Vascular development depends on transforming growth factor β (TGFβ), but whether signalling of this protein is required for the development of endothelial cells (ECs), vascular smooth muscle cells (VSMCs) or both is unclear. To address this, we selectively deleted the type I (ALK5, TGFBR1) and type II (TβRII, TGFBR2) receptors in mice. Absence of either receptor in ECs resulted in vascular defects in the yolk sac, as seen in mice lacking receptors in all cells, causing embryonic lethality at embryonic day (E)10.5. Deletion of TβRII specifically in VSMCs also resulted in vascular defects in the yolk sac; however, these were observed at later stages of development, allowing the embryo to survive to E12.5. Because TGFβ can also signal in ECs via ALK1 (ACVRL1), we replaced ALK5 by a mutant defective in SMAD2 and SMAD3 (SMAD2/3) activation that retained the ability to transactivate ALK1. This again caused defects in the yolk sac vasculature with embryonic lethality at E10.5, demonstrating that TGFβ/ALK1 signalling in ECs cannot compensate for the lack of TGFβ/ALK5-induced SMAD2/3 signalling in vivo. Unexpectedly, SMAD2 phosphorylation and α-smooth muscle actin (SMAα, ACTA2) expression occurred in the yolk sacs of ALK5–/– embryos and ALK5–/– embryonic stem cells undergoing vasculogenesis, and these processes could be blocked by an ALK4 (ACVR1B)/ALK5 inhibitor. Together, the data show that ALK5 is required in ECs and VSMCs for yolk sac vasculogenesis; in the absence of ALK5, ALK4 mediates SMAD2 phosphorylation and consequently SMAα expression.
The initiation of blood vessel formation in the embryo is called vasculogenesis: the differentiation of undifferentiated precursor cells into endothelial cells (ECs) that assemble into a vascular network (Risau, 1997). After the primary vascular plexus is formed, more ECs are generated, which form new vessels by sprouting or splitting from their vessel of origin in a process termed angiogenesis. ECs can initiate but not complete angiogenesis. Proper maturation requires interactions of ECs with themselves and with the extracellular matrix (ECM), as well as the recruitment of surrounding mesenchymal cells or pericytes, which differentiate into vascular smooth muscle cells (VSMCs). VSMCs protect new vessels against rupture or regression (Hanahan, 1997).
Transforming growth factor β (TGFβ) is thought to play a pivotal role during vascular remodelling and the resolution phase of angiogenesis but is considered to be an inhibitor of angiogenesis. In culture, TGFβ has been shown to inhibit the proliferation and migration of ECs, and to stimulate ECM accumulation and differentiation of mesenchymal cells (Hirschi et al., 1998; Madri et al., 1992; Sawdey et al., 1989). However, TGFβ has also been reported to stimulate angiogenesis in vivo and in ECs in vitro (Fajardo et al., 1996; Koh et al., 1995; Roberts et al., 1986; Yang and Moses, 1990), although TGFβ dose-dependence might reconcile these differential effects (Goumans et al., 2003b).
The action of TGFβ is mediated by two transmembrane serine/threonine kinase receptors – type I and type II. After ligand binding, type II receptors (TβRII, TGFBR2) phosphorylate and activate type I receptors, which then recruit and phosphorylate downstream signalling mediators, called SMADs. There are eight distinct SMAD proteins, which can be divided into three classes: the receptor regulated (R-)SMADs, including SMAD1, SMAD2, SMAD3, SMAD5 and SMAD8; the common mediator (Co-)SMAD, SMAD4; and the inhibitory (I-) SMADs, which includes SMAD6 and SMAD7. R-SMADs oligomerize with SMAD4, translocate to the nucleus and regulate the transcription of target genes (Feng and Derynck, 2005; Massague and Gomis, 2006). Activin receptor-like kinase 5 (ALK5, TGFBR1) is a widely expressed TGFβ type I receptor (Franzen et al., 1993) that phosphorylates SMAD2/SMAD3. Mice null for Smad3 survive into adulthood, unlike Smad2-null mice, which die in utero at midgestation (Datto et al., 1999). This suggests that SMAD3 is not required for embryonic development and cannot compensate for the severe defects in Smad2 mutant mice. A second TGFβ type I receptor, ALK1 (ACVRL1), is expressed in ECs (Roelen et al., 1994) and, by contrast, phosphorylates SMAD1 and SMAD5 (Goumans et al., 2002; Goumans et al., 2003b; Lebrin et al., 2004; Oh et al., 2000). TGFβ also binds type III co-receptors, β-glycan and endoglin; endoglin, like ALK1, is also expressed predominantly in ECs (Gougos and Letarte, 1990).
