Peptide growth factors from the fibroblast growth factor (FGF) and transforming growth factor-β families are likely regulators of mesoderm formation in the early Xenopus embryo. Although basic FGF is found in the Xenopus embryo at the correct time and at sufficient concentrations to suggest that it is the FGF-type inducer, the lack of a secretory signal sequence in the basic FGF peptide has raised questions as to its role in the inductive process. We show here that Xenopus basic FGF can ectopically induce mesoderm when translated from injected synthetic RNA within the cells of a Xenopus embryo. Basic FGF produced in this manner is able to induce the formation of both dorsal and ventral mesoderm with the type of mesoderm formed dependent on the inherent dorsal-ventral polarity of the animal hemisphere. Surprisingly, although Xenopus basic FGF produced from the injected mRNA has a potent mesodermalizing effect on animal hemisphere cells, virtually no phenotypic effect is observed with intact embryos. These results suggest that the role of Xenopus basic FGF is to specify the size of the marginal zone, and synergistically with a dorsally localized prepatterning signal, to initially establish the dorsal-ventral axis of the mesoderm.

Determination of both the dorsal-ventral and anterior-posterior embryonic axes are the first major steps in the development of an embryo. Genetic and molecular analyses in Drosophila have demonstrated that the axes are determined by processes occurring during oogenesis and embryogenesis through a combination of maternal gene products acting both from within and from the outside of the oocyte (Levine, 1988; Nusslein-Vollhard and Roth, 1989; Schupbach, 1987; Stevens et al., 1990; Stein et al., 1991). Vertebrate embryos, in contrast, specify their embryonic axes after fertilization through the use of diverse and still poorly understood mechanisms (Eyal-Giladi, 1984; Gerhart et al., 1989). In amphibian embryos, fertilization-dependent rotation of the egg cytoplasm is necessary to determine both anterior and dorsal axial positions, although it is still unclear how this mechanical movement is translated into a biochemical pathway (Scharf and Gerhart, 1980; Scharf and Gerhart, 1983; Vincent and Gerhart, 1987). One likely target for the rotation mediated events are the factors used in intercellular communication during early development. As first demonstrated by the experiments of Spemann and Mangold (Spemann and Mangold, 1924), these signalling factors are used in amphibian embryos to regulate both the formation of cell types and the specification of the body plan. One well-studied example of this phenomenon is the induction of mesoderm in Xenopus embryos which is formed and patterned under the influence of signals emanating from the vegetal hemisphere (Nieuwkoop, 1969a; Nieuwkoop, 1969b; Dale and Slack, 1987). Recent evidence indicates that these signals are likely to be members of the fibroblast growth factor and transforming growth factor-β families since addition of either basic fibroblast growth factor (bFGF) or supernatant from an activin-secreting cell line to an animal hemisphere explant from a Xenopus embryo converts cells that would normally form ectoderm into a variety of mesodermal cell types (Smith, 1987; Slack et al., 1987; Kimelman and Kirschner, 1987; Rosa et al., 1988; Smith et al., 1990; Sokol et al., 1990). Histological analysis of these animal caps has shown that a high concentration of activin induces the formation of dorsal mesodermal tissues whereas a low concentration produces ventral mesoderm (Green et al., 1990). In contrast, basic FGF at both high and low concentrations only induces animal caps to form ventral mesoderm. In studies that used Xenopus homeobox-containing genes as markers of anterior and posterior position, activin was shown to provide anterior positional information whereas bFGF specified the future posterior regions of the embryo (Ruiz i Altaba and Melton, 1989; Cho and De Robertis, 1990). Further advances in this area have been hampered by uncertainty as to the identity and location of the different inducing factors within the early embryo.

Recent experiments with a dominant negative mutant of the FGF receptor have clearly demonstrated that a FGF family member must play a central role in the regulation of mesoderm development in Xenopus (Amaya et al., 1991). However, the identity of the natural ligand is still uncertain. Basic FGF is currently the best candidate for the endogenous ligand since a transcript encoding bFGF has been found in the oocyte, and the bFGF protein is found both in the oocyte and in the early embryo (Kimelman et al., 1988; Slack and Isaacs, 1989). Although bFGF protein is present within the Xenopus embryo at sufficient levels to suggest that it is the natural mesoderm inducer, the lack of a secretory signal within the peptide sequence raises the question as to whether the Xenopus basic fibroblast growth factor (XbFGF) protein contained within the cells of the dividing embryo can function in the inductive process.

Previous studies on the role of bFGF in Xenopus mesoderm induction involved addition of the purified protein to animal cap explants. Since these experiments could not test whether XbFGF produced within the embryonic blastomeres is able to induce mesoderm, we have injected synthetic RNA encoding XbFGF into the one-cell embryo and then removed animal cap explants seven hours later. The results demonstrate that Xenopus bFGF can function effectively as a mesoderm inducer when the protein is produced within the cells of the early embryo. Surprisingly, the in vivo synthesized bFGF is a potent inducer of dorsal mesoderm unlike the predominantly ventral mesoderm formed in response to exogenously added XbFGF. The type of mesoderm produced in response to the injected bFGF RNA is determined by the responding animal cap, with dorsal mesoderm formed from dorsal animal cap cells and ventral mesoderm produced from ventral animal hemisphere cells. Finally we show that intact embryos injected with RNA encoding XbFGF develop without major abnormalities, suggesting that formation of mesoderm may be regulated by events occurring during the gastrula stages.

