Tbx2 is essential for patterning the atrioventricular canal and for morphogenesis of the outflow tract during heart development

T-box genes encode a family of transcription factors that are molecularly and evolutionarily related by the presence of a DNA-binding T-box domain(Bollag et al., 1994). Members of the T-box gene family are found in all metazoans, ranging from hydra to humans, and are notable for the crucial roles they fulfil during embryonic development (Adell et al.,2003; Papaioannou,2001; Showell et al.,2004). The idea that T-box genes perform crucial developmental functions stems from several observations. First, most T-box genes exhibit specific and dynamic expression profiles during embryogenesis(Papaioannou, 2001). Second,mutations in several T-box genes have been associated with genetic developmental disorders in humans (Bamshad et al., 1997; Basson et al.,1997; Braybrook et al.,2001; Lamolet et al.,2001; Yagi et al.,2003). Third, targeted mutagenesis of T-box genes in mice results in severe developmental phenotypes(Chapman and Papaioannou,1998; Herrmann et al.,1990; Naiche and Papaioannou,2003) that frequently model definitive aspects of related human disorders (Bruneau et al.,2001; Davenport et al.,2003; Jerome and Papaioannou,2001; Lindsay et al.,2001).

Several T-box genes have been shown to play essential roles in heart development. At 7.5 days post coitus (dpc), the embryonic mouse heart exists as a crescent-shaped field of cells located in anterior splanchnic mesoderm. As the embryo undergoes turning, the cardiac crescent fuses into an anteroposteriorly oriented linear heart tube. The heart tube grows and expands, and around 8.5 dpc begins a complex morphogenetic looping process that eventually brings the posterior, venous aspect of the tube to a rostral position dorsal to the outflow tract. As looping progresses through 9.5 dpc,chamber formation and septation begins such that, by 10.5 dpc, the heart has transformed from a linear tube into a four-chambered structure with prospective right and left atria and ventricles(Kaufman and Bard, 1999).

Chamber formation involves at least two important processes. First, two myocardial domains on the outer curvature of the heart tube differentiate as either an atrial or ventricular chamber. Importantly, part of the heart tube,including the outflow tract (OFT), inner curvature, atrioventricular canal(AVC) and inflow tract (IFT), escapes this developmental chamber program(Moorman and Christoffels,2003). Simultaneously, an epithelial-to-mesenchymal transformation occurs on the inner walls of the AVC and OFT. These mesenchymal cells and their cardiac jelly matrix form two sets of opposing structures called endocardial cushions. The endocardial cushions grow, fuse, and serve as the precursors of the valves and contribute to septation. Subsequently, cardiac septation leads to division of the outflow tract into two separate outlets,and division of the atrium and ventricle into right and left chambers(Moorman and Christoffels,2003).

Of the six T-box genes known to be expressed in specific patterns during cardiogenesis in mouse, Tbx1, Tbx2, Tbx3, Tbx5, Tbx18 and Tbx20, targeted mutagenesis has demonstrated that Tbx1 and Tbx5 have essential roles during cardiac development(Braybrook et al., 2001; Bruneau et al., 2001; Chapman et al., 1996; Christoffels et al., 2004; Habets et al., 2002; Hoogaars et al., 2004; Jerome and Papaioannou, 2001; Kraus et al., 2001a; Kraus et al., 2001b; Lindsay et al., 2001). Heterozygous Tbx1 mutants display abnormal aortic arch artery remodelling, and homozygous mutants fail to septate the outflow tract(Jerome and Papaioannou, 2001; Lindsay et al., 2001). Heterozygous Tbx5 mutants have conduction and septation defects,accompanied by reduced embryonic expression of the cardiac factor genes connexin40 (Cx40; Gja5 – Mouse Genome Informatics) and natriuretic precursor peptide type A (Nppa, formerly known as atrial natriuretic factor or Anf). Homozygous Tbx5 mutants develop hypoplastic left ventricles and atria, with altered embryonic expression of several additional cardiac factors (Bruneau et al., 2001).

Previous work has shown that Tbx2 expression is first detected in the mouse embryo at 8.5 dpc in the allantois(Mahlapuu et al., 2001), and at 8.75 dpc in OFT, AVC and IFT myocardium(Christoffels et al., 2004; Habets et al., 2002). At 9.5 dpc, Tbx2 is expressed in the myocardium of the OFT, inner curvature,AVC and IFT (Christoffels et al.,2004; Habets et al.,2002). A similar pattern of expression is observed in the developing chick heart (Gibson-Brown et al., 1998b; Yamada et al.,2000). Additionally, Tbx2 is expressed at 9.5 dpc in the optic and otic vesicles, and in the naso-facial mesenchyme, and later in the developing limbs and other internal organ primordia such as the lungs and genitalia (Chapman et al.,1996; Gibson-Brown et al.,1996; Gibson-Brown et al.,1998a; Gibson-Brown et al.,1998b).

