The TGFβ type II receptor plays a critical role in the endothelial cells during cardiac development



TGFβ signalling is required for normal cardiac development. To investigate which cell types are involved, we used mice carrying a floxed Type II TGFβ receptor (Tgfbr2fl) allele and Cre-lox genetics to deplete this receptor in different regions of the heart. The three target tissues and corresponding Cre transgenic lines were atrioventricular myocardium (using cGata6-Cre), ventricular myocardium (using Mlc2v-Cre), and vascular endothelium (using tamoxifen-activated Cdh5(PAC)-CreERT2). Spatio-temporal Cre activity in each case was tracked via lacZ activation from the Rosa26R locus. Atrioventricular-myocardial-specific Tgfbr2 knockout (KO) embryos had short septal leaflets of the tricuspid valve, whereas ventricular myocardial-specific KO embryos mainly exhibited a normal cardiac phenotype. Inactivation of Tgfbr2 in endothelial cells from E11.5 resulted in deficient ventricular septation, accompanied by haemorrhage from cerebral blood vessels. We conclude that TGFβ signalling through the Tgfbr2 receptor, in endothelial cells, plays an important role in cardiac development, and is essential for cerebral vascular integrity. Developmental Dynamics 239:2435–2442, 2010. © 2010 Wiley-Liss, Inc.


The mammalian heart develops in a complex series of co-ordinated steps, involving differentiation of cardiac progenitor cells, specification and septation of the cardiac chambers, and formation of valves (Olson, 2006; Srivastava, 2006). TGFβ signalling has been shown to be important at an early stage in cardiac development, with most previous investigations focussing on its role during formation of the cardiac cushions. These cushions, which develop between embryonic day (E) E9.5 and E11.5 in the mouse, are essential precursors of the mature valves, and intimately involved in septation. It is well known that TGFβ signalling promotes endothelial to mesenchymal transition (EMT), and subsequent migration of mesenchymal cells into the cardiac jelly (Armstrong and Bischoff, 2004). Consistent with this role, it has been shown that loss of the TGFβ2 ligand is associated with deficient ventricular septation (Sanford et al., 1997; Bartram et al., 2001), and that TGFβ1 mutant mice exhibit disorganised valves, albeit only in certain genetic backgrounds (Letterio et al., 1994). More recent advances in mouse genetics have permitted in vivo analysis of TGFβ signalling during cardiac development in a cell type–specific manner, and have confirmed the importance of the main type I and type II TGFβ receptors (Jiao et al., 2006; Sridurongrit et al., 2008). Overall, evidence produced to date suggests a critical role for these TGFβ receptors in endothelial cells (ECs), a role that is required to promote formation of the cardiac cushions (Sridurongrit et al., 2008).

During TGFβ signalling, a TGFβ ligand binds first to the TGFβ type II receptor (TGFBR2) at the cell surface, causing it to activate, by phosphorylation, an associated TGFβ type I receptor, TGFBR1, also known as ALK5. This, in turn, phosphorylates SMAD2 and SMAD3 proteins, enabling their migration to the nucleus (in association with SMAD4) to regulate transcription of TGFβ-responsive genes. As TGFBR2 is one of the key receptors required in this pathway, we chose to investigate the role of TGFβ signalling in the developing heart by inactivating this receptor using a previously described floxed Tgfbr2 mouse (Leveen et al., 2002). Our goal was to use a Cre-lox approach to deplete Tgfbr2 in a tissue-specific manner. The selection of target tissues was based on a combination of available Cre lines, and the known spatio-temporal expression patterns of the three TGFβ ligands (TGFβ1, 2, 3) and the main TGFβ receptor proteins (Tgfbr1&Tgfbr2) during cardiac development. TGFβ1 is expressed in the endocardium from E8 (Akhurst et al., 1990), while the TGFβ2 ligand is expressed in the myocardium of the atrioventricular (AV) canal at E10 (Camenisch et al., 2002; Molin et al., 2003). TGFβ3 is not expressed until E11, and is limited to epicardium and the mesenchymal cells in the cardiac cushions (Camenisch et al., 2002; Molin et al., 2003). The Tgfbr2 receptor is expressed in the primitive heart tube at E8, and in myocardium, cardiac endothelial cells, and the cardiac cushions, from approximately E10 (Roelen et al., 1994; Wang et al., 1995; Mariano et al., 1998; Mummery, 2001). Co-localisation of Tgfbr1 and Tgfbr2 has been observed in the developing endocardium and myocardium (Lawler et al., 1994) and a knock-in mouse that carries a lacZ reporter in the Tgfbr1 gene has shown Tgfbr1 expression in cardiac myocytes and trabecular muscles at E13.5 (Seki et al., 2006). In addition, phosphorylated Smad2 (pSmad2) is present in developing ventricular myocardium, consistent with, although not exclusive to, active TGFβ signalling in this tissue (de Sousa Lopes et al., 2003).

