negative and excluded from the subsequent analysis. It has been reported that, in zebrafish embryos, the penetrance of the effects of morpholinos is variable and much lower than 100% in many cases(Heasman, 2002). Similarly, in our experimental system, the efficiency of morpholinos judged to be positive was variable depending on the morpholino. We estimated that the efficiency(the rate of embryos showing a developmental defect in the injected embryos)was in the range 60-100%, depending on the morpholino, although the reason for this variation is unknown. In Table 2, morpholinos with efficiency of 80% or higher were considered to have strong penetrance, while those with lower efficiency were considered to have weak penetrance.

Phenotypes obtained in the present screen were classified into two major classes, the `disorganized body plan' class and the `abnormal tail' class(Table 1). The first major phenotypic class, the `disorganized body plan' class, was characterized by gross morphological defects. Of 40 genes with a morpholino-induced phenotype,14 genes belonged to this phenotypic class (Tables 1 and 2; Fig. 2). Neither gastrulation nor neurulation occurred normally during the development of these embryos. In severe cases, even the cleavage pattern became atypical. The morphological features of the larvae of the `disorganized body plan' phenotypic class was varied depending on the morpholino (Fig. 2).

The second major phenotypic `abnormal tail' class was characterized by a tail with abnormal morphology. Of the 40 genes for which we detected a morpholino-induced phenotype, 26 genes gave rise to this phenotypic class(Tables 1 and 2; Fig. 3). Embryos showing a phenotype of this class were further subdivided into two groups according to the length of their tail. Embryos of the `short tail' phenotypic group had a tail whose length was less than half that of the normal tail. Of the 26 genes in the `abnormal tail' class, eight had this phenotype. For example, embryos injected with Ci-Bra morpholino(Fig. 1B) were grouped into this phenotypic class. Embryos with a longer tail were categorized into the`bent tail' phenotypic group because of their kinked tail (e.g. Fig. 3A). Eighteen genes were associated with this phenotype.

Some embryos showing the `abnormal tail' phenotype also exhibited two prominent phenotypes related to the formation of the nervous system. One was the `loss of the sensory pigment cells' phenotype (e.g. Fig. 3C), which was displayed by three genes associated with the `short tail' phenotype and six genes with the `bent tail' phenotype. The other was the `loss of the palps' phenotype(e.g. Fig. 3D), which was displayed by two genes that were associated with the `bent tail' phenotype and one gene that also gave the `loss of the sensory pigment cells' and `short tail' phenotypes. The palps are an adhesive organ located at the anterior tip of the larva, an organ containing peripheral sensory neurons essential for signal transmission for metamorphosis.

To provide an overview of the outcome of the present screen and to illustrate the morphological deficiencies of embryos, several interesting examples are described in the following two sections. Fourteen genes were associated with the `disorganized body plan' phenotypic class(Table 1). They consisted of 12 DI-class genes, including eight genes for Zn-finger proteins, one DII-class gene and one C-class gene that is a Ciona intestinalis homologue of Pbx (Table 2). Several intriguing examples in this phenotypic class are described.

cieg015c02 encodes a protein that contains C2H2 type Zn-finger motifs and shows sequence similarity to the human hypothetical protein XP 030938(Table 2). ESTs for this gene were detected in fertilized eggs and cleaving embryos. Eggs injected with its morpholino developed into rough-surfaced aggregates of cells(Fig. 2A). They were not enclosed by a smooth-surfaced epidermis layer. The cells that comprised them were loosely connected, so that the aggregates were very fragile. In spite of such extreme abnormality, the expression of markers for some endodermal and mesodermal tissues was detected in them, suggesting that they did not simply expire. Rather, it seems likely that a mechanism involved in cell adhesion or epidermis formation is affected in the cieg015c02 knockdown embryos.

cieg040a13 encodes a protein that has a Zpr1 motif(Galcheva-Gargova et al.,1998) and is similar to the human Zn-finger protein 259(Table 2). ESTs for this gene were scored in fertilized eggs, cleaving embryos and young adults. Eggs injected with morpholino complementary to this gene developed into deformed embryos (Fig. 2B). The size of the cells constituting these embryos was significantly larger than that of any cells comprising normal embryos, suggesting that the process of cell division was affected in these embryos.

