ABSTRACT
The TGFβ family member activin induces different meso-dermal cell types in a dose-dependent fashion in the Xenopus animal cap assay. High concentrations of activin induce dorsal and anterior cell types such as notochord and muscle, while low concentrations induce ventral and posterior tissues such as mesenchyme and mesothelium. In this paper we investigate whether this threshold phenomenon involves the differential effects of the two type I activin receptors ALK-2 and ALK-4. Injection of RNA encoding constitutively active forms of the receptors (here designated ALK-2* and ALK-4*) reveals that ALK-4* strongly induces the more posterior mesodermal marker Xbra and the dorsoanterior marker goosecoid in animal cap explants. Maximal levels of Xbra expression are attained using lower concentrations of RNA than are required for the strongest activation of goosecoid, and at the highest doses of ALK-4*, levels of Xbra transcription decrease, as is seen with high concentrations of activin. By contrast, the ALK-2* receptor activates Xbra but fails to induce goosecoid to significant levels. Analysis at later stages reveals that ALK-4* signalling induces the formation of a variety of mesodermal derivatives, including dorsal cell types, in a dose-dependent fashion, and that high levels also induce endoderm. By contrast, the ALK-2* receptor induces only ventral mesodermal markers. Consistent with these observations, ALK-4* is capable of inducing a secondary axis when injected into the ventral side of 32-cell stage embryos whilst ALK-2* cannot. Co-injection of RNAs encoding constitutively active forms of both receptors reveals that ventralising signals from ALK-2* antagonise the dorsal mesoderm-inducing signal derived from ALK-4*, suggesting that the two receptors use distinct and interfering signalling pathways. Together, these results show that although ALK-2* and ALK-4* transduce distinct signals, the threshold responses characteristic of activin cannot be due to interactions between these two pathways; rather, thresholds can be established by ALK-4* alone. Furthermore, the effects of ALK-2* signalling are at odds with it behaving as an activin receptor in the early Xenopus embryo.
INTRODUCTION
The TGFβ family member activin induces animal pole tissue of blastula stage Xenopus embryos to form mesoderm rather than ectoderm (reviewed by Slack, 1994), and recent evidence using a dominant-negative activin receptor implicates activin as an endogenous mesoderm-inducing factor (Dyson and Gurdon, 1996). One remarkable property of activin is that it is capable of inducing different mesodermal cell types at different concentrations. Thus, low concentrations induce the formation of ventral and posterior cell types, and activate the expression of ventral and posteriorly expressed genes, while high concentrations induce the formation of dorsal and anterior tissues and activate the expression of dorsal and anteriorly-expressed genes (Green and Smith, 1990; Green et al., 1992, 1994; Gurdon et al., 1994; Symes et al., 1994; Wilson and Melton, 1994; Gurdon et al., 1995). The ability of activin to specify different cell types at different concentrations lends support to models proposing that tissue patterning can occur through the action of a diffusible ‘morphogen’ (Gurdon et al., 1994, 1995; Jones et al., 1996a).
Although precise interpretation of activin concentration in disaggregated-reaggregated animal cap explants requires cell-cell interactions and takes some time to occur (Green et al., 1994; Symes et al., 1994; Wilson and Melton, 1994), recent evidence suggests that a more direct and rapid interpretation of activin concentration is made in intact caps. When activin-soaked beads are sandwiched between animal cap explants, cells that are close to the activin source (and hence receive high doses of activin) express the dorsal marker goosecoid, whilst those at a greater distance activate Xbra (Gurdon et al., 1994, 1995; Jones et al., 1996a).
The mechanism by which cells distinguish between different concentrations of activin is unknown. Activin signals are transduced by heteromeric complexes of two serine-threonine kinases known as type I and type II receptors (reviewed by Mathews, 1994). Current models propose that ligand-induced association of the type I and type II receptors results in the phosphorylation and activation of the type I receptor which then initiates downstream signalling cascades (Wrana et al., 1994; Wieser et al., 1995). At least seven different type I receptors have been characterised in vertebrate cells, and tissue culture binding assays demonstrate that two of these, ALK-2 (ActR-I) and ALK-4 (ActR-IB), bind activin (Attisano et al., 1993; Cárcamo et al., 1994; ten Dijke et al., 1994; Tsuchida et al., 1996). Both ALK-2 and ALK-4 are expressed in the early Xenopus embryo (Kondo et al., 1996; Chang et al., 1997), as is another type I receptor, XTrR-I, which possesses a kinase domain with 98% identity to the human TGF-β receptor ALK-5 (Mahony and Gurdon, 1995).
