Effects of heterodimerization and proteolytic processing on Derrière and Nodal activity: implications for mesoderm induction in Xenopus

Induction of the mesoderm takes place prior to gastrulation and represents the earliest known cell-cell interaction in the Xenopus laevis embryo. Several secreted growth factors from the FGF and TGFβ families have been identified as candidate mesoderm inducers based on their expression patterns and their ability to convert ectoderm to mesodermal fates (reviewed by Slack, 1994; Harland and Gerhart, 1997). These candidate molecules include bFGF, eFGF, Activin, Xnr1, Xnr2, Xnr4, Xnr5, Xnr6, Derrière and Vg1 (Kimelman and Kirschner, 1987; Slack et al., 1987; Smith et al., 1990; Thomsen et al., 1990; Thomsen and Melton, 1993; Jones et al., 1995; Joseph and Melton, 1997; Sun et al., 1999; Takahashi et al., 2000). Embryos expressing dominant negative Activin type I or type II receptors fail to form mesoderm, indicating that an Activin-like signal is essential for mesoderm formation in Xenopus (Hemmati-Brivanlou and Melton, 1992; Chang et al., 1997).

Recent work has shown that the cascade leading to mesoderm induction is initiated by the localized determinant VegT (reviewed by Kimelman and Griffin, 2000). VegT (also known as Brat and Xombi; the zygotic isoform is known as Antipodean) is a maternally supplied T-box transcription factor whose mRNA is stored in the vegetal cortex of the Xenopus egg (Lustig et al., 1996; Stennard et al., 1996; Zhang and King, 1996; Horb and Thomsen, 1997). It is essential for the mesoderm-inducing activity of endoderm, and the amount of mesoderm generated by Xenopus embryos depleted of maternal vegT transcripts is reduced by up to 90% (Zhang et al., 1998; Kofron et al., 1999). VegT promotes derrière, xnr4 and activin βB expression in a cell-autonomous manner (Clements et al., 1999; Kofron et al., 1999; Sun et al., 1999; Yasuo and Lemaire, 1999), showing that these mesoderm-inducing signals in Xenopus are a target of VegT. Although many mesoderm-inducing signal candidates have been identified, the importance of each and the degree to which they act redundantly has not been fully addressed.

Numerous studies in Xenopus have implicated Nodal-related factors (Xnrs) in mesoderm formation (reviewed by Schier and Shen, 2000). Six Xnrs have been identified to date and all but Xnr3 exhibit potent mesoderm-inducing activity (Jones et al., 1995; Smith et al., 1995; Joseph and Melton, 1997; Takahashi et al., 2000). In VegT-depleted embryos, Xnr1, 2 and 4 can all rescue the formation of head, trunk, and tail mesoderm (Kofron et al., 1999). All Xnr proteins characterized to date are initially expressed as zygotic transcripts following mid-blastula transition (MBT) in the endoderm and mesoderm. Transcripts accumulate in a dorsal to ventral gradient, with highest levels seen in dorsal cells (Jones et al., 1995; Agius et al., 2000; Takahashi et al., 2000). The strongest evidence that Nodal signaling is required in Xenopus mesoderm formation comes from studies exploiting the short form of the secreted extracellular antagonist Cerberus (Cer-S). While full length Cerberus is a multifunctional antagonist of Nodal, BMP and Wnt signaling (Bouwmeester et al., 1996; Hsu et al., 1998; Piccolo et al., 1999), the C-terminal Cer-S fragment lacks anti-BMP and anti-Wnt activities and has been suggested to be a Nodal-specific inhibitor (Piccolo et al., 1999). Cer-S blocks mesoderm induction by endodermal explants and inhibits mesoderm formation in whole embryos (Agius et al., 2000). It is difficult to fully reconcile the effects of Cer-S with those of a dominant negative Xnr2 cleavage mutant construct (CM-Xnr2), which has also been described as a specific antagonist of Nodal signaling. While CM-Xnr2 leads to anterior truncations in Xenopus, it fails to block mesoderm formation either in embryos or explants (Osada and Wright, 1999).

Findings in zebrafish and mice confirm that Nodal proteins constitute an important part of the mesoderm-inducing signal but indicate that other factors are probably involved, especially in posterior mesoderm formation. Mice homozygous for a Nodal mutation have severe defects in mesoderm formation. However, a significant fraction of Nodal^–/– embryos still express molecular markers of nascent and posterior mesoderm (Conlon et al., 1994). Zebrafish mutant for the Nodal-related genes cyclops and squint fail to express the dorsal mesoderm marker goosecoid and lack dorsal expression of the mesodermal marker ntl/brachyury. At later stages, the only mesodermal derivatives detected are several somites in the tail (Feldman et al., 1998). Loss of one-eyed pinhead (oep), which encodes an essential extracellular cofactor for Nodal signaling, leads to an almost identical phenotype (Gritsman et al., 1999; Yeo and Whitman, 2001). Similarly, mice mutant for cripto (a murine homologue of oep) have severe defects in mesoderm formation, with brachyury expression mislocalized proximally and a complete loss of later markers of axial, paraxial and lateral plate mesoderm (Ding et al., 1998).

These results suggest that while Nodal proteins are an essential component of the mesoderm-inducing signal in vertebrate embryos, other TGFβs probably play a significant role. One likely candidate is Derrière, a TGFβ family member closely related to Vg1. In Xenopus, derrière is first expressed following MBT in a pattern reminiscent of vegT. In animal cap explants, derrière transcripts are induced in response to VegT and derrière can induce both its own expression and that of vegT, suggesting an autoregulatory feedback loop. Although derrière has been implicated in posterior development and left-right asymmetry, it can efficiently induce both dorsal and ventral mesodermal fates in animal caps (Sun et al., 1999; Hanafusa et al., 2000). Finally, a dominant negative derrière cleavage mutant (cm-der) has been shown to block the ability of vegT to rescue endodermal gene expression and blastopore formation in vegT depleted Xenopus embryos (Xanthos et al., 2001).

To evaluate the contribution of TGFβ family members other than Nodal proteins to mesoderm induction, we have undertaken an analysis of derrière activity in the early Xenopus embryo. We have compared the early embryonic expression pattern of derrière with other likely components of the mesoderm-inducing signal and have characterized the activity of derrière in the gastrula stage embryo. We have also analyzed the effect of CM-Der on mesoderm induction and find that it completely inhibits the formation of both dorsal and ventral mesoderm. Interference with Derrière function also affects the transcriptional feedback loop previously documented for Nodal and Derrière expression (Takahashi et al., 2000; Onuma et al., 2002) Our analysis of the role of Derrière in mesoderm induction has led to several surprising observations. First, we find that not all TGFβ cleavage mutants act as expected, with unprocessed forms Xnr2 retaining long range mesoderm inducing activity. We also find that Derrière can directly interact with Nodal proteins and BMPs through the formation of heterodimeric ligands. We present our analysis of Derrière function and its interaction with other members of the TGFβ superfamily.

