XBMPRII, a novel Xenopus type II receptor mediating BMP signaling in embryonic tissues

Determination of mesodermal cell fates in the marginal zone of the early Xenopus laevis embryo is thought to be specified by their location within a dorsoventral gradient of instructive signals (for review see Slack, 1994). Early models postulated that the signals establishing this instructive gradient emanated from the dorsally located Spemann’s Organizer. However, recent evidence suggests that at least some dorsally derived signals are diffusible long-range inhibitors of ventralizing signals, and that it is their degree of attenuation of a dominant ventralizing influence that instructs marginal zone fate (e.g. Piccolo et al., 1996; Zimmerman et al., 1996).

Many lines of evidence implicate BMP4 as an instructive ventral signal. First, BMP4 is a ventral mesoderm inducer (Dale et al., 1992; Jones et al., 1992; Koster et al., 1991) that can suppress the dorsal mesoderm-inducing activities of activin and noggin (Dale et al., 1992; Jones et al., 1992, 1996; Re’em-Kalma et al., 1995). Second, BMP4 expression patterns suggest a close linkage with ventral specification. During gastrulation, for example, transcripts become restricted from the presumptive neural plate and dorsal marginal zone, while being maintained in ventral regions of the embryo (Fainsod et al., 1994; Hemmati-Brivanlou and Thomsen 1995; Schmidt et al., 1995). Third, the secreted organizer molecules noggin and chordin can functionally inhibit BMP4 through direct binding (Piccolo et al., 1996; Zimmerman et al., 1996). These latter findings suggest a mechanism for generating a BMP4 activity gradient in the marginal zone despite the initially uniform distribution of BMP4 RNA. In this model, the local concentration of active BMP4 in the marginal zone is regulated by the ratio of BMP4 to noggin/chordin and this serves to instruct the different dorsoventral mesodermal cell fates. Future modifications to this model may arise from the finding that BMP4/7 heterodimers are more active than either homodimer (Suzuki et al., 1997), although the existence of endogenously occurring heterodimers is under investigation. Nevertheless, strong genetic evidence for an early requirement for BMP4 signaling in vertebrate embryogenesis comes from the observation that most homozygous null BMP4 mouse embryos arrest development during gastrulation and lack embryonic mesoderm (Winnier et al., 1995).

BMP4 signals are also thought to be important in maintaining an epidermal cell fate within uninduced ectoderm (for review see Hemmati-Brivanlou and Melton, 1997). Prolonged dissociation of Xenopus ectodermal cells causes their neuralization, but an epidermal fate is maintained when sufficient BMP4 protein is added to the medium bathing the dispersed cells (Wilson and Hemmati-Brivanlou, 1995). Furthermore, disrupting BMP signaling within intact animal caps by several methods, i.e. inhibiting receptor activity (Wilson and Hemmati-Brivanlou, 1995; Xu et al., 1995), adding BMP4 inhibitory proteins (Piccolo et al., 1996), or overexpressing dominant-negative BMP4 isoforms (Hawley et al., 1995), results in direct neural induction. The clearing of BMP4 RNA from the presumptive neural plate (Hemmati-Brivanlou and Thomsen, 1995; Schmidt et al., 1995) is consistent with these findings.

As TGFβ-related molecules, BMP signals are presumably transduced through receptor complexes consisting of type I and type II transmembrane serine/threonine kinases (for review see ten Dijke et al., 1996, Hogan, 1996). In general, type II receptors primarily bind the ligand, while type I receptors, after cross-phosphorylation by ligand-bound type II receptors, initiate nuclear signal transduction by activating downstream targets like the Smad proteins (Massague, 1996). BMP receptors differ somewhat from this paradigm in that both type I and type II receptors can apparently bind ligand independently, although high affinity binding and signal transduction require both receptor subtypes (Liu et al., 1995; Nohno et al., 1995; Rosenzweig et al., 1995). Additional complexity may be superimposed on this type I/type II heteromer model based on the findings of fairly promiscuous associations amongst several different type I and type II receptors, and that some receptors bind multiple ligands (Liu et al., 1995; Rosenzweig et al., 1995).

Although past studies on TGFβ/BMP signaling pathways during Xenopus embryogenesis have focused mainly on the ligands, a complete understanding of how these molecules exert their effects in vivo will require a similar analysis of their receptors, including details of their expression patterns during relevant embryonic processes, and their ligand-binding and signaling specificity. For many TGFβ-related molecules likely to be key developmental regulators (e.g. Vg1 and nodal-related factors), specific receptors have not been identified. Moreover, for the identified receptors, only limited information exists on their spatiotemporal expression patterns, particularly during those stages of development when basic decisions regarding cell fate and axial specification are being made.

Recently, a novel type II receptor that binds BMP2, BMP4 and BMP7, but not activin or TGFβ, was identified from mammalian cells (Kawabata et al., 1995; Liu et al., 1995; Nohno et al., 1995; Rosenzweig et al., 1995). This receptor, known as BMPRII, T-ALK, or BRK3, has until now only been characterized in vitro. Given the particular suitability of the frog embryo system for functional analyses in vivo, we have cloned and characterized the Xenopus homolog, XBMPRII. We report its expression pattern during embryogenesis, including the first published description of any BMP receptor during stages when dorsal and ventral signals are being actively interpreted to determine mesodermal cell fate. We describe experiments indicating that XBMPRII mediates the effects of BMP4 on mesodermal patterning and suppression of neural fate. Our data suggest substantial differences between the in vivo effects of a dominant-negative isoform of XBMPRII, tBRII, and a dominant-negative type I BMP receptor, tBR (Graff et al., 1994; Suzuki et al., 1994). Moreover, we show that XBMPRII differs significantly from the activin type II receptor XAR1, since XAR1 can mediate both activin and BMP signaling in vivo, whereas XBMPRII appears to be specific for BMPs.