Embryos and growth factors

Fertilized embryos were prepared as previously described (Newport and Kirschner, 1982). After the jelly coat was removed with 2% cysteine-HCl, pH 7.8, the eggs were washed in 0.1× MMR (1× MMR is 0.1 M NaCl, 2 mM KC1, 1.0 mM MgSO4, 2.0 mM CaCl2, 5.0 mM Hepes and 0.1 mM EDTA). Fertilized eggs were separated from unfertilized eggs at the one-cell stage by gently testing the rigidity of the egg. For ultraviolet light treatment, embryos were irradiated on their vegetal hemispheres for 60 or 90 seconds with 254 run fight (model UVG-11, UVP inc.) prior to RNA injection. Recombinant Xenopus bFGF was prepared as described (Kimelman et al., 1988) and used at 50 or 250 ng/ml. XTC-MIF collected from the XTC cell line (Smith, 1987) was activated before use by treatment at 90°C for 5 minutes. It was used at a 1:10 dilution.

Isolation of animal caps

The upper portion of the animal hemisphere from injected or uninjected embryos was manually separated at stage 9 except as noted, with care taken to remove adhering vegetal cells. The explants were placed in individual wells of a 96-well plate that had been coated with 1% agarose and filled with l× MMR containing 50 μg/ml Gentamicin (Sigma). Where applicable, XbFGF or XTC-MIF was added to the buffer. The explants were harvested at stage 19–21 for RNA analysis or stage 35–40 for histology. One-fifth of a sample of five explants was analyzed in the RNAase protection assays.

Microinjection of embryos

The DNA sequences encoding XbFGF from bp 336 (the initiation codon) to bp 875 (73 bp after the termination codon) were excised from the 4.5 kb Xenopus bFGF cDNA (Kimelman et al., 1988). A BglII site was inserted at the 5’ end and a BornHl site was placed at the 3’ end. This was then inserted into the unique BglII site of the pSP64T translation vector (gift of D. Melton). Capped synthetic RNA was synthesized with SP6 (Krieg and Melton, 1984). 3 ng of this RNA was injected into one-cell Xenopus embryos as described (Moon and Christian, 1989). For injection of HB7 RNA, the HB7-15 cDNA (D. K., unpublished results) was inserted into the EcoRI site of Bluescript II SK+ (Stratagene). RNA was synthesized by linearizing the plasmid with EcoRV and synthesizing capped RNA using T3 polymerase. Following the injection, embryos were maintained for two hours at 16°C in 0.1 × MMR with 5% Ficoll, then transferred to 0.1 × MMR and placed at 23°C.

RNA isolation and analysis

RNA was prepared from whole embryos or animal cap explants by homogenization in a buffer containing proteinase K (Krieg and Melton, 1984). The ethanol precipitate was dissolved in DEPC-treated water and reprecipitated for at least 8 hours with an equal volume of 8 M LiCl at —20°C. For northern blot analysis, RNA from 5 embryos was dissolved in 6 μl of DEPC-water. Half of this sample was electrophoresed on a formaldehyde agarose gel, transferred to Duralon (Stratagene) and immobilized by UV cross-linking. The filter was probed with the coding region of Xenopus bFGF described above. RNAase protection was performed with a 500 bp antisense transcript of the -cardiac actin gene (Sargent et al., 1986) synthesized with SP6 polymerase in the presence of [32P]UTP (Krieg and Melton, 1984). The labelled transcript was hybridized to RNA from either one-half or a whole animal cap, or 10 μ g tRNA as a control. The conditions of RNAase treatment (Krieg and Melton, 1984) were modified so that the concentrations of RNAase A and T1 were 30 μ g/ml and 1.5 μ g/ml, respectively, and the incubation was changed to one hour at 30°C.

Histology

Animal caps were fixed in Bouin’s reagent then washed and stored in 70% ethanol at room temp. 7 μm sections were cut from samples embedded in paraffin and placed on gelatin-subbed slides. Following rehydration, the sections were stained by J. Cooke’s procedure (Green et al., 1990), with a few modifications. The rehydrated sections were hydrolyzed at 60°C in 1 N HC1 for 10 minutes. Following a 30 second water rinse, the sections were treated for 1 hour in Schiffs reagent (Baxter). Following a 10 minute water rinse, the sections were immersed for 2 – 3 minutes in 0.05% Light Green Yellowish, 0.05% glacial acetic acid, 0.7% Orange G, and 0.35% phosphotungstic acid. After a brief water rinse, the sections were bleached in 70% ethanol until the nuclei were clearly visible under a light microscope. This was followed by 30 seconds incubation in 95% ethanol, dehydration and mounting. The final sections were multicolored allowing the visualization of different cell types as described (Green et al., 1990).

Analysis of induction by injected bFGF

The Xenopus bFGF protein is encoded by a 4.5 kb mRNA transcribed during oogenesis and again after the gastrula stage during embryogenesis (Kimelman et al., 1988). Four small open reading frames precede the initiation codon for the XbFGF polypeptide, and approximately 4 kb of 3’ untranslated sequence follows it. The upstream regions may be involved in translational regulation since they prevent translation of the bFGF protein in vitro from a RNA copy of the full-length cDNA (Kimelman and Kirschner, 1989). Therefore, the sequence encoding the XbFGF protein was inserted into a Xenopus β -globin translation vector, such that the initiation codon for XbFGF is the first AUG codon in the synthetic RNA.

