Regulation of a Vascular Plexus by gata4 Is Mediated in Zebrafish through the Chemokine sdf1a

Using the zebrafish model we describe a previously unrecognized requirement for the transcription factor gata4 controlling embryonic angiogenesis. The development of a vascular plexus in the embryonic tail, the caudal hematopoietic tissue (CHT), fails in embryos depleted of gata4. Rather than forming a normal vascular plexus, the CHT of gata4 morphants remains fused, and cells in the CHT express high levels of osteogenic markers ssp1 and runx1. Definitive progenitors emerge from the hemogenic aortic endothelium, but fail to colonize the poorly vascularized CHT. We also found abnormal patterns and levels for the chemokine sdf1a in gata4 morphants, which was found to be functionally relevant, since the embryos also show defects in development of the lateral line, a mechano-sensory organ system highly dependent on a gradient of sdf1a levels. Reduction of sdf1a levels was sufficient to rescue lateral line development, circulation, and CHT morphology. The result was surprising since neither gata4 nor sdf1a is obviously expressed in the CHT. Therefore, we generated transgenic fish that conditionally express a dominant-negative gata4 isoform, and determined that gata4 function is required during gastrulation, when it is co-expressed with sdf1a in lateral mesoderm. Our study shows that the gata4 gene regulates sdf1a levels during early embryogenesis, which impacts embryonic patterning and subsequently the development of the caudal vascular plexus.


Introduction
A characteristic feature of organ morphogenesis is the coordinated cell migration of groups of cells, resulting in the appropriate organ position, shape and size. The gata4 gene is known to play key roles in the development and maintenance of several organ systems, including those comprising cardiovascular, reproductive, and digestive tissues [1,2]. The gross phenotype of the mouse Gata4 mutant is defective embryonic folding, caused indirectly due to unknown alterations in extra-embryonic endoderm [3,4]. Conditional mouse mutants [5,6,7] and studies in zebrafish [8] defined functions for gata4 in early embryonic organ formation, including heart and liver. The mutants display morphogenetic defects, rather than a failure in cell specification or differentiation, because those functions are compensated by the sister genes gata5 and gata6 [9,10]. The morphogenetic genes that gata4 regulates remain largely unknown, although they include cell cycle regulators [11] and signaling molecules including BMPs and WNT inhibitors [12], which can function by non-cell-autonomous mechanisms. Given the pleiotropic nature of the gata4 mutant phenotype, the question arises whether a common morphogenetic program controlled by gata4 functions in different organ systems, or if varied tissue-specific programs are downstream of gata4 to control development of distinct organ systems.
We developed a loss-of-function model for gata4 in the zebrafish system using anti-sense morpholinos to block expression of the gene during embryogenesis [8]. An advantage of the zebrafish model is that it lacks the requirement of gata4 in extra-embryonic tissues, which in the knockout mouse leads to early embryonic lethality. Zebrafish embryos depleted of gata4 are defective for normal heart tube growth and looping, with a phenotype remarkably similar to mouse embryos lacking Gata4 in the embryo proper, and rescued for extra-embryonic function by tetraploid complementation [8,13]. In contrast to mouse mutants, the morphant fish embryos continue to survive even in the absence of a normal heart, and this allowed us to identify an additional function for gata4 in gut-derived organ growth. The finding that gata4 is needed for growth of both heart and gut-derived organs is not unexpected, since the gene is expressed in progenitors that contribute to these organ systems from early stages of embryogenesis. Here we revisit the phenotype of gata4 morphant embryos, and describe for the first time a function for gata4 in the development of the vascular system, specifically the caudal hematopoietic tissue (CHT). We show that gata4 regulates angiogenesis that normally forms this plexus during a ''fetal-like'' intermediate stage of hematopoiesis. This observation was unexpected because gata4 is not obviously expressed in the hematopoietic progenitors or the CHT.
Considering that the function of gata4 might be indirect, we discovered another previously uncharacterized phenotype, occurring in the lateral line of gata4 morphants. The lateral line is an ectodermal placode-derived system comprised of a set of pressure-sensitive sensory organs, the neuromasts, distributed in a stereotypic pattern along the body surface [14]. A major determinant of the collective cell migration required for lateral line development is the chemokine stromal-derived factor 1 or sdf1a (also called cxcl12a), mediated by cxcr4/cxcr7 chemokine receptor signaling [15,16,17,18]. This ligand is well characterized as a chemo-attractant controlling the migration of neurons and primordial germ cells, lymphocyte trafficking, and hematopoietic stem cell (HSC) homing [19,20,21]. In the lateral line, the pathway coordinates the migration and timely deposition of collective groups of cells that form neuromasts. Therefore, chemokine signaling also regulates tissue migration and adherence in order to facilitate organ system development. We find that the novel functions for gata4 in angiogenesis and sensory organ development can both be ascribed to deregulated expression of the sdf1a gene.

Ethics Statement
The only animals used were zebrafish. All experiments using zebrafish were previously approved following full review of the Weill Cornell Medical College Institutional Animal Care and Use Committee (IACUC), on file as protocol 2011-0100, which expires February 2015. Weill Cornell Medical College operates its NIH approved Animal Care and Use program under the Animal Welfare Assurance number A3290-01.

