Tbx2 and Tbx3 Regulate the Dynamics of Cell Proliferation during Heart Remodeling

Background The heart forms from a linear tube that is subject to complex remodeling during embryonic development. Hallmarks of this remodeling are the looping of the heart tube and the regionalization into chamber and non-chamber myocardium. Cardiomyocytes in the future chamber myocardium acquire different cellular and physiological characteristics through activation of a chamber-specific genetic program, which is in part mediated by T-box genes. Methodology/Principal Finding We characterize two new zebrafish T-box transcription factors, tbx3b and tbx2a, and analyze their role during the development of the atrioventricular canal. Loss- and gain-of-function analyses demonstrate that tbx3b and tbx2a are necessary to repress the chamber-genetic program in the non-chamber myocardium. We also show that tbx3b and tbx2a are required to control cell proliferation in the atrioventricular canal and that misregulation of cell proliferation in the heart tube influences looping. Furthermore, we characterize the heart phenotype of a novel Tbx3 mutation in mice and show that both the control of cell proliferation and the repression of chamber-specific genetic program in the non-chamber myocardium are conserved roles of Tbx3 in this species. Conclusions/Significance Taken together, our results uncover an evolutionarily conserved role of Tbx2/3 transcription factors during remodeling of the heart myocardium and highlight the importance of controlling cell proliferation as a driving force of morphogenesis.


INTRODUCTION
The T-box (Tbx) family of transcription factors is represented in all metazoans [1] and is characterized by a highly conserved DNA binding domain, the T-box domain. Several Tbx genes are expressed during embryonic development in specific domains and have been implicated in the formation of a variety of organs, including the heart. The existence of several human congenital syndromes caused by disruption of human TBX genes, demonstrates their importance in morphogenesis. The Holt-Oram syndrome is caused by TBX5 haploinsufficiency and is characterized by heart and hand anomalies [2]. Mutations in TBX1 have been implicated in the Velocardiofacial or DiGeorge syndrome in which congenital heart defects are present [3]. Several Tbx genes are expressed in the heart in discrete and overlapping regions and have been shown to be important for the development of such complex organ.
The embryonic linear heart tube undergoes complex remodeling, including looping and regionalization, to form the final vertebrate multichambered heart. Initially, the primary myocardium is morphologically homogenous throughout the extent of the linear heart tube. Concomitantly with the beginning of looping, the heart tube becomes regionalized into chamber myocardium, the atria and the ventricles, and non-chamber myocardium, the outflow tract (OFT), the atrioventricular canal (AVC), the inflow tract (IFT) and the inner curvatures. The chamber myocardium becomes fast conducting myocardium whereas the non-chamber myocardium retains slow conduction, thereby participating in the establishment of synchronized beating in the embryonic heart [4]. Additionally, the myocardium of the future chambers starts to grow out of the primitive heart tube during looping. Consequently, the non-chamber myocardium forms constrictions in the heart tube, such as the AVC, that prevent backflow of blood before valve formation. A variety of chamber and non-chamber specific genes have been characterized. The expression of chamber markers, such as atrium natriuretic factor (Anf), connexin 40 and 43 (Cx40, Cx43), Chisel and bone morphogenetic protein 10 (Bmp10), comes on concomitantly with the start of chamber growth in the mouse heart tube [5,6]. The myocardium fated to become non-chamber myocardium expresses a different set of molecular markers. Bmp2 and Tgfb2 are expressed in the OFT, AVC and IFT since early looping stages and are important for further development of the nonchamber myocardium [7,8]. Thus, differentiation of the primary myocardium and looping are the first steps of heart remodeling.