To determine if this RNA was stable throughout the period of mesoderm induction, fertilized Xenopus eggs injected with XbFGF RNA were harvested at various times after injection and analyzed on a northern blot. Although the synthetic RNA is eliminated by the tailbud stage (Fig. 1, lane 5), there is only a small decrease between the time of injection and the mid-blastula transition six hours later (Fig. 1, lanes 1 and 2).

Fig. 1.

Stability of synthetic XbFGF RNA. RNA was isolated at various times from whole embryos injected with synthetic XbFGF RNA. The RNA was analyzed on a northern blot hybridized with a labelled probe made from the XbFGF cDNA. RNA was harvested just after injection (stage 1), at stage 8, stage 10, stage 15 and stage 21. The endogenous XbFGF transcripts are not visible under these conditions.

Fig. 1.

Stability of synthetic XbFGF RNA. RNA was isolated at various times from whole embryos injected with synthetic XbFGF RNA. The RNA was analyzed on a northern blot hybridized with a labelled probe made from the XbFGF cDNA. RNA was harvested just after injection (stage 1), at stage 8, stage 10, stage 15 and stage 21. The endogenous XbFGF transcripts are not visible under these conditions.

Even at stage 10, which marks the end of the competence period for induction by bFGF, there is still detectable synthetic XbFGF RNA (Fig. 1, lane 3).

The ability of XbFGF protein to induce mesoderm when synthesized in vivo from injected RNA was measured using explants of the animal cap. Fertilized Xenopus eggs were injected prior to the first cleavage with 3 ng of synthetic XbFGF RNA. Embryos that cleaved abnormally or showed morphological deformations were removed. The upper portion of the animal hemisphere (the animal cap) was then dissected at stage 9, and left to develop until sibling embryos reached stage 19–21. At stage 9, animal caps from injected embryos were indistinguishable from those removed from uninjected embryos demonstrating that the injection of XbFGF RNA did not have a major effect on the cellular morphology of the embryos at this stage. For comparison, animal caps removed from uninjected embryos were treated with either purified recombinant XbFGF (Kimelman et al., 1988) or supernatant from the activin-secreting XTC cell line (XTC-MIF; Smith, 1987). As observed previously, untreated animal caps healed into uniform spheres by the neurula stage, whereas caps treated with XbFGF were slightly elongate (Fig. 2A, B; Slack et al., 1987; Kimelman and Kirschner, 1987). Addition of XTC-MEF to the animal cap produced a more dramatic phenotype with extensive elongation (Fig. 2C). Surprisingly, animal caps from embryos injected with XbFGF RNA also produced very elongated structures (Fig. 2D), although they were typically not as extreme as those seen with the addition of XTC-MIF.

Fig. 2.

Animal cap induction by injected XbFGF and added growth factors. Control animal caps (panel A), animal caps to which XbFGF (panel B), or XTC-MIF (panel C) was added, or animal caps from embryos injected with XbFGF RNA (panel D) were photographed at the equivalent of stage 18 in control embryos.

Fig. 2.

Animal cap induction by injected XbFGF and added growth factors. Control animal caps (panel A), animal caps to which XbFGF (panel B), or XTC-MIF (panel C) was added, or animal caps from embryos injected with XbFGF RNA (panel D) were photographed at the equivalent of stage 18 in control embryos.

Since the extent of elongation is not a quantitative measure of induction, the formation of muscle was determined by measuring α-cardiac actin mRNA levels (Mohun et al., 1984). Although this assays only the induction of one type of mesodermal tissue, it provides a convenient measure for the formation of dorsal mesoderm (Kimelman and Kirschner, 1987; Gurdon et al., 1985). As previously reported, addition of XbFGF to animal caps caused a poor induction of cardiac actin RNA, in one experiment below detectable levels (Fig. 3, lanes 3 and 6). This variability in cardiac actin induction produced by exogenous XbFGF has been observed by us and others (unpublished results; (Green et al., 1990), see Fig. 3). This may account for the frequent weak elongation of animal caps in response to added XbFGF. Both addition of XTC-MIF or injection of XbFGF RNA resulted in comparably high levels of cardiac actin gene expression in two separate experiments (Fig. 3, lanes 2, 4, 5, and 6). Both addition of XTC-MIF and injection of XbFGF RNA produced much more consistent levels of cardiac actin mRNA in these assays.

Fig. 3.

Induction of muscle by injected XbFGF RNA. RNA was isolated from control animal caps (lane 1), animal caps from XbFGF-injected embryos (lanes 2 and 5), and animal caps to which XbFGF (lanes 3 and 6) or XTC-MIF (lanes 4 and 7) was added. Lanes 1 through 4 are from one experiment and 5 through 7 from a second experiment. RNAase protection was used to measure the amount of α-cardiac actin RNA (arrow) in each sample. Each lane represents one embryo equivalent. The fragments at the bottom of the gel are due to protection by cytoskeletal actin, which provides a measure of the amount of RNA present in each sample.

Fig. 3.