A BamHI-EcoRI cDNA fragment encoding C-terminal sequences of mouse BMPRII (gift from M. Kawabata and H. Moses) was used to screen ∼2 million pfu of a dorsal lip cDNA library (Blumberg et al., 1991) at a stringency giving single-copy bands on Xenopus genomic Southern blots. Clone ADL12 encodes amino acids 392 to 1020 at the C terminus of XBMPRII. An ADL12-based screen of a stage 28-30 λZAPII cDNA head library (Hemmati-Brivanlou et al., 1991) gave an additional ∼400 bp of 5′ sequence. PCR amplification of this region and rescreening the head library led to two independent cDNAs. Clone C6 was sequenced on both strands using the Sequenase II kit (USB) and shown to contain a 3.4 kb open reading frame representing the entire protein coding sequence, with approximately 150 bp and 350 bp of 3′ and 5′ untranslated sequence, respectively.

A dominant negative human BMPRII (hDNTALK) was generated by PCR amplifying the human BMPRII clone CL4-1-1 (Kawabata et al., 1995), truncating the receptor 10 amino acids after the transmembrane domain. This fragment was MscI/BstEII-digested and exchanged for the β-globin insert in pSP64TXβm (Krieg and Melton, 1984). A similarly truncated fragment (tBRII) was made from XBMPRII clone C6. The amplified fragment was EcoRI/XbaI-digested and ligated into pCS2+ (Turner and Weintraub, 1994). While screening for XBMPRII, initial experiments, including generation of secondary axes, rescue of UV ventralization, and neural induction in animal caps, were done with hDNTALK. After cloning XBMPRII, experiments were repeated several times with tBRII with similar results; subsequent experiments used only the Xenopus construct. Figs 3 and 5A represent results generated with hDNTALK constructs. All experiments were repeated at least twice with similar results, except for those shown in Fig. 7, which gave the same result twice.

Fig. 1B shows that XBMPRII is expressed both maternally and zygotically. Transcript levels decreases during early gastrulation (stage 10.25), but then increase from the gastrula through late tadpole stages. The spatial expression pattern of XBMPRII (Fig. 2) was determined by whole-mount in situ hybridization. At stage 9, XBMPRII signal is stronger over the animal hemisphere (Fig. 2A), but since detection of transcripts by whole-mount in situ hybridization is often difficult in vegetal cells, we used RNAse protection (Fig. 2C) to detect XBMPRII RNA in dissected animal and vegetal explants. This analysis confirmed a higher level of XBMPRII RNA in animal relative to vegetal cells.

In situ hybridization was performed as described previously (Harland, 1991) with improvements communicated by the author. Antisense XBMPRII probe was made from EcoRI-linearized ADL12 plasmid and T7 RNA polymerase; sense probes were made from XhoI-digested ADL12 and T3 RNA polymerase. Probes for HoxB9, BMP-4, nrp-1, Xotx2, Krox20, XEn2, Xbra and goosecoid were prepared as described previously; specific details may be obtained upon request.

Xenopus eggs were fertilized in vitro, injected and manipulated as described (Kay and Peng, 1991), and staged according to Nieuwkoop and Faber (1967). Animal caps were explanted at stage 8-9 in 1× RSB and cultured on agarose in 0.75× NAM until collection, when tissues were either frozen (RNA analysis) or fixed (histology). Vegetal explants were isolated at stage 8 as described (Gamer and Wright, 1995). Dorsal and ventral marginal zones (DMZs and VMZs) were isolated at stage 10.25-10.5, with the blastopore lip marking the dorsal side. DMZs corresponded to an ∼60° arc of marginal zone tissue centered on the dorsal lip midline, VMZs were cut analogously from the region opposite the DMZ and both were cultured in 0.75× NAM until collection.

Capped mRNA was made with the mMessage mMachine kit (Ambion). hDNTALK and tBRII plasmids were linearized with BamHI or XhoI, respectively, and transcribed with SP6 RNA polymerase. During injection, embryos were immersed in 5% Ficoll dissolved in 0.1× RSB. For animal cap experiments, 10 nl was injected at the 1-to 2-cell stage. For injections at the 2-to 8-cell stage, 5 nl/blastomere was injected. Dorsally or ventrally targeted injections at the 4-to 8-cell stage used pigmentation differences to determine the appropriate site of injection.

Embryos were fixed in 70% ethanol overnight, dehydrated in an ethanol series, washed in xylene, soaked overnight in xylene:paraplast (Oxford Labware; 1:1 ratio) and paraplast embedded. Sections (5-10 µm) were stained with hematoxylin/eosin (Sigma/Surgipath).

