The pattern of congenital heart defects arising from reduced Tbx5 expression is altered in a Down syndrome mouse model

Nearly half of all individuals with Down Syndrome (DS) have some type of congenital heart defect (CHD), suggesting that DS sensitizes to CHD but does not cause it. We used a common mouse model of DS, the Ts65Dn mouse, to study the contribution of Tbx5, a known modifier of CHD, to heart defects on a trisomic backgroun. Mice that were heterozygous for a Tbx5 null allele were crossed with Ts65Dn mice. Thoraxes of progeny were fixed in 10% formalin, embedded in paraffin, and sectioned for analysis of CHD. Gene expression in embryonic hearts was examined by quantitative PCR and in situ hybridization. A TBX5 DNA binding site was verified by luciferase assays. Mice that were heterozygous for a Tbx5 null allele were crossed with Ts65Dn mice. Thoraxes of progeny were fixed in 10 % formalin, embedded in paraffin, and sectioned for analysis of CHD. Gene expression in embryonic hearts was examined by quantitative PCR and in situ hybridization. A TBX5 DNA binding site was verified by luciferase assays. We crossed mice that were heterozygous for a Tbx5 null allele with Ts65Dn mice. Mice that were trisomic and carried the Tbx5 mutation (Ts65Dn;Tbx5+/−) had a significantly increased incidence of overriding aorta compared to their euploid littermates. Ts65Dn;Tbx5+/− mice also showed reduced expression of Pitx2, a molecular marker for the left atrium. Transcript levels of the trisomic Adamts1 gene were decreased in Tbx5+/− mice compared to their euploid littermates. Evidence of a valid binding site for TBX5 upstream of the trisomic Adamts1 locus was also shown. Haploinsufficiency of Tbx5 and trisomy affects alignment of the aorta and this effect may stem from deviations from normal left-right patterning in the heart. We have unveiled a previously unknown interaction between the Tbx5 gene and trisomy, suggesting a connection between Tbx5 and trisomic genes important during heart development.


Background
Congenital heart defects (CHD) comprise the most common congenital anomaly in live births [1]. There is an especially high incidence of CHD in Down syndrome (DS) as 40 -50 % of individuals with trisomy for human chromosome 21 (Hsa21) are affected [2][3][4][5][6][7]. In particular, atrioventricular septal defect (AVSD) occurs in about 20 % of people with DS, a frequency about 2000-fold higher than in the population at large [8]. Since half of those with DS have a normal heart, additional genetic and environmental factors must interact with DS to cause CHD. Thus, dosage effects of Hsa21 genes comprise a complex modifier that, in conjunction with other risk factors such as single-gene variants, alters outcomes of heart development. Trisomy can be thought of as a genetically sensitizing condition that de-stabilizes normal heart development, unveiling roles of disomic modifiers of CHD [8,9]. Attempts to identify genes predisposing to CHD in DS have understandably focused on Hsa21, while there has been little consideration of disomic modifiers that may contribute to this increased risk.
We consider the role of Tbx5, a known contributor to heart development, as a modifier and assess its interaction with trisomy. Tbx5, a transcription factor, has well-described effects in cardiac and limb development [10,11]. Mutations in this gene are associated with Holt-Oram syndrome and about 85 % of affected individuals have a structural heart defect and/or abnormalities in the cardiac conduction system. Holt-Oram patients most often present with atrial septal defects (ASDs) and ventricular septal defects (VSDs) [12]. Tbx5 has a well-studied role in the morphogenesis of the four heart chambers. It is expressed in the left ventricle (LV) and both atria during chamber maturation and septation [13]. Ectopic expression in the right ventricle (RV), or deletion of Tbx5 in the left ventricles (LV) of mice suppresses formation of the ventricular septum, resulting in formation of a single ventricle [14]. Its importance in heart development suggests that variants affecting Tbx5 expression might affect heart development on the "sensitized" trisomic background.
Animal models provide important information for understanding the pathogenesis of CHD and the molecular mechanisms that give rise to these conditions. Orthologs of many genes on Hsa21 are found on mouse chromosome 16 (Mmu16), with smaller subsets on Mmu10 and Mmu17 [15]. The most widely studied DS mouse model, Ts65Dn, is trisomic for a segment of Mmu16 containing about half of the mouse genes orthologous to Hsa21 [16]. The freely segregating extra chromosome carrying these genes also contains genes from Mmu17 that are not conserved with Hsa21, thus this is not an exact model [17,18]. Ts65Dn mice display a number of the features of DS, including cardiac abnormalities, although these occur at a lower frequency than in people with DS [9,19]. Our lab has previously identified the Creld1 and Hey2 genes as disomic modifiers of septal development on this trisomic background [9]. Haploinsufficiency for either of these disomic modifiers alone did not affect heart development, but on a trisomic background the frequency of maldevelopment was increased significantly. The Ts65Dn model thus sensitizes heart development to other genetic perturbations. We used the Ts65Dn mouse model here to examine the role of Tbx5 in heart development.
A mouse model with a null allele for Tbx5 has been described [20]. Homozygous null Tbx5 −/− embryos (Black Swiss/SvJ background) die by embryonic day 10.5 (E10.5) and lack cardiac looping and endocardial cushion formation, among other severe defects [20]. The viability of and defects observed in Tbx5 +/− mice are greatly influenced by genetic background. Bruneau et al. report a 10 % frequency of the Tbx5 null allele on a 129SvEv/129SvJ background at birth and 28 % on a Black Swiss/129SvJ background, instead of the expected Mendelian ratio of 50 %. Deviation from the expected frequency indicates that prenatal loss has occurred; the different frequencies in different mouse strains suggest that genetic background contributes to the penetrance and expressivity of heart phenotypes in this situation [20]. Therefore, the effects of Tbx5 dosage are susceptible to additional genetic modifiers. The molecular mechanisms by which Tbx5 influences heart development are incompletely described and possible interactions between Tbx5 and genes on Hsa21 are unknown. We hypothesize that Tbx5 acts as a genetic modifier to alter CHD in Ts65Dn mice. Here we provide evidence of an interaction between Tbx5 and trisomy and the effects of that interaction on trisomic gene expression and left-right patterning of the heart.