Deletion of either the TGFβ1 ligand, TβRII, ALK1, ALK5 or endoglin in mice severely disrupts vasculogenesis in the yolk sac and leads to embryonic lethality at midgestation (Arthur et al., 2000; Carvalho et al., 2004; Dickson et al., 1995; Larsson et al., 2001; Leveen et al., 2002). This demonstrates an essential role for TGFβ signalling in vasculogenesis in vivo. However, it is not yet clear whether ALK5 is actually expressed in ECs or only in VSMCs (Seki et al., 2006), in part because mRNA levels of some TGFβ receptors are so low. We sought to address this here by comparing embryos lacking TβRII or ALK5 entirely with embryos in which either of the receptors had been deleted specifically in ECs or VSMCs. The results showed that specific absence of the individual receptors in either ECs or VSMCs caused defective yolk sac vasculogenesis and embryonic lethality, as has previously been described in the complete knockout, although embryos lacking TβRII in VSMCs died at a later embryonic stage and with a less-severe phenotype. This indicated that ALK5 is necessary in both ECs and VSMCs for vascular development and confirmed the co-expression of ALK5 and ALK1 in ECs of the yolk sac. The absence of phosphorylated SMAD2 (PSMAD2) and α-smooth muscle actin (SMAα, ACTA2) in VSMCs of mice in which the receptors were absent only in ECs strongly supports previous evidence that any disruption of TGFβ signalling in ECs causes defective processing of the TGFβ they secrete, which secondarily affects VSMC differentiation (Carvalho et al., 2004).
Replacement of ALK5 with a mutant defective in SMAD2 phosphorylation that had retained the capacity to transactivate ALK1 in vitro (Goumans et al., 2003b), again resulted in mice with a phenotype resembling the complete ALK5 knockout. This demonstrated that, in vivo, TGFβ/ALK1-signalling in ECs cannot compensate for lack of TGFβ/ALK5-induced SMAD2/SMAD3 signalling, extending previous observations in ECs in vitro (Goumans et al., 2003b).
Surprisingly, we found that, although PSMAD2 was absent in VSMCs when ALK5 was deleted in ECs, it was detectable in complete-knockout mice. We therefore hypothesized that a mechanism of compensatory upregulation might underlie SMAD2 phosphorylation and identified activin/ALK4 (ACVR1B) signalling among possible candidates.
Lack of TGFβ signalling in ECs is responsible for embryonic lethality in TGFβ receptor mutant mice
Mice lacking TβRII or ALK5 die at around E10.5 because of aberrant vasculogenesis in the yolk sac (Larsson et al., 2001; Oshima et al., 1996). Mice lacking endoglin, which is expressed in ECs but not in VSMCs and pericytes of the yolk sac, show a very similar phenotype to the TGFβ1, ALK5 and TβRII single-mutant mice, including defective smooth muscle cell differentiation (Carvalho et al., 2004). However, it is not yet known whether these similarities are the result of abnormal TGFβ signalling in ECs, pericytes or both cells types. To address this, we deleted TβRII and ALK5 selectively in ECs and pericytes using tissue-specific Cre-loxP deleter mice.
