A vegetally localized T-box transcription factor in Xenopus eggs specifies mesoderm and endoderm and is essential for embryonic mesoderm formation

Pattern formation in the embryos of most animals is guided by localized factors in eggs and inductive interactions that occur between cells of the developing embryo. Evidence has been gathered for decades supporting the existence of localized factors, or cytoplasmic determinants, in the eggs of a variety of phyla, and inductive interactions have been described in embryos throughout the animal kingdom (Slack, 1991a). For example, in Drosophila the localization of bicoid and nanos mRNAs to the anterior and posterior poles of the egg, respectively, is essential for normal anteroposterior axial patterning (Struhl, 1989). Furthermore, after cellularization of the Drosophila blastula, the TGFβ-related growth factor DECAPENTAPLEGIC (DPP) mediates inductive interactions that pattern the dorsal-ventral axis, and DPP also performs other important patterning functions in later development (Gelbart, 1989). In another example, the localization of glp-1 mRNA to the AB blastomere in the C. elegans embryo is required for normal development of cells derived from that blastomere, but inductive interactions are required to establish the normal fates of the daughter cells (Moskowitz et al., 1994).

The existence of localized determinants in the eggs of chordates such as tunicates and amphibians has been inferred by classical embryonic manipulations such as the culture of tissue explants (Gilbert, 1994; Slack, 1993), or sometimes direct visualization (e.g. the yellow crescent of tunicates; Whitaker, 1977). More directly, in the vertebrate Xenopus laevis, mRNAs localized to the animal or vegetal pole of the egg have been isolated. Two of these, Vg1 and Xwnt-11, are vegetally localized and encode secreted growth factors with mesoderm-inducing and/or -patterning activities (Ku and Melton, 1993; Melton, 1987a). Much of vertebrate development is guided by inductive interactions, and the first of these to occur in the Xenopus embryo establishes the mesoderm. Mesoderm induction can be triggered by growth factors in the FGF family and by TGFβ-related factors, such as BMPs, Activin, Vg1, and nodal (reviewed by Kessler and Melton, 1994; Jones et al., 1995). Nonetheless, cells isolated from different regions of the blastula marginal zone autonomously express early mesodermal marker genes, in accord with their spatial position, which provides evidence that maternal determinants may also influence formation of the mesoderm (Lemaire and Gurdon, 1994).

We sought to identify new mesoderm-specific genes in Xenopus and, in the process, uncovered a gene that is expressed throughout the mesodermal germ layer. The gene is a new member of a family of transcription factors, called T-box genes, that function in embryonic pattern formation in a wide spectrum of animals. Mouse Brachyury, or T, was the first member isolated (Herrmann et al., 1990) and, henceforth, a variety of related genes have been isolated in metazoans from nematodes through mammals (Herrmann, 1995a), including Xenopus Brachyury (Xbra) (Smith et al., 1991) and eomesodermin (Ryan et al., 1996). All contain a DNA-binding domain of about 200 amino acids and about 40-50% identity, referred to as the T domain (Kispert, 1995). Outside of the T domain, there is no significant sequence identity among the various T domain proteins and little information regarding how they regulate transcription.

T-box genes function in a variety of developmental processes and the founding member of the family, Brachyury (T), has been extensively studied in mice, zebrafish and frogs (reviewed in Herrmann, 1995a). In all of these organisms, the Brachyury gene is expressed throughout the nascent mesodermal germ layer at blastula and early gastrula stages, and later it is specifically expressed in the notochord (Schulte-Merker et al., 1992; Smith et al., 1991; Wilkinson et al., 1990). Consistent with its embryonic expression, Brachyury functions in mesodermal patterning. Mice that are heterozygous for Brachyury display defects in trunk and tail development, and homozygous mice are severely defective in mesoderm formation and die early in utero (Herrmann, 1995b). In zebrafish, the no tail (ntl) gene encodes Brachyury, and heterozygous ntl mutations result in phenotypes that resemble heterozygous mutations of mouse T (Schulte-Merker, 1995). In Xenopus, Brachyury (Xbra) is capable of triggering ventral mesoderm differentiation when ectopically expressed in animal caps (Cunliffe and Smith, 1992) and the inhibition of Xbra function in vivo by a transcriptional repressor form of the protein blocks posterior mesoderm formation in a manner quite similar to the mouse and zebrafish T mutations (Conlon et al., 1996). A second Xenopus T-box gene, eomesodermin, has also been isolated recently (Ryan et al., 1996), and its expression pattern is similar to that of Xbra, but eomesodermin is not expressed in the notochord. Eomesodermin can induce mesoderm and a dominant-interfering version blocks normal Xenopus mesodermal patterning.

