Twisted gastrulation enhances BMP signaling through chordin dependent and independent mechanisms

Bone morphogenetic protein (BMP) signaling is regulated by complex interactions of inhibitors, proteases and other proteins(Balemans and Van Hul, 2002). The BMP inhibitor Chordin was first identified in Xenopus as a factor capable of rescuing axis formation in ultraviolet treated embryos(Sasai et al., 1994), and was subsequently shown to bind BMPs and prevent them from activating their receptors (Piccolo et al.,1996). Many components of the BMP signaling pathway are functionally conserved in Drosophila, including short gastrulation (sog), a homolog of Chordin, and decapentaplegic (dpp), a homolog of a subset of the vertebrate BMPs (Francois and Bier,1995; Holley et al.,1995; Schmidt et al.,1995).

Chordin contains four cysteine-rich (CR) domains, similar to those found in a number of extracellular matrix proteins(Sasai et al., 1994). High-affinity binding of Chordin for BMPs resides in its CR domains(Larraín et al., 2000);CR1 and CR3 have significant binding affinity as individual domains, although approximately one-tenth that of the full-length (FL) protein, and the biological activity of FL Chordin and its fragments roughly parallels their binding affinities. These data predict that cleavage of Chordin into smaller fragments would significantly reduce its biological activity as a BMP inhibitor.

tolloid (tld) was first identified as a zygotic Drosophila gene required for dorsoventral (DV) patterning, and homologous to vertebrate Bmp1(Shimell et al., 1991). Tld cleaves Sog at two sites, presumably explaining its ability to enhance Dpp signaling by decreasing the levels of a Dpp inhibitor(Marques et al., 1997). In addition to Bmp1, several other Tld-related vertebrate enzymes have been identified. Some of these similarly cleave Chordin in vitro(Blader et al., 1997; Piccolo et al., 1997; Scott et al., 1999), and when misexpressed in the embryo can antagonize Chordin function(Goodman et al., 1998; Piccolo et al., 1997).

Direct genetic evidence for the importance of vertebrate Tld proteins in regulating Chordin has been less clear. The zebrafish DV patterning gene mini-fin (mfn; tll1 – Zebrafish Information Network) encodes a Tld-related enzyme(Connors et al., 1999), which cleaves Chordin in vitro (Blader et al.,1997). However, mfn mutants have no phenotype during gastrulation, when chordin expression is highest, and the mutation has not been correlated with abnormalities in Chordin cleavage in vivo. In mouse, the Bmp1 and Tll1 genes have each been knocked out(Clark et al., 1999; Suzuki et al., 1996),revealing functions in heart septation, body wall closure and collagen processing. However, none of these roles have been correlated with abnormalities in Chordin cleavage in the single mutant mice. Analysis of cells from single and double mutants has shown redundant function for the two enzymes in cleaving Chordin and other substrates(Pappano et al., 2003). Genetic analysis in vertebrates is clearly complicated by the fact that multiple enzymes cleave Chordin, and redundancy masks the importance of this regulatory mechanism.

The Drosophila twisted gastrulation (tsg) gene is required for peak Dpp signaling in the dorsal embryo, and also cooperates with Sog to inhibit Dpp signaling (Mason et al., 1994; Mason et al.,1997). Vertebrate Tsg genes have been reported both to enhance(Oelgeschlager et al., 2000; Zakin and De Robertis, 2004)and inhibit (Blitz et al.,2003; Chang et al.,2001; Ross et al.,2001; Scott et al.,2001) BMP signaling. To reconcile these findings, a model has been proposed in which Tsg acts in two steps, first enhancing the binding of Chordin to BMP, then after cleavage helping to displace the Chordin fragments(Larraín et al., 2001). According to this model, the amount of Tld activity determines the balance between these two counteracting functions. In vitro, Tsg increases cleavage of mouse Chordin at the two identified Tld cleavage sites and at an additional,intermediate site not used in the absence of Tsg(Scott et al., 2001). A similar cleavage of Sog occurs in the Drosophila embryo, generating a more potent Dpp inhibitor termed `Supersog'(Yu et al., 2000). If such a cleavage occurs in vivo for the vertebrate protein, it would ascribe a positive role to both Tld and Tsg in Chordin regulation.

