Expression cloning of a Xenopus T-related gene (Xombi) involved in mesodermal patterning and blastopore lip formation

In amphibians, as in many other vertebrates, both maternal and zygotic proteins are required to establish the basic body plan. During the first cell cycle in Xenopus, the thin cortex of the egg rotates approx. 30° relative to the dense vegetal cytoplasm, and maternal dorsal determinants become localized or activated in the region of the embryo where vegetal cortex mixes with animal cytoplasm. These maternal determinants initiate a series of cell-autonomous and inductive interactions that, by early blastula stages, lead to the formation of a group of dorsovegetal cells known as the Nieuwkoop center (Kimelman et al., 1992). Prior to the onset of zygotic transcription at the mid-blastula transition, at least two different types of presumptive mesoderm tissues have been induced at the marginal zone between the embryo’s vegetal and animal hemispheres. Dorsal mesoderm has been induced by proteins secreted by the Nieuwkoop center, and ventral mesoderm has been induced everywhere else in the marginal zone by vegetal proteins that are active even in the absence of cortical rotation. In Xenopus, members of both the transforming growth factor beta (TGFβ) and fibroblast growth factor (FGF) families of secreted signaling proteins have been implicated in mesoderm induction (reviewed by Kimelman et al., 1992 and Slack, 1994). TGFβ family members such as activinβA and activin βB are found in the early Xenopus embryo, and are capable of inducing dorsal mesoderm. FGF family members such as embryonic FGF (eFGF) and basic FGF (bFGF) are also found in the early embryo; in contrast to TGFβ members, they are only capable of inducing mesoderm of more ventral character. The primary pattern of mesodermal gene expression established during blastula stages, when the embryo forms a hollow ball of cells, is extensively refined during gastrula stages by interactions between zygotically-expressed proteins released by opposing dorsal and ventral organizing centers, and by rec-iprocal interactions within the rest of the mesoderm (Gerhart et al., 1991; Kimelman et al., 1992; Harland, 1994).

Gastrulation is the highly dynamic morphogenetic process by which the three germ layers, initially present on or near the surface of the embyro become repositioned, with mesoderm placed between exterior ectoderm and interior endoderm. In Xenopus, the movement of mesodermal and endodermal cells into the interior of the embryo occurs through a dynamic and transient morphogenetic structure known as the blastopore lip. Little is known of the molecular signals that direct the formation of the blastopore lip, although the changes in cell behavior have been well-studied (for review see Keller, 1991). The first visible sign of lip formation is the apical constriction of pigment in superficial dorso-vegetal epithelial cells located above the Nieuwkoop center, signifying the reduction of apical surface area as cells take on the typical appearance of bottle cells. At the onset of gastrulation, these dorsal bottle cells descend into the embryo, allowing anterior mesoderm to migrate into the interior of the embryo and initiating dorsal blastopore lip formation (Hardin and Keller, 1988). As gastrulation proceeds, bottle cells form laterally and ventrally, preceeding a dorsal to ventral wave of blastopore lip formation, and a dorsal to ventral wave of tissue involution. By the end of gastrulation, dorsal, lateral and ventral mesoderm has involuted through the circular blastopore lip to form the meso-dermal mantle between the ectoderm and the endoderm. The mesoderm formed during this process undergoes complex mor-phogenetic movements that ultimately establish the elongated and bilaterally symmetric body axis. In Xenopus, the blastopore ultimately forms the posterior end of the embryo’s digestive tract.

In the mesoderm, tissue-specific growth factors induce expression of a variety of transcription factors, which in turn induce expression of a different subset of growth factors in responding tissues. Understanding the interplay between growth factor signaling and tissue-specific transcription has been a major focus of vertebrate developmental biology. One of the better-characterized transcription factors involved in mesoderm specification is brachyury (T), a member of the T-box family of DNA-binding proteins (Smith et al., 1991). Homologs of brachyury are found in all chordates and appear to be involved in the differentiation of notochord, a dorsal mesoderm derivative, in all cases where they have been examined (Yasuo et al., 1995). In Xenopus, brachyury is induced by TGFβ and FGF family members, is expressed during gastrula stages throughout the mesoderm as well as in cells fated to form the notochord, and appears to be part of the pathway by which ventral mesoderm is induced (Smith et al., 1991, 1995).

It is at present unclear whether mesoderm itself plays a role in directing blastopore lip formation, although mesoderm-inducing proteins such as brachyury have been reported to affect cell movements around the blastopore (Isaacs et al., 1994). Here we report the expression cloning of a new member of the T-box family of transcription factors (Xombi) that, unlike brachyury, induces ectopic structures that resemble the blastopore lip, as well as both ventral and dorsal mesoderm. Xombi is expressed prior to brachyury in a dorsal to ventral wave in the mesoderm, preceding a similar wave of blastopore lip formation at the border between the mesoderm and endoderm. In addition to promoting morphogenesis, Xombi may participate in a TGFβ-inducible, FGF-dependent positive feedback loop involving brachyury that serves to refine meso-dermal pattern early in gastrulation. In contrast to brachyury, Xombi is expressed maternally at high levels in the oocyte’s vegetal cortex and cytoplasm, and thus may also play a role in the earliest signaling events that induce endoderm and mesoderm and establish dorsal-ventral polarity.