General methods are described in Sive et al. (Sive et al., 2000). For comparison of derrière expression with other mesoderm- and endoderm-specific transcripts, embryos were allowed to develop to the indicated stages and fixed for 1 hour. A subset of the fixed embryos were then bisected prior to in situ hybridization using a razor blade. Gastrula stage embryos were oriented such that the plane of section passed through the dorsal blastopore lip. For lineage tracing, 200 pg of nuclear β-galactosidase mRNA was co-injected and visualized following fixation (Sive et al., 2000) with the modification that 6-chloro-3-indolyl-β-D-galactoside (red-gal; Research Organics) was used in place of X-gal.

The general strategy for generating TGFβ cleavage mutants has been described previously (Lopez et al., 1992; Hawley et al., 1995). The original Xnr2 cleavage mutant (CM-Xnr2) was a gift from Christopher V. E. Wright (Osada and Wright, 1999). Additional Xnr2 cleavage mutant constructs were made using CM-xnr2 in pCS2(+) vector as a template. To generate the Xnr2 double cleavage mutant (DCM-Xnr2), the putative cryptic proteolytic cleavage site RGVR (AGA GGA GTA AGG) at residues 209-212 was altered to ALDA (GCT CTA GAC GCA) by PCR. An upstream cleavage mutant (UCM-Xnr2), in which the cryptic proteolytic cleavage site is mutated but the canonical site is not, was created by replacing the upstream region of wild-type Xnr2 with the equivalent region of DCM-Xnr2 in pCS2 by DraIII restriction digest.

To generate the Xnr2 mature domain with an altered proteolytic cleavage site fused to the Activin pro domain (proAct-CM-Xnr2), the coding region of the original CM-xnr2 construct was first inserted into pCS107 (Baker et al., 1999) and the xnr2 pro domain was excised using the 5′ BstBI restriction site and a SalI site introduced into the altered cleavage site. The pro domain of activin βB was amplified from pCSKA Activin βB plasmid (Thomsen et al., 1990) by high fidelity PCR and inserted into the BstBI/SalI pCS107 CM-Xnr2 vector. All PCR reactions used high fidelity Pfu polymerase (Stratagene) and the final constructs were checked for alterations both by sequencing and whole embryo activity assay.

HA- and Flag-tagged proAct-xnr2 and proAct-derrière constructs were made with the activin βB pro domain in place of the wild-type pro domain to increase proteolytic processing efficiency. The HA-tagged constructs used the pCS2(+) proAct-HA-Xnr1 cassette vector described by Piccolo et al. (Piccolo et al., 1999). The xnr1 mature domain was excised using a XhoI restriction site immediately downstream of the HA-tag and a 3′ XbaI site. The mature domains of xnr2 and derrière were amplified by high fidelity PCR and cloned into the proAct-HA cassette.

To generate the equivalent Flag-tagged constructs, a Flag epitope containing a 5′ SalI site and a 3′ XhoI site was introduced at an endogenous XhoI restriction site immediately downstream from the proteolytic cleavage sequence in activin βB in pSP64T (Thomsen et al., 1990). The resultant construct contains the sequence RIRKRGL(DYKDDDDK)LE (Activin proteolytic cleavage site underlined; Flag-tag in parentheses) and includes an XhoI restriction site following the Flag epitope. The mature domains of xnr2 and derrière were amplified by PCR as previously described, and cloned into the XhoI/XbaI proAct-Flag cassette to create the proAct-Flag-Xnr2 and proAct-Flag-Derrière constructs.

Wild-type Derrière and Xnr2 proteins, which possess endogenous pro and mature domains, were also tagged with Flag and HA epitopes. To generate these constructs, the derrière- and xnr2-coding regions were subcloned into pCS108 (A gift from Mustafa Khoka, UC Berkeley) and a SalI site was introduced by PCR mutagenesis immediately downstream of the proteolytic cleavage site. Sequences encoding either the Flag or HA epitope and containing a 5′ SalI site and a 3′ XhoI site were then inserted at the SalI restriction site. The resultant constructs are Flag-Derrière: RAKRV(DYKDDDDK)LD; HA-Derrière: RAKRVD(YPYDVPDYA)LD; HA-Xnr2: RRPRRVD(YPYDVPDYA)LD (proteolytic cleavage sites underlined; epitope tags in parentheses).

The cleavage mutant form of derrière (CM-der) (Sun et al., 1999) was cloned into pCS107 (Baker et al., 1999) and a single copy of the Glu-Glu epitope inserted at a SalI restriction site in the altered proteolytic cleavage sequence. The introduction of the Glu-Glu-tag changed the altered cleavage site from GVDG to GVD(GEYMPMEG)VDG (Glu-Glu-tag in parentheses). As with the Xnr2 cleavage mutant constructs, all epitope tagged constructs were checked by sequencing and activity assay for the presence of PCR induced mutations.

Capped synthetic mRNAs were transcribed using the mMessage mMachine kit (Ambion). The open reading frames of Xenopus derrière and CM-der (Sun et al., 1999) were amplified by high fidelity PCR and cloned into EcoRI/XhoI digested pCS107. The creation of other constructs is described above. Xenopus derrière, CM-der and Glu-Glu-tagged CM-der mRNAs were made from pCS107 (Baker et al., 1999) vectors linearized with AscI and transcribed by SP6 RNA polymerase. cerberus short-form (cer-S) (Piccolo et al., 1999), proact-HA-der, and proact-HA-xnr2 mRNAs were made from pCS2(+) linearized with NotI and transcribed by SP6. bmp4-flag (a gift from Jan L. Christian, Oregon Health Sciences University), CM-xnr2, DCM-xnr2 and proact-CM-xnr2 mRNAs were made from pCS2(+) linearized with Asp718 and transcribed with SP6. Proact-flag-der, proact-flag-xnr2 and proact-xnr2 were made from pSP64T (Krieg and Melton, 1984) linearized with EcoRI and transcribed with SP6. Nuclear β-galactosidase mRNA was made from pSPnucβGal (Smith and Harland, 1992) linearized with XhoI and transcribed by SP6. All mRNAs were injected as indicated.

For animal cap assays, one-cell embryos were injected with synthetic mRNAs as indicated at the animal pole. Animal cap ectoderm was dissected at stage 8 and cultured in 75% NAM solution (Peng, 1991). RNA was harvested at stages 11 (Condie and Harland, 1987) and 20 and RT-PCR was performed as described by Wilson and Melton (Wilson and Melton, 1994). Primer sets and amplification conditions have been described for ef1α, xbra, gsc, muscle actin and n-cam (Wilson and Melton, 1994); chordin (Sasai et al., 1994); odc (Hudson et al., 1997); cerberus and nrp-1 (Mariani and Harland, 1998); vegT (Zhang et al., 1998); xnr1 and xnr2 (Kofron et al., 1999); and xvex1 (Shapira et al., 1999). Primers used for derrière were U, 5′-TGG TTA CAT GGC AAA CTA CTG C-3′ (nucleotide positions 825-846); D, 5′-GAA TGA GAA CAA TCT CCA AAG C-3′ (nucleotide positions 1070-1091; note that the open reading frame encoding Derrière stops at nucleotide 1065, placing the downstream primer in the 3′ UTR and allowing for selective amplification of endogenous derrière mRNA). Other primer sequences were obtained from http://cbrmed.ucalgary.ca/pvize/html/WWW/Marker_pages/primers.html. For all RT-PCR reactions either ef1α or odc were used as loading controls.