We assessed the role of Tld in regulation of Chordin in the zebrafish gastrula and examined the effect of Tsg on Chordin cleavage and function. Through alterations of conserved residues near the cleavage sites, we created Chordin mutants that retain BMP-inhibitory activity but are resistant to Tld cleavage at one or both sites. RNAs encoding wild-type and cleavage mutant(CM) Chordins were used to rescue zebrafish chordin/dino(din; chd – Zebrafish Information Network) mutant embryos. Prevention of cleavage at either site enhances the ability of Chordin to rescue din mutants, confirming the importance of cleavage as a regulatory mechanism. We also provide evidence that the product of the downstream cleavage event is a stronger BMP inhibitor than the FL protein,suggesting a positive role for cleavage in Chordin regulation. We show that Chordin cleavage is extremely rapid in vivo, and that redundant enzymes cleave Chordin in mfn mutants. However, we find no evidence of alternative cleavage sites being used in either the wild-type or CM proteins. Furthermore,we show that endogenous Tsg decreases steady state Chordin in the embryo,reducing its effectiveness as a BMP inhibitor. We reassessed the tsg1(tsga – Zebrafish Information Network) morphant phenotype, both in wild-type and din mutant backgrounds; we find that the predominant effect of Tsg in the zebrafish gastrula is to enhance BMP signaling, and that it can also do so independently of Chordin.

Fish were cared for using standard methods(Westerfield, 1995). The din^tt250 and mfn^tm124a alleles were maintained through intercrossing of heterozygous or homozygous mutant fish. For all injection experiments, clutches of mutant embryos were obtained by intercrossing din^-/- or din^-/-;mfn^-/- fish.

The QuikChange site-directed mutagenesis kit (Stratagene) was used to introduce mutations into the zebrafish chordin coding sequence,starting with a previously described injection construct in the pCS2+ vector(Miller-Bertoglio et al.,1997). The sequence surrounding the upstream cleavage site was altered to TTCTTCAACAACAGA, and surrounding the downstream cleavage site to ATGTTGCAGGCGAACGGG (altered bases are in bold). All constructs were verified by sequencing.

The six-copy Myc tag was excised from the pCS2+MT vector with ClaI and XbaI and inserted following the wild-type and CM-chordincoding sequences. To verify that the tagged RNAs gave rise to full-length,stable proteins, they were subjected to in vitro translation and the resulting proteins detected by western blotting for Myc (data not shown). In initial experiments, similar amounts of tagged and untagged RNAs rescued dinmutants, showing that the Myc sequences did not adversely affect protein activity or stability (data not shown). Therefore, subsequent rescue experiments were performed with the C-terminal tagged constructs.

To construct N-terminal tagged chordin vectors, a linker encoding the first 29 amino acids of Chordin was synthesized and inserted into the BamHI site of pCS2+MT. Then the remainder of the wild-type and CM-chordin coding sequences were amplified by PCR and inserted into the EcoRI and XbaI sites downstream of the Myc tags. Although placement of the Myc tag at the N-terminus does somewhat destabilize the protein, these RNAs also rescue din mutants at levels comparable with the untagged RNAs, showing that the proteins are functional.

To construct the vector encoding the N+I fragment containing the mutation of the upstream cleavage site (N+I_A), the sequence encoding amino acids 1 to 849 of CM_A was amplified and inserted into the pCS2+MT vector between BamHI and ClaI.

As a control for quantification of Myc-tagged Chordins, a construct was made encoding a GFP-Myc fusion protein with the signal peptide of Chordin added at its N terminus. This RNA was co-injected with tagged chordinRNAs, and the amount of its product used as an internal standard for quantification.

