Lethal arrhythmias in Tbx3-deficient mice reveal extreme dosage sensitivity of cardiac conduction system function and homeostasis

We created graded increments in Tbx3 mRNA production by combining different gene-targeted Tbx3 alleles: Tbx3^GH, Tbx3^G,Tbx3^flox, Tbx3^Δflox, and Tbx3^N (Fig. 1 A and B). These incremental changes in Tbx3 mRNA revealed that prenatal survival is markedly sensitive to Tbx3 dosage.

The Tbx3^GH allele was hypomorphic relative to the WT allele in terms of mRNA production (Fig. 1B), and there was markedly decreased Tbx3 mRNA and protein in Tbx3^GH/N compound heterozygotes (Figs. 1B and 2E–H). This is caused, at least in part, by aberrant splicing into GFP, which we could detect immunohistochemically and by RT-PCR. Although the location of Tbx3 expression was the same in hypomorphs and controls, the Tbx3⁺ domains were smaller in the mutants and staining intensity per cell was decreased (Fig. S1). Most Tbx3^GH/N hypomorphs (background 50% Sv129; 50% C57Bl6) died between embryonic day (E) 13.5 and E15.5; none survived to birth (Fig. 1C). Tbx3^GH/N mutants had limb abnormalities similar to Tbx3^tm1Pa/tm1Pa mutants (14) (Fig. S2).

Removal of the hygromycin resistance gene from Tbx3^GH generated the Tbx3^G allele (Fig. 1A). Tbx3^G was also hypomorphic and the level of Tbx3 mRNA in Tbx3^G/G homozygotes was intermediate to that in Tbx3^+/N heterozygotes and Tbx3^GH/N compound heterozygotes (Fig. 1B). Notably, this increase in Tbx3 function permitted 80% of Tbx3^G/G mice to survive to birth. Progressive postnatal loss resulted in an 11% survival rate to adulthood (Fig. 1C).

These findings in hypomorphs are distinct from the previously reported Tbx3 loss-of-function (LOF) mouse mutants: Tbx3^tm1Pa/tm1Pa, Tbx3^Cre/Cre, and Tbx3^neo/neo (13–15). There is significant phenotypic variability among the previously reported lines, the etiology of which is unknown. Our Tbx3^N allele deletes the DNA binding domain similar to the allele reported by Davenport et al. (14), which die between E11 and E16, yet most of our Tbx3^N/N mutants died before E10.5 (56% of expected mutants were present at E10.5, 25% of those recovered were dead; n = 11 litters). Tbx3^N/N cardiac morphology was similar to that reported by Ribeiro et al. (15). Both sexes of our Tbx3^+/N mice demonstrated reduced fertility, and mothers had poor nurturing, likely because of the requirements for Tbx3 in genital and mammary development (8, 14); thus, Tbx3^N/N mutants were exceedingly difficult to generate and not further evaluated. We therefore converted a unique conditional allele (Tbx3^flox) into a null allele (Tbx3^Δflox) by using deleter-Cre mice (16). Tbx3 transcript and protein were undetectable in Tbx3^{Δflox/Δflox} hearts (Figs. 1B and 2A–D), and survival of Tbx3^{ΔfloxΔ/flox} mice was similar to other Tbx3 reported LOF mutants, with none surviving to birth (Fig. 1C).

Tbx3^tm1Pa/tm1Pa mouse mutants reportedly die from yolk sac degeneration (14). However, we detected no morphologic defects in the yolk sacs of Tbx3^GH/N or Tbx3^G/G mutants (Fig. S2). Other potential causes of midgestational death include hematopoietic failure, placental insufficiency, and cardiac failure. There was no evidence of the anemia or liver hypoplasia (Fig. S2) reported in Tbx3^Cre/Cre LOF mutants (17). No differences were detected in embryo and placenta weights, or placental histology (Fig. S2), between Tbx3^GH/N and Tbx3^+/+ control littermates. Thus, we could not attribute the lethality of these mutants to yolk sac, hematopoietic, or placental defects. The timing of death and their edematous appearance (Fig. S2) suggested cardiovascular insufficiency as the likely pathogenic mechanism.