Induction of muscle by injected XbFGF RNA. RNA was isolated from control animal caps (lane 1), animal caps from XbFGF-injected embryos (lanes 2 and 5), and animal caps to which XbFGF (lanes 3 and 6) or XTC-MIF (lanes 4 and 7) was added. Lanes 1 through 4 are from one experiment and 5 through 7 from a second experiment. RNAase protection was used to measure the amount of α-cardiac actin RNA (arrow) in each sample. Each lane represents one embryo equivalent. The fragments at the bottom of the gel are due to protection by cytoskeletal actin, which provides a measure of the amount of RNA present in each sample.

To assay for the presence of other mesodermal cell types, histologically stained sections of animal caps injected with XbFGF RNA were also examined. Injection of XbFGF RNA induced the formation of notochord cells in 17% of 46 animal caps examined (Fig. 4A) and frequently produced large blocks of muscle tissue (Fig. 4B). In contrast, addition of XbFGF only produces small diffuse clumps of muscle cells at concentrations as high as 200 ng/ml (Ruiz i Altaba and Jessell, 1991; our unpublished results) and it has not been observed to induce notochord, the most dorsal of the mesodermal tissues (Slack et al., 1987; Green et al., 1990). The extent of notochord formation by XbFGF injection was still well below the 80% level that we observed with XTC-MIF, which may account for the greater extent of elongation in animal caps caused by the XTC factor. These results demonstrate that XbFGF is able to induce dorsal mesoderm in animal caps when synthesized from ectopically injected synthetic RNA.

Fig. 4.

Histological analysis of animal caps from XbFGF-injected embryos. Animal caps were removed from embryos injected with XbFGF and left to develop until the equivalent of stage 36. Examination of these sections revealed vacuolated notochord cells (panel A) and large blocks of muscle (arrow, panel B). Neither extensive neural tissue nor cement gland was observed in these sections.

Fig. 4.

Histological analysis of animal caps from XbFGF-injected embryos. Animal caps were removed from embryos injected with XbFGF and left to develop until the equivalent of stage 36. Examination of these sections revealed vacuolated notochord cells (panel A) and large blocks of muscle (arrow, panel B). Neither extensive neural tissue nor cement gland was observed in these sections.

Dorsal-ventral differences

We next asked if animal cap cells along the dorsalventral axis showed inherent differences in their ability to respond to injected XbFGF RNA. The presence of the future dorsal side can be resolved unambiguously at the beginning of gastrulation (stage 10) since the first site of invagination occurs on the dorsal side of the embryo. Therefore, animal caps from embryos injected with XbFGF RNA were dissected into dorsal and ventral halves at stage 10, using the blastopore lip as a reference. RNA was isolated from the dissected halves at stage 19 and the extent of muscle formation was monitored by measuring cardiac actin mRNA levels. Cardiac actin mRNA was not detected at this stage in whole animal caps dissected at stage 10 from uninjected embryos, demonstrating that in normal embryos neither the dorsal nor ventral animal cap pieces at stage 10 are capable of producing muscle (Fig. 5, lane 1). Isolation of whole animal caps at stage 10 from embryos injected with XbFGF RNA resulted in the usual high level of cardiac actin expression (Fig. 5, lane 2). A similar level of cardiac actin was observed when the dorsal half of a stage 10 animal cap was assayed (Fig. 5, lane 3), whereas very little cardiac actin was detected in the ventral half of the animal cap (Fig. 5, lane 4). These results demonstrate that the animal cap must differ along the dorsal-ventral axis in its ability to respond to the XbFGF protein, a pre-existing bias toward induction that we refer to as the prepattern.

Fig. 5.

Prepatterning in the animal cap. Animal caps from control and XbFGF RNA-injected embryos were removed at stage 10 and assayed for cardiac actin expression at stage 19. The animal caps were also divided into dorsal and ventral halves based on the position of the blastopore lip. The same result was observed in two independant experiments. Lane 1, whole animal caps, uninjected embryos; lane 2, whole animal caps, XbFGF injected; lane 3, dorsal half, XbFGF injected; lane 4, ventral half, XbFGF injected.

Fig. 5.

Prepatterning in the animal cap. Animal caps from control and XbFGF RNA-injected embryos were removed at stage 10 and assayed for cardiac actin expression at stage 19. The animal caps were also divided into dorsal and ventral halves based on the position of the blastopore lip. The same result was observed in two independant experiments. Lane 1, whole animal caps, uninjected embryos; lane 2, whole animal caps, XbFGF injected; lane 3, dorsal half, XbFGF injected; lane 4, ventral half, XbFGF injected.

Rotation of the egg cytoplasm prior to the first cleavage has been shown to be necessary for establishing the final dorsal-ventral axis in the complete embryo (Vincent and Gerhart, 1987). In order to see if this rotation was necessary for the responsiveness of animal caps to XbFGF, the rotation of the cytoplasm was blocked by irradiating the vegetal hemisphere of the egg with ultraviolet light (Scharf and Gerhart, 1980) prior to the injection of XbFGF RNA. Since the effectiveness of the irradiation is variable between embryos from different frogs, two doses of irradiation were used. In the experiment shown in Fig. 6, irradiation for both 60 and 90 seconds eliminated any dorsal structures from control embryos, although 90 seconds of ultraviolet irradiation caused a higher degree of mortality (data not shown). As shown above, animal caps from normal uninjected embryos did not produce detectable levels of cardiac actin mRNA (Fig. 6, lane 1), whereas animal caps from irradiated embryos expressed a low level of cardiac actin (Fig. 6, lanes 2 and 3), an effect we have repeatedly observed. Injection of XbFGF RNA into unirradiated eggs produced a high level of cardiac actin expression (Fig. 6, lane 4), that was reduced by ultraviolet irradiation to the level seen in uninjected embryos (Fig. 6, lanes 5 and 6). The inability of XbFGF to induce the expression of cardiac actin in irradiated animal caps is not due to non-specific damage to the animal hemisphere since these cells are still able to form mesoderm when treated with activin A (Sokol and Melton, 1991). Therefore the initial rotation of the egg, which sets up the dorsalventral axis, is necessary for the high-level induction of muscle in response to XbFGF RNA injection.