Tbx5 plays a role during chamber formation despite being expressed throughout the extent of the heart tube. It has been shown that Tbx5 synergistically interacts with Nkx2.5 in the activation of the ANF regulatory element, responsible for expression of Anf in the cardiac chambers [9]. A similar regulatory element, composed of TBE and NKE binding sites, has been found in the promoters of other chamber-specific genes, such as Cx40 and Cx43 [10]. The lack of expression of chamber specific genes in non-chamber myocardium has been attributed to the expression of Tbx2 which is mutually exclusive with that of chamber myocardium markers [11]. Tbx2 cooperates with Nkx2.5 and represses transcription from ANF regulatory element, serving as inhibitors of the chamber-specific genetic program. Supporting this model, misexpression of Tbx2 under the control of b-MHC promoter results in the lack of chamber growth and downregulation of Anf, Chisel and Cx40 [12]. However, mice with targeted mutations in Anf, Chisel or Cx40 have normal heart development and reach adulthood [13][14][15][16], suggesting that the absence of their expression in b-MHC::Tbx2 mice does not account for the lack of chamber growth. Recently, it has been shown that Tbx2 represses transcription of the cell cycle gene Nmyc1. The ectopic expression of Tbx2 throughout the heart tube has been suggested to cause the hypoplastic phenotype observed in Tbx20 mutants [17][18][19]. The chamber-specific genetic program is not initiated in Tbx20 mutants, suggesting that downregulation of Tbx2 in the chamber myocardium is crucial for normal heart development. Although Nmyc1 mutant hearts are hypoplastic [20,21], the cardiac chambers are formed and the overall phenotype is much less severe than that of Tbx20 mutant or b-MHC::Tbx2 hearts, suggesting that regulation of additional downstream targets of Tbx2 may be involved in chamber and non-chamber myocardium differentiation. Moreover, ablation of Tbx2 causes defects in the AVC and a failure in OFT septation in only a quarter of the homozygous embryos [22], indicating that other factors are able to compensate for the lack of Tbx2.
Tbx3 is co-expressed with Tbx2 in the heart in looping stages and biochemical studies suggest Tbx3, like Tbx2, is capable of repressing transcription from the ANF element in chamber specific genes [10]. In humans, haploinsufficiency of TBX3 causes the Ulnar Mammary syndrome (UMS), a pleiotropic disorder that typically presents defects in limb, mammary gland, tooth, hair and apocrine gland development [23]. This autosomal dominant disorder is fully penetrant, however there is a large variability in the clinical presentation among affected individuals and cardiac anomalies have recently been reported [24,25]. Mice lacking Tbx3 are not viable and display a wide range of onset of lethality, between E10.5 and E16.5, which correlates with the variability of clinical presentation in UMS patients [26]. Although Tbx3 is expressed in restricted domains in the mouse and human hearts from early stages [10], heart formation has not been specifically addressed in Tbx3 mutant embryos.
The processes of heart looping and chamber formation are less well studied in other vertebrates. In zebrafish, it is known that the initial broad expression of bmp4 and versican become restricted to the myocardium at the level of the AVC during initiation of looping [27,28]. As for chamber markers, zebrafish anf is expressed in chamber myocardium after looping is well under way [29]. However, the mechanism whereby heart remodeling occurs in zebrafish remains to be elucidated.
The simplicity and accessibility of the two-chambered heart in zebrafish allow a deeper analysis of the complex morphogenetic events involved in heart development. In this report, we show that novel Tbx2 and Tbx3 zebrafish homologs prevent initiation of the chamber-specific genetic program and growth in non-chamber myocardium. We also analyze cardiac defects in mice Tbx3 mutant embryos and show that Tbx3 plays a role in heart remodeling. Together our findings reveal the importance of a tight equilibrium between cell proliferation and differentiation in order to achieve proper organogenesis.

RESULTS
Zebrafish tbx3b and tbx2a are required for heart looping and AVC formation We isolated two novel Tbx genes in zebrafish, which belong to the Tbx2 subfamily. Based on sequence similarity and expression patterns we named them tbx3b and tbx2a (Fig. S1). tbx3b encodes a putative protein of 694 amino acid residues with overall similarity to human TBX3 of 67.8% and to mouse Tbx3 protein of 67.3%. tbx2a encodes a putative protein of 687 amino acid residues with 61.8% similarity to mouse Tbx2, 63.1% to human TBX2 and 78.2% to zebrafish tbx2b. During heart development, tbx3b is first expressed throughout the extent of the heart tube and becomes restricted to the AVC and OFT at 33 hpf (Fig. 1A,B). Zebrafish tbx2a is first weakly expressed throughout the extent of the heart tube becoming later restricted to the AVC (Fig. 1C,D). Thus, the domain of expression of tbx3b and tbx2a in the heart is similar to that of the mouse counterparts [10,22].
