Differential Regulation of Transforming Growth Factor β Signaling Pathways by Notch in Human Endothelial Cells*

Notch and transforming growth factor β (TGFβ) play critical roles in endothelial-to-mesenchymal transition (EndMT), a process that is essential for heart development. Previously, we have shown that Notch and TGFβ signaling synergistically induce Snail expression in endothelial cells, which is required for EndMT in cardiac cushion morphogenesis. Here, we report that Notch activation modulates TGFβ signaling pathways in a receptor-activated Smad (R-Smad)-specific manner. Notch activation inhibits TGFβ/Smad1 and TGFβ/Smad2 signaling pathways by decreasing the expression of Smad1 and Smad2 and their target genes. In contrast, Notch increases SMAD3 mRNA expression and protein half-life and regulates the expression of TGFβ/Smad3 target genes in a gene-specific manner. Inhibition of Notch in the cardiac cushion of mouse embryonic hearts reduces Smad3 expression. Notch and TGFβ synergistically up-regulate a subset of genes by recruiting Smad3 to both Smad and CSL binding sites and cooperatively inducing histone H4 acetylation. This is the first evidence that Notch activation affects R-Smad expression and that cooperative induction of histone acetylation at specific promoters underlies the selective synergy between Notch and TGFβ signaling pathways.

During heart development, a subset of endocardial cells undergoes endothelial-to-mesenchymal transition (EndMT) 4 and migrates into the cardiac cushion to initiate valve formation (1). EndMT is regulated by multiple signaling pathways, including TGF␤ and Notch (1). Although both pathways play critical roles in cardiovascular development (2)(3)(4), their func-tional interaction in endothelial cells remains to be fully investigated. We have previously shown that Notch and TGF␤ synergistically induce expression of SNAIL and HEY1 in endothelial cells (5), both of which play roles in cardiac cushion development (6,7), suggesting functional integration between Notch and TGF␤ signaling pathways in endothelial cells during heart development.
TGF␤ is a multifunctional growth factor that is involved in many biological processes, including proliferation, differentiation, and apoptosis (8,9). The TGF␤ signal is transmitted through specific transmembrane type I and type II serine/threonine kinase receptors. Upon TGF␤ binding, the constitutively active TGF␤ type II receptor recruits and phosphorylates TGF␤ type I receptor, and the latter phosphorylates receptor-activated Smads (R-Smads), including Smad1, Smad2, Smad3, Smad5, and Smad8. The phosphorylated R-Smads then form a complex with a common Smad, Smad4, and translocate into the nucleus to regulate target gene expression through interaction with other cofactors (10). In endothelial cells, TGF␤ binds two distinct type I receptors, ALK1 (activin receptor-like kinase 1) and ALK5, to activate ALK1/Smad1/5/8 and ALK5/Smad2/3 signaling pathways. These two pathways regulate different genes and exert opposing biological functions in endothelial cells (11,12).
The evolutionarily conserved Notch signaling pathway determines cell fate by regulating multiple cellular processes, including proliferation, differentiation, and apoptosis (13,14). In mammals, four Notch receptors (Notch1-Notch4) and five ligands (Dll1 (Delta-like 1), Dll3, Dll4, Jagged1, and Jagged2) have been identified. Notch signaling is initiated by ligand binding, which triggers proteolytic cleavage of the transmembrane receptor and release of the Notch intracellular domain (NICD). Translocation of NICD into the nucleus results in association with the DNA-binding protein CSL and recruitment of coactivators, such as MAML (Mastermind-like) to initiate transcription (15)(16)(17).
Cross-talk between the Notch and TGF␤ pathways has not been studied in endothelial cells, where both the Smad1/5/8 and Smad2/3 pathways can be activated in the same cell via ALK1 and ALK5 receptors, respectively (11,12). Both synergy and antagonism between Notch and TGF␤ signaling have been reported in other cell types, and the interaction between Notch and TGF␤ signaling appears to be cell typeand context-dependent (18 -24). Further, in previous studies, Notch signaling was activated by overexpression of the constitutively active NICD. In the current studies, we have attempted to understand the functional consequences of coordinate TGF␤ and Notch activation at physiologic levels in the endothelium.