The high levels of TGFβ2 ligand expression in the myocardial cells of the AV canal at E10 (Camenisch et al., 2002; Molin et al., 2003) indicate that TGFβ signalling may be occurring in this region during this stage of heart development. We, therefore, sought to determine the role of the Tgfbr2 receptor in this tissue using the cGata6-Cre mouse, which drives Cre expression from a Gata6 enhancer sequence in the AV canal myocardium from E8.5 (Davis et al., 2001). Using this Cre line, the BMP type I receptor, Alk3, has previously been shown to be important for the development of the cardiac valves and conduction tissues (Gaussin et al., 2002, 2005). However, the role of TGFβ signalling in these processes has not yet been determined. Furthermore, in light of evidence that Tgfbr2 is important in endothelial cells during cardiac development (Jiao et al., 2006), and to bypass the embryonic lethality at E10.5 to E11.5 seen in the Jiao et al. (2006) study, we used the tamoxifen-inducible Cdh5(PAC)-CreERT2 line (Mahmoud et al., 2010). This permitted us to control the timing of endothelial Cre activity, and to inactivate Tgfbr2 in ECs after E10.5. Finally, following the unexpected finding that loss of Tgfbr2 expression in atrial and ventricular myocardium generated only a mild phenotype in cTnT-Cre/Tgfbr2fl/fl mice (Jiao et al., 2006), we sought to verify this phenotype using an independent ventricular myocardial-specific Cre line. For this experiment, we chose the Mlc2v-Cre mouse, in which Cre has been knocked into the Mlc2v locus, an early ventricular-restricted marker (Chen et al., 1998). Using this Cre-lox approach, we aimed to analyse the role of TGFβ signalling in the atrioventricular canal and ventricular myocardium, and in the endothelium, during cardiac development.


To establish that Cre was expressed in the expected cell-specific locations in the heart, we took advantage of the Rosa26R line (R26R), which expresses lacZ in cells containing active Cre protein, as well as in their descendant daughter cells (Soriano, 1999). Monitoring LacZ expression in cGata6-Cre/R26R mice at different stages of cardiac development, we observed a similar spatio-temporal pattern of expression to that previously reported for this transgenic line. cGata6-Cre is first expressed as the atrioventricular canal begins to form at E8. Lineage tracing experiments have shown that this tissue contributes to valvar leaflet and conduction tissues in adult hearts (Davis et al., 2001; Gaussin et al., 2005). In cGata6-Cre/R26R embryos, lacZ was expressed in the AV canal myocardium, and can be clearly seen at E10.5 (Fig. 1A,B). In addition, we observed that cGata6-Cre activated LacZ expression occurred in the atrial vestibular musculature supporting the leaflets of the tricuspid valve in adult hearts (Fig. 1C). When Tgfbr2 was depleted in AV canal myocardium and its derivatives in cGata6-Cre/Tgfbr2fl/fl embryos, the septal leaflet of the tricuspid valve was significantly shorter in mutants than in controls at E17.5 (Table 1, Fig. 1D–F). The average length of the septal leaflet in the mutants was 167 μm, whereas the average length in control mice was 238 μm (P = 0.007) (Fig. 1F). Analysis of Mendelian ratios (not shown) demonstrated there was no loss of viability in cGata6-Cre/Tgfbr2fl/fl embryos or adults, suggesting this abnormality had no major detrimental effects on cardiac function.