Pigmented and albino X. laevis embryos were obtained by in vitro fer-tilization, dejellied and cultured at 18°-19°C in 0.1× Marc’s modified Ringers (MMR) containing 50 μg/ml gentamicin (Peng, 1991). Embryos were ventralized to a dorsoanterior index (DAI) of 0-1, as previously described (Kao and Danilchik, 1991), by briefly exposing (40 to 50 seconds) their vegetal surface to UV light about 30 minutes after fertilization. Embryos were dorsalized to a DAI of 9-10, as pre-viously described (Kao and Danilchik, 1991), by treating them for 10 minutes with 0.3 M lithium chloride when they reached the 64-to 128-cell stage. Staging of embryos was carried out according to Nieuwkoop and Faber (1967).

Ectodermal explants were cut from the apex of the animal pole of mid-blastula stage embryos (stage 8.5-9) using an electric cutting tool (Xenotek Engineering, Belleville, IL). The approx. 0.16 mm² square explants were cultured in 1× MBSH (pH 7.4; Peng, 1991) and changes in their developmental fate were evaluated by using reverse tran-scriptase-PCR (RT-PCR) to examine marker gene expression.

The functional assay used to identify Xombi is similar to the assay used by LeMaire et al. (1995) to identify the homeobox-containing protein siamois. Library RNA was injected into the ventral side of an early embryo, and the embryo was visually scored later in development. Unlike Lemaire et al. (1995), who injected complete library RNA or RNA pools representing 100,000 clones into the vegetal region of a ventral blastomere and then scored for the formation of a secondary body axis, we injected small pools of library RNA (100 to 1000 clones) into either the vegetal or animal pole of a ventral blas-tomere and then scored relatively early in development for any visible change in embryo morphology, including but not limited to secondary axes. By significantly enhancing the likelihood of proteins with moderate or weak activity being detected, a small pool screening approach may allow this type of synthetic screen in Xenopus to reach saturation. Although it is labor intensive to prepare and screen small pools, the increased effort seems compensated for by the increased frequency with which active cDNAs are functionally identified.

A cDNA library was constructed in plasmid pCS2+ (Turner and Weintraub, 1994) from gastrula stage embryos (stage 10 to 10.5) that had been hyperdorsalized to a dorsoanterior index (DAI) of 8-9 by LiCl treatment (Lustig and Kirschner, 1995). Aliquots from the library (containing between 100 and 1000 independent transformants) were used to amplify plasmid DNA, which in turn was used to synthesize pools of GpppG-capped RNA in vitro (Wormington, 1991; Lustig et al., 1996). Approximately 2 ng of library RNA was injected into the animal or vegetal region of a single ventral blastomere of a four-cell embryo. Ventral blastomeres were distinguished from dorsal blastomeres based upon their larger size and darker pigmentation and the presence of a sperm pigment trail. Embryos were visually scored for alterations in cell fate at specified times during gastrula (stage 10-12), neurula (stage 17-19) and tailbud (stages 30-32) stages of early development. To isolate the active clone in each positive pool, the pool was subdivided into smaller pools (Lustig et al., 1996), which were again re-tested for activity. Positive pools were progressively subdivided in this manner until a single active cDNA was obtained.

The first active Xombi cDNA clone that we isolated (approx. 1.5 kb in length) contained the entire T-box DNA-binding domain but, as no upstream stop codons were found, appeared to lack the 5′ end of the open reading frame. To clone the full-length coding sequence, we first used PCR to identify library pools that contained the 3′ end of the Xombi gene, and then determined which of these pools contained the longest Xombi clone. Ninety six pools of library cDNA, each con-taining 1000 cDNAs, were assayed by PCR for Xombi. A 650 nt region within the open reading frame was amplified (forward primer: 5′ TTC CAG AAG CTC AAA CTC AC 3′; reverse primer: 5′ TGT GTT GGA ATG ACA TGA AG 3′). The twelve positive pools that contained Xombi were then re-amplified by PCR using a forward primer that hybridizes to the SP6 promoter in the plasmid and a reverse primer that hybridizes about approx. 90 nt downstream of the 5′ end of the partial Xombi clone. Products as large as approx. 450 nt were amplified from several of the pools (of 1000 clones). Sequence analysis indicated that 50 bases at the 3′ end of several of the PCR products exactly matched the 5′ end of the original Xombi clone, suggesting that we had amplified part or all of the missing 5′ sequence. Rather than amplifying the full-length Xombi cDNA by PCR (which can introduce unwanted mutations) and then subcloning the amplified product, we used PCR as an assay to progressively subdivide the original pool of 1000 clones to a single Xombi clone (Lustig and Kirschner, 1995). The new sequence extended the open reading frame 56 amino acids upstream to a potential translation start site methionine with several upstream stop codons. From these results we conclude that we have isolated the complete open reading frame of the Xombi cDNA. When RNA synthesized from the longer Xombi cDNA clone is injected into embryos, it has similar activities as RNA synthesized from the partial Xombi cDNA that was initially isolated and characterized in this study.