The expression pattern of tbx3b and tbx2a suggested that they might play a role in heart regionalization in zebrafish. Morpholino oligonucleotides (MO) were used to knock down tbx3b and tbx2a in a transgenic line expressing GFP in the heart, mlc2a::GFP [30]. The efficacy of the MOs to inhibit translation of tbx3b and tbx2a mRNA were confirmed by co-injecting tbx3b:EGFP or tbx2a:EGFP fusion constructs and examining fluorescent signal (Fig. S2). Injection of tbx3b MO resulted in a growth delay from 36 hpf onwards, absence or reduction of fin buds and misformation of the otic vesicle and lethality past 3 dpf (I.R. and J.C.I.B., unpublished observations). In control embryos at 48 hpf the AVC was a tight constriction between the ventricle and the atrium and these two chambers were at the same anterior-posterior level (Fig. 1E). In stage-matched tbx3b morphant embryos, the atrium-ventricle border was wide and looping was delayed (Fig. 1F). By 3 dpf, the pericardiac cavity was severely swollen, looping was incomplete and the presumptive AVC remained wide (Fig. 1J). Loss of function of tbx2a resulted in similar heart defects, with incomplete looping and lack of AVC constriction (Fig. 1G,L). However, in contrast to tbx3b morphants, the fin buds and the otic placodes developed normally in tbx2a morphants and there was no growth delay (I.R. and J.C.I.B., unpublished observations), suggesting that these genes play specific and non-overlapping roles in the development of other organs. Upon co-injection of both tbx3 and tbx2a MOs we observed that fin bud was absent and the otic placode was reduced (Fig. S2). In the heart of double morphants, the AVC failed to form, and there was no indication of looping (Fig. 1H,M). Although development of other organs was not more affected in double morphants than in tbx3b MO-injected embryos, the heart phenotype was more severe when both tbx3b and tbx2a were knocked down, implying that tbx3b and tbx2a play non-redundant roles in cardiac development.
To test if the absence of the AVC constriction and looping were convoyed by defects in the heart regionalization, we assessed the expression of AVC markers in zebrafish. The expression domains of bmp4 and bmp2b in OFT, AVC and IFT of the heart were not altered in single and double morphant embryos ( Fig. 2A-D, data not shown). The expression domain of notch1b, which normally becomes restricted to the endocardium at the level of the AVC and OFT at 45 hpf [31], was expanded into the ventricle endocardium of tbx3b and tbx2a morphant embryos at 48 hpf (Fig. 2E,F,G), as well as in stage-matched double morphant embryos (Fig. 2H). Zebrafish anf is first expressed throughout the primitive heart tube and later becomes expressed exclusively in the chambers (our unpublished observations) [29]. In tbx3b, tbx2a, and double morphants, however, anf transcripts continued to be expressed throughout the extent of the cardiac myocardium at 48 hpf (data not shown). Zebrafish bmp10 is expressed in the heart in a dynamic pattern reminiscent of zebrafish anf as well as in the otic vesicle and ventricle of the brain (Fig. S3). In tbx3b and tbx2a morphant embryos, bmp10 was downregulated from the AVC as in control embryos ( Fig. 2I-L). However, in double morphants bmp10 was present throughout the extent of the heart tube ( Fig. 2M). Together these results show that the initial specification of chamber and non-chamber myocardium occurs in single and double morphants, but further development is impaired.

Ectopic presence of Tbx3 and Tbx2 results in the absence of chamber growth
We next asked if the ectopic expression of Tbx3 and Tbx2 in chamber myocardium was able to prevent chamber formation.
