Ttrap is an essential modulator of Smad3-dependent Nodal signaling during zebrafish gastrulation and left-right axis determination

Nodal is the major signal in the establishment of left-right (LR) asymmetry during vertebrate development (Reissmann et al., 2001; Schier,2003; Yost, 2003; Raya and Belmonte, 2004). The binding of Nodal to its receptor complex (Alk4/ActRIIB/co-receptor EGF-CFC)activates Smad2/3, and phosphorylation of these Smads by Alk4 increases their affinity for Smad4. The resulting Smad complexes accumulate in the nucleus and participate in transcription of target genes by cooperating with various activators, repressors and chromatin modulators(Massagué, 2000; van Grunsven et al.,2005).

The best characterized Smad2/3 partners in Nodal-Activin signaling(Stemple, 2000) are FoxHI(Fast1) and Mixer (Hill, 2001; Whitman, 2001; Attisano,2001). Analysis of zebrafish mutants for FoxHI (schmalspur, sur)(Pogoda et al., 2000) and Mixer-like (bonnie and clyde, bon)(Kikuchi et al., 2000)revealed that the individual and combinatorial mutant phenotypes do not represent all aspects of Nodal signaling(Kunwar et al., 2003). This could be due to several reasons, including the possibility that additional players in Smad signaling remain to be identified. Nodal signaling studies in fish have focused on the role of Smad2/FoxHI and identification of its targets, whereas the situation is less clear with regard to Smad3(complexes).

Within the context of an antisense screen in zebrafish using morpholino oligomers (morpholinos, Mos) (Summerton and Weller, 1997; Nasevicius and Ekker, 2000), we identified Ttrap (TRAF and TNF receptor-associated protein) (Pype et al.,2000) as a regulator of embryogenesis. Human TTRAP interacts with TNF receptor (TNFR) family members and TNFR-associated factors (TRAFs) and inhibits NF-κB activation in TTRAP-overproducing cells(Pype et al., 2000). TTRAP has also been termed EAPII - ETS-associated protein II - revealing a possible dual role of this protein within the cytoplasm and nucleus(Pei et al., 2003). TTRAP belongs to the family of divalent cation-dependent phosphodiesterases, with highest homology to APE1, an endonuclease involved in DNA repair and transcription factor activation(Rodrigues-Lima et al.,2001).

The in vivo role of Ttrap has not yet been described. We show that Ttrap controls gastrulation movements and LR axis determination in zebrafish via Smad3-mediated regulation of e-cadherin, which is known to be regulated by the repressor snail and modulate cell movements (epiboly and convergent extension) in fish embryos (Babb and Marrs, 2004; Kane et al.,2005; Shimizu et al.,2005). We also uncovered a possible role for e-cadherinin the organization of dorsal forerunner cells (DFCs) during formation of Kupffer's vesicle (KV), a signaling center essential for establishing LR asymmetry.

Danio rerio stocks were maintained at 28.5°C under standard aquaculture conditions. Embryos were staged by hours post fertilization (hpf)using staging criteria as described(Westerfield, 1995).

A cDNA encoding full-length Ttrap (DQ524846) was isolated from a 5′cap-selected, normalized zebrafish embryo cDNA library by EST screening. It is 24 bp longer at the 5′ end than two other ttrap cDNAs (BC083404, BC097117, GenBank). MOs and mRNAs were injected into 1- to 2-cell-stage embryos. Plasmids containing cDNAs were linearized, column purified and subjected to in vitro transcription (Ambion mMessage mMachine High Yield Capped RNA kit), followed by poly(A)-tailing. For DFC-specific knockdowns, embryos were injected into the yolk between 2 and 4 hpf with fluorescence-tagged MOs. The degree of fluorescence in DFCs and forming KV were visually controlled using microscopy. For MOs used in this study, see Table S1 in the supplementary material.

Around 3000 antisense morpholino oligomers (MOs) targeting the putative start codons and/or 5′ untranslated regions (UTRs) of zebrafish mRNAs were designed using GenBank cDNA sequences and 5′ expressed sequence tags (ESTs) generated from normalized, full-length (5′ cap-selected)cDNA libraries encompassing various developmental stages and adult tissues. Targeted sequences were selected randomly. For all MOs that induced a phenotype at one concentration, additional concentrations were tested in a second round of screening (concentration range: 1-8 ng). A second MO(non-overlapping with the first MO sequence) was designed for each gene that was deemed `interesting' based on knockdown phenotype, protein structure, and functional data, and tested again for phenotypic specificity. For the screen in which TTRAP was initially identified as affecting vascular development,MO-injected embryos were subjected to in situ hybridization analysis at 28 hpf using a flk1 probe and subsequently in separate knockdown experiments analyzed by live imaging to observe blood flow and outgrowth of vessels.

In vitro transcription of digoxigenin-RNA probes and whole-mount in situ hybridization were performed according to Hauptmann and Gerster(Hauptmann and Gerster,1994).

Embryos were scored manually using light and fluorescence stereomicroscopy[Stemi-2000C and Lumar V12 (Zeiss), and MZ100FLIII (Leica)]. Digital images were captured using an AxioCam MRc5 and processed with Axiovision 4.5 Software(Zeiss).

