ABSTRACT
The Spemann organizer has long been recognized as a major source of patterning signals during the gastrula stage of amphibian embryogenesis. More recent evidence has suggested that the ventral side of the embryo also plays an important role in dorsal-ventral patterning during gastrulation through the action of signaling factors such as BMP-4. Bmp-4 is closely related to the Drosophila decapen-taplegic (dpp) gene, and like Bmp-4, dpp is excluded from the neurogenic region. Recently we showed that Bmp-4 functions in an analogous role to that of dpp in Drosophila, suggesting that the mechanism of dorsal-ventral patterning in Xenopus and Drosophila embryos may be conserved. To further test this hypothesis, RNA of the Drosophila short gastrulation (sog) gene was injected into Xenopus embryos, since sog has been shown genetically to be an antagonist of dpp function. Overexpression of sog RNA in Xenopus dorsalizes the embryo by expanding neurogenic and dorsal paraxial tissue. When ectopically expressed on the ventral side of the embryo, sog induces a partial secondary axis. In addition, sog partially rescues embryos ventralized by ultraviolet irradiation. Since sog induces many similar changes in gene expression to that caused by truncated BMP receptors, we suggest that sog functions in part by opposing BMP-4 signaling. The recent identification of a possible Xenopus sog homolog, chordin, in conjunction with these results supports the hypothesis that dorsalventral patterning mechanisms are conserved between these two species.
INTRODUCTION
Fertilization of the Xenopus egg initiates a cortical rotation that leads to the activation of a signaling cascade on the dorsal side of the embryo using maternal components (Gerhart et al., 1989). This process establishes a signaling center within the dorsal vegetal region that induces formation of the Spemann organizer at the dorsal equator of the embryo (Kimelman et al., 1992; Kessler and Melton, 1994). Maternal signals sent from the vegetal hemisphere also induce mesoderm throughout the equator of the cleavage-stage embryo.
During the late blastula and gastrula stages, further patterning of the mesoderm and ectoderm occurs. Several zygotic transcripts encoding intercellular signaling factors are expressed in the Spemann organizer, such as noggin (Smith and Harland, 1992), chordin (Sasai et al., 1994) and follistatin (Hemmati-Brivanlou et al., 1994). Previous models proposed that the gastrula-stage embryo was patterned solely by signals emanating from the organizer, with the resulting type of tissue determined by the relative distance of each cell from the organizer (Smith and Slack, 1983; Dale et al., 1985; Smith et al., 1985; Dale and Slack, 1987). Recent work, however, has shown that signaling molecules expressed on the ventral side of the embryo, such as BMP-4 and Xwnt-8, are also likely to be involved in patterning the mesoderm (Moon and Christian, 1992; Sive, 1993; reviewed in Harland, 1994; Schmidt et al., 1995). While ectopic expression of both molecules ventralizes Xenopus embryos, BMP-4 has a stronger effect, eliminating all dorsal and anterior-posterior structures (Dale et al., 1992; Jones et al., 1992). Elimination of BMP-4 signaling throughout the embryo with a truncated BMP receptor does not alter the expression of organizer-specific genes (Suzuki et al., 1994; Schmidt et al., 1995), but instead causes ventral mesoderm to adopt a more dorsal fate (Graff et al., 1994; Maeno et al., 1994; Suzuki et al., 1994; Schmidt et al., 1995). In addition, ectopic expression of a truncated BMP receptor on the ventral side of the embryo leads to the formation of a partial secondary axis containing paraxial mesodermal tissue, such as muscle, but lacking dorsal axial mesoderm, such as notochord and head mesoderm (Graff et al., 1994; Suzuki et al., 1994). Since Bmp-4 transcripts are localized to the ventral side of the gastrula-stage embryo, and excluded from a broad domain approximating the region of the future neural plate (Fainsod et al., 1994; Schmidt et al., 1995), it has been suggested that, in addition to patterning the ventral mesoderm (Graff et al., 1994; Maeno et al., 1994; Suzuki et al., 1994; Schmidt et al., 1995), BMP-4 acts to limit the dorsalizing potential of signals produced by the organizer, and therefore has a role in patterning the lateral mesoderm (Schmidt et al., 1995).
We also observed that ectopic expression of a truncated BMP receptor throughout the embryo leads to the formation of neural ectoderm on the ventral side of the embryo (Schmidt et al., 1995). Conversely, ectopic expression of BMP-4 leads to a reduction in the amount of neural ectoderm (Schmidt et al., 1995). These results are similar to those observed in Drosophila for the decapentaplegic gene (dpp), the product of which is a TGF-β molecule that is closely related to BMP-4 (Padgett et al., 1987). Elimination of dpp leads to a loss of dorsal structures and an expansion of ventral ectoderm from which the nervous system derives (Ray et al., 1991; Arora and Nusslein-Vollhard, 1992), whereas ectopic expression of dpp leads to the reduction of ventral ectoderm (Ferguson and Anderson, 1992b). Furthermore, dpp expression, like that of Bmp-4 is excluded from the lateral neuroectodermal region of the embryo (St. Johnston and Gelbart, 1987; Wharton et al., 1993). These similarities suggest that these two genes may play analogous roles in establishing the dorsal-ventral axis in both organisms, by delimiting the boundaries of the neural ectoderm (Schmidt et al., 1995).