The sequence of tsg1-MO1 has been previously described(Ross et al., 2001). The non-overlapping MO5 (CGCCGAACTCTGAGCTGAGCAGAAC), the four-base mismatch to MO1(CTCATGTTGATGATGAACACCGCAT) and the five-base mismatch to MO5(CCCCCAACTCTCAGCTCAGCACAAC) were gifts from M. Mullins.

RNAs for injection were transcribed with the mMessage mMachine Sp6 kit(Ambion) from NotI linearized templates. RNAs were quantified by spectrophotometry and the amounts confirmed by agarose gel electrophoresis. For each injection experiment, the RNAs encoding different versions of Chordin were synthesized, purified and quantified in parallel. RNA injections were performed as previously described (Fisher and Halpern, 1999). On the following day, embryos were scored using standard phenotypic indicators of excess or decreased BMP signaling(Hammerschmidt et al., 1996; Mullins et al., 1996). For the rescue experiments, `ventralized' embryos resembled din mutants(small brain and somites, excess blood in ventral tail, multiplicated fin folds), `rescued' embryos appeared wildtype or mildly dorsalized (Class 1, or partially absent ventral fin fold) and `dorsalized' embryos were those in Class 2-5 (more severe tail defects or truncations, tail curled on top of yolk, some or all somites expanded to encircle the yolk).

Some morphant embryos were fixed at 80% epiboly or 8 to 12 somite stages,and in situ hybridizations performed as previously described(Miller-Bertoglio et al.,1997), using bmp4, chd, evel, gsc, gata2, krox20(egr2b – Zebrafish Information Network) and myod as markers indicative of DV patterning.

RNA encoding Tsg1 was transcribed as above and co-injected to rescue the tsg1 morphant phenotype. The construct containing the Xenopus tsg1 coding sequence, with the signal peptide replaced with that of ECM protein BM40/SPARC, was a gift from M. O'Connor.

Embryos were collected and homogenized in ice cold extraction buffer [250 mM sucrose, 4 mM EDTA, 100 mM NaCl, 10 mM Tris-HCl (pH 7.6)] with added protease inhibitor cocktail (Roche). After centrifugation, the supernatant was incubated with agarose-coupled 9E10 anti-Myc antibody (Santa Cruz Biotech). The pellet was collected by centrifugation, washed four times with RIPA buffer(PBS, 1%NP-40, 0.5% sodium deoxycholate, 0.1% SDS) and one time with 50 mM Tris (pH 6.8). The pellet was resuspended in electrophoresis sample buffer and analyzed by western blotting. Protein gel electrophoresis and immunoblotting were performed according to standard protocols(Harlow and Lane, 1988; Westerfield, 1995). 9E10 anti-Myc antibody (Santa Cruz Biotech) was used to detect tagged Chordin fragments, and blots were visualized with the ECL Plus kit (Amersham). Some blots were scanned with a Storm Phosphorimager; individual bands were quantified using the local average method.

Sequences surrounding the two Tld cleavage sites in Chordin are conserved in several vertebrate species, including zebrafish(Scott et al., 1999). We altered conserved residues at the two sites, to make them resistant to cleavage (Fig. 1A). RNAs encoding cleavage mutant Chordin with the first, second, or both, sites altered (CM_A, CM_B, and CM_AB-chordin)were injected to rescue Chordin-deficient din^-/- embryos(Fig. 1B,C) as described previously (Fisher and Halpern,1999). In initial experiments, successively lower amounts of each RNA were injected until optimal rescue was achieved. CM_AB RNA was∼10-fold more potent than wild-type chordin in the rescue assay,demonstrating the importance of Tld cleavage in vivo as a mechanism to limit Chordin activity. We predicted that partial cleavage products of Chordin would have reduced BMP-binding affinity, and therefore the partial cleavage mutants would be intermediate in efficacy between wild-type and CM_AB. CM_B was ∼3-fold more potent than wild type, fitting this prediction. However, CM_A was more potent than CM_AB, a result we have repeated in several independent experiments and with multiple batches of RNA (see also Fig. 3A, Fig. 4A).