For qRT-PCR, 15-20 embryos were pooled and RNA extracted (Tri-pure, Roche)and purified (RNeasy RNA purification columns, Qiagen). RNA extraction on single embryos was performed in a similar fashion. RT was performed using MuMLV reverse transcriptase (Revert-aid, Fermentas), oligo-dT and random primers. Real-time qPCR on single embryos was performed on ABI7000 using the SYBRgreen amplification reagent (Eurogentec). For cdh1, we used PCR primers: F, 5′-ATGATGTGGCGCCCACTTT-3′ and R,5′-CCGGTCGAGGTCTGTACTGAG-3′. PCR on whole-embryo cDNA was performed with primers: F, 5′-TGCTCATTGCTCAGGTGACTTT-3′ and R,5′-TTCTTGTTTGCCCAGCTGTTC-3′ to amplify a 251-bp region of ttrap cDNA, and primers: F, 5′-GCCTTCCTTCCTGGGCATGG-3′and R, 5′-CCAAGATGGAGCCACCGAT-3′ for a 251-bp region ofβ-actin cDNA.

To check wild-type and mutant TTRAP synthesis, 250 pg of mRNA made from pCS2-huTTRAP or pCS2-huTTRAP^T88A,T92A were injected into one-cell embryos. To test the efficacy of knockdown in vivo, 80 pg mRNA from pCDNA3-HA-zfttrap were injected either alone or together with 16 ng Ttrap^SCMO or Ttrap^MO. Western blot analysis was carried out on sonicated extracts from 20-30 pooled embryos; extracts were immunoblotted and proteins detected using anti-TTRAP, anti-HA or anti-tubulin antibodies.

2 μg MycAlk4-pCDNA3 or HA-Smad2/3/4-pCDNA3 were transfected into HEK293T cells, together with 2 μg FlagTTRAP-pCS2 or FlagTTRAP-frame-shift-pCS2(control). Co-immunoprecipitation studies of FlagTTRAP with Alk4 and Smads were performed as described (Pype et al.,2000). To test the interaction between TTRAP and Smad3, HEK293T cells were transfected with 2 μg TTRAP-pCS3 or TTRAP^T88A,T92A-pCS3 together with either HA-Smad3-pCDNA3 or FlagTTRAP-frame-shift-pCS2 (control).

Reporter constructs were injected into the cytoplasm of one-cell embryos. From a large collection of injected embryos, 15-20 (one set) were randomly selected and re-injected with 16 ng Ttrap^MO and control^MO. Embryos were allowed to develop to shield stage, lysed(100 μl passive lysis buffer, Promega) and 10 μl lysate was aliquoted in triplicate into 96-well plates. Lysates were incubated with two volumes luciferin (Promega) and measured for luciferase activity. The readouts from one triplicate set were averaged and treated as one data point. Fold induction was calculated by dividing the mean value for Ttrap^MO by the mean value for control^MO embryos. The pGL3 control assay was performed three times and the ARE-luciferase assay eight times. Statistical analysis was with the Student's unpaired t-test. The same conditions for experiments +/- sqt or cyc RNA were used (see above), and measurements performed in triplicate. Statistical analysis was carried out using ANOVA (One-way analysis of variance).

Strep-TTRAP-Myc-His was purified from HEK293T cells (Streptactin beads,IBA, Göttingen) and incubated with 250 ng Alk4 (Upstate, Lake Placid) and[γ-³²P]ATP for 10 minutes at 30°C. The reaction product was separated by PAGE, blotted and exposed to film. For in vivo phosphorylation, pStrepTTRAPMycHis was transfected into HEK293T cells with or without an expression vector encoding constitutively active Alk4. TTRAP was isolated using Ni-affinity purification under denaturing conditions.

PhosphoTTRAP was reduced (10 mM DTT, 45 minutes, 60°C), alkylated (35 mM iodoacetamide, 30 minutes, 24°C; 15 mM DTT, 30 minutes, 24°C) and separated from contaminating proteins by 10% tricine SDS-PAGE. The band of interest was excised and the protein digested (trypsin, overnight, 37°C). The resulting peptide mix was analysed by nanoLC-MS/MS, consisting of a precursor 79(-) ion scan to signal the presence of putative phosphopeptides and a product(+) ion scan to determine the phosphorylated residue. Nano LC-MS/MS was performed on a Dionex Ultimate capillary liquid chromatography system coupled to an Applied Biosystems 4000 QTRAP mass spectrometer. Peptides were separated on a PepMap C18 column developed with a 30 minute linear gradient (0.1% formic acid, 6% acetonitrile/water-0.1% formic acid, 40%acetonitrile/water).

Ttrap was initially identified in a screen on the basis of defects in cardiovascular development, as visualized by in situ hybridization for flk1 (Liao et al.,1997) and live analysis in the endothelium-specific transgenic eGFP line Tg(fli1:egfp)y1 (Lawson and Weinstein, 2002). In addition to vascular outgrowth defects,zebrafish embryos with a MO-mediated knockdown of Ttrap (Ttrap^MO)displayed pericardial edemas and abnormal blood circulation in trunk and tail(not shown). These abnormalities were consistently associated with other gross morphological defects, prompting us to titer MO doses in an attempt to uncouple dysmorphology from vasculature defects. Closer inspection of Ttrap morphants uncovered heart-looping defects, suggesting its involvement in LR patterning.

