Abstract
The Xenopus homologue of Brachyury, Xbra, is expressed in the presumptive mesoderm of the early gastrula. Induction of Xbra in animal pole tissue by activin occurs only in a narrow window of activin concentrations; if the level of inducer is too high, or too low, the gene is not expressed. Previously, we have suggested that the suppression of Xbra by high concentrations of activin is due to the action of genes such as goosecoid and Mix.1. Here, we examine the roles played by goosecoid and Mix.1 during normal development, first in the control of Xbra expression and then in the formation of the mesendoderm. Consistent with the model outlined above, inhibition of the function of either gene product leads to transient ectopic expression of Xbra. Such embryos later develop dorsoanterior defects and, in the case of interference with Mix.1, additional defects in heart and gut formation. Goosecoid, a transcriptional repressor, appears to act directly on transcription of Xbra. In contrast, Mix.1, which functions as a transcriptional activator, may act on Xbra indirectly, in part through activation of goosecoid.
INTRODUCTION
The body plan of the Xenopus embryo is specified through the asymmetric distribution of maternal determinants followed by a series of inductive interactions (Harland and Gerhart, 1997). The first such interaction is mesoderm induction, in which signals from the vegetal hemisphere of the embryo act on overlying equatorial cells and cause them to become mesoderm rather than ectoderm (Harland and Gerhart, 1997). The best candidates for endogenous mesoderm-inducing factors include members of the TGF-β superfamily, including activin and Vg1 (Harland and Gerhart, 1997; Slack, 1994). Of these factors, the most intensively studied is activin, which is capable of inducing different endodermal and mesodermal cell types in a concentration-dependent manner. Thus, low concentrations of activin induce ventral mesoderm and high concentrations activate genes normally expressed in anterior endodermal tissues (Green et al., 1992; Gurdon et al., 1996).
In an effort to understand mesoderm induction and the concentration-dependent effects of activin, we have studied the regulation of Xenopus Brachyury (Xbra). At the early gastrula stage, Xbra is expressed throughout the marginal zone of the embryo and, as gastrulation proceeds, transcripts are lost from newly involuted mesoderm but persist in the notochord (Smith et al., 1991). Expression of Xbra is induced in explants of animal pole tissue by activin, but stable activation occurs only in a narrow window of activin concentrations (Gurdon et al., 1996); if levels are too low, or too high, the gene is not expressed. This phenomenon may underlie the restriction of Xbra expression to the marginal zone of the embryo. Levels of activin, or an activin-like molecule, may be too high in the vegetal hemisphere, and too low in the animal hemisphere, for expression of Xbra to occur, but levels in the equatorial region may be just right. The concentration-dependent response of Xbra to activin may therefore represent a useful model for the problem of germ layer specification during early development.
Previous work has suggested that the downregulation of Xbra expression at high concentrations of activin is due to repression of transcription by the homeobox-containing genes goosecoid and Mix.1 (Latinkic et al., 1997). Both genes are induced by high concentrations of activin (Gurdon et al., 1996), and overexpression of either causes downregulation of Xbra, both in the embryo and in explants of animal pole tissue (Artinger et al., 1997; Latinkic et al., 1997). The effects of goosecoid and Mix.1 are likely to occur at the level of transcription, because they can also repress Xbra reporter constructs (Latinkic et al., 1997).
Here we examine the roles played by goosecoid and Mix.1 in normal development, first in the control of Xbra expression and then in the development of the mesendoderm. Consistent with the model outlined above, inhibition of the function of either gene product leads to transient ectopic expression of Xbra. Such embryos later develop dorsoanterior defects, suggesting that the activities of goosecoid and Mix.1 are both required for normal head development. As well as having reduced heads, embryos in which Mix.1 function is inhibited have additional defects in heart and gut formation, suggesting that Mix.1 has a broader role in the development of dorsoanterior endoderm.
Our data are consistent with the idea that Goosecoid, a transcriptional repressor, acts directly on transcription of Xbra. In contrast, Mix.1 functions as a transcriptional activator, and probably acts on Xbra indirectly, in part through activation of goosecoid. Coexpression of Mix.1 and goosecoid in animal cap explants leads to the synergistic induction of the endodermal marker XSox17α, another gene induced by high concentrations of activin. Together, these observations suggest that Mix.1 and goosecoid act together to promote dorsoanterior endodermal differentiation and to suppress expression of mesodermal genes like Xbra.
