Conservation and Diversification of an Ancestral Chordate Gene Regulatory Network for Dorsoventral Patterning

Formation of a dorsoventral axis is a key event in the early development of most animal embryos. It is well established that bone morphogenetic proteins (Bmps) and Wnts are key mediators of dorsoventral patterning in vertebrates. In the cephalochordate amphioxus, genes encoding Bmps and transcription factors downstream of Bmp signaling such as Vent are expressed in patterns reminiscent of those of their vertebrate orthologues. However, the key question is whether the conservation of expression patterns of network constituents implies conservation of functional network interactions, and if so, how an increased functional complexity can evolve. Using heterologous systems, namely by reporter gene assays in mammalian cell lines and by transgenesis in medaka fish, we have compared the gene regulatory network implicated in dorsoventral patterning of the basal chordate amphioxus and vertebrates. We found that Bmp but not canonical Wnt signaling regulates promoters of genes encoding homeodomain proteins AmphiVent1 and AmphiVent2. Furthermore, AmphiVent1 and AmphiVent2 promoters appear to be correctly regulated in the context of a vertebrate embryo. Finally, we show that AmphiVent1 is able to directly repress promoters of AmphiGoosecoid and AmphiChordin genes. Repression of genes encoding dorsal-specific signaling molecule Chordin and transcription factor Goosecoid by Xenopus and zebrafish Vent genes represents a key regulatory interaction during vertebrate axis formation. Our data indicate high evolutionary conservation of a core Bmp-triggered gene regulatory network for dorsoventral patterning in chordates and suggest that co-option of the canonical Wnt signaling pathway for dorsoventral patterning in vertebrates represents one of the innovations through which an increased morphological complexity of vertebrate embryo is achieved.


Introduction
Establishment of a dorsoventral (DV) axis is a key event in early development of any bilaterian animal embryo. The crucial step in DV axis formation is specification of the dorsal and ventral mesoderm. In vertebrates, the establishment of the organizer involves activation of several genes [1]. Their protein products are mostly transcription factors (such as Otx2, XFD1, Goosecoid) or secreted proteins (such as ADMP, Nodal, Noggin, Chordin). The organizer secreted proteins Chordin and Noggin are capable to inactivate BMP signaling molecules that play a key role in the induction and maintainance of ventral and lateral mesoderm. Conversely, the expression of Chordin is negatively regulated by Bmp2 and Bmp4 proteins through their targets, ventralizing homeobox genes Vent and Vox [1]. Recently it was demonstrated that the expression patterns of genes mediating DV patterning in early development are highly conserved between basal chordates (cephalochordate amphioxus) and vertebrates [2]. Orthologues of the vertebrate organizer-specific genes such as Goosecoid, Chordin, Nodal are expressed in early chordate embryo [2]. Amphioxus ventral-specific genes encoding Bmp signaling molecules, and Hex, Evx and Vent transcription factors demonstrate expression patterns homologous to their vertebrate counterparts [2,3]. It was shown previously that teleost and amphibian Vent proteins can suppress the expression of dorsal genes during early development [4,5,6]. Xvent-2 (also known as Xvent-2B, Xom, Xbr-1 and Vox) directly represses the Goosecoid promoter in Xenopus embryo [6]. Noting mutually exclusive expression of AmphiVent1 and Amphi-Chordin [2], it can be suggested that AmphiVent1 is likewise able to antagonize expression of organizer-specific genes as do its vertebrate homologues [7]. During the gastrula stage, AmphiVent1 is expressed throughout the mesendoderm [3]. By late gastrula, it is down-regulated ventrally but remains expressed dorsolaterally in the paraxial mesoderm. Then, at the mid-neurula stage, AmphiVent1-expressing ventral mesoderm forms as outgrowth from the somites [3,8]. The developmental expression of amphibian and teleost Vent genes during gastrula stages is most conspicuous in ventral mesoderm and is down-regulated in the regions of organizer, chordamesoderm and neural plate [5,7]. At the neurula stage, amphioxus as well as vertebrate Vent genes are expressed along the edges of the neural plate, in the tail bud/ proctodeal region, and in the foregut [3,9]. Even though there appears to be a temporal difference between the ventral expression of AmphiVent1 and vertebrate Vent genes during early develop-ment, their dorsal expression is similar as exemplified by downregulation at the dorsal lip of the blastopore and neural plate [3]. It is interesting to note that within the animal kingdom Vent genes are present in chordates only. Moreover, although in humans the Vent-like homeobox gene has been described, no Vent gene has been found in the mouse. Amphioxus genome contains two Vent genes, which are situated on the same chromosome in close proximity of each other [10].
Multiple transcriptional inputs are likely required for the correct regulation of Vent genes. Among those Bmp-mediated activation of vertebrate Vent genes is well documented. Bmp2 and Bmp4 activate Xvent-2 promoter via Smad1 in Xenopus [11,12] and in P19 murine embryonal cells [13]. This activation is mediated synergistically by OAZ zinc finger transcription factor, which can interact with MH2 domains of Smad1 and Smad4 proteins in response to the Bmp signal [13]. Recent investigations reveal that in addition to Bmp, the canonical Wnt signaling pathway plays an important role in patterning of ventral mesoderm in Xenopus and zebrafish. In the zebrafish embryo Wnt8 directly activates Vent and Vox genes through b-catenin [14]. Both Xenopus Xvent-1 and Xvent-2 genes contain conserved Lef/Tcf binding sites in the promoter. Xwnt-8 protein can activate Xvent-1 promoter and the activation depends on the functional Lef/Tcf binding site [15]. Likewise, transgenic analysis of Xvent-2 promoter revealed that mutation of the Lef/Tcf binding site decreases expression of the reporter gene [16].
In this study we have investigated the role of AmphiVent1 homeodomain protein in the molecular events responsible for DV patterning in amphioxus. We have specifically focused on three main areas: the role of Bmp and canonical Wnt signaling in AmphiVent1 gene regulation, functional properties of Vent proteins, and identification of direct targets of AmphiVent1 transcription factor. Using luciferase reporter assays in P19 murine embryonal cells we have demonstrated Bmp-mediated activation of the AmphiVent1 59genomic non-coding regions (putative promoter) via Smad1/Smad4 proteins. Similar to vertebrate Xvent-2B gene promoter, AmphiVent1 promoter responsiveness to Bmp signaling is co-stimulated by zinc finger transcription factor OAZ. Furthermore, reporter plasmids where expression of GFP is controlled by Xenopus Xvent-2B and amphioxus AmphiVent1 promoters show a highly similar expression pattern in transgenic medaka embryos. We found that AmphiVent1 protein acts as a transcriptional repressor with the repression domain located at its N-terminus that appears to interact with groucho family co-repressor Grg4. As in the case of its vertebrate orthologues, AmphiVent1 protein can suppress the activity of amphioxus Chordin and Goosecoid gene promoters. Our data thus provide evidence for a remarkable conservation of Bmp-triggered gene regulatory network mediating DV patterning in vertebrates and basal chordates. On the other hand, our data suggest an increased complexity of DV pattern regulation in vertebrates. The canonical Wnt signaling regulatory input for ventral-specific gene expression appears to be lacking in cephalochordates (this study) [17] and has likely been co-opted in vertebrates.

