ERK- and JNK-signalling regulate gene networks that stimulate metamorphosis and apoptosis in tail tissues of ascidian tadpoles

During embryonic development, apoptosis plays a central role in the successful outcome of the developmental program, and has been studied using genetic and cell-biological means in many model organisms (for a review, see Meier et al., 2000). Our ability to study the full extent of processes such as apoptosis during embryonic development has been advanced by the sequencing of the genomes of model organism and by the subsequent development of DNA-chip technologies coupled with gene-silencing-based approaches that can provide a mass of testable information about the dynamics of gene regulatory networks. Programmed cell death has been found to operate in all multicellular animals studied so far, including cnidarians, nematodes, insects, ascidians,amphibians, fish, birds and mammals (Cikala et al., 1999; Ellis et al.,1991; Steller,1995; Chambon et al.,2002; Vaux and Korsmeyer,1999; Jacobson et al.,1997). Metamorphosis represents a dramatic example of apoptosis and occurs throughout the animal kingdom(Vaux and Korsmeyer, 1999; Jacobson et al., 1997). In basal chordates such as ascidians, metamorphosis is characterized by a period of intense cell reorganization and remodelling that is dependent on programmed cell death events (Chambon et al.,2002; Jeffery,2002).

Although the execution of apoptosis occurs through the activation and function of a highly conserved family of cysteine proteases, termed caspases,present from hydra to man (Cikala et al.,1999; Thornberry and Lazebnik,1998), the initiation of apoptosis also depends on the activation of extrinsic and/or intrinsic death pathways(Ashkenazi and Dixit, 1998). In addition, the apoptotic process is also regulated by many intracellular signalling pathways, including the mitogen-activating-protein kinase (MAPK)pathways (Deng et al., 2003; Lin and Dibling, 2002; Chen et al., 2005).

MAPK proteins play both pro-apoptotic and anti-apoptotic roles depending on the cellular environment (Tibbles and Woodgett, 1999; Davis,2000; Chang and Karin,2001). Conventional MAPK proteins consist of three family members(Johnson and Lapadat, 2002):the extracellular signal-regulated kinase (ERK)(Seger and Krebs, 1995); the c-Jun NH2-terminal kinase (JNK); and p38(Waskiewicz and Cooper, 1995). ERK1-2 was initially reported to be a prosurvival factor(Wada and Penninger, 2004; Johnson and Lapadat, 2002),although a recent study suggests that it also has a key pro-apoptotic role in neuronal apoptosis induced by potassium (K⁺) withdrawal(Subramaniam et al., 2004). JNK can exhibit either pro- or anti-apoptotic functions, depending on the cell type, the nature of the death stimulus, the duration of JNK activation and the activity of other signalling pathways (for a review, see Liu and Lin, 2005). Similar to JNK, the involvement of p38-MAPK in apoptosis is also diverse(Sarkar et al., 2002).

MAPK plays a pro-apoptotic role during ascidian metamorphosis. The activation of the Ciona intestinalis MAPK ERK (Ci-ERK) in tail cells precedes the wave of apoptosis, suggesting that the phosphorylated form of Ci-ERK transduces the death-activating signal in tail tissues during metamorphosis (Chambon et al.,2002). Moreover, inhibition of Ci-ERK blocks metamorphosis(Chambon et al., 2002). Finally, it was recently reported that programmed cell death in C. intestinalis larvae also correlates with JNK activity(Tarallo and Sordino, 2004). However, it is not yet fully known what transcriptional events are targeted by MAPK to induce apoptosis in vivo. The simplicity of the ascidian tadpole,which consists of approximately 2600 cells, the rapid rate of development and the predictable wave of apoptosis during tail regression make the ascidians an amenable system in which to study apoptosis. In addition, the recent development of comprehensive microarrays (75% representation of the genome),coupled with extensive in situ gene expression profiles during embryogenesis(Satou et al., 2002)(http://ghost.zool.kyoto-u.ac.jp/indexr1.html),make the ascidian a useful organism for identifying the gene-regulatory network that controls the onset of metamorphosis and also for identifying the subset of genes regulated by MAPKs that induce apoptosis. Here, using ascidian metamorphosis as a model, we set out to identify the genes downstream of either JNK- or ERK-activity that control the onset of apoptosis during development. In addition, through a gene-silencing-based approach, we demonstrate that one of the genes identified by our screen, Ci-sushi,links JNK activation to the wave of apoptosis that precedes tail regression. As well as identifying 110 genes regulated by either JNK or ERK, we describe how the activation of JNK in the CNS controls apoptosis in adjacent tissues(notochord, epithelia and muscle), and propose a model for the synchronous apoptosis that occurs in these tissues.

Adult C. intestinalis were collected in the bay of Roscoff(Finistère, France) or were cultivated at the Maizuru Fisheries Research Station of Kyoto University; at the Field Science Center of Tohoku University; or at the International Coastal Research Center of the Ocean Research Institute, the University of Tokyo, Japan. Oocytes and sperm were obtained by dissection of gonoducts and cross-fertilization was performed in plastic Petri dishes. Embryos were cultured at 18°C in 0.2 μm filtered seawater containing 100 U/ml penicillin and 100 μg/ml streptomycin sulphate. Tadpole-type larvae hatched at approximately 18 hours of development(18 hpf). They then swam freely for more than 7 hours (25 hpf). The onset of metamorphosis started at approximately 8 hours after hatching (26 hpf), when juveniles adhered to plastic dishes, and the resorbtion of the tail began 2 hours after the attachment (28 hpf).