In Drosophila, several mutants have been identified that have defects in embryonic dorsal-ventral patterning. One of these mutants, short gastrulation (sog), behaves genetically as an antagonist of the other genes involved in dorsal-ventral patterning of the early embryo (Ferguson and Anderson, 1992a; Wharton et al., 1993; Francois et al., 1994). sog has recently been cloned and is predicted to encode a secreted or membrane-bound protein (Francois et al., 1994). sog acts non-cell-autonomously (Zusman et al., 1988) over a distance of at least 12 cells (Francois et al., 1994), and is proposed to be an antagonist of dpp function (Ferguson and Anderson, 1992b; Wharton et al., 1993; Francois et al., 1994). We therefore decided to test whether similarities in the mechanisms of dorsal-ventral patterning in Xenopus and Drosophila extended beyond the similarities between dpp and BMP-4 by asking whether sog could regulate patterning in Xenopus embryos. We show here that ectopic injection of RNA encoding sog, like ectopic expression of a truncated BMP receptor, leads to the formation of a partial secondary axis that contains muscle but lacks a notochord. In addition, sog, like the truncated BMP receptor, can induce a partial axis in embryos ventralized by uv irradiation, causing the formation of neural structures as anterior as the hindbrain. Ubiquitous ectopic expression of sog leads to the expansion of Hairy II expression to the ventral side of the embryo, demonstrating that sog can regulate the size of the neurogenic region. With the recent identification in Xenopus of a dorsally expressed intercellular factor (chordin) with limited homology to sog (Sasai et al., 1994), these results suggest that the mechanism of dorsal-ventral patterning may be conserved between these evolutionarily distant organisms.
MATERIALS AND METHODS
Embryos and uv irradiation
Fertilized Xenopus embryos were prepared as described previously (Newport and Kirschner, 1982). After the jelly coat was removed with 2% cysteine (pH 7.8), the eggs were washed in 0.1× MMR (1× MMR is 0.1 M NaCl, 2 mM KCl, 1.0 mM MgSO4, 2.0 mM CaCl2, 5.0 mM Hepes, and 0.1 mM EDTA). In the uv-irradiation experiments, vegetal poles of Xenopus embryos were uv irradiated for 50 seconds within the first 30 minutes after fertilization (Scharf and Gerhart, 1980).
Construction of sog expression vectors
A sog cDNA (Francois et al., 1994) was digested with BssH2 and filled in with the Klenow fragment of DNA polymerase I. This removes the first 535 bp of the 5′ UTR, and all potential initiation codons in front of the predicted initiation codon. The sog fragment was excised at the 3′ end with NotI and inserted into the EcoRV and NotI sites of Bluescript SK+ (Stratagene). This vector was digested with NotI and filled in with the Klenow fragment of DNA polymerase I. The fragment was excised with SalI and inserted into the XhoI and SnaBI sites of CS2+ (Turner and Weintraub, 1994). This vector (Sog25) produced wild-type sog RNA.
To produce the easter-sog expression vector, a fragment of the easter gene (in a vector kindly provided by R. DeLotto (unpublished; see Smith and DeLotto, 1994 for the primers used to clone the easter signal sequence) containing the initiation codon and the subsequent 20 amino acids was fused in-frame to the valine at amino acid 99 of the sog cDNA (Francois et al., 1994). This removes the potential membrane-spanning domain from sog.
RNA synthesis and microinjection
RNAs were synthesized using the mMessage mMachine kit (Ambion) according to the manufacturer’s protocol provided. The RNAs were purified by one extraction with phenol:chloroform (1:1) followed by two rounds of concentration and separation in Microcon 100 micro-concentrators (Amicon) to separate the RNA from unincorporated nucleotides. RNA was microinjected as previously described (Moon and Christian, 1989). 5-10 nl RNA were injected per blastomere. Embryos were injected in either 2 or 4 blastomeres of 4-cell embryos or in 1 vegetal cell at the 8-cell stage. The dorsal side of four-cell embryos was identified by pigment and cell size differences between dorsal and ventral blastomeres at this stage (Nieuwkoop and Faber, 1967). Embryos were fixed in 1× MEMFA (Harland, 1991).
In situ hybridization and probe synthesis
Whole-mount in situ hybridization was performed using digoxigenin-labeled RNA probes (Harland, 1991), with a variety of modifications (R. Harland, personal communication). Antisense probes were syn-thesized from cDNAs encoding Xnot, gsc, MyoD, Xwnt-8, Xbra, Krox-20 and Hairy II, using plasmids linearized with HindIII, EcoRI, HindIII, BamHI, EcoRV, EcoR1 and BamHI, respectively. SP6 RNA polymerase was used for the gsc probe and T7 was used for the Xbra, Hairy II, MyoD, Krox-20, and Xwnt-8 probes.
Histology
After MEMFA fixation and storage in methanol, embryos were embedded in Paraplast. Embedded embryos were sectioned at 10 μm and mounted in Permount (Kelly et al. 1991). Some sections were stained with hematoxylin and eosin prior to mounting. Whole-mount embryos and sections were photographed using Kodak Ektachrome 160T film.
RESULTS
Ectopic expression of sog induces a partial secondary axis in Xenopus embryos
To test whether sog can function in Xenopus, a cDNA encoding sog (Francois et al., 1994) was inserted into a Xenopus expression vector. Since the sog cDNA has a large 5′ untranslated region (UTR) containing potential translation-initiation sites, the 5′ UTR was removed to provide optimal translation of sog in Xenopus embryos. Injection of 4-6 ng of synthetic RNA, produced from this construct, into the 2 dorsal cells of 4-cell Xenopus embryos had no visible effect (data not shown), whereas injection of this amount of RNA into the ventral side produced embryos with a variety of secondary axes (Fig. 1 and data not shown; 73% had secondary axes, n=141, 6 experiments). In some embryos, the secondary axis emerged from the primary axis in the lateral region of the embryo (Fig. 1, 3rd embryo from the top). In other embryos, the secondary axis was distinct from the primary axis, forming on the ventral side of the embryo and including a secondary tail (Fig. 1, top two embryos). In no case was a complete secondary axis containing a head observed. In addition, secondary dorsal lips were not observed during early gastrulation. Instead, the secondary axes became apparent during late gastrulation and neurulation. These results are very similar to the results observed when RNA encoding a truncated BMP receptor is injected into the ventral region of Xenopus embryos (Graff et al., 1994; Suzuki et al., 1994).