Viability of Tbx5 +/− mice is dependent on genetic background
Crosses between B6.Tbx5 +/− male mice and B6C3.Ts65Dn females were established to examine the role of Tbx5 as a  [20] modifier of CHD. Genetic background of the Tbx5 mice affected viability (Table 1 and [20]). At birth, Tbx5 genotypes appeared at Mendelian ratios on a B6 x C3H (75 % B6, 25 % C3H) background ( Table 2), but by weaning, the frequency of the Tbx5 +/− genotype was 21 % rather than the expected 50 %. On a B6 background, the frequency of the Tbx5 +/− genotype at weaning was 11 %. Thus, genetic factors appear to contribute to perinatal lethality associated with Tbx5 +/− . We examined Tbx5 expression in hearts at E11.5 and confirmed the expected down-regulation in Tbx5 heterozygous null mice (Fig. 1). The original characterization of these mice demonstrated Tbx5 RNA levels substantially less than 50 % at E8.5 [21], and the very low expression seen here at E11.5 reflects that difference, as well. Surprisingly, Tbx5 mRNA was also reduced in Ts65Dn. There may be an additive effect between the two conditions, as the measured levels of Tbx5 mRNA were lower still in Ts65Dn;Tbx5 +/− mice.
Trisomy affects patterns of CHD in Tbx5 +/− mice Progeny of the Ts65Dn x Tbx5 +/− crosses were collected within hours of birth, prepared for histology and assessed for the presence of CHD. All animals examined for CHD were on the B6 x C3H (75 % B6, 25 % C3H) background. There is some loss of Ts65Dn fetuses during late gestation and the observed frequency of 42 % trisomic pups at P0 (Table 2) was in the expected range [22,23]. Tbx5 +/− mice were recovered at Mendelian ratios within hours of birth. CHD is highly penetrant in Tbx5 +/− mice and we saw only a slight overall increase in the percentage of heart defects when the null allele occurred on a trisomic background (Table 3 and Additional file 1: Table S1). However, the pattern of effects was altered significantly by trisomy.