A total of 85% of Tbx3^GH/N E13.5 to E14.5 hypomorphs had large interventricular foramina (n = 7) compared with the closed foramina in 85% of littermate controls (n = 13). Atrial–ventricular alignments were normal, and only rare cases of ventricular–arterial malalignment were present (n = 2). Both these defects are well tolerated prenatally in many different mouse models (18), thus eliminating cardiac structural abnormalities as the cause of universal Tbx3^GH/N embryonic lethality.

In utero echocardiography was performed serially on Tbx3 mutants of different genotypes (Tbx3^{Δflox/Δflox}, Tbx3^GH/N, and Tbx3^G/G) and their littermates from E10.5 to E15.5. No Doppler blood flow abnormalities in umbilical or yolk sac vessels or at the AV or outflow tract cushions were detected, nor were there overall differences in heart rate or cardiac contractile function at these stages (Table S1). However, we identified severe arrhythmias, as detailed below.

Asynchronous atrial and ventricular contraction (an arrhythmia phenotype called AV block) was reproducibly detected in 2D (Movie S1 and Movie S2) and on Doppler echocardiography (Fig. 3A) in Tbx3^GH/N hypomorphic embryos. This arrhythmia was not observed in littermates or in litters of WT embryos, indicating that it is attributable to decreased Tbx3 (Fig. S3). Echocardiographic detection of embryonic AV block has not previously been reported in mice to our knowledge, so we confirmed the electrophysiologic basis for asynchronous contractions by developing a method to directly record embryonic ECGs (Fig. 3B).

AV block occurred intermittently; embryos were found to have AV synchrony alternating with AV block of variable conduction efficiency (2:1–5:1 block) during approximately 5 min of interrogation (Fig. S3). AV block was detected exclusively in mice with the lowest Tbx3 expression levels (Tbx3^{ΔfloxΔ/flox}, Tbx3^GH/N, and Tbx3^G/G) and is most closely associated with fetal demise. In serial studies, AV block and other arrhythmias were identified at E11.5 to E13.5 in mutants with embryonic demise occurring before the next daily study (n = 5) or by the time of euthanasia (n = 2; E15.5). In another cohort of embryos, AV block was detected on 2 to 3 consecutive days between E11.5 and E14.5 (n = 7 mutants), indicating that AV block was not merely a reflection of imminent death. We conclude that inadequate cardiac output due to AV block and ventricular bradycardia causes embryonic death of Tbx3-deficient mutants. AV block was first detectable at E11.5 in Tbx3^GH/N hypomorphs, and occurred in 50% of these embryos at E12.5. With the incrementally higher Tbx3 expression in Tbx3^G/G mutants, AV block was not detected until E13.5 or later. Tbx3^{ΔfloxΔ/flox} null mutants also had AV block. They suffered from earlier demise, and thus few litters were serially studied.

Atrial arrhythmias were seen in all three classes of Tbx3 mutants, and in their littermates (Fig. 3C and Fig. S3). As analysis of four litters of WT embryos between E10.5 and E13.5 revealed rare bradycardia, but never atrial arrhythmias or AV block, the atrial arrhythmias in mice heterozygous for Tbx3 mutations are further evidence of the profound dosage sensitivity of the developing conduction system to Tbx3.

Direct ECG recordings of Tbx3^G/G newborns revealed a prolonged QRS duration and multiple arrhythmias including atrial bradycardia, atrial arrhythmias, and second-degree AV block (Table 1 and Fig. 3D) in the absence of contractile dysfunction. No differences were detected in short-axis dimensions of left ventricular chamber or wall thickness between WT (n = 19) and Tbx3^G/G (n = 12) littermates. In contrast to our observations at embryonic and early fetal stages, heart rates were significantly slower in Tbx3^G/G newborns compared with WT littermates, and fractional shortening was increased, possibly in compensation for their relative bradycardia (Table 1). We evaluated Tbx3^G/G mutants recovered at birth for anatomical malformations. One fourth of the mutants evaluated by serial sections had structural malformations (n = 3 of 13), including tetralogy of Fallot, ventricular septal defect, and interrupted inferior vena cava with azygous continuation. Structural defects are less severe and frequent than those reported in Tbx3^−/− and Tbx3^Cre/Cre embryos (19, 20).

