The Homeodomain Transcription Factor Xvent-2 Mediates Autocatalytic Regulation of BMP-4 Expression in XenopusEmbryos*

Like other genes of the transforming growth factor-β family, the BMP-4 gene is regulated by an autocatalytic loop. In Xenopus embryos this loop can be ectopically induced by injection of BMP-2 RNA. However, cycloheximide treatment subsequent to BMP-2 overexpression revealed that BMP signaling is not direct but requires additional factor(s). As putative mediator we have identified Xvent-2 which is activated by BMP-2/4 signaling and, in turn, activates BMP-4 transcription. Using promoter/reporter assays we have delineated Xvent-2 responsive elements within the BMP-4 gene. We further demonstrate that Xvent-2 which has recently been characterized as a transcriptional repressor can also act, context dependent, as an activator binding two copies of a 5′-CTAATT-3′ motif in the second intron of theBMP-4 gene. Replacement of Xvent-2 target sites within thegoosecoid (gsc) promoter by theBMP-4 enhancer converts Xvent-2 caused repression ofgsc to strong activation. This switch is obviously due to adjacent nucleotides probably binding a transcriptional co-activator interacting with Xvent-2. A model is presented describing the mechanism of BMP-4 gene activation in Xenopus embryos at the early gastrula stage.

Like other genes of the transforming growth factor-␤ family, the BMP-4 gene is regulated by an autocatalytic loop. In Xenopus embryos this loop can be ectopically induced by injection of BMP-2 RNA. However, cycloheximide treatment subsequent to BMP-2 overexpression revealed that BMP signaling is not direct but requires additional factor(s). As putative mediator we have identified Xvent-2 which is activated by BMP-2/4 signaling and, in turn, activates BMP-4 transcription. Using promoter/reporter assays we have delineated Xvent-2 responsive elements within the BMP-4 gene. We further demonstrate that Xvent-2 which has recently been characterized as a transcriptional repressor can also act, context dependent, as an activator binding two copies of a 5-CTAATT-3 motif in the second intron of the BMP-4 gene. Replacement of Xvent-2 target sites within the goosecoid (gsc) promoter by the BMP-4 enhancer converts Xvent-2 caused repression of gsc to strong activation. This switch is obviously due to adjacent nucleotides probably binding a transcriptional co-activator interacting with Xvent-2. A model is presented describing the mechanism of BMP-4 gene activation in Xenopus embryos at the early gastrula stage.
Induction and patterning of germ layers in vertebrate embryogenesis depends on intercellular signaling and intracellular signal transduction pathways triggered by various growth factors or growth factor-like molecules. Investigations performed with Xenopus laevis embryos have shown that bone morphogenetic protein 4 (BMP-4), 1 a member of the transforming growth factor-␤ superfamily, is a key signal for ventralizing the mesoderm, for the inhibition of dorsalizing and neuralizing factors and for converting ectodermal to epidermal cell fate (1,2). Thus it is evident that the activation of the BMP-4 gene at the ventral side of late blastula/early gastrula stage in Xenopus embryos is of general importance for dorso/ventral pattern formation. However, the molecular nature of factors being responsible for the zygotic activation of this gene in vivo is still not clear. Even in the case of invertebrate homologues, like the Drosophila gene decapentaplegic (dpp), the mechanism governing the initial activation at blastoderm stage is not completely understood. It could be shown that the second intron contains elements which contribute to the correct spatial blastoderm pattern (3)(4)(5). A gradient of the protein dorsal, the homologue of c-Rel in vertebrates, suppresses dpp in the ventral half, but the factors involved in transcriptional activation of dpp in the dorsal half remain to be elucidated. Subsequent expression of dpp in visceral mesoderm is regulated by ultrabithorax (ubx) which itself is up-regulated by dpp (6,7) and, at dorsal closure, the dpp target Fos (FosD) cooperates with Jun (JunD) by regulating the expression of dpp (8).