One possible explanation for the efficacy of CM_A is that cleavage at the upstream site is prerequisite to downstream cleavage. To follow the processing, we injected chordin with Myc epitope tags added to the C terminus into din^-/- embryos, which were lysed and subjected to immunoprecipitation and western blotting(Fig. 1D, left blot). The mutations introduced to the Tld sites largely prevented cleavage as predicted,and cleavage at the two sites occurred independently. The major product from wild-type Chordin was consistent in size with the predicted C-terminal cleavage product. Quantitation of a similar blot showed that in early gastrulation, 6 hours after injection, 93% of the Myc tag was in the C-terminal fragment (data not shown), demonstrating robust, rapid cleavage activity. By contrast, CM_AB gave rise to a single prominent band representing FL protein 20 hours after injection. A small amount of lower molecular weight protein could be seen; however, this did not seem to represent cleavage at the normal site, as it appears as a smear rather than a sharp band on longer exposure (data not shown). The C-terminal fragment was the predominant product from CM_A, showing that cleavage at the downstream site is unaffected by mutation of the upstream site. Similarly,from CM_B the major band was slightly smaller than FL protein,produced by cleavage at the upstream site. Interestingly, following injection of the same amounts of RNA, more protein accumulated from CM_B than from the other constructs, a result observed consistently in multiple experiments (see also Fig. 2A). This suggests that a sequence near the N terminus, removed by cleavage at the upstream site, destabilizes Chordin protein.

Although studies on mouse Chordin showed cleavage at an alternative site in the presence of Tsg (Scott et al.,2001), our examination of the C-terminal cleavage products failed to reveal any novel fragments. However, alternative cleavage might be evident only by examination of N-terminal fragments. Therefore, we constructed chordin RNAs with Myc epitope tags added to the N terminus,immediately following the signal peptide. When these RNAs were injected into din mutants, the major fragments corresponded to predicted cleavage products (Fig. 1D, right blot);the band at ∼50 kDa is non-specific and also observed in negative control samples not injected with Myc-tagged RNA. In conclusion, we see no evidence of alternative cleavage products by examining proteins labeled at either the C or N terminus, even when cleavage is prevented at both of the normal Tld sites.

Because of the enhanced accumulation of labeled protein from CM_B, we wanted to also compare the stability of the N-terminal fragments. They generally appeared less stable than C-terminal fragments, but we directly tested the effect of Myc tag position on stability. We injected two groups of embryos with the same amount of CM_AB RNA, labeled at either the C or N terminus, and processed the samples in parallel. Similar amounts of labeled proteins were detected 5 hours after injection, but substantially less N-terminal labeled protein at 20 hours (data not shown). Therefore, we could not compare the stability of FL Chordin and the cleavage products over longer periods. However, over shorter time periods, there did not appear to be greater accumulation of labeled protein from CM_Athan from CM_AB (Fig. 1D and data not shown).

We examined the products of Chordin cleavage at several time points after injection. For wild-type Chordin, almost all detectable protein was in the form of the small C-terminal fragment 8 hours after injection(Fig. 2A). On longer exposure,small amounts of FL protein and I+C fragment were seen, and the amounts decreased slightly from 8-14 hours. For CM_A a similar amount of FL protein was detected at the time points examined(Fig. 2A). For CM_B,significantly more total Myc tag was detected at all time points, as noted above (Fig. 2A). Although in the blot in Fig. 2A it is impossible to resolve the faint band representing FL protein in the CM_B samples, upon longer electrophoresis of additional samples, we verified that a similar amount of FL protein is present in embryos injected with CM_B, CM_A or wild-type chordin at 6 hours(Fig. 2C). Over the period examined, we did not observe large differences in cleavage kinetics for the different forms of Chordin, but because of the rapidity of cleavage, we cannot exclude subtle differences.