Two distinct genes, cicl045c22 and cieg003h01, gave similar, remarkable phenotypes. cicl045c22 encodes a protein that contains C2H2 type Zn-finger motifs and is similar to the human hypothetical protein XP_ 041139(Table 2). ESTs for this gene were detected at all the developmental stages examined. Eggs injected with cicl045c22 morpholino or cieg003h01 morpholino developed into flattened larvae with similar appearance (Fig. 2C,D). Here, the results of suppression of cieg003h01 function are described in detail.

cieg003h01 encodes a protein that has multiple putative transmembrane motifs and is similar to the human hypothetical protein FLJ12377(Table 2) and its homologs are present in the genomes of mice, flies, worms, plants and yeasts(Table 2). ESTs for this gene were detected at the stages of fertilized egg, cleaving embryo and larva, and its maternal message is distributed evenly in the egg cytoplasm (data not shown). Eggs injected with cieg003h01 morpholino developed into flattened larvae with trunk-like and tail-like regions(Fig. 2D, Fig. 4D), although the cleavage pattern became atypical as early as the eight-cell stage. The trunk-like region appeared to comprise a layer of epidermal cells with no visible development of the endoderm and nervous system(Fig. 2D). Whole-mount in situ hybridization revealed the expression of Ci-Epi1, an epidermal cell differentiation marker (Fig. 4A,B), and histochemistry demonstrated the production of muscle AchE (Fig. 4C,D). In ascidian embryos, the differentiation of epidermis and muscle takes place autonomously dependent on maternally provided egg cytoplasmic information (e.g. Nishida, 2002). This suggests that the genetic cascades leading to the differentiation of epidermis and muscle are independent of cieg003h01. However, in the trunk-like region of experimental embryos, neither the endoderm assessed by histochemical detection of ALP (Fig. 4K,L) nor the nervous system assessed by Ci-ETR expression(Fig. 4G,H) differentiated. In the tail-like region, the differentiation of notochord was inhibited(Fig. 4I,J) and that of mesenchyme was partially suppressed (Fig. 4E,F). The differentiation of notochord and mesenchyme is dependent on the endoderm in Ciona embryos(Imai et al., 2000; Imai et al., 2002a). The specificity of the action of cieg003h01 morpholino was assessed by a rescue experiment with synthetic cieg003h01 mRNA. Co-injection of the morpholino and the mRNA abrogated the loss of ALP expression caused by the injection of cieg003h01 morpholino (Fig. 4N). In some cases, the co-injection resulted in the formation of larvae with a deformed but distinct tail region(Fig. 4M).

Interestingly, the morphology of larvae injected with cieg003h01 morpholino or cicl045c22 morpholino was similar to that of Ciona larvae injected with morpholino that targets β-catenin, an essential regulator of the animal-vegetal specification of the ascidian embryo(Imai et al., 2000). Furthermore, the profile of differentiation of the six tissues we examined was identical in cieg003h01 morpholino-injected embryos and β-catenin morpholino-injected embryos. These findings raise the interesting possibility of a relationship between cieg003h01, cicl045c22 and β-catenin.

cilv008a08 is the single Ciona intestinalis counterpart of vertebrate Pbx genes (Wada et al.,2003) (Table 2). ESTs for this gene were detected in all the developmental stages examined(Table 2). Eggs injected with cilv008a08 morpholino developed into ball-shaped larvae(Fig. 2E). Pbx proteins are TALE class homeodomain transcription factors that act as cofactors for Hox class homeodomain transcription factors and play multiple roles during development according to their partner proteins(Mann and Chan, 1996). The severe defects observed in the cilv008a08 morpholino-injected embryos suggest that cilv008a08 may play a similar crucial role.

In addition to the genes described above, nine genes gave the `disorganized body plan' phenotype. Although they are not described in detail here, embryos injected with morpholino against each of them are shown in Fig. 2F-N.