Zebrafish Strains
Zebrafish were raised and staged according to Westerfield [22]. The tg(gata1:dsRed) line [23] was provided by Shou Lin (UCLA, Los Angeles, CA). The tg(fli1:egfp) [24] was purchased from the Zebrafish International Resource Center (Eugene, OR). The tg(claudinb:gfp) line [17] was provided by James Hudspeth (Rockefeller University, New York, NY). The tg(cd41:gfp) strain [25] was provided by Robert Handin (Brigham and Women's Hospital, Boston, MA). The dominant-negative isoform for gata4 was generated essentially as described [26], but using the zebrafish gata4 cDNA clone for isolation of the DNA-binding domain, cloned in frame with the SR domain from mouse. A region of the zebrafish gata4 cDNA encompassing both zinc fingers and the entire DNA-binding domain (from amino acids 131-338) was isolated by PCR using primers that incorporated a stop codon and permitted in-frame fusion at the N-terminus with sequences encoding the strong repression domain (SR; amino acids 1-69) of the mouse Mxi1 gene, or a valine-proline mutation of the SR domain, SR-P [27]. PCR primers for gata4 were F: (TE2477), GAAGATCTTGGACGGCTTCGCAC.
R: (TE2476), AACTCGAGTTAGGATCCGCTTGGAGA. The chimeric cDNA encoding a protein called SR-Gata4 was confirmed by sequencing. The chimeric cDNA was cloned into the middle entry vector in the Gateway Cloning Kit using the In-Fusion TM Advantage PCR kit (Clontech) and primers: F: (TE2542), TATAGGGCGAATTGGGTACCATG-GAGCGGGTGCGGATGATCAA, R: (TE2543), GGGAACAAAAGCTGGAGCTCTTAG-GATCCGCTTGGAGAGCCCATA and recombined with the hsp70I promoter vector and IRES driving RFP vector into the tol2 destination vector. DNA was co-injected with RNA encoding Tol2 recombinase, and potential F0 founders were tested for transmission of the transgene, and inducible expression of the sr-gata4 RNA. Several independent lines were isolated and maintained as tg(hsp70:sr-gata4). Characterization of SR and SR-P constructs was performed as described [26]. For gel mobility-shift experiments, the probe (top strand) was 59-TCCATCTGATAAGACT-TATCTGCTGCC. The same (unlabeled) sequence was used as specific competitor, while the control mutant sequence (top strand) was: 59-TCCATCTCTTAAGACTTAAGTGCTGCC.

Morpholino Microinjection
All morpholinos were purchased from Gene-Tools, LLC. The sequence of the gata4 translation blocker is (59-TCCACAGGT-GAGCGATTATTGCTCC-39). It was injected in concentrations ranging from 10-12 ng, which is sufficient to reproducibly generate the full range of morphant phenotypes in essentially 100% of the embryos. The morpholino targeting sdf1a is also a translation blocker, validated in multiple previous studies [16,17,18] and shown to phenocopy an sdf1a mutant; its sequence is 59-TTGAGATCCATGTTTGCAGTGTGAA-39. For rescue experiments, embryos were injected first with the gata4-specific morpholino, and a cohort of these morphants injected subsequently with the sdf1a-specific morpholino at concentrations between 0.25-4ng, which is below the threshold for generating an sdf1a morphant phenotype.

Microscopy
All in situ hybridization and 4-Di-2-Asp images were taken with a Nikon SMZ 1500 microscope using a Spot Insight Firewire digital camera and advanced Spot imaging software (Morell Instruments Company, NY). All other images were acquired with a Zeiss Observer.Z1 microscope and a Zeiss Axiocam digital camera (Carl Zeiss MicroImaging). The Axiovision40 V4.8.1 software was used to collect Z-stacks of the images which were then flattened by using the extended focus feature. For confocal analysis, fixed embryos were mounted in 1% low melt agarose (National Diagnostics) dissolved in water. Confocal images were taken with an Olympus Fluoview Microscope with a 606 lens and analyzed using Fluoview software. Images were processed with ImageJ (1.42 q), Imaris (Bitplane Inc.), and Adobe PhotoshopCS4 software.

Quantitative PCR
Total RNA was isolated from embryos using the RNeasy kit (Qiagen). First-strand cDNA synthesis was performed using the Superscript III kit (Invitrogen). The cDNA was analyzed with the Light Cycler 480 II SYBR green master mix (Roche), and using the Light Cycler 480 II machine (Roche). All samples were prepared in triplicate, and each experiment was repeated at least 3 times using independent batches of embryos. The PCR cycle conditions were 95uC for 5 minutes, (95uC for 10 seconds, 54uC for 10 seconds, and 72uC for 15 seconds) for 40 cycles. The data were analyzed to determine Cp values using the 2 DDT method [31], normalized to transcript levels for b-actin. Relative morphant transcript levels were quantified considering wild-type levels as 1.

