hrT is required for cardiovascular development in zebrafish

The developmental roles of the T-box genes have been an area of intense research for many years (Papaioannou and Silver, 1998; Tada and Smith,2001). Much of our current understanding of the critical role of T-box genes comes from the analysis of phenotypes produced by genetic mutations. Importantly, several TBX genes are causally implicated in human congenital malformations. For example, the cardiovascular defects in DiGeorge syndrome are caused by haploinsufficiency of TBX1(Merscher et al., 2001); the ulnarmammary syndrome is a pleiotropic disorder characterized by defects in limb, apocrine gland, tooth and genital development because of haploinsufficiency of TBX3(Bamshad et al., 1997); and Holt-Oram syndrome, a developmental disorder of upper limb malformation and cardiac septation defects, is associated with haploinsufficiency ofTBX5 (Li et al.,1997; Bruneau et al.,2001).

The family of T-box transcription factors shares a phylogenetically conserved DNA-binding domain, which is required for specific DNA sequence recognition. The functional activity of these genes are mediated by binding to their cognate regulatory sites within the promoter and enhancer regions of genes, leading to the activation or repression of gene expression depending upon the type of T-box gene and the context within the promoter(Carreira et al., 1998;He et al., 1999;Sinha et al., 2000;Tada and Smith, 2001). Members of this gene family have been shown to have selective patterns of expression and play many key roles in patterning and specifying the development of a variety of tissues (Basson et al.,1999; Merscher et al.,2001; Lamolet et al.,2001; Bruneau et al.,2001; Tada and Smith,2001).

We recently described a novel zebrafish T-box gene, hrT(tbx20 — Zebrafish Information Network), which is expressed in the developing heart and dorsal aorta(Griffin et al., 2000;Ahn et al., 2000). hrTis expressed in the cardiogenic mesoderm of the anterior lateral plate from the beginning of segmentation, and continues to be expressed in the heart field until at least 72 hours (Griffin et al., 2000; Ahn et al.,2000). The onset of hrT expression in the lateral plate is earlier than the first expression of nkx2.5 or tbx5, and coincident with the start of nkx2.7 expression(Lee et al., 1996;Begemann and Ingham, 2000).hrT is thus expressed during all the key stages of cardiac development, which include cardiac cell fate specification, morphogenesis,looping of the heart tube and chamber formation (Srivastavia and Olson, 2000;Yelon et al., 1999;Stainier, 2001). In addition,hrT begins to be expressed in the dorsal aorta at the 15-somite stage, in addition to sites of expression in the hindbrain, eye and at the anal opening (Griffin et al.,2000; Ahn et al.,2000).

To investigate the developmental role of hrT, we usedhrT-specific morpholino antisense oligonucleotides to produce zebrafish with a reduced amount of HrT. We find that hrT plays an important role in the later development of the heart, including normal cardiac looping and the division of the cardiac chambers. Specifically, we show thathrT is required to regulate tbx5 expression levels correctly during the stages of cardiac looping. In addition, we find that hrTis required for the morphogenesis of the dorsal aorta, resulting in embryos lacking blood circulation. Interestingly, the vascular defects in hrTmorphant embryos are similar to midline mutants such as floating head(flh), which we show lack expression of hrT in vascular progenitors. Taken together, these data indicate that hrT helps to mediate essential functions in vascular morphogenesis downstream of the midline mesoderm. This study provides the first demonstration of the crucial role for hrT in cardiovascular development and sheds new insight into the mechanism regulating the expression of tbx5 during cardiogenesis.

Zebrafish embryos were obtained by natural spawning of adult AB strain zebrafish. Embryos were raised and maintained at 28.5°C in system water and staged as described (Westerfield,1995). Morpholino antisense oligonucleotides were purchased from Gene-Tools (Corvallis, OR). A stock solution was prepared as described(Nasevicius and Ekker, 2000). At the one-cell stage, each embryo was injected with approximately 1 nl volume of morpholino oligonucleotide (0.2 to 5.0 ng)using a Picospritzer II (Parker Hamnifin Corporation). Embryos were collected at the appropriate stages and fixed in 4% paraformaldehyde, pH 7.0, in phosphate-buffered saline (PBS), overnight at 4°C. Fixed embryos were dechorionated, washed three times with PBS and stored in methanol at-20°C.