Candidate trisomic gene interaction with Tbx5
The simplest explanation for the increased incidence of overriding aorta in Ts65Dn;Tbx5 +/− mice is an interaction between Tbx5 and a trisomic gene(s) that is regulated by this transcription factor. We considered 109 genes that are trisomic in Ts65Dn mice [24] and generated a list of candidate trisomic genes based on expression patterns (Additional file 5: Figure S1). Sixty-four of these genes are expressed in the heart during development and 43 of them contained regions that associated with TBX5 in a ChIP study [25]. However, examination of the genomic sequences of these 43 revealed a T-box binding element (TBE) in only six (Adamts1, Dyrk1a, Rcan1, Ripply3, Sh3bgr and Wrb). We used the Transcription Element Search System program (TESS; see Methods) to identify potential TBEs in the first exons and/or promoter There is a significant difference between euploid and trisomic Tbx5 heterozygous mice in the occurrence of overriding aorta (p = 0.0004) c The difference between euploid and trisomic Tbx5 heterozygous mice in AVSD is approaching significance (p = 0.07)  regions of those genes [26] (see Methods). Based on prior reports of possible effects on heart development, evolutionary conservation, the presence of TBEs and binding sites for other heart specific transcription factors, we selected Adamts1 for further investigation.
Adamts1 encodes an extracellular matrix protease. Transcript levels were increased in Tbx5 +/− mice (p = 0.04) (Fig. 3a), suggesting that Tbx5 may act as a repressor of Adamts1. Using TESS, we found a putative TBE upstream of the Adamts1 locus and tested it using a luciferase reporter assay. The putative site, located 229 bp upstream of the Adamts1 transcription start site, was identical to the canonical TBE (RGGTGTVR) [27]. He et al. [25] found that this region is bound by TBX5 in a chromatin immunoprecipitation assay. A 238 bp region containing the putative site and located 121 bp upstream of the Adamts1 transcription start site, was amplified by PCR and cloned into the pGL3 luciferase vector (Additional file 6: Figure S2).
We transfected the construct into cells stably over expressing Tbx5 (Additional file 7: Figure S3). Luciferase levels were increased in cells transfected with the Adamts1-luc construct compared to cells transfected with empty pGL3 vector (p = 1.28 × 10 −4 ) (Fig. 3b). Introduction of siRNA directed at Tbx5 reduced this effect (p = 3.35 × 10 −4 ). The putative TBE was mutated to inhibit TBX5 binding and luciferase levels decreased compared to the construct with the intact TBE (p = 1.51 × 10 −4 ).
Tbx5 is generally considered to be an inducer of transcription, consistent with the increase in luciferase signal when TBX5 binds the upstream Adamts1 region included in the construct. However, Adamts1 expression is increased when Tbx5 is reduced in Tbx5 +/− mice (Fig. 3a). This discrepancy is likely due to the involvement of distal elements in transcriptional regulation that were not included in the construct and/or the lack of critical co-regulators in 3T3 cells. It is not surprising that in vivo results would differ in the presence of the entire promoter, as well as other transcription factors and co-regulators.

Atrial isomerism in Ts65Dn;Tbx5 +/− mice
Overriding aorta and Gerbode's defect are seen more often in Ts65Dn;Tbx5 +/− mice than in their Tbx5 +/− littermates or euploid controls. We hypothesized that these might be accompanied by atrial isomerism. In this situation both atria are mirror images of one another and lack the typical morphological and molecular characteristics of either the right or left atrium. Normally, Pitx2 is specifically expressed in the left atrium and it is absent or ectopically expressed in the right atrium in cases of atrial isomerism [28]. To determine whether atrial isomerism occurred in Ts65Dn;Tbx5 +/− mice, in situ hybridization for Pitx2 was performed on coronal paraffin sections of E13.5 thoraxes. Pitx2 expression can be seen in the left atrium of the heart in euploid embryos, but that expression is much weaker in the mutants (Fig. 4a,c). These results were confirmed by quantitative PCR of whole hearts from E13.5 embryos (Fig. 4e), where Pitx2 expression was reduced in Ts65Dn;Tbx5 +/− compared to euploid littermates (p = 0.01). The expected expression pattern was observed in other organs (Additional file 8: Figure S4). The weak Pitx2 expression in the atria of Ts65Dn,Tbx5 +/ embryos suggests that these embryos have defects in laterality (i.e., atrial isomerism). Thus, atrial isomerism likely contributes to the significant increase in overriding aorta in the Ts65Dn;Tbx5 +/− animals.