Gata4 Regulates Vascular Development
In a previous study [8], we noted that by 4 days post fertilization (dpf) there is a marked absence of blood flow through the improperly looped but still beating heart of the gata4 morphant embryo. This was unexpected since gata4 is not thought to be expressed in the hematopoietic system. Furthermore, although the morphant heart is abnormal, it continues to beat, and the embryonic morphant circulatory system is at least grossly normal. Therefore, we investigated the basis for the hematopoietic deficit. The initiation of hematopoiesis appears normal in gata4 morphants based on in situ hybridization experiments to detect transcripts for the erythroid transcription factor gata1 (not shown), and by imaging RFP+ cells in tg(gata1:dsRed) reporter fish ( Fig. 1A-D). Also, in situ hybridization experiments using an l-plastin probe showed that early myelopoiesis is also not affected by loss of gata4 (Fig. 1E,F). Therefore, the initial generation of ''primitive'' waves of hematopoiesis is not regulated by gata4. Although the embryonic vasculature including angiogenesis of the intersomitic vessels (ISVs) is grossly intact (see Fig. 2), the heart defect could lead to a loss of blood flow and pooling of the blood at later stages. We used quantitative flow cytometry experiments to compare the number of erythroid cells in control and gata4 morphant embryos derived from tg(gata1:dsRed) reporter fish. In morphants, the number of RFP+ cells is significantly decreased by 4 dpf to a level on average of 53% the number in control embryos (n = 14 independent batches, p,0.001). The embryos live for several more days, but hematopoiesis does not recover. As in the mouse, several ''waves'' of definitive hematopoietic progenitors have been described during early zebrafish embryogenesis [32]. Between 30-36 hpf, erythro-myeloid committed progenitors (EMPs) arise in the posterior blood island, derived from lmo2+ lateral mesoderm [33]. Around this same time, cells that are thought to be the long-term adult HSCs begin to emerge from the floor of the dorsal aorta and subsequently migrate to the same region of the posterior tail, which has matured into a more vascularized niche, termed the CHT [34]. These aorta-derived HSCs do not contribute to erythropoiesis until a later stage compared to the time we document defects in hematopoiesis for the gata4 morphant (4 dpf). This suggests that the gata4 morphants have a late defect in primitive erythropoiesis, correlating with an intermediate stage during which time the CHT plexus develops.
To investigate this intermediate stage, the development of the CHT was evaluated in gata4 morphants using embryos derived from double transgenic tg(fli1:gfp; gata1:dsred) fish, to follow the codevelopment of CHT vasculature and circulating erythroid cells [34]. As shown in Fig. 2 (panels A,B) the CHT of gata4 morphants appears normal at 28 hpf, and erythroid cells circulate through the tail as in controls. However, by 3 dpf there is a significant disruption in the morphology of the CHT in gata4 morphants compared to controls (Fig. 2C,D). Circulation does not extend deeply into the tail and is limited to a much more proximal region of the trunk (note the length of the blue bracket indicating the morphant tail that lacks circulation in Fig. 2D). The vascular plexus is significantly reduced in size and extent of branching. The phenotype can be seen already by 48 hpf, when a clearly defined plexus between the dorsal aorta and caudal vein is seen in control embryos (Fig. 2E, denoted by the white arrows), which is not seen in the morphant embryos (Fig. 2F). Confocal microscopy was used to evaluate with higher resolution angiogenesis associated with the CHT. As shown in Fig. 3, the normal CHT plexus develops with a series of looped structures (indicated by asterisks in Fig. 3A). These structures are missing in the gata4 morphants, which develop a smooth unfenestrated tissue lacking looped structures (Fig. 3, panels B,C). Thus, while the primitive hematopoietic and vasculogenic programs initiate normally, gata4 morphants are disturbed subsequently for development and vascularization of the CHT, and this correlates temporally with a relative loss of circulating blood cells by 4 dpf. The failure to form a normal CHT plexus is not simply an indirect defect caused by a poorly functioning heart in the gata4 morphant. Two independent morphants that also display similar cardiac morphogenetic phenotypes and poor circulation (gata5 and tbx2a) have normal CHT plexus formation (Supplemental Fig. S1). Previously, Choi et al. demonstrated using defined mutants (tnnt2 and smo) that circulation is not required for the formation of the plexus in the CHT, although it can affect the subsequent stability of the structure [35].