Results
59genomic non-coding regions of amphioxus Vent genes are activated by Bmp signaling Two Vent-like genes, AmphiVent1 [3] and AmphiVent2 [10], can be identified in the genome of cephalochordate amphioxus (http://genome.jgi-psf.org/Brafl1/Brafl1.home.html). Both of them encode the Q50 homeodomain protein with the Ventspecific T47 substitution [18] (Fig. 1A). Likewise, two Vent genes are present in the zebrafish genome while four Vent genes are present in Xenopus laevis. Higher number of Vent genes in Xenopus laevis may be caused by a recent duplication of its genome as only two Vent genes are found in another frog Xenopus tropicalis (http:// www.ensembl.org). Only a single Vent gene is found in the genome of humans and chimpanzee. Interestingly, a functional copy of a Vent gene has been lost from the mouse genome; only a fragment of the Vent-type homeodomain in the mouse genome can be identified (this study; see discussion). Phylogenetic analysis suggests that independent lineage-specific duplication is responsible for the increased copy number of Vent genes (Fig. 1A). Recent lineage-specific duplication of AmphiVent1 and AmphiVent2 is consistent with high sequence similarity (83% nucleotide identity within respective ORF's; EMBOSS Pairwise Alignment at http:// www.ebi.ac.uk/Tools/emboss/align/). Since AmphiVent2 has not been previously characterized at all we performed an expression analysis using real-time quantitative RT-PCR. As shown in Fig. 1B AmphiVent2 displays similar but not identical temporal and quantitative regulation of mRNA expression as compared to AmphiVent1.
It is well established that a relatively short (approximately 300 bp) promoter (59genomic non-coding region) of XVent-2B gene is sufficient for Bmp-mediated regulation [12,13]. Given the known position of vertebrate Vent genes in the gene regulatory network governing DV patterning, we hypothesized that amphioxus Vent genes might be regulated by Bmp signaling. To test this possibility, we isolated approximately 1 kb of 59genomic noncoding regions of AmphiVent1 (21230/+20) and AmphiVent2 (2912/+22) genes putatively containing promoters and generated luciferase reporter gene constructs. We next tested their activity in P19 cells in the absence and presence of exogenous Bmp (heterologous human BMP2 was used in this study unless indicated otherwise). The promoter of Xenopus Xvent-2B gene known to be activated by Bmp signaling in the embryo and in P19 cells [12,13] was used as a control in all experiments. We observed BMP2induced stimulation of AmphiVent1-luc and AmphiVent2-luc reporter gene activity in P19 cells that was comparable to that of Xvent-2B-luc (Fig. 1B). In contrast, the 59genomic non-coding region (2300/+65) of the orthologous human VENTX2 gene was not inducible by BMP2. However, it is very likely that the 59genomic non-coding region of VENTX2 used in our study did not contain a complete promoter and so Bmp-responsive elements might have been missing. Applying different doses of BMP2 (from 12 ng/ml up to 400 ng/ml) resulted in rather similar promoter inductions (Fig. S1A). Similar results were obtained by using human BMP4 or BMP7 for the pathway stimulation (Fig. S1B). In the same experimental setting, AmphiVent1 promoter was not stimulated by treatment with either human TGF-b or human activin (Fig. S1C) that are, together with Bmp, members of the TGF-b super-family [19]. Combined, our data show that 59genomic non-coding regions of AmphiVent1 and AmphiVent2 genes contain functional regulatory regions that are stimulated by Bmp signaling in P19 cells like their Xenopus counterparts and are therefore referred to as promoters in this manuscript.

Bmp responsiveness of AmphiVent1 promoter is mediated by Smad transcription factors
We next decided to molecularly dissect AmphiVent1 promoter regulation. We have chosen AmphiVent1 since the corresponding gene has previously been characterized and represents an important marker of ventral mesoderm in amphioxus [2,3]. We tested if Bmp responsiveness of the AmphiVent1 gene promoter is mediated by the members of the Smad group of proteins. Cotransfection of common partner human Smad4 with receptor- activated human Smad1 into P19 cells resulted in activation of the AmphiVent1 promoter ( Fig. 2A). To explore whether intrinsic Smad4 DNA-binding activity is required for promoter activation, we used a Smad4-deficient cell line, MDA-MB-468. Wild-type Smad4, but not DNA-binding-deficient mutant Smad4-D4 [20], was able to induce promoter activity when cotransfected with Smad1 (Fig. 2B). We identified six putative Smad-binding elements (SBE; cAGAC) in the promoter of AmphiVent1 (Fig. 2C). To functionally analyze SBE's within the AmphiVent1 promoter, 59truncated promoter fragments (2750/+20, 2350/+20 and 2150/+20) were cloned upstream of the luciferase reporter gene (Fig. 2C). These truncated reporter genes were transfected into P19 cells and cells were stimulated by BMP2. We have observed a gradual decrease of Bmp responsiveness of reporter genes that was directly correlated with the extent of promoter truncation (Fig. 2C). The stimulatory effect of BMP2 was completely abolished only in the case of AmphiVent(2150)-luc reporter gene construct, which does not contain any putative SBE's (Fig. 2C). These results show that in vitro all putative SBE's are relevant for Bmp-mediated inducibility of the AmphiVent1 promoter. We next confirmed these data by a mutational analysis of the AmphiVent1 promoter. Reporter gene constructs containing point mutations of individual SBE's within AmphiVent1 promoter were generated. We found that destroying any single SBE does not have a significant effect on Bmp responsiveness (data not shown). Only when all six SBE's were mutated, the Bmp responsiveness of AmphiVent1 reporter gene was completely lost (Fig. 2C). We were further interested in whether the upstream promoter region of AmphiVent1 is able to function as an autonomous Bmp response element (BRE). Such BRE activity was previously ascribed to a specific region of Xenopus Xvent-2B promoter [13]. To this end two reporter gene constructs were generated that contained either a cluster of three proximal (2669/2218) or three distal (21214/2669) SBE's upstream of a minimal promoter (constructs designated pTAZ-BRE/P and pTAZ-BRE/D, respectively). Constructs were tested for their activity in P19 cells in the absence and presence of exogenous human BMP2. Although deletion and mutation analyses revealed that all SBE's are functional within the context of the natural AmphiVent1 promoter, only the cluster of proximal SBE's can function as an autonomous BRE when fused to a heterologous promoter (Fig. 2D). Taken together, our data suggest that the promoter of amphioxus AmphiVent1 gene is directly activated by the Smad-mediated Bmp signaling pathway.