Of the 40 positive genes, 26 genes belonged to the `abnormal tail'phenotypic class. They consisted of 15 DI-class genes, six DII-class genes,two A-class genes, two B-class genes and one C-class gene, and included six genes encoding Zn-finger proteins (Table 2).

cieg022g05 encodes a protein that has four SANT motifs(Aasland et al., 1996) and shows sequence similarity to the human SNAP190 protein, the function of which is unknown (Table 2). ESTs for this gene were detected in all the developmental stages examined. Eggs injected with cieg022g05 morpholino developed into larvae that had bent tails and lacked the sensory pigment cells (Fig. 3A). Remarkably similar phenotypes were obtained when eggs were injected with morpholino that targets cieg015b23, which encodes a protein similar to the human KIAA1629 protein (Fig. 3B), or cicl057g13, which encodes a protein similar to the human hypothetical protein FLJ12890 (Fig. 3C), although the average length of the tail varied depending on the morpholino. On the other hand, injection of a morpholino designed against cilv006d23, which encodes a protein with a BTB motif(Zollman et al., 1994) that is similar to the human hypothetical protein MGC2628(Table 2), also led to larvae with a bent tail; the larvae developed the sensory pigment cells but lacked the palps (Fig. 3D).

cicl002e04 encodes a protein that belongs to the Zic family of C2H2-type zinc-finger proteins (Table 2). ESTs for this gene were detected in embryos at the cleavage, gastrula/neurula and tailbud stages. Zic family genes play crucial roles in development in a wide range of animals (Nagai et al.,1997). Eggs injected with cicl002e04 morpholino developed into larvae that had short tails and had palps but no sensory pigment cells(Fig. 3E). Recently, Zic family genes were isolated from Ciona savignyi (Cs-ZicL)(Imai et al., 2002b) and Halocynthia roretzi (HrzicN)(Wada and Saiga, 2002), and translational suppression experiments using a specific morpholino showed that larvae injected with Cs-ZicL or HrzicN morpholino exhibited defects in notochord and neural tube development. A molecular phylogenetic analysis of ascidian Zic family genes suggests that cicl002e04 is one of the multiple Ciona intestinalis counterparts of Cs-ZicL and HrzicN (Yamada et al.,2003). Consistent with this, the knockdown phenotypes reported for Cs-ZicL and HrzicN were similar not only to each other but also to those of animals injected with cicl002e04 morpholino.

Here, we describe in detail the results obtained with citb018a09. citb018a09 encodes a protein that has a Ring finger motif and a WWE domain,which shows sequence similarity to the human hypothetical protein DKEZp434O1427 with unknown function (Table 2). ESTs for this gene were detected in all the embryological stages but not in the adult (Table 2). Eggs injected with citb018a09 morpholino (a form that was produced against 25 nucleotides, including the first ATG codon) developed into larvae that had lacked the sensory pigment cells(Fig. 5B). Whole-mount in situ hybridization and histochemistry revealed the occurrence in the morpholino-injected embryos of differentiation markers for epidermis(Ci-Epi1), endoderm (ALP), muscle (AchE), notochord(Citalin) and mesenchyme (ciad005j06) (data not shown). Expression of pan-neural marker (Ci-ETR) indicated that citb018a09 is not required for neural tube formation (Fig. 5D,E). Because citb018a09 is likely to be involved in the formation of the sensory organs, we also examined the effects of its functional suppression on the photosensory organ. The expression of four genes involved in the function of the photosensory organ, Ci-opsin3(Fig. 5F,G), Ci-arr(Fig. 5H,I), Ci-opsin1(data not shown) and Ci-Gαi1a (data not shown)appeared in citb018a09-morpholino-injected embryos as in normal tailbud embryos. Co-injection of citb018a09 morpholino and citb018a09 mRNA failed to rescue the formation of pigment cells. However, a second form of citb018a09 morpholino which was produced against the 5′UTR also gave the same results as the first morpholino. These results suggest that this gene is not likely to be involved in the formation or function of the photosensory organ,but rather in pigment cell formation.

Besides the six genes described above, 20 other genes showed the `abnormal tail' phenotype. Of them, 14 genes gave rise to this phenotype but not the`loss of the sensory pigment cells' or `loss of palps' phenotypes. For example, depletion of cieg017l16, which encodes a protein with an HLH motif,resulted in larvae that had a short tail and a bent tail, respectively, but that developed the sensory pigment cells and the palps(Fig. 3F).