Based on the cardiac expression profile of Tbx2, and on evidence that Tbx2 can act as a transcriptional repressor(Chen et al., 2004; Sinha et al., 2000), a model has been proposed whereby Tbx2 regionalizes chamber differentiation to the prospective ventricle and atrium by repressing these programs in the OFT, inner curvature, AVC and IFT at 9.5 dpc(Christoffels et al., 2004; Habets et al., 2002). In vitro reporter assays and transgenic analyses in mice have shown that Tbx2 can repress the transcription of Nppa, Cx40 and connexin43(Cx43; Gja1 – Mouse Genome Informatics), cardiac genes whose expression is specifically restricted to the developing chambers(Chen et al., 2004; Christoffels et al., 2004; Habets et al., 2002). Additionally, transgenic embryos in which Tbx2 is ubiquitously expressed throughout the heart tube, under the control of a βMHCpromoter fragment, exhibit arrested cardiac development at looping, and a failure of chamber-specific myocardial gene expression, including Nppa (Christoffels et al.,2004).

An alternative, yet compatible, hypothesis suggests that Tbx2regulates cellular proliferation and/or survival via transcriptional repression of downstream targets, such as p19^ARF and p16^INK4a from the cyclin-dependent kinase inhibitor (Cdkn)2a locus, p15^INK4b from Cdkn2b, and p21 from Cdkn1a (Jacobs et al.,2000; Lingbeek et al.,2002; Prince et al.,2004). A senescence bypass screen using prematurely senescing Bmi1^–/– murine embryonic fibroblasts identified TBX2, with further analysis showing that senescence bypass was likely achieved by downregulation of p19^ARF expression and p53 protein levels. TBX2-expressing fibroblasts also exhibited reduced p16^INK4a and p15^INK4b transcription(Jacobs et al., 2000). Subsequent work showed that the p19^ARF promoter contains a functional T-box-binding element (Lingbeek et al., 2002). Others have shown that Tbx2 can specifically regulate transcription of p21 (WAF)(Prince et al., 2004). These results, in combination with the observation that TBX2 is amplified,and sometimes overexpressed, in a subset of primary breast tumors, breast tumor cell lines and pancreatic cancer cell lines(Barlund et al., 2000; Jacobs et al., 2000; Mahlamaki et al., 2002), has led to the hypothesis that Tbx2 regulates cell proliferation or apoptosis through p21, p15^INK4b, p16^INK4a, and/or p19^ARF and p53.

We have used targeted mutagenesis of Tbx2 in mice to gain a greater insight into the function of this gene during normal embryonic development. We engineered an ∼2.2 kb deletion of the endogenous Tbx2 locus, including part of the T-box, to generate a null allele. Mice heterozygous for the targeted locus appear normal and fertile, whereas homozygous mutant embryos exhibit lethal cardiovascular defects, revealing a crucial role for Tbx2 during cardiac development. Abnormal expression of cardiac chamber markers Nppa, Cx40 and chisel (Csl; Smpx – Mouse Genome Informatics) is observed in the AVC of homozygous mutants at 9.5 dpc. At 11.5-12.5 dpc, surviving homozygous mutants exhibit abnormal OFT septation and other cardiac remodeling defects, although at 9.5 dpc, markers of neural crest (NC) cells and cardiac progenitor populations contributing to the arterial pole of the heart are normally expressed. All homozygous mutants are dead by 14.5 dpc. Further studies addressing the role of Tbx2 as a regulator of the cell cycle reveal that loss of Tbx2 function in mouse embryos is not sufficient to deregulate cell proliferation through p21, p15^INK4b,p16^INK4a, p19^ARF or p53.

Mouse Tbx2 genomic clones in Lambda FIXII were isolated from a 129/SvJ genomic library (Stratagene, La Jolla, CA). 2.9 kb of upstream and 6.6 kb of downstream homologous DNA sequence were ligated to a loxP-flanked PGK-neo PGK-thymidine kinase selection cassette, using XbaI and XhoI sites, respectively, to generate a Tbx2-targeting construct(Fig. 2A). The construct was designed to produce an ∼2.2 kb deletion removing 207 bp of exon 1 and all of exon 2, both containing T-box coding sequence. A β-actin diphtheria toxin negative-selection cassette was placed at the 5′ end of the targeting construct. A linearized targeting construct was electroporated into R1 ES cells (Nagy et al.,1993) to generate targeted cell lines. Targeting was assessed by Southern analysis on EcoRV genomic digests, using both 5′ and 3′ external probes. A targeted line was electroporated with pIC-Cre to remove the PGK-neo PGK-thymidine kinasecassette, resulting in the Tbx2^tm1Pa allele. Germ-line chimeras were generated by injection of targeted, Cre-excised ES cells into C57BL/6NTac host blastocysts.