Expression of brachyury, Xwnt8, goosecoid and EF-1α in ectodermal explants or whole embryos was analyzed by RT-PCR, as previously described (Lustig and Kirschner, 1995; Lustig et al., 1996). The sequence of the forward Xombi primer was 5′ TGT CAG CTA TAC TGC ATA CC 3′ and the sequence of the reverse primer was 5′ TGC AAT CTT CCC AAC GCA C 3′. These primers amplify an approx. 250 nt region in the 3′ region of the Xombi gene.

Single and double whole-mount in situ hybridization of albino or pigmented embryos was carried out as described (Harland, 1991) using digoxigenin-11-UTP-labeled probes (Boehringer), except for the brachyury probe which was labeled with fluorescein-12-UTP. 4-nitroblue tetrazolium chloride (NBT), 5-bromo-4-chloro-3-indolyl-phosphate, 4-toluidine salt, (BCIP) or magenta-phosphate were used as substrates for the color reactions. Antisense probe for Xombi was synthesized with T3 RNA polymerase using EcoRI-linearized plasmid as template. Antisense probes for brachyury (Smith et al., 1991), goosecoid (Blumberg et al., 1991) and Xwnt8 (Christian et al., 1991) were synthesized as described previously. To localize Xombi transcripts, overstained embryos were embedded in JB-4 plastic resin and cut into 7 μm sections.

Most images were obtained using incident light or transillumination on a Zeiss axiophot microscope using 2.5× or 5× objectives. Some images were acquired on a Wild M8 stereomicroscope. All images were captured using a 3-color video rate CCD camera con-trolled by Northern Exposure software (Phase 3 Imaging Systems). Brightness, contrast and color balance correction were performed using Adobe Photoshop.

We have carried out a simple visual screen for genes that are capable of respecifying embryonic tissue. In order to identify genes with relatively weak activity, small pools of library RNA (representing only 100 to 1000 cDNA clones) were assayed for the ability to cause any gross change in morphogenesis or tissue specification when microinjected into one cell of a four-cell Xenopus embryo. Out of approx. 500 pools screened, we have thus far identified seven different genes capable of respecifying ventral tissue (Table 1). Four of these genes had been previously identified: siamois (Lemaire et al. 1995), noggin (Smith and Harland, 1992), Xwnt8 (Christian et al., 1991; Smith and Harland, 1991) and β-catenin (Guger and Gumbiner, 1995). Three others were not in the GenBank database as of September, 1996: a new member of the TGFβ family of nodal-related secreted proteins; a new member of the homeodomain family of DNA-binding proteins; and a new member of the T-box family of DNA-binding proteins. Because the T-box gene is more closely related to the Drosophila optomotor blind (omb)/Tbx2 subclass of T-box proteins than to the brachyury subclass, we have named it Xenopusoptomotor blind (Xombi). This report will focus on the functional characterization of Xombi’s unusual activities and expression pattern. The other two new genes will be the subjects of separate reports.

The phenotype of embryos injected ventrally with Xombi RNA is distinct from that of embryos injected ventrally with RNA encoding siamois (Lemaire et al., 1995), noggin (Smith and Harland, 1992), Xwnt8 (Smith and Harland, 1991) or β-catenin (Guger and Gumbiner, 1995). Vegetal injection of these latter transcripts induce an ectopic Nieuwkoop center and thus generate partial or complete secondary axes containing axial tissue (e.g. notochord, somites) often indistinguishable from axial tissue in the primary axis (not shown). By contrast, vegetal or animal injection of Xombi RNA does not induce secondary axis formation. Ectopic Xombi expression does, however, induce an ectopic site of cell invagination (Fig. 1) that forms about the same time as the endogenous blastopore lip. Invaginations are detected when Xombi transcripts are microinjected into the marginal zone or animal pole of the embryo. Vegetal expression does not induce invagination within the deep endoderm but does occasionally lead to premature pigment accumulation at a ventral site within the endogenous blastopore lip (not shown).

The formation of the ectopic invagination site was preceded by changes in cell behavior similar to those observed at the endogenous blastopore lip. In both, the accumulation of apical pigment in scattered superficial epithelial cells first demarks the future site of invagination. This is followed soon after by the formation of a line of pigment at the ectopic site that, although wider and darker, is similar in appearance to the blastopore pigment line in the endogen-ous lip (Fig. 1A). By stage 10.25, when bottle cells in the endogenous lip have invaginated into the embryo to create the blastoporal groove (Hardin and Keller, 1988), invagination is also frequently observed within the ectopic pigment line. The greater the amount of Xombi RNA injected into the embryo, the greater the size of the invagination and the extent of cell movement through it. By the end of gastrulation, the ectopic invagination site is highly pigmented but otherwise similar in appearance to the endogenous blasto-pore lip (Fig. 1B). Most embryos injected ventrally with Xombi transcripts form completely normal head structures by tailbud stages but have severe posterior abnormalities, exhibiting substantial enlargement of the proctodeum and extensive apical pigment accumulation (not shown).