Drosophila sog induces a secondary axis in Xenopus embryos. The top three embryos were injected ventrally with sog RNA. The top two embryos have secondary tails and the third embryo’s secondary axis extends from the primary axis. The bottom embryo is an uninjected control embryo. All embryos are at stage 38.
Drosophila sog induces a secondary axis in Xenopus embryos. The top three embryos were injected ventrally with sog RNA. The top two embryos have secondary tails and the third embryo’s secondary axis extends from the primary axis. The bottom embryo is an uninjected control embryo. All embryos are at stage 38.
In order to understand the effects of ectopic sog expression on early patterning and to compare these effects with those elicited by a dominant negative BMP receptor, we examined the expression of the head mesodermal marker, goosecoid (gsc; Cho et al., 1991) and the presumptive notochord marker, Xnot (von Dassow et al., 1993). Expression of sog on the ventral side of the embryo caused weak ventral expression of gsc (Fig. 2B), but this level of expression was well below the normal level of gsc expression found in the dorsal organizer region (Fig. 2A,B; none had duplicated gsc expression, n=20). Similarly, ventral injection of sog RNA caused a slight enhancement of Xnot expression on the ventral side (Fig. 2D), however, this expression appeared to be an enhancement of the non-notochordal Xnot expression, which appears to define the limit of involution (von Dassow et al., 1993) (none had ventral duplication of Xnot expression, n=28).
Injection of sog RNA alters early gene expression. (A,B) Vegetal view of gsc expression in stage 10 embryos. Dorsal is up. (A) Uninjected embryo. (B) Embryo injected ventrally with 6 ng sog RNA. Gsc expression is only weakly enhanced on the ventral side. (C,D) Vegetal view of Xnot expression in stage 11 embryos. Dorsal is up. (C) Uninjected embryo. (D) Embryo injected ventrally with 6 ng sog RNA. Xnot expression is enhanced on the ventral side, however only in the non-notochordal domain of expression. (E,F) MyoD expression in mid-neurula-stage embryos. Anterior is up. (E) Uninjected embryo, dorsal view. (F) Embryo injected ventrally with 4 ng sog RNA, dorsolateral view. An additional domain of MyoD expression extends from the primary paraxial domain of expression (arrowhead). No midline clearing is present in this secondary axial expression. (G,H) Xwnt-8 expression in mid-neurula-stage embryos. Dorsal is up. (G) Uninjected embryo, dorsoposterior view. Xwnt-8 expression is found ventrally and laterally, and adjacent to the closed blastopore. (H) Embryos injected ventrally with 6 ng sog RNA, posterior view. Ventral Xwnt-8 expression has been eliminated and the lateral expression has been extended along the sides of the new secondary axis (2°). The primary axis is up (1°).
Injection of sog RNA alters early gene expression. (A,B) Vegetal view of gsc expression in stage 10 embryos. Dorsal is up. (A) Uninjected embryo. (B) Embryo injected ventrally with 6 ng sog RNA. Gsc expression is only weakly enhanced on the ventral side. (C,D) Vegetal view of Xnot expression in stage 11 embryos. Dorsal is up. (C) Uninjected embryo. (D) Embryo injected ventrally with 6 ng sog RNA. Xnot expression is enhanced on the ventral side, however only in the non-notochordal domain of expression. (E,F) MyoD expression in mid-neurula-stage embryos. Anterior is up. (E) Uninjected embryo, dorsal view. (F) Embryo injected ventrally with 4 ng sog RNA, dorsolateral view. An additional domain of MyoD expression extends from the primary paraxial domain of expression (arrowhead). No midline clearing is present in this secondary axial expression. (G,H) Xwnt-8 expression in mid-neurula-stage embryos. Dorsal is up. (G) Uninjected embryo, dorsoposterior view. Xwnt-8 expression is found ventrally and laterally, and adjacent to the closed blastopore. (H) Embryos injected ventrally with 6 ng sog RNA, posterior view. Ventral Xwnt-8 expression has been eliminated and the lateral expression has been extended along the sides of the new secondary axis (2°). The primary axis is up (1°).
In contrast to the effects of ventral sog RNA injections on gsc and Xnot expression, MyoD expression was greatly altered. MyoD is normally expressed in the paraxial mesoderm along both sides of the dorsal midline or presumptive notochord in neurula-stage embryos (Fig. 2E; Frank and Harland, 1991). Ectopic expression of sog RNA on the ventral side of Xenopus embryos resulted in a new region of MyoD expression (Fig. 2F; 71%, n=14). The extent and position of the new region of MyoD expression was variable, likely presaging the varied secondary axes observed in later-stage embryos (Fig. 1). In addition, the ectopic MyoD site did not have an axial midline clearing as is characteristic of normal MyoD expression (Fig. 2E). The expression of the ventral-lateral marker, Xwnt-8, was also altered by the ventral expression of sog. Xwnt-8 is expressed during neurulation in the ventral and lateral mesoderm (Christian et al., 1991; Smith and Harland, 1991), and in two regions flanking the closed blastopore (Fig. 2G). Ventral injections of sog RNA eliminated the ventral expression of Xwnt-8 and extended the lateral expression along the induced secondary clearing (54% had no ventral Xwnt-8 expression, n=13). The two strong sites of expression flanking the blastopore remained (Fig. 2H). These results are similar to those observed after injection of RNA encoding a truncated BMP receptor which leads to ventral expression of MyoD in neurula-stage embryos and the elimination of Xwnt-8 expression (Schmidt et al., 1995).
