The Tlx-2 homeobox gene is a downstream target of BMP signalling and is required for mouse mesoderm development

Mesoderm formation and patterning during gastrulation are fundamental for establishing the basic body plan of the mammalian embryo (Hogan et al., 1994). During gastrulation, active proliferation and rearrangement of epiblast cells leads to the formation of a primitive streak in the posterior midline, where prospective mesodermal cells delaminate from the ectoderm and ingress into the embryo to form a mesoderm layer (Hogan et al., 1994; Tam et al., 1993). The processes of generation, patterning, proliferation and differentiation of prospective mesodermal cells in the primitive streak are believed to be highly coordinated by an uncharacterized molecular cascade initiated by mesoderm-inducing signals.

In vertebrates, multiple growth factors, including members of the fibroblast growth factor (FGF), Wnt and transforming growth factor-β (TGFβ) families, are involved in regulating mesoderm formation (Beddington and Smith, 1993; Smith, 1995). Strong candidates for mesoderm-inducing signals are members of the TGFβ superfamily, such as Vg1, activin, nodal, nodal-related factors (Xnr) and bone morphogenetic proteins (BMP2, 4 and 7) (Smith, 1995). A critical role for these factors in Xenopus has been suggested by the observation that overexpression of dominant negative forms of their receptors leads to abnormal mesoderm formation and patterning (Smith, 1995). In the mouse, targeted mutations of BMP4 (Winnier et al., 1995) or the BMP type I receptor, Bmpr (ALK3) (Mishina et al., 1995), lead to defects in mesoderm formation and mutations in nodal affect primitive streak and mesoderm (Conlon et al., 1994; Zhou et al., 1993), demonstrating the requirement for the BMP signalling during mammalian mesoderm development. However, despite the compelling evidence for the important roles of TGFβ-related factors in vertebrate mesoderm formation, little is known as to how these signals execute their function during this developmental process.

Transcriptional responses to TGFβ-related signals in the early Xenopus embryo are necessary to pattern mesoderm. Recent studies in Xenopus have identified several immediate early genes, including homeobox genes such as Mix.1, Mix.2, Xvent-2 and Xom, that respond to BMP or activin signals (Mead et al., 1996; Huang et al., 1995; Chen et al., 1996; Onichtchouk et al., 1996; Ladher et al., 1996). The putative role of these genes in mesoderm development has been inferred from the abnormal patterning induced by their overexpression. However, the requirement for these genes in mediating developmental events remains to be determined. In the case of mammalian gastrulation, the downstream target genes required for the biological function of TGFβ-related signals during mesoderm formation have not been identified.

We describe in this paper the function of Tlx-2, a murine homeobox gene related to the human T-cell leukemia oncogene HOX11 (Raju et al., 1993; Dear et al., 1993; Kennedy et al., 1991; Hatano et al., 1991; Dubé et al., 1991; Lu et al., 1991). Recent studies have revealed roles of HOX11-related genes during formation of the mouse spleen and the enteric nervous system (Roberts et al., 1994; Dear et al., 1995; Hatano et al., 1997; Shirasawa et al., 1997). We demonstrate here that Tlx-2 is a critical downstream target of BMP signalling that is required during gastrulation in the mouse. We show that activation of the Tlx-2 promoter by BMP signalling is mediated by the BMP type I receptors, ALK3 or ALK6 (Hogan, 1996), and Smad1 (Hoodless et al., 1996; Graff et al., 1996; Liu et al., 1996; Thomsen, 1996). Furthermore, we demonstrate that deletion of the Tlx-2 gene by targeted mutation leads to defects in development of the primitive streak and mesoderm. These results thus define a BMP/Tlx-2 signalling pathway that is essential for mammalian mesoderm formation and illustrate a molecular mechanism for the function of TGFβ-related factors during this developmental process.

The homeobox fragment of HOX11 was used as a probe to screen a mouse 129/svj genomic library. To clone the Tlx-2 cDNA, the sequence of the first exon was isolated from the genomic clone and used as the probe to screen an E10.0 mouse embryonic cDNA library (Novagen).

For each BMP2 induction experiment, one litter of CD-1 mouse embryos at day 6.5 was collected in PBS and divided into two morphologically even groups. The control group was incubated in DMEM medium with 0.1% fetal bovine serum, while the experimental group was incubated in the medium with 0.1% fetal bovine serum and BMP2 (10 ng/ml). After 3 hours of incubation, embryos were rinsed in PBS and processed for whole-mount in situ hybridization.