Figure 1.

cGata6-Cre expression and phenotype of cGata6-Cre/Tgfbr2fl/fl embryos. A,B: cGata6-Cre/R26R embryo at E10.5 shows LacZ expression (blue) in AV canal myocardium (arrows) in whole mount view (A) and in a transverse section through the atrioventricular cushion tissue (B). C: cGata6-Cre is expressed in the atrioventricular node (outlined in red) and the overlying atrial vestibular myocardium (black arrows) at the orifice of the tricuspid septal leaflet in an adult heart. D,E: Examples of transverse heart sections of cGata6-Cre/Tgfbr2fl/fl (E) and control (D) embryos at E17.5 to show the shorter tricuspid septal valve leaflet in the mutant heart compared with the age-matched control. Leaflet lengths were measured, as indicated by the green arrows, from the hinge point of the leaflet to the distal tip through multiple serial sections of the heart, taking care to make the measurement in between the sites of insertion of the tendinous cords to the free edge of the leaflet. F: Statistical analysis of the length of the tricuspid septal leaflet through multiple serial sections of the heart of Gata6-Cre/Tgfbr2fl/fl mutant (n = 8) and control embryos (n = 8) showed this leaflet was significantly shorter in mutant than in control hearts. *P < .05. cm, cushion mesenchyme; lv, left ventricle; ra, right atrium; rv, right ventricle; tsl, tricuspid septal leaflet; v, ventricle.

Table 1. Summary of Embryonic Cardiac Defects in Tissue Specific Knockouts of Tgfbr2
GenotypeTarget TissueAge of analysisPhenotypeFrequency
cGata6-Cre/Tgfbr2fl/flAV myocardiumE17.5Short septal leaflet of tricuspid valve8/8 (100%)
Mlc2v-Cre/Tgfbr2fl/ΔVentricular myocardiumE14.5Common AV valve, common arterial trunk, hypoplastic left ventricle.1/10 (10%)
Perimembranous VSD and overriding aorta1/10 (10%)
Normal8/10 (80%)
VE-Cadherin(PAC) CreERT2 Tgfbrfl/ΔEndothelial/endocardial post E11.5E14.5Slit-like perimembranous VSD2/13 (15%)
Perimembranous VSD7/13 (54%)
Perimembranous VSD and overriding aorta2/13 (15%)
Perimembranous VSD, overriding aorta and double outlet right ventricle1/13 (7%)
Normal1/13 (7%)

As the Mlc2v-Cre line was generated using a knock-in approach, Cre expression was expected to recapitulate endogenous Mlc2v expression, which occurs in a ventricularly restricted manner from E8 (Chen et al., 1998). Cre activity was examined using the R26R transgene. Although expression was mainly restricted to the ventricles, it occurred in a patchy heterogeneous pattern (Fig. 2A,B). The phenotypes of the Mlc2v-Cre/Tgfbr2fl/Δ embryos, examined at E14.5, were also variable, with only two out of the 10 embryos analysed showing an abnormal phenotype (Table 1). In one of the affected embryos, there was a perimembranous ventricular septal defect (VSD), but the other had more severe defects, including a common atrioventricular valve, a common arterial trunk, and a hypoplastic left ventricle (Fig. 2C).

Figure 2.

Mlc2v-Cre expression and phenotype of Mlc2v-Cre/Tgfbr2 fl/Δ embryos. A,B: Mlc2v-Cre is expressed in ventricular myocardium seen in whole mount view at E9.5 (A) and in transverse heart section at E16.5 (B) of Mlc2v-Cre/R26R embryos. Note that “patchy” lacZ expression is evident in the ventricular myocardium (B). C,D: Heart of one abnormal Mlc2v-Cre/Tgbfr2 fl/Δ embryo at E14.5 with a hypoplastic left ventricle and common arterial trunk (asterisk) over-riding the interventricular septum (C). An age-matched control heart is shown for comparison (D). Abbreviations are the same as for Figure 1; a, atrium; ivs, interventricular septum; la, left atrium.