Discussion
The interaction between Tbx5 dosage and trisomy resulted in a significant increase in defects of aortic alignment and revealed a potential effect on left-right patterning of the heart. The action of genetic modifiers in heart development is evident from the differences in penetrance and patterns of heart phenotypes in DS, in DS mouse models and when the Tbx5 null allele is bred onto different mouse genetic backgrounds. In addition, Holt-Oram patients who inherit the same TBX5 mutation have variable heart phenotypes, indicating that additional factors affect development [29]. Previous work from our lab and others has shown how interactions between trisomy and disomic modifier genes, both in humans and in mouse models, can adversely affect heart development [8,9]. For example, VEGF-A, HEY2 and the matricellular protein CRELD1 each have an impact, but they represent a small subset of genes in which variants contribute risk of CHD. Individuals with DS and mutations in TBX5 displaying OA and a number of additional defects have been described [30]. Isolated OA is not frequent in DS, however it is a component of Tetralogy of Fallot which occurs in 1-8 % of DS births, far higher than in the population at large [2]. Our results demonstrate an interaction between altered Tbx5 expression and trisomy with increased OA as a primary outcome.
Disruptions in the left-right patterning pathway frequently lead to the kinds of heart defects seen here. OA is considered to be a milder form of double outlet right ventricle (DORV), the most common defect seen in Pitx2 knockout mice [31], arising in part due to aortic shifting [32]. Reducing the level of Pitx2 expression is sufficient to cause septal and valve defects [33]. Mice lacking the Pitx2 and Cited2 genes which are involved in left-right patterning exhibit phenotypes similar to the ones reported, including Gerbode's defect and overriding aorta [32][33][34][35][36][37]. Pitx2 is an important determinant of left- right asymmetry of the heart and gut [28,33,[37][38][39][40] and confers left atrial identity; its misexpression can result in atrial isomerism [28,39,41]. T-box family transcription factors are known to regulate Pitx2 expression [42], and Pitx2 regulates Tbx5 expression in the abdominal wall during development [43]. Ts65Dn;Tbx5 +/− embryos have reduced Pitx2 expression, which can contribute to atrial isomerism and aortic defects.
The left-right signaling pathway is affected when trisomy and Tbx5 haploinsufficiency are combined. We suggested previously that a universal response deficit to Sonic hedgehog signaling due to trisomy could explain many of the clinical features of DS [44]; hedgehog signaling plays a regulatory role in left-right patterning and can regulate Pitx2 expression [45][46][47]. Expression of Sonic hedgehog in the second heart field is essential for proper growth of the dorsal mesenchymal protrusion, without which AVSD occurs [48]. A response deficit to SHH in Ts65Dn could contribute to the altered localization of Pitx2 expression seen here.
Finally, we discovered that mice with reduced Tbx5 dosage show increased expression of the Adamts1 gene. Adamts1 is important in the development of the extracellular matrix in endocardial cushions [49], and is a regulator of the VEGF pathway [50]. It is interesting that individuals with DS who have an AVSD show an excess of deleterious variants in VEGF pathway genes relative to those with DS who do not have CHD [8]. Reduced Tbx5 dosage may contribute to the risk of heart defects by a role in transcriptional regulation of the Adamts1 gene, a regulator of the VEGF pathway.
Defects in left-right patterning likely contribute to the significant increase in overriding aorta in Ts65Dn;Tbx5 +/− mice relative to either genetic defects in isolation. The mechanism likely involves reduced expression of Pitx2, a gene with known roles in laterality. In addition, we demonstrated that a change in Tbx5 dosage affects transcription of Adamts1, suggesting it is a target of TBX5 in normal heart development in vivo.

Conclusions
We have shown that Tbx5 dosage and trisomy interact to affect the outcome of heart development, potentially altering left-right patterning of the heart. We provided evidence of weak Pitx2 expression, important for leftright patterning, in the Ts65Dn;Tbx5 +/− mice. We have also provided evidence that TBX5 plays a role in regulation of the trisomic Adamts1 gene. A valid TBX5 binding site was shown upstream of the Adamts1 locus and mutation of the site significantly affected transcription.

Animal husbandry
Animals were maintained in a virus and antibody-free facility with food and water ad libitum. Ts65Dn mice (B6EiC3H-a/A-Ts(17 16 )65Dn) were obtained from The Jackson Laboratory and maintained as an advanced intercross on the B6;C3H background. Tbx5 heterozygous null mice (Tbx5 +/− ) [20] were kindly provided by Dr. Jonathan Seidman. We backcrossed the Tbx5 +/− mice onto a C57BL/6J background for at least five generations. All data was taken after mice had been backcrossed for at least five generations. Progeny in Tables 1  and 2 were the result of separate experiments. Progeny in Table 1 were obtained by crossing Tbx5 +/− mice on a B6 background with wild type B6 mice (obtained from Jackson Laboratory) or B6;C3H mice bred in our lab. Progeny in Tables 2 and 3 were obtained at P0 via crosses between Ts65Dn females (B6;C3H) and Tbx5 +/− males on a B6 background. All procedures were approved by the Institutional Animal Care and Use Committee.
(See figure on previous page.) Fig. 4 Pitx2 expression in E13.5 WT and trisomic, Tbx5 +/− embryos. In situ hybridization of coronal sections of wild type (a,b) and Ts65Dn;Tbx5 +/− (c,d) mice. Pitx2 expression can be seen in the left atrium of the WT animals (a). Pitx2 expression is much weaker in the trisomic, Tbx5 +/− mutants (c). Scale bar equals to 100 μm. RA/LA: right/left atrium. In situ hybridization results were verified through quantitative PCR in E13.5 hearts (e). Error bars are standard deviation from the mean

