Ectopic retinoic acid signaling affects outflow tract cushion development through suppression of the myocardial Tbx2-Tgfβ2 pathway

Congenital heart diseases (CHDs) are a common human birth defect, affecting nearly 1 in 100 live births (Hoffman, 1995), and ∼30% of CHDs are due to cardiac outflow tract (OFT) malformations (Gruber and Epstein, 2004). These OFT malformations result from abnormal morphogenesis of specific components, particularly the endocardial cushion tissue, which is the primordia of the aortic septum. During OFT development, some of the endothelial cells lining the OFT region are activated by signaling molecules secreted from the adjacent myocardium to undergo the endothelial-mesenchymal transformation (EMT). Subsequently, the transformed mesenchymal cells invade the matrix-rich cardiac jelly to form conotruncal endocardial cushion tissue (Markwald et al., 1975; Sakabe et al., 2005), and then neural crest cells also migrate through the pharyngeal arches to fill the OFT cushions (Hutson and Kirby, 2007). Abnormal EMT or neural crest migration causes several congenital heart defects, such as transposition of the great arteries (TGA), double outlet right ventricle (DORV), and persistent truncus arteriosus (PTA). Therefore, understanding the mechanisms underlying cushion development is very important for clarifying the pathogenesis of CHDs.

The signals required for endocardial EMT have been investigated previously using in vivo, as well as in ex vivo culture systems. In mouse embryos, members of the Tgfβ superfamily, Tgfβ2, Bmp2 and so on, are expressed in the OFT and atrioventricular canal (AVC) myocardium and appear to be necessary for EMT, because a lack of these genes results in the loss of endocardial cushion tissue formation (Ma et al., 2005; Jiao et al., 2006). Moreover, in vitro collagen gel assays have revealed that Tgfβ2 and Bmp2 play a role in activating the endocardium to promote EMT (Camenisch et al., 2002; Sugi et al., 2004). Although both factors have similar functions as inducers of EMT, Bmp signaling also regulates the target gene expressions in the myocardium, including Tbx2. Tbx2, a member of the T-box transcription factor family, is expressed specifically in the OFT and AVC myocardium (Harrelson et al., 2004). Tbx2 is known to be an important factor during cushion formation, as the corresponding knockout mouse shows the ectopic expression of chamber-specific genes in the OFT and AVC myocardium, and also shows abnormal cushion formation (Harrelson et al., 2004). In addition to Tbx2, Tbx1, Tbx3 and Tbx20 are also involved in OFT septation. Tbx1-deficient mice show abnormalities in both OFT septation and pharyngeal arch development that are similar to those associated with human 22q11 deletion syndrome (Baldini, 2004). The mice lacking Tbx3, which is a paralog of Tbx2, showed severe OFT defects, including connection of both the aorta and pulmonary trunk to the right ventricle, aortic arch artery anomalies, and abnormal communications between the right atrium and left ventricle (Mesbah et al., 2008). Tbx20 is ubiquitously expressed in the developing heart, but the corresponding mutant mice show a complete loss of the endocardial cushion in the OFT and AVC regions (Takeuchi et al., 2005). Hence, Tgfβ signaling and/or T-box transcription factors are crucial for the endocardial cushion formation, although the molecular mechanisms underlying the endocardial EMT remain unclear.

In addition to these inherited factors, maternal environmental changes, such as virus infection, smoking, and alcohol or drug intake have also been reported to induce CHD (Zhu et al., 2009). A high concentration of maternal retinoic acid (RA) is known to be a potent teratogen that induces abnormal left-right axis formation, somitogenesis, and limb formation (Vermot and Pourquie, 2005; Sirbu and Duester, 2006; Lee et al., 2010). However, RA is also known to be essential for normal development, because animal models lacking RA signaling, such as vitamin A-deficient (VAD) rats, or mice deficient in retinoic acid receptors (Rar/Rxr) or an RA synthesis enzyme (Retinaldehyde dehydrogenase-2:Raldh2) show a variety of severe developmental defects, including heart malformations (Wilson and Warkany, 1950; Kastner et al., 1994; Mendelsohn et al., 1994; Gruber et al., 1996; Ghyselinck et al., 1998; Ryckebusch et al., 2008), thus suggesting that an appropriate amount of RA is important during heart development. Our previous studies have shown that maternally administered RA within a narrow developmental window induces the TGA phenotype in mouse embryos, suggesting that the RA affected the expression of multiple genetic factors responsible for TGA (Nakajima et al., 1996a; Yasui et al., 1997). However, the causal relationship between RA and genetic factors remains unclear at present.