HSCs are derived from hemogenic endothelium at the floor of the dorsal aorta [36,37], and can be phenotypically identified by expression of HSC markers including cd41 [38] and runx1 [39]. To further evaluate circulation into the CHT, we analyzed the development of progenitors in morphant embryos from the cd41:gfp transgenic reporter line and confirmed that GFP+ cells emerge from the dorsal aorta similar to wildtype embryos. However, unlike control embryos, the GFP+ cells are never detected subsequently associated with the CHT (Supplemental Fig. S2). By 48 hpf, runx1+ progenitors in control embryos are found in the CHT or in circulation. However, at this same stage, runx1+ cells remain scattered along the midline of morphant embryos (Fig. 4A,B; see asterisks in panel B'). In addition, a strong patch of runx1 staining is found at 48 hpf associated with the CHT, which is never seen in control embryos (Fig. 4B, arrow). In addition to its function as a cell-autonomous regulator of HSC [30,39], runx1 is expressed in prechondrocytic tissue that forms embryonic cartilage, and in periosteal and perichondral membranes of bone [40,41]. We evaluated the expression pattern for various known bone marrow stromal factors and found a striking accumulation of transcripts for osteopontin (spp1), also associated with this runx1+ domain of the morphant CHT at 48 hpf (Fig. 4C,D). Therefore, when gata4 is depleted, definitive progenitors are born in the hemogenic endothelium. We do not know the fate of these progenitors, but based on the abnormal vascularization of the CHT, it appears that they fail to seed the CHT. At the same time, the CHT region expresses aberrantly at least some markers normally associated with osteochondrogenic tissue.

The sdf1a Expression Pattern is Altered in gata4 Morphants
In addition to spp1, the expression patterns for additional known stromal factors were evaluated in the gata4 morphants, including ang1, jagged1, ncad, sdf1a, and sdf1b. The expression patterns for ang1, jagged1, ncad, and sdf1b appeared unchanged in the gata4 morphant (data not shown). Only spp1 showed the aberrant pattern in the CHT, but an abnormal expression pattern was also found for the chemokine sdf1a (Fig. 5). Transcripts for sdf1a are normally found between 24-48 hpf at defined structures of the head, pronephric duct, and in a single stripe of mesoderm running down either side, for the length of the fish [15,16]. In wildtype control fish these mesodermal stripes are clearly evident at 30 hpf (Fig. 5A), while transcript levels are reduced to near background levels in the tail by 35 hpf (Fig. 5C). The stripes are evident, albeit often patchy at 30 hpf in gata4 morphant embryos, and embryos showed enhanced relative expression levels in the pronephric duct (Fig. 5B). More strikingly, the transcript levels persist in the gata4 morphants and strong staining, relative to the control embryos, remains in the lateral stripes and pronephric duct at 35 hpf (Fig. 5D). Thus, sdf1a-dependent signaling that is normally downregulated, may be inappropriately persistent in the gata4 morphant embryo. We compared expression levels by quantitative RT-PCR in isolated tail tissue from control and morphant embryos at 40, 48, and 72 hpf, representing the period when normal CHT development fails in the morphants (Fig. 5E). Fully consistent with reporter fish results and in situ hybridization data, in morphant tissue there is a significant loss of globin transcripts (consistent with a defect in primitive red blood cell maturation), and a major decrease in transcripts for hematopoietic progenitors, including lmo2 and cmyb, while both runx1 and sdf1a levels are significantly increased in the tails of morphant embryos.

Gata4 Regulates the Development of the Posterior Lateral Line
The changes in sdf1a expression pattern suggested a possible stromal-based defect caused by gata4 depletion, but does not ascribe a functional role. While Sdf1 is known for its function in mammalian hematopoiesis, it is also well characterized in zebrafish as a key regulator of the lateral line, a mechanosensory organ system [42]. When analyzing gata4 morphant embryos around 3-4 dpf, we discovered that they fail to initiate an ''avoidance'' or ''startle'' response, for example when touched by a pipet. The morphants are not paralyzed and swim about of their own volition. Mediation of the startle response is a major function for the lateral line. The lateral line is comprised of neuromasts, sensory organs that contain central rosettes of hair cells, similar to those found in the mammalian inner ear, surrounded peripherally by supporting cells, and innervated by sensory neurons from the cranial ganglion. The neuromasts of the posterior lateral line are derived from a primordium originating from the cephalic placodes, located posterior to the otic placodes [14]. At around 22 hpf, a primordium on each side of the embryo, consisting of ,100 cells, migrates distally along the lateral body wall, reaching the tip of the tail by 48 hpf. During the migration, groups of cells from the trailing edge of the primordium are deposited, every 5-6 somites, on each side of the embryo. These cells, the proneuromasts, differentiate into either hair cells or supporting cells. The neuromasts integrate mechano-sensory signals that control balance and the startle response. Therefore, the gata4 morphants display a behavior expected for embryos with a defective lateral line.
To assess the integrity of the lateral line, we first examined the development of the neuromasts. Control or morphant larvae were stained at 5 dpf in 4-Di-2-ASP, a cationic vital fluorophore that enters hair cells of the neuromasts through mechanoelectricaltransduction channels (Fig. 6A,B). The gata4 morphant embryos display a very low number of tail neuromasts (Fig. 6B). Although variable, the anterior lateral line is less affected, and neuromasts are often found in the head regions. However, most morphant embryos lack all or most of the trunk neuromasts that comprise the posterior lateral line, and this explains their failure to generate a startle response.