Larvae were fixed for 20 minutes with 3.7% formaldehyde in filtered seawater. Embryos were permeabilized with 0.2% Triton in phosphatebuffered saline (PBS) and washed once in PBS. Phosphorylated ERK was visualized by incubating larvae for 1 hour at room temperature with a monoclonal antibody raised against the active form of MAPKs ERK1 and ERK2 (dual-phosphorylated`HTGFLT(p)EY(p)VAT' peptide) (BD Transduction Laboratories). Embryos were then washed three times in PBS 0.008% Triton and incubated for 1 hour at room temperature with FITC-conjugated donkey-anti-mouse immunoglobulin as a secondary antibody (Jackson ImmunoResearch Laboratories) diluted 1:500 in PBS. Larvae were then washed three times in PBS, once in tris-buffered saline(TBS), were rinsed in distilled water and were then mounted in Vectashield containing DAPI (Vecta Laboratory). The slides were analyzed with a Zeiss Microscope. Phosphorylated JNK and p38 were both detected with a MAP Kinase Activation Monoclonal Antibody kit (BD Transduction Laboratories). Appropriate secondary antibodies were TRITC-conjugated donkey-anti-mouse or FITC-conjugated donkey-anti-mouse immunoglobulin (Jackson Laboratories).

The JNK inhibitor SP600125 (Sigma-Aldrich) was dissolved to DMSO and added at a final concentration of 10 μM immediately after hatching. The MEK inhibitor U0126 (Promega) was dissolved in DMSO and added either once (6μM) immediately after hatching, or once at this time-point (6 μM)followed by every 6 hours thereafter (6 μM at each administration). Embryos were cultured at 18°C during 48 hours. Treated and control larvae were scored for signs of metamorphosis. Results were the mean of three independent experiments.

For activated-ERK analysis, larvae were sonicated on ice in RIPA lysis buffer [150 mM NaCl, 50 mM Tris-Cl (pH 7.6), 5 mM EDTA, 0.5% NP-40, 1 mM PMSF,1 mM orthovanadate, 0.5% sodium deoxycholate, 0.1% SDS] supplemented with Complete Protease Inhibitor Cocktail Tablets (Roche Molecular Biochemicals). Lysates were then clarified by centrifugation. Samples were diluted in sample buffer (Laemmli, 1970) and incubated at 85°C for 5 minutes. Total proteins were separated on 12.5%SDS gels and transferred onto PVDF membranes. The blots were blocked with 5%milk powder in PBS-Tween, were incubated for 1 hour with a monoclonal anti-dual-phosphorylated ERK (BD Transduction Laboratories) that was diluted 1:500 in PBS-Tween, were washed in TBS-Tween, were incubated for 1 hour with the secondary antibody (HRP-conjugated anti-mouse IgG antibody diluted 1:10,000) and were then washed in PBS-Tween. Labelled proteins were detected using the Supersignal West Pico Chemiluminescent Substrate (Pierce). Additional analysis by western blotting with different antibodies was performed using 12.5% SDS polyacrylamide gels containing 0.13% bisacrylamide. The various primary antibodies used were MAP Kinase Activation Sampler kit and MAP Kinase Sampler kit (BD Transduction Laboratories), and, as a secondary antibody, we used horseradish peroxidase linked to rabbit antibodies directed against mouse-immunoglobulin Fc fragments (Sigma, Saint Louis, MO).

A C. intestinalis microarray (oligonucleotide-based chip version 1) was made by Agilent Technologies (ink-jet-based Sure Print technology). A total of 21,939 features on the chip consisted of 21,617 independent 60-mer oligonucleotide probe sequences. These probes represent 22,445 cDNA/EST sequences selected from over 450,000 C. intestinalis ESTs and 4062 cDNAs (Satou et al., 2002)(http://ghost.zool.kyoto-u.ac.jp/indexr1.html),with at least one sequence per gene. Information on the C. intestinalis genome (Dehal et al.,2002) indicates that 17,490 of these features represent 11,903(75.1% of the total 15,852) C. intestinalis gene models, according to computational analysis. Manual checking indicated that approximately 5% of the remaining 3949 gene models are estimated to have representative features also. The additional 4127 features not corresponding to any gene models represent approximately 3500 independent cDNA clusters. As a result, the total features on the array are estimated to cover approximately 75-80% of the genes in the C. intestinalis genome.

Total RNA of tadpole-type larvae was extracted 7 hours after hatching (25 hpf) using TRIZOL reagent and was then treated with DNase (Promega). Labelling of amplified cRNA with either cyanine-3 CTP or cyanine-5 CTP (Perkin-Elmer/NEN Life Sciences) was performed using 250-ng aliquots of total RNA with a Low RNA Input Fluorescent Linear Amplification kit (Agilent Technologies), following the manufacturer's instruction. The quality and size distribution of targets was determined by the RNA 6000 Nano Lab-on-chip Assay (Agilent Technologies),and quantification was performed using a NanoDrop microscale spectrophotometer(NanoDrop). A set of cRNA targets from each sample was assembled into a hybridization reaction using an In Situ Hybridization kit (Agilent Tech.). Each hybridization was compared with that of the 0-hour larvae sample (18 hpf). Hybridization was performed twice with a dye swap. Hybridized microarrays were washed according to the manufacturer's protocol and then scanned using a G2565BA Micro-Array Scanner System with SureScan technology(Agilent Technologies).