The aim of our present study was to determine the molecular mechanisms underlying the normal OFT development using this RA-induced TGA model. We identified gene expression changes in the RA-treated OFT region using several microarray analyses and in situ hybridization studies, and the results showed that Tbx2 expression was greatly downregulated in the RA-treated OFT myocardium. We subsequently found that Tbx2 upregulates Tgfb2 expression, and that defective EMT caused by the excess RA was rescued by the addition of Tgfβ2 in an organ culture system. Taken together, our present data indicate that ectopic RA signaling suppresses myocardial Tbx2-Tgfβ2 pathway in the early stages of OFT cushion development, which is one of the key morphogenetic events to establish the normal ventricle-arterial connection.

ICR mice were purchased from Japan CLEA Co. (Tokyo). The generation of R26R reporter mice has been reported previously (Soriano, 1999). The RARE reporter mouse was kindly provided by Dr H. Hamada (Osaka University). The Tbx2 reporter transgenic mouse (Tbx2-lacZ) has been described previously (Kokubo et al., 2007). The generation of the TGA mice embryo has been reported previously (Nakajima et al., 1996b).

Approximately 200 outflow tract regions were isolated from control and RA-treated embryos [embryonic day (E) 9.5] and total RNA was purified using the RNeasy Mini Kit (Qiagen), according to the manufacturer's instructions. Microarray analyses were performed using GeneChip Mouse Genome 430 2.0 (Affymetrix) at Kurabo Industries (Osaka, Japan). The data discussed in this publication have been deposited in Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE33472.

The InsituPro system (M&S Instruments) was used to conduct a whole-mount in situ hybridization analysis in accordance with the manufacturer's instructions. Digoxigenin-labeled RNA probes for Tbx2, Anf (Nppa – Mouse Genome Informatics), Chisel (Smpx – Mouse Genome Informatics), Wnt11, Sema3c, Mlc2v (Myl2 – Mouse Genome Informatics) and Tgfb2 were prepared using standard methods (Roche).

Total RNA was extracted from the outflow regions of 20 mouse hearts using ISOGEN (Nippon gene). The cDNA was synthesized from 0.1 μg of total RNA, and PCR was carried out in 15 μl reaction mixtures using the One-Step RT-PCR system (Qiagen). The primers used were as follows: Tbx2 (5′-AAGCAGGCTTCAGCGGTCA-3′ and 5′-TCAAATCCAGGGATTCCAGAGTG-3′); Tgfb2 (5′-CGAGACCAAATACTTTGCCACAAAC-3′ and 5′-CCATGAAGCTTCGGCAGACA-3′); Bmp2 (5′GGGCTGATCTGGCCAAAGTACTAA-3′ and 5′-TTATGAGGGCCCACAAGATAATCAA-3′); Bmp4 (5′-TTCCTGGTAACCGAATGCTGA3′ and 5′-CCTGAATCTCGGCGACTTTTT-3′). Previously described primers were used to amplify Bmp6 (Watanabe et al., 2006), and Gapdh (Kokubo et al., 2004).

For the chromatin immunoprecipitation (ChIP) assay, an expression vector for 3×Flag-RARα or 3×Flag-RXRα was transfected into C2C12 cells using Lipofectamine LTX (Invitrogen). After 48 hours, cells were cross-linked with 10% formaldehyde for 10 min. Immunoprecipitation was performed using a SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (Cell signaling). For PCR, 1 μl of the template from 30 μl of DNA and 26 cycles of amplification were used with the promoter-specific primers. The antibody used in ChIP experiments was an anti-Flag M2 antibody from Sigma.