In the present screen, 200 Ciona intestinalis genes expressed during development were examined with respect to their possible function using specific antisense morpholinos. Forty genes were thereby shown to produce visible embryonic defects when their translation was suppressed. The phenotypes obtained in the present study were classified into two major classes, the `disorganized body plan' class and `abnormal tail' class. The first major class of phenotype, the `disorganized body plan' class, is characterized by gross morphological defects. Because embryos with a phenotype of this class exhibited abnormality from a considerably early stage of development, genes of this class may be involved in an early step of embryogenesis or a fundamental cellular process. However, possible causes for the `abnormal tail' phenotypes include failure of the differentiation and/or morphogenetic movement of tissues composing the tail, such as the notochord,muscle, dorsal nerve cord, ventral endodermal strand or posterior epidermis,although further analysis will be required to distinguish among these possibilities. Consistent with this idea, embryos injected with Ci-Bra morpholino also showed the `short tail' subclass phenotype and expression of a notochord maker was lost in the injected embryos. Among larvae with the `abnormal tail' phenotype, `loss of pigment cells' phenotype and`loss of palp' phenotype were detected. The sensory pigment cells are components of the photoreceptor and gravity sensor that are formed in the sensory vesicle. In the ascidian embryo, a primary defect in neural tube formation (e.g. failure of neural tube closure) often results in winding of the tail. The embryos showing both the `bent tail' and `loss of the sensory pigment cells' phenotypes may represent such cases. The palps are an adhesive organ located at the anterior tip of larvae, and this organ contains peripheral sensory neurons essential for signal transmission for metamorphosis. Previous studies have shown that correct cellular inductive interactions are required for the palp formation(Lemaire et al., 2002). Such interactions may be disrupted in embryos with the `loss of palps' phenotype. Altogether, the genes identified in the present screen will provide tools for investigating multiple and distinct steps of ascidian morphogenesis.

It has been shown that morpholinos often exhibit nonspecific side effects on embryogenesis in zebrafish (Heasman,2002). One of the conceivable reasons for this is unexpected targeting of other genes by the morpholinos. However, this is not likely to have been the case in the present screen, because we confirmed that, for each morpholino that induced any phenotype studied here, there was no gene that had a complementary sequence around or upstream of the initiation codon in the Ciona intestinalis genome other than the original target gene (data not shown). Another possible reason for this type of side-effect of morpholinos is nonspecific effects of morpholinos or contaminants. As mentioned above, however, the alteration of the morphology induced by a given morpholino seemed to be specific and distinct from that induced in embryos injected with Ci-Bra morpholino. Further support for the specificity of the action of morpholinos was obtained by rescue experiments. Therefore, it is likely that the phenotypes obtained in the present study were caused by the specific actions of the morpholinos rather than nonspecific side effects of morpholinos. The successful identification here of two genes that are expected to be involved in development, a Zic family gene (cicl002e04) and a homologue of Pbx (cilv008a08), also suggest that the present screen was effective.

In the present simple screen of visible features, morpholino-injected embryos were judged to be positive or tentatively negative based on the observation of the morphology of the resultant larvae. As a result, 160 genes(out of the 200 genes examined) were judged to be tentatively negative. However, the present method may overlook some genes with developmental functions. If we examine the effects of suppressing gene activity with more sophisticated techniques and specific probes that can detect changes at the cellular and molecular level, we may be able to more efficiently identify genes with developmental functions using the present methodology.

Large-scale screens for developmental genes using a reverse genetic approach have been conducted in the nematode C. elegans. A systematic analysis of C. elegans chromosome I by RNAi resulted in the functional identification of 13.9% of the genes analyzed(Fraser et al., 2000), whereas a similar analysis of cell division-related genes on chromosome III showed that ∼6% of genes analyzed were necessary for cell division(Gönczy et al., 2000). More recently, a systematic functional analysis of most of the C. elegans genes encoded in the genome using RNAi demonstrated that 10.3% of the examined genes showed specific phenotypes(Kamath et al., 2003). In zebrafish, it has been estimated from chemical mutagenesis screens that∼3-5% of the genes encoded in the genome can be mutated to yield some developmental defects that can be identified by a visual screen of embryos(Haffter et al., 1996). Compared with the findings of those screens, the efficiency of the present Ciona intestinalis morpholino-mediated experimental system, i.e. 20%of genes analyzed, seems appreciably higher. However, an important difference between the C. elegans studies and the present Ciona study should be noted. In the above-mentioned studies, a non-biased pool of genes was screened, while in the present study the genes to be tested were selected based on the features of their expression pattern and/or structure before the application of the screen. This probably accounts for the difference in the efficiency between the two screens.