Autoregulatory loops have also been reported for transforming growth factor-␤ (9) and Xenopus BMP-4 (10). While autoinduction of transforming growth factor-␤ involves activatory protein 1, it is unknown, whether the autoregulatory loop of BMP-4 is direct or requires additional factors. We here show that the activation of the Xenopus BMP-4 gene depends upon BMP signaling, but this activation is not observed in the presence of cycloheximide, i.e. in the absence of protein synthesis. In search of putative mediators we have analyzed the role of transcription factors activated in ventral mesoderm, like the homeodomain proteins Xvent-1 (closely related to Xvent-1B and PV.1) (11-13), Xvent-2 (identical or closely related to Vox, Xom, Xbr, and Xvent-2B) (12, 14 -17), and the zinc finger factor GATA-2 (18,19). While all these genes are known to be activated by ectopic expression of BMP-4, only Xvent-2 up-regulates BMP-4 transcription in vivo, and hence is a candidate to function within the autocatalytic loop.
Deletion mutant/reporter gene assays of the Xenopus BMP-4 gene (20,21) have shown that enhancer elements located within the second intron and the 5Ј-flanking region contribute to transcriptional activation by BMP signaling. We now have further delineated these regions and demonstrate by co-injection experiments that the same regions which respond to BMP signaling are activated by Xvent-2. The direct interaction of Xvent-2 with a corresponding target site within the second intron was demonstrated by mobility shift and DNase I footprint experiments. In contrast to the previous characterization of Xvent-2 as a transcriptional repressor (22)(23)(24) we here document that this factor can additionally work as a transcriptional activator. This dual activity is context dependent and obviously requires a co-activator interacting with Xvent-2 and binding to an adjacent target site. The results suggest a model, in which the autoregulatory loop of BMP-4 is triggered by maternal BMP-2 activating Xvent-2 and is subsequently maintained by BMP-4 via Xvent-2 as a mediator. It is consistent with the observation that Xvent-2 and BMP-4 show identical spatial expression patterns throughout embryogenesis (14,25).

EXPERIMENTAL PROCEDURES
Preparation of Deletion Mutants-All deletion mutants were synthesized as described previously (20) by PCR using DNA of phage XB as template. To facilitate the directed cloning into the luciferase pGL3 basic vector (Promega), the forward primers were 5Ј elongated with a restriction site for BamHI (upstream) or KpnI (intron) and all reverse primers with one for HindIII (upstream) or SmaI (intron). Indicated numbers for all mutants refer to the BMP-4 gene sequence (EMBL AC: AJ005076). Goosecoid promoter DNA fragments were amplified from genomic DNA by using primers which were either KpnI or XhoI elongated (forward) and either XhoI or HindIII elongated (reverse). Forward Ϫ226: 5Ј-GGGGTACCCATTAATCAGATTAACGGTGAGC-3Ј; forward Ϫ103: 5Ј-CCGCTCGAGGATGAGTCTCCTCTCACCCC-3Ј; reverse ϩ3: 5Ј-CCCAAGCTTGTCCTCTCCCATCTGTGCCTC-3Ј; reverse Ϫ128: 5Ј-CCGCTCGAGCAAACTAATCCACTCCATTAGG-3Ј. Fragments were subcloned either directly or as cassettes in combination with corresponding fragments of the BMP-4 gene in pGL3.
Whole Mount in Situ Hybridization-RNA probes were synthesized from cDNA templates using the DIG labeling kit (Roche Molecular Biochemicals). Whole mount in situ hybridizations were performed with staged embryos as described (26)  Embryo Injections and Luciferase Assay-5Ј promoter deletion and intron-2 fusion constructs were injected at 20 pg/blastomere into twocell stage embryos in both blastomeres or into four-cell stage embryos in the dorsal or in the ventral blastomeres. When indicated, RNA was co-injected into individual blastomeres. Embryos were collected at stage 12.5 (staging according to Nieuwkoop and Faber (27)) and frozen in liquid nitrogen. Embryos injected with luciferase reporter constructs were processed as described previously (20).
Cycloheximide Treatment-Embryos were injected at the four-cell stage with indicated RNAs and grown until stage 7. Cycloheximide concentration for whole embryos was determined to be sufficient at 30 g/ml. Embryos were kept in cycloheximide until control embryos had reached stage 10.5 and then fixed for in situ whole mount hybridization.