Dll4 (Delta-like 4) is the major Notch ligand expressed in endothelial cells (25), and Dll4 activation of Notch plays an important role in cardiovascular development (26). Here, we report for the first time that in endothelial cells, Notch activation by either NICD expression or co-culture of Dll4-express-ing cells regulates TGF␤ ALK1/ Smad1, ALK5/Smad2, and ALK5/ Smad3 signaling pathways by differentially affecting the expression of these R-Smads. Notch activation decreases the expression of Smad1 and Smad2 and their target genes, whereas Notch increases expression of Smad3 and regulates target genes in a gene-specific manner. Notch not only increases SMAD3 mRNA levels but also prolongs Smad3 protein half-life. In vivo endothelium-specific inhibition of Notch signaling by the expression of dominant negative MAML1 reduces Smad3 expression in the cardiac cushion of mouse embryonic hearts. Importantly, we demonstrate that simultaneous activation of TGF␤ and Notch signaling recruits Smad3 to both Smad binding elements (SBEs) and CSL binding sites and cooperatively increases histone H4 acetylation at promoters in which consensus cis elements of both pathways are present. Our findings provide a molecular mechanism for the synergistic induction of a subset of TGF␤ and Notch target genes, and demonstrate the crucial role of this interaction in vivo.
Cell Culture, Transfection, and TGF␤1 Treatment-The human microvascular endothelial cell line HMEC-1 (human microvascular endothelial cells (HMEC); Centers for Disease FIGURE 1. NICD affects the expression and TGF␤-induced phosphorylation of R-Smads. HMEC were transduced with empty vector or NICD. A, protein expression of R-Smads in whole cell lysates was examined by immunoblotting. B, R-Smad band intensity was measured by densitometry and normalized to tubulin. The relative density of R-Smad proteins was expressed as the -fold changes relative to the vector control and shown as the mean Ϯ S.E. of three or four independent experiments. *, p Ͻ 0.05. C, HMEC transduced with empty vector or NICD were left untreated (UT) or treated with 2.5 ng/ml TGF␤1 for 2 h. R-Smad mRNA level was examined by qRT-PCR and normalized to GAPDH. mRNA levels were expressed as -fold changes relative to the untreated vector samples and shown as mean Ϯ S.E. of three independent experiments. *, p Ͻ 0.05. D, HMEC transduced with empty vector or NICD were left untreated or treated with 2.5 ng/ml TGF␤1 for 1 h. The amount of total and phosphorylated R-Smad proteins in whole cell lysates was examined by immunoblotting. Tubulin was used as a loading control. E, R-Smad protein level in cytosolic and nuclear fractions was examined by immunoblotting. Tubulin and poly(ADP-ribose) polymerase (PARP) were used as loading controls for cytosolic and nuclear fractions, respectively.
Preparation of Whole Cell Lysates and Cytosolic and Nuclear Fractions-Whole cell lysates were prepared using modified radioimmune precipitation buffer as described previously (27). For cytosolic and nuclear fraction preparations, cells were lysed using Buffer A (10 mM HEPES-KOH, pH 7.8, 1.5 mM MgCl 2 , 10 mM KCl) containing 0.5% Nonidet P-40 and a protease inhibitor mixture (PIC; Roche Applied Science), incubated on ice for 10 min, and centrifuged at 12,000 ϫ g for 15 min. The supernatant was collected as the cytosolic fraction. The pellet was resuspended in Buffer C (50 mM HEPES-KOH, pH 7.8, 50 mM KCl, 300 mM NaCl, 0.1 mM EDTA, 10% glycerol) containing 1% Triton X-100 and 1ϫ PIC, incubated with rotating at 4°C for 20 min, and centrifuged at 12,000 ϫ g for 10 min. The supernatant was collected as the nuclear fraction. Proteins in cell lysates and cytosolic or nuclear fractions were quantified using the DC protein assay and used for immunoblotting.
RNA Isolation and Quantitative Reverse Transcription (qRT)-PCR-Total RNA was isolated using the GenElute TM Mammalian Total RNA Miniprep Kit (Sigma) and DNasetreated before cDNA was synthesized using SuperScript II reverse transcriptase reagent (Invitrogen) in the presence of RNase inhibitor. qRT-PCR was carried out using the SYBR green method on an Applied Biosystems 7900HT (Applied Biosystems, Foster City, CA). Sequences of the primers for qPCR are listed in supplemental Table 2.