These observations, together with recent work on Drosophila (Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994), prompted us to ask whether differential affinities and signalling specificities of these two receptors may account for the concentration-dependent effects of activin during mesoderm induction (see New et al., 1997). For example, if one receptor activates Xbra and the other goosecoid, and if the former possesses higher ligand affinity than the latter, then this would be consistent with current observations – low concentrations of activin would signal through the higher-affinity receptor and activate Xbra, while occupation of the lower-affinity receptor, and activation of goosecoid, would need high concentrations of activin. One requirement of this model is that the effects of the goosecoid-activating lower-affinity receptor should be ‘dominant’ over the Xbra-activating higher-affinity receptor.
To test this hypothesis we have expressed constitutively active forms of the ALK-2 and ALK-4 receptors (ALK-2* and ALK-4*) in animal caps and analysed their abilities to induce mesodermal markers. A single amino acid substitution in the juxtamembrane activation domain (the GS domain) of the type I receptor results in constitutive ligand-independent activation of the downstream signalling cascade (Wieser et al., 1995; Attisano et al., 1996; Hoodless et al., 1996). This provides a convenient means by which to investigate the signalling specificity of the type I receptors in the absence of their ligands – an important point, because the same ligand has been shown to associate with more than one type I receptor in tissue culture assays (Attisano et al., 1993; Cárcamo et al., 1994).
Expression of these activated receptors in the early Xenopus embryo reveals that whilst both induce mesoderm, only ALK-4* induces dorsal markers and does so in a dose-dependent fashion similar to activin. By contrast, ALK-2* activates ventral markers, and antagonises the effects of ALK-4*. These results are in direct contrast to the model presented above, and indicate that the threshold responses observed in response to activin need not involve interplay between the ALK-2 and ALK-4 receptors; rather, thresholds can be established by ALK-4 signalling alone. The data are also inconsistent with the idea that ALK-2 behaves as an activin receptor in the early Xenopus embryo.
MATERIALS AND METHODS
Xenopus embryos, microinjection and dissection
Xenopus embryos were obtained by in vitro fertilisation (Smith and Slack, 1983). They were maintained in 10% Normal Amphibian Medium (NAM: Slack, 1984) and staged according to Nieuwkoop and Faber (1967). Xenopus embryos at the one-to two-cell stage or at the 32-cell stage were injected with RNA dissolved in 10 nl or 1 nl water, respectively, as described by Smith (1993). For animal cap assays, embryos were dissected and cultured in 75% NAM.
In vitro transcription
Constitutively active murine ALK-2 and ALK-4 (ALK-2* and ALK-4*) constructs are described by Jones et al. (1996a). RNA was synthesized according to Smith (1993). The mutated kinase-defective ALK-2* clone, containing a K to R substitution at amino acid 235 (AVKIF to AVRIF), was constructed using PCR and was checked by sequencing. Note that mouse and Xenopus ALK-4 intracellular domains are 95% identical and 98% similar, whilst the ALK-2 homologues are 90% identical and 94% similar. Furthermore, the biological activities of human and Xenopus ALK-4 have recently been shown to be indistinguishable (Chang et al., 1997).
RNA isolation and RNAse protection assays
RNAse protection analysis was carried out as described by Jones et al. (1995), using RNAse T1 alone for all samples. Probes included cardiac actin (Mohun et al., 1984), EF-1α (Sargent and Bennett, 1990), goosecoid (Cho et al., 1991), Xbra (Smith et al., 1991), Pin-tallavis (Ruiz i Altaba and Jessell, 1992; O’Reilly et al., 1995), and Xhox3 (Ruiz i Altaba and Melton, 1989). An endodermin probe was constructed by excising an XhoI/SspI fragment corresponding to nucleotides 1928–2207 from a 5 kb endodermin cDNA kindly provided by Dr E. M. De Robertis. The fragment was cloned into pBSKS+. To make a probe, the plasmid was digested with XhoI and transcribed with T7 RNA polymerase.