Deletion in ECs
To determine the requirement of TGFβ signalling in ECs for vasculogenesis, we first deleted TβRII and ALK5 independently in these cells by crossing transgenic mice expressing Cre-recombinase under control of the vascular-EC-specific Tie-1 promoter with floxed TβRII and floxed ALK5 mice, respectively. Inclusion of the floxed Rosa26 allele allowed retrospective monitoring of Cre-activity (Soriano, 1999). Genotyped offspring from Tie1-Cre transgenic mice and floxed TβRII or floxed ALK5 intercrosses identified heterozygous and wild-type pups only, indicating that homozygous Tie1-Cre-TβRIIfl/fl and Tie1-Cre-ALK5fl/fl mice died during development (Table 1). Embryos from these crosses were therefore collected at embryonic day (E)8.5, E9.5 and E12.5. lacZ staining in the yolk sacs at E9.5 confirmed the efficiency and specificity of the transgenic Cre. Wild-type and heterozygous embryos were morphologically similar (Fig. 1A,D), whereas Tie1-Cre-TβRIIfl/fl and Tie1-Cre-ALK5fl/fl embryos exhibited distinct abnormalities in the vasculature of the yolk sac at E9.5 (Fig. 1B,C,E,F). By E12.5 only TβRIIfl/fl and ALK5fl/fl embryos were recovered (Tables 1 and 2), indicating that Tie1-Cre-TβRIIfl/fl and Tie1-Cre-ALK5fl/fl mice died at around E10.5, similar to the conventional TβRII and ALK5 mutants. The yolk sacs of mutant embryos lacking TβRII or ALK5 specifically in ECs also lacked networks of vessels at all stages examined (not shown), and the embryos were retarded and exhibited pericardial effusion in the heart (Fig. 1B,E), phenocopying the gross morphological defects seen in embryos lacking TβRII, ALK5 and endoglin in all cells. At E9.5, some embryos were approximately 1-day retarded compared with the wild type (Fig. 1C,F).
|.||wt/wt; – .||wt/wt; Cre .||wt/fl; – .||wt/fl; Cre .||fl/fl .||fl/fl; Cre .||Total .|
|.||wt/wt; – .||wt/wt; Cre .||wt/fl; – .||wt/fl; Cre .||fl/fl .||fl/fl; Cre .||Total .|
fl, floxed TβRII; wt, wild-type TβRII
|.||wt/wt; – .||wt/wt; Cre .||wt/fl; – .||wt/fl; Cre .||fl/fl .||fl/fl; Cre .||Total .|
|.||wt/wt; – .||wt/wt; Cre .||wt/fl; – .||wt/fl; Cre .||fl/fl .||fl/fl; Cre .||Total .|
fl, floxed ALK5 wt, wild-type ALK5
Deletion in VSMCs
To determine whether disruption of TGFβ signalling in VSMCs also contributed to the early vascular defects, we analyzed embryos from SM22-Cre-TβRIIfl/fl intercrosses. Again, only wild-type and heterozygous pups were present within the litters born. Embryos were therefore collected for examination at E9.5, E12.5 and E16.5 (Fig. 2). β-galactosidase staining confirmed the expression of SM22 and efficiency of Cre recombinase in VSMCs. Besides in the yolk sac, amnion and heart (Zhang et al., 2001), lacZ-expressing cells were also detected in the brain, dorsal aorta and vessels between the somites at E9.5 (Fig. 2B,D and not shown). Interestingly, at E9.5, embryos lacking TβRII in VSMCs did not exhibit overt yolk sac defects, and resembled wild-type and heterozygous embryos (Fig. 2A,C). However, at E12.5, the yolk sacs of these embryos were pale and anaemic, with obvious vasculature defects (Fig. 2E,F). The embryos were also retarded and some showed underdeveloped hearts and brains (Fig. 2G,H). At E16.5, only resorbed embryos were found, indicating that SM22-Cre-TβRIIfl/fl mice died between E12.5 and E16.5.
Is ALK5 kinase activity required in ECs for vasculogenesis in mice?
Selective deletion of the predominant TGFβ receptors showed that their mediation of TGFβ signalling in both ECs and VSMCs/pericytes was necessary for normal vasculogenesis in the yolk sac. In ECs in culture, TGFβ can activate a second type I receptor, ALK1, provided that ALK5 is also present to mediate activation of SMAD2/SMAD3. In order to determine whether ALK1 is sufficient to mediate TGFβ signalling in ECs in vivo, we generated a knock-in mouse expressing an ALK5 mutant defective in SMAD2 activity obtained by replacing the aspartic acid residue 266 by an alanine residue, ALK5(D266A). It has been shown previously in vitro that this mutant is unable to activate SMAD2/SMAD3 but is capable of transactivating ALK1 in a TGFβ-dependent manner and induces SMAD1/SMAD5 activation (Itoh et al., 2003).