Tbx3-injected embryos developed at comparable rates as the GFP-injected embryos. At 60 hpf, 44.5% of the embryos overexpressing Tbx3 presented a swollen pericardiac cavity and a delay in heart looping. Most notably, however, was the lack of chamber growth, which, added to the increasing pericardial swelling due to low circulation, gave rise to a thin heart tube that stretched from the base of the head to the yolk ( Fig. 3B; compare to 3A), a condition previously described as pipe-like heart [27]. In 60% Tbx2-injected embryos the pericardiac cavity was swollen from 48 hpf onwards (Fig. 3C). A similar pipe-like heart phenotype was   44.5% of Tbx3-injected embryos and 60% of Tbx2-injected embryos present a pipe-like heart that lacks chamber growth and looping. (D, E, F) Zebrafish bmp10 is strongly expressed in the chamber myocardium in wild type embryos, but is significantly downregulated from the heart in Tbx3-or Tbx2-injected embryos. a, atrium; v, ventricle. doi:10.1371/journal.pone.0000398.g003 visible at 3 dpf. In line with the lack of chamber growth, zebrafish bmp10 was significantly downregulated in the heart in Tbx3-and Tbx2-injected embryos (Fig. 3D,E,F). Transcripts of anf were not downregulated from the AVC of Tbx3-and Tbx2-overexpressing hearts at 48 hpf and presented levels of expression similar to the wild type 24 hpf heart (data not shown). The overexpression of Tbx3 or Tbx2 did not affect the expression of notch1b or bmp4 (data not shown). Thus, the specification and differentiation of nonchamber myocardium is not altered in the presence of ectopic Tbx3 or Tbx2 in the heart tube. However, the presumptive chambers do not grow from the primitive heart tube, and maintain levels of expression of chamber markers characteristic of primary myocardium.
tbx3b and tbx2a are required to establish the pattern of proliferation during heart looping The absence of chamber growth in Tbx3-and Tbx2-injected embryos prompted us to investigate the cell division and apoptosis levels during heart looping. In wild type embryos between 31 and 48 hpf, no significant numbers of apoptotic cells, as evaluated by acridine orange staining, were observed in the heart (data not shown). Additionally, no differences were observed between Tbx3or Tbx2-injected and wild type embryos in the number of apoptotic cells in the heart tube (data not shown). To study the pattern of proliferation in the looping heart we gave a one-hour pulse of BrdU by injecting a BrdU solution in the pericardiac cavity. In 31 hpf wild type embryos the heart presented no signs of looping and chamber formation. The BrdU-positive cells were equally distributed throughout the extent of the heart tube ( Fig. 4A). At 33 hpf, the anterior region of the heart tube, the future ventricle, has jogged to the right, allowing a rough distinction between the two future chambers and the future AVC. At this stage, the BrdU positive cells were more concentrated in the future chamber regions, leaving the AVC region devoid of BrdU-positive cells (Fig. 4B). At 36 hpf the initial phase of heart looping is already completed and it is possible to clearly identify the ventricle and the atrium, as well as the OFT and the AVC. Thus, a dynamic pattern of cell proliferation along the heart tube accompanies chamber outgrowth, and a crucial differentiation step occurs at 33 hpf, concomitant with the rightward jogging of the future ventricle.
The pre-looping heart tube of Tbx3-injected embryos appears normal in size and morphology (data not shown). However, at 33 hpf there were no signs of looping and the dividing cells were homogenously distributed throughout the heart tube, with no obvious separation between ventricle and atrium (Fig. 4C). Moreover, the total number of BrdU-positive cells in the heart was significantly lower when compared to wild type embryos (Fig. 4H). Similar altered numbers and distribution of BrdUpositive cells were observed in the hearts in Tbx2-injected embryos at 33 hpf (Fig. 4D,H). These results indicate that the lack of chamber outgrowth in the presence of ectopic Tbx3 and Tbx2 is due to a decrease in the cell proliferation levels in the future chamber myocardium. In 33 hpf stage-matched tbx3b or tbx2a morphants, the distribution of dividing cells was not regionalized as in control hearts (Fig. 4E,F,G). The 33 hpf heart tube of double morphant embryos displayed an homogenous distribution of BrdU-positive cells and the total number of dividing cells was slightly higher than in wild type (Fig. 4H). Differences in the number of BrdU-positive cells in the AVC myocardium may be difficult to detect due to the small number of cells in this region. Together these results demonstrate that tbx3b and tbx2a are crucial for the establishment of the dynamic pattern of cell proliferation in the heart tube and further show that perturbation of this pattern may have drastic consequences for heart looping.