While many gata4 morphants lack all of the 4-Di-2-ASP staining trunk neuromasts, occasional neuromasts did stain. Notably, many of the morphant larvae displayed differentiated neuromasts at the extremity of the tail. Therefore, it seemed unlikely that the relative lack of trunk neuromasts is caused by a failure of the posterior lateral line primordium to migrate. Rather, we hypothesized that gata4 Regulates Angiogenesis via sdf1 PLOS ONE | www.plosone.org the primordium migrates, but that proneuromasts fail to be deposited or to differentiate normally. The migrating primordium is followed by a neurite that will innervate the future neuromasts. We reasoned that if the primordium migrates distally, the posterior lateral line neurons should be present in the gata4 morphant embryos. To test this hypothesis, larvae at 5 dpf were analyzed by whole-mount immunohistochemistry using an anti-acetylated tubulin antibody that detects neurons (Fig. 6C,D). Indeed, the staining confirmed the existence of the posterior lateral line neurons, extending along the body length to the tail tip. This pattern suggests that the primordium migrated normally along the lateral line path. However, there was a complete lack of normal branching of these neurons toward the presumed sites of neuromast deposition, suggesting that the defect in posterior lateral line development is due to a failure in the deposition of neuromasts by the migrating primordium. We further evaluated the migration of the primordium directly using the tg(claudinb:gfp) reporter fish [17]. In these embryos the primordium is imaged live, seen as a migrating patch of GFP+ cells, confirmed in both control and gata4 morphant embryos (Fig. 6E,F). As expected, the morphants generally failed to deposit proneuromast clusters. In morphant embryos the primordium often migrated along its normal route (as shown here), but it was always significantly delayed, and in some cases the primordium wandered off its normal lateral line route. Thus, in gata4 morphants the primordium forms and migrates, but migration rate and neuromast deposition is significantly disturbed. In panels A and B, 2 independent embryos are shown representing the range of patterns detected. Red arrows pointing down show the strong stripes of expression in lateral mesoderm in control embryos, which is nearly gone by 35 hpf. In contrast, the pattern is less clear in mesoderm of 30 hpf morphant embryos, but strongly persists by 35 hpf, and is much enhanced in the region associated with the pronephric tubules, particularly at 35 hpf (black arrows pointing up in panels B and D). Views are lateral, anterior to the left. (E) Tails including the CHT region were isolated from control and gata4 morphant embryos at 40, 48, or 72 hpf and processed by qPCR to measure relative levels of ae3globin, runx1, lmo2, cmyb, and sdf1a. The Y axis plots the relative transcript levels of morphants compared to controls (controls set at a value of 1). Each sample included .30 embryos, and data was collected in 3 independent experiments. Student's T-test was used to determine significance (*p,0.05, **p,0.01, ***p,0.001). doi:10.1371/journal.pone.0046844.g005

Knockdown of sdf1a can Rescue the Lateral Line and CHT Defects Caused by Depletion of gata4
The unexpected defect in lateral line development, associated here with enhanced transcript levels for sdf1a, prompted us to test if sdf1a is epistatic to gata4, and therefore functionally implicated in one or more of the phenotypes. If increased levels of sdf1a, caused by depletion of gata4, are functionally relevant, then reduction of sdf1a in the gata4 morphant should abrogate the phenotype. In this case it is important to reduce sdf1a levels, but not below the threshold needed for its normal function. To test this, we used a morpholino that has been previously validated for knockdown of sdf1a and titrated by micro-injection to define the amount that causes reproducible lateral line defects similar to the sdf1a mutant [18]. We found lateral line development was normal in embryos injected with up to 4 ng of the sdf1a morpholino, while embryos that had been injected with 8 ng of the sdf1a morpholino fail to develop a lateral line. We therefore injected embryos first with the gata4-specific morpholino and then subsequently injected a cohort of these morphants with up to 4 ng of the morpholino targeting sdf1a. As little as 0.5 ng of the sdf1a morpholino was sufficient to effectively rescue the lateral line defect caused by loss of gata4 ( Fig. 7A), as shown by restoration of the 4-Di-2-ASP staining pattern. This amount of sdf1a morpholino was not sufficient to rescue CHT development (not shown). However, injecting 4 ng of sdf1a morpholino in the background of the gata4 morphant was able to effectively rescue circulating blood levels, demonstrated with the tg(gata1:dsRed) transgenic reporter (Fig. 7B). This rescue correlates with enhanced vascularization and blood flow into the CHT. Therefore, through modulating sdf1a expression levels, gata4 normally regulates lateral line and angiogenesis in the CHT. In contrast, we were unable to consistently rescue heart development in gata4 morphants through manipulation of sdf1a levels (Supplemental Fig. S3).