We also recorded ECGs on the rare surviving adult Tbx3^G/G mutants (50% FVB, 25% Sv129, 25% C57Bl6) and found a shortened PR interval and marked heart rate variability compared with WT and Tbx3^G/+ mice (Table 2). Nine percent of adult Tbx3^G/G mutants (n = 2 of 22) had sudden death. Telemetric 24-h ECG recordings revealed a variety of arrhythmias including sinus bradycardia, sinus pauses, premature atrial contractions, junctional rhythm, AV block, likely ventricular preexcitation, premature ventricular contractions, and ventricular couplets consistent with global CCS dysfunction (Table 2, Fig. 3, and Fig. S4). Our interpretation of the rhythm shown in Fig. 3I is that there is electrical preexcitation of the ventricle; an alternative interpretation we cannot exclude with the available data are AV dissociation. The median QRS duration was significantly longer in Tbx3^G/G mice compared with Tbx3^+/+ mice, reflecting the widened QRS observed in the 50% of mutants with likely ventricular preexcitation. The trend toward a shorter median PR interval may also result from preexcitation in mutants. One third of these animals had bradycardia, resulting in a median HR of 608 beats/min inTbx3^G/G mutants compared with 691 beats/min in Tbx3^+/+ (P = 0.06).

Tbx3^Cre/Cre knock-in mutants were reported to have abnormally small SA nodes (13). We quantified the developing SAN in Tbx3^GH/N and Tbx3^G/G mutants by using Hcn4 IHC in E12.5 embryos and newborns, respectively. SAN volume in Tbx3^GH/N hypomorphs was 60% of that in Tbx3^+/+ (0.017 ± 0.001 mm³ vs. 0.027 ± 0.003 mm³; P = 0.05; Fig. S5), and in Tbx3^G/G mutants it was 45% of that in Tbx3^+/+ (0.049 ± 0.017 mm³ vs. 0.107 ± 0.024 mm³; P < 0.01; Fig. S5). We investigated whether this was caused by decreased proliferation or increased cell death and found no significant difference between Tbx3^GH/N and Tbx3^+/+ in the number of BrdU-positive cells in the Hcn4⁺ SAN at E11.5 or E12.5, nor did we detect apoptosis. This agrees with previous studies (13, 21, 22) and suggests that a threshold of Tbx3 is required to establish and/or maintain cells with SAN identity. As the quantity of nodal tissue influences its ability to drive the surrounding myocardium (23, 24), we posit that bradycardia and atrial arrhythmias (Fig. 3 C–E) are related to the smaller SAN of Tbx3 hypomorphic mutants.

To investigate molecular causes of lethal arrhythmias in Tbx3^GH/N hypomorphs, we studied cardiac gene expression and compared it with that previously reported from Tbx3^Cre/Cre mutants (13, 19). Expression patterns of known working myocardial genes Cx40, Nppa, Tbx18, and Tnni3 mRNAs were not significantly changed in Tbx3^GH/N E12.5 hearts. In contrast, Cx43 is a direct target of Tbx3 repression (13) and was ectopically expressed at the crest of the ventricular septum in mutants (Fig. 2 I–J). This is the nascent AVB and a region where Tbx3 mRNA and protein levels were severely decreased in Tbx3^GH/N hypomorphic mutants (Fig. 2H). The expression patterns of Hcn4, Tbx2, Tbx5, and Cacna2d2 (CCS markers) were preserved in Tbx3^GH/N hypomorphs. These findings reveal that the level of residual Tbx3 expression and protein function in Tbx3^GH/N hypomorphs was sufficient to preserve most of the known complementary gene expression boundaries between working and conduction myocardium, but not to prevent CCS dysfunction and lethal arrhythmias.

In the rare surviving adult Tbx3^G/G mutants, histology and connexin protein expression patterns of the CCS were so severely disrupted that no structures with the histologic or molecular identity of the SAN (Fig. 4 G and J vs. Fig. 4 A and D) or AVB (Fig. 4L vs. Fig. 4 C and F) could be detected. These regions were abnormally positive for the working myocardial marker Cx43.