Whole-mount in situ hybridization was performed using digoxigeninlabeled antisense RNA probe and visualized using anti-digoxigenin Fab fragments conjugated with alkaline phosphatase (Roche Molecular Biochemicals) as described (Griffin et al.,1998). Riboprobes were made from DNA templates, which were linearized and transcribed with either SP6 or T7 RNA polymerase. Embryos were processed and hybridized as described(Griffin et al., 1998), except that 10 μg/ml of proteinase K in PBS/0.1% Tween-20 was used for 10 to 30 minutes depending the age of the collected embryos.

Whole-mount immunostaining was performed using monoclonal antibodies MF20(generous gift of Dr Stephen Hauschka) and S46 (generous gift of Dr Frank Stockdale). Embryos were processed as previously described(Westerfield, 1995), except that the pericardium of each embryo was punctured before treatment with proteinase K using a fine gauge syringe needle. The color reaction was visualized with goat anti-mouse antibodies conjugated to horseradish peroxidase and 3,3′-Diaminobenzidine tetrahydrochloride(Polysciences).

Whole-mount in situ hybridized embryos were washed in water several times,and placed in a series of washes in 25%, 50%, 75% and 100% ethanol. The completely dehydrated embryos were cleared in acetone and then embedded in Paraplast II (Tissue Tek). Sections 7 μm thick were cut and mounted on glass with the same embedding plastic. The mounted slides were covered with coverslips and incubated at 60°C to dry.

RNAs were synthesized from Asp718 linearized CS2-hrT-GFP(details available on request) and CS2-GFP (generous gift of Dr Jeff Miller) templates using the mMessage Machine kit (Ambion) and dissolved in RNase-free sterile water. RNA (0.1 ng) was injected in the presence or absence of 1.5 ng hrTMO(1) into one cell zebrafish embryos. The expression ofhrT-GFP and GFP were analyzed at the shield stage using green fluorescent microscopy.

Synthetic capped mRNA transcripts were synthesized by SP6 in vitro transcription (mMessage machine; Ambion) of an Asp718 linearized CS2 template containing an insert of the coding region of hrT fused to the glucocorticoid receptor ligand binding domain (GR-hrT; details available on request). GR-hrT mRNA was dissolved in RNase-free sterile water and 1 nl volume of RNA at a concentration of 0.2 mg/ml was injected into one cell zebrafish embryos. The GR-hrT protein was activated by adding 0.1% volume of 100 mM dexamethasone in 100% ethanol to give a final concentration of 100 μM dexamethasone. Control treated embryos were put into 0.1% ethanol at the same time.

For photography, whole-mount in situ hybridized embryos were post-fixed in 4% paraformaldehyde, washed three times with PBS, dehydrated with methanol,cleared in methyl salicylate and mounted onto a glass slide with Permount as described (Melby et al.,1997). Plastic tissue sections and whole-mount in situ hybridized embryos were photographed on an Axioplan microscope (Zeiss) using a digital camera (Dage). Photo images were cropped and assembled using the Photoshop program version 5.5 (Adobe).

To determine the role of hrT in zebrafish development, we designed a morpholino antisense oligonucleotide (Nasevicius and Ekker, 2000;Heasman et al., 2000),hrTMO(1), that was designed to block the translation of hrT(Fig. 1A). In addition, we designed a control morpholino antisense oligonucleotide, Control-MO, which is a modified sequence of hrTMO(1) with four nucleotide changes(Fig. 1A). At 1.5 ng/embryo of hrTMO(1), we consistently obtained morphants that were generally normal except that they had dysmorphic hearts, cardiac edema, no blood circulation (65%,n=100), while injection of the same amount of Control-MO produced a severely attenuated effect (Table 1), demonstrating that these results were not due to an injection artifact. In addition, these phenotypes were not observed after injection of over ten other morpholino oligonucleotides(Lee and Kimelman, 2002) (K. Griffin and D. K., unpublished; S. Chen and D. K., unpublished; W. Clements and D. K., unpublished).

The abnormal cardiac development was visible in hrTMO(1)-injected zebrafish embryos at 24 hours post fertilization (hpf)(Fig. 1B,C). By 48 hpf, the heart tube of uninjected zebrafish embryos developed into clearly separate chambers (Fig. 1F,H), whereas the heart tube in hrT morphants remained in a tubular structure with no obvious morphological distinction between the chambers(Fig. 1G,I). This morphological alteration was also associated with abnormal contractility of the heart (see Movies). The hearts of hrT morphants show slower cardiac rhythm than those of their uninjected siblings.