The ability of Xombi to induce ectopic invagination sites is not dependent upon dorsal signals that are activated by the cortical rotation during the first cell cycle. Visible invaginations form in 85% of UV-ventralized embryos injected with Xombi RNA (10, 50 or 125 pg) in an animal pole blastomere (n=55) but never in uninjected embryos (n=57) or embryos injected in a vegetal pole blastomere (n=59). Outwardly these invagination sites are indistinguishable from those formed in normal embryos, except that they form slightly later, by stage 10.5 to 11. Genes that induce Nieuwkoop center formation in normal embryos, including siamois, noggin, Xwnt8 and β-catenin, completely restore the dorsal-ventral body axis in UV-ventralized embryos. By contrast, Xombi expression only weakly rescues the body axis in ventralized embryos, with injected embryos typically having dorsoanterior indices (DAI; Kao and Danilchik, 1991) less than 2 (data not shown).

An examination of sagittal sections through ectopic blasto-pore lips reveals that they are often similar in appearance to endogenous blastopore lips (Fig. 1C-E). At the lowest active doses of Xombi, very little cell movement occurs, and the resulting invagination site is similar in appearance to the dorsal blastopore groove at stage 10+. With higher doses, extensive cell movements are observed at the ectopic site, with the invagination extending in multiple directions, with frequent internal bending (not shown).

Sequence analysis revealed that the 1938 nt cDNA insert had an open reading frame of 455 amino acids (Fig. 2A) with a predicted molecular mass of 52×10³. Comparison of the predicted protein sequence with the GenBank protein database revealed that Xombi is a member of the T-box family of DNA-binding proteins (Bollag et al., 1994). Members of the T-box family are highly related within a region of between 160 to 200 amino acids in length (the T-box) that is necessary and sufficient for DNA binding (Kispert and Hermann, 1993). Within its T-box, Xombi is most closely related (Fig. 2B) to mouse Tbx2 (57% identity; Bollag et al., 1994), Drosophila optomotor blind (54% identity; Pflugfelder et al., 1992) and Xenopus brachyury (49% identity; Smith et al., 1991). Outside its putative DNA-binding domain, Xombi shows no obvious homology to other members of the gene family. The motif Ser-Pro (S-P-X-X)_n, which is found in many DNA-binding proteins (Suzuki, 1989), is found distributed throughout the putative Xombi protein. There is also a consensus MAP kinase phosphorylation site (P-X-S/T-P; Gonzalez et al., 1991) within the putative DNA-binding domain.

A combination of RT-PCR and whole-mount in situ hybridization analysis was used to analyze the spatial and temporal expression of Xombi during early Xenopus development. High levels of maternal transcripts are detected by RT-PCR in unfertilized eggs and cleavage-stage embryos (Fig. 3). Transcripts are found in punctate structures in the vegetal cytoplasm of stage III oocytes (Fig. 4A) and in the cortex region just under the oocyte plasma membrane (Fig. 4B). Transcripts are also evenly distributed outside the cortex, throughout the vegetal pole (Fig. 4C). This expression pattern persists throughout the remainder of oogenesis (over weeks to months). Transcripts are found in the vegetal hemisphere and nucleus of the stage VI oocyte (Fig. 4D) and are also observed in the vegetal half of two-cell, eight-cell and 64-cell embryos (not shown).

Blastopore formation is a highly dynamic process, with the blastopore lip forming dorsally by stage 10+ to 10.25, extending laterally by stage 10.5 and extending ventrally by stages 11 to 11.5 (Nieuwkoop and Faber, 1967). The zygotic expression of Xombi within the marginal zone preceeds blasto-pore lip formation at its vegetal border. Zygotic transcripts are detected very soon after the mid-blastula transition (stage 9) in the nuclei of cells located throughout the marginal zone (not shown). About 90 minutes prior to the onset of gastrulation, a high level of Xombi transcript accumulates in dorsal marginal zone cells (Fig. 4E). When pigment first accumulates in the bottle cells that demark the future site of blastopore lip formation (stage 10−), Xombi transcripts are detected dorsally and laterally, and (in most but not all embryos) ventrally as well. By the time the dorsal blastoporal groove is visible (stage 10+), Xombi transcripts are detected at high levels throughout the entire circumference of the marginal zone (Fig. 4F).

Although Xombi transcripts are detected in marginal zone cells, they are not found in the anterior aspect of the forming dorsal lip or in the bottle cells themselves (Fig. 4G). This restriction is observed as early as stage 10, when the blasto-pore pigment line first become visible (Fig. 4F) and is observed around the entire lateral and ventral marginal zones by mid-gastrula stages (Fig. 4H). An examination of sections of mid-gastrula stage embryos stained in whole mount confirms that Xombi transcripts are excluded from a small region of cells directly adjacent to the blastopore lip (Fig. 4I) as well as from the superficial involuted cells that line the archenteron and are fated to form the lining of the digestive tract. In split-open embryos, involuted and non-involuted cells can be readily dis-tinguished and thus examined for the presence of Xombi transcripts. It does not appear that Xombi transcripts are present in invaginated bottle cells or deep cells that have undergone involution. They are only detected in superficial and deep cells of the marginal zone that have yet to involute (Fig. 4J).