Smad3 Expression in Mouse Cardiac Cushion Cells-Tet OS -dominant negative MAML1 (dnMAML1) transgenic mice were generated in house and will be described elsewhere. 5 dnMAML1 mice were crossed with VE-cadherin-tTA mice (28) (gift of L. Benjamin, Harvard Medical School, Boston, MA), and in the double transgenic offspring, inducible expression of dnMAML1 blocks Notch activity in an endothelial cell-specific manner. Inhibition of Notch activity by dnMAML1 expression was initiated at embryonic day 8.5 (E8.5) or E9.5 by removal of tetracycline from the drinking water, cellularization of the cardiac cushion was examined at E10.5 by staining cells with DAPI, and cell number was quantified using ImageJ Software (National Institutes of Health). To examine Smad3 protein expression in the cardiac cushion cells, Notch activity was inhibited by dnMAML1 expression at E9.5, and Smad3 protein expression was examined at E10.5 by immunofluorescence staining. Smad3 protein expression was quantified using ImageJ software and normalized to total cell numbers by counting the DAPI-stained nuclei in the same area.
Co-immunoprecipitation Assay-To examine the interaction between Smad3 and NICD, HEK 293T cells were transiently transfected with pLNCX-N4IC-HA, MSCV-pac-Smad3, or both. After 48 h, cell lysates were prepared using lysis buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1% Nonidet P-40, and 0.1% deoxycholate) containing protease inhibitors. After preclearing, cell lysates were incubated with anti-Smad3 or anti-HA antibody at 4°C overnight. The immunoprecipitation mixture was then incubated with protein A-agarose (for Smad3 antibody) or protein G-agarose (for HA antibody) beads for 2 h at 4°C. After three washes with lysis buffer, the beads were resuspended with 20 l of 2ϫ SDS sample buffer and heated to 95°C for 5 min. The supernatant was loaded onto 10% SDS-polyacrylamide gels for immunoblotting.
Analysis of Smad3 Protein Turnover-HMEC transduced with empty vector or NICD were treated with 50 g/ml of cycloheximide and cell lysates were prepared at the indicated times. Smad3 protein levels were examined by immunoblotting. The relative density of each Smad3 band compared with tubulin was expressed as the percentage of the untreated sample.  Chromatin Immunoprecipitation Assay-HMEC with or without Dll4 coculture were left untreated or treated with 2.5 ng/ml TGF␤1 for 1 h for Smad3 chromatin immunoprecipitation (ChIP) and 2 h for acetyl-histone H4 and trimethylated histone 3 on lysine 4 (H3K4Me3) ChIP. Cells were then crosslinked using 1% formaldehyde and harvested following lysis. After sonication, the lysates were diluted, and equal amounts of chromatin were used for ChIP with anti-Smad3, anti-acetylhistone H4, or anti-histone H3 (trimethyl Lys 4 ) and rabbit IgG as a negative control. Enrichment of DNA around the SBE or CSL binding sites (for Smad3 ChIP) or the proximal promoter region and/or the 5Ј-end of the genes (for acetyl-histone H4 and H3K4Me3) was detected using qPCR and normalized against the respective input DNA. Primer positions and sequences are listed in supplemental  Tables 3 and 4.
TGF␤ stimulation induces the phosphorylation of R-Smads, which then form protein complexes with Smad4 and translocate to the nucleus. Since NICD affects the expression of R-Smads, we expected that TGF␤-induced phosphorylation of R-Smads and nuclear translocation would be altered accordingly. As shown in Fig. 1D, NICD decreased phospho-Smad1 and phospho-Smad2 levels and increased phospho-Smad3 levels consistent with changes in total R-Smad protein expression. The differential effect of Notch activation on TGF␤-induced nuclear translocation of Smad1, Smad2, and Smad3 was confirmed by immunoblotting for Smad1, Smad2, and Smad3 in cytosolic and nuclear fractions after TGF␤ treatment (Fig. 1E).