We have isolated a third example of a T-box gene from Xenopus, which we have named Brat. The gene is expressed maternally and its transcripts are localized to the vegetal pole of the egg and cleavage-stage blastula, making Brat perhaps the first example of a localized maternal transcription factor. The Brat gene is also expressed zygotically throughout the early mesoderm beginning at gastrulation. Consistent with its expression patterns, we demonstrate that ectopic expression of Brat induces a wide array of mesoderm, from ventral blood through dorsal mesoderm of the Spemann organizer. Mesoderm induction by Brat occurs in a graded fashion, with the character of the mesoderm shifting from ventral to dorsal as the dose of Brat is increased. Brat also induces endoderm, consistent with its maternal localization to the vegetal pole. We demonstrate that Brat is essential for the formation of mesoderm and for proper axial patterning by inhibiting its activity with a dominant-inhibitory form of the protein.

A stage 10⁺ cDNA library was constructed using the Stratagene ZAP express kit and screened with a subtracted ventral stage 10⁺ cDNA probe. Subtraction of common cellular sequences was performed using the Clontech photoactivatable biotin labeling kit. Maternal mRNA served as the driver in the subtraction, and first strand stage 10⁺; ventral cDNA was the tracer. The non-subtracted first strand ventral cDNA (0.95 μg) was labeled with ³²P-dATP by random priming with a kit (Boehringer). Approximately 3.5×10⁷ cts/minute of probe was used to screen 1.4×10⁵ plaques on nylon filters. Hybridization conditions were 5× SSC, 5× Denhardt’s, 0.5% SDS, 0.1 mg/ml salmon sperm DNA at 65°C. Filters were washed in 0.1× SSC, 0.1% SDS at 65° C for 30 minutes. Positive plaques were picked and pooled into 12 groups and screened a second time at a lower density (2,000 pfu per plate) with the same probe to isolate individual candidates. 100 positive plaques, selected at random, were converted to plasmids and used to produce digoxigenin-labeled probes for whole-mount in situ hybridization analysis on stage 10.5 embryos to determine whether they were expressed in the mesoderm. One of these was a partial-length cDNA of Brat. To isolate a full-length Brat cDNA a BamH1 200 base pair (b.p.) fragment of the original isolate was used to rescreen the ventral cDNA library, as above.

DNA sequencing was performed on 5′ and 3′ deletions of Brat with a Sequenase kit (U.S. Biochemical). Deletion clones of Brat were produced with the Promega Erase-a-Base kit. To construct CS2-Brat, full-length Brat cDNA was excised from pBK-CMV by SalI/XhoI digestion, polished with Klenow and subcloned into the StuI site of CS2⁺ (Rupp et al., 1994). The dominant-negative Brat-En^R clone was constructed by inserting an 860 b.p. ClaI-Bsu36I N-terminal fragment of Brat cDNA (excised from the CS2+ vector) into the ClaI-EcoRI site of pCS2-ENG-N (unpublished gift of D. Kessler), which contains the first 298 amino acids of the Drosophila engrailed protein, nucleotides 169-1064. Capped synthetic mRNAs for microinjection experiments were synthesized with SP6 polymerase using the Ambion mMessage Machine kit. Templates for synthesis of synthetic mRNA were cut as follows: CS2-Brat with NotI, CS2-Brat-En^R with SacII, BUT13 (Xbra) (Rao, 1994) with EcoRI, LacZ pSP64T with XbaI. β-gal staining was performed according to Vize et al. (1991).