Despite the fact that brachyury expression is maintained throughout the entire marginal zone during gastrulation, by the middle of gastrulation (stages 10.5 to 11) Xombi expression recedes from the dorsal lip region of the blastopore (Fig. 4G,H). This difference is apparent in mid-gastrula stage embryos that have been double stained by whole-mount in situ hybridization for Xombi and for brachyury, which at these stages is expressed in notochord-forming cells that have already involuted over the lip (Smith et al., 1991). Xombi transcripts and brachyury transcripts co-localize in the entire marginal zone (dark blue stain) with the exception of a narrow stripe of involuted dorsal cells (magenta stain) that express brachyury only (Fig. 4K).

The overall level of Xombi transcript in the embryo declines considerably at the end of gastrulation (Fig. 3), but Xombi tran-scripts are nevertheless detected at low levels until the tadpole stage (stage 38). During neurulation Xombi continues to be expressed in the ventral and lateral but not dorsal domains of the circumblastoporal collar (Fig. 4L), the region of the embryo that generates the bulk of the posterior mesoderm. During neurula and tailbud stages, faint diffuse staining is also detected in the posterior and anterior ends of the embryo (Fig. 4L,M).

By the early tailbud stages, Xombi transcripts become clustered in patches of cells found along the dorsal midline of the trunk (Fig. 4M). Between ten and twenty of these patches are typically observed, about half of which are paired sym-metrically across the dorsal midline. These patches of Xombi expression are detected at least until the tadpole stage of development, although it is not clear whether expression in any one persists over time. Sections of tailbud stage embryos localize Xombi expression to the region of the neural tube where interneurons are found, at its lateral edge above the midline (Fig. 4N).

UV treatment of fertilized eggs or immersion of 64-cell embryos in lithium chloride causes Xenopus embryos to take on uniformally ventralized or dorsalized fates, respectively (Kao and Danilchik, 1991). Both aberrant types of embryos still undergo bottle cell and blastopore formation and at least limited tissue involution. Xombi is expressed at high levels in both dorsalized and ventralized embryos (Fig. 5). In dorsalized embryos, Xombi transcripts are expressed around the entire marginal zone at least 90 minutes prior to the onset of gastru-lation (Stage 9.6; Fig. 5D), the same time they are detected in the dorsal marginal zone in normal embryos (Fig. 5A). In sibling ventralized embryos, Xombi is expressed in a punctate, nuclear pattern (Fig. 5G), the same pattern of expression detected concomitantly in the ventral marginal zone of normal embryos (Fig. 5A). Bottle cells form by stage 10.5 in dorsalized embryos but, as in normal embryos, Xombi transcripts are never detected in the anterior aspect of the forming lip or in the bottle cells themselves (Fig. 5E). By stage 10.5, Xombi transcripts are expressed at high levels around the entire marginal zone of ventralized embryos, although bottle cells have not yet formed (Fig. 5H). During all gastrula stages of development, Xombi staining is slightly less intense in ventralized embryos than in dorsalized embryos and is restricted to the marginal zone. By the beginning of neurulation (stage 13), Xombi expression drops considerably in normal (Fig. 5C) and dorsalized (Fig. 5F) embryos and is confined to a small region of cells surrounding the constricted blastopore. In ventralized embryos, Xombi expression is maintained at high levels even after the end of gastrulation (Fig. 5I).

To assess its effect on cell differentiation, Xombi was expressed in presumptive ectodermal explants otherwise fated to form epidermis. The explants were cultured until sibling embryos reached the mid to late gastrula stages, and then gene expression was analyzed by RT-PCR. Xombi induces the dose-dependent expression of three Xenopus genes that are zygotically expressed within the marginal zone (Fig. 6A): brachyury, a pan-mesodermal gene (Smith et al., 1991); Xwnt8, a ventro-lateral mesodermal gene encoding a secreted signaling protein (Christian et al., 1991; Smith and Harland, 1991); and goosecoid, a dorsal-specific mesodermal gene encoding a homeodomain protein (Blumberg et al., 1991). The lowest active dose of Xombi RNA (30 pg) induces brachyury expression and low level Xwnt8 expression, but not goosecoid expression. Higher RNA doses induce all three genes: brachyury expression is maximal at 100 pg; Xwnt8 expression is maximal at 300 pg and goosecoid expression is maximal at 1000 pg. None of the three mesodermal genes are expressed in control uninjected explants. We also left intact embryos that had been injected with Xombi RNA in both sides of the animal pole. Whole embryos ectopically expressing Xombi, express higher levels of brachyury, Xwnt8 and goosecoid than do control sibling embryos (not shown).

Xombi is thus capable of ectopically inducing mesoderm in explanted tissue and intact embryos. To determine where mesoderm forms relative to the ectopic invagination, normal or ventralized embryos were injected with Xombi RNA at the four-cell stage, harvested at early gastrula stages and stained by double in situ hybridization for Xwnt8, brachyury, or goosecoid, to identify mesodermal cells, and then for Xombi, to identify the site of RNA injection. In over 50 embryos analyzed, the region of ectopic expression of Xombi closely overlapped the region where ectopic mesoderm is induced (not shown). In the example shown in Fig. 6B, the expression of Xwnt8 occurs on the animal pole side of both the ectopic invagination and the endogenous blastopore lip. Similar to the endogenous blastopore lip, which defines the vegetal limit of the involuting mesoderm, ectopic invaginations induced by low doses of Xombi generally form at the vegetal border of the ectopic mesoderm.