Yeast Transcriptional Assay (28)-Xvent-1 and Xvent-2 were integrated in-frame into the BamHI recognition site of the bacteria/yeast shuttle vector pAS2 (CLONTECH). Subsequently, the constructs have been transformed according to established procedures (yeast protocols handbook, CLONTECH) into yeast strain Y187 employing selection markers trp1 and leu2 (29) and plated on drop-out plates missing tryptophan. Colonies were grown overnight in minimal medium in the absence of tryptophan and the next day inoculated in YEPD medium. Cells were harvested at 0.6 OD (600 nm), disintegrated, and assayed for reporter gene activity by means of a colorimetric assay employing ONPG (o-nitrophenyl-␤-D-galactopyranoside) as a substrate at 420 nm (30). The amount of expression was verified by Western blot analysis due to a hemagglutinin A epitope tag provided by pAS2 and by using mouse anti-HA monoclonal antibodies (Roche Molecular Biochemicals). Visualization was performed with goat anti-mouse antibodies using the ECL detection kit (Amersham Pharmacia Biotech).
Mobility Shift Assays and DNase I Footprinting-The isolation procedure of bacterially expressed Xvent-2 homeodomain protein and the conditions for mobility shift and DNase I footprint experiments are as described previously (31).

Activation of BMP-4 Expression in the Gastrula Stage
Embryo-Since the related proteins BMP-2 and BMP-4 are known to bind and activate the same receptors (32, 33), we first investigated whether the BMP-4 autoregulation in Xenopus might be initiated by maternal BMP-2. Injection of BMP-2 RNA into the dorsal blastomeres of four-cell stage embryos with subsequent in situ whole mount hybridization for BMP-4 transcripts reveals ectopic activation of BMP-4 in the most dorsal mesoderm, the dorsal lip or Spemann organizer (Figs. 1, A and B). In contrast, radial injection of the truncated BMP type I receptor RNA leads, in a dose dependent response, to an inhibition of BMP-4 transcription (Fig. 1C), thereby demonstrating that BMP signaling is required for BMP-4 gene activation. Thus, the initial activation of the BMP-4 autoregulatory loop could be triggered by endogenous BMP-2. The BMP-2 gene is maternally transcribed and high levels of transcripts are detected until the early gastrula stage (34,35); BMP-2 protein is present at the late blastula/early gastrula stage (36, 37), 2 when BMP-4 transcription is initiated.
However, treatment of BMP-2 injected embryos with cycloheximide (CHX) prior to midblastula transition prevents transcription of the BMP-4 gene (Fig. 1D). As controls we used the Xvent-2 gene (Fig. 1E) which is a direct target for BMP signaling and the Xvent-1 gene (Fig. 1F) which does not directly respond to BMP-2/4 (12). Thus it seems that the autoregulatory loop is not direct but requires additional newly synthesized proteins. Such factors should be activated by BMP-2/4 signaling and, in turn, up-regulate the BMP-4 gene. Xvent-1 was excluded, since it is neither directly activated by BMP-4 ( blastomeres of four-cell stage embryos which were subsequently analyzed for the presence of BMP-4 transcripts. Fig. 1, G and J, show that in case of Xvent-2 ventral and lateral expression of BMP-4 is expanded and leads to circumferential expression around the blastoporus, but not in the case of GATA-2. Vice versa, dorsal injections of BMP-4 RNA or BMP-2 RNA lead to a strong increase of Xvent-2 or GATA-2, respectively (Fig. 1, I and L). In conclusion, only Xvent-2, but not GATA-2, can be regarded as potential activator of the BMP-4 gene and might participate in the autocatalytic regulation of this gene.
Upstream and Intron Enhancers Regulate the BMP-4 Gene-We have previously shown by injection of promoter/ reporter DNA constructs that the Xenopus BMP-4 gene is activated by several enhancer elements within the 5Ј-flanking region and within the second intron (20). The upstream region between Ϫ206 and Ϫ156 responded to BMP signaling, in that a Ϫ206/ϩ54 mutant but not a Ϫ156/ϩ54 mutant showed a significant increase in reporter gene activity after co-injection with BMP-4 RNA and a down-regulation following co-injection with truncated BMP receptor RNA. Noteworthy, a mutant containing 116 nucleotides of the upstream region fused to complete intron-2 exhibited higher reporter gene activity than approximately 5 kilobases of the 5Ј-flanking region. Thus, it is evident that intron-2 contains an enhancer which significantly contributes to the activation of the BMP-4 gene. Moreover, this mutant was stimulated by co-injection with either BMP-4 or BMP-2 RNA, whereas co-injection with chordin or truncated BMP receptor RNA led to a drastic decrease of reporter gene activity.