Whole-mount in situ hybridization was performed as described previously (Wilkinson, 1992). Two antisense DIG-riboprobes, derived from different regions of the Tlx-2 cDNA, were generated for the hybridization. Both probes yielded the same expression pattern, whereas the corresponding sense probes did not produce specific signals. The ribo-probes of T and Evx-1 were prepared as described (Ang and Rossant, 1994; Dush and Martin, 1992).

A BamHI fragment (4.7 kb) containing the first exon was isolated from the genomic clone, and cloned into pGEM7 to create pGEM7-Tlx-2/4.7 used for the vector construction. A StuI-NcoI fragment (1.0 kb) spanning from an upstream StuI site to a NcoI site located at the initiation codon was used as the 5′ homologous arm, and was fused with the lacZ-coding sequence in frame, by cloning into the NcoI/SpeI (blunted) sites of a lacZ-pBSKS (+) plasmid. A fragment containing the 5′ arm and lacZ was then released by NotI (blunted) and SalI digestions and cloned into SacII (blunted)/XhoI sites of the pGN plasmid (Le Mouellic et al., 1992) to replace the lacZ sequence in pGN, creating pGN-Tlx-2(5′)-lacZ. A NotI (blunted)/NsiI fragment (2.6 kb) derived from pGEM7-Tlx-2/4.7, extending from a NotI site downstream of the first exon into the NsiI site in the polylinker, was used as the 3′ homologous arm after cloning into the BamHI (blunted)/NsiI sites in pGN-Tlx-2(5′)-lacZ. The vector was linearized by NsiI digestion.

R1 ES cells were used to introduce the targeted mutation. ES cells were handled, transfected and selected according to established methods (Joyner, 1993). The targeted cell lines were first identified by Southern blot, with BamHI digestion of genomic DNA and the 5′ external probe (Fig. 3), and then confirmed by Neo and lacZ internal probes (data not shown). Of 124 G418-resistant colonies screened, 8 were found to have a correctly targeted allele, an overall frequency of approximately 1 in 15.

The ES cells from three targeted clones were used to generate chimeric mice by aggregating with CD-1 morulae (Joyner, 1993). The resulting chimeras were bred to CD-1 females and offspring were tested by Southern analysis for the presence of the targeted allele. Chimeras from two cell lines transmitted the mutation through the germline. Heterozygous offspring were interbred to generate homozygous progeny. Genotypes were determined by either Southern blot or PCR as indicated in the figure legends.

Adult mice and embryos after E9.5 were genotyped by Southern blot analysis. Early stage embryos were genotyped by PCR with DNA extracted from cultured ectoplacental cones. Three primers were used in each reaction (Fig. 1a): (a) 5′TCACCTTCCCCTGGATGGAC3′;(b)5′CCACGTCGGATTGAACAGAG3′; (c) 5′ATCGCCTTCTTG-ACGAGTTC3′. The PCR products were resolved on a 1.5% agarose gel. The sizes of the amplified fragments are: a+b=173 bp; b+c=325 bp (Fig. 3c).

An 1.6 kb promoter sequence was isolated by BamHI/NcoI digestion (NcoI was blunted with Mung Bean Nuclease) from a Tlx-2 genomic clone and cloned into the BglII/HindIII sites (HindIII was blunted with Klenow fragment) of the pGL2-Basic plasmid (Promega) to construct the pTlx-Lux reporter plasmid. The 1.6 kb promoter fragment extends upstream from the first ATG codon of Tlx-2, which was destroyed by Mung Bean Nuclease.

P19 cells were cultured in α-MEM containing 7.5% calf serum and 2.5% fetal calf serum. Cells were transiently transfected with the reporter plasmid (pTlx-Lux) and the indicated constructs using the calcium phosphate-DNA precipitation method as described previously (Macías-Silva et al., 1996). To induce the reporter construct, cells were incubated overnight with BMP2 at the indicated concentrations. Luciferase activity was measured using the luciferase assay system (Promega) in a Berthold Lumat LB 9501 luminometer. Transfection efficiency was normalized by co-transfection with CMV-β-galactosidase which was measured using an ONPG colorimetric assay.