E9.5 embryos were fixed in 4% paraformaldehyde for 30 minutes, embedded in OCT compound, and sectioned at a thickness of 8 μm. Phospho-Smad 1/5/8 signaling was detected by incubation with a primary rabbit anti-phospho-Smad 1/5/8 antibody (Cell Signaling), followed by treatment with a secondary Alexa Fluor 594 anti-rabbit antibody (Molecular Probes). The whole-mount X-Gal staining of embryos has been described previously (Saga et al., 1992).

Tbx2 reporter constructs (Tbx2-Luc) have been described previously (Kokubo et al., 2007). A 3.8 kb genomic DNA fragment upstream of the Tgfb2 start codon was amplified with PrimeStar Polymerase (Takara), and cloned into the pGL3-basic vector (Promega). Reporter constructs (200 ng) were individually co-transfected with expression vectors for Smad5, 3×Flag-Tbx2, and the constitutively active form of Alk3 (Bmpr1a – Mouse Genome Informatics). The C2C12 cells were lysed 24 hours after transfection, and then the luciferase activities were measured using the Dual Luciferase Assay Kit (Promega).

The method of collagen gel culture has been described previously (Bernanke and Markwald, 1982). The OFT region was resected from each E9.5 mouse heart and explanted onto a collagen gel lattice. After incubation for 12 hours, the cultures were treated with or without 100 ng/ml recombinant Tgfβ2 (R&D). After 72 hours (the total incubation time), cultures were assessed for EMT under Hoffman modulation optics.

Previous studies have indicated that a high concentration of maternal RA affects embryonic development, including cardiogenesis, and Yasui et al. reported that intraperitoneal injection of all-trans RA at day 8.5 of gestation induced the TGA phenotype in mice (Yasui et al., 1995). To examine the OFT septation process in RA-treated embryos, we analyzed the embryos on E9.5-14.5. At E9.5, the outflow tracts in RA-treated embryos were found to be reduced in size compared with those of control embryos treated with DMSO only (Fig. 1A,B, arrows in 1A′,B′). At E10.5, the endothelial cells in the proximal (conus) OFT region of the control embryos were transformed into mesenchymal cells and invaded the cardiac jelly, resulting in the formation of the OFT cushion tissue (Fig. 1C, arrows in 1E). By contrast, only a few migrating mesenchymal cells were found in the proximal OFT region of RA-treated embryos (Fig. 1D, arrows in 1F).

As OFT defects are also induced by aberrant cardiac neural crest cell migration, we investigated whether cardiac neural crest cell migration occurred normally in the RA-treated embryos. As shown in Fig. 1E,F, cardiac neural crest cells (arrowheads) were observed in the distal (truncal) region of the control and RA-treated outflow. These results were also confirmed by the neural crest cell tracing experiment using Rosa26 reporter (R26R) mice crossed with Wnt1-cre mice (Fig. 1K-L), thus suggesting that cardiac neural crest cell migration still occurred in RA-treated embryos. To further explore the inhibition of endocardial EMT, we traced the endothelial-derived cells in the OFT region using R26R mice crossed with Tie2-Cre mice. This tracing experiment clearly showed that the endothelial-derived mesenchymal cells (lacZ-positive cells) were absent in the RA-treated OFT region, whereas lacZ-positive cells were observed in the control OFT (Fig. 1G,G″,H,H″). By contrast, lacZ-negative mesenchymal cells (neural crest cells) were consistently observed in both control and RA-treated embryos (Fig. 1G,G′,H,H′) indicating that the endocardial EMT, but not the migration of neural crest cells, is affected by excess RA during OFT development. It should be noted that the endocardial EMT in the AVC region was observed to be normal in both types of embryos (Fig. 1I,J,L).

An ink injection analysis of E14.5 hearts demonstrated that the aorta (Ao) connected to the left ventricle (LV), and the pulmonary artery (Pa) connected to the right ventricle (RV), in control embryos (Fig. 1M). By contrast, the RA-treated embryos showed the opposite connections, namely that the Ao connected to the RV, and the Pa connected to the LV, which closely resembles human TGA (Fig. 1N). These observations indicate that a single dose of maternal RA induces the TGA phenotype, in which the OFT shortening and the reduction of endocardial-derived mesenchyme in the proximal OFT cushions are evident.