Gata4 Regulates sdf1a during Gastrulation, with Subsequent Consequences for CHT Development
While a genetic interaction between gata4 and sdf1a was shown by the rescue experiments, the result is enigmatic, since neither gata4 nor sdf1a is obviously expressed in the CHT. While sdf1a transcripts persist in the morphant tail during CHT development, gata4 does not appear to be expressed in the tail. Although we cannot rule out low levels of expression, gata4 and sdf1a are co- expressed at a much earlier stage of development, in lateral mesoderm [43,44,45]. To determine when gata4 function is required for CHT development, we needed to generate a new animal model to control gata4 activity in a conditional manner, since morpholinos knockdown gata4 function throughout embryogenesis. For this purpose we generated a dominant-negative isoform of gata4 and created transgenic lines of fish that could be induced to express it at any time of embryogenesis by heat shock. Our strategy was the same as one we used previously in published studies to block the function of gata4 during Xenopus development [26]. Sequences encoding the well-defined DNA-binding domain of zebrafish Gata4 (amino acids 131-338) were fused with a strong-repressor domain derived from the amino terminal (amino acids 1-69) murine MXI1 transcriptional repressor protein, or a point mutant version of SR, SR-P (Supplemental Fig. S4). The SR domain of MXI1 (but not SR-P) interacts with the general chromatin repressor protein SIN3 to mediate suppression of MYC oncogenic activity [27], and when fused to a heterologous DNAbinding domain, the SR domain targets SIN3 to gene-specific binding sites leading to repression [46]. We used gel mobility-shift assays, and reporter assays to show that SR-Gata4 blocks coexpressed Gata4 from activation of target genes ( [26], Supplemental Fig. S5).
In preliminary experiments we also tested the homologous SR domain derived from zebrafish mxi1, but found no differences compared with the murine-derived domain (not shown). Injection of 250-1000 pg RNA encoding SR-Gata4 into fertilized eggs caused dose-dependent developmental defects including cardiomyopathies, whereas injection of RNA encoding SR-P-Gata4 at equivalent doses was much less active (not shown). We cloned the cDNA encoding SR-Gata4 downstream of the hsp70I heat-shock promoter in a tol2-dependent transgenic destination vector (Supplemental Fig. S6), and used this to generate lines of fish that express SR-Gata4, dependent on heatshock.
Embryos derived from the tg(sr-gata4) line appeared entirely normal when grown at standard 28.5C temperature. When embryos were heat-shocked for one hour at 37-39C SR-specific transcripts were readily detected by qPCR (not shown). Transgenic embryos that were heat shocked (but not sibling nontransgenic, or transgenic but non-heat-shocked embryos) developed heart and endoderm defects that phenocopy the gata4 morphant embryos, as expected (Supplemental Fig. S6). Importantly, when heat-shocked for one hour during gastrulation, the majority of the transgenic embryos developed with a fused CHT that failed to generate a vascular plexus (Fig. 8). Heat-shock of the embryos at later stages, including 10 or 14 hpf during somitogenesis was still sufficient to generate cardiac looping defects, indicating that gata4 function remains important during these stages for heart development. In contrast, embryos heatshocked at the later stages are much less likely to develop defects in the CHT plexus (Fig. 9A). Importantly, heat-shock of the tg(sr-gata4) embryos at 3 hpf was sufficient to cause a modest but significant increase in sdf1a expression levels during subsequent stages of epiboly (Fig. 9B), when sdf1a signaling is known to control the interaction of endoderm and overlaying lateral mesoderm. Therefore, although the vascular plexus does not form until later, it appears that gata4-dependent regulation of sdf1a expression levels during gastrulation is important for mesoderm patterning relevant to CHT development.

Discussion
In addition to its roles in cardiogenesis, gut organogenesis, and gonadal development, we discovered previously unknown functions for the transcription factor gata4 in two additional organ systems: the vascular system and the lateral line. Recently, a functionally redundant role for gata4/5/6 genes was shown for the development of primitive myeloid lineages [47]. However, that reflects a patterning role in lateral mesoderm for the generation of anterior primitive hemangioblasts, distinct from the non-redundant role of gata4 we describe here. With respect to blood development, embryos depleted of gata4 do not develop a normal vascular niche that is required during the transition from primitive to definitive lineages (the CHT), and this can explain the failure in these embryos to expand or maintain early progenitors and a lack of blood as the initial wave of primitive cells declines. The lateral line defect prompted an evaluation of sdf1a, which we found to be overexpressed in the gata4 morphants. Functional epistatic tests showed that both lateral line and CHT defects can be reversed in the gata4 morphants by depleting sdf1a levels, thus confirming a common, presumably non-cell-autonomous, signaling defect. In contrast, the cardiac and gut phenotypes, which may require gata4 cell-autonomously, are not rescued by reduction of sdf1a levels.