Ttrap^MO embryos (4 ng) displayed hallmarks of perturbed LR patterning. Observation of live embryos and whole-mount in situ hybridization(WISH) using cardiac myosin light chain 2 (cmlc2)(Yelon et al., 1999) revealed that 64% of Ttrap^MO embryos exhibited either reversed or no heart looping and occasionally cardia bifida by 48 hpf(Fig. 1; Table 1). We also scored Ttrap^MO embryos at 28 hpf, before morphological chamber specification. In these embryos, the direction of heart jogging was also randomized (Table 2). This indicates that the heart looping defects were not simply due to improper cardiac differentiation but rather to either a general defect in LR-axis determination or failure of cardiac primordial cells to read LR positional cues.

All three MOs, which target different regions of the Ttrap transcript (see Table S1 in the supplementary material), induced heart looping defects (see Table S2 in the supplementary material). A subset of embryos also displayed gastrulation defects (thickened germ ring) for all MOs tested singly (see Table S3 in the supplementary material). A higher penetrance of gastrulation defects was obtained with a 1:1 mix of MO1 (8 ng) and MO2 (8 ng). This mix induced thickening of the germ ring in 40-72% (separate experiments) of shield-stage embryos (Fig. 2A,B). Up to half of such embryos possessed a less distinct or absent shield. This combinatorial MO dose reduced Ttrap levels significantly(see Fig. S1 in the supplementary material). The phenotype could be partially rescued by injection of human (hu) TTRAP mRNA (see Table S4 in the supplementary material). The gastrulation defect is reminiscent of that observed in embryos where Nodal signaling is perturbed, particularly those with a knockdown of lefty1 and lefty2, which results in unrestricted Nodal signaling in the organizer(Feldman et al., 2002). The lack of a shield has also been reported for the homozygous Nodal mutant squint (sqt) (Dougan et al., 2003).

The germ ring defect was followed by incomplete convergent extension (CE)movements as determined by analysis of paraxial protocadherin(papc) (Yamamoto et al.,1998) and myogenic differentiation (myod)(Weinberg et al., 1996),marking paraxial and adaxial mesoderm, respectively(Fig. 2C-F). Epiboly was also hampered in Ttrap^MO embryos, with some undergoing yolk cell lysis between late gastrula stage (90% epiboly) and early somitogenesis (not shown). This lysis was probably caused by constriction of marginal cells in their attempt to achieve blastopore closure, despite their `regressed' position relative to the vegetal pole (Fig. 2I). Of the embryos that did not undergo yolk cell lysis, some managed to achieve full or partial (Fig. 2H) blastopore closure. Ttrap^MO embryos(Fig. 2K) displayed a shortened anterior-posterior axis, microcephaly, micropthalmia, and high incidence of pericardial edemas at 24 hpf, and cardia bifida (not shown) by 48 hpf. Together, these later phenotypes are similar to those resulting from CE movement defects (Matsui et al.,2005). The cardiac looping phenotype is unlikely to result from a general perturbation of gastrulation movements because at a lower MO dose (4 ng), the looping defects and cardia bifida were still observed in normally gastrulating embryos.

As depletion of Ttrap might also have randomized organ situs, we tested markers normally expressed left of the midline and markers of visceral organs(Fig. 3; Table 3): bone morphogenetic protein-4 (bmp4) for cardiac primordium(Chen et al., 1997; Schilling et al., 1999), paired-like homeodomain transcription factor-2 (pitx2)(Bisgrove et al., 1999; Campione et al., 1999) and southpaw (spaw) for left lateral plate mesoderm (LPM)(Long et al., 2003), forkhead box-A3 (foxA3) for liver, pancreas, gut(Odenthal and Nusslein-Volhard,1998; Alexander et al.,1999), and lefty1 (lft1) for the left dorsal diencephalon (Liang et al.,2000) (Fig. 3). These markers were either absent or visibly reduced in level of expression,expressed bilaterally or on the right side in Ttrap^MO embryos(Fig. 3A-E,F). LR-asymmetry defects were therefore not restricted to the heart, but observed along the entire rostral-caudal axis.

The observed phenotypes suggest a role for Ttrap in signaling of Nodal or perhaps Bmp (Whitman and Mercola,2001; Branford and Yost,2004; Kishigami and Mishina,2005). However, a direct involvement of Ttrap, at least in early Bmp signaling, is less likely because dorsal-ventral patterning did not appear to be affected in Ttrap^MO embryos, as evidenced by morphological assessment and normal expression of chordin and bmp2b (not shown).

We analyzed ttrap expression via RT-PCR(Fig. 4A) and WISH(Fig. 4B-E). Transcripts were detected in all blastomeres during cleavage stages, indicating that ttrap mRNA is maternal (Fig. 4A,B). Ubiquitous expression was observed during blastula stages and throughout gastrulation (Fig. 4A,C). Shortly after shield formation, weak expression in DFCs emerged (Fig. 4D). Between somite stages 5 and 9, expression in the tailbud and around the KV was detected above the more uniform expression in the embryo(Fig. 4E).