One characteristic of the secondary axis formed as a result of blocking BMP-4 signaling with a dominant negative BMP receptor is the absence of notochord in the ectopic axis (Graff et al., 1994; Suzuki et al., 1994). The absence of axial midline clearing of MyoD expression in secondary axes induced by ventral injection of sog RNA suggests that the secondary axes formed by sog RNA injections lack notochords as well. In agreement with this deduction, the secondary axes of embryos injected ventrally with sog RNA did not express Xbra (Fig. 3B; none had axial expression, n=23), which is normally expressed in the notochord of late gastrula and neurula-stage embryos and in a posterior region surrounding the blastopore (Fig. 3A; Smith et al., 1991). Xbra expression, however, was clearly present in the primary axis of sog RNA-injected embryos (Fig. 3B). We also examined histological sections of stage 38 embryos that developed secondary axes as a result of ventral sog RNA injections. While the primary axis had a clearly visible vacuolated notochord, notochord cells were not observed in the secondary axis in any of the sections (Fig. 3C,D and data not shown; n=8). The secondary axis included neural tissue and muscle cells that either fused with the somites on one side of the primary axis (Fig. 3C), or were distinct from the primary axis (Fig. 3D). These results indicate that sog can induce a partial secondary axis by changing the fate of lateral or ventral mesoderm to a more dorsal fate.
Secondary axes induced by sog RNA lack notochords. (A) Posterior view of Xbra expression in an uninjected mid-neurula-stage embryo. Xbra is expressed both posteriorly, surrounding the closed blastopore, and in the presumptive notochord. Dorsal is up. (B) Posterior view of a mid-neurula-stage embryo injected ventrally with 6 ng sog RNA. Posterior Xbra expression remains (although slightly altered) and the notochord expression along the primary axis (1°) is present. No axial expression of Xbra is seen in the secondary axis (2°). Dorsal is up. (C) Transverse section near the tail of a stage 38 embryo (Fig. 1, third embryo from the top). The vacuolated notochord (no) is visible in the primary axis, as are the somites (s) and neural tube (ne). The secondary axis (to the lower right) has no notochord, but instead contains neural tissue (ne) as well as somitic tissue (s) that is fused across the midline of the secondary axis, and with the somitic tissue on one side of the primary axis (arrowhead). (D) Transverse section of a stage 38 embryo in which the secondary axis is opposite and distinct from the primary axis (embryo similar to Fig. 1, top embryo). The secondary axis is on the ventral side of the embryo (bottom of picture), and does not contain a notochord. Labeling as in C.
Secondary axes induced by sog RNA lack notochords. (A) Posterior view of Xbra expression in an uninjected mid-neurula-stage embryo. Xbra is expressed both posteriorly, surrounding the closed blastopore, and in the presumptive notochord. Dorsal is up. (B) Posterior view of a mid-neurula-stage embryo injected ventrally with 6 ng sog RNA. Posterior Xbra expression remains (although slightly altered) and the notochord expression along the primary axis (1°) is present. No axial expression of Xbra is seen in the secondary axis (2°). Dorsal is up. (C) Transverse section near the tail of a stage 38 embryo (Fig. 1, third embryo from the top). The vacuolated notochord (no) is visible in the primary axis, as are the somites (s) and neural tube (ne). The secondary axis (to the lower right) has no notochord, but instead contains neural tissue (ne) as well as somitic tissue (s) that is fused across the midline of the secondary axis, and with the somitic tissue on one side of the primary axis (arrowhead). (D) Transverse section of a stage 38 embryo in which the secondary axis is opposite and distinct from the primary axis (embryo similar to Fig. 1, top embryo). The secondary axis is on the ventral side of the embryo (bottom of picture), and does not contain a notochord. Labeling as in C.
sog partially rescues axis formation in uv-irradiated embryos
Irradiation of Xenopus embryos with ultraviolet (uv) light during the first cell cycle inhibits the formation of the dorsal organizer, producing embryos that lack all dorsal structures (Malacinski et al., 1977; Scharf and Gerhart, 1980). It is advantageous to study axis-inducing properties in these ventralized embryos, since the morphological constraints imposed by the primary axis often obscure the full properties of a secondary axis. Whereas the majority of uninjected, uv-irradiated embryos developed without any axial structures (Fig. 4A,E and data not shown; n=660, 7 experiments), which is scored as zero on the dorsoanterior index (DAI; Kao and Elinson, 1988), uvirradiated embryos injected with sog RNA were typically rescued to a maximum DAI value of 2 (Fig. 4A,C and data not shown; n=329, 7 experiments), although rare embryos with a higher DAI value were observed. In contrast, wild-type embryos (Fig. 4B) are scored as having a DAI value of 5. These results are very similar to those observed previously with the injection of RNA encoding a truncated BMP receptor into uvirradiated embryos (Graff et al., 1994).
sog partially rescues uv-irradiated embryos. (A) DAI scores obtained in uninjected (top), sog RNA-injected (middle), and ea-sog-injected (bottom) embryos. Each chart shows the range of values between completely ventralized (DAI=0) and normal (DAI=5) embryos, and the percentage of embryos with each DAI value. (B-D) Stage 34 embryos, anterior is to the left. (B) Uninjected embryo. (C). Uv-irradiated embryo injected with 4 ng wild-type sog RNA (DAI=2). (D) Uv-irradiated embryo injected with 4ng ea-sog (DAI=2). (E) Uninjected, uv-irradiated embryo (DAI=0). Closed blastopore is to the right.
sog partially rescues uv-irradiated embryos. (A) DAI scores obtained in uninjected (top), sog RNA-injected (middle), and ea-sog-injected (bottom) embryos. Each chart shows the range of values between completely ventralized (DAI=0) and normal (DAI=5) embryos, and the percentage of embryos with each DAI value. (B-D) Stage 34 embryos, anterior is to the left. (B) Uninjected embryo. (C). Uv-irradiated embryo injected with 4 ng wild-type sog RNA (DAI=2). (D) Uv-irradiated embryo injected with 4ng ea-sog (DAI=2). (E) Uninjected, uv-irradiated embryo (DAI=0). Closed blastopore is to the right.