In lateral line development the mechanism of sdf1a function has been studied extensively [48]. Active sdf1a-dependent cxcr4 signaling is required at the leading edge for directed migration of the entire primordium, while cxcr7 interprets sdf1a levels distinctly at the trailing edge where the release of proneuromast clusters is facilitated. In gata4 morphants the primordium migrates, consistent with the presence of sdf1a. However neuromasts fail to be deposited, which is consistent with excessive Sdf1a ligand and the inability of the cxcr7 receptor to interpret the gradient needed for deposition. The sdf1a-dependent signaling impacts the relative adhesion of interacting tissues. During gastrulation this controls the extent of anterior endoderm migration, relative to associated mesoderm [44,45]. This serves to ''tether'' the tissues in an appropriate position, at least in part through the stimulated expression of integrins, thus facilitating spatial control of subsequent inductive signals across germ layers. While the previous work documented that disruptions in this signalingdependent germ layer interaction alters endoderm patterning and subsequent endoderm organogenesis, our data indicate that it also affects mesoderm patterning and subsequent vascular development of the CHT. While gastrulation and lateral line development involve collective cell migrations, it was possible that the defect in hematopoiesis could be due to a cellular chemo-attraction issue, reflecting a function more similar to that described in bonemarrow homing of hematopoietic stem cells [49]. Indeed, the expression and function of sdf1a in the adult zebrafish kidney is well conserved with the mammalian bone marrow program [50]. The cxcr4a gene facilitates regional patterning of the early aortic endothelium [51], although in this case mediated by sdf1b. Inappropriate levels of Sdf1a ligand in the region of the pronephros could in principle disrupt the normal migration pattern of definitive progenitors to the CHT. However, our data suggest that the defect can be explained instead by a mesoderm patterning defect leading to morphological disruption in the CHT vasculature. By 36 hpf we found in gata4 morphants aberrant expression of spp1 and runx1 in the CHT, again suggesting that normal mesoderm patterning was disrupted by loss of gata4 function. This tissue fails to be fully vascularized by 3 dpf, and so hematopoietic progenitors would be unable to seed the tissue regardless of misplaced signaling cues. To gain insight into when gata4 expression is required to regulate CHT development we needed to develop a strategy for conditional depletion of gata4. This was accomplished through expression of a dominant-negative isoform of Gata4 (SR-Gata4) under the control of the heat shock promoter. The CHT failed to form a normal vascular plexus when gata4 function was disturbed during gastrulation, while it formed normally if SR-Gata4 was expressed later during somitogenesis. Cardiac development was disrupted when SR-Gata4 is expressed at either time point. Since the SR (repressor) isoform phenocopied loss of gata4, including enhanced levels of sdf1a, the result is consistent with a normal function for gata4 to activate gene(s) that restrict normal sdf1a expression levels (indirectly). Interestingly, a recent cell-tracking analysis in the zebrafish cmyb mutant documented defects in migration of the HSC from the dorsal aorta, which was attributed at least partly to elevated expression levels of sdf1a [52]. This fits well with the HSC phenotype observed in the gata4 morphants, suggesting that mesoderm patterning may not be the only cause for the failure of HSCs to migrate to the CHT.
We speculate that one role of the CHT may be functionally homologous to the mammalian liver, as a transition zone to support hematopoiesis prior to establishment of the adult stem cell populations in the bone (mammals) or kidney (fish). The same gene, gata4, is an essential regulator of liver growth and development in both mammals and fish, and therefore may represent an ancient link to the evolution of hematopoietic niches. Whether the gata4-sdf1a axis is conserved in relation to murine fetal hematopoiesis is not known, although Palis and colleagues [53] showed that hematopoietic progenitors in the fetal liver are responsive to an SDF1 gradient. Furthermore, in separate studies we used a murine embryonic stem cell model that allows conditional control of Gata4 expression [12]. When murine Gata4 was expressed for 6 hours starting at day 2 in embryoid body (EB) cultures, the top most down-regulated gene was Sdf1, based on microarray analyses, and confirmed by qPCR (not shown). Thus, it seems likely that Gata4 has a conserved role for restricting Sdf1 during embryonic development. Figure S1 The CHT defect caused by depletion of Gata4 is not seen in other models of embryonic cardiomyopathy. The heart and caudal tail regions are shown in representative 48 hpf tg(gata1:dsred; fli1:egfp) uninjected control embryo (top), or embryos that had been injected with morpholinos targeting gata5 (middle) or tbx2a (bottom). In the trunk images, white arrows indicate examples of characteristic fenestrated structures that start to form around 32 hpf in the CHT plexus, but fail to do so in the gata4 morphant. The control embryo has a normally looped heart, while the gata5 morphant has an unlooped heart-string (indicated by the double-headed arrow), and the tbx2a morphant has an extended dysmorphic atrium (indicated by the yellow arrow). Morpholinos were used under conditions that generate reproducible cardiomyopathies and poor circulation with pericardial edema, as documented in our own and others studies: gata5, 59-AAGATAAAGCCAGGCTCGAATACAT, 10 ng/embryo; tbx2a, 59-CGGTGCATCCAACAAACGTAGTGAA, 5 ng/ embryo. (TIF) Figure 9. SR-Gata4 blocks CHT plexus formation and is sufficient to enhance sdf1a levels during epiboly, when expressed during gastrulation but not when