To confirm that these morphant phenotypes were specifically due to the inactivation of hrT function, we first investigated the efficacy of hrTMO(1) in blocking the translation of the hrT gene. To do this, we constructed a fusion between the coding region of hrT and theGFP reporter gene, hrT-GFP, which includes the binding region of hrTMO(1). Injection of 0.1 ng hrT-GFP RNA in the presence of 1.5 ng hrTMO(1) resulted in the absence of GFP protein expression(Fig. 1J-M; 100%,n=55). By contrast, there was no effect on the expression of GFP in embryos co-injected with 0.1 ng GFP RNA and 1.5 ng hrTMO(1)(Fig. 1N-Q; 90%,n=45). These experiments demonstrate that hrTMO(1) is able to specifically block the translation of the hrT gene via the binding sequence for hrTMO(1). To demonstrate further the morphant phenotypes were the result of specific loss of the hrT gene function, we designed a second morpholino oligonucleotides, hrTMO(2), which binds to the hrTtranscript at a different region than hrTMO(1)(Fig. 1A). At a range of 6.0 to 12.5 ng, the hrTMO(2)-injected zebrafish embryos exhibited the same phenotype as that observed with hrTMO(1) (Fig. 1D,E), and the resulting phenotypes were dose dependent(Table 1). At higher doses,both morpholino oligonucleotides caused nonspecific effects (data not shown). Thus, we conclude that the specific morphant phenotypes produced by hrTMO(1)result from the specific inhibition of HrT.

hrT is expressed throughout all stages of heart formation including cardiac cell fate specification, morphogenesis, looping of the heart tube and chamber formation (Griffin et al., 2000; Ahn et al.,2000). These different key stages have been well characterized with molecular markers, including nkx2.5, tbx5, cardiac myosin light chain 2 and ventricle myosin heavy chain(Lee et al., 1996;Serbedzija et al., 1998;Begemann and Ingham, 2000;Yelon et al., 1999). These markers were used as in situ hybridization probes to assess gene expression changes during cardiac development in embryos injected with hrTMO(1) to determine when defects in hrT morphants occurred.

We first examined the pattern of nkx2.5 expression during early cardiac development, starting from the specification of cardiac progenitors to the formation of a linear heart tube. One of the initial steps in cardiogenesis is the establishment of the heart field in the anterior region of the lateral plate mesoderm. These cardiac precursors are characterized by the bilateral expression of nkx2.5(Lee et al., 1996;Serbedzija et al., 1998). InhrT-morphant embryos, nkx2.5 expression at the 12-somite stage in the anterior lateral plate mesoderm was unaffected by the depletion of hrT, indicating that hrT is not required for establishment of bilateral sets of cardiac precursors(Fig. 2A,B). Similarly, usingnkx2.5 expression to visualize the cardiac primordium, heart development appeared normal in hrT-morphant embryos through the stages of fusion of the bilateral heart fields and leftward jogging of the heart (Fig. 2C,D). We also compared the relative distributions of cardiac-myosin light chain-2(cmlc2), a pan-cardiac marker and ventricular myosin heavy chain (vmhc), to examine the formation of atrial and ventricular precursors (Yelon et al.,1999). At the 17-somite stage and at 24 hpf, expression ofcmlc2 and vmhc in injected embryos was indistinguishable from uninjected embryos, indicating that the specification of ventricular and atrial precursors proceeded normally in hrT-morphant embryos (data not shown). We conclude therefore that the early expression of hrT in cardiac precursors (up to 24 hpf) is not essential for the early differentiation or morphogenesis of the heart.

At 36 hpf, we used cmlc2 to visualize the process of cardiac looping in hrT morphants. cmlc2 is expressed in the atrial and ventricular regions, outlining the development of the whole heart. In uninjected embryos, the process of cardiac looping is marked by a rightward bending in the ventricular region (Fig. 2E). This morphological signature of cardiac looping was not observed in hrT morphants (Fig. 2F). Thus, HrT plays an essential role in cardiac morphogenesis that is not evident until the cardiac looping stage, significantly later than the onset of hrT expression in cardiac precursors.