We next examined whether the excess RA could affect the RA signaling activity in the embryonic heart using a sensitive reporter for RA-mediated transcriptional activation (the RARE-lacZ mouse line) (Rossant et al., 1991). The RARE-lacZ transgene contains a multiple RA response element (RARE) coupled to a basal promoter, and this reporter reflects only RA signals. In E8.5 embryos, just before RA injection, lacZ expression could be observed in the somitic paraxial mesoderm, but not in the heart-forming regions (Fig. 2A). In E8.75 and E9.5 control embryos (exposed to DMSO on E8.5), the heart appeared to be undergoing normal development, and reporter gene expression was not detectable in the developing heart region (dotted line in Fig. 2B, arrow in 2D). However, a strong lacZ expression was observed in the developing heart at 6 hours after RA injection (E8.75; dotted line in Fig. 2C), and this ectopic expression persisted throughout the OFT cushion-forming stage. In addition, the ectopic lacZ expression was found to be specifically accumulated in the OFT region, indicating its involvement in abnormal OFT development (Fig. 2E, arrowhead). These observations suggest that ectopic RA signaling affected cushion-related gene expression during heart development.

Given our observation that RA treatment alters the RARE-reporter activity, we hypothesized that ectopic RA signaling changed the expression of genes that are crucial to OFT cushion development. To identify the candidate genes involved in the altered development of the OFT in the RA-induced TGA model, we isolated the RNA from E9.5 control and RA-treated OFT regions, and performed a microarray analysis. In this screening, we focused on the changes in the expression of transcription factors, and found that Tbx2 was largely downregulated in the RA-treated OFT region (Fig. 3A). To confirm the downregulation of the Tbx2 gene, we then performed an in situ hybridization analysis using E9.5 control and RA-treated embryos. Tbx2 expression was detectable in the developing OFT region of control embryos as described previously (Camenisch et al., 2002; Harrelson et al., 2004; Zhou et al., 2007) (arrow in Fig. 3B), whereas this signal was not observed in the RA-treated OFT region (Fig. 3C, arrowhead).

To explore the outcomes of the downregulation of Tbx2, we examined the expression of chamber-specific genes, such as Anf (Nppa) and Chisel, in the OFT regions, because previous reports have shown that loss or downregulation of Tbx2 facilitates the expression Anf and Chisel in OFT regions (Harrelson et al., 2004). Reduced expression levels of Anf and Chisel were normally found in the control OFT region, whereas ectopic expression of Anf and Chisel was detected in the RA-treated OFT region by in situ hybridization analysis, as observed in the Tbx2-null mouse (Anf: arrow and arrowhead in Fig. 3D,E; Chisel: data not shown). As these results might suggest the presence of abnormal cardiac patterning in the OFT region, we analyzed the expression of region-specific marker genes, such as Wnt11 and Sema3c for the OFT, and Mlc2v for the ventricular myocardium (Fig. 3F-K). The expression boundary of these markers was not affected by RA treatment, indicating that the normal transcriptional program within the OFT myocardium was disrupted by the downregulation of Tbx2, without affecting the regional identity.

To examine whether the RA-induced Tbx2 downregulation preceded the abnormal OFT development, we analyzed the temporal expression patterns of Tbx2 in the control and RA-treated hearts from E8.5 to E9.0 by in situ hybridization. At E8.5, Tbx2 was marginally expressed (or not detected) in the developing OFT region (arrow in the top panel of Fig. 3L). At 6 hours after DMSO or RA injection (E8.75), Tbx2 expression was detectable in the DMSO-treated OFT region, but seemed to be downregulated in the RA-treated OFT region (arrow and arrowhead in the middle panel of Fig. 3L). At 12 hours after RA injection (E9.0), the downregulation of Tbx2 was more clearly observable compared with the control embryo in the developing OFT region (arrow and arrowhead in the bottom panel of Fig. 3L). No phenotypic changes were observed between the control and RA-treated embryonic hearts from E8.5 to E9.0, suggesting that the downregulation of Tbx2 occurs before the onset of abnormal OFT development.