Effects of NICD on TGF␤ Target Gene Expression are Pathway-and Gene-dependent-Since NICD affects the protein level, TGF␤-induced phosphorylation, and nuclear localization of R-Smads, we next examined the functional effects of NICD on the expression of TGF␤ target genes. qRT-PCR showed that FIGURE 3. Effects of Dll4-induced Notch activation on TGF␤ signaling. HMEC cocultured with Dll4-expressing HMEC at a 1:1 ratio were left untreated (UT) or treated with 2.5 ng/ml TGF␤1 for 1 h (B and C) or 2 h (A, D, E, F, and G). A, the mRNA level of Smad1, Smad2, Smad3, and Smad5 was examined by qRT-PCR and normalized to GAPDH. mRNA levels were expressed as -fold changes relative to the untreated vector sample and shown as mean Ϯ S.E. of three independent experiments. *, p Ͻ 0.05. B, the amount of total and phosphorylated R-Smad proteins in whole cell lysates was examined by immunoblotting. Tubulin was used as a loading control. C, the R-Smad proteins in cytosolic and nuclear fractions were examined by immunoblotting. Tubulin and poly(ADPribose) polymerase (PARP) were used as loading control for cytosolic and nuclear fractions, respectively. D-G, the mRNA level of target genes of TGF␤/Smad1 (D), TGF␤/Smad2 (E), and TGF␤/Smad3 (F and G) was examined by qRT-PCR and normalized to GAPDH. mRNA levels were expressed as -fold changes relative to the untreated vector samples and shown as mean Ϯ S.E. of three independent experiments. *, p Ͻ 0.05. activated Notch significantly reduced the basal and TGF␤-induced levels of ALK1/Smad1 target genes SMAD6, ID1, and ID2 ( Fig. 2A), and two ALK5/Smad2-dependent targets, MMP2 and NET1 (29,30) (Fig. 2B).
In contrast to inhibition of Smad1 and Smad2 targets, NICD alone up-regulated the ALK5/Smad3 target genes PAI1, CTGF, and CYR61 (29,31,32). Costimulation of Notch and TGF␤ only slightly induced (Ͻ2-fold) expression of PAI1, CTGF, and CYR61 compared with TGF␤ alone (Fig. 2C), suggesting a possible common signaling pathway through induction of Smad3. In contrast, a synergistic effect on the expression of ANKRD1, another TGF␤/Smad3 target gene (33), and HEY1 was observed between NICD and TGF␤ (Fig. 2D), similar to the effect we have previously noted for Snail (5). These results suggest that the effect of Notch activation on the expression of TGF␤/ Smad3 targets is gene-specific.
Effects of Dll4-induced Notch Activation on the Expression of R-Smads and Their Target Genes-To examine whether ligand-induced Notch activation would have similar effects as NICD on TGF␤ signaling pathways, we transduced HMEC with Dll4, the major Notch ligand in endothelial cells (25), and cocultured these cells with parental HMEC at a 1:1 ratio. In keeping with the NICD findings, qRT-PCR analysis showed that Dll4-induced Notch activation decreased SMAD1 and SMAD2 mRNA expression by 40 Ϯ 4.8% and 17 Ϯ 2.5%, respectively, and increased SMAD3 mRNA expression 122 Ϯ 3.2% but did not affect SMAD5 mRNA expression (Fig.  3A). Consistent with mRNA expression, Smad1 and Smad2 protein levels were decreased and Smad3 protein levels were increased by Dll4-induced Notch activation (Fig.  3B). Additionally, TGF␤-induced phosphorylation and nuclear translocation of Smad1 and Smad2 were also decreased and that of Smad3 was increased by Dll4-induced Notch activation (Fig. 3, B and C). Given that the nuclear levels of the R-Smads were proportional to cytosolic expression, it is likely that the changes in R-Smad phosphorylation and nuclear translocation are a direct reflection of R-Smad levels.