Sog has a large hydrophobic region at its amino terminus, which could function to anchor it to the outside of cells where it may be released by proteolysis (Francois et al., 1994). To test whether this region is essential for sog function, we removed the hydrophobic region and substituted the secretory peptide from the Drosophila easter gene (Chasen and Anderson, 1989), producing ea-sog. Injection of ea-sog RNA into uv-irradiated embryos caused a partial rescue of the dorsal axis (Fig. 4D and data not shown; n=177, 4 experiments). At a variety of doses tested, ea-sog RNA was equally as effective as wild-type sog RNA at producing an ectopic axis (Fig. 4A,C,D and data not shown), indicating that the amino-terminal hydrophobic region is not necessary for sog function in Xenopus embryos.
Since uv-irradiated embryos rescued by sog RNA injection had, in some cases, a more extensive and apparently more complete body axis than the secondary axis induced by sog RNA in wild-type embryos (compare Fig. 1 with Fig. 4C,D and Fig. 5J,K), we asked whether they were able to form a notochord by examining the expression of MyoD and Xbra in neurulastage embryos. Whereas uv-irradiated embryos did not express MyoD (Fig. 5C; 6%, 1 embryo had very slight expression, n=17), consistent with their ventralized phenotype, ectopic sog expression activated MyoD expression (Fig. 5B) (100% had MyoD expression, n=13). However, there was no clearing along the dorsal midline as in control embryos that were not uv irradiated (Fig. 5A). In addition, we examined Xbra expression in uv-irradiated neurula-stage embryos injected with sog RNA. Uninjected uv-irradiated embryos maintained the posterior expression of Xbra but lacked notochordal Xbra expression (Fig. 5F; none had axial expression, n=10). When injected with sog RNA, uv-irradiated embryos retained the posterior Xbra expression and elongated to form an axis, but they did not have dorsal axial Xbra expression (Fig. 5E; none had axial expression, n=10) as was found in normal control embryos (Fig. 5D), indicating a lack of notochord in uv-irradiated, sog-injected embryos. This was confirmed by histological sections that showed fusion of the somites across the dorsal midline in these embryo (Fig. 6; n=11).
Gene expression in uv-irradiated embryos rescued by sog RNA injection. (A-C) MyoD expression in early neurula-stage embryos. (A) Uninjected embryo. (B) Uv-irradiated embryo injected with sog RNA. MyoD is expressed in the induced axis. There is no midline clearing of expression. (C) Uninjected, uv-irradiated embryo. (D-F) Xbra expression in mid-neurulastage embryos. (D) Uninjected embryo. (E) Uv-irradiated embryo injected with sog RNA. Posterior Xbra expression is present, however no notochordal Xbra is present in the induced axis. (F) Uninjected, uv-irradiated embryo. Xbra is expressed around the blastopore, but there is no axial Xbra expression. (G,H) Hairy II expression in mid-neurula-stage embryos. (G) Uninjected embryo. (H) Uv-irradiated embryo injected with sog RNA. Hairy II expression is found in a stripe around the neural plate, and demonstrates a reduction in the size of the anterior neural region. No dorsal midline expression of Hairy II is observed. (I) Uninjected, uv-irradiated embryo. In the above panels, the uninjected and sog RNA-injected embryos are shown in a dorsal view with anterior up, whereas the uninjected, uv-irradiated embryos are shown in a posterior view. (J) Krox-20 expression in a stage 27 control embryo (top embryo; black arrowhead marks the two stripes of expression in rhombomeres 3 and 5) and in a uv-irradiated embryo injected with sog RNA (bottom embryo). In this embryo, both stripes of Krox-20 expression were observed (white arrowhead). (K) Cement gland was present in some uv-irradiated embryos injected with sog RNA (bottom embryo, black arrowhead). In this embryo, only a single stripe of Krox-20 expression was observed.
Gene expression in uv-irradiated embryos rescued by sog RNA injection. (A-C) MyoD expression in early neurula-stage embryos. (A) Uninjected embryo. (B) Uv-irradiated embryo injected with sog RNA. MyoD is expressed in the induced axis. There is no midline clearing of expression. (C) Uninjected, uv-irradiated embryo. (D-F) Xbra expression in mid-neurulastage embryos. (D) Uninjected embryo. (E) Uv-irradiated embryo injected with sog RNA. Posterior Xbra expression is present, however no notochordal Xbra is present in the induced axis. (F) Uninjected, uv-irradiated embryo. Xbra is expressed around the blastopore, but there is no axial Xbra expression. (G,H) Hairy II expression in mid-neurula-stage embryos. (G) Uninjected embryo. (H) Uv-irradiated embryo injected with sog RNA. Hairy II expression is found in a stripe around the neural plate, and demonstrates a reduction in the size of the anterior neural region. No dorsal midline expression of Hairy II is observed. (I) Uninjected, uv-irradiated embryo. In the above panels, the uninjected and sog RNA-injected embryos are shown in a dorsal view with anterior up, whereas the uninjected, uv-irradiated embryos are shown in a posterior view. (J) Krox-20 expression in a stage 27 control embryo (top embryo; black arrowhead marks the two stripes of expression in rhombomeres 3 and 5) and in a uv-irradiated embryo injected with sog RNA (bottom embryo). In this embryo, both stripes of Krox-20 expression were observed (white arrowhead). (K) Cement gland was present in some uv-irradiated embryos injected with sog RNA (bottom embryo, black arrowhead). In this embryo, only a single stripe of Krox-20 expression was observed.
uv-irradiated, sog-injected embryos do not contain a notochord. Transverse section through a stage 37/38 uv-irradiated embryo injected with sog RNA at the 8-cell stage. The somites (s) are fused across the dorsal midline beneath the neural tissue (ne).