MATERIALS AND METHODS
Plasmid constructs
All recombinant DNA manipulations were performed by standard techniques (Sambrook et al., 1989). Full construction details and maps of all constructs are available on request.
A goosecoid cDNA (Blumberg et al., 1991) was cloned as a HindIII-EcoRI fragment in its reverse orientation into pcDNA3 (Invitrogen) to create pCMV-csg. GscVP16 was constructed by adding two copies of the VP16 minimal transcriptional activation domain (amino acids 413-454; gift of Dr J. Brickman) to a pcDNA3-gsc construct.
A Mix.1 cDNA (Rosa, 1989) was cloned into pcDNA3 as a BamHI-ApaI fragment to create the antisense construct pCMV-1.xiM, or as a HindIII-BamHI fragment into a derivative of pcDNA3 containing two HA tags (B. V. L., unpublished) to create a wild-type overexpression construct. Mix.1-EnR and Mix.1HD-EnR were constructed by using PCR to fuse the Mix.1 coding sequence with a double-haemagluttinin (HA)-tagged Engrailed repressor domain (Conlon et al., 1996, and M. Tada, personal communication). Junctions created by cloning were verified by sequencing.
Reporter constructs
P3 (top strand: 5′-agctTGAG/TCTCTAATTGAATTACTGTACA; bottom strand: 5′-agctTGTACAGTAATTCAATTAGACTCA) or P3C (top strand: 5′-gatcCTGAGTCTAATCCGATTACTGTACG; bottom strand: 5′-gatcCGTACAGTAATCGGATTAGACTCAG) oligonucleotides were annealed and cloned into the HindIII or BglII sites, respectively, of pGL3Promoter (Promega), which contains the SV40 minimal promoter. Clones were isolated that contained two head-to-tail inserts of each oligonucleotide. (P3)6/luc was obtained by cloning 6 copies of the P3 site into a reporter containing the E4 minimal promoter (kind gift of M. Tada). A goosecoid promoter fragment (Watabe et al., 1995) was obtained by genomic PCR and cloned into pGL3Basic to create −300gsc/luc (gift of Niall Armes and Masa Tada). −207gsc/luc and −190gsc/luc were also created by PCR. The nucleotide co-ordinates designate the most 5′ base pairs of the goosecoid promoter retained in the construct and for both constructs the following 3′ primer was used: 5′-GACCTCGAGCTCTCCCATCTGTGCCTCTTC-3′. PCR products were digested with MluI and XhoI and cloned into the same sites of pGL3Basic.
Xenopus embryos and microinjection
Fertilisation, culture and microinjection of Xenopus embryos were as described (Latinkic et al., 1997). They were staged according to Nieuwkoop and Faber (1975).
RNAase protection assays
RNAase protection assays were carried out as described (Smith, 1993), except that rapid aqueous hybridisation was used (Mironov et al., 1995). Probes included Xbra (Smith et al., 1991), goosecoid (Armes and Smith, 1997; Cho et al., 1991), EF-1α (Sargent and Bennett, 1990), ornithine decarboxylase (ODC) (Isaacs et al., 1992), chordin (Howell and Hill, 1997) and XSox17α (Hudson et al., 1997).
Whole-mount in situ hybridisation and immunocytochemistry
Whole-mount in situ hybridisation was carried out essentially as described (Harland, 1991). Probes included Xbra (Smith et al., 1991), goosecoid (Cho et al., 1991) and XMLC2 (Chambers et al., 1994). Whole-mount staining with monoclonal antibodies MZ15 (Smith and Watt, 1985) and 12/101 (Kintner and Brockes, 1984) was performed as described (Smith, 1993). Injected cells were labelled by coinjecting nuclear lacZ RNA followed by X-Gal staining, or by co-injecting biotinylated dextran (Molecular Probes) and detecting with ExtrAvidin-Alkaline Phosphatase (Sigma), using Fast Red as a substrate (Boehringer Mannheim).
In vitro transcription
In vitro transcription using SP6 or T7 RNA polymerase was as described (Smith, 1993).