expressed later during somitogenesis. A: The graph shows the percentage of embryos that form fenestrated structures typical of a normal CHT plexus. Embryos derived from crossing the sr-gata4 transgene into the tg(fli:egfp) background were either left at normal temperature (control) or heat shocked at 37C for one hour starting at 3 hpf, 5 hpf, or 14 hpf, as indicated. Because there is some variation in normal embryogenesis, the clear formation of at least 5 fenestrations, imaged at 48 hpf, was used as criteria for plexus formation. Over 80% of the embryos heatshocked at 3 hpf failed to form a plexus (*p,0.001), while essentially all the embryos heat-shocked at 14 hpf were normal. Embryos heatshocked at 5 hpf showed a trend toward defective plexus formation, but it was not statistically significant. For each sample, n.90, with data compiled from at least 3 independent experiments. B: Transgenic tg(hsp70I:sr-gata4) embryos were heat-shocked for one hour at 37C starting at 3 hpf, and batches of embryos were collected at the times indicated for analysis of sdf1a levels by qPCR. After normalization to beta-actin, the products were plotted relative to control non-heat shocked embryos, given an arbitrary value of 1. By 6 hpf there is a statistically significant increase in sdf1a transcript levels, which continues through at least 10 hpf. ** p,0.01, ***p,0.001. For each sample, n = 20-30 embryos, and the data is compiled from 4 independent experiments. doi:10.1371/journal.pone.0046844.g009 Figure S2 Progenitors that develop in association with the dorsal aorta fail to populate the CHT. Shown are representative embryos (n .50) derived from tg(cd41:gfp) transgenic reporter fish, in which GFP expression marks the wave of definitive progenitors that are born in the floor of the dorsal aorta (indicated by the arrows in A, B). At least some of these migrate and seed the CHT of control embryos (boxed area in A). While the appearance of GFP+ cells from the dorsal aorta appears normal in the gata4 morphant embryos (arrows in B), GFP+ cells are never found in the morphant CHT (equivalent boxed area in B). A' and B' are enlarged views of the dashed boxes shown in A and B. Views are lateral, anterior to the left. (TIF) Figure S3 Modulation of Sdf1a is not sufficient to rescue the cardiac defect of gata4 morphants. A) Control uninjected embryos at 2 dpf have normal looped hearts, as visualized in the tg(myl7:gfp) transgenic reporter background. B) Hearts fail to loop in the gata4 morphant embryos. C) Likewise, the hearts fail to loop properly in gata4 morphants that are co-injected with the sdf1a morpholino. In the example shown here, 2 ng of sdf1a morhoplino was used, but we also failed to reproducibly rescue cardiogenesis with higher or lower doses. (TIF) Figure S4 Schematic illustration of the cDNA encoding a dominant-negative SR-Gata4. Shown on top are the previously studied versions of SR-Gata4 that was used to block Xenopus Gata4 function (ref. 26). The SR-P domain contains a single base change causing a valine to proline missense mutation, largely eliminating the repressor activity. PCR was used to generate the analogous sequence from the zebrafish cDNA, which was sub-cloned in frame with the SR or SR-P sequences. On the right is an autoradiogram after SDS-PAGE of lysates following in vitro transcription/translation, showing the correct expression of approximately 31 kD predicted molecular weight proteins encoding the zebrafish DNA-binding domain, and the somewhat smaller Xenopus version. The control lane sample omitted RNA. (TIF) Figure S5 The SR-Gata4 proteins bind with specificity to a cognate GATA-binding site element. Shown is an autoradiograph following electrophoresis of a specific probe containing a strong GATA-binding consensus element, following incubation with lysates generated using the in vitro translated Xenopus SR-Gata4 (SR-xG4), the zebrafish SR-Gata4 (SR-zG4), or the zebrafish SR-P-Gata4 (SR-P-zG4) proteins. For each set, the first lane lacked competitor DNA. Subsequent lanes had samples with 106 or 206 molar concentration of the specific (S-comp) or mutated non-specific (NS-comp) double-stranded competitor oligomers added to the reaction. Note that 106 of the specific competitor nearly abolishes the complex binding to the probe, while 106 of the non-specific DNA has no effect. Probe indicates the migration of free probe compared to the higher DNA-protein complexes. The larger zebrafish-derived proteins show a higher shift. (TIF) Figure S6 Conditional expression of SR-Gata4 phenocopies the gata4 morphant. A) The schematic shows the transgenic construct used to generate lines of tg(hsp70I:sr-gata4) zebrafish. In the context of a tol2-flanked destination vector, the hsp70I promoter was used to direct expression of SR-Gata4, followed by an IRES that allows co-expression of RFP (cherry). An myl7 promoter directing expression of GFP to the heart is also present as a marker for transgenesis. B) Shown are representative examples of three cohorts of embryos obtained by crossing zebrafish heterozygous for the transgene, imaged under fluorescence with the green channel. The top embryo is GFP+, indicating that the embryo carries the transgene. This embryo was also heatshocked at 37C for one hour, in this case at 5, 24, and 27 hpf. The embryo in the middle was treated the same, but is GFP-, indicating that this sibling does not carry the transgene. The bottom embryo is GFP+ (transgenic) but was kept at 28.5C (no heat shock, HS-). Note that only the top embryo shows the characteristic gata4 morphant phenotype: an improperly looped heart and pericardial edema (red arrow) and a failure in the yolk stalk extension (yellow arrow). C) The same embryos were imaged under the red channel, showing that only the top embryo expresses RFP, since it is the only embryo that is both transgenic and heat shocked. Essentially every transgenic embryo shows this heat-shock dependent phenotype (n.100). (TIF)