uv-irradiated embryos rescued by sog RNA injections form melanocytes, indicating that partial rescue of neural tissue had occurred in the absence of notochord (not shown). To confirm this, we examined the expression of Hairy II (Turner and Weintraub, 1994), which borders the neural plate during neurulation, and is also expressed along the dorsal midline (Fig. 5G). Whereas uv irradiation eliminated Hairy II expression (Fig. 5I; none had Hairy II expression, n=17), ectopic injection of sog RNA into uv-irradiated embryos induced the expression of Hairy II (Fig. 5H; 100% had Hairy II expression, n=17). The expression of Hairy II bordering the neural plate was recovered, although the anterior region of expression was shortened and narrowed compared to normal embryos (Fig. 5G,H). However, the expression of Hairy II along the dorsal midline was absent. To determine whether anterior neural tissue such as hindbrain was present, we examined the expression of Krox-20 in uv-irradiated embryos injected with sog RNA. Krox-20 is expressed in rhombomeres 3 and 5 of untreated embryos (Bradley et al., 1992; Fig. 5J), and therefore is a useful marker for the extent of anterior-posterior axis formation. In uv-irradiated embryos injected with sog RNA, one or two stripes of Krox-20 expression were observed (Fig. 5J; 71%, n=28), demonstrating that hindbrain was present in these embryos. Intriguingly, cement gland, which is presumed to be a more anterior structure than the eye, was observed in some experiments, despite the absence of a complete head (Fig. 5K; 54% had cement glands in the experiment shown here, n=28).
Ubiquitous sog expression dorsalizes Xenopus embryos
In normal embryos, the neural region is confined to the dorsal side of the embryo. In our previous study, we expanded the expression of the neural marker, Hairy II, around the entire circumference of the embryo by injecting a truncated BMP receptor in each cell of 4-cell embryos (Schmidt et al., 1995). This treatment also leads to the circumferential expression of MyoD, which is normally found only on the dorsal side of neurula-stage embryos in the paraxial mesoderm (Fig. 7A; 100% of normal embryos had only dorsal expression of MyoD, n=17; Frank and Harland, 1991; Schmidt et al., 1995). Injection of sog RNA into each blastomere at the 4-cell stage expanded the expression of MyoD to the ventral side of neurula-stage embryos, forming a thick continuous band of expression in the posterior of the embryo (Fig. 7B; 92% had ventral expression of MyoD, n=12). Sections of the anterior region of these embryos showed normal paraxial expression of MyoD (Fig. 7C), whereas posteriorly, MyoD was expressed in a thick layer, circumscribing the dorsal-ventral axis except for the dorsal midline (Fig. 7D). These results indicate that the entire posterior mesoderm, with the exception of the notochord at the dorsal midline, was converted to a uniform paraxial fate. Hairy II expression borders the neural plate in normal neurula-stage embryos (Fig. 7F; none had ventral expression, n=14), but this stripe was found to encircle sog RNA-injected embryos, continuing around to the ventral side of the embryo (Fig. 7G,H; 67% had ventral expression, n=12). The ventral expression of Hairy II was confirmed by sectioning these embryos (Fig. 7I). The change in the expression patterns of these two genes appeared identical to those changes seen in embryos injected with a truncated BMP receptor (Schmidt et al., 1995).
Ubiquitous expression of sog dorsalizes the embryo. (A,B) Lateral view of MyoD expression at the mid-neurula stage in (A) an uninjected embryo and (B) an embryo injected with sog RNA in each blastomere at the 4-cell stage. Anterior is to the left. MyoD expression extends around the entire ventroposterior region of the injected embryo. (C,D) Transverse section of an embryo similar to the one in B. (C) Anterior section. MyoD is expressed in the paraxial mesoderm flanking the presumptive notochord (no). (D) Posterior section. MyoD is expressed circumferentially, except in the dorsal midline, the region of the presumptive notochord (no). (E) Dorsal view of late neurula-stage embryos. Top embryo is an uninjected control embryo. Bottom three embryos were injected in each blastomere at the 4-cell stage with ea-sog RNA (identical results were observed with wild-type sog RNA). The posterior region of the embryo was considerably elongated and narrowed. Anterior is to the left. (F-H) Hairy II expression at the late neurula stage in an uninjected embryo (F) and in an embryo injected with sog RNA in each blastomere at the 4-cell stage (G,H). Dorsal view in F, lateral view in G, and ventral view in H; anterior is to the left in all three panels. The stripe of Hairy II expression bordering the neural plate in the injected embryo continues around to the ventral side (arrowhead). (I) Transverse section of a sog-injected embryo stained for HairyII as in G and H, cut at the level of the stripe that circumscribes the embryo. HairyII expression was found to continue around the ventral side of the embryo in the ectodermal layer of cells. Expression of Hairy II in the floorplate was retained (arrowhead). (J) Top: Control stage 38 embryo; bottom: stage 38 embryo injected with sog RNA (third embryo from the top in E). A normal ventral tail fin has not formed.
Ubiquitous expression of sog dorsalizes the embryo. (A,B) Lateral view of MyoD expression at the mid-neurula stage in (A) an uninjected embryo and (B) an embryo injected with sog RNA in each blastomere at the 4-cell stage. Anterior is to the left. MyoD expression extends around the entire ventroposterior region of the injected embryo. (C,D) Transverse section of an embryo similar to the one in B. (C) Anterior section. MyoD is expressed in the paraxial mesoderm flanking the presumptive notochord (no). (D) Posterior section. MyoD is expressed circumferentially, except in the dorsal midline, the region of the presumptive notochord (no). (E) Dorsal view of late neurula-stage embryos. Top embryo is an uninjected control embryo. Bottom three embryos were injected in each blastomere at the 4-cell stage with ea-sog RNA (identical results were observed with wild-type sog RNA). The posterior region of the embryo was considerably elongated and narrowed. Anterior is to the left. (F-H) Hairy II expression at the late neurula stage in an uninjected embryo (F) and in an embryo injected with sog RNA in each blastomere at the 4-cell stage (G,H). Dorsal view in F, lateral view in G, and ventral view in H; anterior is to the left in all three panels. The stripe of Hairy II expression bordering the neural plate in the injected embryo continues around to the ventral side (arrowhead). (I) Transverse section of a sog-injected embryo stained for HairyII as in G and H, cut at the level of the stripe that circumscribes the embryo. HairyII expression was found to continue around the ventral side of the embryo in the ectodermal layer of cells. Expression of Hairy II in the floorplate was retained (arrowhead). (J) Top: Control stage 38 embryo; bottom: stage 38 embryo injected with sog RNA (third embryo from the top in E). A normal ventral tail fin has not formed.