The persistence of LR defects (despite normal gastrulation movements in embryos injected with lower dose of Ttrap^MO), together with its expression in the KV, suggests that Ttrap is involved in node formation and/or function. However, defective shield formation in Ttrap^MO embryos also implied that forerunner cell fate could be affected, but DFCs are still present in Ttrap^MO embryos (see below). In addition, some Ttrap^MO embryos still have a (less distinct) shield, which may be sufficient to induce DFCs. Since Nodal signaling is important for node function (Essner et al.,2005), we addressed the role of Ttrap by exploiting a special feature of zebrafish: at midblastula transition, the syncytium between yolk and animal cells closes except for cytoplasmic bridges connected to DFCs, the cells that will form KV (Cooper and D'Amico, 1996; Essner et al.,2005). These channels remain open until ∼4 hpf. Between 2 and 4 hpf, MO injection results in DFC-specific knockdown(Amack and Yost, 2004);fluorescence-tagged MOs allow for visual control and selection of DFC-specific(DFCMO) injected embryos.

Ttrap^DFCMO embryos gastrulated normally, yet still displayed randomized heart looping or cardia bifida (48 hpf; Table 4), and again asymmetry markers were missing, expressed bilaterally or unilaterally on the opposite side (Table 5). Analysis of Ttrap^DFCMO embryos revealed that the KV was either absent or smaller (Fig. 4G), and normal in control^DFCMO embryos (Fig. 4F), even at 16 ng. We confirmed the KV phenotype using the node marker chemokine receptor-4 (cxcr4a)(Thisse et al., 2001)(Fig. 4H,I). These findings indicate that Ttrap plays a role in establishing LR asymmetry by regulating the formation of KV.

Flag-TTRAP plasmid was co-transfected with plasmids encoding HA-tagged Smad2/3/4 into HEK293T cells, and proteins were immunoprecipitated with anti-Flag antibody, and the precipitates resolved by SDS-PAGE, followed by western blot analysis for HA-Smad. TTRAP associated only with Smad3(Fig. 5A-C). We also tested co-immunoprecipitation between human TTRAP and mouse Alk4. Myc-Alk4 and Flag-TTRAP were co-produced, precipitated with anti-Flag antibody and the precipitates analyzed via blotting for Myc-Alk4. We observed binding of Alk4 to TTRAP (Fig. 5D). A triple complex between TTRAP, Alk4 and Smad3 was not detected, indicating mutually exclusive association of TTRAP with Alk4 and Smad3 (data not shown).

TTRAP could serve as substrate for ALK4 kinase. Purified TTRAP was incubated with human ALK4 and the reaction product analyzed by SDS-PAGE followed by autoradiography. ALK4 phosphorylated TTRAP in vitro(Fig. 6A). The band migrating at the position of TTRAP was excised and analyzed by LC-MS/MS. One TTRAP peptide was phosphorylated either on T88 and T92, or on T92 only(Fig. 6B). T88 in TTRAP is highly conserved across species, whereas T92 is exclusive to human, dog and chicken (see Fig. S2 in the supplementary material). We tested in vivo phosphorylation of Strep-TTRAP by co-expression with mouse Alk4 in HEK293T cells, and affinity-purified TTRAP. In this preparation TTRAP-specific peptides were reproducibly found in both the singly (T92) and doubly phosphorylated form (T88/T92) (not shown).

We then mutagenized TTRAP by substituting Thr88 and Thr92. TTRAP^T88A,T92A was still capable of associating with Smad3 (see Fig. S3 in the supplementary material) and we therefore tested this mutant in vivo. As mentioned previously, huTTRAP mRNA can partially rescue the Ttrap^MO phenotype (see Table S4 in the supplementary material) and therefore huTTRAP can substitute for zfTtrap. TTRAP^T88A,T92A-RNA-injected embryos were completely normal(Fig. 6C) despite detection of TTRAP^T88A,T92A protein in these embryos between 2 and 6 hours after injection at levels similar to wild-type TTRAP (see Fig. S3 in the supplementary material). To rule out the possibility that endogenous Ttrap was somehow rescuing any potential effect(s) of TTRAP^T88A,T92A, we co-injected TtrapMO with TTRAP^T88A,T92A RNA. This yielded a similar penetrance of gastrulation defects (absence of shield and thickened germ ring)in embryos injected with TtrapMO alone (35%, n=40 for TtrapMO+TTRAP^T88A,T92A mRNA (300 pg) versus 38% (n=42) for TtrapMO only; Fig. 6C,D, and not shown). Thus, mutant TTRAP^T88A,T92A is not able to rescue the TtrapMO phenotype, suggesting that phosphorylation of Ttrap on Thr88 and Thr92 is essential for Ttrap function.