Similar to NICD, Dll4-induced Notch activation inhibited the basal and TGF␤-induced mRNA levels of the ALK1/Smad1 target genes SMAD6, ID1, and ID2 (Fig. 3D) and ALK5/Smad2 target genes MMP2 and NET1 (Fig. 3E). Dll4 minimally affected the basal mRNA levels of CYR61, CTGF, and PAI1 (Fig. 3F). As with NICD, Dll4-induced Notch activation synergized with TGF␤ upon inducing ANKRD1 and HEY1 expression (Fig. 3G) as well as on two classical direct Notch targets, HEY2 and HEYL (see supplemental Fig. 1). To ensure that the effects we observed in coculture of parental HMEC and Dll4-expressing HMEC reflect Dll4-induced Notch activation in parental HMEC, we flow-sorted the cocultured cells into green fluorescent protein-positive (Dll4expressing HMEC cells) and green fluorescent protein-negative (parental HMEC) populations after TGF␤ treatment and examined mRNA expression of R-Smads and TGF␤ target genes in these two populations by qRT-PCR. The results confirmed that the effect of Dll4 on TGF␤ signaling is due to Notch activation, since the effects were more pronounced in parental HMEC (receiving the Notch signal) than in Dll4-expressing HMEC (see supplemental Fig. 2). Taken together, our data indicate that, similar to NICD, Dll4-induced Notch activation has an inhibitory effect on Smad1 and Smad2 pathways but a positive effect on Smad3 levels, with synergistic induction of a subset of Smad3 target genes.

Inhibition of Notch Activity Reduces Smad3 Expression in the Mouse Embryonic Heart-A subset of endocardial cells initiates
EndMT at E9.5 in the atrioventricular canal, which is regulated in part by Notch and TGF␤ pathways (1,5). To investigate whether Notch activity affects the expression of Smad3 during EndMT in vivo, we examined Smad3 expression in cardiac cushions of embryonic hearts from mice inducibly expressing the pan-Notch inhibitor, dnMAML1, from an endothelial promoter. 5 As the mesenchymal cells are derived from the endocardium, Notch inhibition carries through into the mesenchymal cells of the cardiac cushion (34). 5 We first examined the effect of Notch inhibition on cellularization of the cardiac cushion. Notch activity was inhibited by inducing expression of dnMAML1 at E8.5 or E9.5, and cardiac cushion cells were stained by DAPI and quantified at E10.5. Inhibition of Notch activity by dnMAML1 for 24 and 48 h resulted in a reduction of cell numbers in cardiac cushion by 46 Ϯ 17.9 and 81 Ϯ 11.3%, respectively, compared with littermate controls, demonstrating that inhibition of endocardial Notch activity blocks EndMT (Fig. 4, A and B). We then examined Smad3 expression in these cardiac cushion cells. Notch activity was inhibited by inducing expression of dnMAML1 at E9.5, and Smad3 protein expression in cardiac cushion cells was examined at E10.5 by immunofluorescent staining. There was reduced nuclear Smad3 localization in the cardiac cushion cells (Fig. 4C) and an overall reduction in Smad3 protein expression of 34 Ϯ 9% per cell (Fig. 4D) when Notch activity was inhibited for 24 h compared with littermate controls. Thus, Notch activation regulates Smad3 expression in a physiologic context in vivo.
Notch Activation Stabilizes Smad3 Protein-Several pieces of data suggested to us that Notch also regulates Smad3 at the post-transcriptional level. First, Smad3 protein levels appeared disproportionately increased compared with the mRNA levels by NICD (Fig. 1, A and B compared with C) or by Dll4 (Figs. 3B and 6 compared with Fig. 3A). Additionally, TGF␤ stimulation increased SMAD3 mRNA to a similar level as that induced by Notch activation (Fig. 1C (131 Ϯ 7.2%) and Fig. 3A (129 Ϯ 11%)), but the increase of Smad3 protein was only observed in cells with active Notch and not in TGF␤-stimulated cells (Figs. 1D and 3B). To test whether NICD could increase Smad3 stability, we treated HMEC transduced with vector or NICD with 50 g/ml cycloheximide to block new protein synthesis