Smad proteins co-operate with zinc finger protein OAZ in amphiVent1 promoter activation OAZ is a 30-zinc finger (ZF) protein, which associates with Smad1 in response to BMP2, allowing selective recognition of the BRE in Xenopus Xvent-2 promoter [13]. ZF's 6-13 of OAZ bind directly to the BRE of Xvent-2 promoter whereas ZF's14-19 at the C-terminus of OAZ interact with Smad1 and Smad4 (Fig. 3A). The human OAZ protein (hOAZ) is homologous to Xenopus and amphioxus OAZ transcription factors [13] (data not shown). To investigate whether OAZ is involved in Bmp-dependent regulation of AmphiVent1 promoter, we cotransfected human hOAZ with constitutively active human receptor caAlk2 into P19 cells. Expression of caAlk2 is known to trigger Bmp signaling, thus mimicking addition of a Bmp ligand [21]. Transfection of hOAZ cDNA alone did not stimulate the AmphiVent1 reporter gene. As expected, expression of the constitutively active caAlk2 alone activated the AmphiVent1 reporter gene about 4-fold (Fig. 3B). Cotransfection of the hOAZ expression vector together with caAlk2 resulted in potentiation of caAlk2-mediated response (12fold) (Fig. 3B), suggesting that, although OAZ is present in P19 cells [13], it appears to be a limiting factor for Bmp-dependent regulation. Finally, cotransfection of expression vectors encoding caAlk2, hOAZ, Smad1 and Smad4 into P19 cells led to remarkable activation of the AmphiVent1 reporter gene (70-fold activation) (Fig. 3B). These results suggest that transcription factor OAZ mediates Bmp regulation of the AmphiVent1 gene in a similar way as it does in the case of the Xvent-2 gene. To provide an insight into the molecular mechanism of OAZ-mediated regulation of AmphiVent1 promoter we used a dominant-negative hOAZ construct. It is known that ZF's 6-13 constitute the DNA-binding domain of hOAZ, which is however lacking the ability to interact with SMAD's and thereby to activate target promoters [13]. Based on these properties hOAZzf6-13 was previously used as a dominant-negative protein [13]. We cotransfected hOAZzf6-13 together with caAlk2 into P19 cells and we examined the responses of Xvent-2 and AmphiVent1 reporter genes. In congruence with our previous results, caAlk2-induced activation of AmphiVent1 and Xvent-2 reporter genes was significantly suppressed by cotransfection of the hOAZzf6-13 construct (Fig. 3C). The most likely explanation of the observed suppressive effect of hOAZzf6-13 is that the dominant-negative form of hOAZ competes with endogenous OAZ expressed in P19 cells for DNA-binding on AmphiVent1 promoter. It was shown previously that the same mechanism, namely DNA-binding displacement of endogenous OAZ by hOAZzf6-13, was responsible for attenuation of Bmpmediated activation of Xvent-2 [13]. Summarized, our data indicate deep homology in the molecular mechanisms of Bmpmediated regulation of chordate vent genes.

Amphioxus Vent genes are not regulated by Wnt/bcatenin signaling
It was shown that Wnt/b-catenin signaling directly regulates Xvent-1B and Xvent-2B genes via binding of the Lef/Tcf/b-catenin complex to their promoters [15,16]. To analyze a plausible role in the regulation of AmphiVent1 and AmphiVent2 genes, we first searched for Lef/Tcf binding motifs within their promoters. We found putative Lef/Tcf elements (59-CTTTGTT-39) in both AmphiVent1 (position 2596/2590) and AmphiVent2 (position 2450/2446) promoters (Fig. 4A). Promoters of Xenopus Vent genes, however, contain conserved consensus Lef/Tcf binding sequences in a more proximal position (265/259 in Xvent-1B promoter and 276/270 in Xvent-2B promoter, respectively) ( Fig. 4A). In contrast, the 59genomic non-coding region of human VENTX2 (2248/+65) does not contain any Lef/Tcf binding sequence (data not shown). To examine whether promoters of AmphiVent1, AmphiVent2, Xvent-1B and Xvent-2B genes and 59genomic non-coding region of VENTX2 gene are responsive to canonical Wnt signaling, we cotransfected their reporter genes into 293T cells together with N-terminally truncated b-catenin (b-cateninDN). It is well established that b-cateninDN is a constitutively active form of b-catenin (proteolytically stabilized), which is able to interact with endogenous LEF/TCF transcription factors, thus mimicking activation of canonical (Wnt/b-catenin) signaling. Mouse Sp5 promoter is known to be responsive to canonical Wnt signaling and was used as a positive control [22]. Only the activity of Xvent-1B and Xvent-2B promoters was significantly stimulated by cotransfection of human b-cateninDN (Fig. 4B). Conversely, cotransfection of b-cateninDN with AmphiVent1-luc, AmphiVent2-luc and VENTX2-luc did not lead to any significant stimulation of the respective reporter genes. On the contrary, cotransfection of b-cateninDN with Amphi-Vent2-luc resulted in a modest but significant repression of the reporter gene. Similar data were obtained when Wnt3Aconditioned medium was applied to 293T cells transfected with the individual vent reporter genes (Fig. S2). Next we mutated Lef/ Tcf binding sites in AmphiVent1, AmphiVent2, Xvent-1B and Xvent-2B promoters and performed cotransfections with b-cateninDN into 293T cells. As shown in Fig. 4C, responsivness of mutated Xvent-1B and Xvent-2B promoters to b-cateninDN was abolished indicating that single Lef/Tcf binding sites in these promoters mediate canonical Wnt signaling. As expected, mutating putative Lef/Tcf binding sites in AmphiVent1 and AmphiVent2 promoters did not have any significant effect in reporter gene assays as compared to wild-type constructs. Cotransfection of b-cateninDN with AmphiVent2mut-luc (like AmphiVent2-luc, see above) resulted in a modest repression of the reporter gene suggesting an indirect type of regulation. These results confirm previously published data showing direct regulation of the two Xenopus vent genes by canonical Wnt signaling [15,16]. Our data suggested that AmphiVent1 and AmphiVent2 genes are not directly regulated by canonical Wnt signaling. However, our conclusions were based on rather limited 59genomic non-coding regions that might be sufficient for Bmp-responsivness but not necessarily for responsiveness to canonical Wnt signaling. To provide more definitive answer about the possible role of Wnt/b-catenin signaling in the regulation of AmphiVent1 and AmphiVent2 expression we pharmacologically manipulated Wnt pathway in vivo in the developing amphioxus embryos. To activate the canonical Wnt signaling, we used 6-Bromoindirubin-39-oxime (BIO), a potent and less toxic inhibitor of glycogen synthase kinase-3b (GSK-3b) as compared to lithium (Li + ) [23]. BIO was added to developing amphioxus embryos at blastula stage and embryos were allowed to develop until mid-neurula stage at which point mRNA was isolated and gene expression interrogated by real-time quantitative RT-PCR. Amphioxus FoxQ2 and Axin genes were used as controls to test for effectiveness of Wnt pathway stimulation. It was previously shown that AmphiFoxQ2 expression is downregulated upon pharmacological manipulation of canonical Wnt signaling (Li + administration; [24]). Axin is a functional component of Wnt/bcatenin signaling that associates directly with b-catenin, GSK-3b and APC and is implicated in down-regulating Wnt signaling [25]. Vertebrate Axin2 is a direct target of Wnt/b-catenin signaling whose expression is induced by activated Wnt signaling and acts therefore in a negative feedback loop [26,27]. Axin2 is currently the most reliable and frequently used natural readout of Wnt/bcatenin signaling in vivo (http://www.stanford.edu/group/nusselab /cgi-bin/wnt/reporters). As shown in Fig. 4D, constitutive activation of Wnt/b-catenin signaling in the developing amphioxus embryos resulted in marked increase of AmphiAxin expression and downregulation of AmphiFoxQ2 expression. In contrast, however, expression of AmphiVent1 and AmphiVent2 has not been significantly upregulated in the presence of activated Wnt/bcatenin signaling.
Available evidence thus suggests that although canonical Wnt signaling plays a prominent role in the early establishment of ventral mesoderm in Xenopus and zebrafish, amphioxus does not use this pathway for specification of the ventral fate. In summary, our data argue that, in contrast to the situation with Bmp signaling, participation of the Wnt/b-catenin signaling pathway in regulation of Vent genes is not conserved among chordates.