We also noticed that injection of sog RNA caused a striking morphological change in the shape of the embryo. Whereas injection into the ventral side causes some morphological perturbation of neurula-stage embryos (Fig. 2F), injection of sog RNA throughout the embryo caused the posterior region of the embryo to elongate and narrow considerably (Fig. 7E, lower three embryos) as compared to control embryos (Fig. 7E, upper embryo). Cells of the neuroectoderm and dorsal mesoderm, including presumptive muscle, converge and extend on the dorsal side of normal embryos, causing more extensive gastrulation movements on this side of the embryo (Keller et al., 1992). In embryos injected with sog RNA, convergence and extension may occur throughout the mesoderm and overlying ectoderm, resulting in the distorted shape of the embryos observed here. At later stages, several phenotypes were apparent including varying loss of ventral tail fin (Fig. 7J) and truncation of the posterior region of the embryo (not shown).
BMP-4 antagonizes the effects of sog
Our results suggested that sog functions to antagonize the effects of ventralizing factors such as BMP-4, since injection of sog RNA mirrored the results observed with truncated BMP receptors. We therefore asked whether BMP-4 could reverse axis induction by sog. To measure this quantitatively, we used the uv-irradiation assay described above. Uv irradiation of embryos eliminated dorsal axial structures (data not shown; average DAI=0.06, n=83), which was rescued by the injection of sog RNA (Fig. 8, middle embryo; average DAI=1.7, n=18). Co-injection of RNA encoding sog and varying doses of RNA encoding BMP-4 into uv-irradiated embryos completely eliminated the ectopic axis induced by sog with as low as 300 pg of Bmp-4 RNA (Fig. 8 bottom embryo; at 300 pg of Bmp-4 RNA the average DAI=0.0, n=13). A dose of 100 pg of BMP-4 RNA was no longer as effective at eliminating the partial axis induced by sog (data not shown). These results demonstrate that ectopic expression of low levels of BMP-4 can antagonize the effects of sog.
BMP-4 prevents the axis-inducing capabilities of sog in uv-irradiated embryos. Top: uninjected embryo; middle: uv-irradiated embryo injected with 5 ng sog RNA (DAI=2); bottom: uv-irradiated embryo co-injected with 5 ng sog RNA and 300 pg Bmp-4 RNA (DAI=0). Uninjected, uv-irradiated embryos were similar to the bottom injected embryo and had DAI values of 0 (data not shown). All embryos are at the equivalent of stage 37.
BMP-4 prevents the axis-inducing capabilities of sog in uv-irradiated embryos. Top: uninjected embryo; middle: uv-irradiated embryo injected with 5 ng sog RNA (DAI=2); bottom: uv-irradiated embryo co-injected with 5 ng sog RNA and 300 pg Bmp-4 RNA (DAI=0). Uninjected, uv-irradiated embryos were similar to the bottom injected embryo and had DAI values of 0 (data not shown). All embryos are at the equivalent of stage 37.
DISCUSSION
Drosophila sog dorsalizes Xenopus embryos
Our previous studies with BMP-4 and a truncated BMP receptor suggested that the role of BMP-4 in Xenopus may be analogous to that of the Drosophila dpp gene product in dorsal/ventral patterning of the early embryo. To further investigate the conservation of patterning mechanisms between these species, we examined the functional capabilities of the Drosophila sog protein in Xenopus embryos, since genetic evidence indicates that sog functions to antagonize dpp (Ferguson and Anderson, 1992a; Wharton et al., 1993; Francois et al., 1994).
Our results demonstrate that sog can function in Xenopus embryos, and suggest that sog may function in part by opposing BMP-4 signaling. Ectopic expression of sog through-out the marginal zone dorsalizes the embryo, causing circumferential expression of the neural plate marker, Hairy II, and the dorsal/paraxial mesoderm marker, MyoD. The phenotype of tadpole-stage embryos is consistent with a loss of ventral structures. In addition, the unusual elongated shape of neurulastage embryos injected with sog RNA indicates that convergence and extension circumscribes the entire posterior region of the embryo, whereas in normal embryos it occurs predominantly on the dorsal side. These results mirror those seen in embryos injected with RNA encoding a truncated BMP receptor in which neural and paraxial tissue extend around to the ventral side of the neurula-stage embryo, and very similar morphological changes are observed (Schmidt et al., 1995).