Embryos were fixed in Dent’s solution (Kay and Peng, 1991) overnight in a Petri dish, or in MEMFA (Harland, 1991) for 2 hours on a nutator. Pigmented embryos were bleached in 10% H₂O₂/70% methanol under fluorescent light. Immunohistochemistry was performed as described (Hemmati-Brivanlou et al., 1991). The 12/101 and MZ15 antibodies were used at dilutions of 1:500 and 1:750, respectively, and the secondary goat anti-mouse/HRP-linked antibody (Jackson Immunoresearch) was used at a dilution of 1:1000. In some cases, color reactions were enhanced by adding 0.04% NiCl₂.

A 190 bp HindIII fragment from clone C6 was subcloned into pBluescript, XhoI-linearized and antisense RNA made with T7 RNA polymerase. Digestion after hybridization was with RNAses A and T₁. Probes for EF1α, NCAM and Xbra were synthesized as described previously (information available upon request). RNA was isolated by SDS/proteinase K digestion and LiCl precipitation.

RNA was isolated from embryonic tissues and RT-PCR performed as described (Chang et al., 1997). Specific primers for FGFR, goosecoid, NCAM, Xbra, Xwnt-8, muscle actin, Krox20, HoxB9 and Xotx2 were synthesized according to sequences reported (references available upon request). Primer pairs were retested to determine conditions under which transcript detection was in the linear range of amplification.

Overlapping cDNAs encoding XBMPRII were isolated by screening Xenopus libraries with a human BMPRII probe, of which clone C6 contained a 3.4 kb open reading frame with short 5′ and 3′ untranslated regions. Based on the similarity in size and structure of the protein conceptually translated from C6 (Fig. 1A), the in-frame stop codons at the 5′ and 3′ ends of its open-reading frame, and the biological activities described below, we conclude that C6 contains the full protein-coding sequence of XBMPRII (XenopusBMP Receptor type II; GenBank accession number U81958). Northern blot analysis with a C6 cDNA probe on poly(A)⁺ RNA from embryonic stages 9, 15 and 22 detected transcripts of ∼11 kb (data not shown), similar to the size of human BMPRII mRNA (Kawabata et al., 1995).

By pregastrula stages, XBMPRII expression is no longer widespread but becomes localized to the torus of prospective mesodermal cells at the equatorial region of the embryo. Some stage 10⁻ embryos display a uniformly intense ring of staining, but in slightly more advanced embryos (stage 10⁺), staining is substantially reduced in the dorsalmost region of the marginal zone, the site of the future Organizer (Fig. 2D). At early gastrula, XBMPRII signal is detected in preinvoluting mesodermal cells, being excluded from the superficial layer. Expression does not extend to the involuting vegetalmost margin of the dorsal lip (Fig. 2E), in contrast to the pan-mesodermal marker, Xbra, whose expression in the mesodermal precursors at this stage extends to the dorsal lip (data not shown; Smith et al., 1991).

At neurula stages, XBMPRII expression continues in prospective mesoderm (Fig. 2G,H) but not the superficial ectoderm (Fig. 2L). Weak staining along the dorsal midline is visible at this stage (data not shown), although the loss of signal following sectioning prevented attributing this to notochordal or neural tube expression. By early tailbud, XBMPRII is primarily expressed in two ventrolateral mesodermal regions: one surrounding the future proctodeum (Fig. 2J,K) and extending dorsally towards the tail bud (Fig. 2J), and an anterior focus around the stomodeum/heart anlage (Fig. 2J). Weaker expression is apparent along the entire neural tube (Fig. 2K), with higher levels in the developing brain and around the eye (Fig. 2J). At the late tadpole stages (Fig. 2N,O), XBMPRII expression expands to include the neural tube, branchial arches, tail bud, eye, olfactory placode, otic vesicles, head mesenchyme and brain, but is absent from the notochord and somitic derivatives (data not shown). Fig. 2F,I,M,P shows that XBMPRII expression is reminiscent of that of BMP4 (Fainsod et al., 1994; Hemmati-Brivanlou and Thomsen, 1995; Schmidt et al., 1995), a finding consistent with the hypothesis that XBMPRII mediates BMP4 signaling during mesodermal and neural development in Xenopus embryogenesis. However, while BMP4 is expressed in the ventral blood islands (Fig. 2P) and migrating neural crest (Fainsod et al., 1994), XBMPRII was not detected in these areas (Fig. 2N).

Dominant-negative isoforms of TGFβ, BMP type I and activin receptors, which prevent the reciprocal receptor phosphorylation that initiates signal transduction, are generated by deleting the intracellular kinase domain (Chen et al., 1993; Graff et al., 1994; Hemmati-Brivanlou and Melton, 1992;). Previously, it had been reported that ventral expression of a human dominant-negative BMPRII resulted in secondary axes that lacked anterior structures (Ishikawa et al., 1995). Thus, to explore the function of the frog cognate in vivo, we injected Xenopus embryos with RNA encoding a similarly truncated XBMPRII isoform (Fig. 1A), referred to as tBRII, and analyzed the resulting effects in more detail. Injecting tBRII RNA into the two dorsal blastomeres at the 4-cell stage resulted in anterodorsalized embryos with enlarged heads and diminished or absent posterior structures (data not shown), while ventral injections induced partial secondary axes that could be distinguished from the primary axis by the absence of a visible floorplate (Fig. 3B). The secondary axis resulted from a bifurcation of the neural tube, but contained no notochord or longitudinally arrayed somite blocks, although they did contain muscle tissue (Fig. 4D). While the secondary axes did not have proper heads, anterior structures were often seen, including otic vesicles, mucus-secreting cement glands (Fig. 4A) and the occasional dense focus of pigmented tissue probably representing melanized retinal epithelium (data not shown). We infer that tBRII overexpression locally blocks the function of BMP4 (and/or related ligands), with ventral delivery causing secondary axis induction and dorsal injection resulting in hyperdorsalization. We tested whether secondary axis induction was a specific effect of tBRII activity by coinjecting mRNA encoding the normal (wild-type) XBMPRII receptor and tBRII. As shown in Table 1, increasing the dose of wild-type receptor progressively reversed, and eventually completely abrogated, the effects of tBRII.