To confirm that Tbx2 transcriptional activity is downregulated by ectopic RA signaling, the Tbx2 transcriptional activity was visualized using a lacZ-reporter transgenic mouse line (Tbx2-lacZ), which contains the 6 kb upstream region of the Tbx2 gene and mimics endogenous Tbx2 expression in the OFT region (Fig. 4A, arrow), the AVC region (arrowhead in Fig. 4C), and the eyes, as previously reported (Kokubo et al., 2007). In the RA-treated transgenic mouse embryos, lacZ expression was detectable in the AVC (Fig. 4D, arrowhead) and in the eyes at the same level as in the controls, but no signal was observed in the OFT region (Fig. 4B, arrow), suggesting that the downregulation of Tbx2 occurred at the transcriptional level, and that the 6 kb upstream region is sufficient to respond to ectopic RA signaling. These observations gave rise to a question about how Tbx2 transcriptional activity is influenced by ectopic RA signaling. To address this issue, we tested the possibility that the downregulation of Tbx2 transcriptional activity occurs as a consequence of the suppression of Bmp-Smad signaling by ectopic RA signaling, because Tbx2 expression is known to be upregulated by Bmp-Smad signaling (Yamada et al., 2000; Shirai et al., 2009; Singh et al., 2009). However, phosphorylated Smad 1/5/8 was observed to be normal in the nuclei of the OFT myocardial and endocardial cells in both the control (Fig. 4E-E″) and RA-treated mice (Fig. 4F-F″), thus indicating that ectopic RA signaling has little effect on Smad phosphorylation.

We next hypothesized that the Tbx2 transcriptional activity might be directly suppressed by ectopic RA signaling. As RA generally exerts its function by associating with retinoid nuclear receptors, such as the RAR and RXR, which can form heterodimers and bind to RAREs to regulate target gene expression, we searched for a RARE in the 6kb upstream region of the Tbx2 gene. The RAREs consist of direct repeats of AGGTCA-like sequences, each separated by five bases (Perissi and Rosenfeld, 2005). We found a single RARE in the 210-bp downstream region of the Tbx2 transcriptional start site, a region that is highly conserved among mammals and birds (Fig. 4G). To confirm whether this RARE sequence is functional for RA receptors, we performed a ChIP assay with ectopic Flag-tagged RARα and RXRα in C2C12 cells. The region containing the RARE on the Tbx2 promoter region was amplified by PCR using an anti-Flag antibody, indicating that the RA receptors bound to this RARE consensus site (Fig. 4H). To evaluate whether the Tbx2 suppression occurs via the RARE site, we generated several Tbx2 reporter mouse lines harboring a mutated RARE (AGCCCA instead of AGATCA), as shown in Fig. 4K. Surprisingly, in these transgenic lines, the reporter gene expression was not detectable in the OFT region, and it appeared to decrease in the AVC region, even in the non-RA-treated (control) embryos (Fig. 4I, arrow; 4J, arrowhead), thus suggesting that the transcriptional regulation of Tbx2 through the RARE is required for the normal development of the OFT.

We further tested the possibility that the suppression of Tbx2 transcription by excess RA was dependent on the RARE using a luciferase reporter assay system in the C2C12 cell line. As previously reported, the expression of a Tbx2 reporter gene containing the 6 kb upstream region of Tbx2 (Tbx2-Luc, Fig. 4K) was found to be upregulated via the co-transfection of a constitutively active form of ALK3 (caALK3) and Smad5 (Kokubo et al., 2007) (blue bar in Fig. 4L). This transcriptional activity was downregulated in a dose-dependent manner by the addition of RA (Fig. 4L, blue bar) similar to that observed in the Tbx2-lacZ reporter mouse. However, the transcriptional activity of the Tbx2 reporter construct with a mutated RARE (Tbx2-Luc-mRARE, containing the same mutated RARE sequence used for transgenic mouse production) was augmented by phospho-Smad5 (Fig. 4K,L). Of interest, this promoter activation was unaffected by the addition of high levels of RA (Fig. 4L, red bars), thus indicating that ectopic RA signaling directly suppresses the Tbx2 promoter activity through a functional RARE site. Therefore, our data strongly suggest that RA signaling is involved in the regulation of Tbx2 promoter activity during OFT development.