Following tamoxifen activation of the Cdh5(PAC)-CreERT2 /R26R line at E11.5, lacZ expression was seen in the endothelial cells lining the valves and coronary vessels, as well as throughout the endocardium (Fig. 3A–C). Endothelial-to-mesenchymal transition is almost complete in the cushions by the time CreERT2 was activated at E11.5, and consequently the cushion mesenchyme shows low levels of lacZ expression (Fig. 3A,D). We expect, therefore, that Cdh5(PAC)-CreERT2 /Tgfbr2 fl/Δ mice treated with tamoxifen from E11.5 will maintain TGFβ signalling capacity in cushion mesenchyme, whilst suffering a major loss of activity in endothelial cells throughout the cardiovascular system. The mutant embryos (Cdh5(PAC)-CreERT2 /Tgfbr2 fl/Δ) did not survive beyond E15.5. Because of this, we analysed the cardiac morphology 24 hr earlier, at E14.5. In 12 of the 13 hearts examined, we found perimembranous ventricular septal defects, which varied in size from small “suture line” defects in two embryos (Fig. 4A,B) to large defects, found in 10 out of 13 mutants (Fig. 4C,D). In embryos with the small suture line defects, there was no evidence of fusion between adjacent masses of cushion mesenchyme (Fig. 4A,B). Of the 10 embryos with large perimembranous ventricular septal defects, three also had an overriding aorta, including one with a grossly abnormal heart with a double outlet right ventricle (Table 1). In addition, all mutant embryos (13/13) had severe cerebral haemorrhage, readily visible externally at E14.5 (Fig. 5B). Examination of transverse sections through the head revealed extensive cerebral haemorrhage (Fig. 5D).

Figure 3.

Cdh5(PAC)-CreERT2 expression. A–D: Transverse sections of X-Gal-stained Cdh5(PAC)-CreERT2/R26R hearts at E15.5 following tamoxifen treatment. A: LacZ is expressed in the valve leaflet endocardium (red arrows) whilst the valve cushion mesenchyme (cm) is largely lacZ negative. B, C: LacZ is strongly expressed in coronary vessels (green arrows, B) and trabecular endocardium (arrowheads, C). D: The cushion mesenchyme (cm) in the outflow tract is negative for lacZ expression.

Figure 4.

Cardiac phenotype of Cdh5(PAC)-CreERT2/Tgfbr2 fl/Δ embryos. A–D: Transverse sections through Cdh5(PAC)-CreERT2/Tgbfr2 fl/Δ hearts at E14.5 showing perimembranous ventricular septal defects. The boxed areas in A and C are shown at high power in B and D, respectively. A small “suture line” VSD is shown in A and B (arrow) whilst a large VSD is illustrated in C and D (asterisk). E,F: Transverse section through an age-matched control heart shows that septation of the ventricles is normally complete at this stage and that tamoxifen treatment had no detectable effect on normal heart development. Abbreviations are the same as for Figure 1.

Figure 5.

Cerebral phenotype of Cdh5(PAC)-CreERT2 /Tgfbr2 fl/Δ embryos. A,B: Whole mount views of E14.5 control and mutant (Cdh5(PAC)-CreERT2/Tgbfr2 fl/Δ) embryos show clearly visible cerebral haemorrhage (B, arrow) compared with control (A). C,D: Transverse brain sections of E14.5 embryos show extensive haemorrhage in mutant embryos (D) whilst tamoxifen-treated controls were unaffected (C). Llv, left lateral ventricle; rlv, right lateral ventricle.


We have shown, using a floxed Tgfbr2 mouse (Leveen et al., 2002) and three different transgenic Cre lines that express Cre recombinase in different regions of the developing heart, that TGFβ signalling is important for normal development of the tricuspid valve, and for fusion of the cardiac cushions during closure of the embryonic interventricular communication. In addition, we have demonstrated that endothelial-specific TGFβ signalling is required for integrity of the cerebral vasculature.

The AV canal myocardial Cre line, cGata6-Cre, expresses Cre in the atrial vestibular musculature supporting the septal leaflet of the tricuspid valve, a site consistent with abnormalities of the septal leaflet noted in the AV canal myocardial Tgfbr2-specific KO mice (Fig. 1). The myocardium involved in the development of the tricuspid valve derives from two sources, the tricuspid gully complex and the developing supraventricular crest (Lamers et al., 1995). Our data show that the cGata6-Cre-labelled cells contribute to the atrial vestibular portion of the tricuspid gully and that loss of TGFβ signalling in the atrial vestibular musculature supporting the tricuspid valvar leaflet has an effect on its length. In contrast, the myocytes that derive from the AV canal at the mitral valve orifice do not appear to depend on TGFβ signalling for their normal development. The shortened leaflet that we observed in cGata6-Cre/Tgfbr2fl/fl mutants is almost the reverse phenotype of the abnormally long leaflets of the tricuspid and mitral valves reported in cGata6-Cre/Alk3fl/fl mutants (Gaussin et al., 2002, 2005). Thus, loss of BMP signalling through Alk3, and loss of TGFβ signalling through Tgfbr2 in AV canal myocardial derivatives, may have opposing signalling effects, resulting in complementary defects in the formation of the valvar leaflets. There was no loss of viability in either the Tgfbr2 or Alk3 AV canal myocardial-specific KOs, suggesting no major deleterious effect on cardiac function.