The arrhythmias observed in Tbx3^GH/N hypomorphs indicate that other genes besides those known to distinguish working and conducting myocardium are required for CCS function and must be disrupted by decreased levels of Tbx3. Given the overlapping expression of Tbx2 and Tbx3 in the AVC (19, 25), the linkage of Tbx5 and Tbx3, and the known role of Tbx2 and Tbx5 in CCS function (1, 6, 26), we determined expression of Tbx2 and Tbx5 by quantitative RT-PCR (qRT-PCR) in E10.5 whole hearts, and in microdissected E12.5 atrial/AVC tissue in our allelic series of Tbx3 mutants. We detected no changes in the levels of these transcripts (Figure S6 A and B).

We performed an unbiased, genome-wide expression analysis with Agilent microarrays to identify dysregulated genes downstream of Tbx3 that could be involved in the molecular bases of embryonic AV block. Gene profiles from AVC tissue before arrhythmia onset (E10.5) of Tbx3^GH/N hypomorphs and Tbx3^+/+ controls was compared (array data deposited in Gene Expression Omnibus database; accession no. GSE24122). Analysis revealed 958 transcripts significantly changed by more than 1.5 fold.

We validated the microarray findings by using qRT-PCR to assay genes related to ion handling, with known cardiac expression, or relevant to AVC development or conduction (1, 3, 4). As expected, Tbx3 was strongly decreased (−6.6-fold change by qRT-PCR) in Tbx3^GH/N hypomorphs. A number of genes with the potential to affect conduction system function were highly dysregulated in Tbx3^GH/N mutant AVC. Transcripts for Kcne3 (8.5-fold), a potassium channel regulatory subunit; Chac1 (7.1-fold), a cation transport regulator-like protein; Adora2a (2.3-fold), encoding the adenosine 2A receptor, which contributes to heart rate regulation in adult mice (27); Kcnj4 (1.9-fold), encoding a potassium channel; and Atp1b4 (1.8 fold), encoding a sodium-potassium ATPase, were all significantly increased. Scn7a (−3.5-fold), a voltage-gated sodium channel; Gjc1 (−1.5-fold), which encodes the slowly conducting gap junction Cx30.2; Atp1a2 (−1.5-fold), encoding a sodium-potassium ATPase; and Nkx2.5 (−2.8-fold) and Gata4 (−1.6-fold), which have demonstrated roles in human and mouse arrhythmias (1, 6), were decreased. Bmp10, Irx1, Irx3, Tnni2, Nppa, and Nppb transcripts are expressed in trabeculated ventricular working myocardium (28–31); however, consistent with our hypothesis that Tbx3 is required for establishing the identity of CCS myocardium (as detailed earlier and in refs. 13, 19), transcript levels for these genes were increased in the Tbx3^GH/N AVC. Periostin, encoded by Postn, is expressed by the AV valve progenitor cells in the AV cushions (32). Decreased Postn transcripts (−2-fold) in Tbx3^GH/N AVC may directly result from decreased Tbx3 in AV cushion mesenchyme (included in the tissue assayed), or reflect altered signaling from AVC myocardium to valve progenitors.

We compared transcript levels of the four most dysregulated genes between the AVC and LV to examine differential regulation in these two tissues (Fig. S6C). The transcript levels of Tbx3, Kcne3, and Scn7a were lower, and Chac1 levels were higher, in WT LV vs. AVC. These relationships were preserved in Tbx3^GH/N hypomorphs, indicating that, whether Kcne3, Chac1, and Scn7a are direct or indirect targets of Tbx3 regulation, Tbx3 deficiency affects their expression similarly in both tissues.