Histology
We collected the progeny of Ts65Dn x Tbx5 +/− crosses at postnatal day 0 (P0) within hours of birth. We euthanized pups and removed and fixed thoraxes in 10 % formalin for at least 48 h. Tissues were embedded in paraffin, sectioned at 7 μm and stained with hematoxylin/eosin using standard methods. The heart morphology in each animal was scored under a dissecting stereomicroscope (Nikon SMZ1500, Japan) by at least two individuals blinded to genotype. Pictures were taken using the NIS-Elements Br software (Nikon, Japan).
Virtual Histology™ was performed on six Ts65Dn;Tbx5 +/− mice (Numira Biosciences, Salt Lake City, UT). Virtual Histology™ is a micro-CT based method that allows for high resolution imaging. Specimens were stained in a proprietary solution before being scanned at 20 μm resolution with a volumetric micro-CT instrument. Using this method tissues are left intact and imaging analysis can be used to create "virtual" histological sections.

Identification of candidate genes
We examined 109 genes that are trisomic in Ts65Dn mice for localization of expression in the heart using the following databases: EMAGE gene expression database [51], VisiGene image browser [52], the Chromosome 21 gene expression atlas [53], and the MGI gene expression database [54]. Trisomic genes expressed in heart during development, according to the above databases, were compared to a list of genes bound by TBX5 in a chromatin immunoprecipitation (ChIP) study [25]. We generated a list of trisomic candidate genes expressed in heart and bound by TBX5 (Additional file 5: Figure S1, Additional file 9: Table S2). From the list of 43 genes, we chose 6 as top candidates. The top 6 genes were the only genes out of the 43 which had a TBE in their promoter regions, were bound by other heart specific transcription factors, had high evolutionary conservation and were previously implicated in heart development. The Transcription Element Search System [26] was used to search for the TBE (RGGTGTVR) in regions of the 43 candidate genes that had a known or speculated role in heart development [27].

Creation of Tbx5 constitutively over-expressing cells
We obtained a Tbx5 cDNA clone from the I.M.A.G.E. consortium and subcloned it into the pcDNA3.1+ vector (Life Technologies, Carlsbad, CA), putting it under the control of the CMV promoter. NIH-3T3 cells were transfected with the pcDNA-Tbx5 construct using Lipofectamine 2000 (Life Technologies, Carlsbad, CA). Stably transfected cell lines were selected with 1 mg/ ml Geneticin (Gibco, Carlsbad, CA). We maintained the cells in Dulbecco's Modified Eagle Medium with high glucose (4.5 g/L) (Life Technologies, Carlsbad, CA), 10 % fetal bovine serum, 1 mg/ml Geneticin, 1X glutamine and penicillin/streptomycin at 37°C and 5 % CO 2 .

Functional studies and luciferase assay
We amplified the portion of the promoter region of Adamts1 that was bound by TBX5 in a ChIP study [25], by PCR with the insertion of cut sites for KpnI and XhoI. PCR primers are as follows, F: 5′-GGCGCTTATGGTACCTGGT-CACACTTTTTTTGG-3′; R: 5′-GGCGCTTATCTCGAG-CACCTTCACAGAGGCTCA-3′. The amplified region was subcloned into the pGL3 basic vector (Promega, Madison, WI). Site directed mutagenesis was used to mutate the putative TBE site found upstream of the Adamts1 promoter. The following primers were used to mutate the putative TBE site and insert an EcoRI restriction site, Primer 1: 5′-CACAGCTCGTCACTCTGGGAATTCAAGA CGCCGAAACAGCGCTG-3′; Primer 2: 5′-CAGCGC TGTTTCGGCGTCTTGAATTCCCAGAGTGACGAGC TGTG-3′. We transfected the constructs into NIH-3T3 cells constitutively over expressing Tbx5 (Additional file 7: Figure S3). Tbx5 siRNA and scrambled siRNA were purchased from Life Technologies (Carlsbad,CA). Luciferase levels were measured in a 1450 Wallac Jet MicroBeta liquid scintillation and luminescence counter (Perkin-Elmer, Waltham, MA) using the Dual Luciferase Assay Kit (Promega, Madison, WI).