Members of the Smad family of intracellular effectors appear relatively specific for distinct signal transduction pathways (for review, see Massague, 1996). Smad1 overexpression in Xenopus embryonic cells mimics ventralizing signals such as those provided by BMPs, while Smad2 mimics dorsal induction such as that caused by activin. It has recently been found that Smad1 or Smad2 activity is greatly increased on coexpression with their common heteromeric partner, Smad4 (Candia et al., 1997). Therefore, to test whether tBRII-induced secondary axes were caused by interference with BMP signaling, we coinjected tBRII RNA together with Smad1/Smad4 RNAs on the ventral side of the embryo. With 1.25 ng of each RNA per embryo, tBRII alone induced secondary axes in 88% (n=26) of the embryos, while 89% (n=36) of those coinjected with Smad1/Smad4 contained only one axis (data not shown). Together with the failure of Smad2/Smad4 to cause this reversal when coinjected with tBRII, this strongly suggests that tBRII specifically interferes with BMP-like ventralizing signals.

The tBRII-induced secondary axes are highly disorganized, and contain abundant neural crest-derived melanocytes and multiple tubules resembling floorplate-less neural tubes (Fig. 3E,F). The presence of extensive neural tissue throughout the secondary axes was confirmed molecularly by whole-mount analysis (Fig. 4B) for the pan-neural marker nrp-1 (Richter et al., 1990). We characterized the neural tissues present in the induced axes more precisely using markers specific for different anteroposterior (A/P) domains of the neural tube. These were: HoxB9, a spinal cord marker (Wright et al., 1990), En2, a marker of the midbrain/hindbrain junction (Hemmati-Brivanlou et al., 1991), Krox20, a marker of hindbrain rhombomeres 3 and 5 (Bradley et al., 1992), and Otx2, a marker of anterior neural structures including the forebrain and midbrain (Blitz and Cho, 1995). Each of these markers was expressed in approximately the correct position relative to each other along the secondary axis (Fig. 4E,F,G,H).

In contrast to these results for tBRII, it has been reported that a truncated BMP type I receptor, tBR, has no effect when injected dorsally into the 4-cell embryo and that secondary axes induced by ventral injections lack anterior structures (Graff et al., 1994; Suzuki et al., 1994). We therefore compared directly the in vivo effects of these similarly truncated receptors. In agreement with the previous reports, our tBR-induced secondary axes lacked anterior structures (cement glands or otic vesicles) and notochord, although muscle tissue was present (data not shown). However, while tBRII-induced axes expressed a broad range of A/P neural markers including the anterior markers Otx2/Krox20/En2, tBR-induced axes expressed only the posterior neural marker, HoxB9 (n=9 per marker), even at the highest RNA doses tested (1 ng/embryo; data not shown). Therefore, although these two components are presumed to function in the same signaling complex in vivo, we conclude that there is a substantial difference in the A/P extent of the neural tissue induced in the secondary axes by similarly truncated type I and type II BMP receptors.

Non-neural tissues were also detected in the ectopic axes induced by either tBRII or tBR. The muscle-specific 12/101 antibody (Kintner and Brockes, 1984) showed the presence of disorganized muscle tissue located in the posterior regions of all secondary axes tested (Fig. 4D; n=7 for each RNA). These axes lacked notochordal tissue (Fig. 4C; n=15), as defined by the absence of Xbra expression, which marks notochord and tail organizer (Smith et al., 1991), and the lack of immunostaining with the MZ15 antibody (data not shown), which is specific for an extracellular matrix protein on the notochord surface (Smith and Watt, 1985).

We conclude that inhibiting BMP signaling via tBRII in ventral regions of the embryo induces dorsolateral mesoderm (muscle), but not more dorsal mesoderm (notochord). The presence of extensively patterned neural tissue in the secondary axis, in the absence of dorsal mesoderm, suggests either its direct induction, as seen in other examples of disrupted BMP signaling (Hawley et al., 1995, Wilson and Hemmati-Brivanlou, 1995, Xu et al., 1995, Piccolo et al., 1996), or its induction by the dorsolateral mesoderm of the secondary axis.