Ablation of Tbx3 in mice causes cardiovascular defects
In order to investigate if the role of Tbx3 in zebrafish heart morphogenesis is conserved in mammals, we analyzed the cardiovascular development in mice embryos lacking Tbx3 function. Mice with a targeted mutation for Tbx3 were generated (Materials and Methods, Fig. S4). Homozygous Tbx3 null embryos presented poorly developed yolk sac vasculature, truncated hindlimbs and were smaller than their littermates, similar to what has been described for another Tbx3 mutant [26].
In Tbx3 mutant embryos, we observed a large phenotypic variability in the time of onset and degree of the heart defects. In 4/40 homozygous embryos collected 9.5 days post coitum (E9.5), a delay in heart looping was evident (Fig. 5A,B). In 33% of the Tbx3 mutant embryos collected at E10.5 (n = 21), the pericardiac cavity was swollen and the heart incompletely looped (Fig. 5C,D). The defects in looping ranged from a mild delay to a completely unbent tube (Fig. 5D). The constriction at the AVC was also enlarged when compared to wild type or heterozygous littermates. At later stages, all Tbx3 mutant embryos were readily distinguishable from their littermates in that they were smaller and presented a reduced hindlimb bud (our unpublished observations). 43% Tbx3 mutant embryos recovered at E11.5 (n = 7) presented a swollen pericardial cavity with an unlooped heart (Fig. 5E,F). These observations suggest that there is a great phenotypic variation in the severity of the cardiovascular defects in Tbx3 mutant embryos. Nevertheless, the most severe mutants display cardiac anomalies comparable to the heart defects seen in tbx3b morphant embryos.
To test if the AVC is correctly specified in Tbx3-mutant embryos, we analyzed the expression pattern of Bmp2, which is expressed in the AVC and is crucial for the formation of atrioventricular valves in the mouse [8]. Bmp2 transcripts were present in the AVC of Tbx3 mutant embryos at E9.5 (Fig. 5I,J), suggesting that this non-chamber region is properly specified. To test if further development of the AVC was occurring normally, we analyzed TGFb2 and Smad6 which are expressed in the OFT and the AVC at the level of the myocardium and endocardium, respectively, and are required for cushion mesenchyme formation [32,33]. In Tbx3 null mutant embryos, both TGFb2 and Smad6 were downregulated in the nonchamber myocardium (Fig. 5L-O). Expression of Msx2, another marker of AVC myocardium, was downregulated in Tbx3 mutant embryos at E9.5 (data not shown). Thus, the early specification of the AVC seems to occur normally, however further differentiation of this myocardium is perturbed in Tbx3 null mutants.
Next we analyzed the expression of chamber myocardium markers, Anf and Bmp10. In the most severe class of Tbx3-null mutants, Anf was upregulated in the region between the ventricle and the atrium (data not shown). Bmp10 is normally expressed in the chambers at E9.5 marking the future trabeculating myocardium [6]. In Tbx3-null homozygous embryos Bmp10 transcripts were found ectopically in the OFT and the AVC at E9.5 (Fig. 5G,H), suggesting that in the absence of Tbx3, Bmp10 is activated in non-chamber myocardium. We found NKE-and TBE-sites in the promoter region of human BMP10 (Fig. 5P), which suggest that Bmp10 is subject to similar transcriptional regulation as Anf and Cx40 [4]. Tbx5 is required for chamber formation [34] and was normally expressed in Tbx3 mutant embryos (data not shown). The analysis of the expression patterns of chamber and non-chamber markers in Tbx3 mutant embryos advocates that the initial specification of non-chamber myocardi-um occurs normally in these embryos. In spite of this, further development of the AVC is not achieved and instead, chamber markers are upregulated. Moreover, the most severe class of Tbx3null mutant embryos displays heart defects that are similar to zebrafish tbx3b morphants with respect to morphology and the degree of development, indicating that the mechanisms of AVC formation are conserved among vertebrates.