To determine which Tbx3⁺ regions are required for normal embryonic AV conduction, we used two spatially restricted Cre lines, Mef2c-AHF-Cre and cGata6-Cre (33, 34), and our unique Tbx3 conditional floxed allele (Tbx3^flox; Fig. 1A) to generate Tbx3 conditional mutant embryos. We visualized the entire Tbx3⁺ myocardial population (Fig. 5 A–C), using the conditional Tbx3 GFP reporter allele (Tbx3^G) recombined with αMHC-Cre (35) and compared the GFP-positive domains to those detected with the mT/mG reporter (36) after Cre activity from Mef2c-AHF-Cre or cGata6-Cre. Mef2c-AHF-Cre is active in the right ventricle and outflow tract, including the ventricular septal crest (nascent AVB), but not in the AVC myocardium (nascent AV node; Fig. 5 D–F) (34, 37). A total of 75% (n = 9) of Tbx3^flox/flox;Mef2c-AHF-Cre mutants echoed from E12.5 to E13.5 had AV block or were dead, whereas all littermates had normal AV conduction (Table 3). cGata6-Cre activity is restricted to and variable within AVC myocardium (33) (Fig. 5 G–L). AV block was detected in 67% (n = 6) of Tbx3^flox/flox;cGata6-Cre embryos echoed from E12.5 to E14.5 but not in littermates (Table 3), implicating the AVC myocardium in arrhythmia pathogenesis. Normal rhythm in 33% of Tbx3^flox/flox;cGata6-Cre embryos is likely attributable to variable Cre activity (33). Thus, AV block occurs after disruption of Tbx3 from either of two distinct domains of the developing CCS: the nascent AVB or AVC myocardium, which contains the nascent AV node.

Tbx3 is expressed in adult CCS tissues (13), leading us to question whether Tbx3 has roles in maintaining structure or function of the mature CCS. To address this, we used an inducible, globally expressed Cre-recombinase (CAGGCre-ER) (38). We administered tamoxifen to Tbx3^flox/flox;CAGGCre-ER (n = 5) and littermate controls (n = 3) at 2 to 3 mo of age and monitored their cardiac rhythms before, during, and weekly for 10 wk after tamoxifen induction by using telemetric ECG. Telemetry data were mined for changes in heart rate, ECG intervals, and arrhythmias.

Rhythm abnormalities observed in adult conditional mutants included an increased incidence of second-degree AV block (Fig. 6 A and B) and sinus pauses (Fig. S7). Second degree AV block was rarely seen in control mice (Tbx3^+/+;CAGGCre-ER and Tbx3^flox/flox); they had, at most, three episodes per hour of AV block, and no increase after tamoxifen treatment. This indicates that neither tamoxifen nor the Cre genotype causes AV block. Baseline incidence of AV block or sinus pauses (before tamoxifen) was not different in Tbx3^flox/flox;CAGGCre-ER mutants compared with controls. After tamoxifen treatment, we observed an increased frequency of AV block within 1 wk, which peaked at 2 to 6 wk. The frequency of AV block decreased with time, and a statistically significant excess persisted in the mutants 9 and 10 wk after Cre induction (Fig. 6B). SAN pauses were more frequent in Tbx3^flox/flox;CAGGCre-ER compared with controls (0–16 vs. 0–6 pauses per hour).

Two pairs of adult mutant (Tbx3^flox/flox;CAGGCre-ER) and control (Tbx3^flox/flox) animals were killed 2 wk after tamoxifen treatment to assess the impact on Tbx3 production by IHC. Although Tbx3 protein was not eliminated in the Tbx3^flox/flox;CAGGCre-ER hearts, the number and density of Tbx3⁺ cells was decreased in the region of the Hcn4⁺ AV node and AVB (Fig. 6 C–F).

Three targeted mutations of Tbx3 were generated in mice: a null allele (Tbx3^N), a hypomorphic allele (Tbx3^GH), and a floxed conditional allele (Tbx3^flox). Additional alleles were generated by recombining the GH and flox alleles with FLPe or Cre-recombinase (Tbx3^G and Tbx3^Δflox, respectively). Lines were maintained on a 50% C57BL/6, 50% SV129 background for embryo and newborn studies, and crossed into FVB (25% C57BL/6; 25% SV129; 50% FVB) for adult studies. Experiments were conducted in compliance with institutional animal care and use committee standards.

Tbx3^N has a 4.6-kb deletion starting from an EcoRI site 920 bp 5′ of the translational start site (ATG) to an EcoRI site 3.7 kb 3′ of the ATG, deleting exons 1 to 3. The targeting vector also included a FRT-flanked neomycin (neo^r) selection cassette at the BglII site 3.2 kb 5′ of the ATG.