By using serial deletion mutants we have now further delineated the nucleotide element which contributes to the observed activity. Fig. 2A shows the deletion mutants which have been used and a compilation of reporter activities generated by various deletions of the upstream or intron regions after injection into both blastomeres of two-cell stage embryos in the absence or presence of co-injected BMP-2/4 RNA. While the 5Ј border of the BMP responsive element within the upstream region was  Fig. 2B). B and C, while coinjections of the ϩ1815 (B) or ϩ1738/ ϩ1969 intron mutant (C) with 400 pg of Xvent-2 yield more than 2-fold increase of reporter activity, 400 pg of Xvent-1 has no effect. Co-injection of 1 ng of Xvent-2 (P40) or 1 ng of chordin RNA leads for both of these mutants to a significant decrease of reporter activity, which can be rescued by 400 pg of Xvent-2 RNA. Note also, that the activation obtained with 300 pg of BMP-2 RNA is strongly inhibited by co-injection with 1.3 ng of Xvent-2 (P40) RNA. reduced to 20 bp between positions Ϫ176 and Ϫ156 (Fig. 2B), major activity of intron sequences was localized between positions ϩ1815 and ϩ1918 (Fig. 2C).
Xvent-2 as Mediator of BMP Signaling-We next investigated whether the observed stimulations by BMP-2/4 can also be obtained by Xvent-2. Fig. 3A shows reporter gene activities determined after co-injection of 5Ј flanking and intron mutants with Xvent-2 RNA. Both the upstream and the intron mutants behaved similarly or even identically when co-injected with BMP-4 or Xvent-2 RNA, respectively (compare with Fig. 2, B and C). By using a 200-bp intron deletion mutant (I-2⌬ϩ1750/ ϩ1951) we observe a reduced basal activity and a loss of stimulation both for BMP-2/4 (not shown) as well for Xvent-2 (Fig.  3A). These findings suggest that Xvent-2 serves as the required component in mediating the autocatalytic regulation of the BMP-4 gene. Additional support for this notion was rendered by the finding that the ϩ1815 mutant was strongly inhibited both by chordin or a dominant negative Xvent-2 mutant (P40) (22), respectively, but could be rescued in both cases by coinjection with wild type Xvent-2 (Fig. 3B). Moreover, the activation obtained with BMP-2 was drastically reduced by coinjection with the P40 mutant. This negative interference supports the notion that Xvent-2 is required for the activation of BMP-4, and it is consistent with the observation that Xvent-2 (P40) inhibits BMP-4 transcription in vivo (data not shown).
The activation observed with Xvent-2 is specific as it was never observed with Xvent-1. This result does not agree with previous studies using the Xvent-1 related PV.1 (21) but is consistent with the finding that only Xvent-2, but not Xvent-1, is able to activate BMP-4 transcription (38). An intron mutant comprising only 232 nucleotides (ϩ1738 to ϩ1969), but including the BMP response element, behaved similarly, in that it was activated by Xvent-2 and inhibited by chordin or Xvent-2 (P40), respectively (Fig. 3C). In conclusion, these results are consistent with Xvent-2 as a mediator of BMP signaling on the Ϫ176/Ϫ156 upstream and ϩ1815/ϩ1951 intron regions.