The results of the co-immunoprecipitation(Fig. 5) suggest a role for Ttrap in modulating Alk4/Smad3 signaling. To test whether binding of TTRAP to Smad3 directly modulates Smad3 activity, we performed reporter assays using a Smad2/3-responsive element in vivo. ARE-luciferase reporter(Yeo et al., 1999) (or pGL3 as negative control) was co-injected with Ttrap^MO into embryos, and luciferase measured in lysates from pooled embryos at shield stage. Co-injection resulted in ∼fivefold higher Smad2/3 activity over control^MO (mean value 4.6±2.5, P≤0.01; Student's unpaired t-test) (Fig. 7A). To test whether this increase is dependent on Nodal signaling, we repeated the assay in the presence of Sqt or Cyc. The addition(via RNA injection) of either ligand to Ttrap^MO embryos potentiated ARE-luciferase tenfold above the activity detected with Sqt (or Cyc) but without Ttrap^MO (Fig. 7B). Thus, Ttrap appears to negatively modulate Nodal signaling in vivo.

We also tested whether expression of Smad3 targets was misregulated in Ttrap^MO embryos. In contrast to Smad2, there is a striking paucity for known Smad3 targets with regard to Nodal/Alk4 signaling. In Xenopus, one reported target of both Smad2 and Smad3 is Mix-2 (Yeo et al.,1999). The fish mutant bon harbors a mutation in the mixer-like gene and displays cardia bifida(Chen et al., 1996; Stainier et al., 1996; Kikuchi et al., 2000). The characterization of the bon promoter and its activation by Smad2/3 has not been reported. We observed an increase in bon staining in Ttrap^MO embryos (Fig. 7C,D). We also tested whether Smad3 RNA injection would upregulate bon. In zebrafish two smad3 genes exist, smad3a(Dick et al., 2000) and smad3b (Pogoda and Meyer,2002). Because a gastrulation phenotype has been reported for Smad3b RNA-injected embryos and smad3b is expressed in the tailbud region (i.e. in the vicinity of KV), we focused on smad3b. Importantly, Smad2 overproduction does not result in gastrulation defects(Muller et al., 1999; Dick et al., 2000). In embryos overexpressing Smad3b (Smad3b^OE), endogenous bon was strongly upregulated (Fig. 7E,F; and data not shown).

To determine whether modulation of Smad3 activity by Ttrap depends on Alk4 signaling, we soaked embryos from dome stage onwards in the Alk4/5/7 inhibitor SB431542 (Inman et al., 2002),50 μM of which phenocopied the cyc;sqt double mutant by shield stage (Feldman et al.,1998) (not shown) and abolished bon expression in Ttrap^MO embryos (Fig. 7G,H). Thus, the upregulation of bon in Ttrap^MO embryos appears to depend on Alk signaling. Ttrap mRNA levels are not affected by SB431542 (not shown). Morphological observation of live Smad3b^OE embryos and WISH for papc and myodrevealed CE and epiboly defects similar to Ttrap^MO embryos(Fig. 7L-N). In addition,overexpression of Smad3b in DFCs also induced a low percentage of heart-looping defects (see Table S5 in the supplementary material). Since targeting of mRNA to DFCs has not been previously reported, we initially determined whether DFC-specific expression could be achieved, using eGFP RNA. Fluorescence was detectable in the node at the time of KV formation, and all embryos developed normally. However, in the majority of embryos, only part of the KV was fluorescent, indicating that not all DFCs were targeted or expressed eGFP. Therefore, DFC-RNA overexpression may not be as efficient as DFC-MO injections (see Fig. S4 in the supplementary material). Nevertheless,about 20% of DFC-Smad3b mRNA-injected embryos displayed heart-looping defects.

If Ttrap knockdown increases Smad3 activity, then simultaneous reduction of Smad3 in Ttrap^MO embryos should rescue the Ttrap knockdown phenotype. Indeed, graded double knockdowns of Ttrap-Smad3b rescued up to 70%of Ttrap^MO embryos with gastrulation and node formation defects(Fig. 8A-C and Tables S6-S8 in the supplementary material). Consistent with our co-immunoprecipitation data,Ttrap-Smad2 double knockdowns did not rescue gastrulation defects in Ttrap^MO embryos (see Table S9 in the supplementary material). Single Smad2 knockdown resulted in a curved, shortened body axis, anterior truncation, and loss of floorplate (see Fig. S5 in the supplementary material), reminiscent of mutant sur(Pogoda et al., 2000; Sirotkin et al., 2000) and similar to the phenotype described for Smad2 knockdown in Xenopus(Rana et al., 2006).

Nodal regulates its own expression and that of its antagonists(Bisgrove et al., 1999; Meno et al., 1999; Cheng et al., 2000; Feldman et al., 2002). We tested whether Ttrap knockdown would alter expression of cyc, sqt(Rebagliati et al., 1998a; Rebagliati et al., 1998b) and lft1. Ttrap was neither required for initiation nor maintenance of cyc, sqt and lft1 expression at germ ring stage and 70%epiboly (not shown). This is consistent with our data showing no detectable interaction between TTRAP and Smad2, since expression of cyc, sqt and lft1 is regulated primarily by Smad2/FoxHI(Schier, 2003). Ttrap may therefore play a role in distinguishing between Smad2 and Smad3 signaling.