Luciferase and β-galactosidase assays
Dual-luciferase assays on NIH3T3 and animal cap extracts were carried according to the manufacturer’s recommendations (Promega), essentially as described (Latinkic et al., 1997). In experiments where β-galactosidase was used as a reference, enzymatic assays were performed as described (Sambrook et al., 1989).
Cell culture and transfections
NIH3T3 mouse embryo fibroblasts were cultured in Dulbecco’s Modified Eagle’s Medium (Sigma) supplemented with 10% newborn calf serum, 2 mM L-glutamine, 100 U/ml penicillin and 0.1 mg/ml streptomycin (Sigma). Calcium phosphate transfections were performed as described (Sambrook et al., 1989) in 6-well plates. 5 μg DNA per well was used. Unless indicated otherwise, this comprised 4 μg of pcDNA3 or the indicated derivative, 0.5 μg of reporter plasmid, and 0.5 μg of pRL-TK (Promega) or EF-1α/lacZ as a reference plasmid. Cells were analysed 3 days after transfection for luciferase activity as described above. Each sample was transfected in duplicate.
Immunofluorescence
Indirect immunofluorescence was performed with anti-HA mouse monoclonal antibodies and secondary anti-mouse-FITC antibody. Bright-field and fluorescent images were electronically overlaid.
Electrophoretic mobility shift assays
Proteins for use in binding reactions were translated in the TNT coupled transcription-translation system, according to the manufacturer’s recommendations (Promega). Electrophoretic mobility shift assays were performed as described (Latinkic et al., 1997). In experiments where the identity of complexes was tested by the addition of anti-HA antibodies (1 mg/ml; Boehringer Mannheim), 1 μl of antibody was added after addition of probe, and samples were incubated for an additional 15-20 minutes on ice. The probe derived from the Xbra2 promoter, and the non-specific competitor, were as described (Latinkic et al., 1997). The sequence of the DE and PE is shown in Fig. 4B; annealed oligonucleotides had 5′-GATC single-stranded overhangs.
RESULTS
Interference with goosecoid function causes ectopic expression of Xbra
Misexpression of goosecoid in Xenopus embryos or in activin- or FGF-treated animal caps suppresses transcription of Xbra (Artinger et al., 1997; Latinkic et al., 1997). Since goosecoid can bind to the Xbra promoter (Artinger et al., 1997; Latinkic et al., 1997), and can repress Xbra reporter constructs in a heterologous system in a sequence-specific manner (Latinkic et al., 1997), it is likely that this repression occurs in a direct fashion.
To investigate whether goosecoid regulates Xbra expression during normal development, we inhibited the function of the gene in two different ways. In the first, we interfered with the ability of goosecoid to repress transcription (Mailhos et al., 1998) by adding to it the VP16 transcription activation domain (Fig. 1A; see Materials and Methods). The resulting gscVP16 construct differs from that recently described by Ferreiro et al. (1998) because it includes the entire coding region of goosecoid (in an effort to increase specificity) and because we use two copies of a minimal VP16 activation domain, which in our hands is less toxic than the entire activation domain. GscVP16, like wild-type goosecoid (Latinkic et al., 1997), binds to nucleotides −172 to −154 of the Xbra2 promoter (not shown).
The ability of gscVP16 to interfere with the function of wild-type goosecoid was tested in NIH3T3 cell transient transfection assays using a luciferase reporter construct (pP3C-SV40/luc) in which two P3C sites (see Materials and Methods), to which goosecoid (Wilson et al., 1993) and gscVP16 (data not shown) bind, are positioned upstream of the SV40 minimal promoter. Fig. 1B shows that gscVP16 does not activate pP3C-SV40/luc, but does interfere with the ability of wild-type goosecoid to repress it, even at a ratio of 1:3. Transfection of different quantities of gscVP16 (0.1-4.0 μg) suggests that the inability of gscVP16 to activate transcription is unlikely to be due to squelching effects (not shown). Rather, it is likely that the VP16 domain, positioned at the C terminus of the protein, interferes with the N-terminally located repression domain (Mailhos et al., 1998). The apparent lack of transcription activation by gscVP16 is an advantage in our studies, because its effects should be restricted to preventing goosecoid-mediated repression; it will not exceed this remit by inappropriate activation of goosecoid targets.