Xvent-2 Interacts with a BMP-4 Gene Enhancer-To test whether the biological effects observed upon co-injection of deletion mutants with Xvent-2 RNA reflect direct binding of this homeobox protein to DNA we have performed gel retardation assays and DNase I footprint analyses with the Xvent-2 homeodomain. While bacterially expressed Xvent-2 homeodomain readily shifts the intron 2 region which had been shown to elicit a biological response (Fig. 4A), we repeatedly failed to demonstrate such an behavior with the corresponding upstream region (data not shown). This result can be explained by the assumption that an additional factor which might be synthesized under control of BMP signaling via Xvent-2 is binding to this region. Fig. 4B shows a DNase I footprint of the Xvent-2 homeodomain protein on both strands of the intron 2 region, for which we could demonstrate a stimulation upon Xvent-2 co-injection and band retardation. The protected region corresponds for both strands and contains two copies of a motif 5Ј-CTAATT-3Ј. This binding site defines an AT-rich homeobox target sequence and it is rather similar to the 5Ј-CTATTT-3Ј motif we have previously shown to bind to Xvent-1 (31). Even more important, this motif is fully compatible to the recently published Xom (Xvent-2) target sequence derived from random oligonucleotides by PCR selection (24) revealing two copies of a TAAT/ ATTA motif separated by six or seven nucleotides and the first copy most frequently being found as 5Ј-CTAATT-3Ј. In the case of the BMP-4 enhancer, the distance was found to be six nucleotides, but the core motif 5Ј-CTAATT-3Ј is two times directly repeated, whereas the PCR approach is reported to yield in 75% of investigated sequences in antiparallel orientation. Search for this element in the human and mouse BMP-4 genes (39 -42), which contain 5 instead of 3 exons, revealed the existence of a conserved 13-bp element within the proximal promoter region (Fig. 4C). This region has also been implicated in transcriptional activation (40,42). However, mammalian orthologues to Xvent-2 have so far not been isolated; thus, it remains an open question whether these sites are necessary in mammals and whether they bind to similar or other homeodomain proteins.
The Core Enhancer is Required for Binding but Not Suffi-  (25,75, and 225 ng) of Xvent-2 HD were added prior to DNase I digestion in lanes 3-5. Protected regions (indicated by lines) correspond for both strands and contain two copies of the motif 5Ј-CTAATT-3Ј (underlined). Gene sequence is shown from nucleotide positions ϩ1820 to ϩ 1879. C, best fit alignment of the X. laevis Xvent-2-binding site with the human and mouse BMP-4 gene (number of matches: 13; allowed mismatches: 1). Note that a 13-bp match (including one T to C transition) is found within the proximal promoter of both the human (accession number U43842) and the mouse BMP-4 genes (accession number L47480). cient for Activation-We have further delineated the enhancer by additional 5Ј and 3Ј deletions to a 62-bp fragment extending from nucleotide positions ϩ1823 to ϩ1884. This fragment is shifted by the Xvent-2 homeodomain and, after fusion to the Ϫ116/ϩ54 basal promoter, results in a distinct increase of reporter gene activity upon co-injection of Xvent-2 (Fig. 5). Thus, all molecular and biological data suggest that this region contributes to the activation and autocatalytic regulation of the BMP-4 gene. Accordingly, deletion of the two core motifs (ϩ1842 to ϩ1857) led to a loss of binding, to reduced basal activity, and loss of stimulation by Xvent-2. Interestingly, dissection of the 62-bp fragment into a 5Ј fragment (28 bp: ϩ1823/ ϩ1850) and a 3Ј fragment (33 bp: ϩ1852/ϩ1884) revealed that, while both fragments contain a core motif and still bind to the Xvent-2 homeodomain, only the 5Ј fragment responds to coinjection of Xvent-2. These findings demonstrate a necessity of 5Ј-flanking nucleotides for transcriptional activation and that the core motif is required for Xvent-2 binding but not sufficient to up-regulate the BMP-4 gene.
Xvent-2 Serves as Transcriptional Activator-Recent reports demonstrated that constructs containing VP16 or GAL4 activation domains (GAL4-AD) fused to Xvent-2 behave as antimorphs in acting as repressors for ventral and as activators for dorsal genes (22)(23)(24). Accordingly, Xvent-2 was postulated to serve as a repressor. To explain the activation of ventral genes by Xvent-2, a mechanism was suggested according to which Xvent-2 acting as a repressor inhibits transcription of a yet unknown ventral suppressor.