Previous studies show that the sinoatrial region, the AVC and, to a lesser extent, the OFT present lower levels of cell proliferation compared to the chamber myocardium [35,36]. Since the AVC constriction is unusually wide in Tbx3-null mutants and there is ectopic activation of the chamber genetic program, we analyzed whether the pattern of cell proliferation was altered in these mutants. Consistent with previous findings [35], wild type E9.5 embryos labeled with a one-hour pulse of BrdU present a lower density of BrdU-positive cells in the OFT and the AVC when compared to the ventricle or the common atrium (Fig. 6A, data not shown). Tbx3 mutant embryos subjected to the same treatment displayed densities of proliferating cells in the OFT similar to wild type embryos. However, the density of proliferating cells at the level of the AVC myocardium was comparable to that of the ventricle myocardium, rather than to that of the wild type AVC myocardium (Fig. 6A,B and data not shown). In addition to the analysis of BrdU-labeled cells, we also examined the pattern of mitotic cells in the looping heart by detection of phospho-Histone3-positive cells. Exhaustive counting of the phospho-Histone3-positive cells in each section revealed that proliferation levels were not altered in the OFT, the ventricles or the common atrium at E9.5 in Tbx3 mutant embryos (n = 2/3) relative to wild type counterparts (n = 3/4; data not shown). In spite of this, the number of dividing cells was significantly increased in the AVC of Tbx3 mutant embryos (n = 3/3) when compared to wild type embryos (4/4; Fig. 6C). These results suggest that Tbx3 is an important regulator of the rate of proliferation in the AVC and reveal that the role of Tbx3 is conserved in zebrafish and mouse heart development.

DISCUSSION
In this report, we have studied the role of two novel zebrafish genes, tbx3b and tbx2a, in the process of heart remodeling and the associated differential cell proliferation. Incomplete looping was observed both after gain-and loss-of-function experiments. The pattern of cell proliferation along the heart tube was perturbed in either case, suggesting that differential rates of cell proliferation are important for proper looping to occur. In the absence of tbx3b and tbx2a, the broadening of the AVC constriction may hamper heart tube bending by abolishing the flexion point. On the other hand, the chamber growth out of the primary heart tube may be a permissive factor for the bending of the heart tube along with other intrinsic and extrinsic biomechanical driving forces [37], as evidenced by the absence of looping in the gain-of-function experiments. Supporting this hypothesis, heartstrings mutant and hrT morphants display an absence of chamber formation and impaired heart looping [27,38]. Perturbation of heart looping in fog1 morphants does not correlate with lack of chamber growth [39], suggesting that chamber growth is not sufficient for proper looping to occur. In mice, lack of Tbx5 and Nkx2.5 results in an hypoplastic heart and absence of looping [11,34], indicating that lack of looping in the absence of chamber growth is conserved in mice. Tbx3 and Tbx2 may also be involved in other aspects of the process of heart looping not related to chamber growth. It has been previously shown that Tbx genes can regulate extracellular matrix composition and cell adhesion [40], which are important factors in the bending of the heart tube [41].
Frequently during embryonic development, two adjacent tissues present disparate rates of cell division and the maintenance of these discrepancies is important for proper morphogenesis to occur. In the heart, the rate of cell proliferation is higher in chamber myocardium compared to the non-chamber myocardium in mouse [35,36] and there is a correlation between absence of chamber formation and lower levels of cell proliferation [42,43]. Analysis of the pattern of cell proliferation along the extent of the zebrafish heart tube revealed that cell proliferation is dynamic throughout heart development, increasing upon initiation of chamber formation (this study). We show that tbx3b and tbx2a are necessary to break the uniform distribution of dividing cells in the zebrafish heart by repressing cell proliferation. T-box transcription factors have been previously implicated in the control of cell proliferation in several developmental contexts [19,44,45] and in one case, this control has been shown to be direct: Tbx2 downregulates expression of Nmyc1 in the context of heart development [43]. Thus, one possible mechanism for cell proliferation control in the non-chamber myocardium by tbx2a and tbx3b is by regulation of expression of cell cycle genes.