In addition to the absence of looping, the distribution of cmlc2expression at 36 hpf also revealed a defect in chamber morphology. In uninjected embryos, the dense cmlc2 expression observed in the ventricle contrasts with the relatively diffuse appearance of the atrium(Fig. 2E). In injected embryos,however, atrial expression of cmlc2 appeared similar to ventricular expression (Fig. 2F). This defect could either be due to collapse of the atrium, so that there was an apparent increase in the density of cmlc2 expression there, or because atrial chamber identity was defective. Consistent with the latter possibility, we observed that vmhc expression in hrTmorphants was no longer specific to the ventricle and was now also detected in the atrium (Fig. 2G compare with 2H), suggesting that hrT may play a role in maintaining chamber-specific patterns of gene expression.

At 48 hpf, when morphologically distinct cardiac chambers begin to form in untreated embryos, hrT morphants did not exhibit clearly distinct physical boundaries between the chambers. To examine the chamber identity inhrT morphants, we analyzed cardiac chamber formation using the monoclonal antibodies MF20 to detect a myosin chain common to both chambers(Fig. 2K), and S46 which detects an atrial-specific myosin epitope(Fig. 2M)(Yelon et al., 1999;Evans et al., 1988). Staining of the hrT morphants with MF20 indicated the presence of two distinct chambers (Fig. 2L). This result was confirmed with the atrium-specific antibody S46, which revealed a clear atrioventricular boundary (Fig. 2N). Despite the absence of morphologically distinct cardiac chambers in the hrT morphants, the heart was characteristically divided into an atrium and ventricle at this stage. Thus, the depletion ofhrT function does not prevent the acquisition of anteroposterior fates within the heart tube, although there are clear alterations to gene expression within the atrium.

Our initial study of hrT suggested that it might be genetically upstream of tbx5 as the expression of hrT precedes the expression of tbx5 in the heart field(Griffin et al., 2000). Consistent with this idea, mice lacking the tbx5 gene have normal expression of tbx20, the mouse ortholog of hrT(Bruneau et al., 2001). Becausetbx5 is a key transcription factor that regulates cardiac morphogenesis and gene expression within the heart field(Basson et al., 1999;Horb and Thomsen, 1999;Bruneau et al., 2001), we wished to determine the relationship between HrT and tbx5 expression. In hrT morphants at 33 hpf, we observed a dramatic upregulation oftbx5 expression in the embryonic heart(Fig. 3E,F; 44%,n=34). However, there was no change in tbx5 expression at earlier times (Fig. 3A-D), or in the developing pectoral fin buds where tbx5 is expressed buthrT is not (Fig. 3G,H).

As depletion of HrT increased the expression levels of tbx5 in the heart, we predicted that overexpression of hrT would decrease the expression of tbx5. In order to regulate the expression ofhrT, we constructed a fusion between the coding region ofhrT and the ligand-binding domain of the glucocorticoid receptor (GR-hrT; gr — Zebrafish Information Network). As shown previously, these fusion proteins are inactive until the hormone dexamethasone is added (Kolm and Sive,1995; Tada et al.,1997). For the experiments shown here, dexamethasone was added at the 12-somite stage (Fig. 3I-L), but the same results were observed when the hormone was added at earlier stages (data not shown). We also determined that the same concentration of dexamethasone does not cause any apparent developmental defect in the developing uninjected embryos (data not shown). Induction of GR-HrT had no effect on the morphological appearance of the heart field or on the expression of tbx5 before the stages of cardiac looping (data not shown). However, at 30 hpf, we observed a significant downregulation oftbx5 expression in the embryos after induction of GR-HrT(Fig. 3I,J; 55%,n=40), as well as morphological alterations in the heart at later stages. While the hearts of control embryos underwent the normal process of heart looping, the hearts in GR-HrT-induced embryos did not loop (data not shown). The effects of hrT overexpression on tbx5 were specific to the heart, as the levels of tbx5 expression in the fin buds of both GR-HrT-induced and control embryos were the same(Fig. 3K,L). Both the overexpression and the morpholino antisense experiments demonstrate that HrT acts to regulate the levels of tbx5 expression during the stages of cardiac looping. As normal cardiac morphogenesis and gene expression requires a precise regulation of the levels of tbx5 (reviewed by Hatcher and Beeson, 2001), our results suggest that a major role for HrT is to modulate the levels of tbx5 during cardiac looping.