Since these conclusions apparently contradict our present model for Xvent-2 as an activator, we have prepared a GAL4-AD/Xvent-2 fusion construct in pCS2 (23) to analyze the effect of microinjected RNA on the BMP-4 gene promoter. Our results confirm the previous findings, but the following experiments demonstrate that Xvent-2 has a dual role in that it can also serve as transcriptional activator. First, all intron deletion mutants lacking the Xvent-2 responding region show a reduced activity as compared with the wild type sequence which is not compatible with a suppressor model. Second, while ventral injection of GAL4-AD/Xvent-2 fusion construct leads to an already described duplication of posterior axis (23), dorsal injection causes a loss of anterior structures suggesting ventralizing activity (Fig. 6A). Third, co-injection of GAL4-AD/Xvent-2 at low concentration with the intron-2/reporter mutant leads to a stimulation of reporter activity, whereas, at higher concentration, an inhibition is observed (Fig. 6B). Fourth, this concentration dependent dual effect can also be observed for the wild type BMP-4 gene. In situ hybridizations show that low concentrations lead to a weak but distinct activation at the dorsal side, while high concentrations applied to the ventral side even lead to clearance of BMP-4 transcripts (data not shown). Thus, at least at low concentrations, this construct behaves as an activator of BMP-4 gene transcription. Fifth, the difference between Xvent-2 and Xvent-1 regarding their activatory potential was finally also demonstrated in the yeast GAL4 assay (28). Fusions of the two proteins to the GAL4 DNA-binding domain revealed that Xvent-2, but not Xvent-1, like the GAL4  (1), ϩ1823 to ϩ1850 (2), ϩ1852 to ϩ1884 (3), and ϩ1823/ϩ1884 deleted from ϩ1842 to ϩ1857 (4) were subjected to gel mobility shifts with the Xvent-2 homeodomain as described in the legend to Fig. 4. Note that there is no shift when both copies of the core motif (underlined) are deleted but that one copy is sufficient for binding. C, fragments were subcloned in front of the Ϫ116/ϩ54 basal BMP-4 promoter fused to the luciferase reporter. 20 pg of DNA constructs were co-injected with 500 pg of Xvent-2 RNA into dorsal blastomeres of four-cell stage embryos. Reporter activity of the complete fragment is set as 100%. Note that neither the internal deletion nor the 3Ј fragment containing one copy of the core motif can be activated by  activation domain is a very potent activator of the LacZ reporter gene (Fig. 6C). This result is not simply due to a lower concentration of Xvent-1 fusion protein as demonstrated by immunoblotting using a monoclonal hemagglutinin antibody. In summary, all these results support the notion that Xvent-2, besides its known suppressor function, can also act as a transcriptional activator.
Activation of the Goosecoid Promoter by the BMP-4 Intron Enhancer-It has recently been shown that dorsal activity of the goosecoid (gsc) gene is repressed by Xom (Xvent-2) and that Xom as transcriptional repressor directly interacts with several target sites located between Ϫ128 to Ϫ226 on the gsc promoter (24). Since Xvent-2 in the case of the BMP-4 gene acts as transcriptional activator, we have asked whether distinct features of the binding region or the basal promoters are responsible for this differential behavior. The corresponding region of the gsc promoter was replaced by the intron enhancer of the BMP-4 gene and, vice versa, the Xvent-2-binding region of the gsc promoter was fused to the basal BMP promoter. These constructs were co-injected with Xvent-2 into dorsal blastomeres of four-cell stage embryos. Fig. 7 demonstrates that the BMP-4 intron enhancer converts Xvent-2 which caused repression of gsc promoter activity to strong activation, whereas the Xvent-2 binding region of gsc leads to an inhibition of the basal BMP-4 promoter. These results convincingly demonstrate the dual role of Xvent-2 as repressor and activator and they render additional evidence that nucleotide motifs adjacent to the core target sites of Xvent-2 determine the mode of regulation. Alternatively, the direct orientation of two core motifs within the intron enhancer instead of antiparallel orientation within the gsc promoter region could explain this differential behavior; however, we estimate this assumption to be less likely, because we observed stimulation already with only one copy of the core motif of the intron enhancer including its 5Ј-flanking region (see Fig. 5).