The Tbx3^GH allele contains a loxA2 site inserted into the BamHI site 5.4 kb 3′ of the ATG. An FRT-flanked hygromycin resistance gene, a second loxA2 sequence, and a splice acceptor/GFP were inserted in an Apa1 site approximately 14 kb 3′ of the ATG. The hygromycin gene was removed by using the FLPe transgene (52) generating Tbx3^G. These alleles express a Tbx3-GFP fusion protein after Cre-mediated recombination.

Tbx3^flox contains two loxP sequences inserted in parallel orientation flanking the first exon of the Tbx3 gene. Cre-mediated recombination generates the Tbx3^Δflox null allele. Tbx3^flox/flox mice survive, have normal limbs, and reproduce normally.

Transgenic Cre lines used for conditional ablation of Tbx3 included: deleter-Cre (16), Mef2c-AHF-Cre (34), cGata6-Cre (33), αMHC-Cre (35), and CAGGCre-ER (38). Activities of these Cre lines in our mice were determined by using mT/mG and Rosa26^lacZ reporter mice (36, 53). CAGGCre-ER was induced by oral gavage of 0.25 mg/g tamoxifen (Sigma Chemical) in peanut oil (10 mg/mL) daily for four doses.

Genotypes were determined by PCR of genomic DNA from ear biopsies or yolk sacs; primer sequences are available upon request. “E” refers to days of gestation beginning from midnight on the day of identification of a copulation plug.

Embryos were imaged in utero by using ultrasound biomicroscopy (Vevo 660; VisualSonics) with a 40-MHz transducer and 23-MHz spectral pulsed-wave Doppler. Heart rate and rhythm were determined from pulsed Doppler waveforms. Ventricular function was determined by fractional area change between systole and diastole from 2D images. Maternal anesthesia was achieved with 1% to 2% isoflurane, maintaining heart rate between 450 and 550 beats/min and rapid shallow breathing. Body temperature was maintained within normal range and duration of maternal anesthesia was less than 1.5 h. Hemodynamics of E10.5 to E15.5 embryos were analyzed at daily intervals. During scanning, the bladder was used as a reference point for the left and right uterine horns, and the relative location of embryos was mapped. After the final scan, laparotomy was performed, and position of the embryos correlated with ultrasound embryo location before genotyping. Newborn mice were studied awake on the morning of their birth. A short-axis view of the ventricles was obtained to determine wall and chamber dimensions.

A unique method to record embryonic ECGs used 240-mm-diameter silver electrodes positioned around the thorax of embryos exposed through abdominal and uterine incisions leaving placental circulation intact; a ground electrode was attached to the anesthetized mother. For newborns and adults, needle electrodes were inserted into proximal limb skin under anesthesia. Electrodes were connected to a DC-powered differential amplifier; signals were filtered at 50 kHz and digitized at 10 kHz with a 12-bit A/D converter (Digidata 1322A; Molecular Devices). Data acquisition/analyses were performed with pClamp software (Molecular Devices).

Long-term ECG monitoring in adults was via two lead telemetry devices (Data Sciences International) implanted intraperitoneally. In hypomorphs and controls, we ascertained 24 h of data 4 d after implantation. For adult Tbx3 conditional ablations and controls, we implanted telemetry devices at 2 mo of age; baseline data were obtained 4 d after implantation, and tamoxifen administration began 1 wk after implantation. Heart rhythm was monitored for 2 h each week for 10 wk after tamoxifen dosing. ECG data were analyzed using Ponemah Physiology Platform software (Data Sciences International).