Fig. 6.

UV irradiation eliminates the expression of muscle caused by injected XbFGF. Embryos were irradiated with ultraviolet light for the times indicated just after removal of the jelly coat and prior to injection of XbFGF RNA. RNA was isolated from animal caps at stage 20 and assayed by RNAase protection with a labelled cardiac actin transcript. Animal caps were collected from uninjected embryos (lanes 1–3) and from XbFGF-injected embryos (lanes 4–6) that were not irradiated (lanes 1, 4), or irradiated for 60 seconds (lane 2, 5) or 90 seconds (lane 3, 6).

Fig. 6.

UV irradiation eliminates the expression of muscle caused by injected XbFGF. Embryos were irradiated with ultraviolet light for the times indicated just after removal of the jelly coat and prior to injection of XbFGF RNA. RNA was isolated from animal caps at stage 20 and assayed by RNAase protection with a labelled cardiac actin transcript. Animal caps were collected from uninjected embryos (lanes 1–3) and from XbFGF-injected embryos (lanes 4–6) that were not irradiated (lanes 1, 4), or irradiated for 60 seconds (lane 2, 5) or 90 seconds (lane 3, 6).

Effect of XbFGF injection on whole embryos

Since injection of XbFGF RNA was able to induce both notochord and muscle tissue in animal cap explants, it was expected to also have a large effect on the whole embryo. To test this, XbFGF RNA was injected as before but the whole embryos were left to develop for three days. As a control, animal caps were removed from six of the injected embryos and the expression of cardiac actin was measured (data not shown). To control for non-specific effects of RNA injection, RNA from a Xenopus homeobox gene HB7 was also injected at the same concentration. RNA from this gene has not been found to cause a specific phenotypic change in Xenopus embryos at any concentration tested (unpublished results).

The injected embryos were scored for the different defects shown in Fig. 7: microcephaly, macrocephaly, and the presence of vesicles on the ventral side. In one experiment tabulated in Table 1, 6 of 112 uninjected embryos showed some degree of either microcephaly or macrocephaly (5.3%). Injection of HB7 RNA caused a similar number of defects (5 of 77; 6.5%). The injected XbFGF RNA caused the greatest number of abnormal embryos with 13 of 106 defective (12.3%). Most of these embryos showed a characteristic phenotype of microcephaly with ventral vesicles. However, 88% of the embryos were phenotypically normal despite the injection of XbFGF RNA. These results demonstrate that the embryo can develop normally even when ectodermal cells have received a mesoderm-inducing signal in an inappropriate location.

Table 1.

Phenotype of injected embryos

Phenotype of injected embryos
Phenotype of injected embryos
Fig. 7.

Phenotypic effects from injection of XbFGF and HB7 RNAs. Embryos were injected with 3 ng of XbFGF or HB7 RNA and allowed to develop until stage 37. Panel A, normal uninjected embryo; panel B, XbFGF RNA-injected embryo showing microcephaly and the presence of ventral vesicles; panel C, HB7 RNA-injected embryo with macrocephaly.

Fig. 7.

Phenotypic effects from injection of XbFGF and HB7 RNAs. Embryos were injected with 3 ng of XbFGF or HB7 RNA and allowed to develop until stage 37. Panel A, normal uninjected embryo; panel B, XbFGF RNA-injected embryo showing microcephaly and the presence of ventral vesicles; panel C, HB7 RNA-injected embryo with macrocephaly.

Injected XbFGF RNA induces mesoderm in the animal hemisphere

The experiments presented here have shown that Xenopus bFGF can induce mesoderm when the XbFGF protein is produced within the cells of the early embryo, demonstrating that XbFGF does not require a classical secretory signal sequence in order to be effective as a mesoderm inducer. Therefore we conclude that the XbFGF protein normally synthesized during oogenesis and present during the early embryonic stages (Kimel-man et al., 1988; Slack and Isaacs, 1989) is likely to have a major role in mesoderm induction. It has not yet been possible to determine definitively whether the XbFGF translated from the injected RNA is actually secreted, or whether it might work within the cell that translates it. However, the recent demonstration that the extracellular domain of the Xenopus bFGF receptor is essential for mesoderm induction (Amaya et al., 1991) argues that XbFGF must cross a membrane in order to be effective.