Chimeras were mated with 129/SvEv/Tac females, to maintain the allele on an inbred background, and with C57BL/6Tac females. C57BL/6×129 progeny were subsequently mated with random-bred ICR (Taconic) females. The first progeny of the chimeras were confirmed to harbor the Tbx2^tm1Paallele by Southern analysis of XhoI genomic digests, using a 5′internal probe that binds a wild-type 5.1 kb fragment and a mutant 2.8 kb fragment (Fig. 2B). Subsequently, mice and embryos were genotyped by 3-primer PCR using the following primers:

1. 5′-CCAGCCAGGGAACATAATGAGG-3′;

2. 5′-CTGTCCCCTGGCATTTCTGG-3′; and

3. 5′-CCTGCAGGAATTCCTCGACC-3′.

This PCR generates a 180 bp product from the wild-type allele and an 88 bp product from the mutant allele (Fig. 2C).

The Tbx2^tm1Pa allele has been maintained both on the 129 inbred background and on a mixed (129/C57/ICR) genetic background. Heterozygous mice were intercrossed to generate homozygous mutant embryos. Embryos were dissected in phosphate-buffered saline (PBS) containing 0.2%bovine albumin (fraction V) (Sigma, St Louis, MO). Embryos for in situ hybridization and immunocytochemistry were fixed in 4% paraformaldehyde overnight, dehydrated in methanol and stored at –20°C. Embryos for histology were fixed in Bouin's fixative, dehydrated in ethanol and stored at 4°C. Yolk sacs were used for genotyping by PCR.

Embryos were collected at 10.5, 11.5 and 12.5 dpc for histology. Paraffin-embedded embryos were sectioned at 8 μm and stained with Hematoxylin and Eosin Y. In situ hybridization was performed according to previously described protocols (Wilkinson,1992), using the following probes: mouse Tbx2, Tbx3, Tbx5,Csl, Cx40, Cited1, MLC2v, βMHC, eHAND and Crabp1;and rat Nppa and Islet1. Stained whole-mount embryos were post-fixed in 4% paraformaldehyde for vibratome sectioning. Embryos were infiltrated with 4% sucrose in PBS, 30% sucrose in PBS, and transferred to embedding mix (0.44% gelatin, 14% bovine serum albumen, 18% sucrose in PBS). Embryos were embedded in fresh embedding mix with the addition of glutaraldehyde (0.25% final concentration). Sections (50 μm) were cut on a Vibratome 1000 Plus Sectioning System (The Vibratome Company, St Louis, MO). Immunocytochemistry with rabbit anti-phospho-histone H3 primary IgG (Upstate Biotechnology, Lake Placid, NY) was performed according to standard protocols(Davis, 1993). Secondary antibody was peroxidase-conjugated goat anti-rat IgG (Jackson Immunoresearch Laboratories, West Grove, PA). Stained whole-mount embryos were post-fixed in 4% paraformaldehyde and embedded in paraffin wax for sectioning. Sections were counterstained with Nuclear Fast Red. Mitotic cells were counted in the heart tube. Alcian Blue cartilage staining was performed as previously described(Jegalian and De Robertis,1992).

Previously characterized 1v-nlacZ-24 transgenic mice(Kelly et al., 2001) were bred with Tbx2^tm1Pa heterozygous males of a mixed 129/C57/ICR background. Tbx2^tm1Pa heterozygous mice carrying the 1v-nlacZ-24 transgene were crossed with Tbx2^tm1Paheterozygous mice to collect embryos at 9.5 and 12.5 dpc. β-Galactosidase staining was performed according to a previously described protocol(Kelly et al., 2001).

At 9.5 and 10.5 dpc, both whole embryos and a dissected trunk region including the heart were collected and stored in RNAlater RNA stabilization reagent at 4°C or –20°C (QIAGEN, Valencia, CA). RNA was extracted using RNeasy Protect Mini kit (QIAGEN, Valencia, CA), and cDNA was reverse-transcribed using SuperScript III First-Strand Synthesis System(Invitrogen Life Technologies, Carlsbad, CA). p19^ARF,p16^INK4a, p15^INK4b and p21 expression were assayed with semi-quantitative real-time RT-PCR performed on a DNA Engine Opticon 2 Continuous Fluorescence Detection System (MJ Research, Waltham, MA). cDNA from adult testis was used as a positive control for p19^ARF+p16^INK4a and p15^INK4b expression. Mandible from a 14.5 dpc embryo was used as a positive control for p21 expression. The following primers were used for the analysis:

1. p19^ARF+ p16^INK4a, 179 bp product, (a)5′-GGTGGTCTTTGTGTACCGCT-3′, (b)5′-GCCACATGCTAGACACGCTA-3′;

2. p21, 281 bp product, (a) 5′-GTACTTCCTCTGCCCTGCTG-3′, (b)5′-CACAGAGTGAGGGCTAAGGC-3′;

3. p15^INK4b, 178 bp product, (a)5′-AGATCCCAACGCCCTGAAC-3′, (b)5′-CTTCCTGGACACGCTTGTC-3′;