and analyzed Smad3 protein expression by immunoblotting. As shown in Fig. 5A, loss of Smad3 protein was much slower in NICD-expressing cells than in the vector control cells, suggesting that NICD increases Smad3 protein stability. In contract, NICD did not increase the stability of Smad1 and Smad2 (Fig. 5A). Next, we examined whether expression of NICD could increase the pro-FIGURE 5. NICD interacts with Smad3 and increases protein stability. A, HMEC transduced with empty vector or NICD were treated with 50 g/ml cycloheximide for various times. Smad1, Smad2, and Smad3 protein level in whole cell lysates was examined by immunoblotting. The density of the Smad3 bands was measured by densitometry and normalized to tubulin. The relative density of Smad3 protein at each time point is shown as the mean Ϯ S.E. of four independent experiments. *, p Ͻ 0.05. B, 293T cells were transiently transfected with FLAG-Smad3 (2 g) with or without cotransfection of various amounts of NICD-HA (1, 2, or 3 g). Empty vector was used to equalize total plasmid concentration for each transfection. FLAG-Smad3 level was examined using anti-Smad3 or anti-FLAG antibodies. NICD expression was detected using anti-HA antibody. Tubulin was used as a loading control. The density of the FLAG-Smad3 and NICD bands was measured by densitometry and normalized to tubulin. The relative density of FLAG-Smad3 and NICD is shown as the mean Ϯ S.E. of three independent experiments. The density of FLAG-Smad3 in cells transfected with FLAG-Smad3 alone and that of NICD in cells transfected with the lowest amount of NICD were designated as 1 for FLAG-Smad3 and NICD, respectively. *, p Ͻ 0.05. C, 293T cells were transiently transfected with Smad3, NICD-HA, or both for 48 h. Physical interaction between Smad3 and NICD was examined in the whole cell lysates by co-immunoprecipitation using anti-Smad3 or anti-HA antibodies. IP, immunoprecipitation; IB, immunoblot.
tein level of ectopically expressed Smad3 whose expression is controlled by the viral LTR rather than the endogenous promoter. Transient transfection of 293T cells with vector or FLAG-Smad3 co-transfected with increasing amounts of NICD showed that NICD increased protein levels of ectopically transfected FLAG-Smad3 (Fig. 5B). Since the expression of FLAG-Smad3 is not controlled by the endogenous Smad3 promoter, the increase of FLAG-Smad3 protein by NICD can be attributed to regulation at a post-translational level. Although we confirmed the physical interaction between NICD and Smad3 by co-immunoprecipitation in transiently transfected 293T cells (Fig. 5C) (18,21), whether the stabilization of Smad3 by Notch is through physical interaction between Smad3 and NICD remains to be further investigated, because NICD has also been shown to interact with Smad1 and Smad2 (21,35).
Synergism between Dll4 and TGF␤ Signaling Requires both ALK5 and Activated Notch-To further study the synergy between Notch and TGF␤ signaling, we pretreated HMEC-Dll4 cocultures with an ALK5 kinase inhibitor SB431542 or a ␥-secretase inhibitor, DAPT, to block Notch activation prior to TGF␤1 stimulation. Immunoblotting showed that the synergistic up-regulation of ANKRD1 and Snail was abolished by either ALK5 or Notch inhibition (Fig. 6). However, ALK5 inhibition did not affect the induction of Smad3 by Dll4 but completely blocked TGF␤-induced Smad3 phosphorylation (Fig. 6), indicating that SB431542 effectively inhibits the TGF␤/ALK5/ Smad3 signaling pathway, independent of the effect of Notch on Smad3 protein levels. In contrast, inhibition of Notch signaling by DAPT abolished the induction of Smad3 expression by Dll4 and reduced TGF␤-induced Smad3 phosphorylation to the level seen by TGF␤ stimulation alone (Fig. 6). These data confirm the functional integration between Notch and TGF␤ signaling pathways and highlight the synergistic effect of Notch on a subset of Smad3-inducible genes.