XbFGF induces dorsal mesoderm

A surprising result of our analysis was the induction of dorsal mesoderm by the injected XbFGF RNA. Previous results using the addition of bFGF protein to animal caps had indicated that XbFGF was mostly an inducer of ventral mesoderm with only weak muscle-inducing properties (Slack et al., 1987; Kimelman and Kirschner, 1987; Green et al., 1990). In contrast, using injected XbFGF RNA, we have consistently observed production of muscle cells, as indicated by the expression of the α-cardiac actin gene, at levels comparable to those produced by XTC-MIF. In addition, notochord, which is among the most dorsal of mesodermal tissues, was observed in 17% of the animal caps derived from XbFGF RNA injected embryos. This demonstrates that XbFGF is capable of inducing all types of mesoderm and is not solely a ventralizing signal as suggested earlier (Slack et al., 1987; Green et al., 1990). The discrepancy between our results and those obtained previously may be explained by the difference in the way that XbFGF was delivered to the animal cap cells. Mesoderm induction in Xenopus embryos has been shown to occur over approximately a seven-hour time span from the 64-cell stage until the beginning of gastrulation (Jones and Woodland, 1987); however, as little as 10 minutes of exposure to added XbFGF protein in the animal cap assay is sufficient to induce ventral mesoderm formation (Green et al., 1990). Hence it is possible that either an early or a long period of exposure to XbFGF, which occurs with the injection of XbFGF RNA, may more closely resemble the true mesoderm-inducing potential of the XbFGF protein.

Since the healing of the animal cap after dissection prevents a long period of exposure to factors added to the culture medium, only the effects of a short pulse of XbFGF are revealed by this type of assay.

Experiments in which the animal cap was separated into dorsal and ventral halves demonstrated that the production of dorsal tissue in response to injected XbFGF occurred only on the dorsal side of the embryo. However, since regions of the animal cap that would normally not form mesoderm were induced to form muscle by injected XbFGF RNA, the animal cap must be patterned in regions outside of the marginal zone, the area that becomes mesoderm in the normal embryo. This effect, which we refer to as the prepatterning of the animal cap, could either be due to the differential response of embryonic cells to XbFGF, or due to the degradation of the XbFGF protein or RNA on the ventral side in the injected embryos. We favor the former interpretation since the same dorsal-ventral prepatterning has also been observed for neural induction by dorsal mesoderm (Sharpe et al., 1987), and for mesoderm and neural induction in response to PIF, an activin A homolog (Sokol and Melton, 1991). However, the effect of the prepattern on mesoderm induction by XbFGF and activin differs in that activin can induce’ muscle on both dorsal and ventral sides of the embryo whereas XbFGF can only induce muscle on the dorsal side. Therefore the prepatterning signal is unlikely to dictate the type of mesodermal or neural tissue that forms in a particular part of the embryo but instead may stimulate the strength of the response to any mesoderm or neural inducing signal.

Effects on whole embryos

Since injection of XbFGF RNA converts many of the ectodermal cells to mesoderm in the animal cap assay, we expected that it would have a deleterious effect in the whole embryo. For example, it has been shown that injection of an extremely high concentration of bFGF protein into the blastocoel converts some of the inner lining of the blastocoel to mesoderm and partially disrupts normal gastrulation of the embryo (Cooke and Smith, 1989). A number of the XbFGF RNA-injected embryos do have minor defects such as microcephaly and macrocephaly that may be gastrulation-related defects. However, similar defects are often seen at a low level with injection of any synthetic RNA. The surprising result was the high number of normal embryos. This was not due to inactivation of the XbFGF protein in most of the embryos, since each of the animal caps isolated from a random sample of the injected embryos formed mesoderm. These results indicate that the embryo can self-regulate to cope with cells that are exposed to XbFGF in inappropriate locations.

Previous studies on whole embryos have shown that to a large extent Xenopus embryos can regulate the amount of tissue dedicated to mesoderm, and even the proportion of cells allocated to each type of mesodermal tissue (Cooke, 1989; Cooke, 1981). Therefore, although inducing factors such as XbFGF and activin may initially specify the mesodermal pattern, these or other later acting factors may also regulate the amount of mesoderm formed and the distribution of mesodermal cell types. We note, however, that the results with injected XbFGF and activin RNAs are quite different; whereas injection of XbFGF does not change the phenotype of the embryo, injection of activin produces a partial second axis (Thomsen et al., 1990). We propose that the difference in these results may be explained by the different roles of these two factors in the specification of the mesoderm: the principal role of’ XbFGF may be to initially establish the region of the embryo that will become mesoderm, whereas activin may be used in subsequent steps as the principal axial inducing signal of the Spemann organizer. The intact embryo appears able to limit the amount of mesoderm that is formed despite the ectopic injection of XbFGF RNA. However, the embryo is not able to overcome errant positional information as when activin RNA is injected into the ventral blastomeres of the early embryo (Thomsen et al., 1990).

Conclusion

The ability of XbFGF to induce both dorsal and ventral mesoderm in isolated animal caps suggests that the formation of the mesodermal pattern in the marginal zone may be accomplished by a dorsally localized prepatterning signal acting synergistically with a uniform signal of XbFGF in this region. Since the prepatterning exists in the animal cap outside of the region that will become mesoderm, the pre-patteming molecule must not itself be a mesoderm inducing agent. Interestingly, ectopically expressed Xwnt-8 has recently been shown to activate a pathway characteristic of such a prepatterning molecule (Christian et al., 1991). Therefore two signals, XbFGF and a Wnt-like factor, could cooperate to initially specify the complete dorsalventral axis of the mesoderm. As these findings differ from the three-signal model of Slack which envisioned XbFGF solely as a ventralizing signal (Slack et al., 1987), it will now be essential to develop experimentally testable models that explain both the recent results with defined factors and classical embryological studies.