Tissues were fixed in 4% paraformaldehyde, embedded in paraplast or frozen in optimal cutting temperature compound, and sectioned at 7 to 10 μm for immunohistochemistry (IHC) and at 10 to 14 μm for in situ hybridization. Probes and methodology of in situ hybridization were described previously (12, 22, 25, 54). Immunohistochemistry was performed with Antigen Unmasking Solution (Vector Labs). Blocking solution consisted of 5% serum (specific to species of secondary antibody) and 0.3% Triton X-100 in PBS solution. Detection was with fluorescent-tagged or biotinylated secondary antibodies [detected with Vectastain Elite ABC reagent (Vector Labs) and ImmPACT DAB (Vector Labs) with eosin or methyl green counterstains]. For fluorescent IHC, nuclei were stained using Hoechst 33342 (H1399, 1:500 in PBS solution; Molecular Probes). Antibody combinations used included: anti-PECAM1 (553370; BD Biosciences) with biotin goat anti-rat Ig (559286, 1:75; BD Pharmingen), anti-BrdU (ab6326, 1:1,000; Abcam) with biotin-SP–conjugated donkey anti-rat IgG (712–065-153, 1:250; Jackson ImmunoResearch), or Alexa Fluor 594 goat anti-rat IgG (A11007, 1:500; Invitrogen), anti-Cx43 (C6219, 1:500; Sigma) with Alexa Fluor 488 goat anti-rabbit IgG (A11008, 1:500; Invitrogen), anti-Cx43 (45256, 1:200; BD Biosciences) with Alexa Fluor 568 donkey anti-mouse IgG (1:250), anti-GFP (A11122, 1:1,000; Invitrogen) with Alexa Fluor 488 goat anti-rabbit IgG (1:500), anti-Hcn4 (APC-052, 1:500; Alomone) with biotin goat anti-rabbit Ig (550338; BD Pharmingen), anti-Hcn4 (NG164345, 1:200; Millipore) with Alexa Fluor 488 donkey anti-rabbit IgG (1:250), anti-Tbx3 (SC17871, 1:500; Santa Cruz) with biotinylated rabbit anti-goat IgG (BA-5000, 1:500; Vector Labs), anti-Tbx3 (B1006, 1:200; Santa Cruz) with biotinylated donkey anti-goat (1:250) detected with TSA Enhancement Kit (NEL702A; Perkin-Elmer), and anti-Cx40 (K2204, 1:200; Santa Cruz) with Alexa Fluor 680 donkey anti-goat IgG (1:250). Apoptosis was assayed in paraffin-embedded sections by using the ApopTag Detection Kit (Chemicon). Embryos bearing the double fluorescent mT/mG Cre reporter were cryosectioned and directly imaged under epifluorescence (36).

We assayed mRNA levels of Tbx3, Tbx2, and Tbx5 in dissected atria and AVC tissues from E12.5 embryos with normalization to hprt by qRT-PCR.

For the microarray expression analysis, we microdissected AVC, including AV cushions and myocardium, from E10.5 Tbx3^GH/N mutant and Tbx3^+/+ control embryos. Tissues were dissected in ice-cold PBS solution and stored in RLT buffer (Qiagen) at −80 °C. Five specimens of each genotype were pooled per sample. Total RNA was extracted from samples (RNeasy Micro Kit; Qiagen). The experiment was run in quadruplicate (four separate biologic replicates) on Agilent mouse whole-genome expression arrays. Agilent two-color LRILAK labeling, the Agilent two-color GE hybridization/wash protocol, and the Agilent 5-μm XDR scanning protocols were carried out the by the University of Utah Microarray Core Facility. The array image data were quantified using Agilent Feature Extraction software (version 9.5.1.1). Subtle intensity-dependent bias was corrected with LOWESS normalization, with no background subtraction. The raw and normalized data sets have been submitted to Gene Expression Omnibus database (accession no. GSE24122). Spots with intensity below background were removed before statistical analysis. Statistical analysis of normalized log-transformed data was performed in GeneSifter (www.geospiza.com).

For qRT-PCR validation, separate samples were made from tissue specimens obtained as described earlier. One hundred micrograms of total RNA was transcribed to cDNA by using the SuperScript III First-Strand Synthesis System (Invitrogen). Quantitative PCR was performed with iQ SYBR Green Supermix on the iCycler system (Bio-Rad), and normalization was to hprt, gapdh, and β-actin. Primer sequences provided upon request.

Quantitative PCR data are presented using the ΔΔC(t) method (55). For the experiments evaluating expression of levels ofTbx3 in the allelic series, ANOVA was performed with Student–Newman–Keuls testing to determine differences between individual groups. t tests were used to compare qRT-PCR measurements for Tbx3^GH/N vs. Tbx3^+/+ validation of microarray results. ECG and echocardiogram measurements were compared by using the Mann–Whitney U test or t test. Statistical analysis was performed by using SPSS Statistics (version 17.0); a significance level of 0.05 was used.