Histology and immunocytochemistry
For histological analysis, specimens were fixed, sectioned and stained as described by Smith (1993). Whole-mount immunocytochemistry with monoclonal antibody 12/101 (Kintner and Brockes, 1984), specific for muscle, was performed as described by Smith (1993).
RESULTS
Constitutively active ALK-4 (ALK-4*) mimics induction of early mesoderm markers by activin
The mesoderm-inducing activity of constitutively active ALK-4 (ALK-4*) was tested by injecting ALK-4* RNA into Xenopus embryos at the one-cell stage and dissecting animal caps at stage 8 (mid-blastula). Caps were cultured to stage 10.5 (early gastrula) and assayed for expression of Xbra and goosecoid, genes which are induced by low and high concentrations of activin, respectively. Low doses of ALK-4* (10 pg RNA) were sufficient to induce expression of Xbra, and although expression levels first increased as more RNA was injected, they declined at the highest amount used (4 ng; Fig. 1). ALK-4* also induced expression of goosecoid, but the dose-response profile for this gene was different. Significant levels of expression needed more RNA (100 pg) than was required for Xbra, and the levels continued to increase as more RNA was injected; no decrease in expression was observed at the highest concentration of ALK-4* RNA as was observed with Xbra (Fig. 1).
Constitutively active ALK-4 (ALK-4*) induces the early mesodermal markers Xbra and goosecoid in a concentration-dependent fashion. Constitutively active ALK-2 (ALK-2*) induces only Xbra. Embryos were injected at the one cell stage with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. Animal cap explants were cut at stage 8.5 and expression of Xbra and goosecoid was assayed by RNAse protection at stage 10.5. EF-1α was used as a loading control; shorter exposures of the gel confirmed equal loading.
Constitutively active ALK-4 (ALK-4*) induces the early mesodermal markers Xbra and goosecoid in a concentration-dependent fashion. Constitutively active ALK-2 (ALK-2*) induces only Xbra. Embryos were injected at the one cell stage with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. Animal cap explants were cut at stage 8.5 and expression of Xbra and goosecoid was assayed by RNAse protection at stage 10.5. EF-1α was used as a loading control; shorter exposures of the gel confirmed equal loading.
High doses of ALK-4* RNA (1 ng or above in these experiments) caused a distinct morphological response in animal caps, in which the sensorial layer became positioned on the outside of the tissue and the pigmented layer aggregated and appeared to sink inside the caps (Fig. 2). This phenomenon, which was visible by the early gastrula stage (stage 10) has also been observed in animal caps following injection of high doses of RNA encoding activin (C. M. Jones and N. A. A., unpublished observations; see Jones et al., 1996a) or Xmad2 (Graff et al., 1996).
Comparison of the effects of ALK-2* and ALK-4* on animal cap explant morphology. Explants from uninjected embryos (A,B) or embryos injected at the one-cell stage with RNA encoding either ALK-2* (C,D) or ALK-4* (E,F) were cultured until early gastrula stage 10.5 (A,C,E) or neurula stage 17 (B,D,F). At the early gastrula stage, explants from uninjected embryos (A), or from embryos injected with 100 pg (C) or 1 ng (not shown) ALK-2* RNA remain spherical. In contrast, explants injected with 1 ng ALK-4* RNA (E) undergo inversion movements. At the neurula stage, explants from uninjected embryos (B), or from embryos injected with 100 pg (not shown) or 1 ng (D)ALK-2* RNA remain spherical. Animal caps derived from embryos injected with 100 pg ALK-4* undergo elongation, reminiscent of activin-treated animal caps (F), while caps from embryos injected with 1 ng ALK-4* RNA remain spherical (not shown).