In the past several years, chemical mutagenesis screening was performed in Ciona savignyi (Moody et al.,1999; Nakatani et al.,1999) and Ciona intestinalis(Sordino et al., 2000). Interestingly, the spectrum of phenotypes found in those studies seems similar to that observed in the present screen. For example, Moody et al.(Moody et al., 1999) reported that the most common phenotype they obtained was `rounded', which seems to be classified into `short tail' according to our categorization. The other phenotypes they described were `short tail', `kinked axis (similar to `bent tail' in the present study)', `no pigment cells' and `early arrest (probably categorized into `disorganized body plan')', all of which are similar to certain of the phenotypes found in the present study, although embryos with the last phenotype show diverse features depending on the morpholino, as discussed above, and thus are not necessarily functionally alike. The similarities of the phenotypes obtained by the chemical mutagenesis screen and our screen may indicate that defects in distinct developmental steps converge to produce several types of abnormal embryos with particular morphological features. Another notable, but probably rare, possibility is that homologous genes are responsible for the similar phenotypes found in the various classes of abnormal manipulated animals. In this respect, the genes identified in the present study represent candidate genes for the mutants obtained in the chemical mutagenesis. Further analysis is required to test these possibilities.

Analysis of the draft genome sequence and the ESTs of Ciona intestinalis suggests that, among a total of 15,852 genes, ∼10,000 genes are common to worms, flies and humans, 2600 genes are probably chordate-specific, and another 3400 have presently been found only in Ciona intestinalis (Dehal et al.,2002). As shown in Table 2, most of the DI-class genes identified in the present screen have vertebrate counterparts with unknown function. Therefore, determining the function of Ciona intestinalis DI-class genes may facilitate our understanding of the function of their vertebrate counterparts. There are at least 2500 DI-class genes in Ciona intestinalis(Satou et al., 2002a). A simple estimate from the results of the present study is that about one-fifth of DI-class genes that are expressed during embryogenesis may be shown to have a developmental function by a morpholino-based approach. It has been shown that most of the genes homologous between ascidians and vertebrates have a conserved function during development(Corbo et al., 2001; Satoh, 2001). Therefore,developmental genes suggested to have a pivotal function based on Ciona intestinalis studies should be analyzed in Xenopus or zebrafish embryos, or should be targeted in mouse embryos.

The cDNA project has characterized transcripts for more than three-quarters of Ciona intestinalis genes(Satou et al., 2002b). In addition, more than 5600 full-length sequences of cDNAs have already been determined. Together with the draft genome sequence, such information should help us to design effective morpholinos. For example, in the case of a gene that gives rise to multiple transcripts by alternative splicing, it is possible for us to design morpholinos more carefully by taking this point into account. The cDNA database also provides a rough estimate of the temporal expression pattern, i.e. the count of the ESTs at each developmental stage(Satou et al., 2003b). Furthermore, the spatial expression profiles of 5000 randomly selected genes have been determined by whole-mount in situ hybridization(Satou et al., 2002a). These expression profiles may help us to interpret the phenotypes of morpholino-injected embryos. In conclusion, Ciona intestinalisembryos provide a powerful experimental system for identifying genes with novel developmental function required for the formation of the chordate body plan (Satoh et al., 2003).

We are grateful to the following individuals for their help in determining the sequence of cDNAs: Kazuko Hirayama, Chikako Imaizumi, Asako Fujimoto,Hisayoshi Ishikawa and Hitoe Metoki. We also thank Kazuko Hirayama and Dr Shigeki Fujiwara for collection of Ciona adults. L.Y. was a pre-doctoral fellow and K.K. was a post-doctoral fellow of the JSPS. The present study was a project (headed by N.S.) of the CREST program(Development, Differentiation and Regeneration), JST, Japan.