RT-PCR analysis was used to detemine whether a number of known mesoderm-inducing proteins are capable of inducing Xombi expression in ectodermal explants. We found that Xombi expression is induced by members of both TGFβ and FGF signaling families and by brachyury, an inducer of ventral mesoderm. Treatment of ectodermal explants with as low as 0.3 ng/ml activin A protein or microinjection as low as 0.25 pg activinβB RNA or 50 pg Xnr1 RNA strongly induces Xombi expression (Fig. 7). Microinjection of 10 pg eFGF RNA induces Xombi expression in ectodermal explants, although even five-fold lower doses were generally ineffective and higher doses were toxic to the injected cells. Brachyury also induces the dose-dependent expression of Xombi (Fig. 7). A low level of Xombi transcript is reproducibly detected in control uninjected explants.

Not all of the signaling proteins that induce Xombi expression in ectodermal explants are also able to induce ectopic invaginations in intact embryos. Whereas the dorsal mesoderm inducers activinβB and Xnr1 induce both Xombi expression and the formation of ectopic invaginations, the ventral mesoderm inducers eFGF and brachyury induce Xombi expression but not invagination. Ectopic invaginations are induced by the same doses of activinβB and Xnr1 that induce Xombi expression, both in explants and in injected embryos assayed for Xombi expression by in situ hybridization (not shown). At the lowest effective doses of activinβB and Xnr1 RNA, the ectopic invaginations are similar in appearance to those induced by Xombi. At higher doses, the invagination sites are larger and more darkly pigmented than those induced by Xombi and the entire region of cells with constricted apical pigment invaginates to form a darkly pigmented pit within the animal pole (not shown).

In ectodermal explants, the induction of some (but not all) mesodermal markers by activin requires an intact FGF signaling pathway (Labonne and Whitman, 1994; Cornell and Kimelman, 1994). Thus, even though FGF itself does not induce ectopic invaginations, it seemed possible that FGF might play a role in mediating the changes in cell behavior that underly their formation. To address this possibility, transcripts encoding activinβB and Xombi were co-injected with transcripts encoding a dominant negative inhibitor of the FGF receptor (XFD), and embryos were scored at mid-gastrula stages for formation of ectopic invaginations. Expression of XFD, but not UHAVϕ, an inactive XFD construct missing just three amino acids in the ligand-binding domain (Amaya et al., 1991), strongly inhibited ectopic lip formation by activinβB and Xombi (Table 2), and partially inhibited ectopic lip formation by higher doses of activinβB (data not shown). ActivinβB RNA-injected embryos frequently formed a clearly defined partial secondary body axis, as previously shown (Thomsen et al., 1990), whereas Xombi RNA-injected embryos almost never did (Table 2).

In the intact Xenopus embryo, the FGF signalling pathway is required for normal gastrulation movements laterally and ventrally, though, surprisingly, not for gastrulation and subsequent head morphogenesis dorsally (Amaya et al., 1991, 1993; Isaacs et al., 1994). To determine whether Xombi expression in the presumptive mesoderm requires FGF signaling in vivo, we injected transcripts encoding XFD into the dorsal or ventral marginal zone and then examined Xombi and brachyury expression by double in situ hybridization analysis just prior to and during gastrulation. Since brachyury expression is inhibited anywhere in the marginal zone that XFD is expressed (Amaya et al., 1993), brachyury expression is used here as a measure of XFD function.

As shown in Fig. 8, a functional FGF signaling pathway is not required for initial expression of Xombi in the dorsal marginal zone, but is required for its continued expression during mid to late gastrula stages. Just prior to the onset of gastrulation (stage 10-), Xombi is expressed at high levels dorsally and throughout much of the marginal zone in uninjected embryos or embryos injected dorsally or ventrally with XFD RNA (magenta stain in Fig. 8A,D,G). At this time embryos have not yet begun to express detectable levels of brachyury (light blue stain). Later, when involution begins dorsally (stage 10.25) and ventrally (stage 11.5), brachyury expression is completely inhibited in the region of the mesoderm where XFD is expressed (dark blue stain is due to the overlap of Xombi and brachyury expression). Xombi expression is partially inhibited by XFD at stage 10.25 (Fig. 8B,E,H), and is almost completely inhibited at stage 11.5 (Fig. 8C,F,I).

Blastopore lip formation exhibits the same dependence on FGF signaling as does Xombi expression. Xombi expression and lip formation occur early in gastrulation at the dorsal side regardless of where XFD is expressed. During mid to late gastrula stages, however, XFD intereferes with the maintenance of Xombi expression and also appears to inhibit the process by which the dorsal blastopore lip extends laterally and ventrally. XFD-expressing embryos exhibit defects in lateral and ventral blastopore lip formation of varying severity; embryos injected dorsally generally have more extreme defects than embryos injected ventrally. The role of FGF in the extension of the blastopore lip is illustrated in Fig. 8F,I. In Fig. 8F, XFD has been injected near the dorsal lip (as denoted by the absence of brachyury expression); there is no FGF signaling in the tissue flanking either side of the lip and lateral extension of the lip is not observed. In Fig. 8I, FGF signaling occurs on one side of the dorsal blastopore lip, but not on the other. In this case, the lip extends laterally on the left side of the marginal zone, where there is intact FGF signaling, but not on the other side, where XFD is expressed.