Comparison of the effects of ALK-2* and ALK-4* on animal cap explant morphology. Explants from uninjected embryos (A,B) or embryos injected at the one-cell stage with RNA encoding either ALK-2* (C,D) or ALK-4* (E,F) were cultured until early gastrula stage 10.5 (A,C,E) or neurula stage 17 (B,D,F). At the early gastrula stage, explants from uninjected embryos (A), or from embryos injected with 100 pg (C) or 1 ng (not shown) ALK-2* RNA remain spherical. In contrast, explants injected with 1 ng ALK-4* RNA (E) undergo inversion movements. At the neurula stage, explants from uninjected embryos (B), or from embryos injected with 100 pg (not shown) or 1 ng (D)ALK-2* RNA remain spherical. Animal caps derived from embryos injected with 100 pg ALK-4* undergo elongation, reminiscent of activin-treated animal caps (F), while caps from embryos injected with 1 ng ALK-4* RNA remain spherical (not shown).
Constitutively active ALK-2 (ALK-2*) receptor strongly induces Xbra but not goosecoid
As with ALK-4*, low concentrations of RNA encoding constitutively active ALK-2 (ALK-2*) efficiently induced expression of Xbra. However, increasing amounts of ALK-2* RNA failed to alter levels of Xbra, and activation of the dorsal marker goosecoid was only very slightly raised against background (Fig. 1). In contrast to ALK-4*, ALK-2* did not induce, at the early gastrula stage, a change in cap morphology (Fig. 2).
Induction of dorsal cell types by ALK-4* and ventral cell types by ALK-2*
Later effects of ALK-4* and ALK-2* were studied by incubating injected animal caps to stage 17 for analysis of cardiac actin and Xhox3 expression, and to stage 35 for histological analysis. Analysis at stage 17 revealed that ALK-4* strongly induces cardiac actin expression at all RNA concentrations tested, together with expression of Xhox3 at low RNA doses (Fig. 3). Injection of 100 pg ALK-4* RNA caused animal caps to undergo activin-like extension movements (see Symes and Smith, 1987), whilst animal caps that had been injected with 1 ng or 4 ng ALK-4* RNA and whose sensorial layer and outer layer had previously become inverted either became compacted or had started to shed cells. As was seen with low doses of ALK-4*, ALK-2* induced expression of Xhox3, but no induction of cardiac actin was observed (Fig. 3). ALK-2*-injected animal caps did not undergo extension movements.
ALK-4* induces the late mesodermal markers muscle-specific actin and Xhox3 in a concentration-dependent fashion. ALK-2* induces only Xhox3. Embryos were injected at the one cell stage with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. Animal cap explants were dissected at stage 8.5 and assayed by RNAse protection at stage 17 for expression of muscle-specific actin and Xhox3.
ALK-4* induces the late mesodermal markers muscle-specific actin and Xhox3 in a concentration-dependent fashion. ALK-2* induces only Xhox3. Embryos were injected at the one cell stage with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. Animal cap explants were dissected at stage 8.5 and assayed by RNAse protection at stage 17 for expression of muscle-specific actin and Xhox3.
Histological analyses were consistent with these molecular data. Explants expressing low levels of ALK-4* (10 pg) usually formed mesenchyme and mesothelium (Fig. 4B), while those injected with 100 pg or more formed muscle and notochord (Fig. 4C). At high concentrations of ALK-4* RNA (≥1 ng), some explants did not resemble ectodermal tissue and nor did they contain characteristic mesodermal cell types (Fig. 4D). It seemed possible that these explants had formed endoderm, because this is the fate of the earliest cells to involute on the dorsal side of the embryo, and these are the cells which are likely to receive the highest levels of dorsal-specific signal. To test this idea, animal caps derived from embryos injected with ALK-4* RNA were cultured to the late gastrula stage (stage 13) and probed for expression of endodermin, which is most highly expressed in endoderm (Sasai et al., 1996) although it is also expressed in notochord and neural plate at gastrula stages. Expression of endodermin was induced by ALK-4*, and levels were particularly high in tissue derived from caps injected with sufficient RNA to promote cap inversion at stage 10 (Fig. 5). Analysis of endodermin expression was also performed at early tailbud stages (stage 24) when high levels of endodermin occur only in endodermal tissues, and the results were similar to those obtained at gastrula stages (data not shown).