Based on the current experiments, we hypothesized that Tbx2 downregulation is the primary cause of endocardial cushion malformation due to excess RA. Although the expression of Tbx2, which is known as a transcription factor, is detectable only in the myocardium during OFT development (Harrelson et al., 2004), it seems unlikely that Tbx2 directly regulates the endocardial EMT. Therefore, we hypothesized that Tbx2 could regulate the endocardial EMT through regulation of myocardium-derived secretory molecules, such as those of the Tgfβ superfamily. According to the RT-PCR analyses using RNA isolated from control and RA-treated OFT regions, we found that Tgfb2 expression was significantly downregulated, similar to the Tbx2 expression in the RA-treated OFT region, but that other genes such as Bmp2, -4, and -6 were unaffected by RA treatment (Fig. 5A). An in situ hybridization analysis also showed that Tgfb2, which was normally expressed in the OFT myocardium, was hardly detected in RA-treated embryos (Fig. 5B,C). As the expression of Bmp2 was not detected in either control or RA-treated OFT regions by the in situ hybridization analysis (Fig. 5D,E), we hypothesized that Tbx2 could directly regulate Tgfb2 expression during OFT cushion development. Using a luciferase reporter assay with a 3.7-kb upstream region of Tgfb2, we demonstrated that upregulation of its reporter activity (1.5-fold, P<0.001) occurred in the presence of Tbx2. Because a GATA binding site was found in the upstream region of Tgfb2, in addition to the T-box binding sites (Fig. 5F), we tested whether the GATA factor was involved in its transcriptional activity. In the presence of Tbx2 and Gata4, we found that there was significant upregulation of the reporter activity (2.7-fold: P<0.001, Fig. 5F, left), the activation of which was depleted by deletion of the binding sites in the reporter construct (Fig. 5F, right), thus suggesting that Tbx2 plays a role as a positive regulator of Tgfb2 expression, and that this occurs synergistically with Gata4 in the developing OFT myocardium.

We then examined whether Tgfβ2 is capable of recovering the endocardial EMT in the RA-treated OFT region using three-dimensional explant culture system. To avoid the inclusion of neural crest cells in the explants, we performed the culture experiment using OFT from E9.5 embryos. OFT explants from control or RA-treated embryos were cultured on collagen gel and incubated with or without recombinant Tgfβ2. After 72 hours, RA-treated OFT explants (RA-OFT) showed endothelial outgrowth on the gel surface, but mesenchymal cells were never seeded into the gel lattice (approximately five to ten mesenchymal cells/explant invaded the collagen gel lattice; Fig. 5H,J). By contrast, when RA-treated OFT explants were cultured with 100 ng/ml recombinant Tgfβ2 (RA-OFT + rTgfβ2), the endothelial cells showed phenotypic changes, and many mesenchymal cells were found in the gel lattice, similar to the control explants (∼35-45 mesenchymal cells/explant; Fig. 5G,I. arrows). In addition, the number of migrating mesenchymal cells significantly increased in the cultures treated with recombinant Tgfβ2 (P<0.001, Fig. 5J), thus suggesting that RA-induced abnormal endocardial EMT can be rescued by the addition of rTgfβ2. In addition, the preferential effect of RA administration on EMT in OFT region but not in AVC region, as seen in section analysis (Fig. 1G-L), had been demonstrated previously by Nakajima et al. (Nakajima et al., 1996b) using the same culture system. Together, these data suggest that the Tbx2-Tgfβ2 pathway regulates the EMT process of OFT region and is also a key component of RA-induced TGA.