To test whether tBRII blocks BMP signaling in animal cap ectoderm and causes its direct neuralization, we explanted caps from embryos injected at the 1-cell stage with 1 ng of tBRII RNA and analyzed them at late gastrula (stage 10.5-11) for expression of the pan-mesodermal marker Xbra, or tailbud (stage 25) for the pan-neural marker NCAM (Kintner and Melton, 1987). NCAM was induced in tBRII-loaded animal caps without coincident mesoderm induction (Fig. 5A). Caps loaded with RNA encoding a dominant-negative activin type II receptor (ΔXAR1; Hemmati-Brivanlou and Melton, 1992) and activin-treated uninjected caps served as positive controls for neural and mesodermal induction, respectively (Fig. 5A; lanes 4 and 5). The NCAM expressed in activin-treated caps is caused by secondary induction by the activin-induced dorsal mesoderm.

The A/P character of the neural tissue induced in the animal caps by tBRII was tested using the region-specific markers described above. We detected expression of Otx2, but not the more posterior markers Krox20 (Fig. 5B) or HoxB9 (data not shown). Thus, we conclude that tBRII blocks BMP signaling in animal cap ectoderm, switching it to an anterior neural fate, consistent with previous reports using different agents to block BMP signaling (Hawley et al., 1995; Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995; Xu et al., 1995).

We also found that coinjection of Smad1 RNA reversed the tBRII-induced neuralization of the explant, while Smad2 induced dorsal mesodermal markers in tBRII-loaded explants (data not shown). This is consistent with the reversal of tBRII-induced secondary axis induction by Smad1, in showing that tBRII specifically interferes with BMP-related signaling pathways.

The morphologically incomplete secondary axes suggested that ventral expression of tBRII cannot induce a fully functional ectopic Spemann organizer. In agreement with this hypothesis, gastrulation-stage embryos that were ventrally injected with tBRII RNA developed neither a secondary dorsal lip, nor ectopic expression of the organizer marker goosecoid (Blumberg et al., 1991; Cho et al., 1991) as assayed by whole-mount in situ analysis (data not shown). We tested whether partial dorsalization of the ventral mesoderm by tBRII might contribute to the non-neural tissues found in the secondary axis. Gastrula-stage dorsal and ventral marginal zones (DMZ and VMZ) were explanted from embryos that were ventrally injected at the 4-cell stage with 3 ng of tBRII RNA (1.5 ng/blastomere) and analyzed at stage 12.5 (early) or stage 25 (late) for expression of goosecoid, the late dorsolateral mesodermal marker actin (Mohun et al., 1984), or for Xwnt8, a ventrolateral mesodermal marker with much lower expression in the DMZ (Christian et al., 1991; Lemaire and Gurdon, 1994). We compared these to explants from embryos injected similarly with tBR RNA, which was previously shown to dorsalize VMZs (Graff et al., 1994; Maeno et al., 1994). As shown in Fig. 6, gsc and actin were induced in VMZs by either tBR or tBRII, coincident with decreased Xwnt8 expression. Thus, truncated type I or type II receptors induce similar DMZ marker expression in VMZ explants, consistent with the appearance of ectopic muscle, but not notochord, in tBRII-induced secondary axes. The failure to detect ectopic gsc expression in whole embryos injected ventrally with tBRII RNA, while gsc was detected in tBRII-loaded VMZs, could reflect the different sensitivities of the two assays, but may also be related to the repressive actions of endogenous ventralizing signals present in the whole embryo.

Because tBRII-induced secondary axes extended more anteriorly than those induced by tBR, we tested whether coexpressing both truncated receptors might affect the completeness of the induced secondary axis. Embryos were injected into two ventral cells at the 4-cell stage with tBR or tBRII RNA, alone or in combination, at doses from 20 to 1000 pg/embryo for each receptor (total combined doses of 40 to 2000 pg; Table 2). While the secondary axes induced by coexpressing tBR and tBRII were essentially indistinguishable morphologically from those induced by tBRII alone (data not shown), tBR/tBRII did synergistically increase the incidence of axis induction relative to injection dose. For example, co-injecting 20 pg of each receptor RNA (40 pg total) induced secondary axes in ∼50% of the surviving embryos. In contrast, 20 pg of tBR or tBRII RNA alone did not induce secondary axes, while doses of 50 pg of tBR or tBRII RNA caused axial induction in only 2% or 18% of the embryos, respectively. In contrast, the degree of neuralization of animal cap ectoderm effected by coexpressing both truncated receptors was similar to that caused by each truncated receptor alone (data not shown.

We then tested whether tBRII could block mesoderm induction by BMP4, and asked whether full-length versions of two different type II receptors, XBMPRII and the activin XAR1 receptor, could reverse this block. Embryos were injected at the 1-cell stage with 2 ng of RNA encoding either tBRII or the truncated activin type II receptor ΔXAR1. Animal caps were then explanted and incubated with or without recombinant human BMP4 (rhBMP4) at 50 ng/ml, and analyzed by RT-PCR at stage 14 for expression of the Xbra, and the ventrolateral mesodermal marker, Xhox3 (Ruiz i Altaba and Melton, 1989), which are both induced in animal caps by BMP4 (Dale et al., 1992; Jones et al., 1992). As shown in Fig. 7, rhBMP4 induced Xbra and Xhox3 (lane 5), and this induction was completely abrogated (Xbra), or substantially diminished (Xhox3), by tBRII, indicating an efficient block to ventral mesoderm induction by rhBMP4. This block was reversed by coinjecting RNA encoding full-length XBMPRII (lane 7) or activin type II receptor (lane 8). ΔXAR1 also eliminated mesoderm induction by rhBMP4 (lane 9), as shown previously (Chang et al., 1997; Hemmati-Brivanlou and Thomsen, 1995). However, both the wild-type activin or BMP type II receptors rescued the ΔXAR1-induced block in Xbra and Xhox3 expression (lanes 10 and 11). The finding that both type II receptors can restore functional BMP4 signaling indicates that these activin and BMP type II receptors are both able to mediate BMP4 signaling in vivo. Thus, XAR1 has the potential to be a shared component in ventralizing and dorsalizing pathways.