The intensity of 21,939 gene features per array was extracted from scanned microarray images using Feature Extraction 7.1 software (Agilent Technologies), which performs background subtractions and dye normalization. The data were analyzed using Excel (Microsoft). Hierarchical clustering was performed using Cluster and Treeview(Eisen et al., 1998).

The gene features were searched against the C. intestinalis draft genome sequence (Dehal et al.,2002) to obtain information about corresponding genes. The features that had several hits in the draft genome sequences were excluded from the present analyses and the remaining 18,222 gene features were picked up to be included in the analyses. Gene features that coincided with a predicted gene model or cDNA sequences from the cDNA/EST database(Satou et al., 2002) and the deduced protein sequences were used to characterize corresponding genes with the Genome browser of JGI(http://genome.jgi-psf.org/ciona4/ciona4.home.html). The cDNA and EST sequences were used in combination with the JGI gene model to obtain more-complete gene models. The GENES Human ENTRY number concerning JGI gene models was used for analysis on the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.ad.jp/kegg)GENES Human database and KEGG PATHWAY database(Kanehisa and Goto, 2000; Kanehisa, 2002).

For mRNA isolation, larvae were lyzed in TRIzol (Invitrogen), according to the supplier's instructions. mRNA was reverse transcribed. For each gene, a set of specific forward (For) and Reverse (Rev) oligonucleotides was designed to amplify a small part of coding region (from 200-600 bp):

C-JUN: (For: CCGTCTTGAAGGGTATGAGC, Rev: GTTCCAGTTTCCGCTTTCTG);

SRF: (For: CCACGACGATGAACATTACG, Rev: GTCGGCGTTTTATGTTCGTT);

MKP: (For: CCACTTTCCAGACCGATTTC, Rev: CCTCACAGGTCCACTCCATT);

Ci-GNRH: (For: TGTGTGTTACTTGTCGTTCTAGCC, Rev: GGATCCGTTGCAAGAGTTGT);

Ci-sushi: (For: TTGCAAGTCTTTGCACAGTTG, Rev: CCAACGGCTGTGATATGTGA);

Ci-ETS: (For: CAAAGCACACCAAGCCAGTA, Rev: GTTGGGGTAGCATGGTTCAT);

Ci-LyOx: (For: TGGGTGGGACTTGAACAC, Rev: TTCCCTCTTGCGTACTTTGG);

Ci-Vwa1a: (For: TGGTTGCAAACAAGAAGCTG, Rev: ATCCTCATTTGCATCGAACC);

Ci-Vwa1b: (For: GCACTGTCGGTTCAGTGTGT, Rev: CCGAAACTAGGTTGCGTTGT);

Ci-Vwa1c: (For: ATACTTCGACCCAGCACGTC, Rev: AAACTCCGTTACGCCTCAGA);

Ci-OATP: (For: CGGTTGGGTTGATCTTGAGT, Rev: GACGATCCCAACTTTTTCCA);

Ci-dhg: (For: GTCACCGTTTCCTCTGAAGC, Rev: GCGCCGTGTATTATGGTCTT);

Ci-Mx: (For: ACCCAGACATTGCAGGAAC, Rev: GAGCCGCTACAATTCTCAGG).

Reverse transcriptase (RT)-PCR was performed on equal amounts of input RNA and cDNA using the amplimers describe above. S26 mRNA was used as a control using the amplimer For: TCCCCTTCTTCCTCAAGCAC and Rev: GCCCAACCACCATCCTGTA. For the control S26, PCR was first performed on cDNA from untreated larvae or from 18 hpf larvae, and the number of cycles adjusted so as to produce a non-saturating signal on ethidium bromide-stained 1.2% agarose slab gels. Semi-quantitative PCR was next performed on cDNA from each time-point or from untreated and treated larvae using the number of cycles determined above. PCR products were run on 1.2% agarose slab gels.

Chorionated C. intestinalis oocytes were vigorously pipetted to remove the outer follicular cells. These partially dechorionated oocytes were then fertilized and transferred to cleaned Petri dishes (cleaned with ethanol and rinsed in distilled water) and allowed to settle by gravity. As the oocytes settled, a substantial proportion fixed to the plastic Petri dishes. Those zygotes that became fixed were injected (injection-pipette holder mounted at ∼30°), with the injection needle controlled by a hydraulic micromanipulator (Narishige) coupled to a high-pressure injection device(Narishige IM300). Ci-sushi morpholino antisense DNA (Gene Tools) at a pipette concentration of 1 mM was mixed 10:1 with 20 mM Texas Red-Dextran 10 kDa (Molecular Probes) and injected into the zygotes (∼1% injection volume) after extrusion of the second polar body and before the first cleavage. The following day (at 18 hpf), those tadpoles that were fluorescent were isolated and further analyzed through metamorphosis and TUNEL labelling. As a control, we used standard control morpholino (Gene Tools). Results were the mean of three independent experiments.

Whole-mount in situ hybridization was performed by using digoxigenin(DIG)-labelled antisense probes as described previously(Satou, 1999). RNA probes were prepared using a DIG RNA-labelling kit (Roche). Control embryos hybridized with a sense probe did not show any signal above background levels.

