TGF-β and activin inhibit formation of ESC-derived endothelial sheets
According to a previous report (
Yamashita et al., 2000), undifferentiated mouse ESCs were cultured for 3 d on collagen IV–coated dishes with 10% FCS to induce Flk1+ cells, which were purified by cell sorting using anti-Flk1 antibody. After an additional 3 d of culture of Flk1+ cells with 10% FCS, we obtained cells positive for a mural cell marker, α-smooth muscle actin (SMA; a). When 30 ng/ml VEGF was added, we were able to obtain platelet-endothelial cell adhesion molecule 1 (PECAM1)–positive sheets of endothelial cells. The remaining cells surrounding the sheets were positive for SMA.
To examine the effects of TGF-β superfamily signals on ESC-derived in vitro vasculogenesis, we studied the expression of various TGF-β superfamily signaling components by semi-quantitative RT-PCR analysis (). ESC-derived Flk1+ cells (Flk1+), mural cells (MC), and mixed populations of endothelial and mural cells (EC + MC) expressed transcripts for most components of TGF-β superfamily signaling pathways, suggesting that they are capable of responding to BMPs, TGF-βs, and activins. Next, we examined the effects of the TGF-β superfamily proteins on the in vitro vascular differentiation of ESC-derived Flk1+ cells. Although BMP7 did not exhibit significant effects, TGF-β and activin led to the decrease in PECAM1+ sheets of endothelial cells ( a). Moreover, the effects of TGF-β and activin on PECAM1+ sheets of endothelial cells were reversed by the addition of anti–TGF-β neutralizing antibody and follistatin, a natural extracellular inhibitor of activin, respectively.
Goumans et al. (2002) showed that the response of mouse embryonic endothelial cells to TGF-β is biphasic, with ALK-1–mediated positive effects on growth and migration at low doses (0.25–0.5 ng/ml) and ALK-5–mediated negative effects at high doses (2.5 ng/ml and higher). To examine whether TGF-β exhibits biphasic effects in ESC-derived vascular cells, different doses of TGF-β were added to the culture of Flk1+ cells in the presence of VEGF ( b). The formation of endothelial sheet was significantly inhibited at 0.03 ng/ml, and its inhibition reached a maximum at 1 ng/ml, which remained at this level at higher doses ( b). These results suggest that TGF-β does not exhibit biphasic effects in the ESC-derived endothelial sheet formation.
TGF-β receptor kinase inhibitors enhance growth and integrity of ESC-derived endothelial cells
Next, we attempted to enhance formation of endothelium by inhibiting endogenous TGF-β and activin signaling. First, we added anti–TGF-β neutralizing antibody, follistatin or both of them to culture containing VEGF in the absence of exogenous TGF-β or activin. Neither of the extracellular inhibitors exhibited significant effects on the formation of endothelial sheets induced by VEGF ( a).
Synthetic inhibitors of signaling kinases have proven to be extremely useful for investigating signal transduction pathways and are potentially of use for the treatment of various human diseases in which growth factor signals play important roles. Recently, synthetic compounds, e.g., SB-431542 (compound 14), that specifically inhibit the kinases of ALK-4, -5, and -7 (a type I receptor for Nodal) have been identified (
Callahan et al., 2002;
Inman et al., 2002). They have no effects on the other, more divergent ALK family type I receptors that bind BMPs, or on other kinases including ERK, JNK, and p38 MAPK. We tested the specificity of SB-431542 for kinase activities of seven mammalian type I receptors for the TGF-β superfamily (Fig. S1, available at
http://www.jcb.org/cgi/content/full/jcb.200305147/DC1), and confirmed that SB-431542 is a specific inhibitor for ALK-4, -5, and –7.
When SB-431542 was added to ESC-derived Flk1+ cells in the presence of VEGF and TGF-β, it reversed the inhibition of formation of endothelial sheets by TGF-β in a dose-dependent manner ( b), suggesting that SB-431542 is capable of inhibiting the ALK-5 kinase activity in these cells. Next, we added SB-431542 to the ESC-derived in vitro vascular differentiation culture containing VEGF in the absence of exogenous TGF-β. Interestingly, the endothelial sheets were significantly enlarged, and formed a fine cobblestone-like structure (, c–e), which was observed in a dose-dependent manner ( c). Similar results were obtained with another ALK-4/-5/-7 kinase inhibitor, compound 13 (
Callahan et al., 2002; unpublished data). These results suggest that inhibition of ALK-4/-5/-7 kinase activities modifies the growth and sheet formation of endothelial cells.