4. Hprt, 195 bp product, (a) 5′-AGCAGTACAGCCCCAAAA-3′, (b)5′-TTTGGCTTTTCCAGTTTCA-3′.

Whole-mount in situ hybridization shows Tbx2 expression in the cardiac crescent of 8.5 dpc mouse embryos(Fig. 1A). Vibratome sectioning of whole-mount stained embryos confirms expression in the cardiac mesoderm at 8.5 dpc (Fig. 1B). Tbx2 expression is detected in the atrium and IFT of 8.5 dpc embryos(Fig. 1C). Hybridization of 9.5 dpc embryos reveals previously unreported areas of embryonic expression,including the septum transversum (Fig. 1D,E,H), bilateral nephrogenic mesodermal cords (intermediate mesoderm; Fig. 1D-F) and ventral body wall mesoderm caudal to the forelimbs(Fig. 1D,E). Expression is also found in pharyngeal arch mesenchyme that contains neural crest cells,including those migrating into the OFT septum(Fig. 1G). Tbx2 is also expressed in the OFT, inner curvature, AVC and IFT of the 9.5 dpc mouse heart (Fig. 1H-L), in agreement with published expression data(Christoffels et al., 2004; Habets et al., 2002). Notably,the highest levels of cardiac expression at 9.5 dpc are found at the AVC on the outer curvature (Fig. 1K,L).

To investigate the developmental functions of Tbx2, a targeted mutation was introduced into mice via embryonic stem (ES) cell-mediated mutagenesis. Successful targeting and Cre-mediated excision of the selection cassette generated the final Tbx2^tm1Pa allele,containing an ∼2.2 kb deletion that includes the first two exons of the T-box coding sequence (Fig. 2A). The presence of this allele in mice was identified by Southern blot, PCR and sequence analysis(Fig. 2B,C).

Heterozygotes containing the Tbx2^tm1Pa allele on a mixed 129/C57/ICR background are viable, fertile, and display no obvious phenotypic abnormalities. No Tbx2^tm1Pa/Tbx2^tm1Pa mutants were recovered postnatally, indicating that the homozygous state is embryonic lethal(Table 1). Homozygous embryos collected between 8.5 and 18.5 dpc from heterozygous intercross matings were present at Mendelian ratios, suggesting that no homozygous mutants are lost due to preimplantation defects (Table 1, χ₂=4.85, P>0.05). However, all homozygous mutants are dead by 14.5 dpc apparently due to cardiovascular insufficiency (Table 2). The mutant phenotype is first discernable at 9.5 dpc in 35% (n=25/72) of homozygous mutants, by the absence of a constriction at the AVC and/or an enlarged and dilated ventricle (Fig. 3A-F). At 10.5 dpc, 26% (n=14/53) of homozygous mutants exhibit abnormal AVC morphology. Inflated pericardial sacs and generalized edema indicate that these embryos are suffering from circulatory distress(Fig. 3H). The hearts have undergone looping, but show the absence or reduction of an atrioventricular constriction (Fig. 3I). Transverse sections reveal that endocardial cushion development is compromised in both the AVC and OFT (Fig. 3J-O). Small endocardial cushions are observed in the OFT, which appears shortened (Fig. 3L),but cushion formation is more severely affected at the AVC, where only minor cushion-like structures from the inner curvature can be identified(Fig. 3O). Other homozygous mutants display normal cushion and atrioventricular development(Fig. 3K,N). Forty-two percent(n=11/26) of homozygous mutants are dead at 11.5 dpc(Table 2). Other Tbx2-expressing tissues are also affected: the facial region is dysmorphic with hypoplastic pharyngeal arches and the eyes have morphology equivalent to that of 9.5 dpc embryos (Fig. 3H). These phenotypic features are currently under study (Z.H. and V.E.P., unpublished).

To address whether the variability of the Tbx2^tm1Pa/Tbx2^tm1Pa phenotype could be attributed to the influence of genetic modifiers on the outbred background, a morphological analysis on an inbred 129 background was performed. Embryos of each genotype were present between 9.5 and 14.5 dpc at Mendelian ratios(Table 1,χ ₂=0.0984). Variable penetrance of a morphological atrioventricular phenotype at 10.5 dpc, and the presence of both normal and dead homozygous mutants at 12.5 dpc, suggest that genetic background plays a minimal role in affecting the range of phenotype(Table 2).

Most homozygous mutants that survive to 12.5 dpc show signs of circulatory distress, including pericardial effusion and generalized edema(Fig. 4A), and all homozygous mutants are dead by 14.5 dpc (Table 2). Gross morphological analysis of dissected 12.5 dpc hearts reveals that many homozygous mutants display abnormal OFT development, such that the base of the aorta is positioned to the right of the pulmonary trunk(Fig. 4B). In wild-type embryos, the OFT becomes divided by the fusion of endocardial cushions,resulting in the formation of two separate outflow tracts exiting from specific chambers: the aorta from the left ventricle, the pulmonary trunk from the right ventricle. Histological analysis revealed that OFT septation is delayed by ∼0.5 dpc in homozygous mutant embryos(Fig. 4D-I), and that the aortic outlet is not aligned with the left ventricle relative to normal 12.5 dpc littermates (Fig. 4J,K). Histology showed that homozygous mutants surviving to 12.5 dpc have normal atrioventricular cushion morphology (data not shown). Defects in aortic arch artery remodeling were also observed. The right 6th arch artery, which normally degenerates by 12.5 dpc, persists in half the homozygous mutants analyzed at this age (n=3/6) (Fig. 4L). A rightward positioned aorta and failure of the OFT to septate properly will lead to double-outlet right ventricle or other OFT anomalies that could contribute to lethality at 13.5-14.5 dpc.