Following a rapid embryogenesis, which lasts 18 hours at 18°C, C. intestinalis larvae hatch and swim within the plankton for a variable period of a few hours. The larvae then become competent to respond to either natural or artificial settlement cues and start a series of rapid morphological changes (Arnold et al.,1997). After settlement, C. intestinalis metamorphose into sessile filter feeders, a process involving a massive reorganization of the body plan and, most notably, apoptosis-dependent tail regression(Chambon et al., 2002). The potential role of the CNS in initiating metamorphosis by selecting sites for settlement was previously proposed(Cloney, 1978; Cloney, 1982).

We chose to focus on MAPK signalling cascades for two reasons. First, we and others have shown that the inhibition of ERK prevents tail regression and apoptosis during ascidian metamorphosis(Chambon et al., 2002; Tarallo and Sordino, 2004). Second, it has previously been reported that MAPK pathways are well conserved in the C. intestinalis genome(Satou et al., 2003; Hotta et al., 2003) (see Table S1 in the supplementary material) and that some components of these pathways are strongly and/or exclusively expressed in the CNS of C. intestinalis during the tadpole stage(Imai et al., 2004) (see also http://ghost.zool.kyoto-u.ac.jp/ST2005.html). We examined whether activation of the three main MAPK pathways occurs during the swimming phase of C. intestinalis larvae. Given the perfect match between the Ciona and human peptide sequences in each of the two proteins (see Fig. S1 in the supplementary material), we used antibodies directed against the human dual-phosphorylated peptides to detect activated and non-phosphorylated Ci-ERK, Ci-JNK and Ci-p38 from C. intestinalisprotein extracts and/or by indirect immunofluorescence(Fig. 1A-C and D). C. intestinalis larvae from hatching (18 hours post-fertilization; hpf) to the onset of metamorphosis (28 hpf) were analyzed by western blotting. The level of total protein extract of each sample was visualized by Coomassie Blue staining (Fig. 1A, upper panel). An activated form of Ci-JNK was present during the swimming phase of the larvae at 20 and 22 hpf. Ci-ERK was activated between 20 and 24 hpf, with peaks at 20 and 22 hpf, and was then activated again at 28 hpf, as we previously described (Chambon et al.,2002). No activation of Ci-p38 was detected at any time during the swimming period. We confirmed that Ci-ERK, Ci-JNK and Ci-p38 protein levels did not vary significantly during this period(Fig. 1A, lower panels).

In order to determine where in the larvae JNK and ERK activation occurred,we used the same antibodies for indirect immunofluorescence studies on larvae fixed at various times. In agreement with protein expression analysis and our activation study (Fig. 2C),Ci-JNK and Ci-ERK were found to be activated at 20 and 22 hpf. No Ci-p38 was detected in any region of the swimming larvae (see Fig. S2 in the supplementary material).

Ci-JNK was phosphorylated in the CNS of larvae(1B,ctivated Ci-JNK was detected in the CNS localized in the trunk of the larvae: specifically, in the posterior part of the sensory vesicle that contains the pigmented ocellus and otolith, the visceral ganglion composed of motoneurons and between these the neck region. The caudal nerve cord also contained activated Ci-JNK.

Ci-ERK was activated in the papillae of Ciona larvae(1C. intestinalis papillae are constituted of secreting cells, axial columnar cells, primary sensory neurons and undifferentiated ectodermal cells(Manni et al., 2004). Activated Ci-ERK was detected in all the papillae, from the anterior part to the extremity of each of the three palps. Some nuclei of the anterior epidermis of the trunk were also positively stained for phosphorylated Ci-ERK. As we previously described, Ci-ERK was also later activated in tail cells at 28 hpf (Chambon et al.,2002).

We next compared the status of apoptosis in larvae that had initiated metamorphosis and that had been treated with MAPK inhibitors. The hatching larvae were treated with a MAPK inhibitor and, after 10 hours (at 28 hpf),were fixed and processed for TUNEL labelling(Fig. 2A). Major differences were observed between larvae treated with the MEK inhibitor (U0126) or JNK inhibitor (SP600125) to those larvae untreated or treated with the p38 inhibitor (SB203580). Consistent with our previous work(Chambon et al., 2002), the untreated larvae that began metamorphosis exhibited TUNEL-positive cells at the tip of the tail (Fig. 2A). No difference was observed when larvae were treated with the p38 inhibitor. By contrast, the inhibition of JNK (SP600125) and ERK (U0126) phosphorylation inhibited initiation of apoptosis at the tip of the tail(Fig. 2A).

Hatched larvae were incubated at 18°C in filtered seawater with DMSO(0.1%), U0126 (6 μM) or SP600125 (10 μM). Inhibition of the JNK cascade with SP600125 completely blocked metamorphosis(Fig. 2D) at 2 days post-fertilization. Similarly, but to a lesser extent, inhibition of MEK with U0126 blocked metamorphosis in a significant proportion of the tadpoles(Fig. 2D)(Chambon et al., 2002). Given that U0126 has a short half-life, U0126 treatment was repeated every 6 hours. In this condition, inhibition of MEK significantly decreased the number of larvae that reached the metamorphosis stage and increased the number of swimming larvae (Fig. 2D). As depicted in Fig. 2D, at 2 days after fertilization, inhibition of the JNK pathway with 10 μM SP600125 induced ∼100% blockade of metamorphosis, and inhibition of the ERK pathway with 6 μM U0126 led to a significant decrease (94%) in the number of metamorphosed larvae. In the control (DMSO treated) group, 95% of larvae underwent metamorphosis.