Genotyping offspring from heterozygous intercrosses resulted only in heterozygous and wild-type pups being identified, indicating that homozygous ALK5(D266A) knock-in mutant mice died during development. Examination of embryos from heterozygous intercrosses showed that mice lacking TGFβ/ALK5-induced SMAD2/SMAD3 signalling exhibited a phenotype similar to the TGFβ ligand and receptor mutant mice. Yolk sacs from wild-type embryos at E9.5 showed well-organized vessels, whereas those from both ALK5–/– and ALK5 knock-in embryos on the same day of development had a honey-comb structure without vessels (Fig. 3A-D) and the embryos died at around E10.5.
Together, the results showed that ECs express both TβRII and ALK5 and that, although TGFβ signalling in VSMCs is important for complete development, the initial phase of angiogenesis occurs in the absence of TGFβ signalling in these cells. Additionally, data from both the Tie1-Cre transgenic crosses and ALK5(D266A) knock-in mice confirmed the ALK5 expression in ECs described previously (Lebrin et al., 2004) and demonstrated that it was not restricted to the smooth muscle cells in the yolk sac.
The similarities in timing and nature of the phenotype in the complete ligand and receptor knockouts and those with a conditional deletion in ECs demonstrated that the lack of signalling in ECs is the earliest and major cause of the vascular abnormalities seen in these mutants.
SMAD2 phosphorylation in yolk sacs of mutants with impaired TGFβ/ALK5 signalling
To determine how the loss of ALK5 activity affected TGFβ signalling in isolated yolk sacs, we compared phosphorylation of SMAD1 and SMAD2 in the yolk sacs of wild-type, ALK5 knock-in and ALK5–/– mice (Fig. 3E-J). It has been shown that SMAD3 is not present in the yolk sac at E8.0-E8.5 (Dunn et al., 2004). Therefore, we carried out the studies focusing on SMAD2 and not SMAD3. In wild-type yolk sacs, PSMAD1 and PSMAD2 were detected in the EC layer (Fig. 3E,F). However, in ALK5–/– embryos, SMAD1 phosphorylation, which would indicate TGFβ signalling via ALK1 or BMP signalling, was not detected in this layer (Fig. 3G) although it was present in yolk sac ECs of ALK5 knock-in mice (Fig. 3I). Unexpectedly, PSMAD2 was detected in ECs of both the knockout (Fig. 3H) and knock-in (Fig. 3J) mutant mice, as in wild type (Fig. 3F), suggesting that SMAD2 is phosphorylated independently of TGFβ/ALK5 signalling. The results also showed that ALK5 is required in vivo for efficient TGFβ/ALK1-induced SMAD1 activation, confirming previous observations in vitro (Goumans et al., 2003b).
Because we observed that, in ALK5 knock-in mice, SMAD2 was phosphorylated in both ECs and VSMCs in the yolk sac, we next determined whether this was also the case in ALK5–/– mice. The assumption was that the phosphorylation of SMAD2 in these cells was mediated by activation of ALK5 by TGFβ and that, in ALK5–/– embryos, there would be no SMAD2 phosphorylation in the mesothelial layer of the yolk sac that would become VSMCs. In fact, SMAD2 was phosphorylated in freshly isolated yolk sacs (Fig. 3 and Fig. 4A,C), although not in the ECs, as would be expected. SMAα was also detected in ALK5–/– yolk sacs (Fig. 4D,F), suggesting that the mesothelial layer had differentiated. As shown previously, this was dependent on SMAD2 phosphorylation (Carvalho et al., 2004); however, it might not only be mediated by ALK5.
Immunohistochemistry for PSMAD2 was also carried out on paraffin sections of E9.5 embryos (supplementary material Fig. S1). Analysis of serial sections through the embryo revealed normal PSMAD2 staining in the neural tube, mesenchymal tissue, heart, arteries, veins, foregut and somites; staining was entirely comparable with wild-type embryos (de Sousa Lopes et al., 2003). The results suggested that loss of SMAD2 phosphorylation was not the cause of the early phenotype within the embryo itself but that this was more likely secondary to the yolk sac defects.