Whole-mount in situ hybridization was used to address the hypothesis that Tbx2 regulates the boundaries of atrial and ventricular development by repressing differentiation in myocardium of the OFT, inner curvature, AVC and IFT. Genes such as Nppa, Csl and Cx40 are among the earliest differentiation markers of working atrial and ventricular myocardium,and are absent from Tbx2-expressing AVC myocardium of 9.5 dpc mouse hearts (Fig. 5A,C,E)(Moorman and Christoffels,2003). Nppa is a hormone involved in the homeostatic regulation of blood pressure and volume in adults(Walther et al., 2002). Previous reports have shown that Tbx2 downregulates expression of an Nppa reporter in transgenic mouse embryos at 10.5 dpc(Habets et al., 2002). Consistent with these results, ectopic Nppa expression was found in the AVC of Tbx2^tm1Pa/Tbx2^tm1Pa mutant embryos at 9.5 dpc (n=4/4) (Fig. 5B). Also, Csl, encoding a regulatory component of muscle cytoskeleton (Palmer et al.,2001), was abnormally expressed in the AVC of homozygous mutants(n=4/4; Fig. 5D). Similar inappropriate AVC expression of Cx40, the gene encoding a gap junction component involved in the electrical coupling of cells and tissues(Delorme et al., 1997), was observed (n=4/4; Fig. 5F). Msg1, encoding a transcription cofactor also known as Cited1, is normally expressed in the ventricle of 9.5 dpc embryos(Fig. 5G)(Dunwoodie et al., 1998). Expression in homozygous mutants (n=3) ranges from normal to expanded expression in the AVC (Fig. 5H). These results suggest that the normal transcriptional program of AVC myocardium has been replaced by that of chamber myocardium.

Other markers were analyzed to assess the status of anteroposterior patterning in the heart tube and the possibility of cross-regulation between T-box genes exhibiting overlapping expression during cardiac development. The expression profile of Tbx5, normally within a continuous myocardial zone from the presumptive inter-ventricular boundary into the IFT(Fig. 5I)(Chapman et al., 1996; Gibson-Brown et al., 1998b),is unaffected in Tbx2^tm1Pa/Tbx2^tm1Pa mutants(n=4/4; Fig. 5J). Myosin light chain 2v (MLC2v; Myl2 – Mouse Genome Informatics) is normally expressed in the OFT, ventricles and AVC at 9.5 dpc(Fig. 5K)(O'Brien et al., 1993). MLC2v expression is normal in homozygous mutant embryos(n=4/4; Fig. 5L).β-Myosin heavy chain (βMHC) expression is initially present throughout the linear heart tube but is subsequently repressed in developing atrial myocardium, such that, at 9.5 dpc, the posterior boundary of expression is found at the AVC/left atrium boundary(Fig. 5M)(Lyons et al., 1990). In Tbx2^tm1Pa/Tbx2^tm1Pa embryos (n=4/4),β MHC expression extends posterior of this border(Fig. 5N), suggesting a possible delay in transcriptional downregulation in the atrium at 9.5 dpc. The expression of eHAND (Hand1 – Mouse Genome Informatics), normally observed in the left ventricle of 9.5 dpc wild-type embryos (Fig. 5O)(Thomas et al., 1998), is normal in homozygous mutants (n=4/4; Fig. 5P). Tbx3expression in the AVC was also assessed and was found to be unaffected in homozygous mutants (data not shown).

Further expression analysis addressed possible explanations for the OFT defects observed in Tbx2^tm1Pa/Tbx2^tm1Pa mutants at 12.5 dpc. Cellular retinoic acid binding protein 1 (Crabp1) is a NC marker gene normally expressed in dorsal-ventral stripes between the neural tube and the pharyngeal arches at 9.5 dpc(Fig. 5Q)(Giguere et al., 1990). Crabp1 expression is normal in Tbx2^tm1Pa/Tbx2^tm1Pa mutants (n=4/4; Fig. 5R). A population of splanchnic mesoderm that will contribute to myocardium at both poles of the heart, including the OFT and right ventricle (RV), is marked by expression of the LIM homeobox gene Islet1 at 9.5 dpc(Fig. 5S)(Cai et al., 2003). Homozygous mutant embryos have normal Islet1 expression (n=4/4; Fig. 5T). The Fgf10enhancer-trap transgene 1v-nlacZ-24 was bred onto the mutant background to assess the integrity of the anterior heart field (AHF) in Tbx2^tm1Pa/Tbx2^tm1Pa embryos(Kelly et al., 2001). The transgene was normally expressed in RV and OFT myocardium, and pharyngeal arch mesoderm, in 9.5 dpc homozygous mutant (n=7/7) and heterozygous control embryos (Fig. 5U,V). Transgenic homozygous mutant embryos at 12.5 dpc confirm a normal contribution of the AHF to the RV and OFT (n=3/3)(Fig. 5W,X).