Xvent-2 Requires a Co-factor for Transcriptional Activation of the BMP-4 Gene-The experimental results suggest that the activatory function of Xvent-2 is context-dependent and requires additional sequence motifs which probably bind to a co-activator. If this hypothesis holds true, we would expect that the Xvent-2 function by itself is not sufficient to up-regulate BMP-4 expression. We therefore have analyzed the ability of Xvent-2 to activate BMP-4 gene transcription in the presence of CHX. RT-PCR of RNA from Xvent-2-injected embryos and from uninjected control embryos grown in the absence or presence of CHX (Fig. 8) as well as an in situ whole mount hybridization of corresponding embryos (data not shown) clearly demonstrate that CHX, in contrast to untreated embryos, significantly diminishes or even prevents BMP-4 transcription. This result allows the conclusion that activation of BMP-4 by Xvent-2 requires a co-activator which is either de novo synthesized or whose recruitment is blocked by treatment with CHX. An alternative could be that Xvent-2 directs synthesis of another homeodomain transcription factor which binds to the same AT-rich motif as found for Xvent-2. However, the demonstrated binding of Xvent-2 to the BMP-4 intron enhancer makes such a hypothesis less likely, but certainly does not rule out binding of other related factors. DISCUSSION We here show that zygotic activation of the BMP-4 gene at late blastula/early gastrula stage can be triggered by BMP-2 which, in vivo, is translated from maternal transcripts being present until the gastrula stage within the embryo (35). In Higher amounts lead to reversal or even to inhibition, as shown by ventral co-injection with 1.5 ng of perblastomere. C, a yeast-based assay was used to determine the transactivatory efficiency of Xvent-1 and Xvent-2. Complete proteins were fused to the GAL4 DNA-binding domain and expressed in pAS2 under control of the ADH promoter. The recombinants were transfected into yeast strain Y187, thereby prompting an interaction between the GAL4 fusions and the genomic GAL1-binding site giving rise to the activation of the LacZ reporter. Efficiency of fusion protein synthesis was controlled by immunoblotting of yeast extracts using monoclonal hemagglutinin antibodies. The enzymatic activity of LacZ was determined after 1 h in arbitrary units relative to the activatory strength of the GAL-4 activation domain.
addition, maternal BMP-4 transcripts being present at a very low level (34,43) could also be translated during early cleavage stages and contribute to zygotic activation of the gene. In line with this, injection of truncated BMP receptor prevents formation of BMP-4 transcripts at the gastrula stage. Thus it is likely that BMP signaling is a major component not only in the maintenance but also in the activation of the Xenopus BMP-4 gene. Similar conclusions can be drawn from zebrafish mutants. At least the maintenance of zBMP-2 and zBMP-4 is affected in swirl (zBMP-2) mutant embryos; zBMP-4 expression in the ventral marginal region depends on zBMP-2 as indicated by the reduced initial expression and subsequent loss of the marginal zone zBMP-4 expression in swirl mutant embryos (44). Also, the requirement of BMP signal transducers has been documented. BMP2b and Smad5 (somitabun: sbn) double mutant analysis and RNA injection experiments have shown that sbn acts downstream of BMP2b signaling to mediate BMP2b autoregulation during early dorsoventral pattern formation (45).
However, cycloheximide treatment of BMP-2-injected Xenopus embryos prior to midblastula transition prevents BMP-4 transcription; thus it seems clear that the activation and/or the autoregulatory loop are not direct but indirect. As putative mediators we have investigated several genes which are activated by BMP-2/4. In turn, corresponding proteins should also be able to activate the BMP-4 gene when overexpressed within the embryo. We found that Xvent-2, but not Xvent-1, GATA-2, or Xwnt-8 (46) 2 fulfill both of these requirements. Therefore, the results suggest that Xvent-2 might be directly involved in the transcriptional regulation of the BMP-4 gene. Also, the activatory potential of Xvent-2 for BMP-4 as well as for Xvent-1 transcription (12) strongly suggests that Xvent-2 does not only work as a repressor as recently suggested (22)(23)(24), but additionally serves as a transcriptional activator. We here demonstrate that the GAL4 activator domain/Xvent-2 fusion protein behaves for the BMP-4 promoter in a dose-dependent manner as a transcriptional activator which then is converted to a repressor at higher concentrations. The inhibition of ventral genes observed after injection at high concentrations might be explained by an artificial activation of dorsal genes, like goosecoid and chordin, which are known to suppress ventral genes. Also, a comparison of the activatory potential of Xvent-1 to that of Xvent-2 in the yeast system clearly indicates that Xvent-2, in contrast to Xvent-1, behaves as an activator. Finally, swapping of the BMP-4 intron enhancer to the gsc promoter converts Xvent-2 from acting as a repressor to a transcriptional activator. However, the activatory potential of Xvent-2 on BMP-4 gene transcription requires an additional co-activator, because FIG. 7. BMP-4/goosecoid promoter swapping. The goosecoid promoter has recently been shown to contain several Xom (Xvent-2)-binding sites (BS) located between Ϫ128 and Ϫ226, thereby mediating Xvent-2 caused repression of activin A-stimulated promoter/reporter activity (24). In contrast, the intron 2 enhancer (ϩ1815/ϩ1892) fused to the basal BMP-4 promoter yields a strong activation upon co-injection of Xvent-2. Promoter swapping of corresponding gsc and intron 2 enhancer regions reveals that not the basal promoters, but that the Xvent-2-binding sites are responsible for repression or activation, depending on the context. Reporter activities determined after injection of 20 pg of DNA constructs into dorsal blastomeres at the four-cell stage are set as 100%, the values obtained upon co-injection of 500 pg of Xvent-2 RNA are given as relative percentage.