Secreted [³⁵S]-labeled proteins were prepared from Xenopus oocytes essentially as described (Sive et al., 2000). For expression of single proteins, 40 ng of synthetic mRNA was injected per oocyte; for co-expression of two proteins, both mRNAs were mixed at equimolar concentrations and a total of 80 ng was injected. Conditioned medium was stored at 4°C. To examine intracellular proteins, oocytes were lysed in ice-cold RIPA buffer (20 μl/oocyte) (Harlow and Lane, 1998) and centrifuged twice for 15 minutes at 4°C at 13,000 g to remove the yolk. After each round of centrifugation, the clear middle layer was retained.

Supernatants and oocyte lysates were immunoprecipitated using the indicated antibodies; all immunoprecipitations used two to five oocyte equivalents (20-50 μl) of either supernatant or lysate. αGlu-Glu (clone Glu-Glu; Covance, Richmond, CA) and αFlag (clone M2; Sigma) immunoprecipitations were carried out in a mixture of 60% RIPA and 40% phosphate-buffered saline (PBS) with the antibody at 2 ng/μl. Anti-HA immunoprecipitations were carried out in 60% RIPA and 40% αHA mouse monoclonal antibody (clone 12CA5) hybridoma tissue culture supernatant. Binding was carried out on ice for 1 hour in a total volume of 500 μl. Antibodies were then bound to protein A beads for 30 minutes at 4°C with end-over-end rotation, centrifuged and washed three times in RIPA buffer. For two-step immunoprecipitations, the αHA monoclonal antibody was first coupled to protein A beads (Harlow and Lane, 1988). Bound proteins were released by boiling for 10 minutes in Laemmli sample buffer. For all immunoprecipitations, immunoprecipitating proteins were resolved by 12% reducing SDS-PAGE.

Although Nodal proteins have been implicated as a major component of the mesoderm-inducing signal in Xenopus, zebrafish and mouse, there is reason to believe that other members of the TGFβ family also play a crucial role. In Xenopus embryos, derrière is a likely candidate for contributing to mesoderm formation based on its activity in naïve ectodermal explants and its expression pattern (Sun et al., 1999). To better understand the role derrière may play in mesoderm induction, we undertook a thorough analysis of its spatial and temporal expression pattern during blastula and gastrula stages.

The timing and distribution of derrière transcripts were compared with well-characterized tissue-specific markers and other members of the TGFβ family implicated in mesoderm formation. The temporal expression of derrière was analyzed by RT-PCR on RNA from whole embryos. As previously reported (Sun et al., 1999), derrière is present only as a zygotic transcript and cannot be detected prior to MBT at stage 8 (Fig. 1A, lanes 1 and 2). Transcripts are first detected at stage 8.5 and rapidly accumulate (Fig. 1A, lanes 3 and 4). Levels plateau by stage 9.5 and remain high throughout gastrulation, but taper off by stage 13 (Fig. 1A, lanes 5-8). The onset of derrière expression coincides with the onset of other presumptive targets of the maternally supplied transcription factor VegT. These include the homeobox genes mix1 (Fig. 1A) (Clements et al., 1999) and bix4 (Fig. 1G) (Tada et al., 1998; Casey et al., 1999) as well as xnr1 and xnr4 (Fig. 1A). Expression of the mesodermal marker xbra is not detected until stage 9.5 (Fig. 1A, lane 4), as expected for a gene whose activation depends on one or more of the zygotically supplied TGFβs. Expression of cerberus, in the dorsal-most region of the mesendoderm (Bouwmeester et al., 1996; Zorn et al., 1999), is also first detected at stage 9.5 and increases to maximum levels during early gastrula stages (Fig. 1A, lane 5) as dorsoventral patterning of the mesoderm commences.

In order to compare the spatial expression pattern of derrière with other TGFβ family members and determine its position relative to the boundaries of the three germ layers, we analyzed embryos by whole-mount in situ hybridization. Embryos were bisected prior to in situ hybridization to reveal expression patterns in deep cells. In agreement with RT-PCR analysis, derrière transcripts are detected at stage 8.5 (Fig. 1B). The entire marginal zone expresses derrière, both superficially in cells of the suprablastoporal endoderm and throughout the underlying mesoderm all the way to the blastocoel. In addition, expression is detected in the vegetal cells of the sub-blastoporal endoderm. Although derrière appears to exhibit a slight dorsal bias, expression is clearly detected throughout the marginal zone. At this stage, derrière expression is highly reminiscent of that of bix4, a direct target of VegT (Fig. 1G). The HMG-domain transcription factor sox17β is expressed by stage 8.5, but is limited to the vegetal endoderm and superficial cells of the marginal zone (Fig. 1E) [see Henry and Melton (Henry and Melton, 1998), and Zorn et al. (Zorn et al., 1999) for additional analysis of endoderm and mesoderm markers in sectioned embryos]. Expression of mesodermal markers such as xbra and gsc is not yet seen at this stage (Fig. 1D,F).

At gastrulation stages 10+ and 11, derrière is strongly expressed throughout much of the marginal zone and in at least some cells of the sub-blastoporal endoderm (Fig. 1H,N). Its expression pattern continues to closely mirror that of bix4 (Fig. 1M,S). Localized transcription of xnr2 is apparent by stage 10+ and shows a strong dorsal bias (Fig. 1I). Relative to derrière, which is expressed at high levels in both the superficial and underlying cells of the mesoderm and endoderm, xnr2 is clearly restricted to the surface of the embryo and appears to be most closely associated with the blastopore lip (Jones et al., 1995). Expression of xnr1, xnr5 and xnr6 is not as restricted to the embryonic surface as is xnr2, but all show the same strong dorsal bias, which probably reflects positive regulation by the dorsal determinant β-catenin (Agius et al., 2000; Takahashi et al., 2000). derrière is expressed more broadly than the mesodermal marker xbra (Fig. 1J,P), particularly at the onset of gastrulation. At stage 10+, the high levels of derrière transcript seen in the marginal zone extend vegetally all the way to the dorsal blastopore lip (Fig. 1H; arrowhead indicates blastopore lip). By contrast, xbra expression terminates several cell diameters above the blastopore lip (Kumano and Smith, 2000). In summary, derrière expression corresponds to what is expected for an endogenous mesoderm-inducing signal: it is initiated prior to the onset of mesoderm formation and is present at high levels in both the dorsal and ventral marginal zone, and in the underlying cells of the vegetal endoderm.

derrière has been previously shown to induce a number of mesodermal markers in animal cap assays and to induce a partial secondary axis when injected ventrally (Sun et al., 1999). To understand the capacity of derrière to induce mesoderm in the context of the whole embryo further, we overexpressed derrière mRNA in either the dorsal or ventral marginal zone. Although even modest doses of xnrs and activin strongly induce dorsal mesoderm and convert ventral fates to dorsal, we found that derriere was much less potent as a dorsal inducer. Both dorsal and ventral injections induce general mesoderm markers such as xbra (Fig. 2B,C). However, dorsal targeting leads to an expansion of dorsal markers such as gsc (Fig. 2E) while ventral targeting leads to an expansion of ventral and lateral markers such as xvent1 (Fig. 2I). The ability of derrière to induce ventrolateral fates in addition to dorsal ones is consistent with the loss of anterior structures reported by Sun et al. (Sun et al., 1999). Head formation requires the suppression of trunk fates, so that an expansion of trunk mesoderm is expected to inhibit head formation. Unlike xnr2, which will strongly induce organizer fates on the ventral side at only 10 pg (P. M. E. and R. M. H., unpublished), derrière does not induce ectopic gsc expression even at 200 pg (Fig. 2F). derrière’s reported ability to induce partial secondary axes when overexpressed ventrally (Sun et al., 1999) could result from the induction of ventrolateral rather than organizer fates.