Injection of sog RNA into the ventral marginal zone at the 4-cell stage produced embryos with partially duplicated axes. Like the results obtained with a dominant negative BMP receptor, these secondary axes contained muscle but not notochord (Graff et al., 1994; Suzuki et al., 1994), at both early and late stages of development. Similarly, sog did not significantly induce the expression of the organizer-specific genes, gsc and Xnot. The ventral and lateral marker, Xwnt-8, shown previously to be eliminated by the overexpression of a truncated BMP receptor (Schmidt et al., 1995), was also eliminated in the ventral region of embryos ventrally injected with sog RNA. Furthermore, Graff et al. (1994) have shown that a truncated BMP receptor is able to partially rescue uv-irradiated ventralized embryos, yielding tadpoles containing muscle but no notochord. Otic vesicles, which form adjacent to rhombomere 4, were also present in these embryos, indicating the partial rescue of head structures. These results are very similar to those obtained here. Uv-irradiated embryos injected with sog RNA were rescued up to a DAI of 2, which corresponds to embryos with normal bodies and little or no obvious head structure. Muscle and neural tissue formed along the body axis in these embryos, but the notochord was lacking, as shown by the absence of dorsal midline clearing of MyoD expression, the lack of Xbra and Hairy II expression along the axis, and the absence of vacuolated notochord cells in histological sections. Krox-20 was expressed in the anterior region in one or two stripes in some embryos, indicating that hindbrain had formed as far anteriorly as rhombomere 3, but the anterior neural plate, as revealed by Hairy II expression, was reduced. Interestingly, cement gland was also observed in the anterior end of some embryos, demonstrating that cement gland induction can occur without the formation of complete heads. In summary, the similarities between the results obtained with sog and truncated BMP receptors suggests that sog may function in part by antagonizing BMP-4 signaling. This hypothesis is supported by our observation that a low dose of exogenous BMP-4 is able to prevent the induction of an ectopic axis in uv-irradiated embryos injected with sog RNA.
chordin is likely to be the Xenopus homolog of sog
In a search for dorsally expressed genes activated by goosecoid, Sasai et al. (1994) isolated an RNA encoding a protein that contains a potential secretory signal sequence. This factor, which they named chordin, is able to induce a secondary axis in normal embryos, and to rescue uv-irradiated embryos when ectopically expressed. chordin has a modest overall homology to sog (27% amino acid identity) and it contains the four Cysrich (CR) regions present in sog (Francois and Bier, 1995). Within these regions, sog and chordin share up to 47% amino acid identity, and all of the Cys residues are conserved. In addition, the relative positions of the CR repeats in these two molecules have been strictly maintained, and each CR of sog is most related to the corresponding CR of chordin, suggesting that the Cys-rich regions have an important role in the function of these two proteins. These domains are similar to a domain found in procollagen, thrombospondin and other proteins. Although a peptide from the procollagen domain in thrombospondin has been shown to block angiogenesis in vivo and in vitro (Tolsma et al., 1993), suggesting that this region has a role in intercellular signaling, this domain remains poorly understood. The only region that is distinctly different between sog and chordin is the hydrophobic region at the amino terminus of sog that might function as a membrane anchor (Francois et al., 1994). This region is absent in chordin. Our results demonstrate that this region neither enhances nor reduces the axis-inducing properties of sog in Xenopus embryos. We are currently testing whether it has an important role in Drosophila.
Following the submission of our work, another group reported that injection of sog RNA into Xenopus embryos causes axis duplication and rescue from uv-irradiation (Holley et al., 1995). Unlike our findings, they observed the formation of notochords in sog-injected embryos, as was observed with injection of chordin RNA (Sasai et al., 1994). These results strengthen the hypothesis that sog and chordin are homologs.
Dorsal-ventral patterning in flies and frogs
Previously, we suggested that BMP-4 in Xenopus has a similar role to that of dpp in Drosophila (Schmidt et al., 1995). Both are expressed opposite the neurogenic region (St. Johnston and Gelbart, 1987; Wharton et al., 1993; Fainsod et al., 1994; Schmidt et al., 1995) and elimination or overexpression of dpp in flies and Bmp-4 in frogs causes analogous changes in dorsalventral patterning (see Introduction for references). The data presented in this paper strongly support the view that basic mechanisms for establishing dorsal-ventral polarity are analogous in two evolutionarily distant species. Consistent with this view, sog and dpp are expressed in opposing regions of the Drosophila embryo and sog behaves as a genetic antagonist of dpp signaling in Drosophila (Ferguson and Anderson, 1992b; Wharton et al., 1993; Francois et al., 1994). Likewise, chordin is expressed during gastrulation in what appears to be a complementary domain of expression to that of Bmp-4 and appears to exert opposite effects to those of Bmp-4 on dorsal-ventral patterning in Xenopus (Sasai et al., 1994). Furthermore, Bmp-4 and dpp have been functionally conserved during evolution since human BMP-4 can substitute for dpp in patterning the dorsalventral axis in Drosophila (Padgett et al., 1993), and dpp is active in vertebrate bone morphogenesis assays (Sampath et al., 1993). The data presented here, and in Holley et al. (1995), provide complementary evidence that sog can at least partially substitute for chordin as a BMP-4 antagonist in Xenopus.
One twist to the evolutionary conservation of dorsal-ventral patterning is that the relative positions of the sog/chordin and dpp/Bmp-4 expression domains are inverted in flies relative to frogs, with sog expressed ventrally and chordin expressed dorsally. This apparent inversion is consistent with a longstanding speculation that the dorsal-ventral axis in vertebrates and invertebrates have been reversed during evolution (Geoffroy St. Hillaire, 1822; Arendt and Nubler-Jung, 1994), although this issue has not been resolved (Jefferies and Brown, 1995; Lacalli, 1995; Peterson, 1995). Our work and that of others demonstrates an analogous function of BMP-4/dpp and chordin/sog in two very divergent species. This suggests that factors such as these may together represent a molecular module that functions to pattern the neural and non-neural ectoderm and that the relationship between these molecules was preserved.
The conservation between at least two factors involved in dorsal-ventral patterning in flies and frogs suggests that other factors may also be involved. For example, in Drosophila, the TGF-β family member, screw, is required together with dpp to pattern the dorsal ectoderm (Arora et al., 1994). It will be interesting to see if additional factors like screw or other regulators of Drosophila dorsal-ventral patterning are required along with BMP-4 to pattern the early Xenopus embryo.
ACKNOWLEDGEMENTS
We wish to thank Sarah Pierce and George von Dassow for critical comments on the manuscript. We would also like to thank Glenda Froelick for her expert advice on sectioning. This work was supported by NIH grants to D. K. (RO1-HD 27262) and E. B. (RO1-NS 29870). J. S. is a recipient of a Howard Hughes predoctoral fellowship.