Our results demonstrate robust endogenous Tld activity in the zebrafish embryo, even prior to gastrulation. The zebrafish DV patterning gene mfn encodes a Tld-related enzyme(Connors et al., 1999), shown to cleave Chordin in vitro (Blader et al.,1997). When wild-type chordin RNA was injected into mfn;din double mutants, an increase in the amount of FL protein was seen (Fig. 2B). However, the majority (>75%) of detectable Myc tag was still on the small C-terminal fragment, showing that redundant enzyme activity compensates for loss of mfn. We also observed slight increases in FL protein for both of the partial CM Chordins (Fig. 2C),indicating that Mfn does not show a strong preference for cleavage either site.

Our rescue data suggest that the N+I fragment of Chordin is a stronger BMP inhibitor than the FL protein. To test this directly, we constructed a version of chordin encoding the N+I fragment containing the mutations of the upstream cleavage site in CM_A (N+I_A). We compared its efficacy in the rescue of din mutants to CM_A and CM_AB-Chordin. The N+I_A fragment was slightly more effective than CM_A (Fig. 3A), and both were more effective than CM_AB. These CM constructs are useful for separating the effects of cleavage and binding,although they admittedly give rise to stable fragments not present endogenously. However, the N+I fragment is normally produced in the embryo from wild-type Chordin, and is present at steady-state levels comparable with the FL protein (Fig. 3B),suggesting that it could significantly contribute to BMP inhibition in the embryo.

It is possible that an additional protein participates in the binding and preferentially increases the affinity of the N+I fragment for BMPs. Although Tsg displaces the N+I fragment from BMP in vitro(Larraín et al., 2001),it has been shown to enhance the binding of other Chordin fragments and the FL protein (Chang et al., 2001; Larraín et al., 2001; Oelgeschlager et al., 2000; Ross et al., 2001; Scott et al., 2001). To test Tsg as a candidate for this additional protein, we depleted embryos of Tsg function using a previously described antisense morpholino directed against tsg1 (tsg1-MO1) (Ross et al., 2001). We rescued din mutants with each chordin RNA in the absence or presence of tsg1-MO1; in every case, the depletion of Tsg resulted in a greater percentage of rescued or dorsalized embryos (Fig. 4A). These data argue that Tsg does not preferentially enhance the binding of the N+I fragment, and further support a general role for Tsg in decreasing the effectiveness of Chordin as a BMP inhibitor.

To determine the mechanism of this effect, we compared the Chordin cleavage products in the absence and presence of tsg1-MO1(Fig. 4B). These experiments were first performed in mfn;din double mutants, to enhance the accumulation of FL protein and more readily reveal alterations in the ratio of cleavage products. For all versions of Chordin, more total protein was observed in the presence of tsg1-MO1, indicating that endogenous Tsg decreases Chordin levels (Fig. 4C). The ratio of FL protein to cleavage products seen in the wild-type, CM_A and CM_B samples was also slightly greater(6-10%) in the presence of tsg1-MO1. This is consistent with previous reports that Tsg enhances Tld cleavage rates(Larraín et al., 2001; Scott et al., 2001; Shimmi and O'Connor, 2003),and further suggests that Tsg does not have a differential effect on cleavage at the two sites. We performed additional experiments in din mutants,to verify the effect of Tsg in the presence of Mfn, and also to examine N-terminal fragments. These experiments were performed with the addition of a control RNA encoding Myc-tagged GFP. When the amount of GFP was used as an internal standard to normalize Chordin, we observed similar increases in total Chordin in the presence of tsg1-MO1(Fig. 4D,E). Experiments performed with a control morpholino showed no increase in Chordin (data not shown). We also did not observe products of cleavage at any site between the A and B sites, even when cleavage was prevented at those sites. When wild-type chordin was co-injected with tsg1 RNA to increase Tsg activity, we still observed no additional fragments, nor did we see a large decrease in total Chordin (Fig. 4D,E). This suggests that our failure to observe alternative cleavage was not due to insufficient Tsg.