To determine if derrière is an essential component of the endogenous mesoderm inducing signal in Xenopus, we used a dominant-negative approach to analyze loss of function. As TGFβs must be proteolytically cleaved after dimerization to form an active ligand, a non-cleavable version can dimerize with endogenous wild-type monomers and prevent proper processing (Fig. 3K) (Lopez et al., 1992). A derrière cleavage mutant (CM-der) has previously been described by Sun et al. (Sun et al., 1999). We injected 2 ng of CM-der mRNA into the marginal zone of a single blastomere at the four-cell stage. This dose of CM-der is enough to inhibit mesoderm induction in animal caps by up to 200 pg of derrière (Fig. 3L, lane 3). Embryos were allowed to develop to mid-gastrula stages and fixed for whole-mount in situ hybridization analysis. The site of mRNA injection was identified by detection of a β-galactosidase lineage tracer. In embryos where CM-der had been targeted to the marginal zone, expression of the general mesodermal marker xbra was completely lost at the site of injection (Fig. 3B). In addition, the blastopore lip either failed to form in the presence of CM-der or was severely disorganized (not shown). The complete inhibition of mesoderm formation caused by CM-der is highly reminiscent of the activity reported for cer-S (Fig. 3D) (Agius et al., 2000) and is consistent with the loss of trunk mesoderm reported by Sun et al. (Sun et al., 1999).

As derrière and nodal genes are linked to mesoderm induction both by their expression patterns and activity, we reasoned that their transcription might be inhibited when mesoderm is blocked by CM-der and cer-S. This was borne out in both animal cap experiments and whole embryos expressing cer-S. Animal caps co-injected with derrière and cer-S failed to express xbra, derriere and xnr1 (Fig. 3L, lane 4). In situ hybridization analysis of embryos injected marginally with cer-S mRNA also revealed a substantial decrease in the expression of both derrière and xnr2 at stage 11 (Fig. 3G,J). In addition to being significantly reduced in level, the boundaries of the strong band of derrière expression normally seen at the marginal zone were much less distinct than in control embryos and more punctate expression was detected in the vegetal endoderm (compare expression in Fig. 3E,G). Injection of CM-der prevents analysis of endogenous derrière, but CM-der appeared to cause a modest reduction in xnr2 transcription (Fig. 3I). These data demonstrate that Derrière and Nodal proteins positively regulate each other at the transcriptional level during mesoderm induction; therefore even reagents that block only one of these molecules through direct interaction (such as Cer-S) still have inhibitory effects on both.

When a cleavage mutant form of xnr2 (CM-xnr2) (Osada et al., 1999) was compared with CM-der and cer-S, striking differences in activity were apparent. Expression of xbra was greatly expanded and could be seen in most of the animal hemisphere (Fig. 3C). CM-xnr2 was also found to upregulate derrière expression (Fig. 3F). These effects on cells far removed from the site of expression are of particular interest because TGFβ cleavage mutants are conventionally thought to function cell autonomously by blocking secretion as well as proteolytic processing.

The induction of mesoderm-specific transcripts by CM-xnr2 indicates that it may retain at least some of the signaling activity of wild-type xnr2. One possible explanation for this is suggested by the presence of a cryptic proteolytic cleavage in the pro region of Xnr2, upstream from the canonical cleavage site. TGFβs are cleaved by members of the proprotein convertase family of endoproteases that recognize the consensus cleavage site RXXR (Bresnahan et al., 1990; Molloy et al., 1992; Dubois et al., 1995; Cui et al., 1998). Recent work has shown that the TGFβ family member Lefty A is processed at two distinct cleavage sites, each of which retains activity when the other is mutated (Ulloa et al., 2001). In addition, BMP4 is cleaved in an sequential manner at two sites, and mutation of the upstream site has important consequences for both the strength and range of the signal (Cui et al., 2001). In Xnr2, cleavage at both sites would result in two pro domain fragments, which we have designated proA and proB (Fig. 4A). Because only the canonical site was altered in CM-xnr2, we reasoned that processing might still be occurring at a cryptic site and generating active ligand consisting of proB and the mature domain. To test this possibility, we altered the second cleavage site at residues 209-212 from RGVR to ALDA, creating a double cleavage mutant xnr2 (DCM-xnr2; Fig. 4A). CM-xnr2 and DCM-xnr2 were tested for mesoderm inducing activity in animal cap assays. Surprisingly, we found that even DCM-xnr2 was still capable of inducing xbra. The overall activity of both cleavage mutants was substantially attenuated, as indicated by the fact that even at 2 ng they induced less xbra than 10 pg of wild-type xnr2 (Fig. 4B, lanes 1-3). Reduced signaling capacity was confirmed by the failure of CM-xnr2 and DCM-xnr2 to induce extreme dorsal mesoderm. Xnr2 is capable of inducing a range of mesodermal fates in a concentration-dependent manner, from ventral and lateral derivatives at very low doses to exclusively dorsal fates at higher doses (Jones et al., 1995). While 10 pg of wild-type xnr2 was sufficient to activate expression of the dorsal marker gsc in animal caps, even 2 ng of the cleavage mutants failed to do so. Taken together, these data indicate that unprocessed Xnr2 can transduce a Nodal-like signal and cleavage mutant forms of the protein should not be considered authentic dominant negatives. In contrast to Xnr2, we found that CM-Der is completely inactive in mesoderm induction assays (Fig. 4B, lane 4).

In order to assess the importance of the Xnr2 cryptic cleavage site for normal Nodal signaling, we constructed an upstream cleavage mutant (UCM-Xnr2) in which the cryptic site is mutated while the canonical site is left unaltered. The activity of each was evaluated over a range of doses in both animal caps and whole embryos. In animal cap assays, both wild-type Xnr2 and UCM-Xnr2 were found to induce mesodermal markers such as xbra and vegT at similar doses (Fig. 5A). When overexpressed in whole embryos both molecules gave rise to identical phenotypes at equivalent doses (Fig. 5B-E). We conclude that proteolytic processing at the canonical cleavage recognition sequence can occur in the absence of upstream processing and is sufficient for normal (or nearly normal) activity.