The majority (8/10) of the hearts from the ventricular myocardium–restricted Tgfbr2 KO embryos (produced using the Mlc2v-Cre line) were normal. The heart of one embryo, however, was grossly abnormal, exhibiting a common AV valve, common arterial trunk, and hypoplastic left ventricle (Fig. 2C). One further embryo had a perimembranous VSD with an overriding aorta. The low incidence of defects in the Mlc2v-Cre/Tgfbr2 fl/Δ embryos is similar to that reported for the myocardial-restricted Tgfbr2 KO embryos generated using cTnT-Cre/Tgfbr2fl/fl mice where only 8% of embryos were affected, one with a VSD and one with a VSD plus double outlet right ventricle (Jiao et al., 2006). Two independent models targeting Tgfbr2 expression in the myocardium, therefore, produced normal cardiac development in the majority of cases. Although it remains possible that the low frequency of defects was due to the non-uniform expression of Mlc2v-Cre in our study, the fact that myocardial-specific depletion of the related receptor Tgfbr1 also has no cardiac phenotype (Sridurongrit et al., 2008), suggests it is more likely that TGFβ signalling is not essential for development of the cardiac myocytes.

By using an inducible endothelial Cre to inactivate Tgfbr2, we were able to bypass the embryonic lethality at E10.5 to E11.5 caused by the constitutively expressed Tie2-Cre or Tie1-Cre lines (Jiao et al., 2006; Carvalho et al., 2007). We, therefore, activated Cre in ECs from E11.5, which also corresponds to the stage at which endothelial to mesenchymal transformation is almost complete. This presumption was confirmed by the lack of lacZ expression in mesenchyme of the arterial valves, and low levels of expression of lacZ in the atrioventricular valves in Cdh5(PAC)-CreERT2 /R26R embryos (Fig. 3A,D). In 93% of endothelial-specific Tgfbr2 KO mutants, we observed perimembranous VSDs, often in combination with delayed maturation of the AV cushions. We were unable, however, to detect any difference in proliferation or apoptosis of cells in the AV cushions or associated endothelial and endocardial cells (not shown). Our evidence, therefore, points to a failure in remodelling of the atrioventricular cushions. The cushion fusion defects may indicate a role for TGFβ signalling in delamination of endothelial cells and eventual fusion of the atrioventricular and outflow cushions to close the interventricular communication. On the other hand, the larger VSDs are more likely to be due to failures in gross remodelling of the cushion mesenchyme across the atrioventricular junction, an essential step during formation of the membranous portion of the ventricular septum (Wessels and Sedmera, 2003). Such remodelling is likely to involve signalling in endothelial/endocardial cells. We observed defects that would be consistent with this hypothesis, including delayed fusion of cushions, and slight misalignment of muscular ventricular septum with the inferior atrioventricular cushion. Both these defects are known to contribute to defective ventricular septation (Webb et al., 1998). Alternatively, it is also formally possible that the high frequency of VSDs in these mutants may be a secondary consequence of cerebral haemorrhage leading to reduced blood flow that might, in turn, affect maturation of the heart. On the other hand, a primary defect caused by reduced TGFβ signalling in the heart would be consistent with the increased frequency of VSDs seen in embryos that are non-haemorrhagic, but are deficient in betaglycan, an auxiliary receptor that promotes TGFβ2 signalling (Stenvers et al., 2003; Compton et al., 2007). In one of these studies, Betaglycan null embryos died at E14, which was thought to be a result of lack of coronary vessels (Compton et al., 2007). In contrast, coronary vessel development appeared to be unaffected in the endothelial-specific Tgfbr2 knockout mice in our study (data not shown). Further investigation of the underlying cause of these VSDs is required to better understand the cellular mechanisms involved in valvoseptal morphogenesis, especially in light of the fact that VSDs are the most common congenital heart defects in newborn babies, and this is a highly penetrant phenotype in our model.