We are very grateful to John Gerhart for helping us to think through the process of mesoderm induction. We thank Steve Hauschka, Tim Schuh, Jennifer Northrop, George von Dassow and Rob Cornell for their detailed criticisms of this manuscript, and Jan Christian and Randy Moon for sharing their unpublished results. D.K. wishes to thank Marc Kirschner for teaching him to think more like an embryologist than a molecular biologist. This work was supported by grant HD 27262 from the N.I.H. to D.K.

Amaya
,
E.
,
Musd
,
T. J.
and
Kirschner
,
M. W.
(
1991
).
Expression of a dominant negative mutant of the FGF receptor disrupts mesoderm formation in Xenopus embryos
.
Cell
66
,
257
70
.
Cho
,
K. W. Y.
and
De Robertis
,
E. M.
(
1990
).
Differential activation of Xenopus homeobox genes by mesoderm-inducing growth factors and retinoic acid
.
Genes Dev
.
4
,
1910
1916
.
Christian
,
J. L.
,
Olson
,
D. J.
and
Moon
,
R.
(
1991
).
Xwnt-8 modifies the character of mesoderm induced by bFGF in isolated Xenopus ectoderm
.
EM BO J. (in press)
.
Cooke
,
J.
(
1981
).
Scale of body pattern adjusts to available cell number in amphibian embryos
.
Nature
290
,
775
778
.
Cooke
,
J.
(
1989
).
Xenopus mesoderm induction: evidence for early size control and partial autonomy for pattern development by onset of gastrulation
.
Development
106
,
519
29
.
Cooke
,
J.
and
Smith
,
J. C.
(
1989
).
Gastrulation and larval pattern in Xenopus after blastocoelic injection of a Xenopus-derived inducing factor: experiments testing models for normal organization of mesoderm
.
Dev. Biol
.
131
,
383
400
.
Dale
,
L.
and
Slack
,
J. M. W.
(
1987
).
Regional specification within the mesoderm of early embryos of Xenopus laevis
.
Development
100
,
279
295
.
Eyal-Giladi
,
H.
(
1984
).
The gradual establishment of cell commitments during the early stages of chick development
.
Cell. Diff
.
14
,
245
55
.
Gerhart
,
J. C.
,
Danilchick
,
M.
,
Doniach
,
T.
,
Roberts
,
S.
,
Rowning
,
B.
and
Stewart
,
R.
(
1989
).
Cortical rotation of the Xenopus egg: consequences for the anteroposterior pattern of embryonic dorsal development
.
Development, Suppl
.
107
,
37
51
.
Green
,
J. B. A.
,
Howes
,
G.
,
Symes
,
K.
,
Cooke
,
J.
and
Smith
,
J. C.
(
1990
).
The biological effects of XTC-MIF: quantative comparison with Xenopus bFGF
.
Development
108
,
173
183
.
Gurdon
,
J. B.
,
Fairman
,
S.
,
Mohnn
,
T. J.
and
Brennan
,
S.
(
1985
).
Activation of muscle-specific actin genes in Xenopus development by an induction between animal and vegetal cells of a blastula
.
Cell
41
,
913
922
.
Jones
,
E. A.
and
Woodland
,
H. R.
(
1987
).
The development of animal cap cells in Xenopus: a measure of the start of animal cap competence to form mesoderm
.
Development
101
,
557
563
.
Kimelman
,
D.
,
Abraham
,
J. A.
,
Haaparanta
,
T.
,
Palisi
,
T. M.
and
Kirschner
,
M. W.
(
1988
).
The presence of fibroblast growth factor in the frog egg: its role as a natural mesoderm inducer
.
Science
242
,
1053
1056
.
Kimelman
,
D.
and
Kirschner
,
M.
(
1987
).
Synergistic induction of mesoderm by FGF and TGF-/J and the indentification of an mRNA coding for FGF in the early Xenopus embryo
.
Cell
51
,
869
877
.
Kimelman
,
D.
and
Kirschner
,
M. W.
(
1989
).
An antisense mRNA directs the covalent modification of the transcript encoding fibroblast growth factor in Xenopus oocytes
.
Cell
59
,
687
696
.
Krieg
,
P.
and
Melton
,
D.
(
1984
).
Functional messenger RNAs are produced by SP6 in vitro transcription of cloned DNAs
.
Nucl. Acids Res
.
12
,
7057
7070
.
Levine
,
M.
(
1988
).
Molecular analysis of dorsal-ventral polarity in Drosophila
.
Cell
52
,
785
6
.
Mohun
,
T. J.
,
Brennan
,
S.
,
Dathan
,
N.
,
Fairman
,
S.
and
Gurdon
,
J. B.
(
1984
).
Cell type-specific activation of actin genes in the early amphibian embryo
.
Nature
311
,
716
721
.
Moon
,
R. T.
and
Christian
,
J. L.
(
1989
).
Microinjection and expression of synthetic mRNAs in Xenopus embryos
.
Technique
1
,
76
89
.
Newport
,
J.
and
Kirschner
,
M. W.
(
1982
).
A major developmental transition in early Xenopus embryos. I. Characterization and timing of cellular changes at the midblastula stage