Compensatory mechanisms resulting from deficient TGFβ/ALK5 signalling
The loss of ALK5 in early development might result in the upregulation of another signalling pathway able to phosphorylate SMAD2 and SMAD3, namely that mediated by ActR-IB (ALK4) (Massague, 1996; Whitman, 1998), a type I receptor that can bind activin (Attisano et al., 1993; ten Dijke et al., 1993) and nodal (Gu et al., 1998). To investigate whether this was in fact the case, we quantified the expression level of ALK4 mRNA in yolk sacs from wild-type, ALK5–/– and Tie1-Cre-ALK5fl/fl embryos using real-time PCR. ALK4 mRNA was significantly upregulated in ALK5–/– yolk sacs compared with wild-type littermates (Fig. 4G), whereas, in Tie1-Cre-ALK5fl/fl yolk sacs, the levels were lower than in wild type (Fig. 4H). We next examined the co-localization of ALK4 and PECAM in ECs and SMAα in pericytes/VSMCs both from yolk sacs and embryoid bodies (EBs) derived from embryonic stem (ES) cells by double immunostaining (Fig. 5A-F; Fig. 6). Mouse ES cells can undergo programmed differentiation to form ECs that organize into mature vessels resembling those in the yolk sac when grown as EBs in the presence of VEGF and FGF (Doetschman et al., 1985; Keller et al., 1993; Wang et al., 1992). Staining indicated co-localization of ALK4 and SMAα in VSMCs, and demonstrated that, in the yolk sac and EBs, ALK4 is mainly, although not exclusively, expressed in these cells (Fig. 5F and Fig. 6). Interestingly, the number of SMAα positive VSMCs was slightly, but significantly, higher in ALK5–/– yolk sacs when compared with wild-type yolk sacs (Fig. 5G).
To investigate whether loss of ALK5 affected the ability of EBs to undergo vascular development, we immunostained cryosections of ALK5–/– EBs with PECAM-1 and SMAα (Fig. 6A,B). Similar to the wild-type ES cells, ALK5–/– ES cells formed EBs containing both ECs and VSMCs, reflecting the results we observed in the ALK5–/– embryos; however, ECs were observed in very large numbers (not shown), possibly due to a lack of growth inhibition conferred by TGFβ/ALK5 signalling (Goumans et al., 2003a; Lebrin et al., 2004).
In order to determine whether signalling mediated by SMAD2 was still active in ALK5–/– EBs and able to induce VSMC differentiation, the expression of SMAα was compared by immunostaining in wild-type and ALK5–/– EBs. Comparable SMAα staining was observed between wild-type and mutant EBs (Fig. 7A-B′), suggesting that a compensatory mechanism might be active in ALK5–/– mice. To determine whether this might be mediated by activin/ALK4, we used an inhibitor of ALK4/ALK5/ALK7, SB-431542 (Goumans et al., 2003b). Treatment with SB-431542 led to complete loss of SMAα expression in both wild-type and ALK5–/– EBs (Fig. 7C-D′), suggesting that ALK4 is indeed responsible for the expression of SMAα that was observed in the ALK5–/– EBs and possibly also in the ALK5–/– yolk sacs.
TβRII and ALK5 signal in yolk sac endothelial and smooth muscle cells
To investigate the function of TGFβ in cells of the developing vasculature, we generated mice that lacked TβRII or TβRI/ALK5 specifically in ECs or VSMCs and compared them with the conventional TβRII and ALK5 knockouts. EC-specific ALK5–/– mice died at around E10.5, exhibiting pale and anaemic yolk sacs that lacked proper vascular networks, very similar to the phenotype observed in the conventional mutants. It has recently been shown that embryos from TβRIIfl/fl mice crossed with those carrying tie-2-Cre, another endothelium-specific promoter (Schlaeger et al., 1997), have severe anaemia and defective vasculogenesis, consistent with TβRII–/– embryos (Jiao et al., 2006). These data confirm that TGFβ signalling in ECs is essential for precise regulation of blood vessel assembly and consequently normal embryonic development. They also suggest that the impaired VSMC differentiation reported in both ALK5–/– and endoglin–/– embryos is a direct consequence of functional defects in ECs.