Functional diversification of amphioxus Vent gene promoters: a case of possible regulation by dorsalspecific forkhead transcription factors By analyzing the proximal regions of AmphiVent1 and AmphiVent2 promoters in silico using Family Relations software (http://family. caltech.edu) we identified highly conserved sequence motifs that are 80% similar within a sliding 20 bp window ( Fig. 5A and data not shown). A FoxD-binding element GTAAC was found within this region in the AmphiVent2 gene whereas the promoter of AmphiVent1 contains a single nucleotide change (GcAAC) in the FoxD motif (Fig. 5A). In amphioxus AmphiFoxD is expressed in the axial mesendoderm within the dorsal lip of blastopore at early gastrula stage [28]. Taking into account the expression pattern data and our in silico analysis we hypothesized that AmphiVent genes might be targets of AmphiFoxD. From this point of view, it is interesting to note that a negative regulation between FoxD and vent genes has been described in Xenopus [29]. XFD-19, the Xenopus FoxD ortologue, was shown to be suppressed by the Xvent-1 gene and plays a role in DV patterning. In fact, XFD-19 is a dorsal lipspecific transcription factor, which is specifically activated in Xenopus organizer. To investigate whether AmphiFoxD protein can bind to putative sites within the conserved region of AmphiVent1 and AmphiVent2 promoters, double-stranded oligonucleotides derived from the corresponding regions of each promoter were tested by in vitro DNA-binding assay (electrophoretic mobility shift assay, EMSA). AmphiFoxD formed a specific complex with the probe, which corresponded to the AmphiVent2 promoter region (Fig. 5B, C). In contrast, AmphiFoxD did not bind to the probe which corresponded to the AmphiVent1 promoter region, consistent with the observed mutation in the FoxD binding motif or to a nonspecific (unrelated) probe. Binding site specificity was confirmed by EMSA in the presence of increasing amounts of non-specific (unrelated) double-stranded oligonucleotide or AmphiFoxD binding site derived from AmphiVent2 promoter region. As shown in Fig. 5D, only AmphiFoxD binding site can effectively compete for the formation of the complex. To provide further evidence for AmphiFoxD-mediated regulation of AmphiVent2, P19 cells were transfected with luciferase reporters containing AmphiVent1 or AmphiVent2 promoters in the presence or absence of an expression plasmid encoding AmphiFoxD. AmphiFoxD can significantly repress AmphiVent2-luc but not AmphiVent1-luc promoter construct (Fig. 5E). Our data suggest that AmphiVent2, but not AmphiVent1, might be subject to FoxD regulation. In addition, the data exemplify functional diversification of promoter sequences after duplication of vent genes in the amphioxus lineage.

Activation of AmphiVent1 and AmphiVent2 promoters in early developing medaka embryos
We next asked whether the amphioxus AmphiVent1 (1.2 kb), AmphiVent2 (0.9 kb) and Xenopus Xvent-2B (0.3 kb) promoters are Luciferase reporter plasmids containing wild-type, deleted or mutated AmphiVent1 promoter fragments were transfected into P19 cells and cells were stimulated by BMP2. Fold-induction by BMP2 is indicated. Position of individual SBE's is indicated by black ovals, and mutated SBE's by crossed oval. (D) Identification of an autonomous BRE in AmphiVent1 promoter. P19 cells were transfected with reporters containing a minimal promoter fused to either three proximal SBE elements (pTAZ-BRE/P) or three distal SBE elements (pTAZ-BRE/D). Reporter genes were stimulated by exogenous human BMP2 (50 ng/ml). *P,0.05, **P,0.01, ***P,0.001. doi:10.1371/journal.pone.0014650.g002 activated in medaka embryos. Corresponding EGFP reporter constructs p817-AmphiVent1, p817-AmphiVent2 and p817-Xvent-2B were injected into medaka embryos at the single cell stage and their transient expression was monitored during early embryogenesis. At early gastrula stage AmphiVent1 (Fig. 6A-A9), Xvent-2B (Fig. 6B-B9), and AmphiVent2 (Fig. S3A-A9) promoters were activated throughout the dorsal blastoderm of the embryo, demarcating the region of the most dorsal embryonic shield, where dorsal mesodermal marker Chordin is expressed (Fig. 6C) [30]. The EGFP signal from the p817-AmphiVent1 (Fig. 6D-E9) and p817-AmphiVent2 ( Fig. S3B-C9) constructs remained evident during gastrulation and its pattern resembled the activation of EGFP in the embryos injected with p817-Xvent-2B (Fig. 6 F-G9). At midgastrula stage the strongest EGFP signal was observed laterally from growing embryonic shield. Neither AmphiVent1 nor Xvent-2B promoter was activated in the area of the embryonic shield, where dorsal mesodermal marker Goosecoid is expressed in the medaka embryo (Fig. 6H) [31]. We detected ventrolateral expression of EGFP driven by the Xvent-2B promoter (Fig. 6G). In contrast, the AmphiVent1 promoter was not activated ventrally in medaka embryo at mid-gastrula stage (Fig. 6E). It is interesting to note that, in contrast to the Xvent-2B promoter, AmphiVent1 and   AmphiVent2 promoters were activated dorsolaterally, but not ventrally, in the medaka embryo. These observations are in agreement with known expression pattern data for vertebrate and cephalochordate Vent genes. The expression of Xenopus vent genes and AmphiVent1 was demonstrated to be highly similar in dorsolateral but not ventrolateral mesoderm [3]. Whereas somites and their derivatives originate from dorsolateral mesoderm, ventrolateral mesoderm gives rise to the heart and other components of circulatory system, which is generally much simpler in cephalochordates than in vertebrates.
Taken together, our data suggest that the AmphiVent1 promoter is correctly regulated in the vertebrate embryo and that its spatiotemporal activity is by large similar to the activity of Xvent-2B promoter in the same context.

All Vent proteins function as transcriptional repressors and interact with groucho co-repressors
Xenopus Xvent-2B and zebrafish Vent genes were shown to act as transcriptional repressors [4,6,11]. Besides, activating function of Xvent-2 was described [15,32]. To examine the transcriptional properties of chordate vent proteins, a Gal4 reporter assay was employed. Plasmids encoding Gal4 fusions with AmphiVent1, AmphiVent2, Xvent-1b, Xvent-2b and VENTX2 were cotransfected with Gal4-dependent reporter gene. As shown in Fig. 7A, all vent proteins strongly repressed expression of the reporter gene when tethered to the promoter via Gal4 binding sites. We further focused on identifying the functional domains within AmphiVent1 that mediate transcriptional repression. AmphiVent1 and Amphi-Vent2 show high amino acid sequence homology within the entire open reading frame (80% identity, 84% similarity) suggesting similar molecular properties. From the two amphioxus vent proteins we selected AmphiVent1 since it is encoded by a previously characterized gene that represents an important marker of ventral mesoderm in amphioxus [2,3]. To this end, Gal4 fusion constructs encoding different domains of AmphiVent1 were cotransfected together with the Gal4-dependent reporter plasmid into 293T cells. The Gal4 fusion proteins containing the Nterminus or the homeodomain repressed transcription 21-fold and 10-fold, respectively (Fig. 7B). In contrast, the Gal4 fusion protein containing the C-terminus of AmphiVent1 activated transcription 2.2-fold (Fig. 7B). These results suggest that overall, AmphiVent1 acts as a transcriptional repressor and has strong repression domains at its N-terminus and within the homeodomain. In addition, there is a weak transcriptional activation domain located within the C-terminus of AmphiVent1 like in Xenopus Xvent-2B protein [15]. Next, we tried to identify specific amino acid sequences within the N-terminus which are responsible for the repression function of AmphiVent1. We cotransfected Gal4 fusion constructs encoding AmphiVent1 amino acids 1-116, 23-116, 41-116, 67-116, 1-74 and 1-42, respectively. All fusion proteins strongly repressed expression of the reporter gene (Fig. 7C). These data suggest that the AmphiVent1 protein likely has multiple independent repressor domains within its N-terminus. Sin3A and Groucho family members appear to be widely used cofactors mediating transcriptional repression of many DNA-binding proteins, including those containing a homeodomain [33,34,35,36,37]. We therefore tested a possible interaction of the AmphiVent1 N-terminal domain with these obligatory corepressors. As shown in Fig. 7D, the AmphiVent1 N-terminal domain is able to interact with mouse Grg4 but not with human Sin3A. Interaction of AmphiVent1 with groucho-like co-repressors may be responsible, at least in part, for function of AmphiVent1 as a potent transcriptional repressor.