Because CM-Xnr2 had previously been reported to lack all wild-type activity, including the ability to induce mesoderm (Osada and Wright, 1999), it was important to confirm our observations in additional assays. We therefore made a second attempt to rule out low-level proteolytic processing from any unidentified cleavage sites in the pro region of Xnr2. To do this, we completely replaced the Xnr2 pro region with that of Activin βB, while leaving the cleavage site mutated (proAct-CM-Xnr2) (Fig. 4A, Activin pro region in blue). Previous studies have shown that when the Activin cleavage mutant is translated in Xenopus oocytes, only the unprocessed 51 kDa protein is produced (Hawley et al., 1995). This indicates that Activin contains only a single site for proteolytic processing. Consequently, proAct-CM-Xnr2 should be completely inactive if the activity of CM-Xnr2 and DCM-Xnr2 is dependent upon processing from unidentified sites in the pro domain.

We compared the activity of all three Xnr2 cleavage mutants in whole embryos using in situ hybridization to evaluate mesoderm formation. CM-xnr2 and DCM-xnr2 both induced xbra expression throughout much of the animal hemisphere (Fig. 4I,L). Proact-CM-xnr2 was equally effective at inducing mesoderm in this assay (Fig. 4O). All three cleavage mutant mRNAs, when injected in the animal hemispheres of one-cell embryos, caused expansion of gsc on the dorsal side of the embryos (Fig. 4J,M,P) and xwnt8 (a marker of ventral and lateral mesoderm) in the remainder of the embryos (Fig. 4K,N,Q). Wild-type xnr2 was more efficient than the cleavage mutants at inducing gsc (Fig. 4G), confirming that activity is reduced when proteolytic processing is inhibited. Once again, CM-der showed no mesoderm-inducing activity and actually inhibited xbra expression in some embryos, despite being targeted to the animal pole rather than the marginal zone (Fig. 4R).

It is possible that Xnr2 cleavage mutants could induce mesoderm either by acting as functional Nodal ligands or by heterodimerizing with and inhibiting endogenous Nodal antagonists such as Antivin and Lefty (Meno et al., 1999; Thisse and Thisse, 1999; Cheng et al., 2000). In the former case, Xnr2 cleavage mutants would act as secreted mesoderm-inducing signals, while in the later case their activity would be cell autonomous (because even if the unprocessed ligand were to be secreted, heterodimerization with Nodal antagonists could only occur in the expressing cells). To distinguish between these possibilities, we compared the range of DCM-Xnr2 activity with wild-type Xnr2 and the cell-autonomous mesoderm inducer Smad2 (an intracellular component of the Nodal signal transduction pathway). Embryos were injected at the four-cell stage to limit diffusion of mRNAs to one quadrant of the animal hemisphere. To facilitate lineage tracing, a lacZ/smad2 fusion construct was employed (Baker and Harland, 1996) and nuclear β-galactosidase was co-injected with xnr2 and DCM-xnr2. Mesoderm induction was assayed at stage 11 by in situ hybridization. In smad2-injected embryos, ectopic xbra expression was strictly limited to those cells of the animal hemisphere where Smad2 was also expressed (Fig. 6C). By contrast, both xnr2 and DCM-xnr2 caused ectopic xbra expression throughout the animal hemisphere even in cells far removed from those injected with mRNA (Fig. 6A,B). While it is possible that antagonism of Antivin and Lefty through heterodimerization accounts for some of the activity seen in the Xnr2 cleavage mutants, these data show that the unprocessed ligand also retains long-range signaling activity.

The ability of CM-Xnr2 and DCM-Xnr2 to act as secreted signaling molecules is unexpected, given that cleavage mutant forms of TGFβs are thought to block secretion of the endogenous, wild-type monomers (Lopez et al., 1992). To confirm that Xnr2 cleavage mutants are secreted, we examined expression of wild-type Xnr2, CM-Xnr2, DCM-Xnr2 and CM-Der proteins in Xenopus oocytes. Levels of protein were compared between conditioned oocyte supernatants (containing secreted protein) and lysates (containing intracellular protein). Xnr2, CM-Xnr2 and DCM-Xnr2 were found almost exclusively in the oocyte supernatant, indicating that they are efficiently secreted (Fig. 6E, lanes 1-3). By contrast, CM-Der was not secreted by oocytes and could only be detected in the lysate (Fig. 6E, lane 4). Proteolytic processing was observed in wild-type Xnr2 and CM-Xnr2, indicating that both canonical and cryptic cleavage sites are recognized by proteases. When wild-type Xnr2 was expressed, bands corresponding to both proA and mature domains were observed (Fig. 6E, lane 1; mature ligand indicated by arrowhead). Although a band corresponding to the predicted proB/mature ligand was not observed in CM-Xnr2, the presence of the proA domain indicated proteolytic processing had taken place. As expected, DCM-Xnr2 was present only as an unprocessed precursor protein running at approximately 70 kDa. These results demonstrate that different TGFβ family members have very different activities following proteolytic processing. However, for our purposes, the finding that CM-Der is both inactive and an inhibitor of mesoderm formation enables us to further test its role in normal mesoderm induction.

We wanted to evaluate thoroughly the specificity of CM-Der for Derrière signaling. Previous reports have indicated that although CM-Der is a preferential antagonist of Derrière signaling, it may have some effect on other TGFβ family members. In an assay for animal cap elongation (characteristic of dorsal mesoderm fates), CM-der was found to reduce both the frequency and extent of xnr2-induced elongation (Sun et al., 1999). We confirmed the ability of CM-Der to attenuate Nodal signaling by showing that it can ameliorate the xnr2 overexpression phenotype in whole embryos (Fig. 7A). Normally, the severity of this phenotype increases with the level of xnr2 mRNA (for the purposes of this experiment doses ranged from 1 to 10 pg). When xnr2 is co-injected with 3 ng of CM-der, the whole embryo phenotype is markedly reduced such that 10-fold more xnr2 mRNA is required to produce the same effect (compare 1 pg xnr2 alone with 10 pg xnr2 co-injected with 3 ng CM-der in Fig. 7A).

We used animal cap assays to understand the inhibitory activity of CM-Der on Nodal signaling. Embryos were injected with increasing amounts of xnr2 mRNA either alone or in combination with 2 ng of CM-der. When animal cap explants were analyzed at stage 11, xnr2 mRNA was found to induce xbra expression at both 1 and 10 pg. Consistent with the ability of Nodal ability to induce both dorsal and ventral mesoderm in a dose-dependent manner, gsc and chordin expression (both markers of dorsal mesoderm) were detected at 10 pg, but not at lower doses (Fig. 7B, lanes 1-3). Levels of xnr2 that induced mesoderm formation also activated transcription of other TGFβs including derrière and xnr1. In the presence of CM-der, a ten-fold higher dose of xnr2 mRNA was required for mesoderm induction, consistent with data from whole embryos (Fig. 7B, lanes 4-6). While 10 ng of xnr2 mRNA was sufficient to induce extreme dorsal types of mesoderm on its own, when used in combination with CM-der only the general mesoderm marker xbra was seen. In addition, activation of other TGFβ transcripts was severely reduced, suggesting that the ability of CM-Der to inhibit Nodal signaling is at least partially due to antagonism of Nodal-induced derrière.