To test whether Ttrap regulation of gastrulation movements is due to a direct effect on cell motility or perturbation of mesendodermal cell fate, we transplanted wild-type and Ttrap^MO cells sequentially from the lateral margin into the germ ring of maternal-zygotic one-eyed pinhead (MZoep) embryos (see Fig. S7 in the supplementary material) (Schier et al.,1997; Zhang et al.,1998; Gritsman et al.,1999). In contrast to wild-type cells, Ttrap^MO cells neither internalized nor migrated to regions normally occupied by endodermal progenitors (as observed for wild-type cells that continue to express axial in MZoep mutants)(Gritsman et al., 1999). Thus,it appears that Ttrap function (at least its modulation of Smad3 activity) is non cell-autonomous in this context. Moreover, MZoep mutants are not rescued by Ttrap knockdown (not shown). This implies that Ttrap knockdown cannot compensate for the lack of Smad2 activation in this Nodal-insensitive oep background. This, and our reporter studies, supports our hypothesis that the Ttrap-mediated increase in Smad3 activity is Nodal dependent. In addition, the cell transplantation data implicate a role for Smad3 in specific aspects of Nodal signaling.

Common candidate target genes of Ttrap and Smad3 would help to explain the gastrulation and/or node formation defects in Ttrap^MO embryos. cdh1 posed as candidate because the cdh1 mutant half-baked (hab) displays epiboly and CE defects similar to Ttrap^MO embryos, and cdh1 is expressed in DFCs(Kane et al., 2005; Shimizu et al., 2005).

Because hab:cdh1 null mutants do not survive beyond gastrulation, we again exploited DFC-specific knockdown to determine whether cdh1 plays a role in node formation. We found randomized heart looping, cardia bifida and smaller/absent KV in cdh1^DFCMO embryos(Table 6). Analysis of DFCs using cas/sox32 (Dickmeis et al.,2001; Kikuchi et al.,2001) and prior to node formation revealed that DFCs are present in Ttrap^DFCMO embryos (Fig. 9A,B). Moreover, at shield and 70-80% epiboly stages, no significant difference in DFC number between Ttrap^MO and control^MO embryos was detected (see Fig. S8, Table S10 and Fig. S9 in the supplementary material). However, Ttrap^MO DFCs do not always converge at the midline to form one tight cluster of cells below the shield or later in gastrulation. Rather, cells are more widely dispersed in a broad stripe along the lateral axis of the embryo(Fig. 9B; see Fig. S8, Table S11, and Fig. S10 in the supplementary material), suggesting a defect in cell migration. In line with our DFC data in Ttrap^MO embryos, DFCs are still present in hab:cdh1 mutants(Kane et al., 2005). The same DFC defect was also observed in Smad3b^OE embryos(Fig. 9D) and may reflect an inability of these cells to organize into KV.

We tested whether the DFC defects were due to loss/reduction in cdh1 in Ttrap^MO and Smad3b^OE embryos. At 70%ebiboly (Fig. 9E-H), cdh1 mRNA (as assessed in these deliberately overstained embryos) was absent in DFCs and the anterior axial hypoblast in both cases(Fig. 9F,H). No significant difference in expression within the epiblast could be detected based on macroscopic observation, but qRT-PCR analysis revealed an overall reduction(30%) of cdh1 transcript levels in Ttrap^MO embryos[1.60±0.59, n=30 (Ttrap^MO) vs 2.3±0.13, n=15 (control^MO); P≤0.01, Student's unpaired t-test; not shown). This limited but consistent reduction may be explained by a Ttrap-Smad3-mediated downregulation of cdh1 that is localized only to regions where Nodal signaling is present and/or can be sensed, and may therefore be partly masked by more general expression in the rest of the blastoderm.

Sequence alignment of e-cadherin promoters (not shown) revealed no conserved Smad3 sites, making it unlikely that e-cadherindownregulation occurs via direct binding of Smad3. One E-box containing the Snail-binding sequence CACC was strictly conserved between human, mouse and zebrafish. Several repressors including Snail bind within E-box elements and inhibit e-cadherin expression(Cano et al., 2000; Grooteclaes and Frisch, 2000; Comijn et al., 2001; Hajra et al., 2002; Eger et al., 2000), and Snail is induced by TGFβ or Nodal(Fujimoto et al., 2001; Hajra et al., 2002; Peinado et al., 2003; Gotzmann et al., 2006; Bennett et al., 2007).

