FGF is a prospective competence factor for early activin-type signals in Xenopus mesoderm induction

Inductive interactions are a widely observed phenomenon in metazoan embryogenesis. Proper development depends on the regulation of both the release of inductive signals and of the competence of cells to respond to those signals. While candidates for inductive signals have been identified in many systems, relatively little is known about the regulation of competence. For instance, Nieuwkoop’s conjugate experiments demonstrated that the vegetal hemisphere of the amphibian embryo sends signals that can convert animal hemisphere cells from an ectodermal to a mesodermal fate (Nieuwkoop, 1969a, b). These results lead to the proposal that the formation of the mesoderm at the equator of the embryo was due to an interaction between the inducing signals sent from the vegetal cells and the overlying animal cells which are capable of responding to these signals (reviewed by Kimelman et al., 1992; Sive, 1993). While strong candidates for endogenous mesoderm-inducing factors have been identified in the Xenopus embryo (Kessler and Melton, 1994), it has remained unclear why the vegetal cells are not also converted to a meso-dermal fate.

Recent results have suggested that the transforming growth factor-β (TGF-β) family member, Vg1, may be the inducing signal released by the vegetal cells. Vg1, like the TGF-β family member activin, has potent mesoderm-inducing abilities (Thomsen and Melton, 1993) and Vg1 transcripts are detected throughout the vegetal hemisphere (Weeks and Melton, 1987). Vg1, however, requires a unique post-translational activation event (Thomsen and Melton, 1993) and it is not yet known which regions of the embryo contain active Vg1 protein. An unidentified maternal activin-like activity has also been detected in the Xenopus embryo, but it is not known whether this activity is regionally localized (Asashima et al., 1991).

The Xenopus embryo also contains two members of the fibroblast growth factor (FGF) family, basic FGF (XbFGF) (Kimelman et al., 1988; Slack and Isaacs, 1989) and Xenopus embryonic FGF (XeFGF) (Isaacs et al., 1992), both of which can induce mesoderm when exogenously added to animal hemisphere cells (Kimelman and Kirschner, 1987; Slack et al., 1987; Isaacs et al., 1992). It is unknown why the embryo contains two families of signaling molecules with inducing capability, but experiments with dominant-negative receptors have demonstrated that both FGF and activin-type factors are necessary for normal embryonic development (Amaya et al., 1991; Hemmati-Brivanlou and Melton, 1992).

Two lines of evidence have led us to question whether FGFs normally act as direct mesoderm-inducing agents. First, expression of a truncated activin receptor, which acts as a dominant negative mutant, eliminates all embryonic mesoderm, but it does not prevent exogenous FGF from inducing mesoderm in the animal cap assay (Hemmati-Brivanlou and Melton, 1992). If FGFs acted as direct mesoderm-inducing agents, the truncated activin receptor would not be expected to eliminate all mesoderm from the embryo. In contrast, injection of a truncated FGF receptor eliminates only some of the embryonic mesoderm, producing embryos with aberrant bodies but relatively unaffected heads (Amaya et al., 1991). Second, expression of the truncated FGF receptor prevents exogenous activin from inducing the full spectrum of mesodermal genes in animal caps (Cornell and Kimelman, 1994a; LaBonne and Whitman, 1994), suggesting that FGFs may have a role in activin-mediated mesoderm induction in vivo. FGFs are not required for primary activin signaling, however, since several mesodermal genes are induced to normal levels by activin when FGF signaling is blocked (Cornell and Kimelman, 1994a).

We hypothesized that an absence of FGF signaling in vegetal cells could prevent them from becoming mesoderm, even if they were exposed to locally released activin-type signals. We show here that the addition of FGF to vegetal cells, even at levels which do not induce mesodermal gene expression in animal caps, causes them to express mesodermal markers. Furthermore, we show that vegetal cells release an activin-type signal, but do not themselves activate mesodermal gene expression in response to exogenous activin. We suggest that FGF does not act as a mesoderm-inducing factor in vivo, but instead acts to permit mesoderm induction mediated by an activin-type molecule, possibly Vg1. We propose that the distribution of FGF signaling helps to define the vegetal boundary of the mesoderm.

Xenopus adults and embryos were used (Newport and Kirschner, 1982) and staged (Nieuwkoop and Faber, 1967) as previously described. UV irradiation was applied for 60 seconds within 40 minutes of fertilization with a 4-watt UV lamp from UVP, Inc. (San Gabriel, CA).