AmphiVent1 transcription factor represses AmphiGoosecoid and AmphiChordin promoters
One of the most prominent functions of vertebrate Vent transcription factors is repression of organizer-specific genes such as Goosecoid and chordin. To analyze whether AmphiVent1 acts as transcriptional repressor of AmphiGoosecoid and AmphiChordin genes, we first generated luciferase reporters containing AmphiGoosecoid (2580/+101) and AmphiChordin (21354/+118) promoter regions. Cotransfections of AmphiGoosecoid-luc and AmphiChordin-luc plasmids with an expression vector carrying the AmphiVent1 coding sequence resulted in downregulation of reporter genes 2.8fold and 6.8-fold, respectively (Fig. 8A). Paired-type homeodomains interact with a core TAAT motif as monomers or as homo/ heterodimers with dimer sites containing inverted TAAT core motifs separated by several nucleotides [38,39,40,41]. A synthetic homeodomain reporter gene designated 3xHD(P3)-luc that contains three palindromic homeodomain binding sites (TAAT-cagATTA) was repressed 2-fold by AmphiVent1 (Fig. 8A). In case of 36HD(P3)-luc the two TAAT core motifs are separated by three nucleotides (P3). In order to better define the DNA-binding specificity of AmphiVent1 and spacing requirements, we performed EMSA with AmphiVent1 homeodomain and a series of binding sites (Fig. 8B). Although AmphiVent1 homeodomain was able to interact with a single TAAT motif in the P1/2 binding site, homeodomain dimerization was observed on P2-P4 binding sites. In comparison with a related paired-type homeodomain of Pax6, AmphiVent1 has a conspicuous preference for a three-nucleotide spacer (P3) (Fig. 8B). It was shown previously that the sequence CTAATTG is critical for Xvent-2B binding, and that the binding is enhanced by the presence of an additional ATTA motif six or seven nucleotides 39 of the core TAAT [6]. In addition to multiple TAAT core motifs we found two CTAATTG motifs in AmphiChordin promoter at positions 2574/2580 and 2687/ 2682 ( Fig. 8A and data not shown). The CTAATTG motif is not present in AmphiGoosecoid promoter, but instead, we identified a putative P3-like palindromic homeodomain-binding sequence ATTAttgTAAT at a position 256/268. To investigate whether AmphiVent1 is able to repress AmphiGoosecoid promoter through this binding site, AmphiGoosecoid promoter-containing reporter plasmid with a mutated homeodomain motif (CTTCttgTCCT) was generated and designated AmphiGoosecoid(mut)-luc (Fig. 8C). Binding of the AmphiVent1 homeodomain protein in EMSA was readily detectable to the wild-type AmphiGoosecoid promoter sequence, but not to its mutated version (Fig. 8C, lower right panel). In accordance with this data, AmphiVent1 was not able to repress the AmphiGoosecoid(mut)-luc reporter gene (Fig. 8C, lower left panel). To further corroborate our results we converted AmphiVent1 to an activator by fusing it to a strong transactivation domain derived from the VP16 transcription factor. As shown in Fig. 8C, AmphiVent1-VP16 was able to strongly activate the wildtype AmphiGoosecoid-luc reporter gene, but not the AmphiGoosecoid(mut)-luc in which the homeodomain binding site was mutated. Combined, these data suggest that repression of the AmphiGoosecoid promoter by AmphiVent1 is mediated via P3-like binding site. Likewise, AmphiVent1-VP16 was able to strongly activate the AmphiChordin-luc reporter gene, whereas fusion of AmphiVent1 to the engrailed repression domain (AmphiVent1-EN) generated a transcription factor with properties comparable to wild-type AmphiVent1 (Fig. 8D). It is well established that the homeodomain can both bind DNA and mediate protein-protein interactions [42]. To investigate whether the DNA-binding function of AmphiVent1 homeodomain is critical for the downregulation of both AmphiGoosecoid and AmphiChordin genes, we generated two mutants, AmphiVent1(R53A) and Amphi-  [43] these mutations are predicted to disrupt DNA-binding ability of the paired-type homeodomain such as AmphiVent1. AmphiVent1(R53A) and AmphiVent1(N51Q) proteins were no longer able to repress AmphiGoosecoid and AmphiChordin reporter genes (Fig. 8D) despite normal level of expression (Fig. 8D, inset). Taken together, our data show that the AmphiVent1 homeodomain protein is able to act as a direct transcriptional repressor of AmphiGoosecoid and AmphiChordin genes.

Discussion
Orthologous genes encoding components of Bmp signaling and transcription factors downstream of Bmp signaling (such as Vent and Goosecoid) are expressed in highly similar patterns in vertebrates and a cephalochordate amphioxus [2]. However, the key question is whether the conservation of expression patterns of network constituents necessarily means conservation of functional network interactions. Furthermore, even if part (a core) of the gene regulatory network is evolutionarily conserved, one has to explain how an increased morphological complexity is achieved.
Here, we performed a detailed molecular analysis of gene regulatory network underlying DV patterning in a basal chordate amphioxus. Although we believe that the data presented in this study are relevant for amphioxus it must be emphasized that most of the experiments were performed in vitro using heterologous cell lines and heterologous proteins or heterologous animal model system (medaka) which may induce a bias in the results and influence our interpretation. In addition, since no data on transgenic amphioxus are available for AmphiVent1 or AmphiVent2 promoter constructs used in this study, we do not know whether we have isolated sufficiently large promoter fragments that can account for the entire expression domain of these two genes.

Vent genes in chordate evolution: from indispensableness to damnation
Genetic and molecular studies have identified a remarkably conserved function of the Bmp-Chordin signaling network in animal DV patterning [1]. The system, which relies on production of secreted Bmp inhibitors such as Chordin, antagonizing the activity of Bmps with resulting gradient of Bmp activity along the DV axis, seems to be utilized in invertebrates as well as in vertebrates [1,2,44]. However, an impressive conservation of the key players (Bmp and Chordin) does not eventuate in the conservation of all other components in the network. For example, Vent genes, which are critically important for DV patterning in 'lower' vertebrates [4,7,11,12] and cephalochordates (this study), are apparently dispensable for the Bmp-Chordin signaling network in mammals [45] (this study). It is interesting to note that the number of Vent-like genes differs among the chordate species. Amphioxus genome contains two Vent genes copies (AmphiVent1 and AmphiVent2) [10], which are situated on the same chromosome in close proximity and have likely arosen by a lineage-specific duplication. Three and four Vent-like genes are present in the genome of zebrafish and Xenopus, respectively. A single Vent gene (VENTX) is present in the genome of humans and chimpanzee. In amphibians and teleosts mutational analysis of Vent genes suggested their prominent role in DV patterning [4,9,16]. Products of zebrafish Vent, Vox, Ved and Xenopus Xvent-1/Xvent1b and Xvent-2/Xvent2b genes are homeodomain-containing repressor proteins, which at early stages of development suppress organizerspecific genes Goosecoid and Chordin [4,7,9,11,46]. It was shown that Xenopus Goosecoid promoter region contains the Xvent-2 binding site and is directly repressed by Xvent-2 [6,9]. In our study we have provided evidence for direct repression of amphioxus Goosecoid and Chordin gene promoters by the AmphiVent1 protein.