The Tlx-2 gene was identified during a screen of a mouse genomic library using a probe derived from the homeobox region of human HOX11 (Dubé et al., 1991; Raju et al., 1993). To understand the putative function of Tlx-2 in development, we first determined its expression pattern in mouse embryos by whole-mount in situ hybridization analysis (Wilkinson, 1992). Expression was first detected in E7.0 embryos in a broad domain that correlated with the primitive streak and in the proximal lateral region (Fig. 1b). The expression of Tlx-2 at this stage was observed in ectoderm at the primitive streak, anterior proximal and extraembryonic regions, but not in mesoderm and endoderm (Fig. 1c). However, at E7.5 expression was only seen in the extreme posterior end of the primitive streak and in two symmetrical stripes that extended laterally on either side of the streak (Fig. 1d). Histological analysis of these embryos revealed that the expression of Tlx-2 was localized in both the ectoderm and mesoderm at the posterior end of the primitive streak (Fig. 1e), but in the regions immediately anterior to this, expression in the stripes occurred in the ectoderm and was excluded from the midline and the mesoderm (Fig. 1f). The lateral expression pattern of Tlx-2 thus correlated with the position of the presumed edges of the neural ectoderm in the posterior half of the embryo.

In later embryos at E8.5, Tlx-2 was prominently expressed in the posterior end of the primitive streak and extended laterally on either side, similar to the pattern observed at E7.5 (Fig. 1g). In addition, we consistently observed relatively low expression in the node of E8.5 embryos although such signals were not seen in the node at E7.5 (Fig. 1g,d). By E9.0, expression of Tlx-2 could not be detected at the posterior end of the embryo and after E9.0 expression of Tlx-2 was restricted to the developing PNS (S. J. T. and M. B., unpublished data). These data suggest that Tlx-2 might function during mesoderm formation and patterning of the early mammalian embryo.

BMP signalling is required for mesoderm development in the mouse and has been shown to be expressed in the posterior primitive streak of E6.5-E7.5 mouse embryos (Winnier et al., 1995). Given the similarities between the timing and expression pattern of Tlx-2 and BMP4, we examined whether Tlx-2 is a target gene for BMP signalling in the mouse embryo. To investigate this, we first determined whether exogenous BMP could induce Tlx-2 expression in early mouse embryos cultured in vitro. For these studies we used embryos at E6.5, which is just prior to the time of in vivo initiation of Tlx-2 expression. Embryos were explanted and cultured in vitro in the presence or absence of BMP2. Since BMP2 is highly related to BMP4 and in vitro both these proteins interact with the same receptors and activate similar downstream components of BMP signalling pathways (Hogan, 1996; Hoodless et al., 1996; ten Dijke et al., 1994), we reasoned that BMP2 would have the same effect on in vitro cultured embryos as BMP4 does. Tlx-2 expression in cultured embryos was examined by whole-mount in situ hybridization. Consistent with our observations described above, Tlx-2 expression was not observed in control embryos cultured in the absence of BMP2 (Fig. 2a). However, treatment of embryos with 10 nM BMP2 for 3 hours resulted in evident induction of Tlx-2 expression in the epiblast layer and probably in the mesoderm in approximately 60% of treated embryos (Fig. 2b,c; Table 1). In the epiblast layer, the strongest induction was observed in the posterior region (Fig. 2c). In contrast, Tlx-2 expression was not induced in the endoderm (Fig. 2b,c).

To characterize more fully the induction of Tlx-2 expression by BMP signalling pathways, we isolated a 1.6 kb fragment from the 5′ region of the Tlx-2 promoter and cloned it upstream of a luciferase reporter gene, to create pTlx-Lux. To test the activity of this promoter, we utilized P19 cells, which are derived from a mouse embryonic carcinoma and can be induced to differentiate into derivatives of the three germ layers in response to BMPs and activins (Vidricaire et al., 1994; Hoodless and Hemmati-Brivanlou, 1997). Transient transfection of pTlx-Lux into P19 cells resulted in low basal levels of transcription of the reporter gene (Fig. 2d). However, treatment of the transfectants with varying doses of BMP2 resulted in a 3- to 4-fold induction of the Tlx-2 promoter which reached maximal levels at 2 nM BMP2 (Fig. 2d). Treatment of the cells with activin had no effect on this promoter (Fig. 2d), indicating that the induction is specific for BMPs.