Concomitant with differential increase in cell proliferation in the heart tube is the activation of a chamber-specific genetic program in chamber myocardium. tbx2a and tbx3b repress the transcription of bmp10 in the AVC. Mouse Bmp10 has been shown to be required for chamber growth and maturation [42], indicating that Bmp10 is a chamber-specific gene that is actually required for chamber growth and maturation. In Bmp10-null hearts, the cell cycle inhibitor p57 (kip2) is upregulated and the rate of cell proliferation is lower than in wild type hearts [42]. The presence of putative NKE-and TBE-binding sites in tandem in upstream sequences of human BMP10 suggest that regulation of this chamber marker could be similar to the that of Anf, Cx40 and Cx43 [4]. Thus, regulation of chamber versus non-chamber genetic programs may be achieved by Bmp10, which we show is altered in the absence of tbx2a and tbx3b in zebrafish and Tbx3 in mouse. Downregulation of TGFb2 in the AVC myocardium was observed upon ablation of Tbx3 in the mouse. It has been suggested that defects in cardiac looping are at the origin of cardiac defects of TGFb2 mutant embryos, which display an overriding tricuspid valve and double outlet right ventricle [7]. Furthermore, TGFb2 can trigger cell cycle arrest by induction of p21 expression in other contexts [46]. In the converse experiment, cardiac-restricted expression of a constitutively active form of the type I TGFb receptor Alk5 results in defective heart looping, hypocellularity and upregulation of p21 [47]. These studies suggest that TGFb2 must play role in heart remodeling and TGFb2 downregulation in Tbx3-null hearts may explain the increase in cell proliferation and defect in heart looping observed.
Loss-of-function of tbx2a and tbx3b each independently resulted in malformation of AVC constriction due to altered cell proliferation rates and incomplete looping. Ablation of Tbx2 or Tbx3 independently in mouse also result in AVC malformation [22] and, in the case of Tbx3, in incomplete looping and altered cell proliferation rates in the AVC. These observations imply that the roles of Tbx2 and Tbx3 are conserved in both vertebrates. Upon knock down of both tbx2a and tbx3b we observed absence of AVC constriction and looping altogether and the defects in cell proliferation were more severe, suggesting that tbx2a and tbx3b play additive roles in heart morphogenesis. As repressors of transcription, tbx2 and tbx3 may simply have different transcriptional targets in the nucleus. Low similarity of protein sequences outside the T-box domain [48] may signify association with different cofactors, supporting this hypothesis. The identification of transcriptional targets and binding partners of Tbx2 and Tbx3 would clarify this issue.
Cardiac looping involves a complex set of movements of discrete regions of the heart tube that are correlated with the regionalization of the heart, i.e., the appearance of molecular and cellular characteristics specific for each heart region. A differential rate of cell proliferation along the heart tube results in growth of cardiac chambers out of the primary myocardium thereby creating the AVC constriction. Tbx2 and Tbx3 repress the induction of chamber-specific genes and cell proliferation in the AVC. The AVC constriction may act as a flexion point that facilitates the bending of the heart tube. On the other hand, defects in cardiac looping are at the origin of malformations in the valve s due to a failure in approaching the correct heart regions [7]. Thus, looping and regionalization are deeply interrelated processes that depend on each other to occur successfully. Our studies uncover a crucial role of Tbx3/2 transcription factors in the regulation of these processes and provide an entry-point for understanding the cellular bases that underlie heart tube looping and regionalization.