Whole-mount in situ hybridization was performed as described (Harland, 1991) using BM Purple colorimetric substrate. Antisense digoxigenin probes were synthesized from the following templates with appropriate RNA polymerases: Brat in pBK-CMV was cut with EcoRI and transcribed with T3; Goosecoid, pΔgsc (Cho et al., 1991) was cut with XbaI and transcribed with T3; Xbra, BU-K345 (gift of P. Wilson and A. Hemmatti-Brivanlou), was cut with XhoI and transcribed with SP6. For histological purposes, Brat-injected animal caps were fixed in MEMFA for 1 hour and embedded in paraplast. 10 μm sections were cut and stained with Giemsa. Whole-mount immunohistochemistry was performed as described (Hemmati-Brivanlou and Harland, 1989) using an alkaline phosphatase-conjugated secondary antibody and detected with BM Purple substrate (Boehringer). Antibodies used were 12/101 for muscle (Kintner and Brockes, 1984).

Embryos in 3% Ficoll, 0.5× MMR were injected in the animal pole at the 2-cell stage with synthetic mRNA. Animal caps were cut at blastula stages 8-9 and cultured in 0.5× MMR until harvesting. RNA preparation, RT-PCR assays and primer pairs were as described (Henry et al., 1996; Thomsen, 1996). Brat and Chordin primers were as follows. Brat: upstream 5′-CAG TGC CGG ATT CCG TAT C (1005-1023) and downstream 5′-GAG CTA CTG CTC CTT GTG (1290-1270); Chordin : upstream 5′-TTT CGC AAC AGG AGC ACA GAC (3439-3459) and downstream 5′-TAC CGC ACC CAC TCA AAA TAC (3722-3702). In mesoderm induction assays with proteins, animal caps cut at blastula stage 8-9 were treated with FGF or activin protein in 0.5× MMR, 0.1% BSA, cultured until sibling embryos reached stage 11 and RNA was prepared as above for northern blots. Primer sequences are available on the Xenopus XMMR home page (http://vize222.zo.utexas.edu).

To gain a better understanding of the early events of mesoderm induction and patterning, we sought to isolate new mesodermspecific genes by differential screening. We screened a stage 10⁺ ventrolateral cDNA library with a gastrula stage ventral cDNA pool (details in Materials and Methods), and the expression patterns of 100 positive clones were analyzed by whole-mount in situ hybridization. Several were mesodermspecific and among these one coded for a new member of the T-box family of transcription factors. We named this gene Brat because it is most closely related to the other vertebrate T-box genes, Brachyury and Tbx-2 (Campbell et al., 1995) The Brat cDNA is 2.7 kb in length, and the Brat open reading frame is 456 amino acids (Fig. 1A). Within the T domain, Brat is about 50% identical at the amino acid level to other T domain proteins (Fig. 1B) but, outside of the T domain, Brat is not significantly similar to other members of the T-box family nor to any other proteins. In general, T-box genes do not share significant sequence identity outside of the DNA-binding domain.

A developmental northern blot revealed that Brat transcripts are maternal and present throughout early development until neurula stage 18 (Fig. 2A). The presence of Brat transcripts in Xenopus eggs prompted us to examine their in situ spatial distribution during oogenesis. We found that Brat transcripts are uniformly distributed in stage I oocytes but become vegetally localized by stage II and remain so thereafter in oogenesis (Fig. 2B). The localization of Brat mRNA coincides precisely with the unpigmented, vegetal region of late stage oocytes (compare upper and lower panels of oocytes in Fig. 2B). This domain of localization approximately coincides with that of two other vegetally localized transcripts, Vg1 and Xwnt-11 (Ku and Melton, 1993; Melton, 1987b). Brat mRNA is tightly associated with the vegetal cortex since partial removal of the membrane from the vegetal pole of fixed oocytes abolishes the in situ hybridization signal (Fig. 2B).

Brat transcripts remain predominantly vegetal during the early cleavage stages of embryonic development. A northern blot on animal and vegetal halves of stage 6 embryos (approximately 32-64 cells) showed that Brat message is localized mainly within the vegetal blastomeres, although a low level of Brat mRNA is also present in cells of the animal hemisphere (Fig. 2C). Whole-mount in situ hybridization on 32-cell embryos confirmed that Brat mRNA is contained mostly within the C and D tiers of vegetal blastomeres (data not shown). Cells cleaved from the D tier form endoderm exclusively, while cells of the C tier contribute to both the endoderm and mesoderm (Dale and Slack, 1987). The presence of Brat transcripts in the D tier raises the possibility that it may act as a maternal determinant of endoderm formation. The localization of Brat transcripts within the C blastomeres also suggests that maternal Brat might function in mesoderm specification. Functional studies (below) support these notions.