Immunocytochemistry with an anti-phospho-histone H3 antibody was used to assay cell proliferation in 9.5 dpc embryos. Whole-mount staining showed no differences in the global pattern of phospho-histone H3-positive cells between wild-type (n=2) and homozygous mutant (n=3) embryos. Cell counts in the entire myocardium showed no difference in the percentage of phospho-histone H3-positive cells between wild-type (1.97%, n=17,738 cells) and homozygous mutant populations (1.89%, n=28,155 cells;χ ₂=0.87, P>0.05), demonstrating normal cell proliferation levels in Tbx2^tm1Pa/Tbx2^tm1Pahearts at 9.5 dpc.

In the most developmentally advanced homozygous mutant embryos recovered, a bilateral, hindlimb-specific, digit IV duplication was observed(n=3/3 embryos; Fig. 6A,B). Alcian Blue staining showed two cartilage condensations within the first phalangeal segment of digit IV (n=4/4 hindlimbs; Fig. 6C). Tbx2expression has been reported in the interdigital mesenchyme between digits IV and V at 13.5 dpc (Gibson-Brown et al.,1996).

A growing body of work suggests that Tbx2 can inhibit cell cycle arrest and/or apoptosis by directly repressing p19^ARFtranscription, which then leads to the repression of p53 activity(Jacobs et al., 2000; Lingbeek et al., 2002). p53 exerts its effects on cellular proliferation/survival through multiple downstream targets, including p21, although many of these factors also regulate the cell cycle via p53-independent pathways(Taylor and Stark, 2001). Recent work has led to the hypothesis that Tbx2 can also regulate proliferation/survival through a p19^ARF-p53-independent p21 pathway(Prince et al., 2004). TBX2 expression has also been shown to downregulate p16^INK4a and p15^INK4b transcription in cultured cells(Jacobs et al., 2000). Semi-quantitative real-time RT-PCR was used to analyze p21, p19^ARF,p16^INK4a and p15^INK4b expression levels in homozygous mutant versus wild-type or heterozygous embryos. Normal p21 expression levels were observed in whole 9.5 dpc (n=2) and 10.5 dpc (n=3)homozygous mutants, and also in a dissected trunk region including the heart from 10.5 dpc homozygous mutants (n=3; Fig. 7A). One primer pair was designed to detect the combined expression of p19^ARF and p16^INK4a expression, which is normally undetectable in wild-type embryos (Zindy et al., 1997). No evidence of upregulated p19^ARF+ p16^INK4a expression was found in whole 9.5 dpc (n=3) or 10.5 dpc (n=4)homozygous mutants, or in a dissected trunk region including the heart from 10.5 dpc homozygous mutants (n=4; Fig. 7B). No evidence of p15^INK4b expression, also undetectable in wild-type embryos(Zindy et al., 1997), was found in whole 9.5 dpc (n=4) or 10.5 dpc (n=4) homozygous mutants, or in a dissected trunk region including the heart from 10.5 dpc homozygous mutants (n=4; Fig. 7C). These results indicate that loss of Tbx2 function is not sufficient to upregulate the expression of p21, p19^ARF,p16^INK4a or p15^INK4b in 9.5-10.5 dpc mouse embryos.

A potential genetic interaction between Tbx2 and p53(Trp53) was investigated by a morphological analysis of Tbx2 Trp53/Tbx2 Trp53 double homozygous mutants. Tbx2^tm1Pa heterozygotes on a mixed background were crossed with homozygous 129-Trp53^tm1Tyj mutants(Jacks et al., 1994). If the Tbx2^tm1Pa/Tbx2^tm1Pa phenotype were due to upregulated p53 function, the prediction would be that this phenotype should be rescued in double homozygous mutants. At 10.5 dpc, double homozygous mutants (n=3/8) are edematous with inflated pericardial sacs,dysmorphic faces, hypoplastic arches and delayed optic morphology, similar to Tbx2^tm1Pa/Tbx2^tm1Pa mutants, and at 11.5-12.5 dpc (n=6/8), double homozygous mutants are dead. The similar range of phenotype and time of death among Tbx2^tm1PaTrp53^tm1Tyj/Tbx2^tm1PaTrp53^tm1Tyj mutants and Tbx2^tm1Pa/Tbx2^tm1Pa mutants suggests that there is no major developmental genetic interaction between Tbx2 and p53.