BMP2 signals through heteromeric complexes of type I and type II ser/thr kinase receptors (Hogan, 1996; Attisano and Wrana, 1996). To identify the signalling receptors that potentially mediate BMP2-dependent induction of pTlx-Lux, we first tested whether a kinase-deficient version of ActRIIB [ActRII(KR)] could inhibit BMP signalling. Consistent with previous studies in Xenopus (Hemmati-Brivanlou and Thomsen, 1995), expression of this dominant negative type II receptor suppressed BMP2-dependent induction of the Tlx-2 promoter (Fig. 2e). Furthermore, expression of constitutively activated forms of the BMP type I receptors ALK3 or ALK6 resulted in strong activation of the promoter (Fig. 2e). Induction of the Tlx-2 promoter by activated type I receptors even in the presence of the dominant negative type II receptor (Fig. 2e), suggests that activation of Tlx-2 is not secondary to autocrine signalling by other TGFβ-like factors.

To further characterize the downstream components in this BMP signalling pathway, we assessed whether the MAD-related protein Smad1 could increase BMP2-dependent induction of the Tlx-2 promoter. Smad1 is the mammalian homologue of Drosophila Mad and was previously shown to function in BMP signalling pathways in vertebrates (Hoodless et al., 1996; Graff et al., 1996; Liu et al., 1996; Thomsen, 1996). For these studies, we utilized suboptimal doses of ligand and tested whether coexpression of Smad1 could increase sensitivity of the cells to BMP2. Under these conditions, treatment of control cells with 1 nM BMP2 resulted in a 2-fold induction of the Tlx-2 promoter (Fig. 2f). However, cotransfection with Smad1 increased BMP-responsiveness of this promoter to almost 5-fold over the controls (Fig. 2f). Interestingly, Smad1 expression alone increased basal transcription from the reporter (Fig. 2f), consistent with the observation that overexpression of these proteins can initiate biological responses (Graff et al., 1996). Previous studies in TGFβ and BMP signalling have shown that the C-terminal SSXS motif of MAD-related proteins is the target for receptor-mediated phosphorylation (Macías-Silva et al., 1996; Kretzchmar et al., 1997). To test whether Smad1-mediated induction of Tlx-2 requires phosphorylation of Smad1, we constructed Smad1(Δ458), in which the C-terminal SSXS motif is deleted from the protein. Cotransfection of this construct resulted in an abrogation of the Smad1-dependent increase in Tlx-2 activation by BMP2 (Fig. 2f). These data strongly suggest that receptor-mediated phosphorylation of Smad1 is required for BMP2-dependent induction of the Tlx-2 promoter. Collectively, our results demonstrate that the Tlx-2 gene is a direct target of BMP2/4 signalling pathways during early mouse development.

Previous studies have shown that BMPs play an essential role during early mouse development (Winnier et al., 1995; Mishina et al., 1995). The expression pattern of Tlx-2 and the demonstration that it is a direct target gene of BMP signalling suggest that the protein might be required for downstream functions of BMPs during early mesoderm formation. To directly assess this, we created a mutation at the Tlx-2 locus by gene targeting (Joyner, 1993). The targeted mutation was introduced by replacing most of the first exon starting from the initiation codon with a lacZ sequence and a Neo^r expression cassette (Fig. 3a). From two transfections of R1 embryonic stem cells, eight cell lines with the expected targeted mutation were identified (Fig. 3b; data not shown). Germline transmission of the mutation was obtained from chimeric mice derived from two independently targeted cell lines (Table 2).

In F₂ offspring derived from the two cell lines, no homozygous mice were identified, indicating that the mutation was lethal during embryogenesis (Table 2). The heterozygotes, in contrast, appeared to be normal. To determine the time of embryonic lethality, embryos at different stages were analyzed (Fig. 3c; Table 3). At E6.5, we observed no obvious morphological differences between mutant and wild-type embryos (data not shown). In contrast, at E7.0, we observed delayed development in some of the embryos, while others still appeared morphologically normal (Fig. 4a). However, by E7.5, all homozygous embryos were arrested in development (Fig. 4b) and often appeared as spherical structures (Fig. 4c). In addition, empty decidua were often observed at this stage. At E8.5, mutant embryos were only found as spherical structures or small pieces of embryonic tissues (Fig. 4d,e). By E9.0, only empty yolk sacs were occasionally found for the mutants (Fig. 4f), and no mutant embryos were ever observed after E9.5 (Table 3).

Together, these data demonstrate that the Tlx-2 targeted mutation causes early embryonic lethality at E7.0-E7.5. Since the mutant embryos derived from both cell lines showed the same phenotype, we focused on one mouse line (206) for further analyses.