The observation that CM-Der inhibited a broad range of Xnr2 induced transcripts suggested to us that it may antagonize Nodal signaling directly through the formation of heterodimers. This possibility is suggested by the observation that BMPs readily form heterodimeric ligands within their subfamily (Aono et al., 1995; Hazama et al., 1995) and the recent discovery that Nodal proteins and BMPs might mutually antagonize one another through the formation of heterodimers (Yeo and Whitman, 2001). In addition, a cleavage mutant form of xnr5 has been shown to block the mesoderm inducing activity of multiple TGFβ family members, including derrière, in animal caps (Onuma et al., 2002). We used an in vivo translation assay in Xenopus oocytes to study direct interactions between CM-Der and other members of the TGFβ family. Mature Nodal ligands have proven difficult to detect in either lysates or conditioned media from expressing cells, probably because of inefficient proteolytic processing, instability of the mature ligand or a combination of both factors (Constam and Robertson, 1999). We encountered similar problems with Derrière expression in Xenopus oocytes (not shown), but overcame it by employing HA-epitope tagged Activin/Nodal and Activin/Derrière chimeric proteins. As shown in Fig. 8 (supernatant αHA lanes 1 and 2), when the pro domains of Xnr2 and Derrière were replaced with that of Activin, detectable levels of the mature ligands were immunoprecipitated from oocyte supernatants. Co-injection of an equal amount of CM-der mRNA completely blocked secretion of Xnr2 and Derrière ligands from oocytes (Fig. 8, supernatant αHA lanes 5 and 6), demonstrating that the Derrière cleavage mutant can inhibit secretion of multiple proteins in the TGFβ family. We also found that secretion of Flag-tagged BMP4 was completely inhibited by CM-Der (Fig. 8, supernatant αFlag lanes 3 and 7). To confirm that all mRNAs were being translated in the oocytes, we carried out immunoprecipitations from lysates. All oocytes injected with Glu-Glu-tagged CM-der mRNA expressed an unprocessed protein of ∼50 kDa (Fig. 8, lysate αGlu lanes 4-7). Unprocessed HA-tagged proAct-Derrière and proAct-Xnr2 were also immunoprecipitated from oocyte lysates, even when CM-Der was co-expressed (Fig. 8, lysate αHA lanes 1-2 and 5-6), indicating that inhibition occurs at the level of protein processing and secretion rather than mRNA translation. Very low levels of mature Derrière and Xnr2 ligands were detected only in lysates from oocytes not expressing CM-Der (mature ligands indicated by arrowheads). Similarly, unprocessed BMP4 was present in all oocyte lysates expressing Flag-tagged BMP4, but the mature ligand was lost upon co-expression of CM-Der (Fig. 8, lysate αFlag lanes 3 and 7; unprocessed protein bands are indicated with an asterisk).

To confirm that CM-Der is capable of inhibiting BMP activity, we analyzed its activity in animal cap explants. Untreated ectodermal explants develop in culture into undifferentiated epidermis; however, when BMP signaling is blocked by treatment with BMP-specific antagonists (Lamb et al., 1993; Sasai et al., 1995; Piccolo et al., 1996; Zimmerman et al., 1996), by use of dominant negative BMP ligands (Hawley et al., 1995) or through dissociation of the animal cap (Wilson and Hemmati-Brivanlou, 1995), neural fates are induced (for a review, see Harland, 2000). We reasoned that if CM-Der antagonizes members of the BMP subfamily, it should also be a direct inducer of neural tissue in animal caps. Stage 20 animal caps injected with 2 ng of CM-der mRNA were found to express the general neural markers n-cam and nrp-1, even in the absence of the mesodermal marker muscle actin (Fig. 9, lane 2). This induction occurred at the expense of the epidermal marker epidermal keratin (Fig. 9, lanes 2 and 3). The activity of CM-der was indistinguishable from that of the secreted BMP antagonist noggin (Fig. 9, lane 1).

The most likely mechanism by which CM-Der could inhibit processing and secretion of Xnr2 and BMP4 is through heterodimerization. Multiple strategies were used to determine whether mature Derrière/Nodal and Derrière/BMP heterodimers formed in the Xenopus oocyte expression system. In the first, HA-tagged proAct-Derrière was co-expressed in oocytes with either Flag-tagged proAct-Xnr2 or untagged proAct-Xnr2. HA-tagged Derrière was then immunoprecipitated from the oocyte culture media using αHA antibody and resolved by SDS-PAGE. When expressed by itself, the mature Derrière ligand was present as a smear rather than a single band, migrating between 22 and 28 kDa (Fig. 10A, lane 3). This is most likely to be due to post-translational modifications such as glycosylation. When Flag-tagged proAct-Xnr2 was co-expressed with HA-tagged proAct-Derrière, αHA immunoprecipitation pulled down two distinct protein bands (Fig. 10A, lane 4). The first band in this doublet migrated from 19 to 23 kDa, while the second was a tighter band centered at 25 kDa. When Flag-tagged proAct-Xnr2 was replaced with untagged proAct-Xnr2, the mobility of the upper band shifted by slightly over 1 kDa (the expected size of the Flag epitope), while the lower band remained unchanged. Based on these data, we conclude that the upper band in the doublet is Xnr2 co-precipitating with the HA-tagged Derrière ligand. The equivalent intensities of the lower Derrière band and the upper Xnr2 band indicate that Derrière forms heterodimers with Nodal at least as efficiently as it forms homodimers with itself (Fig. 10A). The differences in mobility observed between Derrière alone and in combination with Xnr2 raise the possibility that heterodimerization affects glycosylation or other post-translational modifications.

The second method used to detect a direct association between the mature Derrière ligand and other members of the TGFβ family was a two-step immunoprecipitation relying on the disulfide-mediated covalent association between subunits. As before, HA-tagged proAct-Derrière was co-expressed in oocytes with various Flag-tagged TGFβs. HA-Derrière was than immunoprecipitated from supernatants using αHA antibody conjugated to protein A sepharose beads and bound proteins were released by boiling. Co-precipitating Flag-tagged proteins were then pulled down in a second immunoprecipitation, this time using αFlag antibody. Single-step immunoprecipitations using αHA or αFlag antibody alone confirmed that all mRNAs were efficiently translated by the oocytes (Fig. 10B, αFlag and αHA panels). Flag-tagged Derrière, Xnr2 and BMP4 were all detected after the two-step immunoprecipitation, demonstrating that Derrière is capable of heterodimerizing with BMPs and Nodal proteins in addition to forming homodimers (Fig. 10B, αHA:αFlag panel, lanes 5-7).