Smad3 Is Recruited to both SBEs and CSL Binding Sites in the Promoters of Target Genes with Combined TGF␤ and Notch
Activation-The finding that only a subset of TGF␤/Smad3 target genes is synergistically induced by the combination of Notch activation and TGF␤ stimulation (Figs. 2D and 3G) prompted us to examine the mechanism underlying the selective synergy between Notch activation and TGF␤ signaling. We selected PAI1 as a primary TGF␤ target gene that is only induced by TGF␤, but not by Dll4, and is not synergistically induced by the combination (Fig. 3F). ANKRD1 and HEY1 were chosen as TGF␤ and Notch target genes that are synergistically induced by combined TGF␤ stimulation and Dll4/NICD (Fig.  3G). Since Smad3 is induced by Notch activation, we first examined whether the combined activation of TGF␤ and Notch pathways affects the level of occupancy of Smad3 on the promoters of these target genes. SBEs have been identified in PAI1 (31), ANKRD1 (33), and HEY1 (19) promoters, and verified or potential CSL binding sites have been identified in HEY1 (36) and ANKRD1 promoters but not in the PAI1 promoter up to 3 kb upstream of the transcriptional start site (see supplemental Table 1). To examine the occupancy of Smad3 on SBEs and CSL binding sites in these genes, we treated HMEC with or without Dll4 coculture with TGF␤1 for 1 h and examined Smad3 occupancy on SBEs and/or CSL binding sites in PAI1, ANKRD1, and HEY1 promoters using a Smad3 ChIP assay followed by qPCR using primers amplifying the SBEs or CSL binding sites. Smad3 ChIP-qPCR results showed that the occupancy of Smad3 on the SBEs in the PAI1 promoter was induced by TGF␤ stimulation and was not further increased by Dll4-induced Notch activation (Fig. 7), consistent with the mRNA expression data (Fig. 3F). Similarly, TGF␤-induced Smad3 occupancy on the SBE in the ANKRD1 promoter was not further increased by Dll4-induced Notch activation, suggesting that synergistic induction of ANKRD1 mRNA expression by both TGF␤ and Notch pathways is not through the regulation of Smad3 occupancy on the SBE in its promoter. In contrast, TGF␤-induced Smad3 occupancy on the CSL binding site in the ANKRD1 promoter was further increased by Dll4-induced Notch activation (Fig. 7), which corresponds to the synergistic induction of ANKRD1 mRNA expression by both pathways (Fig. 3G). Interestingly, Smad3 occupancy on both SBE and CSL binding sites in the HEY1 promoter was induced only when both TGF␤ and Notch pathways were activated (Fig. 7), which is consistent with our expression data indicating that the highest expression of HEY1 mRNA was induced when both pathways were activated (Fig.  3G). These results suggest that the presence of CSL binding sites in Smad3-dependent promoters is the determining factor that allows the recruitment of Smad3 to not only SBEs but also to CSL binding sites by combined TGF␤ stimulation and Notch activation, and this may explain the selective synergy between the TGF␤/ALK5/Smad3 and Notch pathways.
Dll4-induced Notch Activation and TGF␤ Signaling Cooperatively Regulate Histone Acetylation-We next investigated whether histone modification by TGF␤ and Notch could explain the synergistic up-regulation of specific target genes. Since both R-Smads and NICD interact with histone acetyltransferases (37,38) and Smad3 is recruited to both SBEs and CSL binding sites (Fig. 7) in the promoters of ANKRD1 and HEY1 by combined TGF␤ stimulation and Dll4-induced Notch activation, we examined whether TGF␤ and Notch signaling pathways would cooperatively induce histone acetylation in these target genes. HMEC were cultured with combinations of Dll4 and TGF␤1 stimulation for 2 h, and histone H4 acetylation was examined by acetyl-histone H4 ChIP followed by qPCR to amplify the proximal promoter and/or 5Ј-end of target genes. As shown in Fig. 8A, histone H4 acetylation in the proximal promoter and 5Ј-end of PAI1 was induced only by TGF␤ stimulation and not by Dll4, and consistent with the expression data, the combination of both did not further increase histone H4 acetylation. In contrast, and in keeping with the expression data, the combination of Dll4 and TGF␤ stimulation resulted in greater induction of histone H4 acetylation in the proximal promoter and/or 5Ј-end of ANKRD1 and HEY1 genes compared with either Dll4-or TGF␤-induced signaling alone (Fig. 8A). Thus, consistent with the simultaneous recruitment of Smad3 to both SBEs and CSL binding sites in ANKRD1 and HEY1 promoters (Fig. 7), our data demonstrate that combined TGF␤ and Dll4 stimulation induces greater histone acetylation in the proximal promoter and 5Ј-end of these genes.