In addition to the defective heart, hrT morphants do not have circulating blood. To determine the cause of this defect, we examined the process of hematopoiesis and vasculogenesis in embryos depleted ofhrT function. Hematopoiesis occurs in several waves and in distinct locations, giving rise to the terminal differentiation of various hematopoietic progenitor cells, including early macrophages, erythrocytes,megakaryocytes and leucocytes (Zon,1995; Herbomel et al.,1999; Detrich et al.,1995; Thompson et al.,1998). From the onset of gastrulation to 24 hpf in thehrT morphants, we observed no apparent change in the expression of genes that are known to denote the development of different hematopoietic cell lineages (data not shown), indicating that the different hematopoietic cell lineages form in the absence of hrT function. Thus, the absence of circulating blood is not a result of an absence of blood, but is most probably due to a problem with vasculogenesis. This conclusion is supported by the observed blood pooling in the peri-anal region of the hrT morphants at 36 hpf (Fig. 4F). Although the hematopoiesis in the hrT morphants was essentially normal, we did observe an interesting alteration in the pattern of gata1 andgata2 expression. At the 2- to 12-somite stage, gata1expression resides in two stripes flanking the posterior paraxial mesoderm in the developing zebrafish embryos, marking the erythrocyte and megakaryocyte lineages (Detrich et al.,1995). At the 8- to 10-somite stage, we found a `U'-shaped pattern of gata1 expression in hrT morphants, rather than the bilateral stripes of expression in the uninjected sibling embryos, because of ectopic expression of gata1 in the most posterior lateral plate mesoderm (Fig. 4B; 61%,n=81). We observed a similar result for gata2 expression,which marks all of the definitive hematopoietic lineages (data not shown). Interestingly, hrT is expressed in the most posterior region of the lateral plate where the normal bilateral stripes join to form the `U'-shaped pattern in the hrT morphants (Fig. 4C,D), suggesting that HrT normally prevents the hematopoietic fate in this domain.

hrT is expressed in the dorsal aorta(Griffin et al., 2000;Ahn et al., 2000), and thus the blood circulation defects could be due to aberrant formation of the dorsal aorta. To determine if there is a vascular defect when hrT function is impaired, we investigated the formation of the major blood vessels in thehrT morphants. After gastrulation, vasculogenesis begins with the formation of two stripes of endothelial precursors at the lateral edges of the mesoderm, and these cells express fli1, a member of the ETS-domain family of transcription factors (Brown et al., 2000). At 14 hpf, the expression of fli1 in the lateral mesoderm extends along the entire axis during segmentation in two continuous bands, forming a `U'-shaped surrounding the axial and paraxial mesoderm (Fig. 4G). At this stage, we found that the pattern of fli1 expression was normal inhrT morphants when compared with that in the uninjected siblings(Fig. 4G,H). This result suggests that the early specification of endothelial precursors is unaffected when hrT function is disrupted. At approximately 20 hpf, the`U'-shaped pattern of fli1 expression in trunk and tail coalesces in the midline (Fig. 4I). Subsequently, the fli1 expression is found in the walls of major vessels, including the dorsal aorta, axial vein and intersegmental vessels(Fig. 4K). At 20 hpf, the expression of fli1 was detected in a broad distribution pattern around the midline of hrT morphants(Fig. 4J), suggesting that the fusion of endothelial cells to form the dorsal aorta in the midline was disrupted. By 24 hpf, the abnormal expression pattern of fli1 inhrT morphants was even more apparent. In uninjected embryos, thefli1 expression was detected in the dorsal aorta, axial vein and intersegmental vessels (Fig. 4M), whereas in hrT morphants, fli1 was expressed in a single domain running along the midline ventral to the notochord (Fig. 4N; 53%,n=30). In addition, the sprouting of intersegmental vessels was not detected in hrT morphants (Fig. 4N). However, the pattern of fli1 expression in the pharyngeal primordium appeared to be normal(Fig. 4K,L), suggesting that the vasculogenic requirement for hrT is localized to the trunk of the embryos.