However, it appears that in mammals Vent genes have completely abandoned their prominent function within the Bmp-Chordin signaling network [18] (this study). First of all, a functional copy of Vent gene has been apparently lost in the mouse genome. We have identified a sequence encoding a remnant of a homeodomain homologous to exon2 and exon3 of the human VENTX gene that is located on mouse chromosome 7 between Utf1 and Adam8. This region precisely corresponds to a syntenic region where human orthologue VENTX is situated on human chromosome 10. Moreover, the functional role of VENTX is not clear. We have shown that in contrast to AmphiVent1/2 and Xvent1/2 promoters, the promoter of VENTX is not Bmp-inducible. Expression of the VENTX gene has only been detected in immature bone marrow, indicating that the gene may be involved in the maintenance of hematopoietic stem cells in the adult [45]. Its expression is downregulated in lymphocytic leukemias, suggesting a potential role of VENTX in the clinical behavior of hematopoietic malignancies [47]. Structure-function studies have shown that the human VENTX2 protein functions as a transcriptional repressor [45] (this study), a feature conserved among other Vent proteins (this study). Functional equivalence of human VENTX and zebrafish/Xenopus Vent proteins was confirmed by microinjection of VENTX mRNA that resulted in ventralized zebrafish embryo [45]. Recently, a new protein partner for VENTX and Xvent-2 has been identified. Gao et al. have reported that Xvent-2 and VENTX are Lef/Tcf-associated factors [47,48]. However, their interaction with Lef/Tcf results in distinct functional properties. Upon association with Lef/Tcf Xenopus Xvent-2 seems to be an activator, whereas human VENTX protein functions as a suppressor of canonical Wnt signaling. Gao et al. proposed that Xvent-2 interacts with Tcf/Lef directly through its homeodomain and activates gene expression by the N-terminal transactivation domain [47]. More recently, Gao et al. argue that the mechanism of transcriptional repression by VENTX is caused by the disruption of the complex between b-catenin and TCF/LEF factors [48]. Using reporter gene assays we have been able to reproduce transactivation results for vertebrate Vent proteins reported previously [47,48], i.e. transcriptional stimulation in the case of XVent-2 and suppression for VENTX (Fig. S4). However, in the same transcriptional assay AmphiVent1 was not able to influence the Lef/Tcf-responsive reporter gene (Fig. S4) despite high sequence conservation of AmhiVent1, Xvent-2 and VENTX homeodomains (Fig. 1A), which are the presumed interaction domains. One obvious explanation is the lack of a strong transactivation domain in AmphiVent1. This is unlikely since the fusion of AmphiVent1 with strong transcriptional activator VP16 does not lead to detectable activation of the Lef/Tcf reporter gene (Fig. S4). The molecular mechanism of Vent-G9); dorsal (D-D9, F-F9) and lateral (E-E9, G-G9) views show EGFP fluorescence in the blastoderm around the most dorsal region of embryonic shield (Sh). (C, H) Schematic diagram of developing medaka embryo depicting mRNA expression pattern of early dorsal mesoderm markers Chordin (stage 13, C) [30] and Goosecoid (stage 15, H) [31]. Dashed line indicates the borders of the blastoderm. White arrowheads depict the most dorsal embryonic shield of the medaka embryo, where Chordin and Goosecoid are expressed. GR-germ ring, Sh-embryonic shield. doi:10.1371/journal.pone.0014650.g006 mediated modulation of canonical Wnt signaling thus remains unclear.
In conclusion, genome analyses, structure-function and developmental studies point to a dynamic evolution of vent genes. The loss of a presumed ancestral role of Vent genes in chordate DV patterning is particularly intriguing. One plausible explanation of such an impressive damnation of Vent genes in mammals lies probably in marked evolutionary changes in early development of chordates. Sanders and co-workers [18] argue that the lack of Vent gene in the mouse is due to slower development of the mouse embryo. Besides, at very early stages of development mammalian embryo does not need to be polarized for future proper development. In contrast, molecular asymmetry of teleost and amphibian eggs is an inherent condition of their normal development. Many maternal and newly expressed zygotic genes are involved in the highly organized scenario driving the polarization of zebrafish and Xenopus embryo. Thus, Vent genes have likely been recruited as one of the key players in this complicated system of molecular asymmetry. Yu et al. have shown that the expression patterns of Bmp signaling components and its modulators such as Chordin, ADMP, Goosecoid, BAMBI, Tsg display extreme similarity between amphioxus and vertebrate embryos at early stage of development [2]. Amphioxus embryos treated with BMP4 have a phenotype similar to that resulting from overexpression of Bmp2 and Bmp4 in vertebrate embryos [49]. Exogenous Bmp protein treatment repressed expression of markers of the dorsal mesoderm and caused ventralization of the amphioxus embryo [2]. In early vertebrate (Xenopus and zebrafish) embryo Bmp signaling is mediated by Vent genes [4,11]. Detailed molecular analysis of Xvent-2 promoter has revealed BRE in its proximal region [12]. The proximal promoter of Xvent-2 contains Smad and OAZ binding sites, which are known mediators of Bmp4 signaling. Indeed, it was shown that OAZ is able to form a complex with Smad1 and Smad4 upon Bmp stimulation [13]. Furthermore, a cooperative binding of OAZ and Smads to the BRE of Xvent-2 gene promoter leads to promoter activation. Our data support a concept of a highly conserved role of Bmp regulation in the establishment of D/V axis in chordates. Bmp signaling proteins activate amphioxus AmphiVent1 and AmphiVent2 genes. Detailed analysis of the regulation of AmphiVent1 promoter region has revealed that the activation is mediated by Smad1/Smad4 proteins. Furthermore, zinc finger protein OAZ seems to be involved in the activation of AmphiVent1 in a similar way as shown previously for the Xenopus Xvent-2 gene. There is a notable difference in the promoter structure between cephalochordates and amphibian genes with regard to positioning and significance of individual SMAD binding motifs. BRE of the Xenopus Xvent-2 gene is constrained within a short promoter region between nucleotides 2243 to 2191 [13]. This region contains a single Smad binding site, which is crucial for Bmp inducibility of Xvent-2 promoter. In contrast, we have identified six putative Smad binding elements (SBE) in the Bmp inducible region of AmphiVent1 gene promoter between nucleotides 21230 to +20. Deletion and mutational analysis has revealed that none of these SBE's plays a dominant role in the activation of the gene by Bmp signaling since point mutations of individual SBE's did not lead to the complete loss of Bmp inducibility. Bmp-mediated activation was abolished only in the case of simultaneous mutation of all six SBE's within the 21230 to +20 promoter region of AmphiVent1. Therefore, the significance of six SBE's within the AmphiVent1 promoter region appears to be equally balanced. Seemingly this fact is not even influenced by the presence of OAZ binding site in the vicinity of one of the SBE's. It is of note that the human VENTX gene promoter contains a putative SBE, which however does not provide any detectable Bmp inducibility [50] (this study). Our observations provide evidence for a prominent robustness of the cephalochordate AmphiVent1 promoter with regard to Bmp responsiveness through multiplication of SBE's. In contrast, the evolution of vertebrate Vent gene regulation proceeded by a distinct mechanism. The apparent robustness of Bmp-mediated regulation of Xenopus Vent genes was achieved by co-option of Vent binding sites into their promoters. These sites provide a positive autoregulatory loop (Fig. 9), which maintains the Bmp inducibility of Xenopus Vent promoters. We have not found any evidence for such an autoregulatory loop in the case of AmphiVent1 and AmphiVent2 promoters (Fig. S5). Summarized, our data suggest that BRE-containing cis-regulatory sequences for Smad and OAZ transcription factors were likely present in Vent-like homeobox genes of a chordate common ancestor. Taking into account our data, which indicate the ability of amphioxus Vent proteins to directly repress AmphiChordin and AmphiGoosecoid, we propose a deep evolutionary conservation of a Bmp-mediated regulatory module within the gene regulatory network (GRN) controlling chordate DV patterning (Fig. 9).