Noteworthy, a comparison of the temporal expression of Xvent-2 (14) versus that of BMP-4 (43,48) also supports the notion that Xvent-2 serves as a regulatory component for the zygotic activation of BMP-4 gene transcription. Actually, Xvent-2 transcription clearly precedes that of BMP-4 and the spatial pattern of Xvent-2 which is almost identical to that of BMP-4 is consistent with its role in BMP-4 gene activation.
To localize enhancers on the BMP-4 promoter which interact with factors being involved in the autoregulatory loop, we have investigated a series of deletion mutants from the 5Ј-flanking region and the second intron fused to a luciferase reporter gene by co-injection with BMP-2/4, chordin, Xvent-2, and dominant negative Xvent-2 (P40) RNA. The results demonstrate the existence of two sites, one upstream and one in intron 2, responding to Xvent-2. Comparing reporter gene activities obtained by co-injection with BMP-2/4 to those obtained with Xvent-2 we find a correlation for all mutants tested regarding their potential and their extent of stimulation. Thus, on a qualitative and a quantitative level, Xvent-2 mimics the action of BMP-2/4 and seems to be the major player in the autocatalytic loop. Also, down-regulation observed with chordin directly correlates to BMP-2/4 reactive mutants and can be rescued by Xvent-2. This finding supports our previous notion that transcriptional repression of BMP-4 by chordin is solely due to chordin/BMP-4 interaction at the protein level, thereby interfering with the autoregulatory loop (20).
Finally, we have investigated the ability of Xvent-2 to interact with DNA motifs found to be essential in reporter gene activation. We demonstrate that the Xvent-2 responsive element in intron 2 is retarded by the Xvent-2 homeodomain, but we failed in mobility shifts using the upstream Xvent-2 responsive region. Thus, we have to conclude that the action of Xvent-2 on the upstream region is not direct but indirect and requires the action of another, further downstream factor. The intron target was subjected to DNase I footprinting. The result corresponds for both strands and reveals a duplicated 5Ј-CTA-ATT-3Ј motif as a target motif for Xvent-2. This strongly supports previous findings of a Xom (Xvent-2) target consensus sequence derived by a PCR-based oligonucleotide selection procedure containing exactly this motif (24). The fact that the biological effects observed in reporter gene activation assays coincide with the presence of this element strongly supports the notion that this motif serves as a natural Xvent-2-binding site. Moreover, it displays a high degree of conservation to the 5Ј-CTATTT-3Ј motif, which we have recently described as a Xvent-1 target site within the XFD-1Ј promoter (31).
In summary, we show both for the wild type gene and for promoter/reporter constructs that the autocatalytic regulation of the BMP-4 gene is mediated by Xvent-2. Results obtained from biological and molecular investigations are compatible with the notion that the autoregulatory loop of BMP-4 is initially triggered by maternal BMP-2 signals activating Xvent-2 and maintained during early development by Xvent-2. The major contribution of intron 2 to transcriptional activation of BMP-4 coincides with two copies of a 5Ј-CTAATT-3Ј target motif which have the potential to bind to Xvent-2. Although our results do definitely not rule out the possibility that other factors being activated by BMP-4, e.g. msx1 or Xvex-1 (49,50), might participate in the autocatalytic loop, we provide the first insight into the regulatory mechanisms governing the transcription of the BMP-4 gene in Xenopus embryos.