ALK-4* induces a range of mesodermal cell types and perhaps endodermal tissues; ALK-2* induces only ventral mesoderm. Animal caps were dissected from embryos injected with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. They were cultured to the equivalent of stage 35, when they were fixed, sectioned and stained by the Feulgen/Light Green/Orange G technique (Smith, 1993). Control caps (A,E) form atypical epidermis. Explants derived from embryos injected with 10 pg ALK-4* RNA (B) form mesenchyme (Mes), explants derived from embryos injected with 100 pg ALK-4* (C) form muscle (Mus) and notochord (Not), and those derived from embryos injected with 1 ng ALK-4* RNA (D) form an unfamiliar tissue which may be endoderm (see text). Explants expressing ALK-2* form mesenchyme at all doses of RNA. (F) shows a cap derived from an embryo injected with 1 ng ALK-2* RNA. Scale bar in A, 50 μm.
ALK-4* induces a range of mesodermal cell types and perhaps endodermal tissues; ALK-2* induces only ventral mesoderm. Animal caps were dissected from embryos injected with increasing quantities of capped RNA encoding either ALK-2* or ALK-4*. They were cultured to the equivalent of stage 35, when they were fixed, sectioned and stained by the Feulgen/Light Green/Orange G technique (Smith, 1993). Control caps (A,E) form atypical epidermis. Explants derived from embryos injected with 10 pg ALK-4* RNA (B) form mesenchyme (Mes), explants derived from embryos injected with 100 pg ALK-4* (C) form muscle (Mus) and notochord (Not), and those derived from embryos injected with 1 ng ALK-4* RNA (D) form an unfamiliar tissue which may be endoderm (see text). Explants expressing ALK-2* form mesenchyme at all doses of RNA. (F) shows a cap derived from an embryo injected with 1 ng ALK-2* RNA. Scale bar in A, 50 μm.
ALK-4* induces the pan-endodermal marker endodermin. Explants from embryos injected with increasing quantities of RNA encoding ALK-4* were assayed at stage 13 by RNAse protection for expression of the endoderm marker endodermin.
Histological analysis of explants injected with RNA encoding ALK-2* revealed only ventral mesodermal cell types, even after injection of 1 ng ALK-2* RNA (Fig. 4F).
ALK-4* induces secondary axes
ALK-4* induces dorsal derivatives in animal cap explants, raising the possibility that it might induce secondary axes when injected into the ventral side of Xenopus embryos during cleavage stages. Trial experiments revealed that injection into the one-cell embryo, or injection into ventral blastomeres at the 8-cell stage, resulted in severe gastrulation defects. However injection of 150 pg ALK-4* RNA into each B4 blastomere (see Dale and Slack, 1987, for blastomere nomenclature) of the 32-cell stage embryo was successful in generating duplicated axes (Fig. 6B). In a typical experiment, 58% of injected embryos formed secondary axes (n=19). These axes were never complete, although in a separate experiment the secondary axes frequently contained an eye. Staining with the monoclonal antibody 12/101 revealed that secondary axes contained muscle.
ALK-4* but not ALK-2* can induce secondary axes when expressed in ventral blastomeres of the Xenopus embryo. 150 pg of RNA encoding either ALK-2* or ALK-4* was injected into each of the two most ventral B-tier blastomeres of 32-cell stage embryos. Embryos were allowed to develop to stage 33 before being fixed and lightly stained for muscle using the monoclonal antibody 12/101. (A) Control embryo. (B) ALK-4* induces a partial secondary axis (arrow) in 58% of embryos. Axes contained muscle and in some instances a single eye at their anterior end. (C,D) Injection of ALK-2* RNA never produced secondary axes but often produced defects on the ventral side of the embryo. Sometimes the ventral region formed a bulge (C) and sometimes the ventral tissue mass was decreased (arrow in D).
ALK-4* but not ALK-2* can induce secondary axes when expressed in ventral blastomeres of the Xenopus embryo. 150 pg of RNA encoding either ALK-2* or ALK-4* was injected into each of the two most ventral B-tier blastomeres of 32-cell stage embryos. Embryos were allowed to develop to stage 33 before being fixed and lightly stained for muscle using the monoclonal antibody 12/101. (A) Control embryo. (B) ALK-4* induces a partial secondary axis (arrow) in 58% of embryos. Axes contained muscle and in some instances a single eye at their anterior end. (C,D) Injection of ALK-2* RNA never produced secondary axes but often produced defects on the ventral side of the embryo. Sometimes the ventral region formed a bulge (C) and sometimes the ventral tissue mass was decreased (arrow in D).