Dorsal, but not ventral, injection of both gscVP16 RNA and pCMV-gscVP16, in which gscVP16 expression is driven by the CMV promoter, leads to ectopic activation of Xbra (Fig. 2B,C). Embryos allowed to develop to tadpole stages showed a range of anterior deficiencies, including cyclopia and loss of head. Notochord and somite formation, revealed using monoclonal antibodies MZ15 and 12/101, respectively, were essentially normal (Fig. 2E-H).
These results suggest that goosecoid represses expression of Xbra during normal development. To confirm this conclusion, we used an antisense approach in which a plasmid directing expression of antisense goosecoid RNA under the control of the CMV promoter (pCMV-csg) was injected into Xenopus embryos at the 4-cell stage. Expression of Xbra was then analysed at gastrula stages by whole-mount in situ hybridisation. Previous work has shown that antisense goosecoid constructs causes anterior defects in Xenopus embryos, probably by interfering with translation of goosecoid protein (Steinbeisser et al., 1995).
Fig. 2A shows that embryos injected with pCMV-csg display patches of ectopic Xbra expression in anterior regions, suggesting that the reduction of goosecoid activity in these territories causes activation of Xbra. No such patches were observed in control embryos in which empty vector was injected (Fig. 2D). Embryos injected with pCMV-csg, and with antisense goosecoid RNA, lacked anterior structures (Fig. 2I,J, and data not shown).
Together, these observations show that interference with goosecoid function leads to ectopic activation of Xbra, suggesting that goosecoid is involved in repression of Xbra in the dorsoanterior mesendoderm of Xenopus embryos. This result is consistent with previous work indicating that goosecoid represses transcription of Xbra directly (Artinger et al., 1997; Latinkic et al., 1997). In addition, both approaches indicate that goosecoid-like activity is required for normal development of dorsoanterior mesendoderm.
Mix.1 is a transcriptional activator
We next tested the role of Mix.1 in restriction of Xbra expression. Like goosecoid, Mix.1 can suppress expression of Xbra (Latinkic et al., 1997) and, like goosecoid, Mix.1 contains a paired-type homeodomain.
Although it suppresses expression of Xbra in embryos and animal cap assays, Mix.1 has been stated to act as a transcriptional activator (Lemaire et al., 1998; Mead et al., 1996). Our own experiments demonstrate that Mix.1 causes activation of a reporter gene containing six copies of the P3 binding site (two palindromic core sequences TAAT, separated by three nucleotides) placed upstream of a TATA box (see Fig. 6F), as well as activation of a reporter construct containing two copies of P3 placed upstream of the SV40 promoter (Fig. 3A). Mix.1 can also transactivate the Xbra promoter construct −381Xbra2/luc; this requires the homeodomain binding sites within the −381 promoter, arguing that the effect is specific (Fig. 3A).
Thus, when tested in a simple heterologous system, Mix.1 behaves as a transcriptional activator; in contrast, when expressed in embryos, or in animal caps, Mix.1 suppresses expression of Xbra (Latinkic et al., 1997). How does Mix.1 cause suppression of Xbra in the embryo and what is the biological significance of this effect?
Mix.1 activates goosecoid expression
One way in which Mix.1 might suppress expression of Xbra is by potentiating the action of a repressor. Another is that it acts indirectly, through the activation of a transcriptional repressor and a third possibility is a combination of the two models in which Mix.1 induces a repressor whose activity it potentiates. One potential target gene of Mix.1 is goosecoid. Inspection of published data suggests that the two genes are transiently co-expressed in the organiser and in vegetal tissue during normal development (Medina et al., 1997; Vodicka and Gerhart, 1995), and both are induced in animal caps by high concentrations of activin (Gurdon et al., 1996). To compare directly the expression patterns of goosecoid and Mix.1, we dissected Xenopus embryos at the late blastula and early gastrula stages into left and right halves, which were then processed separately for whole-mount in situ hybridisation. Our results show that the two genes are expressed in overlapping domains on the dorsal side, and that goosecoid is expressed in deep dorsoanterior endoderm (Fig. 3B,C).
We next asked whether Mix.1 can induce expression of goosecoid. Fig. 3D shows that misexpression of Mix.1 in Xenopus animal caps is sufficient to induce expression of goosecoid, and that this induction can occur in the presence of cycloheximide, suggesting that it reflects direct transcriptional activation. The efficacy of cycloheximide treatment was confirmed by demonstrating that cycloheximide also inhibits induction of chordin, which is known to be induced indirectly by activin (Howell and Hill, 1997; Sasai et al., 1994) (Fig. 3E).