We describe here the isolation and characterization of Xombi, a novel T-box gene, which has an expression pattern and bio-logical properties that reflect a possible role in blastopore lip formation. Xombi is expressed within the marginal zone soon after the mid-blastula transition in a dorsal to ventral wave of expression, preceding a similar wave of bottle cell and blastopore lip formation at the vegetal border of the marginal zone. Xombi expression also precedes blastopore lip formation in both dorsalized and ventralized embryos, where bottle cell formation and (subsequently) tissue involution occur simultaneously around the ring of marginal zone tissue, a strikingly different pattern than in normal embryos. When expressed ectopically in intact embryos, Xombi mimics the activity of the endogenous blastopore inducer, causing formation of an ectopic invagination that resembles the endogenous blastopore lip. This ectopic invagination frequently forms at the border of the Xombi injection site; the endogenous blastopore lip likewise forms at the border of Xombi expression in the marginal zone. Xombi is thus expressed at the correct time and place and appears sufficient to induce lip formation.

ActivinβB and Xnr1, two secreted TGFβ proteins expressed in the Xenopus embryo during gastrulation (Thomsen et al., 1990; Jones et al., 1995; Lustig et al., 1996), induce both Xombi expression and ectopic invagination, and thus may act upstream of Xombi in the blastopore lip induction pathway.

Ventral mesoderm inducers such as eFGF are also capable of inducing Xombi expression but, surprisingly, do not induce invagination. This may reflect the timing of Xombi induction, lower levels of Xombi induced or perhaps an inhibitory effect of ventral mesoderm that is concomitantly induced. Although FGFs do not induce ectopic invagination, FGF signaling is required for Xombi expression and ventrolateral blastopore lip formation in vivo. The induction of ectopic invagination by both Xombi and activinβB is also dependent upon FGF signaling. It is not yet clear when FGF signaling is required in vivo or whether this requirement reflects a direct role for FGF or an indirect role for FGF in the TGFβ signaling pathway.

Xombi injection recreates many of the characteristics of normal invagination. Ectopic expression of Xombi induces the formation of cells with similar constricted pigmentation as bottle cells, and induces changes in cell behavior that also precede bottle cell formation and involution at the endogenous lip (Hardin and Keller, 1988). These include the accumulation of pigment granules in the constricted apices of superficial epithelial cells, the fusion of the apical ends of these cells to form a pigment line and the invagination of superficial cells within this line to form a groove. Since it is not expressed in bottle cells, but only adjacent to them, Xombi may induce expression of genes encoding secreted proteins that act in a paracrine manner to induce these cell behaviors. Xombi expression in the marginal zone may first lead to bottle cell formation and the initial invagination of the blastopore lip, but subsequent expression of other signaling proteins in the epithelial layer of the organizer, such as the Wnt-inducible secreted protein Xnr3 (Smith et al., 1995), are probably required for the convergence and extension movements that provide the driving force for much of the later tissue involution.

Xombi is expressed in superficial and deep cells of the involuting marginal zone that are fated to form the lining of the archenteron and the mesodermal mantle, respectively. The absence of Xombi staining in involuted tissue and noninvoluted cells close to the lip, however, suggests that Xombi is expressed transiently as the cells move toward the blastopore lip, and is shut off as they near the inflection point. Thus, in addition to the dorsal to ventral traveling waves of Xombi expression and blastopore lip formation, there is a standing wave of Xombi expression that forms at right angles to the traveling waves (Fig. 9). A number of other mesodermal genes expressed during gastrulation, including Xwnt8 (Fig. 6B), also exhibit a similar standing wave of expression. A standing wave of Xombi expression is observed in dorsalized and ventralized embryos, suggesting that the mechanism of down regulation in cells near the lip is independent of normal dorsal-ventral patterning.

Although injection of a bolus of Xombi will induce bottle cell formation and ectopic invagination, in the context of the normal process of gastrulation there is a precise temporal pro-gression from the dorsal to the ventral side of both Xombi expression and lip formation. This progression suggests either inhibitory or activating signals propagating from the expanding site of involution. The strong effect of disrupting FGF signaling on the dorsal side, not on dorsal invagination, which proceeds normally, but on the lateral propagation of involu-tion, which is inhibited, provides evidence for a role for FGF in the propagation of this traveling wave. Propagating waves of morphogenesis are important in developing pattern in such systems as the insect eye (Heberlein et al., 1993; Ma et al., 1993). That Xombi expression is only dependent on FGF signaling during mid to late gastrula stages provides a possible explanation for the differential sensitivity of dorsal invagination and ventrolateral lip propagation to XFD inhibition. Thus, regardless of whether XFD is expressed dorsally or ventrally, Xombi is expressed dorsally and the dorsal blastopore lip forms normally. When FGF signaling is inhibited dorsally, however, Xombi is not expressed laterally during mid to late gastrula stages and gastrulation movements are inhibited.