Similar experiments performed with ALK-2* gave different results. Injection of ALK-2* RNA into the 1-cell stage embryo resulted in ventralisation of the embryo, whilst larger amounts caused a failure to involute properly, the blastopore enlarging instead of shrinking and moving up towards the animal hemi-sphere. Injection of 150 pg ALK-2* RNA into each B4 blastomere at the 32-cell stage never caused axis duplication (0%; n=26 in a typical experiment). Rather, some injected embryos formed a distended belly (Fig. 6C), while others appeared to be slightly deficient in ventral-posterior tissues (Fig. 6D).
ALK-2* suppresses ALK-4* signalling in a dominant and direct fashion
ALK-2*, unlike ALK-4*, cannot induce significant expression of goosecoid and nor can it induce dorsal mesodermal cell types. One explanation of this observation is simply that ALK-2* signals more weakly than ALK-4*, in the same way that the Drosophila type I receptor tkv appears to signal more weakly than sax (Singer et al., 1997). Alternatively, ALK-2* and ALK-4* might have signalling properties that are qualitatively different. To address this question, we have co-expressed the two constitutively active receptors. If ALK-2* is simply less active than ALK-4* it should have little effect on dorsal gene induction by ALK-4*. Preliminary experiments in which the receptors were expressed in combination revealed, contrary to our expectations, that co-expression of the two receptors resulted in gene expression patterns similar to those seen when injecting ALK-2* alone. Activation of muscle-specific actin gene expression by 1 ng ALK-4*, for example, was abolished by co-expression with 50 pg of ALK-2* (Fig. 7). Furthermore, the morphological changes associated with expression of ALK-4* were eliminated by co-expression of ALK-2* (data not shown).
ALK-2* abolishes the dorsalising effects of ALK-4*. Animal caps were dissected from embryos injected with 1 ng of ALK-4* alone or with 1 ng ALK-4* together with increasing quantities of RNA encoding ALK-2*. Explants were analysed at stage 17 for the expression of the dorsal mesoderm marker muscle actin.
ALK-2* abolishes the dorsalising effects of ALK-4*. Animal caps were dissected from embryos injected with 1 ng of ALK-4* alone or with 1 ng ALK-4* together with increasing quantities of RNA encoding ALK-2*. Explants were analysed at stage 17 for the expression of the dorsal mesoderm marker muscle actin.
Suppression of ALK-4* function by ALK-2* was further examined in a time-course experiment (Fig. 8A). Injection of 1 ng RNA encoding ALK-4* caused detectable elevation of goosecoid expression in animal caps within 1.5 hours of dis-section, and transcript levels continued to increase thereafter. Co-expression of ALK-2* prevented even the initial elevation of goosecoid transcription. This inhibition was achieved by injection of very small amounts of ALK-2* relative to ALK-4*, making it unlikely that the inhibitory effects of ALK-2* are due to competition for a common signalling component. Similar results were obtained with the dorsal-specific gene Pin-tallavis (Fig. 8A).
The inhibitory effects of ALK-2* on dorsal mesoderm induction caused by ALK-4* take place rapidly, suggesting that signal interference occurs prior to immediate transcriptional responses. A kinase-inactive ALK-2* is a poor inhibitor of ALK-4* dorsal mesoderm induction. (A) Animal cap explants were dissected at stage 8 from embryos injected with ALK-4* RNA alone or with the indicated combinations of ALK-4* and ALK-2* RNA. Explants were analysed for expression of the early mesodermal markers Pintallavis, goosecoid and Xbra by RNAse protection after 1.5, 2.5, 3.5 or 6 hours of subsequent culture. (B) Animal cap explants were dissected at stage 8 from embryos injected with ALK-4* RNA alone or with the indicated combinations of ALK-4* with either ALK-2* or ALK-2* kinase-inactive RNA. Explants were analysed for expression of the early mesodermal markers Pintallavis, goosecoid and Xbra by RNAse protection at stage 10.5.