To evaluate the RA teratogenicity during heart development, many animal experiments have been performed, and these observations suggested that the teratogenic effect of RA can be attributed to its suppression of myocardial inductive molecules, such as type I collagen or laminin, in the developing heart (Mahmood et al., 1992; Nakajima et al., 1996b), although the molecular mechanisms underlying the abnormal expression of these myocardial molecules was unclear. In our present study, we identified the transcription factor, Tbx2, to be a functional regulator of OFT myocardium-derived molecules. Tbx2 acts synergistically with Nkx2.5 as an inhibitor of chamber-specific genes, such as Anf in the OFT and AVC regions, and also functions as an upregulator of Tgfb2 through its activation domain within the T-box during the endocardial EMT process (Habets et al., 2002; Paxton et al., 2002; Harrelson et al., 2004; Shirai et al., 2009). These observations strongly suggest that the RA-induced abnormal OFT cushions would be derived from suppression of the Tbx2-TGFβ2 pathway. However, neither the single mutant mice of Tbx2 or Tgfb2 have exhibited TGA morphology, indicating that other molecules are required for inducing TGA. Downregulation of Wnt11, which is also known to regulate Tgfb2 expression, in the OFT region was detected in our in situ hybridization analysis, as shown in Fig. 3F,G, although the RA responsive element was not found in the upstream region of the Wnt11 gene, indicating that Wnt11 expression might not be directly suppressed by excess RA. As many of the other genes were listed as being affected by RA treatment according to our microarray analysis, we speculate that other altered conditions in addition to the Tbx2-TGFβ2 pathway would be necessary to establish TGA phenotype. Further analyses are necessary to elucidate the molecular mechanism(s) underlying the RA-induced alterations in OFT.

It has been thought that conotruncal defects can be categorized as TGA, DORV, tetralogy of Fallot, and PTA, all of which are induced by abnormal development of the OFT, such as hypoplastic cushion ridges, altered/arrested rotation of the conotruncus, and misalignment of the septal components (Nakajima, 2010). Adequate numbers of both endothelium-derived mesenchymal cells and neural crest cells migrating into the OFT are essential for normal OFT development. According to observations of the morphology seen in the RA-induced TGA and other mutant mice with conotruncal defects, we predict that the abnormal cushion formation is induced by different components affecting either endocardium-derived cushion mesenchyme or cardiac neural crest cells, and thus resulting in distinct conotruncal malformations. In fact, in Pax3 mutants, the embryonic phenotype might vary according to the extent of reduction of neural crest cells incorporated into the OFT; a large reduction results in PTA, whereas a small reduction results in DORV (Epstein et al., 2000). In our RA-induced TGA model, although endocardium-derived mesenchymal cells are sparse in the proximal cushion ridges and thus proximal cushion ridges appear to be hypoplastic, cardiac neural crest-derived mesenchymal cells are normally distributed in the distal cushion tissues (Fig. 1). In the Perlecan-null heart, the hyperplastic proximal cushion ridges perturb the appropriate alignment between the left ventricle and aortic route; thus the right ventricle to aortic connection is established (Costell et al., 2002). In addition to the hyperplastic or hypoplastic proximal cushion tissues, stunting and rotational defects of the OFT, which may transpose the aorta to the right anterior (ventral) region of the pulmonary trunk, are observed in RA-induced TGA and Pitx2c mutant hearts (Nakajima et al., 1996a; Yasui et al., 1997; Bajolle et al., 2006). However, laterality defects, including an l-looping heart, were never observed in our RA-induced TGA mouse model. From these observations, we also speculated that RA administration might affect differentiation or contribution of second heart field (SHF). Although our microarray experiment used RA-treated OFT material and showed no significant changes in genes expressed in SHF-derived cells, such as Fgf8, Fgf10, Foxc1, Foxc2 and Mef2c, we cannot exclude the possibility that the abnormal SHF-cell lineage differentiation might have contributed to the TGA morphology. Therefore, further studies, such as lineage-tracing analysis using Islet1-Cre mice, are required to determine the involvement of SHF cells in stunting and rotational defects of the RA-treated OFT region. Taken together, these observations strongly suggest that the hypoplastic proximal cushion ridges in association with shortening and rotational defects of the OFT are the part of etiology leading TGA morphology in RA-treated embryos. It remains to be investigated how RA affects the morphogenesis of the OFT, including cushion formation, septation, elongation/rotation and retraction of the conotruncus (Okamoto et al., 2010).

We found in our current experiments that the RA-induced TGA phenotype is associated with the hypoplastic cushion tissue formation, which is observed only in the OFT region, and not in the AVC region (Nakajima et al., 1996b; Yasui et al., 1997). We also found the downregulation of Tbx2 expression or transcriptional activity occurred only in the OFT region, but not in the AVC myocardium. This surprising correlation strongly suggests that the downregulation of Tbx2 is the main cause of the hypoplastic cushion formation. However, it remains unclear as to why ectopic RA signaling affects only the OFT myocardial expression of Tbx2.