Our data are inconsistent with previous reports that depletion of Tsg promotes BMP signaling (Blitz et al.,2003; Chang et al.,2001; Scott et al.,2001) and ventralizes the zebrafish embryo(Ross et al., 2001). To test the possibility that Tsg has different roles that predominate at different protein levels, we re-examined the tsg1 morphant phenotype, injecting different amounts of tsg1-MO1 into wild-type embryos. At a lower MO level, we observed some features suggestive of ventralization in the morphants(Fig. 5B,C,N), including a reduced anterior nervous system and blood accumulation in the ventral tail. However, the morphants did not have multiplicated ventral fin folds, which is a prominent feature of ventralized din and ogon mutants(Hammerschmidt et al., 1996). Importantly, at higher MO levels we observed primarily dorsalized phenotypes in the morphants (Fig. 5I,N),consistent with our rescue data and the effect of Tsg depletion on Chordin levels.

To verify our characterization of the morphant phenotypes, we analyzed tsg1 morphants for markers indicative of DV patterning during gastrulation. We also performed injections with a second, non-overlapping MO targeting the tsg1 5′UTR (tsg1-MO5). At higher MO levels, we observed consistent expansion of dorsal markers and reduction of ventral markers, indicative of decreased BMP signaling, in embryos injected with either tsg1-MO (Fig. 5S-Z). Interestingly, at the lower MO level we did observe a consistent increase in the intensity of gata2 expression(Fig. 5D′), although it was not accompanied by an increase in expression of other ventral markers or a decrease in expression of dorsal markers(Fig. 5A′-C′). This suggests a specific role for tsg1 in limiting blood formation downstream of DV patterning, and may account for the increased blood we observe later in embryos injected with a low amount of MO.

As an additional control for the specificity of the morphant phenotypes, we co-injected tsg1-MO with tsg1 RNA (data not shown). At the lower MO level, 1 pg of tsg1 RNA rescued the phenotype of blood accumulation in the ventral tail, but did not correct the narrowing of the anterior nervous system, supporting our belief that it is non-specific. At the higher MO level, tsg1 RNA also corrected the features of dorsalization. We also performed injections with mismatch control MOs, and observed none of the phenotypes produced with either high or low amounts of the specific MOs.

Another possible explanation for the disparate roles ascribed to Tsg is that it acts on Chordin to decrease its efficacy as a BMP inhibitor, and simultaneously inhibits BMP signaling through an independent mechanism. To test this hypothesis, we injected tsg1-MO1 into din mutants. Surprisingly, this depletion of Tsg partially rescued or even dorsalized din mutants (Fig. 5E,F,L,M). These results provide strong evidence that Tsg also can act independently of Chordin, but to further enhance rather than inhibit BMP signaling.

Assessment of the importance of Chordin cleavage has been complicated in vertebrates by genetic redundancy. Therefore, we used a non-genetic approach,mutating conserved residues at the Tld sites in Chordin to render them resistant to cleavage. The altered Chordins were introduced into Chordin-deficient embryos and their effects on embryonic patterning analyzed. This approach allowed us to test directly the importance of Chordin cleavage,examine the cleavage products under various conditions, and determine the role of Tsg in regulating Chordin cleavage and efficacy in vivo. Although RNA injections are widely used in zebrafish embryos, and often to express mutated gene products, our study combines RNA rescue with structure/function correlations and the biochemical analysis of protein processing. Given the large number of zebrafish mutants with embryonic phenotypes that can be rescued by RNA injection, this general approach should have wide applicability in understanding signaling pathways in early development.

Previously we have shown that chordin RNA injected into embryos at the one-cell stage is detectable by in situ hybridization for ∼10 hours(Fisher and Halpern, 1999). Therefore, at 5 hours after injection there is still RNA available for new protein synthesis. However, at this time point the large majority of Chordin has been cleaved, demonstrating robust endogenous Tld activity even prior to gastrulation. Our data further show that cleavage at the A and B sites occurs independently and that the kinetics of cleavage are not significantly different for wild-type and CM Chordins.