The selectivity of XBMPRII for BMP versus activin signaling pathways in vivo was tested by challenging activin-induced mesoderm induction for its sensitivity to tBRII (Fig. 8). Animal caps from embryos injected with 1 or 2 ng of tBRII RNA at the 1-cell stage were cultured with or without 5 ng/ml activin and compared to animal caps loaded with RNA encoding ΔXAR1. As expected, activin induced profound convergent extension in uninjected caps (Fig. 8A) and induced Xbra and actin (Fig. 8B), while ΔXAR1 dose-dependently blocked extension and marker induction. In contrast, tBRII did not block induction by activin, regardless of the amount injected. By both morphological and molecular criteria, we conclude that tBRII does not crossreact with activin signaling pathways.

We also tested whether tBRII could interfere with signaling by an unrelated ligand, bFGF. Our results were similar to those reported for the truncated activin type II receptor and truncated BMP type I receptor (Hemmati-Brivanlou and Melton, 1992; Suzuki et al., 1995) in that, instead of abrogating bFGF activity, injection of tBRII RNA altered the response of animal caps to bFGF. Caps treated with bFGF (50 ng/ml) normally form vesicular, ovoid structures containing mostly ventral mesoderm. However, the majority of caps receiving both tBRII RNA and bFGF treatment became substantially elongated with the elaboration of large cement glands at either end (data not shown). From histological analysis, we surmised that these caps contained both anterior and posterior neural tissues, the latter being likely responsible for the explant extension. This situation would be explained by a combination of anterior neural induction by tBRII together with partial posteriorization by bFGF, consistent with previous findings (see above mentioned references).

While TGFβ/BMP-related factors are crucial to embryogenesis (reviews by Hogan, 1996; Kingsley, 1994), only a subgroup of receptors mediating their effects have been identified. In addition, the potential for receptor crossreactivity both in vitro and in vivo, and the lack of data on spatial expression at both the RNA and protein level currently make it difficult to assign specific ligand effects in vivo to particular receptor combinations. Our studies here address these issues by providing the first detailed expression analysis of any BMP receptor, XBMPRII, at early stages of embryogenesis when critical patterning decisions are being made. Our data suggest that XBMPRII can mediate the mesodermal patterning and neural suppressing effects of BMP4 in the embryo, and we have demonstrated a qualitative difference between XBMPRII, the only type II BMP-specific receptor known to date, and the type II activin receptor, XAR1, in their crossreaction between dorsalizing (activin) and ventralizing (BMP) signaling pathways. Furthermore, we have shown that similarly truncated BMP type I and type II receptors have recognizably different effects in vivo.

Mammalian BMPRII selectively binds BMP2, BMP4 or BMP7 in vitro (Kawabata et al., 1995; Liu et al., 1995; Nohno et al., 1995; Rosenzweig et al., 1995), although the strength of binding and signal transduction depends upon the ligand (see below). The 63% identity (∼81% similarity) between human and frog BMPRII in the region between the signal sequence and transmembrane domains predicts a similar ligand specificity in vivo. In the Xenopus embryo, BMP2, BMP4 and BMP7 are expressed in somewhat overlapping domains in the animal hemisphere and marginal zone during blastula and gastrula stages, but their expression domains become more distinct as development proceeds (Clement et al., 1995; Fainsod et al., 1994; Hawley et al., 1995; Hemmati-Brivanlou and Thomsen, 1995). From the early gastrula stage, XBMPRII expression appears most similar to that of BMP4, which is especially apparent in the premesodermal region in gastrula embryos where both XBMPRII (Fig. 2) and BMP4 transcripts (Fainsod et al., 1994; Hemmati-Brivanlou and Thomsen, 1995; Schmidt et al., 1995) start to become restricted from the dorsalmost regions.

Although the activin type II receptors have been implicated in BMP signaling (Yamashita et al., 1995), BMPRII/XBMPRII is the only known type II receptor that preferentially binds BMP ligands in vitro. Thus, the heteromer model predicts that BMPRII/XBMPRII, together with type I receptors, would provide the most productive signaling by BMPs. The absence of XBMPRII expression from the animal region of gastrula-stage embryos is consistent with the hypothesis that removing BMP signals, which leads to a loss of epidermalizing signals, is a major component of the neural induction process (Hemmati-Brivanlou and Melton, 1997). By in situ hybridization analysis, XBMPRII RNA is also absent from the ventral ectoderm at early gastrula stages, which argues against its involvement in maintaining the epidermal fate of this region. However, it is possible that the wider expression of XBMPRII RNA throughout the animal region at earlier stages (Fig. 2) might provide these embryonic cells with a pool of stable type II receptors, giving them the potential to respond to BMP ligands that are first secreted after the disappearance of XBMPRII transcripts. In principle, these ligands would include BMP2 and BMP7, for which transcripts are expressed maternally (and thus stand in contrast to BMP4), and are enriched in the animal hemisphere during blastula stages (Clement et al., 1995; Hawley et al., 1995; Hemmati-Brivanlou and Thomsen, 1995). The generation of specific antibodies against different BMPs and XBMPRII would allow direct testing of these possibilities.