Zygotic Brat transcripts first appear in the dorsal marginal zone just prior to the appearance of the dorsal blastopore lip (stage 10–, Fig. 3A). Thereafter Brat expression extends ventrally so that, by stage 10+, the Brat gene is transcribed throughout the marginal zone (Fig. 3B). The timing and distribution of Brat transcripts appear nearly identical to those of Xenopus Brachyury (Xbra) at these early stages (compare Fig. 3A,B with D,E). However, by late gastrulation (stage 12), differences in their expression patterns are apparent (Fig. 3C,F). Both Brat and Xbra transcripts are present throughout the mesodermal ring around the closing blastopore. Brat transcripts, however, are not present in the involuting axial mesoderm that will form the notochord, whereas Xbra transcripts are present (Fig. 3C,F). Late in gastrulation and through neurula stage 15 Brat is expressed in the posterior mesoderm but, by stage 18, its transcripts are not detected (Fig, 2A and data not shown).

Xbra and Brat encode putative DNA-binding factors, and Xbra has been shown to function as a transcriptional activator that can trigger mesoderm formation when ectopically expressed in Xenopus animal caps (Cunliffe and Smith, 1992). We compared the mesoderm-inducing potential of Brat with Xbra and found that Brat indeed induces mesoderm in animal caps, but that the mesoderm-inducing activity of Brat differs from that of Xbra. Brat and Xbra both induced the ventrolateral mesodermal markers XMyoD and Xwnt-8 (Fig. 4A), although Xbra induced Xwnt-8 to a lesser degree than Brat. The two factors show a distinct difference in their activities, however, because Brat induces the goosecoid and XFKH-1 genes that mark the Spemann organizer, the most dorsal mesodermal derivative, but Xbra does not.

We examined mesoderm induction by Brat in greater detail and found that Brat induces a full spectrum of early markers for ventral, lateral and dorsal mesoderm in a dose-dependent fashion (Fig. 4B). Low doses of Brat mRNA (14 picograms, pg) induce ventral and posterior mesodermal genes, such as Xwnt-8, Xhox3, Xlhbox6 and XMyoD. Intermediate doses (42-140 pg) induce XFKH-1, a marker expressed in the organizer and adjacent dorsolateral tissue, and doses at or above 350 pg induce Spemann organizer-specific genes, such as chordin and goosecoid. Note that, as the dose of Brat is increased, the expression of ventral-posterior markers (Xhox3, Xlhbox6 and XMyoD) declines as organizer gene expression increases, reflecting a shift in the proportion of dorsal or ventral mesoderm induced by Brat. All doses of Brat that induce mesoderm also induce the Xbra gene, consistent with the relatively uniform embryonic expression pattern of Xbra across the dorsal-ventral axis of the marginal zone. Brat-injected animal caps also express late stage markers of dorsal and ventral mesoderm (Fig. 4C) in agreement with the induction of early mesodermal markers. At the equivalent of tadpole stage 28, Brat induces the muscle-specific actin gene, a marker for somitic muscle, and alpha globin, a marker of red blood, the most ventral mesodermal tissue.

To confirm that Brat induces mesodermal tissues, as opposed to simply activating mesoderm-specific genes, we examined histological sections of animal caps injected with Brat mRNA. At low doses (50 pg) of Brat, animal caps form vesicles, characteristic of ventral mesoderm (Fig. 5B). At high doses (500 pg) of Brat, the animal caps elongate and form muscle (Fig. 5E) a dorsal mesodermal tissue, as shown by immunohistochemical staining of Brat-injected animal caps with the muscle-specific antibody, 12/101. In CS2-injected animal caps, no muscle tissues form (Fig. 5D), while 33% of Brat-injected animal caps (n=27) form muscle (Fig. 5E). At low doses of Brat, 50-250 pg, little muscle tissue forms. As a positive control, all BVg1-injected animal caps (n=22) form muscle (Fig. 5F). Although Brat can induce organizer-specific markers, as well as muscle, we did not observe the formation of notochord tissue (n>30). These histological examinations confirm our findings with molecular markers that Brat induces ventral and dorsal mesoderm.