Prior to the establishment of the chorioallantoic placenta at E10-E11, the embryo is dependent on the yolk sac for nourishment. Our results are consistent with several studies suggesting that TGFβ, acting by paracrine and autocrine mechanisms, is a crucial mediator of yolk sac vasculogenesis beginning at E8.5 (Larsson et al., 2001; Oh et al., 2000; Arthur et al., 2000; Cohen et al., 2007; Goumans and Mummery, 2000). By this stage of development, both TGFβ and its high-affinity receptor are expressed in the yolk sac vasculature. Moreover, mutations in genes independent of the pathways described here, but which also cause disrupted TGFβ activity, also affect yolk sac vasculogenesis. For example, in mice lacking connexin 43 (GJA1) (Kruger et al., 2000), TGFβ levels are reduced and the embryos die between E9.5 and E11.5 from abnormalities in yolk sac vasculature. This might be a common mechanism adopted in mouse mutants with defective yolk sac vasculature and might provide clues to the molecular cascade underlying the defects.
By the analysis of heterozygous mice in which lacZ had been knocked into the ALK5 locus, it has recently been suggested that, in blood vessels, TGFβ signals specifically in ECs via ALK1 and in VSMCs via ALK5, with ALK5 being undetectable in ECs (Seki et al., 2006). Our examination of Tie1-Cre-ALK5fl/fl mice suggests this is not the case. Although the homozygous ALK5-lacZ mice exhibited the expected defects in the yolk sac and embryonic lethality occurred at around E10.5, detailed analysis of the expression of ALK5 based on β-galactosidase staining in the vasculature was not described before E10.5, either in the yolk sac or in embryonic blood vessels. ALK5 could therefore have been expressed in the early ECs. Furthermore, in the present study, we used the SM22α promoter in transgenic mice to disrupt TGFβ signalling in VSMCs. This promoter is expressed in the yolk sac at a time when the embryo is dependent on the yolk sac vasculature for nutrition and survival (Zhang et al., 2001). Mice lacking TβRII specifically in VSMCs exhibited an aberrant phenotype later in development than the conventional TβRII or ALK5 mutant mice, indicating that disruption of TGFβ signalling in VSMCs is not responsible for the early phenotypes observed in either the TβRII or ALK5 mutants. Moreover, it has been shown independently that homozygous deletion of TβRII by SM22-Cre mice is embryonic lethal, confirming SM22-Cre mice as useful tools for deleting floxed alleles in a large proportion of VSMCs (Frutkin et al., 2006). Additionally, ALK5(D266A) knock-in mice showed that TGFβ-induced ALK1 signalling required TGFβ/ALK5-induced SMAD2 signalling in ECs; if it were not required, the mutant phenotype would have appeared later, similar to the VSMC deletion of ALK5 rather than to its EC deletion. Furthermore, this extends our previous finding in ECs in vitro to at least some EC types in vivo, an essential extrapolation if we continue to use cultures of ECs as predictive models.
Taken together, these results are in agreement with previous studies showing ALK5 expression in ECs (Lebrin et al., 2004). More importantly, they show that TGFβ signalling is required in ECs in the early phase of angiogenesis and in VSMCs during the later phase of blood vessel development.
ALK4 is able to phosphorylate SMAD2 in ALK5–/– mice
An unexpected finding was that SMAD2 could be phosphorylated in embryos and in the yolk sac mesothelial layer that would become VSMCs, in the absence of ALK5. ALK4 mRNA levels were upregulated in ALK5–/– yolk sacs, suggesting that the difference between conditional and complete deletion might have at least in part been caused by adaptation during early development, with ALK4 compensating for decreased ALK5 signalling.
The ability of ALK4/ALK5 kinase inhibitor SB-431542 to inhibit SMAD2 phosphorylation in ALK5–/– EBs supported our suggestion that there is a requirement for ALK4 when the TGFβ pathway is intrinsically suppressed via a mechanism involving the loss of ALK5.