Vegetal explants were cut at stage 9 by first removing the animal cap, then inverting the embryo and cutting away the marginal zone from below. All explants were dissected and cultured in 1× MBS (Kay and Peng, 1991) with 1 mg/ml bovine serum albumin (USB) and 25 μg/ml gentamycin (Sigma). Capped RNA for injection was synthesized (Krieg and Melton, 1984), using a m⁷G(5′)ppp(5′) GTP:GTP ratio of 8:1. 1 ng of RNA encoding XFD or d50 (Amaya et al., 1991) was injected into each blastomere at the 2-cell stage, alone or with 0.2 ng of BVg1 (Thomsen and Melton, 1993). 2 ng of Δ1XAR (Hemmati-Brivanlou and Melton, 1992) or globin (Green et al., 1983) RNA was injected into each vegetal blastomere at the 8-cell stage, 8 ng total. The ras constructs (Whitman and Melton, 1992) were injected at 40 pg per blastomere into three or four vegetal cells at the 8-cell stage, or at 20 pg per blastomere into three or four of the D-tier cells at the 32-cell stage.

At least 10 explants were pooled for each treatment, and RNA from 12 vegetal explants and 2-4 animal explants was subjected to RNase protection using the indicated probes as described (Cornell and Kimelman, 1994a) except that all but muscle actin reactions were done without RNase A (RNase T1 only). An antisense probe for Mix.1 (Rosa, 1989) was generated by removing the XbaI fragment from the full-length cDNA and religating the plasmid. This subclone was digested with Rsa1 and transcribed with T3 RNA polymerase.

Equivalent-sized animal and vegetal pole explants (approximately 1/3 of the embryo volume) were lysed and partially purified on wheat germ agglutinin, separated on 7.5% PAGE and transferred to a nylon membrane (Amaya et al., 1993). FGFR1 analysis was performed on lysates from 20 animal and vegetal explants using affinity purified Xenopus FGFR1 antibody (Amaya et al., 1993). β1-integrin levels were measured using lysates from 10 explants using mAb 8C8 (Gawantka et al., 1992) (a generous gift of P. Hausen) at approximately 5 μg/ml, except that protein was eluted from wheat germ agglutinin beads in sample buffer minus DTT. Horseradish peroxidase-coupled secondary antibodies were used for chemiluminescent detection using the manufacturer’s protocol (ECL, Amersham).

To determine whether vegetal cells respond to mesoderm-inducing factors, animal and vegetal explants were dissected from Xenopus embryos at stage 9 and cultured in XbFGF or activin until control embryos reached stage 17. In agreement with previous results (Green et al., 1990; Sokol et al., 1990; van den Eijnden Van Raaij et al., 1990), activin was a potent inducer and XbFGF only a weak inducer, of muscle actin in animal caps (Fig. 1A) (Kimelman and Kirschner, 1987). In vegetal caps, however, XbFGF induced muscle gene expression, whereas activin did not (Fig. 1A). Similarly, activin and XbFGF both induced expression of the pan-meso-dermal gene Xbra animal caps, but only XbFGF induced Xbra expression in vegetal caps, assayed at stage 10.5 (Fig. 1B).

As a further test, we examined the expression of the MyoD gene in vegetal caps. MyoD has a complex early pattern of expression but, by stage 12.5, it is restricted to presumptive mesoderm, primarily the presumptive muscle (Hopwood et al., 1989; Harvey, 1990; Scales et al., 1990; Frank and Harland, 1991; Rupp and Weintraub, 1991). As previously reported (Harvey, 1990; Rupp and Weintraub, 1991), both activin and XbFGF induced MyoD expression in animal caps (not shown), whereas XbFGF, but not activin, induced MyoD expression in vegetal caps at stage 12.5 (Fig. 1C).