The localization of maternal Brat transcripts within vegetal cells led us to test whether Brat might also function in endoderm specification. Brat-injected animal caps were scored at the equivalent of tadpole stage 28 for induction of the Xlhbox8 gene, a marker of anterior endoderm of the pancreas and liver (Wright et al., 1988), and the Intestinal fatty acid binding protein (IFABP) gene (Shi and Par Hayes, 1994), a general endoderm-specific marker (Fig. 6). Brat mRNA doses at or above 70 pg induce both markers, which prompts us to suggest that Brat might act as a maternally localized determinant for endoderm differentiation.

Mesoderm is induced within the marginal zone of the Xenopus embryo by factors secreted from the vegetal cells at blastula stages. Mesoderm can be induced in blastula stage animal caps in vitro when placed in contact with vegetal cells or exposed to particular growth factors of the TGFβ and FGF families (Kessler and Melton, 1996). Like other mesoderm-specific genes, we anticipated that, in animal caps, Brat would respond to mesoderm-inducing factors, and this prediction holds true. Brat gene expression can be induced in animal caps exposed to basic FGF (bFGF) or activin B proteins, and the level of induction is concentration-dependent (Fig. 7A). The lowest concentration of factor that induced Brat was 6.4 ng/ml for bFGF and 2 ng/ml for activin B and, as the concentration of either factor was increased, we observed a concomitant increase in the level of Brat gene expression (Fig. 7A). The Brat gene can also be induced by BMP-4 or Vg1 (in its functional form, Bvg1; Thomsen and Melton, 1993) when these factors are expressed in animal caps from microinjected synthetic mRNA (Fig. 7B). These results demonstrate that Brat can be induced by growth factors that specify either ventral mesoderm (FGF and BMP-4) or dorsal mesoderm (activin and BVg1), consistent with the expression of the Brat gene throughout the early mesoderm.

The expression of Brat in the marginal zone and its ability to induce mesoderm ectopically in animal caps predict that Brat plays an essential role in the formation of mesoderm in the embryo. To test this hypothesis, we sought to block the function of endogenous Brat with a dominant negative version of Brat. Natural mutations in the mouse Brachyury gene (Herrmann, 1995a) and studies with mutated forms of Xbra (Conlon et al., 1996; Rao, 1994) indicate that a transcriptional activation domain lies within the C terminus of both proteins, outside of the DNA-binding domain. Deletion of the C terminus of Xbra generates a dominant negative mutant protein that can antagonize wild-type Xbra in mesoderm induction assays (Rao, 1994). Furthermore, replacement of the C terminus of Xbra with the transcriptional repressor domain of the Drosophila engrailed protein creates a more potent dominant negative inhibitor of Xbra (referred to as Xbra-En^R), and its expression in Xenopus embryos blocks posterior mesoderm formation (Conlon et al., 1996). This domain of engrailed has been shown to function as a transcriptional repressor (Han and Manley, 1993). When this domain is fused to the DNA-binding portion of a transcriptional activator, it turns it into a transcriptional repressor that can act in a dominant-negative manner to block the function of the endogenous factor (Badiani et al., 1994). We sought to create a similar dominant negative form of Brat, so we replaced its C terminus with the Drosophila engrailed transcriptional repressor domain to form Brat-En^R (Fig. 8A).

The effects of Brat-En^R expression in Xenopus embryos are shown in Figs 8 and 9. By whole-mount in situ hybridization, we found that expression of Brat-En^R in the marginal zone inhibits the expression of general and organizer-specific mesodermal genes (Fig. 8). We injected Brat-En^R mRNA together with LacZ mRNA (to mark the injected cells) into the marginal zone at one end of the first cleavage furrow (Fig. 8B). Injected embryos were cultured to early gastrula, stage 10, and the expression of several mesoderm-specific genes was scored. Control-injected embryos showed a normal pattern of Xbra expression in the marginal zone (Fig. 8C), but in embryos injected with Brat-EnR Xbra expression was eliminated in the injected cells (Fig. 8D,E). Furthermore, Brat-En^R inhibited the expression of goosecoid (gsc) within the territory of the Spemann organizer (Fig. 8F-H). Similarly, Brat-En^R also inhibited Xlim-1 expression in the organizer (data not shown). The inhibition of both pan-mesodermal and Spemann organizer markers by Brat-En^R demonstrates that Brat activity is essential for mesoderm formation across the entire dorsalventral axis of the Xenopus embryo.