There are two possible explanations for this region-specific regulation. First, the timing of RA treatment should be considered. As the Tbx2 expression in the AVC region occurs earlier than that in the OFT region (Habets et al., 2002), the administration of RA at E8.5 might be too late to suppress Tbx2 expression in the AVC region. In fact, previous data indicated that abnormal AVC endocardial EMT was also induced by excess RA when embryos were treated with RA at E6.5 (Yasui et al., 1998). A second possibility is that different sensitivities to RA exist among different heart-forming lineages. In general, the SHF contributes to the OFT, but not the AVC myocardium (Harvey, 2002; Laugwitz et al., 2008). Recent studies using RA synthetase (Raldh2) or RA receptor knockout mice have indicated that SHF but not first heart field (FHF) formation is affected in RA-defective mutant embryos (Vermot et al., 2003; Ryckebusch et al., 2008; Li et al., 2010). Our results also showed that the RARE-reporter gene in RA-treated embryos was more highly expressed in the OFT region compared with other regions of the developing heart (Fig. 2E). Moreover, our unpublished data showed that Tbx2 expression was also detected, but slightly reduced, in the AVC region, even in the embryos that were treated with RA at E6.5. Together, these data strongly support our hypothesis that the cardiomyocytes derived from the SHF are highly sensitive to maternal RA. Further analyses will be necessary to elucidate whether the forced expression of Tbx2 in an OFT myocardium-specific manner can rescue the abnormal cushion formation induced by RA.

We have shown in our present experiments that ectopic RA signaling directly suppresses Tbx2 transcriptional activation. Recently, it has been reported that the expression of Tbx2 in the OFT and in the AVC myocardium is regulated by Bmp/Smad signaling (Yamada et al., 2000). Although Bmps and phosphorylated-Smad1/5/8 were observed in the present study, Tbx2 was not detected in the RA-treated OFT myocardia (Fig. 4A-F″). This finding suggests that factors other than Bmp signaling regulate the Tbx2 expression, and that the RA signaling might participate in the Tbx2 transcriptional machinery. In this regard, the retinoid receptors should be considered. It is already known that heterodimeric retinoid receptors, such as RAR/RXR, are constitutively bound to RARE sequences even in the absence of RA signaling, and that these receptors recruit transcriptional co-factors, such as histone acetyltransferase (HAT) or histone deacetylase (HDAC), which can regulate gene expression through the alteration of the chromatin structure (Perissi and Rosenfeld, 2005). The activation of these RA receptors might induce the replacement of the associated co-factors, thereby switching the target gene expression. Moreover, it has been shown that RAR/RXR-recruited co-factors can interact with phosphorylated Smads, thus resulting in the formation of transcriptional repressor complexes that lead to a negative functional interaction between RA and Smad signaling during chondrogenesis (Zhang et al., 2009).

We therefore speculate that the activated RAR/RXR in an RA-treated embryonic heart might recruit co-repressors or phosphorylated Smads, which then induces the repression of Tbx2 transcription (Fig. 6). Moreover, our in vivo observations using RA reporters with a mutated RARE sequence showed no reporter gene expression even in the control embryos (Fig. 4I,J), thus indicating that even non-activated RAR/RXR might recruit co-activators, and these co-activators are necessary to initiate Tbx2 transcription synergistically with the Smads during normal OFT development (Fig. 6). Indeed, Rar/Rxr KO mice exhibit abnormal heart development, including TGA (Pan and Baker, 2007). Taken together, our present data indicate that the RAR/RXR might regulate Tbx2 expression through modulation of the epigenetic state of the Tbx2 promoter region during OFT development. Further experiments will be necessary to identify the co-factors recruited by non-activated or activated RA receptors.

We thank M. Yanagisawa and A. McMahon for providing the Tie2-Cre and Wnt1-Cre transgenic mice. We also thank Dr Osamu Nakagawa at Nara Medical University for critical discussions. We also thank Mr Ken Inada and Genki Satoh at Nara Medical University for their technical assistance.