To address the developmental functions of Tbx2, we have used targeted mutagenesis in mice to produce a ∼2.2 kb deletion in the endogenous Tbx2 locus, including sequence encoding part of the DNA-binding T-box domain. Our results show that heterozygotes are viable,fertile, and grossly normal, but that homozygous mutants die between 10.5 and 14.5 dpc because of cardiac insufficiency. A quarter of homozygous mutants dissected at 10.5 dpc exhibit signs of circulatory distress, including inflated pericardial sacs and generalized edema, and approximately 40% are dead by 11.5 dpc; the rest are dead by 14.5 dpc. Histological analysis of 10.5 dpc homozygous mutants shows the presence of distinct ventricular and atrial myocardial morphologies, but deficient endocardial cushion formation in the AVC. Importantly, the first observable defect in homozygous mutants presents as abnormal morphology at the AVC and/or left ventricle at 9.5 dpc, a day before the circulatory distress surfaces.

The cardiac phenotype of Tbx5 homozygous null mutant mice(Bruneau et al., 2001), in vitro reporter assays, and transgenic analysis of the regulation of the chamber-specific gene Nppa(Christoffels et al., 2004; Habets et al., 2002) have culminated in the following hypothesis regarding the development of chamber myocardium in the 9.5 dpc mouse heart: Tbx5 activates a chamber differentiation program and Tbx2 spatially restricts this program by repressing a set of downstream target genes in non-chamber myocardium of the AVC and OFT. Tbx5 homozygous mutants exhibit reduced expression of at least two chamber-specific genes, Nppa and Cx40(Bruneau et al., 2001). Biochemical evidence has shown that Tbx5 and Nkx2.5 can specifically and cooperatively bind the promoter of Nppa, synergistically activating reporter expression (Hiroi et al.,2001). Nkx2.5 is a homeodomain transcription factor that is one of the earliest markers of the mouse cardiac lineage and interacts with a number of factors in the cooperative regulation of downstream targets(Bruneau et al., 2000). Nkx2.5 homozygous mutants display reduced expression of a number of cardiac genes, including Nppa(Tanaka et al., 1999). Biochemical experiments have shown that Tbx2 also has the capacity to specifically regulate Nppa in cooperation with Nkx2.5 and that this interaction is preferred in competitive binding assays between Tbx2, Tbx5 and Nkx2.5 (Habets et al., 2002). Tbx2 can also specifically regulate the expression of other chamber-specific genes, including Cx40(Christoffels et al.,2004).

The embryonic expression profile of Tbx2, and both the morphological and molecular aspects of the Tbx2 homozygous mutant phenotype, support a model in which Tbx2-mediated repression localizes chamber differentiation to the prospective ventricle and atrium at 9.5 dpc. Whole-mount in situ hybridization data confirm previously reported results that Tbx2 is normally expressed in myocardium of the OFT,inner curvature, AVC and IFT of the 9.5 dpc mouse heart(Christoffels et al., 2004; Habets et al., 2002). While AVC morphology is normal in a subset of Tbx2^tm1Pa/Tbx2^tm1Pa mutants, all homozygous mutants exhibit ectopic expression of the chamber-specific markers analyzed: Nppa, Csl, Cx40 and Cited1. Importantly, there is no evidence that Tbx2 directly affects anteroposterior patterning of the heart tube, as the posterior expression boundaries of MLC2v and eHAND are unaffected in homozygous mutants. Loss of Tbx2,however, can affect the expression pattern of some genes, such asβ MHC, which fails to be downregulated in Tbx2^tm1Pa/Tbx2^tm1Pa atria.

The abnormal development of the AVC provides a plausible explanation for the lethality observed amongst Tbx2^tm1Pa/Tbx2^tm1Pa mutants at 10.5-11.5 dpc. Insulation of the ventricular and atrial chambers by a distinct myocardial zone is crucial for the mechanical and electrical isolation of chambers that are functionally separate in the mature heart. The molecular characterization described above suggests that Tbx2^tm1Pa/Tbx2^tm1Pa hearts are developing atrial and ventricular chambers whose conduction and contraction are coupled. The lack of functional separation between the developing chambers could explain the lethal cardiac distress that kills many homozygous mutants.

Many Tbx2^tm1Pa/Tbx2^tm1Pa embryos escape the cardiac distress imposed by ectopic chamber differentiation in the AVC at 9.5 dpc. Of those that survive, however, many experience similar distress by 12.5 dpc. The homozygous mutant population displays a range of defects in septation and remodeling of the OFT and aortic arch arteries. Histological analysis indicates misalignment of the aorta and pulmonary trunk with the appropriate ventricles. Although interventricular septum formation is still incomplete by 12.5 dpc, persistence of such misalignment will eventually result in double-outlet right ventricle, where both the aorta and pulmonary trunk emerge from the right ventricle. There are several possible explanations for the defects observed in 12.5 dpc Tbx2^tm1Pa/Tbx2^tm1Pa mutants. First, the defects may be due to misalignment of the OFT and the ventricles, as a secondary consequence of torsional strains on looping imposed by abnormal AVC development. Another possibility is that homozygous mutants surviving to 12.5 dpc experience primary defects in the elongation, septation and rotation of the OFT, such that the ventricular outlets never achieve their final alignment. The fact that Tbx2 is expressed in the OFT from 8.75 dpc until late fetal stages supports the latter hypothesis(Fig. 1G,I)(Christoffels et al., 2004). However, normal expression of Islet1 and the Fgf10enhancer-trap 1v-nlacZ-24 transgene suggests that the population of cells contributing the majority of myocardium to the OFT and RV are present and normal in Tbx2^tm1Pa/Tbx2^tm1Pa mutants. A third possibility is that Tbx2 is required in the cardiac NC for OFT septation, or that Tbx2 targets regulate NC deployment during septation and aortic arch artery remodelling(Kirby and Waldo, 1995),although normal Crabp1 expression in homozygous mutant embryos at 9.5 dpc provides preliminary evidence against this hypothesis. Conditional mutagenesis experiments will be required to resolve this issue.