Although Ci-ERK and Ci-JNK activation was observed in DMSO-treated larvae,the activated Ci-ERK and Ci-JNK labelling was lost at 24 hpf with SP600125 and U0126 treatment, respectively (Fig. 2B). Moreover, incubation with SP600125 or U0126 significantly decreased Ci-JNK or Ci-ERK phosphorylation when compared with the control by western blot analysis (Fig. 2C).

We performed a kinetic study based on cDNA microarray experiments(Azumi et al., 2003) to identify genes that are differentially transcribed during the swimming phase. An overview of the results for each MAPK pathway is displayed in Fig. 3A. Although very few modulations in gene expression were observed immediately after hatching, the situation changed between 26 and 28 hpf, which corresponded approximately to the start of tail absorption (Fig. 3A). Interestingly, the two transcription factors SRF(serum response factor) and c-jun, which are known to be targeted by ERK and JNK, respectively, were upregulated at 26 and 28 hpf(Fig. 3B). The expression of SRF was maximum at these times before declining to its initial level,whereas c-jun expression was highly up-regulated from 26 to 50 hpf(Fig. 3B, upper and middle panel). We confirmed the microarray expression profile of SRF and c-jun by semi-quantitative reverse transcriptase (RT)-PCR(Fig. 3C). SRF is a target of the ERK signalling pathway and provides a positive-feedback loop(Kasza et al., 2005). A similar positive feedback exists for JNK, as demonstrated in hypoxic HepG2 cells, through regulation of the c-jun promoter(Minet et al., 2001). Moreover, the MKP (MAP kinase phosphatase) gene, a key protein of MAPK pathways, was upregulated from 26 to 30 hpf(Fig. 3B, lower panel; Fig. 3C). MKP is a member of the dual specific phosphatase (DSP) protein family, members of which inactivate MAPKs through dephosphorylation. The increase in MKP gene expression could provide a negative-feedback loop on activated MAPK pathways in C. intestinalis larvae; as has been described for the ERK pathway in Drosophila and Xenopus(Gomez et al., 2005).

Because inhibition of either ERK or JNK prevented the wave of apoptosis that preceded tail regression, we undertook a gene-profiling study using oligonucleotide-based microarrays (Yamada et al., 2005; Ishibashi et al., 2005) to determine the transcriptional differences between larvae treated with MAPK inhibitors or with DMSO (control). We chose tadpole larvae that were at 25 hpf for two reasons. First, because both ERK and JNK displayed a peak of activity starting at 20 hpf(Fig. 1); and second, because the major change in gene expression began between 24 and 26 hpf(Fig. 3). Among the many genes where the expression pattern is altered at 25 hpf, we anticipated identifying the subset that is involved in the initiation of the wave of apoptosis,because this is blocked by both inhibitors.

To identify genes regulated by these two MAPK pathways, we determined the expression profile of DMSO-treated larvae aged 25 hpf with identically aged larvae that had been treated with either SP600125 or U0126. Only genes that consistently displayed greater than threefold changes during SP600125 treatment or greater than twofold changes in transcript abundance during U0126 treatment were scored. This conservative approach, which has been used in a variety of microarray studies (e.g. Butler et al., 2003; Munoz-Sanjuan et al., 2002), is likely to miss a number of genes that are differentially expressed in response to MAPK-inhibitor treatment. The genes affected by the MAPK-inhibitor-treatment protocol were categorized into four major classes based on Lee et al. (Lee et al., 1999) (Tables 1, 2). Class A coded for proteins common to many cell types, class B were proteins associated with cell-cell communication, class C coded for proteins that function as transcription factors and other regulatory proteins, and class D were proteins of an unknown function (DI) or with no similarity (DII).

In total, 38 genes were upregulated and 21 genes were downregulated by inhibition of the JNK pathway (Table 1). These included homologues of genes and genes previously shown to be involved in metamorphosis of other ascidians species, as well as of C. intestinalis (Nakayama et al.,2001; Nakayama et al.,2002; Davidson and Swalla,2002; Woods et al.,2004). For example, Ci-meta5, a gene previously shown to be involved in metamorphosis (Nakayama et al., 2001; Nakayama et al.,2002), was identified by our screen.

A total of 12 genes were upregulated and 41 genes were downregulated by U0126 MEK-inhibitor treatment (Table 2). In addition to genes previously shown to be involved in ascidian metamorphosis, treatment with MAPK inhibitors identified other genes not previously implicated in this process. We will come back to some of these genes later in the discussion.

In order to confirm the microarray data, we performed RT-PCR of the genes identified by cDNA chips between larvae treated with MAPK inhibitors or with DMSO (control), including three Vwa1 genes (Ci-Vwa1a,Ci-Vwa1b and Ci-Vwa1c), dehydrogenase(Ci-Dhg), matrix metalloprotease (Ci-Mx), gonadotropin releasing hormone (GNRH; Ci-GNRH), lysil oxidase (Ci-LyOx), ets (Ci-ets) and organic anion transporting polypeptide 14 (oatp14; Ci-oatp) (Fig. 4A,B).