Conserved BMP regulation of chordate Vent genes
In addition to the transcriptionally-based gene regulatory network, experimental analyses have recently revealed another level of Bmp-mediated regulation via formation of a gradient in Xenopus embryo [51]. It was shown that the Bmp activity gradient is defined by a'shuttling-based' mechanism, whereby the Bmp ligands are translocated ventrally through their association with Bmp inhibitor Chordin. This mechanism, which may function in amphioxus as well, represents another way to achieve the robustness of Bmp gradient [52].
The role of Wnt-signaling in the formation of chordate DV axis is not conserved One of the earliest asymmetrical molecular signals in the developing vertebrate embryo is the nuclear localization of bcatenin on the dorsal side. The b-catenin signal induces the expression of Bmp antagonists such as Chordin and Noggin in dorsal mesoderm (Spemann organizer) of early Xenopus embryo [53]. Likewise, in zebrafish, cells of the embryonic shield accumulate b-catenin and thus acquire organizing activity [54]. A significant role of canonical Wnt signaling in the formation of an organizer has not been observed in cephalochordates. Although during early amphioxus development nuclear b-catenin and some The N-terminal repression domain of AmphiVent1 interacts with co-repressor Grg4. Schematic diagram of the construct is shown to the left. GST or GST-AmphiVent1-N proteins were bound to Glutathione-Sepharose beads and analyzed by SDS PAGE (middle panel). Please, note equal levels of protein loaded onto beads. GST pull-down assay with in vitro produced S 35 -labelled luciferase (negative control), Sin3A and Grg4 co-repressor. Input represents 10% of in vitro synthesized proteins used for binding assay. Note that only Grg4 efficiently interacts with the AmphiVent1 N-terminal repression domain. *P,0.05, **P,0.01, ***P,0.001. doi:10.1371/journal.pone.0014650.g007 Wnts are present around the blastopore [2], overactivation of Wnt signaling by treatment of amphioxus embryos with Li + at early stages of development had no effect on the DV axis specification [17].
Loss-of-function experiments indicate an important role of canonical Wnt signaling in the regulation of Vent genes and specification of vertebrate ventral mesoderm [14,55]. Moreover, Wnt8 was shown to be able to activate Xenopus Xvent-2 and Xvent-1 genes as well as zebrafish Vent and Vox [14,15,56]. Using luciferase reporter assays we have confirmed positive regulation of Xvent-1 and Xvent-2 by canonical Wnt signaling (this study, Fig. 4B). It is established that canonical Wnt signaling plays an important role in early mesoderm formation of mammals [57]. We have found that the promoter of the human VENTX gene, likely due to its rapid evolution (see above), is not responsive to canonical Wnt signaling. Taking into account an apparent loss of the mouse Vent gene, the available data point to Vent-independent formation of mesoderm in mammals.
It seems that Wnt/b-catenin signaling plays a dual role in the regulation of Xenopus and zebrafish Vent genes. In addition to stimulation triggered by Wnt8, b-catenin participates in the suppression of vertebrate Vent genes in dorsal lip at early stages of development. The downregulation is achieved by b-catenindependent activation of dorsal-lip-specific transcription factors such as goosecoid, which in turn represses Vent gene expression [7,18,58]. Expression data indicate mutually exclusive patterns of AmphiGoosecoid and AmphiVent1 genes in early amphioxus embryo, which makes the above-mentioned mechanism plausible in amphioxus as well [2].  AmphiChordin-luc and AmphiGossecoid-luc reporters were cotransfected with or without expression plasmids encoding wild-type AmphiVent1, DNA-binding deficient AmphiVent1 (R53A, N51Q), AmphiVent1-EN (artificial repression construct) and AmphiVent1-VP16 (artificial activation construct). Note that only the wild-type, but not the DNA-binding mutants of AmphiVent1 can repress the reporter genes. Equal expression of wildtype and mutant AmphiVent1 proteins is documented by Western blotting using antibody against the N-terminally engineered FLAG epitope (inset). Please, note that as predicted from structure-function analysis AmphiVent1-EN functions as wild-type AmphiVent1 (mediates repression), whereas AmphiVent1-VP16 mediates an opposite effect (strong activation of reporter genes). **P,0.01, ***P,0.001. doi:10.1371/journal.pone.0014650.g008 We were interested in whether the Wnt-mediated regulation of Vent genes is an evolutionarily conserved feature between cephalochordates and 'lower' vertebrates. It was shown previously that the expression patterns of Wnt8 and AmphiVent1 overlap in early amphioxus embryo [2]. At a mid-gastrula stage AmphiVent1 and Wnt8 are co-expressed dorsolaterally in regions of invaginated mesendoderm, which constitute the prospective paraxial mesoderm [2,3,59]. Although we have identified putative Lef/Tcf binding sites in AmphiVent1 and AmphiVent2 gene promoters by in silico analysis, our experimental data in vitro (reporter assays) and in vivo (pharmacological treatment of embryos) do not indicate that the promoters are regulated by canonical Wnt signaling, which is congruent with results obtained by Li+ treatment that has not impact on DV patterning in amphioxus- [17].
It is interesting to note that canonical Wnt signaling is broadly used for the establishment of anteroposterior (A/P) axis in diverse deutorostomes (including frogs, fish, mammals, birds, amphioxus, and echinoderms) and protostomes [60]. In A/P patterning Wnts are important posteriorizing factors, which influence development of multiple tissues. In contrast, the main outcome of DV polarity specification in chordates is the formation of distinct types of mesoderm from which ventral mesoderm of vertebrates gives rise to a large number of derivatives (such as complicated circulatory system). Co-option of a new signaling pathway (Wnt) in mesoderm patterning might be explained by a need for increased complexity and quantity of ventral mesoderm in vertebrate development. For example, vertebrate embryos need to produce an efficient circulatory system without delay. In contrast, the embryo of amphioxus is very small and does not need to deploy a large amount of mesoderm early to set the stage for rapid production of a highly efficient circulatory system [3].
Summarized, the available experimental evidence indicates that the canonical Wnt signaling pathway does not play a role in the establishment of ventral fate in cephalochordate amphioxus (Fig. 9). This is in principle consistent with a secondary loss of an ancestral feature (Wnt-mediated regulation) in cephalochordates. We, however, favor the hypothesis that the co-option of canonical Wnt signaling pathway for the establishment of DV patterning in vertebrates represents one of the innovations through which an increased morphological complexity of vertebrate embryo is achieved.