At 60% epiboly, snail1a (snai1a) expression is restricted to the blastoderm margin and presumptive paraxial mesoderm, but is excluded from the dorsal most region of the shield (presumptive axial mesoderm)(Hammerschmidt and Nüsslein-Volhard,1993; Thisse et al.,2001). DFCs are in close proximity to these dorsal cells and do not express snai1a. We questioned whether the increase in Smad3 activity in Ttrap^MO embryos would be sufficient to cause ectopic expression of snai1a in either axial mesoderm and/or DFCs, thereby contributing to cdh1 downregulation. Knockdown of Ttrap resulted in misexpression of snai1a in axial mesoderm (60% epiboly; Fig. 10A,B) and DFCs (not shown). To determine whether this expression is mediated by Smad3, we performed Ttrap-Smad3 double knockdowns to test for reversion to the normal snai1a domain. Double knockdown resulted in a 58% rescue of embryos with ectopic snai1a in the axial mesoderm(Fig. 10C,D; Fig. 10E for graphical depiction of rescue; data not shown). Curiously, neither the single Smad3b nor Smad3a knockdown (and their combination) resulted in downregulation of snai1a (not shown). Moreover, Smad3b knockdown did not result in any visible gastrulation phenotype/s (our unpublished observations), despite the presence of transcripts throughout early development(Dick et al., 2000; Pogoda and Meyer, 2002). By 24 hpf however, morphological defects similar to the Smad2 knockdown phenotype such as head degeneration, absence of floorplate, and curved, shortened body axis could be observed (see Fig. S6 in the supplementary material). The lack of a phenotype before 24 hpf may be attributed to compensation by either Smad2, Smad3a, and/or Smad2/3 maternal protein/s. Finally, we performed Ttrap-Snail1a double-knockdowns to test for rescue of Ttrap-induced LR defects. This resulted in a 62% rescue of heart looping defects (48 hpf; see Table S12 and Fig. S11 in the supplementary material).

Ttrap knockdown perturbs gastrulation movements. TTRAP binds to Alk4 and Smad3, and TTRAP limits Smad3 transcriptional activity. The similarities in CE and epiboly defects between the Ttrap knockdown embryos and hab:cdh1 mutants led us to investigate the link between ttrap and cdh1. The importance of Cdh1 in controlling CE and epibolic movements during gastrulation is known(Babb and Marrs, 2004; Kane et al., 2005; Shimizu et al., 2005; Ulrich et al., 2005). Our data suggest that: (1) cdh1 is downstream of Alk4-Smad3 and that cdh1 expression may, in part, be controlled by Ttrap through attenuation of Smad3 activity; and (2) that knockdown of Ttrap increases Smad3 activity, which in turn downregulates cdh1, thereby leading to impaired CE and epiboly. Our results also implicate snail as a potential link between increase in Smad3 activity and downregulation of cdh1,possibly via a snail-binding site in the cdh1 promoter.

Knockdown of Ttrap does not appear to affect mesodermal or endodermal cell fate, because bon and cas expression persist in mesendoderm and presumptive endoderm, respectively. This observation was also true for meso/endodermal bhik, mix, ntl and gsc (not shown),including in Ttrap^MO embryos with thickened germ ring. We therefore propose that ttrap is primarily involved in cell migration in early embryos. The results of the cell transplantations (both with wild-type and with Ttrap^MO cells) into a defective Nodal signaling background(MZoep) indicate a possible distinct function for Smad3 in directing cell movement. TTRAP has been implicated in migration of cancer cells through its interaction with ETS1/2 and FLI1 (Pei et al., 2003). However, knockdown of Ets1 in fish did not yield gastrulation defects (our unpublished observations), and fli1expression only begins after gastrulation at 10 hpf(Brown et al., 2000), making fli1 an unlikely Ttrap target with respect to regulation of CE movements and epiboly.

LR-axis determination in fish(Kramer-Zucker et al., 2005)and mouse, and the resulting laterality of heart and viscera, is initiated in part by the action of monocilia residing within the node. This structure consists of a `pit' of cells, with each cell protruding one monocilium, which in mouse is posteriorly tilted at an angle of 60°(Nonaka et al., 2005). The first symmetry-breaking event occurs when these monocilia beat in vortical fashion to direct unidirectional fluid flow, resulting in accumulation of proteins left of the node. Physical models that can mimic such flow have shown that even this small difference in Nodal flow is subsequently converted through reaction-diffusion mechanisms involving Nodal/Lefty proteins into a robust asymmetrical target gene expression(Okada et al., 2005; Hirokawa et al., 2006). This results in activation of target genes in the LPM, which endow `leftness' to this side of the embryo and activate asymmetrical differentiation of organ primordia. However, the mechanism(s) by which these asymmetric signals are translated into morphology is not well understood(Shiratori and Hamada, 2006). Genes expressed specifically in the right LPM also function in LR determination, and other studies have also implicated the intact midline as serving a barrier function between left- and right-sided factors(Roessler and Muenke, 2001; Tabin and Vogan, 2003; Yost, 2003; Raya and Belmonte, 2004; Levin, 2005; Raya and Belmonte, 2006).

Our findings suggest that Ttrap plays a role both in the earliest steps of KV (node) formation and gastrulation. DFC-specific Ttrap knockdown embryos gastrulate normally, yet DFC behavior is abnormal, resulting in a KV that is strongly reduced in size, or absent. The laterality defects observed in Ttrap^DFCMO embryos are consistent with LR defects obtained after DFC ablation (Essner et al.,2005). Moreover, we provide results implicating cdh1 as a possible target of Ttrap and Smad3 for regulating gastrulation movements and KV formation. E-cadherin (cdh1) is an established player in mediating cell (de)adhesion/migration in embryos and invasive tumors. Our results suggest that cdh1 may play a role in LR-axis determination. It must be noted, however, that neither KV nor LR defects have been described for cdh1 mutants to date. It therefore remains to be seen what the precise role is for Cdh1 with regard to LR patterning and DFC migration. Our data suggests that Ttrap regulates cdh1 via Smad3 as opposed to Smad2. This is supported by a recent siRNA study on the epithelial-to-mesenchymal transition (EMT) of proximal-tubule epithelial cells, which showed a Smad3-dependent (and Smad2-independent) downregulation of cdh1 following stimulation of cells with TGFβ(Phanish et al., 2006).