Due to the substantial EST programme coupled with extensive in situ data(Satou et al., 2002)(http://ghost.zool.kyoto-u.ac.jp/indexr1.html),many of the genes identified by our screen had already been described. For example, genes under the control of Ci-ERK, such as rev-erb, are expressed exclusively in the palps(Kusakabe et al., 2002) (see also http://ghost.zool.kyoto-u.ac.jp/cgibin3/photoget2.cgi?CLSTR03308). Selectin, which is also controlled by Ci-ERK, was observed in the palps of competent and attached larvae during metamorphosis of the ascidian Boltenia villosa (Davidson and Swalla, 2002). Among genes that we identified to be controlled by Ci-JNK, Vwa1 had already been detected in the tadpole brain(Satou et al., 2001) (see also http://ghost.zool.kyoto-u.ac.jp/cgi-bin3/photoget2.cgi?CLSTR01650),and Emc in the nervous system and the nerve corde(Imai et al., 2004) (see also http://ghost.zool.kyoto-u.ac.jp/cgi-bin3/photoget2.cgi?cicl010f24). We also conducted a series of whole-mount in situ hybridizations for several of the genes identified by our screen, including Ci-GNRH, Ci-endoglucanase(Ci-endogl), NTKL (Ci-NTKL), Ci-ets, Ci-Vwa1a, Ci-Vwa1ac,Ci-metalloprotease (Ci-Mx) and Ci-oatp (Tables 1, 2). An overview of the results for each MAPK pathway is displayed in Fig. 4.

We previously demonstrated that JNK activation is required for apoptosis induction in tail cells and for tail regression during metamorphosis. In order to identify candidate genes involved in the initiation of the wave of apoptosis, we next identified, by cDNA microarray analysis, genes that are controlled by the JNK pathway. To address this issue, we examined the effect of the functional suppression of JNK-controlled genes on apoptosis induction at the onset of metamorphosis. Among the many genes identified, Ci-Sccpb and Ci-sushi(Table 1) showed, respectively,expression at the tip of the tail and in tail epithelia(Fig. 5A). Out of these two genes, we choose Ci-sushi (Table 1) for gene silencing for two reasons. First, based on its EST count (Satou et al., 2002) and RT-PCR (5,Ci-sushi is expressed between 24 and 26 hpf only, just before the initiation of apoptosis and the onset of metamorphosis; therefore, the Ci-sushi knockdown avoids possible lethal phenotypes due to Ci-sushi having a role in earlier steps of Ciona development. Second, we confirmed the microarray data by RT-PCR and found that Ci-sushi expression is downregulated by Ci-JNK inhibitor (Fig. 5B and Table 1). By contrast, Ci-Sccpb expression is upregulated by SP600125(Table 1). These data thus suggested that Ci-sushi would be an ideal candidate for a functional suppression experiment.

When a morpholino against Ci-sushi was injected into chorionated fertilized eggs, all larvae examined (18/18) were unable to initiate apoptosis in tail cells at the onset of metamorphosis(Fig. 5C). At this stage, most(22/25) of the control larva exhibited TUNEL-positive cells at the tip of the tail, and began tail regression (Fig. 5C). Moreover, as a control for synchronization and methodology,the tunic cells in both morpholino-injected and control larvae were apoptotic at this time (Chambon et al.,2002) (Fig. 5C). These data demonstrate that Ci-sushi is necessary to induce apoptosis in cells that compose the tail of C. intestinalis larvae at the onset of metamorphosis.

The MAPK signalling cascade plays a central role in mediating apoptosis during embryogenesis in invertebrates such as Drosophila(Kuranaga et al., 2002; Moreno et al., 2002) and Caenorhabditis elegans (Gumienny et al., 1999), and, in the mouse, double knockout of the two MAPK family members JNK1 (also known as MAPK8 - Mouse Genome Informatics) and JNK2(also known as MAPK9 - Mouse Genome Informatics) leads to embryonic lethality associated with severe dysregulation in the control of cell death in the hindbrain and forebrain (Kuan et al.,1999; Sabapathy et al.,1999). There are also indications that the MAPK signalling cascade mediated via the ERK family plays a role in the onset of apoptosis during embryogenesis (Kling et al.,2002; Yao et al.,2003). Although the role of JNK (and to a lesser extent ERK) is well established in many experimental systems, it is less clear what the gene targets of JNK (or ERK) are that trigger the onset of apoptosis. We addressed this issue by initially measuring the dynamics of the gene-expression network during the onset of metamorphosis in Ciona. By exploiting the observation that inhibition of the MAPK signalling cascade specifically blocks the onset of apoptosis (Chambon et al.,2002) (this study), we used these pharmacological tools together with a microarray-based approach to identify the subset of genes, controlled by either Ci-ERK or Ci-JNK, that are involved in triggering the onset of apoptosis. This strategy revealed 50 potential target genes that were upregulated and 63 that were downregulated when the MAPK signalling cascades,and therefore apoptosis, were inhibited.

At the end of the period of swimming, ascidian tadpole larvae undergo metamorphosis, which usually begins with settlement through adhesive papillae and subsequent tail regression and loss of adhesive papillae(Cloney, 1978; Cloney, 1982). A larva capable of undergoing these metamorphic changes successfully is termed competent, and the acquisition of metamorphic competence during the larval period has been shown to occur in response to a wide variety of external and endogenous signals (Jackson and Strathmann,1981; Cloney,1982; Davidson and Swalla,2002; Jackson et al.,2002). The Ci-JNK cascade is activated at the time of competence in the tadpole CNS; more specifically, in the pharyngeal rudiment, anterior sensory vesicle, neck, internal neurons of the posterior sensory vesicle,visceral ganglion and nerve corde (Tarallo and Sordino, 2004; Chambon et al., 2002) (this study). More importantly, inhibition of the JNK pathway completely blocked metamorphosis. Similarly, Ci-ERK activation correlates with the time of metamorphic competence. Ci-ERK is activated in:tail muscle, the cytoplasm of proximal palp cells, the stomodeum, the anterior and posterior sensory vesicle, the epidermis overlying the sensory vesicle,the neck region, atrial primordia, the notochord and in the epidermis of the tail (Tarallo and Sordino,2004; Chambon et al.,2002) (this study). Similarly, blocking activation of Ci-ERK inhibited metamorphosis in a significant percentage of the tadpole larvae(Chambon et al., 2002) (this study).