At the phenotypic level, inhibition of Brat function disrupts body patterning. At the early gastrula stage, expression of Brat-En^R inhibited the formation of the dorsal blastopore lip of the organizer when expressed on the dorsal side (Fig. 9B). The inhibition of dorsal lip formation in these embryos is consistent with the inhibition of organizer specific genes by Brat-En^R (above, Fig. 8). Similarly, expression of Brat-En^R mRNA on the ventral side of the embryo inhibited ventral lip formation (Fig. 9C). At tadpole stages, embryonic patterning defects caused by Brat-En^R are clearly evident. Embryos injected with Brat-En^R mRNA on the dorsal side lack a head (or have a very small head rudiment) and develop severe dorsal trunk defects (Fig. 9E) due to incomplete gastrulation and neural plate closure (arrow in Fig. 9E). Ventral expression of Brat-En^R caused the reduction or loss of tail, trunk and ventral structures (Fig. 9F). These results demonstrate that Brat activity is necessary for correct embryonic body patterning.

To demonstrate that Brat-En^R specifically antagonizes the function of Brat, we co-expressed both transcripts in the lateral marginal zone of developing embryos. Trunk defects resulting from the expression of Brat-En^R (Fig. 10B) in the lateral marginal zone were rescued by coinjection of twice as much wild-type Brat mRNA (Fig. 10C). Wild-type Brat also rescued defects produced by dorsal and ventral injection of Brat-En^R mRNA (data not shown). We also tested whether or not Xbra can substitute for Brat and rescue the patterning defects caused by Brat-En^R. We found that Xbra did not rescue Brat-En^R defects at any ratio of Xbra to Brat-En^R, ranging from 0.25:1 to 5:1 (Fig. 10D). The rescue of Brat-En^R phenotypes by wildtype Brat demonstrates that Brat-En^R is a specific dominantnegative inhibitor of Brat in the embryo. Furthermore, the failure of Xbra to rescue dominant-negative Brat phenotypes argues that the two genes do not function in a simply redundant manner in development.

By differential screening, we have isolated a cDNA encoding a T-box transcription factor, named Brat, that displays a unique set of properties. The Brat gene is expressed during oogenesis and its transcripts are localized to the oocyte vegetal pole where they later become incorporated into vegetal cells during embryonic cleavage stages. Zygotic expression of Brat commences just before gastrulation and Brat transcripts are present throughout the marginal zone. Consistent with its expression patterns, Brat can induce mesoderm and endoderm as a function of dose, and blocking Brat activity in the embryo with a dominant-negative version of the protein inhibits mesoderm formation and severely disrupts body patterning. Thus, Brat performs an essential role in Xenopus embryonic development.

The Brat gene is a member of the T-box gene family of DNA-binding proteins, and it shares about 50% identity to other T-box genes within the DNA-binding domain. At least one member of the family, Brachyury functions as a transcriptional activator (Conlon et al., 1996; Kispert, 1995). We have not directly demonstrated that Brat also functions as a transcription factor, but that is almost certainly the case since we can create a dominant-negative version of Brat by fusing its DNA-binding domain to a transcriptional repressor domain of engrailed. Recent results have shown that the C terminus of a gene called VegT, which is nearly identical to Brat, activates transcription in a yeast one-hybrid assay (Zhang and King, 1996).

The developmental expression pattern of Brat suggests it has a dual role in the formation of endodermal and mesodermal tissues. The first phase of its expression is maternal when Brat transcripts become localized to the vegetal pole of oocytes early in oogenesis and, in full-grown oocytes, the transcripts reside mostly in the vegetal cortex. The timing of Brat transcript localization in oogenesis is similar to that of Xwnt-11, Xcat-2 and Xlsirts, which are transported to the vegetal pole via the mitochondrial cloud, referred to as the METRO mechanism (Forristall et al., 1995; Kloc and Etkin, 1995). Unlike Xcat-2 and Xlsirts, however, Brat transcripts are distributed more widely in the vegetal cortex, similar to Vg1 mRNA (Melton, 1987b) which becomes localized via a microtubule-dependent mechanism (Yisraeli et al., 1990). It will be interesting to determine which transport system localizes Brat mRNA.