The unexpectedly mild phenotype of the zebrafish mutant mfn has led to speculation that, at least in zebrafish, Chordin cleavage does not play an important role in DV patterning during gastrulation(Connors et al., 1999; Oelgeschlager et al., 2003; Zakin and De Robertis, 2004). We show instead that redundant enzyme activity compensates for loss of mfn. Both during and after gastrulation, the majority of Chordin protein is cleaved in mfn mutants, although measurably less than in wild type. In mouse, the Tld-related Bmp1 and Tll1 gene products function redundantly to cleave Chordin and other substrates(Pappano et al., 2003). The mfn gene is most closely homologous to Tll1(Scott et al., 1999); we have identified zebrafish ESTs corresponding to a second ortholog of Tll1,which is strongly expressed in the early embryo (J.X. and S.F., unpublished). The product of this gene is a likely candidate for at least some of the Tld activity present in mfn mutants.

Although mutating either cleavage site increased the efficacy of Chordin as a BMP inhibitor, the effect of alterations at the two sites was not equal. Prevention of cleavage at both sites resulted in a stable FL protein ∼10 times more effective as a BMP inhibitor. However, by preventing cleavage at the upstream site, we created a C-terminally truncated Chordin fragment that was even more effective than the stable FL protein. Another protein may participate in Chordin-BMP binding in vivo, selectively increasing the effectiveness of the truncated Chordin. Our data indicate that Tsg is unlikely to play this role, as it decreases the efficacy of all cleavage forms of Chordin. In fact, although there is evidence from a number of in vitro binding studies that Tsg enhances the binding of FL Chordin to BMP(Larraín et al., 2001; Oelgeschlager et al., 2000; Scott et al., 2001), our data indicate that this is not its predominant role in vivo. If it were, then the efficacy of CM_AB would decrease in embryos depleted of Tsg.

The N+I fragment may be more stable than the FL protein, as we observed for the I+C fragment. It is difficult to assess this because of the destabilizing effect of N-terminal Myc epitopes. However, at several time points there appeared to be comparable levels of labeled protein accumulated from the CM_A and CM_AB constructs (see Fig. 1D; data not shown),making this unlikely. We favor the possibility that CR4, which has little BMP binding affinity or biological activity on its own(Larraín et al., 2000; Scott et al., 2001), actually decreases the overall binding affinity of the FL protein.

There is evidence in Drosophila for an alternative cleavage product of Sog with enhanced Dpp inhibitory activity, whose creation is promoted by Tsg (Yu et al.,2000). Tsg also enhances Chordin cleavage at an intermediate site in vitro, suggesting that a parallel event occurs for the vertebrate proteins(Scott et al., 2001). However,we see no evidence of alternative cleavage products, either when cleavage is prevented at both of the normal Tld sites or when tsg1 is overexpressed by RNA injection. We cannot rule out that this cleavage event takes place in a small region of the embryo, or in specific tissues later in development. However, the necessary components (Chordin, Tld enzymes, and Tsg)are all present in the gastrula, and we should be able to detect fragments present even at less than 1% of the total label on our western blots. Therefore, we conclude that such cleavage is not likely to play a significant role in the gastrula.

Tld cleavage might also play a positive role if Chordin fragments have novel activities, independent of BMP binding. However, several lines of evidence argue against this. In particular, the epistatis between dino and the BMP mutants swirl and snailhouse has been examined (Wagner and Mullins,2002). That study confirmed that the BMP mutant phenotypes are epistatic to dino, and importantly discovered no additional phenotypes in the double mutants, as would be expected if Chordin or its fragments had functions independent of BMP. We also find that CM_AB-Chordin is capable of fully rescuing the dinphenotype, which would not be the case if the fragments had independent functions.