Although XBMPRII distribution has not been determined at the protein level, the finding that most XBMPRII RNA is localized to the marginal zone in gastrula-stage embryos suggests that its main function at this time of development lies in transducing BMP signals as part of the mesodermal patterning process. BMP4 signaling is particularly important during this window of embryogenesis, as demonstrated by several findings. Maternally derived BMP4 transcripts are expressed at very low levels (Dale et al., 1992), and zygotic BMP4 transcription occurs mostly after stage 9/10, during which its expression becomes rapidly restricted to the ventral and lateral regions of the marginal zone (Fainsod et al., 1994; Hemmati-Brivanlou et al., 1995; Schmidt et al., 1995). In addition, careful temporal analyses indicate that the ventralizing effects of BMP4 impact dorsoventral mesodermal patterning during and just after gastrulation, but not earlier (Jones et al., 1992, 1996).

Since exogenous BMP4 increases endogenous BMP4 expression (Jones et al., 1992), it has been proposed that BMP4 transcription is under positive autoregulatory control. If XBMPRII expression is also upregulated by ventralizing BMP signals, then blocking functional BMP4, for example, through the actions of noggin and/or chordin (Piccolo et al., 1996; Zimmerman et al., 1996), would also lead to decreased transcription of XBMPRII in dorsal regions. Such a scenario might locally reduce the potential for BMP signaling and facilitate the action of other dorsalizing molecules. Alternatively, the lower levels of XBMPRII expression in dorsal regions may directly result from the earlier actions of the Nieuwkoop center/Spemann organizer in dorsoventral specification. In either case, it is possible that the decreased XBMPRII expression in the DMZ protects these regions against the ventralizing effects of overly strong or residual BMP4 ligands.

Consistent with its expected role as a dominant negative inhibitor of XBMPRII-mediated signal transduction, we found that tBRII induced dorsal marker expression in ventral marginal zone explants. On the contrary, ventral injection of tBRII RNA induced only partial secondary axes and, by this criterion, did not induce an ectopic Spemann organizer. It is likely that the removal of ventralizing signals induces only partial dorsal specification, while complete axial development requires a full complement of Organizer activities, normally arising from Nieuwkoop center signaling. This is supported by the observation that injection of RNA encoding the organizer molecules noggin or gsc induces partial secondary axes (Cho et al., 1991; Smith and Harland, 1992), while Xwnt8 or siamois RNA mimic Nieuwkoop center activity and cause complete axial duplications (Lemaire et al., 1995; Sokol et al., 1991; Smith and Harland, 1991). Moreover, recent studies on Cerberus and lim-1/Xlim-1 indicate that subregions of the Organizer direct the formation of dorsoanterior structures through the activity of a ‘head organizer’ (Bouwmeester et al., 1996; Shawlot and Behringer, 1995). The failure to generate such head organizer tissue would explain the finding that localized injection of tBR or tBRII RNA into UV-ventralized embryos causes only a partial axial rescue, with the resulting embryos having only posterior tissues and rudimentary anterior structures (Graff et al., 1994; and our unpublished results).

Significant crossreaction has been demonstrated amongst several type I and type II receptors. For example, a truncated type II activin receptor, ΔXAR1, also interferes with signaling by BMPs (Hemmati-Brivanlou and Thomsen, 1995; Wilson and Hemmati-Brivanlou, 1995; Yamashita et al., 1995) and Vg1 (Schulte-Merker et al., 1994). Furthermore, XAR1 can be functionally replaced by the type II TGFβ receptor in activin-mediated mesoderm induction (Bhushan et al., 1994). Similarly, the BMPRIA, BMPRIB and ActRI type I receptors can all form complexes with BMPRII in vitro (Liu et al., 1995; Rosenzweig et al., 1995).

Despite the promiscuity in the assembly of signaling complexes and their potential for interacting with different ligands, not all ligand-receptor combinations cause equivalent signal transduction. Of specific relevance to our studies on XBMPRII, in vitro studies show that only low levels of binding and signal transduction arise when mammalian BMPRII/BMPRIB are coexpressed with BMP7, or when BMPRII/ActRI are coexpressed with BMP4 (Liu et al., 1995; Rosenzweig et al., 1995). In contrast, BMPRII/ActRI coexpression elicits high affinity binding and strong signal transduction by BMP7. Together with similar findings for other TGFβ-related receptors (e.g. Attisano et al., 1993), this has led to the idea that independent assortment of type I/type II receptors might produce a multitude of different receptor pairings that mediate distinct downstream effects of a particular ligand. Some support for this idea has come from in vivo studies in Xenopus embryos showing that a truncated activin type I receptor can block the activity of BMP4 as a mesoderm-inducing signal, but not as an epidermalizing factor (Chang et al., 1997). It is clear that information on the preferred type I/type II pairings and their activities in different cell types will further enhance our understanding of the molecular mechanisms for patterning by BMPs.