We tested the ability of the other known Xenopus maternal FGF family member, XeFGF (Isaacs et al., 1992), to induce mesoderm in vegetal cells by injecting RNA encoding XeFGF into the vegetal pole of the 8-cell embryo and isolating vegetal caps at stage 9. XeFGF RNA induced the expression of Xbra in vegetal explants cultured to stage 10.5 (Fig. 2A), as did RNA encoding a secreted form of bFGF (Blam et al., 1988). In addition, injection of RNA encoding an activated ras protein (p21^v-Ha-ras) (Whitman and Melton, 1992), which can induce mesoderm in animal hemisphere cells by activating the FGF signal transduction pathway (Whitman and Melton, 1992; LaBonne and Whitman, 1994), induced Xbra expression in vegetal caps, while an inactive ras (p21(Ser 34)^v-Ha-ras) (Whitman and Melton, 1992) had no effect (Fig. 2A). Injection of RNA encoding the activated ras into the vegetal-most tier (D-tier) of cells at the 32-cell stage, also induced Xbra expression in vegetal explants (Fig. 2B), suggesting that the whole vegetal hemisphere is competent to express mesodermal genes. The level of Xbra induction was lower for the D-tier injection (Fig. 2B), but this probably resulted from fewer cells inheriting the injected RNA or its protein product when compared to the injection at the 8-cell stage. These results demonstrate that culturing vegetal cells in the presence of XbFGF or XeFGF, or activation of the FGF signaling pathway within vegetal cells, is sufficient to induce mesoderm in explanted vegetal tissue. In contrast, activin protein (Fig. 1) and RNA encoding activin (not shown) were unable to induce mesodermal gene expression in vegetal caps.

We hypothesized that the inability of activin to induce meso-dermal gene expression in vegetal cells was due to a lack of FGF signaling. An alternative explanation, however, was that vegetal cells lack a component of the activin signaling pathway. We therefore asked whether activin signaling occurs in the vegetal hemisphere. Previous studies have shown that vegetal cells express high levels of the homeobox gene Mix.1 (Rosa, 1989), in a pattern that is very similar to the distribution of Vg1 transcripts (Weeks and Melton, 1987). In the animal cap assay, Mix.1 is induced by activin, but not by FGF (Rosa, 1989), suggesting that Mix.1 could be directly induced by activin-type signals present in the vegetal hemisphere. We first asked whether disruption of activin-type signaling with a truncated activin receptor (Hemmati-Brivanlou and Melton, 1992), which also blocks Vg1-mediated signaling (Schulte-Merker et al., 1994), would eliminate the expression of Mix.1. As shown in Fig. 3A, Mix.1 is expressed at high levels in explanted vegetal caps. Overexpression of the truncated activin receptor in vegetal cells inhibited Mix.1 expression, whereas injection of a control RNA encoding β-globin had no effect (Fig. 3A). This result demonstrates that activin-type signaling occurs within the vegetal cells and suggests that the distribution of Mix.1 transcripts is directly related to the distribution of Vg1 signals.

Vegetal expression of another activin-responsive gene, goosecoid (gsc), has been detected in vegetal explants (Lemaire and Gurdon, 1994). We confirmed this vegetal expression (Fig. 3A) and, to ensure that the gsc expression was not due to contaminating dorsal mesoderm, we also dissected vegetal explants from embryos that had been ‘ventralized’ with UV light (DAI average = 1, n=20 embryos) (Scharf and Gerhart, 1980). Gsc expression was as strong in vegetal caps from ventralized embryos as it was in vegetal caps fromuntreated embryos, demonstrating that the vegetal expression of gsc does not depend on dorsal axis formation (Fig. 3B). Like the Mix.1 expression, the vegetal gsc expression was inhibited by the truncated activin-receptor (Fig. 3A), further supporting the hypothesis that vegetal cells respond to an activin-type signal.

Since Mix.1 and gsc are expressed in the vegetal region, which we propose has little or no FGF signaling, we asked whether Mix.1, like gsc (Cornell and Kimelman, 1994a), belongs to the subset of genes that can be induced by activin in the absence of FGF signaling. To test this, we used our previous animal cap assay, in which either control animal caps or animal caps expressing the truncated FGF receptor were treated with activin (Cornell and Kimelman, 1994a). As shown in Fig. 3C, Mix.1 and Xbra expression were induced by activin in animal caps from uninjected embryos. However, expression of the truncated FGF receptor dramatically reduced the activin-mediated expression of Xbra, but not of Mix.1. Similarly, co-expression of constitutively processed Vg1 (BVg1) (Thomsen and Melton, 1993) and an inactive mutant of the FGF receptor (d50, Amaya et al., 1991) induced both Xbra and Mix.1 in animal caps (Fig. 3D), but co-expression of constitutively processed Vg1 and the dominant-negative FGF receptor (XFD) severely reduced the Xbra expression (decreased by a factor of 14.4±2.9, n=3 experiments), while reducing the level of Mix.1 expression to a much lesser extent (decreased by a factor of 1.6±0.1, n=3 experiments). Thus for activin- or Vg1-mediated induction of Mix.1, FGF signaling is required to a much lesser extent than for the induction of Xbra. These results suggest that only genes that require little or no input from the FGF signaling pathway will be expressed within the vegetal region.