The Xenopus genes that have been previously cloned using functional assays (Xwnt8, siamois, noggin and Xnr3) all rescue axial structures by ectopically inducing formation of a Nieuwkoop center (Christian et al., 1991; Lemaire et al., 1995; Smith and Harland, 1991, 1992; Smith et al., 1995). Xombi does not induce formation of a similar dorsalizing center but instead appears to act only as a general mesoderm inducer. Xombi transcripts are expressed maternally in the vegetal cortex and cytoplasm. Though the maternal function of these transcripts remains obscure, Xombi could play a role in early steps in mesoderm formation following the cortical rotation. When mesodermal genes begin to be expressed in the marginal zone, zygotic Xombi transcripts colocalize with transcripts of three mesodermal genes it ectopically induces: goosecoid, Xwnt8 and brachyury. Thus the time course and site of Xombi expression, its general mesoderm-inducing properties, its induction by activin A, activinβB, Xnr1 and eFGF in vitro, and its partial dependence on FGF signaling, are all consistent with a model in which Xombi expression is induced during late blastula stages by mesoderm-inducing signals from the Nieuwkoop center and then is maintained during gastrulation by vegetal signals or signals within the marginal zone itself. According to this view, the role of Xombi is not to mediate the induction of goosecoid, Xwnt8 or brachyury by primary mesoderm-inducing signals but rather to maintain and amplify the response to these primary signals.

The extent to which overlapping mesodermal genes within the marginal zone regulate each other’s expression remains unclear. The finding that brachyury and eFGF each induce the expression of the other has led to the proposal that the two genes form a regulatory circuit within the marginal zone and play a role in the maintenance of mesodermal gene expression (Isaacs et al., 1994). In the present study, we show that Xombi and brachyury induce each other’s expression, suggesting this putative positive feedback loop could involve Xombi as well. Xombi is expressed prior to brachyury on the dorsal side, and thus may initiate the loop.

Although Xombi and brachyury are both T-box-containing proteins and induce the expression of each other, they appear to carry out distinct functions during most of early Xenopus development. Unlike Xombi, brachyury is not expressed at high levels maternally, and outside of gastrula stages there is no significant overlap between Xombi and brachyury expression. At the neurula stage of development, Xombi is specifically excluded from notochord-forming cells along the dorsal midline, whereas brachyury is not only expressed in these cells but in fact is believed to be required for notochord formation in most vertebrates (Yasuo et al., 1995). Conversely, brachyury is not detected along the dorsal midline during tailbud stages of development, when Xombi is expressed in cells in a region of the trunk neural tube fated to form interneurons. Finally, ectopic expression of brachyury, unlike ectopic Xombi expression, does not lead to dorsal mesoderm induction or to changes in cell behavior that lead to the formation of ectopic invaginations.

The family of T genes may be derived from an original T gene involved in gut formation in the worm-like animals from which all living animals have evolved (Reuter, 1995; Yasuo et al., 1995). T genes are still involved in gut endoderm development in various phyla today; for example, a T-related gene (Trg) is expressed in the hindgut primordium in arthropods and is necessary for hindgut formation (Kispert et al., 1994). In more highly evolved organisms, T genes are also expressed in other embryonic structures, although many of these, such as the sto-mochord of hemichordates and the notochord of chordates, are believed to be phylogenetically derived from the gut endoderm (Reuter, 1995; Yasuo et al., 1995). A role for Xombi in gut development would not be surprising since recent work has shown that the gut endoderm in Xenopus is induced and patterned by many of the same genes that induce and pattern the mesoderm (Gamer and Wright, 1995; Henry et al., 1996). Although Xombi is not detected at high levels in the gut endoderm after gastrulation in Xenopus, several findings are consistent with the idea that it may be an early marker in the endodermal differentiation pathway. In normal development, Xombi transcripts are expressed in the vegetal pole region of the oocyte, which is fated to form the bulk of the endoderm, and are expressed throughout the superficial involuting marginal zone, which is fated to form the endodermal lining of the gut. Xombi also appears to be part of the pathway that leads to induction of bottle cells, which are fated to form pharyngeal endoderm at the anterior end of the gut. Finally, we have found that Xombi expression occasionally leads to the ectopic formation of hollow tubes several hundred microns in length (unpublished results) that are reminiscent of the endodermally lined tube that forms early in gut development. The nature of these tubes and their relationship to Xombi function in vivo, if any, remain to be determined. The ancient role for T genes in endoderm differentiation and tubulation in gut morphogenesis suggests that Xombi may stand at an interesting intersection point between the evolution of morphogenesis and cell fate determination in complex metazoa.

We are particularly indebted to Julie Theriot and John Gerhart for insightful discussions and advice. We also greatly appreciate thought-ful help and comments on the manuscript by Robert Davis, Randy King, Carol Labonne, Adrian Salic, Julie Theriot and other members of the Kirschner lab. Early screens were carried out with the expert assistance of Mike Wu. We thank Louise Evans for excellent assis-tance with in situ hybridizations and Dianne Purdy, Rebeca Ramos, Madeleine Carey and Teresita Bernal for valuable technical assis-tance. Finally, we thank Jian Zhang and Mary Lou King for sharing results prior to publication. Supported in part by a fellowship from the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation (to K. K.). K. D. L. gratefully acknowledges the support of the Department of Biochemistry and Biophysics at the University of California, San Francisco. This work was supported in part by a generous gift from the Chiron Corporation.