To test the effect of Tsg on Chordin efficacy in our system, and the dependence of the effect on cleavage by Tld at either site, we performed rescue experiments with wild-type and CM-chordin RNAs in the absence or presence of tsg1-MO. In every case, the rescue was more effective under conditions of lowered Tsg, suggesting that Tsg normally acts to suppress Chordin function in the zebrafish embryo. To determine the molecular basis of this effect, we examined Chordin cleavage products resulting from these experiments. The consistent effect of Tsg depletion was to increase steady-state Chordin levels. It has been previously shown that Tsg has the net effect of destabilizing Chordin(Oelgeschlager et al., 2003)and that Drosophila Tsg has a similar effect on Sog(Shimmi and O'Connor, 2003),consistent with our data. However, proposed mechanisms for this destabilization have invoked Tld cleavage. We see a similar effect on levels of wild-type and CM_AB-chordin in tsg1 morphants,suggesting that this effect is in part independent of Tld cleavage. We do observe a low level of cleavage or degradation of CM_AB-Chordin,although it appears not to represent cleavage at the normal site, it may still be mediated by Tld. To resolve this question definitively would require elimination of all Tld activity from the zebrafish gastrula, a challenge given the genetic redundancy.

Contradictory roles have been ascribed to Tsg in modulating vertebrate BMP signaling. However, Tsg has been reported in zebrafish embryos to cooperate with Chordin to inhibit BMPs (Ross et al.,2001). To reconcile our data with these published results, we performed additional Tsg depletions with different levels of tsg1-MOs. By injecting lower amounts, we did produce features suggestive of a ventralized phenotype. However, we did not see multiplicated ventral fin folds in any of the morphants, although this is a sensitive indicator of increased BMP signaling and is a consistent feature of the ventralized din and ogon mutants(Hammerschmidt et al., 1996). We did observe narrowing of the anterior nervous system, but morpholinos can induce non-specific toxic effects in zebrafish(Heasman, 2002), including widespread cell death and neural degeneration(Braat et al., 2001; Lele et al., 2001). In support of this possibility, increased apoptosis occurs in the brains of tsg1morphants (Little and Mullins,2004) and is not seen in ventralized din mutants(Fisher et al., 1997). Interestingly, we did observe increased expression of gata2 in embryos receiving a lower dose of MO. Increased gata2 expression was also previously reported in tsg1 morphants, and cited as evidence of their ventralization (Ross et al.,2001). However, the increase is apparently downstream of alterations in DV patterning, and may point to a specific role for tsg1 in limiting formation of blood or ventral ectoderm. Injection of higher levels of tsg1-MOs dorsalized the morphants, which we confirmed by examination of multiple markers during gastrulation. Our results show that endogenous Tsg enhances BMP signaling in vivo, in part by destabilizing Chordin.

Several previous studies of the effects of Tsg in the embryo relied on RNA overexpression. In our hands, tsg1 RNA dorsalizes both wild-type and din^-/- embryos, showing that the effect does not depend on Chordin (data not shown). This result is apparently at odds with our analysis of tsg1 morphants, and might suggest a normal role for Tsg in inhibiting BMPs. However, Tsg binds BMPs with an affinity comparable with that of individual Chordin CR domains (Chang et al., 2001; Oelgeschlager et al., 2000; Scott et al.,2001). We speculate that, when overexpressed, Tsg can bind BMPs sufficiently to prevent receptor activation, although this does not reflect its normal function. Interestingly, mutated versions of Tsg which do not bind BMPs hyperventralize the zebrafish embryo(Oelgeschlager et al., 2003),consistent with the possibility that the dorsalization seen with overexpression is due to direct BMP binding.

Many proteins in addition to Chordin contain repeated CR domains, and it has been proposed that Tsg could also interact with some of these(Garcia Abreu et al., 2002; Oelgeschlager et al., 2003). We tested the possibility that Tsg decreases the efficacy of Chordin while simultaneously inhibiting BMP signaling through another interaction. However,depletion of Tsg in din mutants ameliorated features of the mutant phenotype and even dorsalized the mutants. Thus, endogenous Tsg enhances BMP signaling both in the presence and absence of Chordin, either through direct interaction with BMPs or in conjunction with other unidentified modulating proteins. Although we cannot rule out that Tsg inhibits BMP signaling under some circumstances, we show that it does not do so in the zebrafish gastrula.

We thank M. Mullins and S. Little for the gift of morpholinos, and for sharing data prior to publication; we thank M. O'Connor for the gift of the tsg1 expression plasmid. This work was funded by the March of Dimes(S.F.).