Conclusion
In conclusion, our data show that there is a remarkably conserved gene regulatory network in which Bmp signaling induces transcriptional repressor Vent, which in turn represses genes encoding transcription factor Goosecoid and signaling molecule Chordin. We propose that this gene regulatory network was a key module recruited in early chordate evolution for establishment of DV patterning. This network remains in use in the present-day chordates such as amphioxus, fish and amphibians.

Ethics statement
Housing of animals and in vivo experiments were performed after approval by the Animal Care Committee of the Institute of Molecular Genetics (study ID#36/2007) and in compliance with national and institutional guidelines (ID#12135/2010-17210).
Cell Culture, Transient Transfection and Luciferase Reporter Assay 293T, MDA-MB468 and P19 cells were purchased from ATCC. 293T and MDA-MB468 cells were cultured in Dulbecco's modified Eagle's medium (SIGMA) supplemented with 10% Fetal bovine serum (PAA laboratories), 2 mM L-glutamine, 100 units/ml penicillin, and 0.1 mg/ml streptomycin (SIGMA). P19 cells were cultured in Dulbecco's modified Eagle's medium (SIGMA) supplemented with 5% Fetal bovine serum (PAA laboratories), 2 mM L-glutamine, 100 units/ml penicillin, and 0.1 mg/ml streptomycin. Cells were passaged every three days and maintained at 37uC in an atmosphere of humidified air with 5% CO2. Cells were plated in 24-well plates 24 hours prior to transfection. Each well was transfected with 100 ng of the reporter gene and 50 ng of the expression vector (when applicable) using Fugene 6 (Roche) according to the manufacturer's protocol. The total amount of DNA transfected per well was 300 ng and was adjusted with pUC18 when necessary. A b-Galactosidase expression plasmid was cotransfected to normalize the transfection efficiency. In some experiments, 24 hours after transfection recombinant human BMP2, BMP4, BMP7, TGFb or Activin B (all from R&D) was added.Unless indicated otherwise, Bmp signaling was stimulated by the addition of human BMP2 at a final concentration of 20 ng/ml. Alternatively, Bmp pathway stimulation was elicited by transfection of 50 ng of the caAlk2 expression vector. Two days after transfection, the cells were lysed in 100 ul of 16passive lysis buffer (Promega). Luciferase reporter assays were performed using Luciferase Reporter assay kit (Promega). b-Galactosidase was detected with Galacto-Star system (Applied Biosystems). All transfection experiments were performed at least three times and a representative experiment is shown. In each experiment triplicate assays were performed; graph values represent the average of triplicates +/2 standard deviation. Statistical significance was determined using Student t-test in Microsoft Excel.

Electrophoretic Mobility Shift Assay (EMSA)
The following double-stranded oligonucleotides were used in EMSA with amphiFoxD (59-39,  Double-stranded oligonucleotides containing indicated homeodomain and FoxD binding sites were radioactively labeled at the 59ends with c 32 PdATP using polynucleotide kinase (Boehringer Manheim) and purified on microspin columns (Amersham Biosciences). 32 P-labeled oligonucleotides were incubated with bacterially-purified 6xHis-tagged or GSTtagged homeodomain proteins in binding buffer (10 mM HEPES at pH 7.7, 75 mM KCl, 2.5 mM MgCl2,0.1 mM EDTA, 1 mM DTT, 20% glycerol, 0.5% BSA, and 0.1 mg/ mL poly-dIdC) for 15 minutes at RT. In some experiments increasing amounts of unlabelled double-stranded oligonucleotides were added to binding reaction to test for specificity. Samples were analysed by 6 % polyacrylamide gel electrophoresis and autoradiography.
GST-Pull Down Assay 35 S-labeled Grg4 was prepared by TNT Quick Coupled Transcription/Translation Systems according to the manufacturer's protocol (Promega). GST fusion expression plasmids were transformed into BL21 CodonPlus (DE3)-RIPL cells (Stratagene). A single colony from the transformation was cultured in 2 ml LB medium containing 50 mg/ml of chloramphenicol and 30 ng/ml of kanamycin overnight at 37uC. The cultures were transferred to 100 ml of LB without antibiotics. The expression of the fusion construct was induced by adding IPTG to a final concentration of 2 mM for 2 hours. The cells were harvested by centrifugation and resuspended in 5 ml of NETN buffer (20 mM Tris pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% NP40). Lysozyme was added to a final concentration of 0.1 mg/ml. The lysates were incubated on ice for 20 min, sonicated and centrifuged to remove the cell debris. The supernatant was incubated with 200 ml of glutathione-Sepharose beads slurry (BD Bioscience) for 1 hour at 4uC. The beads were washed three times by 5 ml of binding buffer (20 mM Tris pH 8.0, 100 mM KCl, 5 mM MgCl2, 0.1 mM EDTA, 20% Glycerol) containing 0.1% NP40. GST fusion proteins bound to the beads were analyzed by SDS-PAGE. Beads containing normalized amounts of fusion proteins were blocked by binding buffer containing 0.05% of NP40 and 5 mg/ml of BSA for 2 hours at 4uC and resuspended in 150 ml of binding buffer containing 0.05% NP40, 1 mg/ml BSA and 100 mg/ml Ethidium Bromide. The beads were incubated overnight at 4uC with 3 ml of 35 S-labeled Grg4. The beads were washed three times with 500 ml of binding buffer containing 0.05% NP40 and boiled with SDS sample buffer. The amount of Grg4 was detected by autoradiography.
Embryo treatment, RNA purification and real-time quantitative RT-PCR (qRT-PCR) Developing embryos of B. floridae were collected into RNA later (Ambion). Some embryos treated with DMSO (control) or 3 mM Wnt signaling activator 6-Bromoindirubin-39-oxime (BIO; Sigma) [23] at blastula stage and allowed to develop until mid-neurula stage. Standard procedures were used for RNA purification and reverse transcription. Briefly, total RNAs were isolated from embryos using the Trizol reagent (Invitrogen); contaminating genomic DNA was eliminated by DNAse I digestion and RNA was repurified using RNeasy Micro kit (Qiagen). Random-primed cDNA was prepared in a 20 ml reaction from 100 ng of total RNA using SuperScript VILO cDNA Synthesis kit (Invitrogen). cDNAs were produced from at least two independent RNA isolations and the PCR reactions were performed in triplicate for each primer set. Control reactions (containing corresponding aliquots from cDNA synthesis reactions that were performed without reverse transcriptase; minus RT controls) were run in parallel. PCR reactions were run using the LightCycler 480 Real-Time PCR System (Roche). Typically, a 10 ml reaction mixture contained 5 ml of LightCycler 480 SYBR Green I Master mix (Roche), 1 ml of primers (final concentration 0.5 mM) and cDNA diluted in 4 ml of deionized water. Crossing-threshold (CT) values were calculated by LightCyclerH 480 Software (Roche) using the second-derivative maximum algorithm. The specificity of each PCR product was analysed using the in-built melting curve analysis tool for each DNA product identified; additionally, PCR products were verified by sequencing. All primers were calculated using Primer 3 computer services at http://frodo.wi.mit.edu/. The housekeeping gene encoding ribosomal protein L32, AmphiRPL32, was used as internal control gene to standardize the quality of different cDNA preparations. Primer sequences were as follows: AmphiRPL32  Figure S2 Wnt3A activates Xenopus Xvent-2B but not AmphiVent1 and AmphiVent2 promoters. Indicated luciferase reporter plasmids were transfected into 293T cells in the absence or presence of Wnt3A conditioned medium. Please, note that fold induction of individual reporter genes was normalized to activation of the promoter-less construct pGL3-basic. **P,0.01, ***P,0.001.