Quantitative analyses of the effects of TGF-β signals on proliferation, differentiation, and integrity of ESC-derived endothelial cells
TGF-β has been shown to have strong growth inhibitory effects on endothelial cells (
Baird and Durkin, 1986). To examine the effects of TGF-β on the proliferation of differentiating vascular progenitor cells, we added TGF-β or SB-431542 to the Flk1+ cells in the presence of 10% FCS and VEGF, and determined the change in cell number after 3 d ( a). TGF-β resulted in a reproducible decrease in cell proliferation, whereas SB-431542 induced an increase in proliferation, suggesting that TGF-β signals inhibit the growth of differentiating vascular progenitor cells. To examine whether this growth inhibitory effect of TGF-β signals is on endothelial cells or mural cells, we added TGF-β or SB-431542 to the Flk1+ cells in the presence of 10% FCS without VEGF, and determined the change in cell number after 3 d ( b and Fig. S2, available at
http://www.jcb.org/cgi/content/full/jcb.200305147/DC1). Addition of TGF-β resulted in a reproducible decrease in the proliferation of mural cells, whereas SB-431542 induced an increase in proliferation, suggesting that TGF-β signals inhibit the growth of ESC-derived endothelial and mural cells.
To further dissect the roles of TGF-β in the differentiation of Flk1+ cells into endothelial and mural cells, we performed quantitative analyses using a limiting dilution assay. When Flk1+ cells were plated at a lower density, they formed individual colonies, and the total number of colonies varied depending on the culture conditions. TGF-β decreased the number of colonies, whereas SB-431542 strongly increased it ( c).
Next, we evaluated the phenotypes of the colonies by immunohistochemical analysis. Culturing Flk1+ cells with 10% FCS resulted in colonies of a single type consisting of pure mural cells positive for SMA. VEGF induced four types of colonies emerging from single Flk1+ cells: PECAM1+ pure endothelial cells with or without sheet structure (EC-sheet and -scattered, respectively); pure mural cells (MC); and mixtures of both (Mix; c). The frequency of pure endothelial cell colonies (EC-scattered and -sheet) was 25%. Addition of TGF-β or SB-431542 reproducibly decreased (22%) or increased (33%) this frequency, respectively, suggesting that TGF-β signals modify the balance of differentiation from vascular progenitor cells.
Although endothelial cells were observed in the presence and absence of TGF-β signals, formation of endothelial sheets was significantly affected ( c). The frequency of sheet formation among pure endothelial colonies was 60% when single Flk1+ cells were cultured in the presence of FCS and VEGF. When TGF-β was added, most endothelial colonies exhibited scattered phenotypes, whereas 86% of endothelial colonies formed sheet structures when SB-431542 was added ( c). In contrast, neither TGF-β or SB-431542 altered the integrity of the mural cell colonies (Fig. S2). Together with the finding that addition of SB-431542 increased the total number of vascular cells/colonies, these results suggest that inhibition of endogenous TGF-β signals results in enhancement of endothelial sheet formation.
Activation of R-Smads by TGF-β superfamily signals in ESC-derived vascular cells
To examine the activity of endogenous TGF-β signals in ESC-derived endothelial cells, we examined phosphorylation of R-Smads by a phospho-Smad1/5 antibody and a phospho-Smad2 antibody. Stimulation of mixed populations of endothelial and mural cells with TGF-β for 1 h resulted in phosphorylation of both Smad2 and Smad1/5 ( a, lane 3), as reported previously in other types of primary endothelial cells (
Goumans et al., 2002).
Next, we examined dose-dependent effects of TGF-β on the phosphorylation of Smads in ESC-derived vascular cells. In mouse embryonic endothelial cells, ALK-5–mediated Smad2 phosphorylation was induced substantially at 0.025 ng/ml of TGF-β and reached a maximum at 0.25 ng/ml, which remained at this level at higher doses, whereas ALK-1–mediated Smad1/5 phosphorylation was induced at low doses (0.25–0.5 ng/ml) of TGF-β but was decreased at 1 ng/ml (
Goumans et al., 2002). We have shown that TGF-β does not elicit biphasic effect on the formation of ESC-derived vascular cells ( b). In agreement with this finding, when different doses of TGF-β were added to the culture of Flk1+ cells in the presence of VEGF, Smad2 phosphorylation was significantly induced and reached a maximum at 0.03 ng/ml of TGF-β, which remained at this level at higher doses; whereas Smad1/5 phosphorylation was significantly induced at 0.03 ng/ml, and reached a maximum at 1 ng/ml, which remained at this level at higher doses ( b). These results suggest that TGF-β does not elicit biphasic effects in the phosphorylation of Smads in ESC-derived vascular cells.
As expected, BMP7 led to phosphorylation of only Smad1/5 ( a, lane 5) whereas activin led to phosphorylation of only Smad2 ( a, lane 7). To evaluate whether TGF-β-induced Smad5 phosphorylation is specific to endothelial cells, we analyzed phosphorylation of Smads by TGF-β in ESC-derived mural cells. Although TGF-β induced Smad2 phosphorylation in mural cells ( a, lane 10), it failed to phosphorylate Smad1/5. BMP led to phosphorylation of Smad1/5 in mural cells ( a, lane 11). These results suggest that TGF-β induced phosphorylation of both Smad2 and Smad1/5 only in ESC-derived endothelial cells.
To confirm that Smad1/5 as well as Smad2 is activated by TGF-β in ESC-derived endothelial cells, nuclear translocation of Smad proteins was studied using an antiphospho-Smad1/5 antibody and an anti-Smad2 antibody. TGF-β induced nuclear translocation of Smad2, whereas BMP failed to do it ( c). In agreement with the results of Western blot analysis, both TGF-β and BMP induced nuclear translocation of phosphorylated Smad1/5 in ESC-derived endothelial cells ( c).