In sections of fli1-stained, uninjected embryos, the dorsal aorta is seen ventral to the notochord and the axial vein is below the aorta(Fig. 4O). In the hrTmorphants, the fli1-expressing cells have not organized into two clear vessels, although we typically saw one lumen above the gut tube(Fig. 4P). As hrT is expressed in the dorsal aorta but not the axial vein, we expect that the major defect is due to the formation of this vessel. Thus, we conclude that the failure of hrT morphants to circulate blood is primarily due to a defect in the formation of the dorsal aorta.

We noticed a strong resemblance between the trunk vascular defects in thehrT morphants and floating head (flh) mutant embryos (Sumoy et al., 1997;Brown et al., 2000;Fouquet et al., 1997).flh is a homeodomain transcription factor expressed in the notochord precursors, and the flh mutation causes an absence of notochord(Talbot et al., 1995). Inflh mutants as in the hrT morphants, blood circulation does not occur, and the blood accumulates in the peri-anal region because of a failure to form the dorsal aorta (Fouquet et al., 1997; Brown et al.,2000; Sumoy et al.,1997). Similarly in the hrT morphants and flhmutants, fli1 is expressed in a single stripe in the midline with no apparent intersegmental vessels.

The similarity between the vascular defects in flh mutant embryos and hrT morphants, prompted us to examine hrT expression inflh embryos. In flh homozygotes, which were identified morphologically and confirmed by the absence of flh expression(Talbot et al., 1995), we observed a complete absence of hrT expression in the vascular progenitors whereas other domains of hrT expression were unaffected(Fig. 5). This is consistent with the possibility that hrT expression in vascular progenitors depends, directly or indirectly, upon a signal from the midline mesoderm

Using two non-overlapping hrT-specific morpholino antisense oligonucleotides, we have demonstrated that hrT is required for normal heart development. Prior to the onset of cardiac looping, cardiogenesis progressed normally in hrT morphants, indicating that there is no requirement for hrT function in the early stages of heart development. We first began to see the effect of depleting HrT function around 33 hpf. In hrT morphants, the embryonic heart showed no ventricle bending, a characteristic of normal cardiac looping. In addition, the expression of two marker genes, cmlc2 and vmhc was aberrant,with abnormal expression of these genes in the atrium. Although the dysmorphic hearts in the hrT morphants appeared to form two distinct chambers,the heart remained as a linear tube and did not form the characteristically looped structure, resulting in a heart with abnormal contractions and a slower rhythm.

Although hrT is expressed very early in the heart field, depletion of its function did not have an apparent effect until 33 hpf. It is possible that this is due to incomplete inhibition of hrT function using the antisense oligonucleotides or due to the presence of a gene that compensates at earlier times for a loss of hrT. However, it is intriguing that a similar effect has been observed with murine tbx5; whiletbx5 is also expressed very early in the heart field(Liberatore et al., 2000),elimination of the tbx5 gene did not have an effect until later times of development (Brunneau et al., 2001). This finding suggests that the T-box genes may not function by themselves in the regulation of heart development,but may need to interact with other factors that are expressed at later times of development.

A key finding in this study is that HrT acts as a regulator oftbx5 expression. Depletion of HrT function causes the cardiac expression of tbx5 to be upregulated, whereas overexpression ofhrT leads to the downregulation of tbx5 expression in the developing heart. Importantly, in both cases, the regulation of tbx5was specific to the heart and did not affect the expression of tbx5in the limb buds. Although we do not yet know whether or not Tbx5 regulateshrT expression in zebrafish, mice lacking the tbx5 gene do not show alterations in the expression of the murine hrT orthologtbx20 (Brunneau et al., 2001), indicating that hrT may function solely upstream of tbx5.

tbx5 has emerged as a key gene regulating heart development from amphibians to mammals (Horb and Thomsen,1999; Liberatore et al.,2000; Bruneau et al.,2001). Patients with Holt-Oram syndrome, a disorder caused by haploinsufficiency of the tbx5 gene, have a high penetrance of cardiac defects (Li et al.,1997; Basson et al.,1999). Similar defects have been found in mice lacking one copy of the tbx5 gene (Bruneau et al.,2001). Mice lacking both copies of tbx5 have more severe defects, including an absence of heart looping and alterations in a subset of cardiac-specific genes, although the early steps in cardiac development appear normal (Bruneau et al., 2001). Curiously, mice overexpressing tbx5 also have cardiac looping defects as well as alterations in the expression of some cardiac genes(Liberatore et al., 2000). These results indicate that the gene dose of tbx5 is crucial for normal cardiac morphogenesis (Bruneau et al., 2001; Hatcher and Basson,2001). These studies fit very well with our observations that inhibition of hrT function or overexpression of hrT, which cause an increase or decrease of tbx5 expression, respectively, leads to defects in cardiac morphogenesis. Importantly, zebrafish tbx5mutants also show a complete absence of heart looping (M. Fishman, personal communication). It will be of great interest to determine if the alterations we observed in cardiac development are due solely to changes in the level oftbx5 expression, or whether hrT has additionaltbx5-independent functions in the heart.