At E7.5, expression of ACTR-IIA (ACVR2A) and ActR-IIB (ACVR2B) is essentially ubiquitous in all three germ layers, with only visceral endoderm being negative for these receptors (Mummery and van den Eijnden-van Raaij, 1993). SMAD2 phosphorylation in ALK5–/– yolk sacs and embryos could be due to activin signalling, and, during gastrulation, ALK4 is also known to bind nodal (Gu et al., 1998). Furthermore, studies of ERBB2 (NEU)-induced tumours showed the presence of phosphorylated SMAD2 in association with the expression of ALK4, suggesting that, although SMAD-dependent TGFβ signalling was absent because of loss of ALK5, activin signalling might be active at the leading edge of these tumours (Landis et al., 2005).
Despite the phosphorylation of SMAD2 and differentiation of smooth muscle cells in ALK5–/– embryos, these mice were not capable of forming an organized vasculature and consequently only developed partially through organogenesis. One explanation for this might be that, although SMAD2 did become phosphorylated, ALK4 upregulation and SMAD2 activation occurred either too late or just transiently, and were therefore not sufficient to rescue complete development. Alternatively, although ALK4 can drive expression of the early marker SMAα, it might not be sufficient for terminal differentiation of mesenchymal cells; this might explain why there are more SMAα-positive cells in ALK5–/– yolk sacs, contributing to the increase in ALK4 expression that they also exhibit.
Together, our results suggest that the defects observed in ALK5 and TβRII mutant mice are caused primarily by a functional defect in ECs, which in turn impairs the differentiation of smooth muscle cells via a TGFβ-dependent mechanism. Combined with our previous data on mice lacking endoglin (Carvalho et al., 2004), in which we showed that defective TGFβ signalling in ECs results in their failure to process TGFβ properly and subsequently to mediate differentiation of VSMCs, we speculate that an underlying mechanism of the phenotypes in other mice with defective yolk sac vasculogenesis might be related to defective TGFβ signalling. Moreover, deletion of one gene might result in activation of compensatory mechanisms in an attempt by the conceptus to survive (Fig. 8). Our results also showed that ALK5 is clearly expressed in ECs of the yolk sac, although this does not exclude the possibility that ECs from other tissues do not express ALK5 or that upregulation might occur if ECs are cultured.
Materials and Methods
Mice were kept on a mixed genetic background. ALK5 mutant mice (Larsson et al., 2001), floxed ALK5 mice (Larsson et al., 2001) and floxed TβRII mice (Leveen et al., 2002) have been described previously. Floxed TβRII, floxed ALK5 and ALK5 mutant mice were crossed with R26R (Zambrowicz et al., 1997) and subsequently crossed with tie1-Cre transgenic mice (Gustafsson et al., 2001) to delete TβRII and ALK5 in ECs. Floxed TβRII mice were also crossed with SM22-Cre (Holtwick et al., 2002) to delete TβRII in smooth muscle cells. The presence of a vaginal plug was designated as E0.5.
Replacement of ALK5 in mice by an ALK5 mutant defective in SMAD2 and SMAD3 phosphorylation
The mutant ALK5 knock-in targeting vector was constructed by site-directed mutagenesis using the original TβRI/ALK5 targeting construct as a template (Larsson et al., 2001). GAC encoding Asp266 in exon 4 was substituted by GCC, which encodes Ala. The ALK5 knock-in targeting vector was linearized with NotI and electroporated into E14 ES cells from 129Sv mice. Cells were selected in the presence of G418 and gancyclovir. Screening for homologous recombination was performed according to Larsson et al. (Larsson et al., 2001). One ES cell line was injected into C57BL/6 blastocysts. Chimeric mice obtained were further crossed for germline transmission. To confirm that mRNA corresponding to ALK5(D266A) was expressed in ALK5 knock-in homozygotes, RT-PCR was carried out using mRNA from spleen and liver in ALK5 knock-in mice. Point mutation in exon 4 was confirmed when cDNAs corresponding to mRNA from exon 3 to exon 5 were amplified.
ALK5–/– ES cell differentiation in EBs
ALK5 wild-type (+/+) and –/–ES cells were routinely cultured in the presence of mouse embryonic fibroblasts (MEFs) in Dulbecco's minimum essential medium (DMEM) supplemented with 20% heat-inactivated fetal bovine serum (FBS), 0.1 mM β-mercaptoethanol and 1000 U/ml recombinant leukaemia inhibitory factor (LIF). For differentiation into EBs, ES cells were cultured as hanging drops as previously described (Slager et al., 1993). Briefly, 800 cells were cultured in 20 μl DMEM supplemented with 20% FBS, 25 ng/ml VEGF, 10 ng/ml bFGF (FGF2), in the presence or absence of 10 μM SB431542, an ALK4/ALK5 inhibitor, in drops hanging from the lids of culture dishes for 5 days. EBs of uniform size thus formed, which were cultured for 15 days in suspension on bacterial dishes coated with 1% agar.