Finally, we used epitope tagged versions of wild-type Derrière and Xnr2 (rather than Activin/Derrière and Activin/Xnr2 fusion constructs) to examine heterodimerization between full-length, uncleaved proteins. This was done to eliminate the possibility that the activin pro domain was responsible for mediating heterodimerization not normally observed with wild-type ligands. This is not an idle concern, as Gray and Mason (Gray and Mason, 1990) have demonstrated that the pro domains of TGFβ ligands support dimerization. As noted previously, we were unable to detect significant levels of the mature Xnr2 and Derrière ligands in supernatants from oocytes expressing wild-type proteins. However, the full-length, uncleaved proteins could be readily extracted and immunoprecipitated from oocyte lysates. Unprocessed (prepro) Derrière is present in the oocyte lysate as a doublet, while prepro Xnr2 is present as a single band; these bands migrate at distinct molecular weights and can be resolved by SDS-PAGE (Fig. 10C, lanes 1-3). When Flag-tagged Derrière and HA-tagged Xnr2 were co-expressed in oocytes and immunoprecipitated using αHA, the prepro Derrière doublet was clearly seen to co-precipitate with prepro Xnr2 (Fig. 10C, lane 5, upper αHA panel). When the reciprocal immunoprecipitation was carried out using αFlag, prepro Xnr2 was pulled down with prepro Derrière (Fig. 10C, lane 5, αFlag panel). In addition, mature Derrière and Xnr2 ligands were detected in oocyte lysates after αHA immunoprecipitation (Fig. 10C, lanes 1-3, middle αHA panel). When HA-tagged Xnr2 and Flag-tagged Derrière were co-expressed and pulled down using αHA, a band migrating at the expected size of the mature Derrière ligand was seen to co-precipitate (Fig. 10C, lane 5, arrowhead). These data, coupled with our observation that CM-Derrière (possessing the wild-type Derrière pro domain) efficiently inhibits the processing and secretion of wild-type BMP4, proAct-Derrière and proAct-Xnr2 (Fig. 9), provide convincing evidence that the naturally occurring forms of diverse TGFβ family members efficiently heterodimerize in co-expressing cells.

Members of the Nodal family have recently emerged as essential components of mesoderm induction in species ranging from frog to mouse. Establishing the full importance of Nodal signaling has been complicated by the fact that there are at least five different Nodal-like proteins expressed in early Xenopus embryogenesis capable of inducing mesoderm. All are expressed in the vegetal endoderm and the marginal zone beginning shortly after MBT, with considerably higher levels detected on the dorsal side of the embryo (Jones et al., 1995; Joseph and Melton, 1997; Takahashi et al., 2000). The ability of the Nodal-specific antagonist Cer-S to completely block mesoderm formation in Xenopus has led to the suggestion that a dorsal to ventral gradient of Nodal signaling may be responsible for mediating mesoderm induction and patterning in response to VegT and β-catenin (Agius et al., 2000). While not contradicting this basic model, we find that other members of the TGFβ superfamily, particularly Derrière, are also likely to play a significant role in this process.

While Derrière has previously been implicated in anteroposterior patterning of the mesoderm during gastrulation (Sun et al., 1999), we suggest that its expression in late blastula stages is indicative of a role in general mesoderm induction in response to VegT activity. More than any of the Nodal family members, derrière transcripts mirror the expression pattern of zygotic vegT and known targets of VegT signaling such as bix4 (Fig. 1). Like the mesodermal marker xbra, derrière is expressed in both the superficial and deep cells of the marginal zone at stage 10+ and transcript levels are more or less equally distributed between dorsal and ventral sides of the embryo. derrière expression is also detected in cells of the sub-blastoporal endoderm, which are able to mediate mesoderm induction. Finally, we find that nodal activity sufficient to induce mesoderm also invariably activates derrière transcription and the Nodal-specific antagonist cer-S blocks derrière expression in the whole embryo (Fig. 3).

Taken together, these data strongly suggest that mesoderm induction is a cooperative process in Xenopus that involves VegT and multiple members of the TGFβ superfamily. Derrière and members of the Nodal family are each capable of inducing one another in animal cap assays (Fig. 3L, Fig. 7B) (Takahashi et al., 2000; Onuma et al., 2002), creating a positive transcriptional feedback loop that is probably mediated in part through the maternal transcription factor FAST-1 (Osada et al., 2000). In addition, each of these ligands can induce Smad2 phosphorylation (an essential step in the transduction of Activin-like mesoderm-inducing signals) and expression of zygotic vegT (Lee et al., 2001). The probable importance of this feedback loop for mesoderm formation is demonstrated by the observation that the lowest level of Xnr2 capable of generating mesoderm is also the lowest level that induces transcription of derrière and other nodal family members (Fig. 7B, lanes 2 and 6). The same is found to be true for mesoderm induction by derrière (not shown). Thus, these TGFβ family members act cooperatively to induce and maintain their mutual expression.

It is also interesting to note that Derrière, unlike Nodal proteins and Activin, does not appear to strongly favor the formation of dorsal over ventral mesoderm even at relatively high doses (Fig. 2). This difference may be due in part to the late induction of Smad2 phosphorylation by Derrière in comparison to Xnrs (Lee et al., 2001). It has been hypothesized that differences in the timing of Smad2 activation may explain how different ligands are able to produce different activities while using the same intracellular pathway (Lee et al., 2001). In Xenopus embryogenesis, Smad2 phosphorylation is initiated dorsally at stage 9 but is essentially equivalent in dorsal and ventral halves of the embryo by the onset of gastrulation (Faure et al., 2000; Lee et al., 2001). This suggests that early dorsal Smad2 phosphorylation arises from Xnr activity, while later Smad2 phosphorylation may rely on Derrière, which is expressed more uniformly along the dorsoventral axis than Xnrs. In this model, early activation of Smad2 leads to organizer fates, while later activation is compatible ventral and lateral fates in addition to dorsal ones.

Ultimately, our observation that induction of dorsal and ventral markers by Derrière is dependent on the context of the embryo emphasizes the importance of the dorsoventral prepattern imposed by β-catenin accumulation (Heasman et al., 1994; Schneider et al., 1996; Rowning et al., 1997). The presence of this dorsal modifier overrides the late activation of Smad2 by Derrière to induce organizer-specific transcripts, while its absence causes ventrolateral gene expression. Thus, while timing of Smad2 activation may contribute to dorsoventral patterning (Lee et al., 2001), early Smad signaling is not necessary for organizer formation in the context of β-catenin activity. Mechanistically, the two pathways are known to converge and synergize at cis-regulatory sequences to induce organizer-specific genes (Watabe et al., 1995). Even elevated levels of Nodal expression on the dorsal side, which may impart specific dorsoventral patterning information, can be attributed to transcriptional activation by β-catenin (Agius et al., 2000) and repression by ventral inducers such as BMP4 (Osada et al., 2000).

Interaction between Nodal proteins and other members of the TGFβ family can occur at the level of ligand heterodimerization as well as transcriptional activation. Here, we have shown in immunoprecipitation assays that the mature Derrière ligand physically associates with Nodal proteins and BMPs (Fig. 10A,B). In addition, unprocessed Derrière and Nodal heterodimers are found in the lysate of co-expressing oocytes (Fig. 10C). Consistent with this finding, CM-Der is capable of blocking expression of processed Xnr2 and BMP4 protein and directly neuralizing ectodermal explants (Figs 8, 9). This extends the findings of Yeo and Whitman (Yeo and Whitman, 2001), who have shown that heterodimers can form between cleavage mutant Nodal and BMPs. While our experiments cannot definitively address whether heterodimerization occurs at physiological levels in the embryo, the efficient interaction between Flag-tagged Xnr2 and HA-tagged Derrière in co-expressing oocytes suggest that heterodimers would form. This conclusion is reinforced by the fact that secretion of Xnr2 from oocytes can be efficiently blocked by co-injection of an equivalent amount of CM-der mRNA (Fig. 8).