Another key aspect of hrT depletion is the absence of circulating blood. By examining the expression of a variety of different blood cell markers at various stages in hrT morphants, we find that there is no observable defect in hematopoiesis. In analyzing vasculogenesis, however, we have observed abnormal trunk vascular development in the hrTmorphants. In hrT morphants, the endothelial marker fli1 is expressed normally at 14 hpf when endothelial cells form a `U'-shape around the axial and paraxial mesoderm, demonstrating that hrT plays no significant role in the specification of endothelial cells. At 24 hpf, the trunk vascular defect is evident by the altered expression of fli1,as well as the absence of normal trunk vessel formation. In addition, no intersegmental vessels sprout from the domain of the aberrant fli1expression. As hrT is expressed only in the dorsal aorta(Griffin et al., 2000;Ahn et al., 2000), we suggest that hrT is required specifically in the dorsal aorta rather than the axial vein, although we do not rule out the possibility of indirect effects on axial vein formation due to a defective dorsal aorta. As the endothelial cells come close to the midline in hrT morphants but do not completely fuse to give the dorsal aorta, our results suggest that hrT plays a critical role in assembling the endothelial cells into a vessel.

We noticed a striking similarity between the hrT morphants andflh mutants with regard to the formation of the vascular system, and we found that hrT expression is specifically eliminated in the dorsal aorta of flh mutants, while hrT expression in other regions is unaffected. As the major defect in flh mutants is an absence of notochord, it has been suggested that formation of the endothelial cells into the dorsal aorta requires a notochord-derived signal such as sonic hedgehog (shh) (Brown et al.,2000; Fouquet et al.,1997; Sumoy et al.,1997). Our results suggest that HrT is an essential component of the response to this signal, and indeed we have observed that hrTexpression is abolished in the hedgehog signaling mutant you-too (D. P. S., K. J. P. G. and D. K., unpublished). As the process of vessel formation is still not well understood, it will be very interesting to determine the targets of HrT in the dorsal aorta.

We observed a very interesting alteration in the expression pattern ofgata1 and gata2 at the 8- to 10-somite stage in hrTmorphants. While gata1 and gata2 are normally expressed in bilateral stripes, in hrT morphants, gata1 andgata2 form a `U'-shaped pattern because of the presence of an extra posterior domain of expression that connects the two bilateral stripes. This extra domain of expression coincides with a region of the embryo that expresses hrT (Griffin et al.,2000; Ahn et al.,2000), indicating that one function of hrT is to prevent the expression of gata1 and gata2 in the most posterior lateral plate of the developing embryo. This ectopic expression ofgata1 and gata2, as well as the increased cardiac expression of tbx5 and vmhc, suggests that an important function of HrT is to negatively regulate gene expression, at least indirectly.

This study has demonstrated a clear role for hrT in zebrafish cardiovascular development. Mutations in the human hrT ortholog,TBX20, are candidates for producing congenital cardiovascular defects that are among the most prevalent defects affecting live births. Since several human T-box genes have been shown to cause birth defects when haploinsufficient (Basson et al.,1999; Li et al.,1997; Bamshad et al.,1997; Merscher et al.,2001), it will be worthwhile examining the TBX20 gene in humans with congenital heart or vascular defects if a disease allele maps to this genetic interval.

We thank Chris Bjornson and Wilson Clements for reading this manuscript;Debbie Yelon, Bernard Thisse and Adam Rodaway for probes; Stephen Hauschka and Frank Stockdale for antibodies; Jeff Miller for the CS2-GFP plasmid DNA; and Dave Raible for help and advice during this project. D. S. was supported by a vascular training grant (T32 07312) and a NRSA fellowship (F32 HD08725). This work was supported by grant 0078303 from the National Science Foundation to D. K.