To investigate whether the induction of goosecoid by Mix.1 occurs directly, we asked whether Mix.1 can activate a -300 base pair goosecoid reporter construct (Watabe et al., 1995), both in NIH3T3 cells (not shown) and in animal caps (Fig. 4A). In each system, over-expression of Mix.1 leads to activation of reporter gene activity. Progressive 5′ deletions of the goosecoid promoter caused a gradual reduction in Mix.1 responsiveness, suggesting that multiple elements are involved (Fig. 4A).
Inspection of the goosecoid promoter sequence (Fig. 4B) reveals two clusters of putative Mix.1 binding sites (Wilson et al., 1993) within the distal and proximal elements, which confer responsiveness to activin and Wnt signalling respectively (Watabe et al., 1995); these regions also appear to be necessary for the response to Mix.1. The distal element (DE) contains a P3 site, deletion of which causes the greatest loss of activity, and one core TAAT site. The proximal element (PE) includes two inverted repeats of the core binding site separated by 7 base pairs. As expected, Mix.1 binds with higher affinity to the DE than to the PE (Fig. 4C).
These results suggest that one mechanism by which Mix.1 suppresses expression of Xbra is through activation of goosecoid. However, it is still possible, as suggested above, that Mix.1 potentiates the repressor action of goosecoid. This question was investigated by measuring goosecoid and Xbra reporter gene activity in NIH3T3 cells in the presence of Mix.1, goosecoid and a combination of the two proteins. At a ratio of 1:1, goosecoid inhibited Mix.1-induced activation of both reporter constructs, but no evidence for potentiation of repression was obtained in this heterologous system (data not shown). It remains possible, however, that Mix.1 does enhance the activity of a repressor such as goosecoid in vivo.
Inhibition of Mix.1-like function causes transient ectopic expression of Xbra
The above results are consistent with the suggestion that Mix.1 regulates expression of Xbra indirectly, through the activation of repressor molecules such as goosecoid. We next asked whether Mix.1 regulates expression of Xbra during normal development. To this end, we first made a construct (pCMV-1.xiM) in which the entire Mix.1 cDNA is driven in the antisense orientation by the CMV promoter. This is essentially the same strategy as used above for goosecoid. The efficacy of the construct was tested by injecting pCMV-1.xiM into the vegetal hemisphere of Xenopus embryos together with the (P3)6/luc reporter construct. pCMV-1.xiM proved significantly to inhibit (P3)6/luciferase activity, presumably due to interference with endogenous Mix.1 function and this interference was reversed by injection of RNA encoding wild-type Mix.1 (Fig. 5A).
At the mid/late-gastrula stage, 75% of embryos injected with pCMV-1.xiM displayed ectopic patches of Xbra expression on the injected side (Fig. 5B,C), whereas all embryos injected with empty vector showed a normal Xbra expression pattern (Fig. 5E).
To test the specificity of the results obtained with pCMV-1.xiM, we devised a second approach in which the activation function of Mix.1 was compromised by fusing it to the Engrailed repressor domain (Conlon et al., 1996). Two constructs were made (Fig. 6A): one included only the homeodomain of Mix.1 and sequences N terminal to it (Mix.1HD-EnR), while the other included the entire open reading frame of Mix.1 (Mix.1-EnR). Like the wild-type protein, both fusions bind the P3 oligonucleotide (Fig. 6B) and both are nuclear proteins (Fig. 6C-E). When tested in NIH3T3 cells on the P3 reporter, both Mix.1HD-EnR and Mix.1-EnR behave as transcriptional repressors and inhibit activation by wild-type Mix.1 (Fig. 6F). Complete inhibition of Mix.1 activity was achieved with a 1:2 ratio of Mix.1HD-EnR or Mix.1-EnR to Mix.1, and partial inhibition was achieved even with a ratio of 1:10, arguing that our interfering reagents act as active repressors (data not shown). Mix.1-EnR also prevents activation of Xbra and goosecoid reporter gene constructs in NIH3T3 cells (data not shown), and it inhibits Mix.1 function in animal cap assays (Fig. 6G). We note that in NIH3T3 cells our repressor fusions inhibited activation not only by Mix.1 but also by the highly related paired-type homeobox protein Bix.1 (Tada et al., 1998) (data not shown).