Having shown that vegetal cells contain endogenous activin-type signaling, we reasoned that the sensitivity of vegetal cells to respond to FGF might be increased relative to animal pole cells, since the animal hemisphere does not contain Vg1 RNA (Weeks and Melton, 1987). To test this hypothesis, animal and vegetal caps were dissected at stage 9 and treated with increasing levels of XbFGF. The expression of Xbra was determined by RNAse protection analysis (Fig. 4A), and the relative levels of Xbra expression were quantitated by phos-phoimager analysis (Fig. 4B). We found that greater than a10-fold excess of XbFGF was required to induce maximal A V levels of Xbra in animal explants compared to vegetal explants.

To ascertain whether vegetal caps are more sensitive to exogenous FGF signals because of a higher level of the FGF receptor, embryos were dissected into animal and vegetal halves and analyzed on western blots with an antibody directed against the FGF receptor. When compared to the integral membrane protein, β-1 integrin, which is homogeneously distributed throughout all blastomere membranes (Gawantka et al., 1992), the FGF receptor protein was more abundant in the animal caps than in the vegetal caps (Fig. 5). Therefore the greater sensitivity of vegetal explants to exogenous FGF cannot be explained by higher levels of the FGF receptor, instead the relative exclusion of the FGF receptor may be an important mechanism by which endogenous FGF signaling is reduced in vegetal cells.

We previously reported that FGF is required for activin-mediated mesoderm induction in an in vitro assay using animal caps (Cornell and Kimelman, 1994a; see also LaBonne and Whitman, 1994). Here we present evidence supporting the hypothesis that the role of maternal FGF is to act as a competence factor at the equator for mesoderm induction by activin-like signals emanating from vegetal cells. First, addition of XbFGF induces the expression of the mesodermal markers Xbra, MyoD and muscle actin in vegetal explants, as does injection of RNA encoding XeFGF or activated ras, while activin does not have this effect. These results show that there is nothing inherent to vegetal cells that prevents them from expressing mesodermal genes. Second, introduction of the truncated activin receptor diminishes Mix.1 and gsc expression in vegetal cells, indicating that activin-type signaling occurs in vegetal cells and, therefore, that it is not a lack of activin-type signaling that prevents mesoderm from arising in vegetal cells. Third, vegetal cells express maximal levels of Xbra in response to a much lower dose of FGF than do animal cells. This difference may be due in part to a possible increased penetrability of growth factor into vegetal explants, which lack the growth-factor-resistant (Green and Smith, 1990) epithelial layer of cells that surrounds animal explants. However, we also find that vegetal cells have far lower levels of the FGF receptor, which would presumably limit the total FGF signal received in the vegetal cells relative to the animal cells at a given dose of FGF. We believe that the increased sensitivity of vegetal cells suggests that, at least at low doses, the FGF added to vegetal cells does not directly induce mesoderm but rather complements an endogenous activin-type signal to induce mesoderm. Our interpretation of all these results is that FGF signaling is excluded from vegetal cells and this prevents activin-type signaling from activating mesodermal gene expression there.

In our previous study, we classified a number of genes, including Xbra, MyoD and muscle actin, as FGF-dependent, since expression of the truncated FGF receptor completely blocked their activin-mediated expression in animal caps. Other genes, including gsc and Xwnt-8, were classified as FGF-independent, since the truncated FGF receptor did not affect or only slightly diminished their induction by activin (Cornell and Kimelman, 1994a). Consistent with these results, Itoh and Sokol (1994) have shown that treating animal caps with heparanase, which destroys heparan-sulfate proteoglycan, a molecule known to be required by the FGF receptor (Yayon et al., 1991), blocked the activin-mediated expression of Xbra and muscle actin, but not of gsc and Xwnt-8.