Smad2 and Smad3 share over 90% identity and a number of overlapping functions, such as the co-regulation of the Nodal targets bon and snail (Bennett et al.,2007). Although our results indicate that bon and snail can be regulated by Smad3, our data do not rule out co-regulation by Smad2, and regulation of these genes most likely occurs via cooperation between these two Smads. Nevertheless, there are structural/functional differences between both Smads, several of which appear to distinguish their actions in vitro (Yew et al., 2004; Uemura et al.,2005; Ju et al.,2006) and in vivo (Dunn et al., 2005; Wang et al.,2006). In addition, the ratio of Smad2 versus Smad3 influences their respective roles as effectors (Dunn et al., 2004; Kim et al.,2005). The functional differences between Smad2 and Smad3 may also depend on their ability to associate with various co-factors that mediate distinct responses to TGFβ (Attisano et al., 2001). These co-factors include Fox proteins(Nagarajan and Chen, 2000) or the Smad-interacting protein Smicl, which primarily modulates Smad3 activity during Nodal-dependent induction of chordin in the Spemann organizer(Collart et al., 2005). Clearly, additional Smad2- and Smad3-specific targets and co-factors remain to be identified. Ttrap may be one such co-factor in Nodal-Alk4-Smad3 signaling. Intriguingly, the Ttrap knockdown phenotype does not entirely mimic the Nodal overexpression phenotype (Toyama et al.,1995). In the latter study, Nodal mRNA injection into fish embryos resulted in axis duplication and ectopic organizer formation, phenotypes not observed in Ttrap knockdown embryos. Thus, Ttrap knockdown does not result in a general over-activation of Nodal signaling, which would also encompass Smad2 activity.

The role of Ttrap as a co-factor for ETS(Pei et al., 2003) suggests that intranuclear Ttrap may play a similar role with regard to Smad3. This is supported by the finding that Ttrap binds to SUMO proteins, which are implicated in a number of cellular processes including transcription(Hecker et al., 2006). In any case, several lines of evidence support the hypothesis of Ttrap as modulator of Alk4-dependent Smad3 activity: (1) the association between TTRAP and Smad3 is mutually exclusive with Alk4; (2) the high degree of overlap between Ttrap^MO and Smad3b^OE phenotypes; (3) the ARE-luciferase data; and (4) the rescue of the Ttrap knockdown phenotype via Smad3 knockdown. However, the functional mechanism underlying this modulation remains to be investigated in detail.

The roles of Smad2/3 as effectors have been extensively characterized in the mouse, including elegant studies that address the effects of changing their ratio in vivo in Nodal-controlled mesoderm formation(Dunn et al., 2004; Dunn et al., 2005). Our data show that Smad3 plays an important role in zebrafish in controlling cell migration and/or (de)adhesion through Nodal-Alk4. Sqt is the most likely ligand here because of its reported role in regulating gastrulation movements and its expression in DFCs (Rebagliati et al., 1998b; Feldman et al., 2000). Ttrap may serve as limiting factor for Smad3, perhaps to maintain a balance between Smad2 (the DNA-binding splice form) and Smad3 signaling, because both are capable of occupying the same promoter sites of target genes for Nodal.

Using biochemical studies and phenotypic analysis in zebrafish, we have uncovered a role for Ttrap as a novel player in TGFβ signaling in vivo. Our findings suggest that this protein is essential for regulating Nodal signaling, at least by limiting the early developmental activity of Smad3. Given that extranuclear Ttrap and Alk4 also interact either directly or are present together in a complex, it is possible that Alk4 itself, via a negative feedback loop involving Ttrap and possibly other co-factors, functions to regulate the level of its own signal transduction cascade. This type of higher order regulation fits in the concept of self-enabled gene response cascades(Massagué et al., 2005)and has been observed in TGFβ-Smad signaling. Although further studies are needed to elucidate the exact mechanism by which Ttrap modulates Alk4-Smad3 activity, our data underscore the importance of tightly fine-tuning TGFβ-Smad pathways in embryos.

We thank D. Stainier, D. Meyer, D. Kane, B. Thisse, M. Rebagliati, J. Yost,A. Zwijsen, L. van Grunsven, S. Plaisance, C. Hill, H. Hamada, J. C. Belmonte,C. P. Heisenberg and H. Peeters for valuable advice and sharing reagents. We thank F. Rosa for valuable advice and making MZoep fish available and K. Schildermans for technical assistance. C.V.E. was supported by VIB and the Desiré Collen Research Foundation. The D.H. group was supported by VIB,IUAP 5/35 and 6/20, and the EC Integrated Project EndoTrack(LSHG-CT-2006-019050).