Immunohistochemistry and immunofluorescence
After isolation, E9.5 yolk sacs were divided in half, cultured in DMEM with or without TGFβ1 (1 ng/ml) for 1 hour at 37°C and then fixed for 30 minutes at room temperature in 2% paraformaldehyde (PFA). EBs cultured for 15 days were fixed in MeOH/DMSO (4:1). Whole-mount immunohistochemistry of yolk sacs and EBs, and immunohistochemistry in paraffin sections of E9.5 embryos were performed as in previously described (Carvalho et al., 2004). Anti-phosphorylated SMAD2 (PS2) (Persson et al., 1998), anti-phosphorylated SMAD1 (PS1) (Persson et al., 1998) and anti-smooth muscle actin (1:400, Sigma) were used as primary antibodies, and biotin-conjugated goat anti-rabbit IgG (for PS2 and PS1) and biotin-conjugated anti-mouse IgG (for smooth muscle actin) as secondary antibodies (DAKO, 1:250).
For immunofluorescence, E9.5 yolk sac or EBs were processed for cryosections as previously described (Bajanca et al., 2004), sectioned at 7 μm, fixed in acetone for 10 minutes at 4°C, then air dried for 30 minutes at room temperature. Sections were permeabilized for 5 minutes with 0.2% Triton X-100 in PBS, blocked with 2% BSA in PBS for 1 hour then incubated with a rat anti-mouse PECAM antibody (clone 13.3, dilution 1/100, BD Biosciences) or a mouse anti-SMAα antibody (clone 1A4, dilution 1/500, Sigma) in the presence of a rabbit anti-mouse ALK4 antibody (Rosendahl et al., 2002) overnight at 4°C. The slides were then washed four times with PBS and incubated for 1 hour with a donkey anti-rat FITC antibody (dilution 1/800; Jackson ImmunoResearch) or a goat anti-mouse FITC antibody (dilution 1/800; Jackson ImmunoResearch) in the presence of a goat anti-rabbit Cya3 antibody (dilution 1/1000; Jackson ImmunoResearch). The slides were then washed four times with PBS and mounted in Mowiol before Leica TCS-SP2 confocal microscope analysis.
Analysis of VSMCs in the yolk sac
Immunostaining for SMAα was used to identify VSMCs in the yolk sac. Positive cells for SMAα were determined in ALK5-deficient and wild-type yolk sacs, and endoderm cells were used as an internal control.
RNA extraction and real-time PCR analysis
RNA from E9.5 yolk sacs was isolated using Ultraspec (Biotecx), according to the manufacturer's protocol; 10 μg PolyI (Sigma) was added as a carrier. After DNase treatment, 1 μg RNA was reversed transcribed as described previously (Roelen et al., 1994). RNA samples that had not been reverse transcribed were used as negative control.
Real-time PCR was performed in a MyiQ single-color real-time detection system (BIO-RAD) as described previously (Carvalho et al., 2004). Primers: β-actin forward primer 5′-CCTGAACCCTAAGGCCAACCG-3′ and reverse 5′-GCTCATAGCTCTTCTCCAGGG-3′, annealing temperature 60.2°C. ALK4 forward primer 5′-CTGAGGACTGCTACGGGAA and reverse 5′-TAAGCGTGCAGGAAGATGT-3′, annealing temperature 56.4°C.
We thank Amanda Barlow for comments and discussion on the manuscript and P. Tjou Sin for technical assistance. F.L. and M.-J.G. are supported by the Dutch Platform for Tissue Engineering (Bsik), M.-J.G. is supported by NWO-VIDI grant 016.056.319, R.L.C.C. and P.B. by EU grant QLG1-CT-2001-01032, and P.t.D. by EU grant `Angiotargetting' and the Dutch Cancer Society.