An important prediction of the hypothesis that FGF signaling complements activin-type signaling only at the equator is that genes that depend on both FGF and activin signals should only be expressed at the equator, while genes that are activin-inducible and FGF-independent should also be expressed in the vegetal hemisphere. Mix.1, for instance, is expressed in the vegetal hemisphere and is shown here to be relatively independent of FGF (Fig. 3A). Gsc and Xwnt-8 were reported to be expressed only at the equator based on whole-mount in situ hybridization studies (Cho et al., 1991; Smith and Harland, 1991); however, whole-mount hybridizations yield artifactually low levels of staining in the vegetal hemisphere (Frank and Harland, 1992). We find expression of gsc (Fig. 3A,B) and Xwnt-8 (not shown) in RNase protection analyses of explanted vegetal tissue, consistent with the RNase protection analysis performed by Lemaire and Gurdon (1994) on explanted D-tier (deep vegetal hemisphere) cells, and with their in situ study showing vegetal expression of Xwnt-8 by hybridization to sections of stage 10.5 embryos. We have also observed vegetal expression of the homeobox gene Xlim-1 (R. C. and D. K., unpublished results), another gene that is induced by activin in the absence of FGF signaling (Cornell and Kimelman, 1994a). In contrast, muscle actin, Xbra and MyoD, which require both FGF and activin-type signals, are not expressed within the vegetal hemisphere (Fig. 1) (Lemaire and Gurdon, 1994) (see below). These results can be explained by the response of genes to FGF. If a gene requires FGF and activin signals, then it will only be expressed at the equator. If it requires only activin signaling, then it will be expressed at the equator and in the vegetal hemisphere.

MyoD at first looks like an exception to this rule since it is expressed in vegetal explants at stage 10.5 (Harvey, 1991; Rupp and Weintraub, 1991). However, in concurrence with our hypothesis, Isaacs et al. recently reported that the stage 10 expression of MyoD is not sensitive to disruption by the truncated FGF receptor (1994).

We propose that activin-type signaling is restricted to the vegetal hemisphere and marginal zone, with the upper limit of activin-type signaling determining the upper limit of the mesoderm (Fig. 6). As we have previously argued, since Xwnt-8 can induce a secondary axis in all regions of the equator (Smith and Harland, 1991; Sokol et al., 1991), an activin-like activity must be present throughout the equator to synergize with the ectopic Xwnt-8 (Kimelman et al., 1992). In addition, the distribution of Mix.1 transcripts in the embryo (Rosa, 1989) suggests that activin-type signaling occurs throughout the vegetal hemisphere, since Mix.1 and gsc expression are inhibited by the expression of a truncated activin receptor (Fig. 3). Finally, vegetal cells expressing the truncated activin receptor appear to adopt a neural fate (Hemmati-Brivanlou and Melton, 1994), arguing for a role for vegetal activin-type signals. The activin-type signal could be Vg1, or another activin-like activity, although the restriction of Vg1 transcripts to the vegetal hemisphere makes it an attractive candidate (Melton, 1987).

We have shown that FGF-signaling affects activin-type signaling (Cornell and Kimelman, 1994a; see also LaBonne and Whitman, 1994). We propose that FGF signaling is present at a low constitutive level in the animal hemisphere and equatorial region, but is present at a much lower level or absent in the vegetal hemisphere (Fig. 6). First, FGF signaling must occur in the equatorial region and in the animal cap because expression of a truncated FGF receptor blocks the normal expression of Xbra in the marginal zone (Amaya et al., 1993; Northrop and Kimelman, 1994) and the activin-mediated expression of Xbra in animal caps (Cornell and Kimelman, 1994a; LaBonne and Whitman, 1994). The activin-mediated induction of Xbra in animal caps is not inhibited by the protein-translation inhibitor, cycloheximide (Smith et al., 1991), which indicates that the FGF required for this process is maternally supplied. This is distinct from the requirement for a zygotic FGF at later times in development (Isaacs et al., 1994).

Two recent studies suggest possible mechanisms by which FGF signaling may be limited in the vegetal hemisphere. First, sensitive immunohistochemical analysis has shown that the concentration of maternal XbFGF protein is much lower in vegetal cells than in animal cells (Song and Slack, 1994). Second, the levels of FGF receptor protein are very low in the vegetal cells compared to the animal blastomeres (Fig. 5). Recently, the maternal FGF receptor mRNA has been shown to be under temporal translational control (Robbie et al., 1995). The limitation of FGF receptor protein levels in the vegetal cells may be achieved by spatial translational control of the FGF receptor mRNA, since this transcript is homogeneously distributed throughout the embryo (Musci et al., 1990). Indeed, the need to limit FGF signaling in the vegetal hemisphere is a likely reason for the embryo to have evolved translational control of the FGF receptor. Interestingly, FGF signaling also inhibits the expression of the endoderm-specific homeobox gene, Xlhbox8, in vegetal caps (C. Wright, personal communication), reinforcing the importance of excluding FGF signaling from the vegetal pole.