Cerebral haemorrhage was seen in all inducible endothelial-specific Tgfbr2 KO embryos. This was also a reported phenotype in endothelial-specific Tgfbr1 KO embryos that die at E13 (Sridurongrit et al., 2008), indicating that TGFβ signalling is critical for cerebral vascular integrity. (In this study, the cardiac interventricular septum and atrioventricular cushions were poorly developed in mutant embryos compared with controls, but it was not possible to formally investigate VSDs because the normal ventricular septum is still remodelling at E13.) Mice without one of the TGFβ ligands (as seen in TGFβ2, TGFβ3, and certain TGFβ1 KO mice) can develop to birth; this was recently shown to be due to ligand redundancy, because mice carrying mutations in both Tgfb1 and Tgfb3 genes show cerebral hemorrhage from E11.5 and subsequent embryonic lethality (Mu et al., 2008; Aluwihare et al., 2009). Interestingly, mice with mutations in alpha-V or beta-8 integrins also develop cerebral haemorrhage (Bader et al., 1998; Zhu et al., 2002). As integrins are required for TGFβ activation (Sheppard, 2005; ten Dijke and Arthur, 2007), this strengthens the evidence for a critical role of TGFβ signalling in maintaining the integrity of the developing cerebral vasculature.

Taken together, our findings support the importance of TGFβ signalling in cardiovascular development. Signalling is required in AV canal myocardial cells for normal development of the tricuspid valve, and in endothelial/endocardial cells for closure of the embryonic interventricular communication, as well as for integrity of the cerebral vasculature.



All animals were maintained according to the requirements of the Animals (Scientific Procedures) Act 1986 of the UK Government. For timed matings, noon of the day of the copulation plug was taken as 0.5 days embryonic development (E0.5). Pregnant females were humanely killed at the appropriate day of gestation for embryo analysis. To activate the Cre recombinase in embryos carrying the Cdh5(PAC)-CreERT2 transgene, pregnant females were given an intraperitoneal injection of 1 mg Tamoxifen (dissolved in peanut oil) at embryonic day 11.5 and on the following two days (E12.5 and E13.5). All the mouse lines used in this work (R26R, Gata6-Cre, Mlc2v-Cre, Cdh5(PAC)-CreERT2 and Tgfbr2fl/fl) and the primers used for genotyping have been previously described (Chen et al., 1998; Soriano, 1999; Davis et al., 2001; Mahmoud et al., 2010; Leveen et al., 2002). The floxed Tgfbr2 allele was converted to a null (Δ) allele by PGK-Cre mediated recombination. As mice that were ubiquitously heterozygous for this allele (Tgfbr2 fl/Δ) had no detectable phenotype, they were used in some of our breeding programs to reduce the number of Cre-mediated recombination events required to remove Tgfbr2 expression and increase the efficiency of the Cre/lox approach.

Tissue Staining and Analysis

Freshly harvested embryonic tissues were fixed in 4% PFA overnight at 4°C and processed to paraffin. For morphological analysis, transverse sections that had been carefully matched for stage and position within the heart were stained with haematoxylin and eosin. For XGal staining, tissues were fixed for 30 min in 0.2% gluteraldehyde, 0.1M Phosphate Buffer, pH 7.2, 5 mM EGTA, 2 mM MgCl2, and 0.02% NP40 and then processed for XGal staining, paraffin sections, and eosin counterstaining as previously described (Arthur et al., 2000). All sections were mounted onto slides with histomount and photographed using a digital camera attached to a Zeiss Axioplan microscope. Tricuspid valve leaflet length in 8 serial sections from each heart was measured using Axiovision software version 4.4. Mean tricuspid septal valve leaflet lengths were compared between eight controls and eight mutants using an unpaired Student's t-test and graphpad Prism statistical software.


This work was supported in part by BHF studentship FS/04/019 and BHF senior fellowship FS/08/001.