In order to determine how defects in Tlx-2 affect early development, we examined histologically the formation of the primitive streak and mesoderm in early wild-type and mutant embryos. Normal E7.0 embryos displayed a well-organized primitive streak with abundant mesodermal cells evident between the ectoderm and endoderm layers (Fig. 5a). In E7.0 mutant embryos (Fig. 5c), in which the size and overall structure appeared to be normal, an undulating non-uniform streak region with little mesoderm was observed (Fig. 5c,d). Unlike the ectoderm in the normal embryos (Fig. 5a,b), which is composed of a single columnar cell layer, the streak ectoderm in these mutants was disorganized, with cells forming a multicell layer (Fig. 5c,d). Occasionally, a few mesoderm-like cells could be detected immediately underneath the deformed primitive streak ectoderm at the posterior end in mutant embryos (Fig. 5d). In addition, the posterior amniotic fold was less developed in these mutants (Fig. 5c,d). In some E7.0 mutants (Fig. 5e,f), the embryos appeared to be smaller with no distinct primitive streak and few if any mesodermal cells. A disorganized ectoderm with multicell layers was also seen in these mutants (Fig. 5f). By E7.5, in normal embryos, the mesoderm layer covers the basal surface of the whole embryonic ectoderm (Fig. 5g). However, in E7.5 mutant embryos that had not yet degenerated, the size of the embryo was smaller, the presumed primitive streak was kinked and the embryonic ectoderm displayed a spherical morphology. These embryos had reduced mesoderm that remained in the posterior region or spread into the extraembryonic cavity (Fig. 5h). Mesodermal cells were not seen in the anterior region of the mutant embryo (Fig. 5i). In mutant embryos at this stage, an allantois-like structure was never seen and we frequently observed a ruffled extraembryonic endoderm with deep folds (Fig. 5h). Together, these results reveal that mutation of Tlx-2 induces specific defects in primitive streak and mesoderm formation.

To further characterize the defects in the mutant embryos, we examined the expression of brachyury, a marker of nascent mesoderm (Wilkinson et al., 1990). In normal embryos, brachyury expression is in the primitive streak during E7.0-E7.5 and in the notochord and posterior end at E8.5 (Fig. 6a-c). In embryos homozygous for the targeted mutation in Tlx-2, we observed initiation of brachyury expression in the posterior end of the embryos at E7.0 (Fig. 6a). However, unlike the normal littermates, brachyury expression failed to expand in the mutants and, at E7.5, remained in a restricted domain of expression on the posterior side (Fig. 6b). Interestingly, in spheroid mutants at these and later stages, we could not morphologically discern the proximal and posterior end but a restricted domain of brachyury expression was observed on one side of these embryos, presumably marking the posterior end (Fig. 6b-d). Of note, these patterns resembled the patterns of mesodermal cells that we observed in histological analyses of mutant embryos with similar spheroid morphology (Fig. 5h,i). Thus, Tlx-2 mutant embryos are deficient in mesoderm formation after the initial induction at the posterior end.

To determine whether the Tlx-2 mutants might also have defects in the ectoderm, we examined the expression of Evx-1, a gene expressed in the ectoderm and mesoderm of the primitive streak region of E7.0 embryos (Dush and Martin, 1992). As shown in Fig. 6e, Evx-1 was specifically expressed in the posterior primitive streak of normal embryos. Evx-1 was also expressed in the mutant embryo although the expression pattern appeared to circle in the proximal region rather than being restricted to the posterior end (Fig. 6e). To further assess development of the streak ectoderm, we examined the expression of Sax-1, an early neurulation marker that is specifically expressed in the ectoderm overlying the primitive streak at E7.5 (Schubert et al., 1995). Similar to Evx-1, Sax-1 expression was also retained in the kinked posterior ectoderm of mutant embryos (data not shown).

To determine if the Tlx-2 mutation affected the anterior mesoderm, we examined the expression of HNF3β in E7.5 embryos by immunohistochemistry. In the wild-type embryo at this stage, HNF3β was expressed in the prechordal mesoderm and in the node (Fig. 6f; Hogan et al., 1994). We observed that HNF3β was expressed in a group of cells in the distorted mutant embryos (Fig. 6f), indicating the initiation of the anterior mesoderm in those mutants.