Zebrafish lines and Microinjection
Wild type zebrafish (AB) and the transgenic mlc2a::GFP line [30] were used in this study. pCS2+ vectors carrying cDNA inserts coding for mouse Tbx3, human TBX2 and green fluorescent protein (GFP) were synthesized using the SP6 mMessage Machine System according to the manufacturer's instructions (Ambion). Capped mRNAs and morpholino oligonucleotides (MO) were injected into one-cell stage embryos as described [49]. MOs were designed to specifically inhibit RNA translation of the targeted gene and were obtained from GeneTools. The sequences were as follows: Control MO: 59-CCTCTTACCTCAGTTACAATTTATA-39 [50], tbx3b MO: 59-TGGATCTCTCATCGGGAAGTCCAG-39, tbx2a MO: 59-ATCGGTGCATCCAAAAAGCCAGAT-39. Injection of the same amount of control MO did not produce any detectable defects.
Cloning of zebrafish tbx3b, tbx2a, and bmp10 genes A zebrafish embryonic cDNA library was screened with low stringency hybridization using the T-box domain of zebrafish tbx5 as a probe. 59RACE-PCR was performed to obtain the open reading frame sequence for tbx3b and tbx2a. The sequences were submitted to NCBI molecular database. The zebrafish cDNA ensembl database (Zv5) was blasted with the cDNA sequence from mouse and human Bmp10. We obtained one cDNA sequence with significant similarity, which was amplified by RT-PCR. By comparing the expression pattern of this gene with the one from mouse Bmp10 we concluded that it is zebrafish bmp10.

Whole Mount in situ Hybridization
Whole-mount in situ hybridization for zebrafish embryos was performed as described [49]. The zebrafish antisense RNA probes for nkx2.5, anf, bmp2a, bmp2b and has2 were obtained by RT-PCR of a cDNA fragment including the 39 UTR, bmp4, tbx5 and versican were previously described [49,51] and hrT was kindly provided by D.Yelon. For mouse embryos, in situ hybridization was performed as described [49]. The mouse antisense probes for Bmp10, Anf, Smad6, and TGFb2 were obtained by RT-PCR of a cDNA fragment including 39 UTR.

Analysis of cell proliferation
To analyze cell proliferation in the zebrafish heart, aprox. 2 nl of 1 mg/ml bromodeoxyuridine (BrdU) was injected in the pericardiac cavity of tricaine anesthetized embryos, which were fixed after one hour. The distribution of BrdU-positive cells along the progressive zone of the 48hpf fin bud confirmed that this approach is reliable (data not shown) [49]. Whole-mount immunofluorescence was performed using standard protocols using anti-BrdU antibody conjugated to fluorescein (Roche). BrdU-injected embryos were then stripped of their head, oriented in solidifying 0.5% agarose and photographed using a BioRad confocal microscope. To analyze cell proliferation in the mouse heart, pregnant female mice were IP injected with BrdU solution at 100 mg/g body weight. Embryos were recovered after 1 h, fixed and embedded in paraffin. Ten-micrometer serial sections were cut and processed for staining. Epitopes were recovered by boiling the slides for 20 minutes in unmasking solution (VectorLabs). Mice immunohistochemistry was performed according to standard protocols. The antibodies used were MF20, 1:100, obtained from the Developmental Studies Hybridoma Bank, anti-BrdU conjugated to fluorescein, 1:100 (Roche), anti-BrdU, 1:200 (Accurate Chemical) and anti-phospho Histone3, 1:300 (Ser10; Upstate).

Generation of Tbx3 KO mice
Targeted disruption of the mouse Tbx3 locus to produce the Tbx3neo allele by homologous recombination in embryonic stem (ES) cells was performed. Details of the targeting strategy and other analyses are shown in Supplemental Figure S4 and Supplemental Methods S1. In brief, a loxP-flanked neo selection cassette was inserted into the first exon of the Tbx3 locus by homologous recombination in J1 ES cells. This neo cassette contains multiple stop codons that prevent translation of Tbx3 mRNA into protein, thus we considered this generated allele as a null mutated allele, and called it Tbx3neo. Germ-line chimeras were generated by injection of targeted ES cells into C57BL/6 host blastocysts. Chimeras were mated with 129t females and the first progeny were confirmed to harbor the Tbx3neo allele by Southern analysis and PCR.