It is interesting to speculate on how and to what degree the activity of heterodimers may differ from that of homodimers. There is compelling evidence from studies of the BMP subfamily that heterodimerization has the potential to modulate activity of TGFβ ligands. BMP2/7 and BMP4/7 heterodimers appear to form preferentially over homodimers and display increased activity in assays for osteogenic differentiation (Aono et al., 1995; Hazama et al., 1995) and mesoderm induction (Suzuki et al., 1997; Nishimatsu and Thomsen, 1998). A complete evaluation of the activity of Derrière/Nodal and Derrière/BMP heterodimers is beyond the scope of the present study; however, the altered electrophoretic mobility of processed Xnr2 and Derrière (Fig. 10A,B) when expressed together strongly suggests differences in post-translational modification. What importance, if any, these differences have for endogenous signaling remains to be determined.

A wide variety of dominant-negative strategies have been developed for carrying out loss-of-function studies in Xenopus because genetic techniques for removing zygotic gene function are not readily available (Lagna and Hemmati-Brivanlou, 1998). In the TGFβ superfamily, proteolytic cleavage mutants have proven especially useful for this task (Lopez et al., 1992; Wittbrodt and Rosa, 1994; Hawley et al., 1995; Joseph and Melton, 1998). All TGFβs are synthesized as an inactive precursor protein in which the pro and mature domains are separated by a proteolytic cleavage site with an RXXR motif. Alteration of this motif inhibits proteolytic processing and heterodimerization between wild-type and mutant monomers in co-expressing cells blocks the release of active dimers. A high degree of specificity has been claimed for these dominant-negative reagents. In this study we report several findings that raise concerns about both the specificity of cleavage mutants and the degree to which inhibition of proteolytic processing in the Nodal family blocks signaling.

Our analysis of the discrepancies between Nodal loss-of-function studies using CM-Xnr2 (Osada and Wright, 1999) and those using Cer-S (Agius et al., 2000) revealed that cleavage mutant forms of Nodal retain some signaling capacity. These data suggest that formation of mesoderm in Xenopus embryos injected with CM-Xnr2 can be explained by incomplete inhibition of Nodal signaling. Such an explanation fits with the phenotypes reported for CM-Xnr2-expressing embryos. Defects include a delay in the formation of the blastopore lip during gastrula stages and a variety of anterior defects, such as microcephaly, anencephaly and shortened body axes, later in development. All of these may be secondary consequences of disorganized mesoderm formation or partial inhibition of involution at gastrulation because of attenuated Nodal activity. The observation that CM-Xnr2 blocked elongation in response to wild-type Nodal proteins is consistent with our findings that the activity of CM-Xnr2 is severely attenuated and, as a result, it fails to induce dorsal mesodermal fates. As dorsal mesoderm undergoes the most extreme convergent extension, and CM-Xnr2 can heterodimerize with wild-type Nodal proteins and block their ability to induce dorsal fates (data not shown), the failure of animal caps to elongate probably reflects a conversion to ventral and lateral mesoderm fates. In addition, cleavage mutant forms of both Xnr1 and 4 were found to possess mesoderm-inducing activities comparable with unaltered ligands, suggesting this may be a general property of the Nodal family (Osada and Wright, 1999).

While the presence of putative proteolytic cleavage sites N-terminal to the canonical cleavage site in Xnr1, 2 and 4 may partially explain the residual activity of cleavage mutants, it cannot be the sole factor at work. A mutant that lacks both RXXR motifs had activity comparable with the single mutant (Fig. 4B). Even completely replacing the Xnr2 pro-domain with that of Activin (which has been shown to be cleaved at a single site in oocyte expression studies) (Hawley et al., 1995) failed to produce an authentic dominant-negative Nodal construct. These data indicate that unprocessed Nodal protein is capable of generating an attenuated Nodal-like signal. A similar activity has been reported for the TGFβ family member Lefty A, which can activate the MAPK pathway as an unprocessed precursor (Ulloa et al., 2001). However, the ability for unprocessed TGFβ proteins to transduce a wild-type signal appears to be the exception rather than the rule. In the same battery of assays (including whole embryos, animal caps, and in vivo translation and secretion),we confirmed that the Derrière cleavage mutant was devoid of all activity. Other studies using cleavage mutant forms of Activin, TGFβ, Vg1, BMPs and Xnr5 and Xnr6 all reach the same conclusion (Lopez et al., 1992; Wittbrodt and Rosa, 1994; Hawley et al., 1995; Joseph and Melton, 1998; Onuma et al., 2002). Consequently, the use of cleavage mutants for conducting loss-of-function studies can be valid, provided the effects of heterodimerization are taken into account.

A second, and potentially more serious, concern with the use of cleavage mutant proteins as dominant negative reagents relates to their specificity. The TGFβ superfamily contains many members and several distinct subfamilies. For a dominant negative reagent to be useful, its inhibitory activity should be as restricted as possible, preferably limited to the specific molecule in question or a few closely related family members. The promiscuous heterodimerization that we have observed between Derrière, Nodal and BMP ligands suggests that precise specificity may be impossible to achieve with cleavage mutants. These concerns are verified by our finding that CM-Der can induce neural fates in animal caps (Fig. 9), an activity that almost certainly arises from promiscuous heterodimerization with endogenous BMPs. At the very least, dominant-negative TGFβ constructs must be carefully evaluated for their effects on other family members, ideally using a combination of in vivo expression and embryonic activity assays.

In conclusion, we report that both the mesoderm-inducing effect of derrière and the promiscuous blocking activity of CM-derrière are consistent with the anteroposterior patterning activity of these reagents (Sun et al., 1999). derrière induces ventral or paraxial mesoderm, with its well-established posteriorizing activity (for a review, see Munoz-Sanjuan and Brivanlou, 2001), and therefore tends to suppress anterior structures. It is therefore an indirect posteriorizing factor as suggested by Sun et al. (Sun et al., 1999). By contrast, CM-derrière blocks the formation of mesoderm, and by blocking BMP activity also induces neural structures, a combination that results in anterior fates (reviewed by Harland, 2000). In more general terms, our findings suggest that interactions between different members of the TGFβ superfamily can be highly complex, involving reciprocal activation at the transcriptional level and direct physical association through the formation of heterodimers. While these multiple levels of crosstalk make it impossible to separate completely the requirement for individual TGFβs in the process of mesoderm formation in Xenopus, the preponderance of evidence points towards a model of cooperative induction in which both Nodal proteins and Derrière are key participants.

We thank Anne Fisher for αHA hybridoma tissue culture supernatant, Jan Christian for providing the Flag-tagged BMP4 prior to publication, and Eddy De Robertis, Hazel Sive and Chris Wright for other constructs used in this work. We thank Chris Wright, Sharon Amacher, Timothy Grammer and Mustafa Khoka for discussion and comments on the manuscript, and all members of the Harland laboratory for their input and advice. This work was supported by the NIH and a graduate student fellowship from the NSF.