In vertebrates and invertebrates, multiple classes of secreted molecules control the growth and differentiation of the embryo. In mammals, BMPs play a central role in skeletal patterning and organogenesis, and in regulating gastrulation (Hogan, 1996). However, little is known as to how these factors execute their function during development. Here we show that the homeobox gene Tlx-2 is expressed during gastrulation in the mouse and is induced by BMP2 in early mouse embryos. We demonstrate that the promoter is responsive to BMP2 signalling in embryonic carcinoma cells and is induced by constitutively active forms of the BMP type I receptors ALK3 or ALK6. BMP2-dependent induction of the Tlx-2 promoter is potentiated by increased expression of Smad1 and is dependent on phosphorylation of Smad1 on the C-terminal SSXS motif. Furthermore, deletion of the Tlx-2 gene in the mouse results in specific defects during gastrulation and embryos arrest development at E7.0-7.5. Mutant embryos initiate mesoderm formation at the posterior end. However, mesoderm cells remain restricted to the posterior region and formation of the primitive streak is disrupted, with cells displaying a disorganized, multilayer morphology. These studies define Tlx-2 as a BMP target gene that is required for primitive streak and mesoderm formation during mammalian gastrulation.

Members of the TGFβ superfamily play a central role in vertebrate gastrulation (Hogan, 1996; Hoodless and Wrana, 1997), but how these signals regulate this developmental process is not yet clear. In this study, we identify Tlx-2 as a critical target gene for BMP signalling during mammalian gastrulation. Tlx-2 is expressed in the posterior primitive streak of E7.0-7.5 embryos in a pattern coincident with that of BMP4. Furthermore, activation of BMP signalling pathways induces Tlx-2 in mouse embryos and regulates induction of the Tlx-2 promoter in P19 embryonic carcinoma cells. Thus, Tlx-2 is a target gene for BMP signalling during mouse development and its rapid induction in whole embryos in response to exogenous BMP2 suggests it may represent a direct target of this pathway. BMP signalling is initiated when the ligand induces heteromeric complex formation between its type II receptors and the type I receptors, ALK3 and ALK6. This results in phosphorylation of Smad1, and likely the highly related Smad5, by the type I receptors followed by association with Smad4 and nuclear translocation (Hoodless et al., 1996; Liu et al., 1996; Kretzchmar et al., 1997; Lagna et al., 1996). An analogous pathway has been defined for Smad2 and Smad3 in response to TGFβ and activin signalling (Macías-Silva et al., 1996; Lagna et al., 1996; Zhang et al., 1997). Importantly, these receptor-regulated Smads appear to function specifically in either BMP or TGFβ/activin pathways (Wrana and Attisano, 1996). For instance, in Xenopus, overexpression of Smad1 or Smad2 induces ventral or dorsal mesoderm, respectively (Graff et al., 1996; Thomsen, 1996; Baker and Harland, 1996). Our characterization of BMP2-dependent regulation of Tlx-2 is consistent with these models. The Tlx-2 promoter was induced by activated ALK3 or ALK6 and was unresponsive to activin signalling. Furthermore, increasing Smad1 expression potentiated BMP2-inducing activity and this was dependent on phosphorylation at the C terminus. Thus, Tlx-2 defines a specific gene target of a BMP/Smad1 signalling pathway. Interestingly, Ladher et al. (1996) recently identified a Xenopus homeobox gene, Xom, as an immediate early gene responsive to BMP4 signalling. Like Tlx-2, Xom belongs to the HOX11 subfamily suggesting the existence of an evolutionarily conserved BMP/Tlx-2 (Xom) signalling pathway that is essential for vertebrate mesoderm development.

It is currently unclear what mechanism underlies Smad1-dependent regulation of the Tlx-2 promoter. Studies on the activin-response element in the Xenopus Mix.2 gene have shown that transcriptional activation requires association of Smad2 with the forkhead-containing, DNA-binding protein FAST-1 (Chen et al., 1996). However, recent evidence from Drosophila suggests that Mad proteins may bind DNA to regulate transcription (Kim et al., 1997). Thus, BMP signalling could drive the direct interaction of Smad1 with response elements in the promoter of the Tlx-2 gene, thus activating transcription. Regardless of the precise mechanism, since the gene is not ubiquitously expressed in all regions that are responding to BMP signals, modifiers of Smad1-dependent regulation of the gene must exist. Consistent with this, we have found that the Tlx-2 promoter is not responsive to BMP signalling in a number of epithelial cell lines (S. Kim, P. A. H. and J. L. W., unpublished data).