We propose that, in the normal embryo, the maternal FGF signal complements the activin-type signaling at the equator, leading to the equatorial expression of mesodermal genes. In the animal cap, the constitutive FGF signal is too low to directly induce mesoderm, but it will complement activin added exogenously under experimental conditions. In the vegetal hemisphere, the absence of FGF signals permits the expression of endodermal genes, which are induced by very high levels of activin-type signaling (C. Wright, personal communication; D. Melton, personal communication). Although the distribution of the activin-type signal is not known, these results suggest that the relative levels of FGF and activin-type signals are used to establish the mesoderm-endoderm boundary, with higher FGF signaling promoting the formation of mesoderm and higher activin-type signaling promoting the formation of endoderm.

The hypothesis presented here helps to explain a number of anomalous findings. First, it explains why the vegetal cells, which act as a source of mesoderm-inducing signals, do not become mesoderm. The lack of FGF signaling prevents the activin-type signals released by the vegetal cells from causing their neighbors to become mesoderm. Second, this view explains why studies with the dominant-negative FGF receptor indicate that FGF signaling is required for the development of dorsal mesoderm (Amaya et al., 1993; Northrop and Kimelman, 1993), even though FGFs are weak inducers of dorsal mesoderm by themselves (Green et al., 1990; Isaacs et al., 1992). The normal role of FGF is to complement the endogenous activin-type signal throughout the equator and not to induce any specific cell type. Third, since this hypothesis proposes that FGF acts only to permit mesoderm induction by activin-type signaling, it explains why a truncated activin receptor can eliminate all embryonic mesoderm without per-turbing the FGF signaling pathway (Hemmati-Brivanlou and Melton, 1992). Finally, this model provides a biological explanation for the observation that activin-mediated mesoderm induction requires FGF signaling (Cornell and Kimelman, 1994a; LaBonne and Whitman, 1994). In this proposal, the interaction between the activin and FGF pathways is important for the establishment of the mesoderm only at the equator of the embryo.

The ideas presented here are consistent with the view that the regulation of pattern within the embryo is under the control of several factors. Maternal FGF is proposed here to be important for activin-type signaling to induce mesoderm at the equator, whereas the zygotic expression of XeFGF is necessary to maintain this expression (Isaacs et al., 1994). Factors such as maternal Wnt and noggin may act to regulate the dorsalventral axis of the embryo at both early (Fig. 6) and late stages of development (Christian and Moon, 1993; Smith et al., 1993). While it is possible that a complicated distribution of one morphogen participates in the patterning of the mesoderm (Gurdon et al., 1994) and the endoderm (Hemmati-Brivanlou and Melton, 1992), we suggest that the embryo utilizes several factors, each of which is present in a relatively simple spatial gradient. While further experiments will be required to resolve these models of Xenopus embryogenesis, clear examples of combinatorial signaling can be found in the development of the C. elegans vulva (reviewed in Cornell and Kimelman, 1994b) and in the chick limb (Niswander et al., 1994). The limb bud represents an intriguing parallel to the study presented here, because in the limb bud, an FGF appears to act as a competence factor for the action of sonic hedgehog (Laufer et al., 1994).

We are very grateful to Celeste Berg, Lynn McGrew, Randall Moon, Richard Palmiter, Sarah Pierce and Tom Reh for reading this manuscript and supplying insightful comments. We thank Doug Melton and Chris Wright for informing us of their unpublished results, and Malcolm Whitman, Ali Hemmati-Brivanlou and Gerry Thomsen for providing constructs. Phosphoimager analysis was carried out by the Markey Molecular Center at the UW. R. C. was supported by a predoctoral training grant from NIGMS. This work was supported by an NIH grant to D. K.

Expression of Xbra and muscle actin in FGF-treated vegetal explants has been independently observed (Gamer, L. W. and Wright, C. E., submitted).