.
Cell
30
,
675
686
.
Nieuwkoop
,
P. D.
(
1969a
).
The formation of the mesoderm in urodelean amphibians I. The induction by the endoderm
.
Roux’s Arch. EnnvMech. Org
.
162
,
341
373
.
Nieuwkoop
,
P. D.
(
1969b
).
The formation of the mesoderm in urodelean amphibians II. The origin of the dorso-ventral polarity of the mesoderm
.
Roux’s Arch. EntwMech. Org
.
163
,
298
315
.
Nusslein-Vollhard
,
C.
and
Roth
,
S.
(
1989
).
Axis determination in insects
. In
Cellular Basis of Morphogenesis. Ciba Foundation Symp
.
144
,
37
55
.
Rosa
,
F.
,
Roberts
,
A. B.
,
Danielpour
,
D.
,
Dart
,
L. L.
,
Sporn
,
M. B.
and
Dawid
,
I. B.
(
1988
).
Mesoderm induction in Amphibians: the role of TGF-β2-like factors
.
Science
239
,
783
785
.
Ruiz I Altaba
,
A.
and
Jessell
,
T.
(
1991
).
Retinoic acid modifies mesodermal patterning in early Xenopus embryos
.
Genes Dev
.
5
,
175
87
.
Ruiz I Altaba
,
A.
and
Melton
,
D. A.
(
1989
).
Involvement of the Xenopus homeobox gene Xhox3 in patterning the anterior-posterior axis
.
Cell
57
,
317
26
.
Sargent
,
T. D.
,
Jamrich
,
M.
and
David
,
I. B.
(
1986
).
Cell interactions and the control of gene activity during early development of Xenopus laevis
.
Dev. Biol
.
114
,
238
246
.
Scharf
,
S. R.
and
Gerhart
,
J. C.
(
1980
).
Determination of the dorsal-ventral axis in eggs of Xenopus laevis: complete rescue of uvimpaired eggs by oblique orientation before first cleavage
.
Dev. Biol
.
79
,
181
198
.
Scharf
,
S. R.
and
Gerhart
,
J. C.
(
1983
).
Axis determination in eggs of Xenopus laevis: a critical period before first cleavage, indentified by the common effects of cold, pressure, and ultraviolet irradiation
.
Dev. Biol
.
99
,
75
87
.
Schupbach
,
T.
(
1987
).
Germ line and soma cooperate during oogenesis to establish the dorsal-ventral pattern of egg shell and embryo in Drosophila melanogaster
.
Cell
49
,
699
707
.
Sharpe
,
C. R.
,
Fritz
,
A.
,
Em
,
D.
and
Gurdon
,
J. B.
(
1987
).
A homeobox-containing marker of posterior neural - the importance of predetermination in neural induction
.
Cell
50
,
749
58
.
Slack
,
J. M. W.
,
Darlington
,
B. G.
,
Heath
,
J. K.
and
Godsave
,
S. F.
(
1987
).
Mesoderm induction in early Xenopus embryos by heparin-binding growth factors
.
Nature
326
,
197
200
.
Slack
,
J. M. W.
and
Isaacs
,
H. V.
(
1989
).
Presence of basic fibroblast growth factor in the early Xenopus embryo
.
Development
105
,
147
153
.
Smith
,
J. C.
(
1987
).
A mesoderm-inducing factor is produced by a Xenopus cell line
.
Development
99
,
3
14
.
Smith
,
J. C.
,
Price
,
B. M. J.
,
Van Nimmen
,
K.
and
Huylebroeck
,
D.
(
1990
).
Identification of a potent Xenopus mesoderm-inducing factor as a homologue of activin A
.
Nature
345
,
729
731
.
Sokol
,
S.
and
Melton
,
D. A.
(
1991
).
Pre-existent pattern in Xenopus animal pole cells revealed by induction with activin
.
Nature
351
,
409
411
.
Sokol
,
S.
,
Wong
,
G.
and
Melton
,
D. A.
(
1990
).
A mouse macrophage factor induces head structure and organizes a body axis in Xenopus
.
Science
249
,
561
4
.
Spemann
,
H.
and
Mangold
,
H.
(
1924
).
Uber induction von embryonalagen durch implantation artfremder organis atoren
.
Roux’s Arch. Dev. Biol
.
100
,
599
638
.
Stein
,
D.
,
Roth
,
S.
,
Vogelsang
,
E.
and
Nusslein-Vollhard
,
C.
(
1991
).
The polarity of the dorsoventral axis in the drosophila embryo is defined by an extracellular signal
.
Cell
65
,
725
35
.
Stevens
,
L.
,
Frohnhofer
,
H.
,
Klingler
,
M.
and
Nusslein-Vollhard
,
C.
(
1990
).
Localized requirement for torsolike expression in follicle cells for the development of terminal anlagen of the Drosophila embryo
.
Nature
346
,
660
3
.
Thomsen
,
G.
,
Woolf
,
T.
,
Whitman
,
M.
,
Sokol
,
S.
,
Vaughan
,
J.
,
Vale
,
W.
and
Melton
,
D. A.
(
1990
).
Activins are expressed in Xenopus embryogenesis and can induce axial mesoderm and anterior structures
.
Cell
63
,
485
493
.
Vincent
,
J. P.
and
Gerhart
,
J. C.
(
1987
).
Subcortical rotation in Xenopus eggs: an early step in embryonic axis formation
.
Dev. Biol
.
123
,
526
39
.