The inhibitory effects of ALK-2* on dorsal mesoderm induction caused by ALK-4* take place rapidly, suggesting that signal interference occurs prior to immediate transcriptional responses. A kinase-inactive ALK-2* is a poor inhibitor of ALK-4* dorsal mesoderm induction. (A) Animal cap explants were dissected at stage 8 from embryos injected with ALK-4* RNA alone or with the indicated combinations of ALK-4* and ALK-2* RNA. Explants were analysed for expression of the early mesodermal markers Pintallavis, goosecoid and Xbra by RNAse protection after 1.5, 2.5, 3.5 or 6 hours of subsequent culture. (B) Animal cap explants were dissected at stage 8 from embryos injected with ALK-4* RNA alone or with the indicated combinations of ALK-4* with either ALK-2* or ALK-2* kinase-inactive RNA. Explants were analysed for expression of the early mesodermal markers Pintallavis, goosecoid and Xbra by RNAse protection at stage 10.5.
To investigate whether inhibition of ALK-4* function is due to active signalling by ALK-2*, an ALK-2* receptor was con-structed which carried a mutation in the kinase domain predicted, by analogy with the type I TGF-β receptor, to destroy kinase activity (Wrana et al., 1994; see Materials and Methods). Co-expression of the mutant receptor with ALK-4* was significantly less effective than the wild-type version in inhibiting ALK-4* induction of dorsal markers (Fig. 8B). For example, 100 pg of wild-type ALK-2* efficiently suppressed induction of goosecoid by ALK-4*, whilst a similar amount of mutant ALK-2* was less effective and had no effect on expression of Pintallavis. This suggests that active signalling from ALK-2* is required for complete inhibition of dorsal sig-nalling by ALK-4*.
We note, however, that mutant ALK-2* did cause some inhibition of ALK-4* function, and also, that in inhibiting expression of Xbra (see Fig. 8B), the effects of 1 ng mutant ALK-2* RNA differed qualitatively from the effects of 1 ng wild-type ALK-2. There are several explanations for the ability of kinase-inactive ALK-2* partially to inhibit ALK-4*-mediated gene induction. For example, kinase-defective and activation-defective receptors can functionally complement one another (Weis-Garcia and Massagué, 1996), so it is possible that wild-type endogenous receptors complement the exogenous kinase-defective activated receptors to produce an active signalling complex. Alternatively, it is possible that although inhibition by ALK-2* occurs by active mechanisms when physiological quantities of receptor are present, when large amounts of inactive receptor are injected, as are required to inhibit ALK-4* function, they do succeed in squelching common components of other signalling pathways. Our data support this second model, because inhibition of ALK-4*-induced dorsal induction by kinase-inactive ALK-2* also results in inhibition of expression of Xbra (Fig. 8B).
DISCUSSION
Activin induces prospective ectodermal tissue to express different mesoderm-specific genes, and to form different meso-dermal cell types, in a concentration-dependent fashion, as well as inducing endodermal tissues at the highest doses (Smith et al., 1988; Green and Smith, 1990; Green et al., 1992, 1994; Gurdon et al., 1994; Symes et al., 1994; Wilson and Melton, 1994; Gurdon et al., 1995; Jones et al., 1996a). The mechanism by which activin achieves this dose-dependent effect is unknown, but by analogy with work on Drosophila (Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994), we recently suggested that threshold formation might depend on two activin receptors with different ligand affinities and different signalling pathways (New et al., 1997). A high-affinity receptor would respond to low activin concentrations and induce the formation of ventral cell types, and a lower affinity receptor would only be activated by high activin concentrations and would induce dorsal tissues. A requirement of this model is that the lower-affinity receptor, which transduces a dorsal signal, would somehow override the action of the high-affinity receptor, which transduces a ventral signal.