Among the genes identified, we will discuss some that we have separated for convenience as either upregulated or downregulated by JNK or ERK activity into four categories. First, there are genes involved in innate immunity. Second,there are genes involved in hormone signalling. Third, there are genes involved in metabolism of the extracellular matrix (ECM) or coding components of the ECM. And fourth, we grouped the remaining genes into one category that we term diverse genes. Before discussing the four categories, it was also gratifying to notice that our screen identified genes that had previously been shown to be specifically expressed during ascidian metamorphosis(Woods et al., 2004; Davidson and Swalla, 2002; Nakayama et al., 2002), such as Ci-meta5, which is downregulated in response to JNK-inhibitor treatment. Ci-meta5 was previously identified by differential screening of a cDNA library of swimming larvae and metamorphosing juveniles(Nakayama et al., 2002). We also identified three genes (glutathione S-transferase, Cytochrome p450 and Gluthathione-requiring prostaglandin D synthase) that are under the control of Ci-ERK and are orthologues of or closely related to genes expressed in papillae of the ascidians

Herdmania curvata that have been implicated in metamorphosis and are downstream genes of the EGF-like Hemps pathway(Woods et al., 2004; Arnold et al., 1997). Although it is not known whether Hemps activates ERK, it is tempting to speculate that it does, because activation of the Ras/Raf/ERK pathway by EGF is well described in many species (Hornstein et al., 2003). Moreover, our identification of genes controlled by Ci-ERK in papillae, the observation that metamorphosis does not occur with MEK inhibition and data on Hemps in H. curvata that shows that it induces settlement and metamorphosis (Eri et al.,1999) are consistent with our observations that one of the effects of the Hemps pathway is to activate the ERK cascade in papillae cells.

In our screen, 20 genes known to be involved in innate immunity were identified. It has been suggested that the activation of innate immunity genes during metamorphosis may represent the maturation of the adult immune system,and may be necessary for phagocytosis and for the re-structuring of larval tissues (Davidson and Swalla,2002). Among the genes controlled by activation of Ci-ERK in papillae, Ci-Vwa1c and Ci-polydom could coordinate papillae resorption during metamorphosis. In the same way, Ci-Pgly, Ci-ficolinand the five genes similar to Vwa1 (Ci-Vwa1a, Ci-Vwa1b, Ci-Vwa1d,Ci-Vwa1e and Ci-Vwa1f) that are controlled by phosphorylation of Ci-JNK in the CNS, could lead to phagocytosis of the visceral ganglion and sensory organs, which has been observed during metamorphosis of many ascidian species (Cloney, 1978). It is also of interest that the modulation of expression of genes such as Ci-Sccp could also enhance cell-cell communication before the extended period of cell reorganization and the co-ordinated massive wave of apoptosis that occurs during tail regression.

Among the genes identified that are controlled by the JNK pathway, two are interesting: Ci-GNRH and Ci-oatp, which are involved,respectively, in the reproductive and thyroid axes. The mouse Oatp14(also known as Slco1c1) was described in the transport of thyroxine across the blood-brain barrier (Tohyama et al., 2004). It is interesting to notice that the role of thyroid hormones in metamorphosis had been reported previously in ascidians(Patricolo et al., 1981; Patricolo et al., 2001), and also in amphibians (Dodd and Dodd,1976; Nakajima et al.,2005) and lamprey (Youson and Sower, 2001). Moreover, in four ascidian species, thyroxin is present in larval mesenchyme and seems to be involved in the control of metamorphosis (D'Agati and Cammarata,2005). The expression of Ci-oatp via JNK activation in the CNS may enhance thyroid signalling in larvae. Concerning GNRH, no report describes any function of this hormone in invertebrate metamorphosis. However,it is possible that GNRH may have a role in lamprey metamorphosis because, in sea lamprey, the level of GNRH increases throughout the stage of spontaneous metamorphosis (Youson and Sower,2001; Youson et al.,2006).

In addition to identifying genes involved in the immune system and hormonal signalling, we also identified a number of genes coding for proteins involved in the composition or processing of the ECM. For example, we identified Ci-LyOx (Table 1, Fig. 4B), which is responsible for the cross linking and deposition of collagen fibres, elastin fibres and Ci-Mx, a matrix metalloprotease(Table 2, Fig. 4A). The regulation of matrix metalloprotease and the ECM remodelling have been shown to affect apoptosis in different systems, including the apoptotic remodelling of the intestine during Xenopus laevis metamorphosis and post-lactation involution of the mouse mammary gland(Nakajima et al., 2005; Fata et al., 2004). Anoikis is apoptosis induced by the loss of, or inappropriate, cell adhesion. It is tempting to hypothesize that one of the inductive signals from Ci-JNK in the CNS controls apoptosis by changing ECM composition. The role of JNK in ECM degradation has already been reported in rat aortic walls(Yoshimura et al., 2005). In the tail of the tadpole, nerve corde is surrounded by matrix, which leads us to speculate that remodelling the ECM could provide a means to coordinate the response of tail cells in promoting either cell death or survival. In support of such a scenario, it was reported that, after a modification in ECM components, activation of the MAPK ERK leads to anoikis-type death(Zugasti et al., 2001). Because Ci-ERK activation precedes apoptosis in tail cells(Chambon et al., 2002), this cell death could be regulated by JNK-controlled anoikis in the tail of ascidian tadpoles.