When SB-431542 was added in the presence of TGF-β or activin, Smad2 phosphorylation was decreased in a dose-dependent manner because of its inhibitory effect on ALK-5 and -4, respectively ( a, lane 4 and 8, respectively; d). We found that Smad1/5 phosphorylation induced by TGF-β was also decreased by SB-431542 ( a, lane 4). Because SB-431542 is not capable of inhibiting Smad phosphorylation by ALK-1 (Fig. S1), these results suggest that ALK-5 kinase activity is required for ALK-1–mediated phosphorylation of Smad1/5 in ESC-derived endothelial cells. Inhibition of Smad1/5 phosphorylation by SB-431542 is specific for ALK-1 because Smad1/5 phosphorylation by BMP was not decreased by SB-431542 ( a, lane 6).
Importantly, phosphorylation of Smad2 was weakly detected in the absence of exogenous ligands ( a, lane 1), suggesting that endogenous TGF-β or activin () acts on these cells. Moreover, addition of SB-431542 resulted in the decrease of phospho-Smad2 in these cells in a dose-dependent manner ( a, lane 2; e). SB-431542 substantially decreased Smad2 phosphorylation at 0.1 μM and reached a maximum at 1 μM, which remained at this level at higher doses ( e). In accordance with this result, the enhancement of endothelial sheet formation by SB-431542 was elevated gradually and reached a maximum at 1 μM ( c). These results strongly implicate the causal link between the states of Smad2 phosphorylation induced by TGF-β signals and endothelial sheet formation.
TGF-β regulates integrity of ESC-derived endothelial cells through claudin-5 expression
Next, we attempted to elucidate the molecular mechanisms by which endogenous TGF-β signals regulate the integrity of endothelial sheets. Integrity of cell–cell contacts is vital for functioning of endothelial cells as barriers and fences between the blood and extravascular components. Endothelial cell–cell junctions such as TJs are thought to control not only normal physiological conditions such as vascular permeability, leukocyte extravasation, and the formation and outgrowth of blood vessels, but also pathological conditions such as vasogenic edema (
Morita et al., 1999).
Because TGF-β signals regulate ESC-derived endothelial sheet formation ( b), we searched for the molecular targets of ALK-5 signals. Recently, we conducted oligonucleotide microarray analysis to identify target genes of ALK-1 and -5 in human umbilical vein endothelial cells (
Ota et al., 2002). One of the genes down-regulated by constitutively active ALK-5 was claudin-5, a component of TJ expressed exclusively in endothelial cells (
Morita et al., 1999).
To study the effects of TGF-β signals on claudin-5 expression in ESC-derived endothelial cells, we performed quantitative RT-PCR analysis against mural cells obtained in the absence of VEGF, and mixed populations of endothelial and mural cells differentiated in the presence of VEGF ( a). Claudin-5 expression was detected in the populations containing endothelial cells. TGF-β suppressed the expression of claudin-5, whereas SB-431542 strongly enhanced its expression ( a). Interestingly, TGF-β signals did not alter the expression of other components of TJs such as claudin-12 ( a;
Tsukita et al., 2001), or VE-cadherin (a component of adherens junctions; unpublished data), suggesting that effects of TGF-β signals on endothelial integrity may be primarily mediated by transcriptional regulation of claudin-5.
The above results were confirmed by immunofluorescence staining of claudin-5 ( b). Claudin-5 was concentrated at cell–cell borders of PECAM1+ endothelial cells ( b, top). When these endothelial cells were treated with TGF-β, expression of claudin-5 was decreased in accordance with the disruption of endothelial sheet structure, and claudin-5 was not detected at cell–cell borders ( b, middle). When these endothelial cells were treated with SB-431542, claudin-5 expression was strongly increased with the enhancement of endothelial sheet formation ( b, bottom).
TGF-β regulates formation of endothelial sheets derived from 8.5-dpc mouse embryos
To examine the roles of TGF-β signals during vascular development in mice, we tested the effects of TGF-β and SB-431542 on the differentiation of Flk1+ cells obtained from 8.5-dpc mouse embryos. When embryo-derived Flk1+ cells were cultured with 10% FCS, >95% cells became positive for SMA ( a). When 100 ng/ml VEGF was added, we were able to obtain PECAM1-positive sheets of endothelial cells. When TGF-β was added to the culture in the presence of VEGF, it inhibited the growth and integrity of endothelial cells. In contrast, SB-431542 strongly facilitated them. Subsequently, the effects of TGF-β signals on claudin-5 expression were examined by immunofluorescence staining ( b). Claudin-5 was concentrated at cell–cell borders of PECAM1+ endothelial cells. When these embryo-derived endothelial cells were treated with TGF-β, claudin-5 expression was decreased, whereas SB-431542 increased the expression of claudin-5, in accordance with alteration of endothelial sheet formation. These results strongly suggest that TGF-β signals play roles in vascular development in mouse embryos similar to those in ESC-derived vascular progenitor cells.