Having established that Mix.1-Engrailed repressor fusions act as transcriptional repressors and inhibit the function of Mix.1 in vitro, we tested their effects on Xbra and goosecoid expression during normal development. Embryos were injected at the 4-cell stage with pCMV-Mix.1-EnR or Mix.1-EnR RNA and analysed for goosecoid or Xbra expression by in situ hybridisation at mid to late gastrula stages. Injection of both constructs leads to an upregulation of Xbra (Fig. 5D and data not shown) and a suppression of goosecoid in a cell-autonomous fashion (Fig. 5F,G). The upregulation of Xbra was transient, and undetectable by stage 13. These observations are consistent with the suggestion that Mix.1 regulates expression of Xbra during normal development, and that this regulation occurs through activation of goosecoid and perhaps other transcriptional repressors.
We have also asked whether proteins such as Mix.1 and goosecoid participate in the repression of Xbra mediated by high doses of activin by using cycloheximide to block their translation. This treatment resulted in expression of Xbra even at high activin concentrations (not shown).
Interference with Mix.1 function causes deficiencies in anterior structures and in endodermal differentiation
Use of an antisense goosecoid construct and gscVP16 confirms that goosecoid activity is required for anterior patterning during Xenopus development (Fig. 2E-J). What is the role of Mix.1? Embryos injected dorsally at the 4-cell stage with pCMV-1.xiM, or RNA encoding antisense Mix.1, develop with dorsoanterior deformities, ranging from mild cyclopia to complete loss of head (Fig. 7). Most of these embryos also have abnormal gut morphology and defective heart formation, as judged by in situ hybridisation using the heart-specific marker XMLC2 (Fig. 7A-C). However, notochord was present in all specimens, indicating that these embryos are posteriorised rather than ventralised (Fig. 7F-H).
Embryos injected on their dorsal sides with RNA encoding Mix.1HD-EnR or Mix.1-EnR displayed similar but distinct phenotypes, with both constructs causing a reduction in dorsoanterior structures. We concentrate here on results obtained with Mix.1-EnR, since it is likely to be more specific (see Discussion). Certainly, the phenotypes of embryos injected with RNA encoding Mix.1-EnR are indistinguishable from those obtained with antisense Mix.1.
Injection of Mix.1-EnR RNA causes a reduction in anterior structures, varying from a slight decrease in head size and cement gland to complete loss of head. Intermediate phenotypes include cyclopia, together with a greatly reduced cement gland. There are in addition defects in posterior endoderm, including a reduction of gut size and inhibition of normal gut coiling (Fig. 7I-K), a phenotype also observed with injection of antisense Mix.1 RNA. Injection of 1.xiM RNA into the vegetal pole region of the embryo often results in deformities in the gut region, while injections in the dorsoequatorial region usually cause a combined head and gut phenotype. As with antisense Mix.1, injection of Mix.1-EnR RNA frequently interferes with heart formation, as revealed by in situ hybridisation using a probe specific for XMLC2.
Since Mix.1-EnR interferes with the formation of dorsoanterior endodermal tissues, we next asked whether wild-type Mix.1 is able to induce early endodermal markers in animal cap explants. While Mix.1 alone induces only very weak but reproducible expression of XSox17α in animal caps, this effect is greatly increased, in a synergistic manner, by co-injection of goosecoid RNA (Fig. 8). We also observe that Mix.1-EnR causes a downregulation of XSox17α expression in whole embryos (not shown).
Mix.1-EnR and M11
The results presented in this paper, like those of Lemaire et al. (1998) suggest that Mix.1 plays a role in development of the endoderm, a conclusion that contrasts with previous work indicating that the gene is required for differentiation of ventral mesoderm (Mead et al., 1996). The interfering Mix.1 construct used by Mead and colleagues (designated M11) introduces a proline between helices two and three of the homeodomain, a mutation that is thought to interfere with DNA binding (Mead et al., 1996). We compared the effects of Mix.1-EnR and M11 by injecting RNA encoding the two proteins into Xenopus embryos and dissecting animal caps at the mid-blastula stage. Animal pole explants injected with RNA encoding M11 form cement glands (Fig. 9; see Lemaire et al., 1998) and express the neural marker N-CAM (not shown), whereas those injected with Mix.1-EnR are indistinguishable from uninjected controls (Fig. 9).