The vegetal localization of Brat mRNA in the egg results in its sequestration within cleavage stage cells that fate map to the endoderm and part of the mesoderm. At the 32-cell stage, the most vegetal tier of cells, the D tier, is fated to form endoderm, while the next layer of cells, the C tier, populate both endoderm and some of the mesoderm (Dale and Slack, 1987). The capacity of Brat to induce both endodermal and mesodermal marker genes in animal caps prompts us to suggest that Brat may function as a localized determinant for the specification of these germ layers. The capacity of vegetal pole explants of amphibian embryos to form rudimentary endodermal tissues (Holtfreter, 1938) and express endoderm-specific genes (Henry et al., 1996) provides indirect evidence for the existence of maternal determinants for the endoderm. Loss-of-function tests will be required to conclusively establish whether Brat is a bona fide endodermal determinant. The existence of maternal determinants for mesoderm is supported by the demonstration by Lemaire and Gurdon (1994) that presumptive mesodermal cells isolated from the blastula marginal zone express mesodermal genes (Xwnt-8 and goosecoid) in a cell autonomous manner, and their expression properly reflects the ventral or dorsal origin of the explanted cells. Our demonstration that Brat induces mesoderm in animal caps would be consistent with the possibility that maternal Brat may act as a determinant for mesoderm formation. Brat may be the first example of a transcription factor that functions as a localized determinant in a vertebrate egg.

When Brat is expressed in animal caps, it can induce a full range of mesoderm, from the most ventral types through that of the Spemann organizer, and the type of mesoderm induced by Brat is dose-dependent. As the amount of Brat is increased, the character of the mesoderm shifts from ventral, through lateral, to dorsal. On the basis of these results, we hypothesize that a gradient of Brat protein (or activity) is generated across the dorsal-ventral axis of the marginal zone, with the low end of the gradient present in ventral cells and the high end present in dorsal cells. The level of Brat protein in cells along the dorsal-ventral axis would specify the particular character of the mesoderm. It is also possible that a gradient of Brat protein is set up along the animal-vegetal axis of the blastula, since a relatively high level of Brat mRNA is contained within vegetal pole cells compared to animal pole cells (Fig. 2C). The high level of Brat in vegetal cells might be sufficient to specify autonomous endodermal differentiation, but the lower level of Brat in animal cells is insufficient to specify mesoderm in isolated animal caps. Similarly, intact marginal zone tissue, isolated prior to the onset of inductive signaling by vegetal cells, do not form mesoderm, so maternal Brat trapped within these cells is also apparently insufficient to trigger mesoderm differentiation in intact tissue. Whether Brat is the agent that triggers mesodermal gene expression in dispersed marginal zone cells remains to be determined. We are pursuing experiments to examine Brat protein localization in development to begin addressing these issues.

Once mesoderm induction is underway in the Xenopus embryo, Brat may perform a pivotal role in the commitment of marginal zone cells to mesodermal fates in response to mesoderm-inducing signals. The vegetal cells are the source of mesoderm-inducing factors (Nieuwkoop, 1969; Slack, 1991b), and there is good evidence that some of these factors might be members of the TGF-β, FGF and Wnt growth factor families (Kessler and Melton, 1994). We have shown that the Brat gene can be induced by FGF, activin, BMP-4 and Vg1, and this ability to be activated by several types of mesoderm inducers may ensure that Brat is expressed throughout the marginal zone, even if the identity or concentration of mesoderm inducer changes along the dorsal-ventral axis. The expression of Brat, in turn, would trigger differentiation of the appropriate type of mesoderm. For instance, in the ventral marginal zone Brat induced by BMP signals would specify blood, whereas in the organizer domain Brat induced by Vg1 or an activin-like signal would specify organizer-specific tissues. Furthermore, since the Brat gene responds in a concentration-dependent manner to activin and FGF, cells positioned across a gradient of such factors might activate the Brat gene in proportion to the inducing signal. Activin in particular can act as a morphogen on animal cap cells and induce different types of mesoderm as a function of concentration (Green et al., 1992; Wilson and Melton, 1994). The level of Brat gene induction might provide the molecular mechanism by which cells convert the signal from a morphogen into a proportional genetic, and ultimately phenotypic, response. In the Xenopus embryo, however, the steady-state levels of Brat transcripts that we observe in situ are nearly equal in marginal zone cells during gastrulation, inconsistent with the presence of a morphogen gradient that directly affects Brat transcript levels. However, perhaps the dynamic appearance of Brat transcripts starting on the dorsal side of the marginal zone and moving ventrally reflects the presence of a transient gradient of mesoderm inducer(s).