Two type I receptors capable of binding activin have been identified in vertebrate cells: ALK-2 and ALK-4 (Attisano et al., 1993; Cárcamo et al., 1994; ten Dijke et al., 1994). We show here that constitutively active forms of the receptors, termed ALK-2* and ALK-4*, cause distinct transcriptional and morphological responses. ALK-4* induces expression of both Xbra and goosecoid. Lower levels of ALK-4* are required to activate Xbra expression than are required to induce goosecoid, and high levels of ALK-4* suppress Xbra expression. Similarly, expression of Xhox3 is induced by low, but not by high, levels of ALK-4*. At the morphological level, inter-mediate doses of ALK-4* induce convergent extension-like movements and high levels cause, at the early gastrula stage, inversion of animal caps. In all these respects, the effects of ALK-4* resemble those of activin itself (Green et al., 1990, 1992; Gurdon et al., 1994, 1995). By contrast, ALK-2* is a poor inducer of goosecoid and high levels of ALK-2* do not suppress expression of Xbra. Furthermore, ALK-2* does not cause changes in animal cap morphology.
These results demonstrate that there are differences in the signals transmitted by these two type I receptors. Their behaviour, however, is not consistent with the idea that they co-operate to assess activin concentration and thus generate threshold responses. Firstly, as described above, ALK-4* itself elicits dose-dependent responses. But secondly, and more importantly, even low concentrations of ALK-2*, which trans-duces a ventral mesoderm-inducing signal, abolish the effects of ALK-4*. This is the reverse of the prediction made above, thus disproving the two-receptor threshold model of New et al. (1997) at least for this pair of tested type I receptors. Indeed, the observed data is hard to reconcile with the binding affinities of the two receptors for activin previously reported, both receptors having KD values of about 100 pM when co-expressed with the ActR-II type II receptor (Attisano et al., 1993; Cárcamo et al., 1994).
The ability of ALK-2* to suppress the effects of ALK-4* resembles, at first sight, that of BMP-4 to inhibit the effect of activin (Dale et al., 1992; Jones et al., 1992), and indeed ALK-2 may function as a BMP receptor (ten Dijke et al., 1994; Liu et al., 1995). A significant difference, however, is that the effect of ALK-2*, when co-expressed with ALK-4*, is immediate (Fig. 8A); there is no transient activation of goosecoid, as is seen with co-expression of activin and BMP-4 (Jones et al., 1996b). This difference may occur, however, because the effects of exogenous BMP-4 require the synthesis of, for example, a type II receptor; a constitutively type I receptor would bypass this requirement.
Low levels of ALK-2* are capable of blocking dorsal induction by ALK-4*. This suggests, firstly, that the different effects of the two receptors when expressed singly are not due to differences in translation efficiencies. The observation also suggests, however, that the inhibition of dorsal induction requires active signalling by ALK-2*. Consistent with this idea, kinase-inactive ALK-2* has weak inhibiting activity compared with wild-type ALK-2* (Fig. 8B).
Together, these results suggest that the concentration-dependent effects of activin occur exclusively through the ALK-4 signal transduction pathway. One possibility is that different levels of ALK-4* activation are translated directly into different levels of activation of the Smad family member Xmad2, over-expression of which mimics the effect of activin in Xenopus embryos (Graff et al., 1996). Indeed, lower levels of Xmad2 RNA are required to activate expression of Xbra in animal caps than are required to induce goosecoid (Graff et al., 1996). The Smad proteins function as transcription activators (Liu et al., 1996), and if levels of extracellular activin concentrations are translated linearly into levels of Xmad2 activity, it may be possible to understand the concentration-dependent effects of activin using paradigms established for the concentration-dependent effects of the Drosophila morphogen Bicoid (Driever and Nüsslein-Volhard, 1989; Driever et al., 1989; Struhl et al., 1989; Ma et al., 1996).
Finally, our results appear inconsistent with the idea that ALK-2 functions as an activin receptor in the early Xenopus embryo. An alternative possibility is that ALK-2 binds to OP-1 (BMP-7) (ten Dijke et al., 1994), but OP-1 is a weak inducer of mesoderm, and unlike ALK-2* it does not inhibit the effects of activin in the animal cap assay (Yamashita et al., 1995). We are currently testing other candidate ALK-2 ligands.
ACKNOWLEDGEMENTS
This work is supported by the Medical Research Council. J. C. S. is an International Scholar of the Howard Hughes Medical Institute. We are grateful to Mike Jones for his helpful comments.