The induction of cement gland and N-CAM expression by M11 suggests that the effects of this reagent are not limited to Mix.1, because Mix.1 is not expressed, to detectable levels, in animal pole tissue (Rosa, 1989). This is discussed below.
DISCUSSION
This work addresses the roles of goosecoid and Mix.1 in the control of Xbra expression and in germ layer specification in Xenopus. Our approach has been to impair the functions of goosecoid and Mix.1. In doing so, in an effort to ensure specificity, we have used two different approaches (antisense RNA as well as dominant-interfering constructs) and have considered our results significant only if the two methods give similar results. In order to achieve maximum specificity, our dominant-interfering constructs contain the entire open reading frames of goosecoid or Mix.1 rather than just the homeodomains. The homeodomain binds particular DNA sequences, but this is not sufficient to account for specificity of action in vivo, which is further refined by protein-protein interactions (Mann and Affolter, 1998). This consideration is particularly important in the case of Mix.1, which has recently been shown to be the founder member of a subfamily of at least six homeodomain-containing proteins with overlapping expression patterns and activities (Ecochard et al., 1998; Henry and Melton, 1998; Rosa, 1989; Tada et al., 1998; Vize, 1996). The dominant-interfering constructs were tested by characterising their DNA binding and transcriptional activities both in Xenopus and in a heterologous system and on a variety of promoters. These experiments have provided the first direct evidence that Mix.1 acts as a transcription activator and that Mix.1-EnR functions as an active repressor.
Consistent with previous work demonstrating that goosecoid and Mix.1 suppress expression of Xbra (Artinger et al., 1997; Latinkic et al., 1997), we find that inhibition of the function of these genes during normal development leads to ectopic expression of Xbra. This effect was transient, indicating that simple de-repression is not sufficient to cause stable activation of Xbra; rather, continued expression of the gene must require region-specific activation signals. The transient activation of Xbra was not able to induce ectopic tail formation, as is seen following the more stable activation of Xbra during gastrula stages that is obtained using hormone-inducible constructs (Tada et al., 1997). We note that activation of Xbra was not observed in a recent study which also investigated the consequences of inhibiting Mix.1 function in early development (Lemaire et al., 1998). This discrepancy may arise from the transient nature of ectopic Xbra activation, which makes it difficult to detect. Consistent with our observations, however, fusion of the Mix.1-like gene Mixer with the engrailed repressor domain proved to cause ‘higher and less concentrated’ expression of Xbra (Henry and Melton, 1998). Together, our results suggest that both goosecoid and Mix.1 play a role in the regulation of Xbra expression during normal development.
Comparing the functions of goosecoid and Mix.1
Our study focuses on the roles of goosecoid and Mix.1 in regulating expression of Xbra, but the results also address the roles of the two genes in the development of the whole embryo. Ours is not the first attempt to study the functions of these two homeobox-containing genes. For goosecoid, as mentioned in Results, Steinbeisser et al. (1995) have already used an antisense approach, and Ferreiro et al. (1998) have employed two ‘antimorphic’ constructs. The first of these ‘antimorphs’ uses an approach similar to our gscVP16, but the construct differs in that it lacks the N-terminal 113 amino acids of goosecoid and uses the entire VP16 activation domain. The second construct, and the main focus of the study, is a Myc-tagged version of goosecoid (MTgsc) which, surprisingly, acts as a powerful transcription activator (Ferreiro et al., 1998). The effects of MTgsc may, therefore, go beyond preventing goosecoid-mediated repression by inappropriately activating the expression of goosecoid target genes.
The phenotypes obtained in the three studies are broadly similar in that all display loss of head, but they differ in significant details. In particular, embryos obtained following expression of MTgsc lack a notochord and are described by Ferreiro et al. (1998) as ventralised. By contrast, notochord formation is normal in the embryos obtained in our study and in that of Steinbeisser et al. (1995) and are best described as posteriorised. These results show that goosecoid function is required in dorsoanterior mesendoderm and not in dorsal mesoderm.