Our results demonstrate that tBRII expression is sufficient to block BMP4-induced mesoderm induction, dorsalize ventral mesoderm and neuralize ectoderm. Together with the overlapping expression patterns of XBMPRII and BMP4, the preferred binding of BMPRII to BMPs in vitro and the ability of the coinjected BMP-specific ventralizing factor, Smad1, to reverse the tBRII effects, the studies presented here imply strongly that XBMPRII mediates BMP signals in vivo. Moreover, we have also tested whether the BMPRII crossreactivity shown by overexpression studies in cultured cells occurs in vivo. ΔXAR1 blocks BMP signaling in animal caps, but tBRII does not cause a reciprocal block in activin signaling, indicating that XBMPRII is relatively specific for ventralizing BMP signals. Furthermore, the ability of both the full-length activin or BMP type II receptors to rescue a block in BMP-mediated mesoderm induction imposed either by ΔXAR1 or tBRII strongly suggests that this type IIB ‘activin receptor’ can, under appropriate circumstances, function as a BMP receptor in vivo. This finding would argue against the hypothesis suggested by New et al. (1997) in which different receptor/ligand affinities result in the transduction of ventral (BMP-like) or dorsal (activin-like) signals specifically through type II or type IIB activin receptors, respectively. However, the activities of different receptor complexes in transducing BMP signals during embryogenesis will depend not only upon their ligand-binding preferences, but also on the expression patterns of the ligands and receptors. The finding that both BMPRII and XAR1 can interact with multiple type I receptors, and that XAR1 can function as a BMP receptor, makes it important to determine the relative amounts and expression patterns of these type II receptors during critical periods of BMP signaling and to determine the effect of different type I/type II pairings on ligand binding and signal transduction in vivo.

Although type I and type II BMP receptors are presumed to function in the same signaling complex, similarly truncated versions of each receptor have different effects on secondary axis induction. Ectopic axes induced by tBRII extend much more anteriorly, and contain Otx2-expressing neural tissues, cement glands and otic vesicles, whereas those induced by tBR do not, even at the highest doses (Suzuki et al., 1994; Graff et al., 1994, and our results). Several not necessarily mutually exclusive possibilities could explain this. First, since the primary function of TGFβ-like type II receptors is to bind ligand and initiate complex formation (ten Dijke, 1996), signaling specificity may be regulated more through type II receptor/ligand recognition than through type I receptor/ligand interactions. If this can be generalized to include the BMP receptors, then any of several coexpressed type I receptors might productively interact with the type II receptor to transduce BMP signals, and tBR might only block a subset of these functional complexes. But, if XBMPRII represents a shared component of all BMP signaling complexes in vivo, tBRII might more effectively inhibit BMP signaling. A second possibility is that tBRII is more stable than tBR in vivo, rendering it intrinsically more capable of inhibiting BMP signaling, either by having a higher probability of inhibiting endogenous signaling complexes or, even without entering into complexes with type I receptors, forming a stable ligand sink that more effectively diverts the ligand from functional signaling complexes.

Similarly, peptide stability may be higher in signaling complexes containing both tBR and tBRII, as compared to those with only tBR or tBRII. If such complexes were to sequester ligand more effectively, this might explain the synergistic increase in the incidence of secondary axis induction upon coexpressing both truncated receptors. Alternatively, since it has been proposed that both BMP receptor subtypes are required in order to achieve high affinity ligand binding (ten Dijke et al., 1996), normal levels of BMP signal transduction may require the cytoplasmic domains of both type I and II receptors for optimal interaction with downstream effectors like Smad1 or Smad 5 (Massague, 1996; Suzuki et al., 1997). The synergistic effects on ventral signaling observed by coexpressing both truncated receptor subtypes may reflect a more effective loss of interaction with the downstream factors, whereas complexes containing only one truncated receptor subtype might still be able to maintain low levels of activity.

To date, XBMPRII/BMPRII remains the only type II receptor specific for the BMP signaling pathway and, compared to other type I/type II receptors for the TGFβ-related ligands, is distinguished by the presence of an extended cytoplasmic tail (Fig. 1). It is possible that this domain is involved in regulating interactions with type I receptors and Smad proteins. For example, specific regions of this domain might serve to convert distinct signals received at the cell surface (for example from different homodimeric or heterodimeric ligands) into associations with different subsets of the Smad proteins and/or other effectors. Such flexibility in signaling might allow a single type II receptor to be adaptable to many different embryonic signaling situations.

We are very grateful to Masa Kawabata and Hal Moses for providing human and mouse T-ALK cDNAs. We thank Ali Hemmati-Brivanlou and Doug Melton for activin type II and BMP type I receptor constructs; Mike Jones and Jim Smith for the MZ15 and 12/101 antibodies, Richard Harland and Ken Cho for whole-mount in situ probes, David Kimelman for purified bFGF, Genetics Institute for recombinant BMP-4 and Genentech for purified activin. Ali Hemmati-Brivanlou and Richard Harland also provided insightful discussions during the progress of this project. We thank Brigid Hogan and Maureen Gannon for critical comments on the manuscript and Mike Ray for superb technical assistance. This work was supported by NIH grant HD-28062 and ACS grant DB53. A. F. is supported by the Vanderbilt University Medical Scientist Training Program.