Our in vitro studies with explanted mouse embryos showed that Tlx-2 was induced by BMP2 throughout the embryonic ectoderm. This suggests that all cells in this layer are competent to respond to BMP during this stage of development. However, while the early expression of Tlx-2 initiates in a broad domain during normal development, at later stages, the gene shows limited expression in two symmetrical wings of expression that bracket the primitive streak. Thus, it is likely that BMP signalling is restricted in the early embryo. There are several potential mechanisms whereby this might occur. Notably, both noggin and chordin are known to bind BMP2 or BMP4 and can function as extracellular antagonists of these factors during early vertebrate and invertebrate development (Zimmerman et al., 1996; Piccolo et al., 1996; Holley et al., 1996). Thus, the domain of BMP4 activity may be restricted by these factors. In addition, in Drosophila and Xenopus, specific gene responses may be controlled by discrete thresholds of activin or dpp (BMP) activity (Green and Smith, 1990; Lecuit and Cohen, 1997) and the distinct domain of Tlx-2 expression that we describe here may reflect such a threshold response to a BMP4 morphogenetic gradient. Finally, Tlx-2 expression may be modified by other signalling pathways during normal development. Thus, at endogenous BMP4 levels, additional signals may be required for Tlx-2 induction and the intersection of these two pathways may yield the observed expression pattern. Interestingly, a similar model involving the intersection of dpp and wg signalling was recently proposed to regulate the restricted expression domain of dll during Drosophila limb development (Lecuit and Cohen, 1997).

Studies in recent years have identified many genes required for mouse gastrulation and mesoderm development (St. Jacques and McMahon, 1996). These include (1) cell signalling molecules such as fibroblast growth factor 4 (FGF4) (Feldman et al., 1995), fibroblast growth factor receptor-1 (FGFR-1) (Deng et al., 1994; Yamaguchi et al., 1994), Notch-1 (Swiatek et al., 1994; Conlon et al., 1995) and TGFβ-like factors (see below), (2) cell adhesion molecules such as β-catenin (Haegel et al., 1995), fibronectin (George et al., 1993), integrin (Fassler and Meyer, 1995; Stephens et al., 1995) and focal adhesion kinase (FAK) (Furuta et al., 1995; Ilic et al., 1995), and (3) transcription factors such as Brachyury (Beddington et al., 1992; Herrmann, 1992; Herrmann and Kispert, 1994), hepatocyte nuclear factors HNF4 and HNF3β (Ang and Rossant, 1994; Chen et al., 1994; Weinstein et al., 1994), Lim1 (Shawlot and Behringer, 1995), Otx2 (Acampora et al., 1995; Ang et al., 1996; Matsuo et al., 1995), MDM2 (de Oca Luna et al., 1995; Jones et al., 1995), eed (Faust et al., 1995; Schumacher et al., 1996) and LCR-F-1 (Farmer et al., 1997). Mutations in these genes can block or disturb the formation or patterning of mouse mesoderm; however, their direct roles in mesoderm induction or specification have often been difficult to distinguish from indirect effects on epiblast cell proliferation and survival. In addition, the functional interaction among these genes during mouse mesoderm development is largely unknown.

We show in this study that mutation of the Tlx-2 gene leads to defects in ectodermal organization and development of primitive streak and mesoderm. In wild-type embryos, Tlx-2 expression was not detected at the onset of gastrulation and, in the Tlx-2 mutant mouse, mesodermal cells expressing brachyury were present and extraembryonic mesoderm was clearly detected. Together, these data indicate that Tlx-2 is not required for the initiation of mesoderm formation at the outset of gastrulation but is required for maintenance of mesoderm production and patterning. The Tlx-2 mutants also displayed defects in the primitive streak and cells in the ectoderm were multilayered and disorganized. These observations suggest that Tlx-2 could be required for specific morphogenetic responses that mediate their ingression through the primitive streak. In the absence of this response, subsequent differentiation into mesodermal cells may be blocked. Consistent with this specificity, two markers of ectodermal differentiation, Sax-1 and Evx-1, were expressed in relatively normal patterns in the Tlx-2 mutant embryos. Alternatively, since cell proliferation has been implicated in initiation and maintenance of gastrulation, mutation of Tlx-2 may disturb this process and block formation of a functional primitive streak.

Previous studies have demonstrated a fundamental role of BMP signalling in the development of the primitive streak and mesoderm in the mouse (Winnier et al., 1995; Mishina et al., 1995). Similar to Tlx-2 mutant embryos, the most severely affected BMP4 mutants that arrest at the egg cylinder stage display defects in mesoderm and primitive streak development (Winnier et al., 1995). In addition, a small amount of extraembryonic mesoderm is observed in both mutants. The phenotypic similarities caused by these two mutations support the view that Tlx-2 and BMP4 function in the same developmental pathway during mesoderm formation. However, in contrast to the Tlx-2 mutant mouse, which all arrest at the primitive streak stage, some BMP4 mutant embryos develop beyond gastrulation. The activation of Tlx-2 expression in these embryos may be attributed to the activity of other related BMPs. In this regard, many BMPs have overlapping biological activities and interact with common receptors. For instance, BMP2 is highly related to BMP4 and both factors bind to common receptors. Similarly, other more distantly related proteins, such as BMP7 and GDF5 can also interact with BMP type I receptors and could thus elicit common responses. This overlap in receptor specificity and biological activity could provide one mechanism whereby BMP4 mutant embryos might be rescued during gastrulation.