The function of Mix.1 has been addressed by Mead et al. (1996), using the M11 construct in which a proline is inserted between helices 2 and 3 of the homeodomain, and more recently by Lemaire et al. (1998), who fuse the engrailed repressor construct to the Mix.1 homeodomain. These papers differ quite dramatically in their conclusions, with Mead et al. (1996) suggesting that Mix.1 is required for ventral mesoderm formation and Lemaire et al. (1998) arguing that it is needed for head development and endoderm formation. Our own data using an antisense approach and an engrailed repressor construct that includes the entire Mix.1 open reading frame agree with Lemaire and colleagues. Like these authors, we found that M11induces cement gland formation in animal caps. Therefore, it may be interfering with the functions of other homeobox-containing genes such as Xvent-1, Xvent-2 (Gawantka et al., 1995; Ladher et al., 1996) and msx1 (Suzuki et al., 1997). Our results do differ in one respect, however, from those of Lemaire et al. (1998), because we see no expansion of mesodermal tissues following interference with Mix.1 function.
We note that, although interference with Mix.1 function affects anterior development, posterior and ventral structures appear normal. Thus, although Mix.1 is expressed in ventral regions of the vegetal pole, it appears not to be required there. The effects of Mix.1 on Xbra and on dorsoanterior development may be mediated, at least in part, through its ability to amplify or maintain expression of goosecoid in anterior endodermal tissue. Thus, the two genes are transiently co-expressed in dorsoanterior endoderm at the early gastrula stage (Fig. 3B,C), Mix.1 can induce expression of goosecoid in animal caps (Fig. 3D) through direct binding to the goosecoid promoter (Fig. 4), and interference with Mix.1 function causes downregulation of goosecoid expression (Fig. 5F,G). These results support the proposed indirect mode of action of Mix.1. The effects of pCMV-Mix.1-EnR on Xbra expression provide further support for this model: if Mix.1 were acting directly on Xbra transcription, then Mix.1-EnR, an active repressor, would have been expected to downregulate Xbra, not activate it.
The phenotypes resulting from interference with Mix.1 and goosecoid function may be due to interference with normal gastrulation movements. Overexpression of wild-type goosecoid, for example, is known to cause inappropriate anterior migration of mesodermal cells (Niehrs et al., 1993). Furthermore, Mix.1 has recently been shown to cause adhesion of animal pole cells and goosecoid acts synergistically with Mix.1 to promote this effect (Wacker et al., 1998). This synergism is reminiscent of the effects of the two genes in inducing Xsox17α (Fig. 8) and is consistent with their ability to form heterodimers on the P3C site (Wilson et al., 1993).
The function of Mix.1 appears not to be restricted to the prospective head because embryos injected with interfering Mix.1 constructs also have defects in heart and gut development (Fig. 7). Unless Mix.1 protein is unusually long-lived, it is likely that Mix.1 is involved in the earliest steps of endoderm formation, because the gene is not expressed after the end of gastrulation (Henry and Melton, 1998; Rosa, 1989). The effects of interference with Mix.1 function on heart development may be indirect; heart development requires an inductive signal from the endoderm (Nascone and Mercola, 1995) and downregulation of Mix.1 function may affect this process.
Even though Mix.1 appears to be required for proper formation of the gut, simple misexpression of the gene is not sufficient to specify endoderm in animal pole tissue (this work and Lemaire et al., 1998). Rather, in combination with Siamois, Mix.1 induces expression of cerberus (Lemaire et al., 1998), which is expressed in anterior endoderm (Bouwmeester et al., 1996), whereas in combination with goosecoid it induces XSox17α (Hudson et al., 1997), a general endoderm marker (Fig. 8).
Together, our results provide evidence that Mix.1 and goosecoid promote endodermal differentiation while suppressing mesoderm, and are required for dorsoanterior development of the Xenopus embryo.
Acknowledgments
We dedicate this paper to the memory of Nigel Holder. This work was supported by the European Science Foundation (B. V. L.) and the Medical Research Council. J. C. S. was an International Scholar of the Howard Hughes Medical Institute. We are grateful to Niall Armes, Josh Brickman, Caroline Hill, Tim Mohun, Masazumi Tada, Hugh Woodland and Len Zon for cDNAs, and to Patrick Lemaire and Hugh Woodland for communicating results prior to publication. We also thank Niall Armes, Masa Tada and Derek Stemple for their helpful comments.