In addition, we also observed that 45.5% of the identified genes controlled by the ERK pathway had no significant identity or similarity with any sequence in the GenBank Database, or matched with hypothetical proteins from different organisms. This observation might be explained by the highly specialized adhesive organs (papillae) in which these genes are expressed. For example,there are eight types of adhesive papillae among ascidians species, which could explain these species-specific components(Cloney, 1978).

We identified Ci-sushi as a target gene of the JNK pathway through the microarray-based approach. Ci-Sushi expression decreased following inhibition of the JNK pathway (SP600125 treatment). Moreover, under physiological conditions, Ci-sushi was only expressed in tail epithelia at 26 hpf, just before the onset of metamorphosis and of apoptosis at the tail extremity. Ci-Sushi encodes a protein containing domains known as complement control protein (CCP) modules, or short consensus repeats(SCR), which exist in a wide variety of complement and adhesion proteins. The abolition of apoptosis initiation by silencing of Ci-sushi with an antisense morpholino oligonucleotide demonstrates that JNK-induced activation of Ci-sushi expression in the tail is required to trigger the onset of apoptosis in tail tissues. This result strengthens the hypothesis that cross-talk exists between the different tail tissues and the CNS before the onset of metamorphosis.

Here, we observed that activation of ERK in the papillae and JNK in the CNS are able to control, either directly or indirectly, apoptosis of different tissues composing the tail. Because the papillae are innervated in C. intestinalis (Manni et al.,2004), it is possible that the coordination of settlement with tail regression is controlled by the CNS. The potential role of the CNS in metamorphosis was previously raised in a review and analysis of ascidian metamorphosis in 1978 by R. A. Cloney. He proposed a preponderant role for the larval nervous system and sensory organs in selecting sites for settlement and in the onset of metamorphosis (Cloney,1978; Cloney,1982). Moreover, he hypothesized that the nervous system, and conduction and diffusion of one or more humoral factors are likely to be involved in metamorphosis (Satoh,1994). Taken together with our previous results on apoptosis-dependent tail regression(Chambon et al., 2002), and with the activation of JNK in the CNS that we report here, these results raise two interesting questions: (1) How can JNK signalling in the CNS control apoptosis of the different tissues composing the tail during its regression?and (2) What is the nature of the JNK inductive signal that leads to apoptosis?

We propose a model (Fig. 5D)whereby the CNS enhances cell-cell communication in adjacent tissues through the expression of genes, such as Ci-sushi or Ci-Sccpb, that are essential for apoptosis-dependent tail regression. In addition, the CNS may also modify ECM composition trough Ci-JNK activation, thus leading to the induction of apoptosis in adjacent tissues that are receptive (the CNS and the endoderm escape apoptosis) (Chambon et al.,2002). The model we propose places the CNS centrally in the coordination of the wave of apoptosis that precedes tail regression during Ciona metamorphosis

Investigation of these questions in C. intestinalis is particularly interesting because this organism contains the genetic rudiments of many vertebrate characteristics and the larvae represent the basic chordate body plan (Satoh, 1994; Satoh, 2003). It is interesting to notice that, during development in humans, cell death has been shown to be an important morphogenetic mechanism for the formation of the vertebral column. For example, some of the notochord exhibits cell death while the remaining cells contribute to the formation of the nucleus pulposus in the human inter-vertebral disc (Saraga-Babic et al., 1994). An excess of cell death in these structures leads to neural tube defects or tailless mutants in mouse(Alles and Sulik, 1990; Grüneberg, 1963). Ciona may therefore provide a pertinent model to study in vivo regulation of apoptosis in these different tissues.

By contrast, apoptosis in mammalian skeletal muscle is a rather rare event. It has been reported as a mechanism for removal of undesired myotubes(mononucleated cells) during development(Sandri and Carraro, 1999). Additionally, apoptotic death of single nuclei in otherwise normal muscle fibre has been shown with an incidence of 0.1%(Sandri et al., 1998) to 0.3%(Migheli et al., 1997)TUNEL-positive nuclei. Muscle satellite cells are believed to form a stable,self-renewing pool of stem cells in adult muscle, where they function in tissue growth and repair. A regulatory disruption of growth, differentiation and apoptosis of these cells is assumed to result in tumour formation(Koleva et al., 2005).

The future use of microarray technology coupled with a gene-silencing strategy should permit a better understanding of apoptosis regulation in the tail of Ciona larvae during metamorphosis and could create avenues of investigation that lead towards a better understanding of cell death regulation during development in mammals.

We acknowledge H. Yasuo and C. Hudson for helpful discussion. The work of J.-P.C. in Nori Satoh's laboratory at Kyoto University was supported by the Japan Society of Promotion of Science (JSPS). A.N. was a postdoctoral fellow of the 21 COE program (A14) of Kyoto University. The present study was supported by Grants-in-aid of MEXT, Japan and Japan Science Technology Agency(JST) CREST project to N.S.