Few Tbx2^tm1Pa/Tbx2^tm1Pa embryos survive past 13.5 dpc, but those that do display bilateral, hindlimb-specific, distal digit IV duplications, revealing a late role for the gene during patterning of the hindlimb autopod. Tbx2 is expressed in the interdigital mesenchyme between developing digits IV and V of 13.5 dpc mouse embryos(Gibson-Brown et al., 1996),and our observations could be compatible with the involvement of Tbx2in regulating apoptosis in this region. Recent work has implicated Tbx2 as a posteriorizing influence during digit identity specification in chick and mouse (Suzuki et al., 2004). Our results, however, do not address this possible role for Tbx2. Further work will be required before the exact role of Tbx2 during limb patterning and digit specification is understood.

Despite overwhelming evidence connecting Tbx2 to the cell cycle(Barlund et al., 2000; Jacobs et al., 2000; Lingbeek et al., 2002; Mahlamaki et al., 2002; Prince et al., 2004), Tbx2^tm1Pa/Tbx2^tm1Pa embryos offer no evidence that any of the implicated pathways are dysregulated during embryogenesis. There is no difference in p19^ARF, p16^INK4a,p15^INK4b or p21 expression levels between wild-type embryos and homozygous mutants at 9.5 or 10.5 dpc, as assessed by semi-quantitative RT-PCR. Additionally, introduction of the Trp53^tm1Tyjmutation into the Tbx2^tm1Pa line failed to rescue the morphological range or presentation of the Tbx2^tm1Pa/Tbx2^tm1Pa phenotype, eliminating the possibility of a specific genetic interaction. Loss of Tbx2 function is therefore not sufficient to upregulate expression of p19^ARF,p16^INK4a, p15^INK4b or p21 in 9.5-10.5 dpc mouse embryos,nor can the Tbx2^tm1Pa/Tbx2^tm1Pa phenotype be attributed to an excess of p53.

Although Tbx2^tm1Pa/Tbx2^tm1Pa embryos do not exhibit precocious p19^ARF, p16^INK4a, p15^INK4bor p21 expression, or abnormal p53 function, these findings do not rule out a role for Tbx2 in regulating the cell cycle. The knowledge that TBX2 is capable of regulating p19^ARF, p16^INK4a,p15^INK4b and p21 expression in vitro(Lingbeek et al., 2002; Prince et al., 2004), and that TBX2-mediated regulation of p19^ARF has an observable biological effect on cellular senescence(Jacobs et al., 2000),combined with the lack of a cell cycle phenotype in 9.5-10.5 dpc Tbx2^tm1Pa/Tbx2^tm1Pa mutants, suggests that compensating factors may participate in the regulation of this pathway. Tbx3 is the most obvious candidate for several reasons. Within the T-box family, Tbx2 and Tbx3 are more closely related to each other than to any other family member, and both can function as transcriptional repressors (Agulnik et al.,1996; Carlson et al.,2001; Sinha et al.,2000). Tbx2 and Tbx3 exhibit extensive expression overlap in many tissues during mouse development, including the AVC and the limbs (Chapman et al.,1996; Christoffels et al.,2004; Gibson-Brown et al.,1996; Hoogaars et al.,2004). Tbx3 has also been implicated in p19^ARF-mediated regulation of the cell cycle and cell death(Carlson et al., 2002; Lingbeek et al., 2002). Therefore, the likely functional overlap between Tbx2 and Tbx3 may account for our observations that p19^ARF,p16^INK4a, p15^INK4b and p21 expression, and p53 function,are normal in Tbx2^tm1Pa/Tbx2^tm1Pa mutants at 9.5-10.5 dpc. We have initiated an analysis of Tbx2, Tbx3 double mutants to investigate potential functional overlap, particularly with respect to cardiac development and cell cycle regulation.

We thank members of the Papaioannou Laboratory for support and helpful criticism. We thank V. Christoffels and A. Moorman for discussion and the exchange of ideas. We also thank Carl de Luca for discussion and assistance performing RT-PCR. Probes were generously donated by V. Christoffels, R. Harvey, L. Micquerol, S. Dunwoodie and D. Srivastava. This work was supported by NIH grants HD20275 (L.M.S.) and HD33082 (V.E.P.). R.G.K. is an INSERM Research Fellow.