In order to establish whether or not Brat is essential for mesodermal patterning in the embryo, we blocked its activity with a dominant-negative version of Brat, called Brat-En^R, which is a chimera of the Brat DNA-binding domain and a transcriptional repressor domain of Drosophila engrailed. We assumed Brat is a transcriptional activator based on the fact that Xbra is an activator (Conlon et al., 1996). Expression of Brat-En^R in embryos blocked the induction of ventrolateral and Spemann organizer mesoderm and severely perturbed body patterning. Importantly, Brat-En^R specifically inhibits the functions of Brat, because coexpression of these proteins (at a 2:1 ratio of Brat to Brat-En^R) rescues the patterning defects caused by Brat-En^R. We therefore conclude that endogenous Brat is essential for the induction and patterning of the entire mesodermal germ layer.

How Brat functions together with other T-box genes in mesodermal patterning is important to establish, and some of our experiments point to functional interactions between Brat and Brachyury. For instance, we have shown that Brat induces the Xbra gene in animal caps, and in the embryo dominantnegative Brat-En^R blocks Xbra expression. Therefore Brat might directly activate transcription of the Xbra gene, but we have not established whether Brat functions in this manner or acts in some indirect way, such as by up-regulating FGF signaling (which maintains Xbra expression in embryos; Kroll and Amaya, 1996; Schulte-Merker and Smith, 1995). Other evidence indicates that Xbra and Brat perform some similar functions in the Xenopus embryo; however, it is unlikely that their roles are simply redundant. In animal cap assays, both genes can induce ventral and lateral mesoderm, but only Brat can induce organizer markers. Our loss-of-function experiments with dominant-negative Brat, and similar experiments with Xbra (Conlon et al., 1996), demonstrate that both proteins are required for ventral-posterior mesoderm formation and body patterning, but only Brat is essential for the formation of the organizer and head. Another important finding that illustrates a lack of functional redundancy between Brat and Xbra is that Xbra does not rescue embryonic phenotypes caused by dominant-negative Brat: Further work is required to fully understand the relationship between Brat and other T-box genes in mesodermal patterning.

Lastly, it is worth considering whether maternal Brat protein might control transcriptional responses to inductive signaling. That is, Brat protein sequestered in blastula cells during cleavage might act as a nuclear mediator of signals from growth factor receptors and activate transcription of mesoderm- or endoderm-specific genes. There is clear evidence that maternal transcription factors facilitate mesoderm-specific gene activation by growth factors. A maternal factor named FAST binds a Mix.2 promoter element in response to activin signals (Chen et al., 1996) and the goosecoid gene is activated by unidentified maternal transcription factors that recognize an activin-responsive element (Watabe et al., 1995). Whether maternal Brat protein is a target of signal transduction pathways triggered by mesoderm or endoderm-inducing factors awaits the identification of Bratbinding sites in growth factor-inducible genes.

We thank S. Nishimatsu for technical advice and comments on the manuscript. We thank D. Kessler for providing his unpublished pCS2-ENG-N vector, and A. Hemmatti-Brivanlou and P. Wilson for providing an Xbra clone for in situ hybridization. We also thank Y. Rao for providing an Xbra plasmid for synthetic mRNA synthesis and E. De Robertis for the goosecoid clone.

Note: At the time our manuscript was submitted three reports on genes very similar, if not identical, to Brat were reported (Lustig et al., 1996; Stennard et al., 1996; Zhang and King, 1996).