The overlap in BMP signalling pathways also raises the possibility that Tlx-2 might be regulated by other BMP-like signals during gastrulation. In particular, nodal is expressed throughout the epiblast at the egg cylinder stage and becomes restricted to the primitive streak at gastrulation (Colon et al., 1994; Zhou et al., 1993). Similar to Tlx-2 mutants, homozygotes with nodal mutations are defective in the formation of the primitive streak and most mesoderm. The expression and phenotypic coincidence between Tlx-2 and nodal mutants suggests that it could either directly, or cooperatively with BMP4, regulate Tlx-2 expression during gastrulation. Recent studies have shown that Nodal expression initiates before streak formation in the most proximal epiblast (Varlet et al., 1997). Therefore, Nodal signals may also activate Tlx-2 expression before gastrulation at a low level that could not be detected by our procedure of in situ hybridization.

BMP2 and BMP4 signal through the type I receptor ALK3, which is expressed in the epiblast and mesoderm during gastrulation (Mishina et al., 1995). However, the phenotype of the ALK3 mutant mouse is more severe than that of Tlx-2, and the embryos arrest development prior to the initiation of mesoderm formation and induction of Tlx-2. In these embryos, the primary defect may be a failure of adequate proliferation of the epiblast prior to gastrulation (Mishina et al., 1995). This phenotypic difference indicates that Tlx-2 might only carry out a subset of the functions of BMP signalling cascades during gastrulation. Since mutations in receptors likely disrupt the activity of several BMP-like factors and their target genes, this suggests that ALK3 signalling pathways mediate multiple essential gene responses during gastrulation. Our studies on Tlx-2 have dissected one of these pathways and define a target gene required specifically for expansion of mesoderm and formation of the primitive streak.

Other target genes are likely to mediate other responses to BMP both during gastrulation and at later stages of development when these factors function in tissues where Tlx-2 is not expressed. Interestingly, Msx homeobox genes have been implicated as BMP target genes in wide range of tissues (Davidson, 1995). These genes show delayed induction in response to BMP treatment (Gañan et al., 1996) and the mutants display a variety of craniofacial and teeth abnormalities that are not phenocopied by any of the BMP knockout mice (Satokata and Mass, 1994; Mass, 1997). Thus, it is unclear whether these genes are direct targets of BMP signalling pathways and whether they are required specifically for BMP function during development. Nevertheless, the data suggest that tissue-specific regulation of homeobox genes by BMP signalling pathways could function at multiple times throughout mouse development. Thus, the competence of the cell to respond to BMPs in a specific way may be determined by its developmental status. These cell-type-specific responses to BMPs could thus provide one mechanism to explain the diverse array of developmental processes that these factors regulate.

In summary, we have identified Tlx-2 as a downstream target gene of BMP signalling during mouse gastrulation and show that it is essential for the development of the primitive streak and mesoderm. These results define a BMP/Tlx-2 signalling pathway that is required during early mammalian development and provide a molecular mechanism for the function of a TGFβ-related factor during mesoderm induction.

We thank Drs Y. Mishina, R. Behringer, C. C. Hui, O. Jin, G. Martin, P. Gruss, P. Chambon, R. Colon and A. McMahon for providing reagents, and C. C. Hui, J. Lightfoot, M. Crackower and R. Cumming for their critical reading of the manuscript. We are grateful to M. Gertsenstein for her help in creating the chimeric mice, K. Raju for his involvement during the early stage of Tlx-2 cloning, and M. Macías-Silva for Smad1(Δ458). This work was supported by grants to M. B. and J. L. W. from the Medical Research Council of Canada. S. J. T. is supported by a RESTRACOM Ph.D. studentship from the Hospital for Sick Children and P. A. H. and J. L. W. are recipients of an MRC Centenial Fellowship and Scholarship, respectively. R. R. M. is an International Research Scholar of the Howard Hughes Medical Institute.