To examine whether combined TGF␤ and Dll4 stimulation also affects H3K4Me3, which is another histone mark for active genes (39), histone H3K4Me3 ChIP-qPCR was performed and showed that TGF␤ stimulation induced H3K4Me3 only in PAI1 and not in HEY1 and ANKRD1 genes (Fig. 8B). Dll4 had no effect on H3K4Me3 in any of these genes (Fig. 8B). Thus, histone H4 acetylation, but not trimethylation of histone H3 Lys 4 , is involved in the synergistic up-regulation of specific Smad3 targets following combined Dll4 and TGF␤ stimulation.

DISCUSSION
In this study, we report that Notch activation differentially alters the expression of R-Smads. TGF␤ activates ALK1/ Smad1/5, ALK5/Smad2, and ALK5/Smad3 signaling pathways in endothelial cells. TGF␤/ALK1 and TGF␤/ALK5 pathways activate different target gene expression and play opposing roles in endothelial cells (11,12). In this regard, ALK1 phosphorylates Smad1/5 and promotes proliferation and migration of endothelial cells, whereas ALK5 phosphorylates Smad2/3 and inhibits proliferation and migration of endothelial cells. Smad2 and Smad3 have been shown to have both overlapping and distinct roles in regulating the expression of TGF␤ target genes and mediating TGF␤ functions in a cellular context-dependent manner (29,40,41). Thus, modulation of the balance of TGF␤ signaling pathways in certain cellular contexts can regulate distinct functional outcomes (42,43). Here, we provide the first evidence that Notch activation not only modulates the balance between TGF␤/ALK1 and TGF␤/ALK5 signaling pathways, but also fine tunes the ALK5/Smad2 versus the ALK5/Smad3 pathways. These data reveal a novel mechanism by which Notch activation modulates TGF␤ signaling pathways in an R-Smad-dependent manner in endothelial cells.
TGF␤ signaling is regulated at multiple levels, including R-Smad expression. It has been shown that endoglin affects TGF␤ signaling by increasing Smad2 protein levels without affecting its mRNA expression in endothelial cells, probably by inhibiting the ubiquitination and proteasome-dependent degradation of Smad2 protein (44). Here, we show that Notch acti- FIGURE 7. Smad3 occupancy on SBE and CSL sites in the promoters of Smad3/CSL target genes. HMEC with or without Dll4 co-culture were left untreated or treated with TGF␤1 for 1 h. Smad3 occupancy on SBE and/or CSL sites in PAI1 (top), ANKRD1 (middle), and HEY1 promoter (bottom) was examined by Smad3 ChIP with IgG as a negative control. ChIP-qPCR was conducted using primers that amplify the SBE or CSL sites (see supplemental Tables 1  and 3 for SBE and CSL sites and primer sequences). Smad3 occupancy on these sites was normalized against the respective input DNA and expressed as a percentage of input DNA. Values were shown as mean Ϯ S.E. of four independent experiments. *, p Ͻ 0.05.
vation not only increases Smad3 mRNA levels but stabilizes Smad3 protein as well. Thus, our data demonstrate that Notch activation regulates Smad3 expression at both transcriptional and post-translational levels.
Dll4 is the major ligand that activates Notch signaling in endothelial cells (25). Our data indicate that Dll4-induced Notch activation has similar inhibitory effects as that of NICD on TGF␤/Smad1 and TGF␤/Smad2 pathways. But the effect of Dll4-induced Notch activation on TGF␤/Smad3 target gene expression is less potent than NICD (compare Figs. 2C and 3F). The discrepancy between Dll4-induced Notch activation and overexpression of NICD on PAI1 expression suggests that the effects of Notch activation on TGF␤ signaling are probably dose-dependent. Nevertheless, both NICD and Dll4induced Notch activation show a strong synergistic effect with TGF␤ signaling in the up-regulation of a subset of Smad3 target genes. Importantly, genes that are synergistically induced by Notch activation and TGF␤ signaling, including HEY1, HEY2, HEYL, and SNAIL, have been shown to play critical roles in embryonic development (5,7,(45)(46)(47). In this regard, targeted deletion of Hey2, or double knockout of Hey1 and Hey2 or Hey1 and HeyL cause embryonic heart defects (7,45,46), and conditional deletion of Snail after E8.0 causes severe cardiovascular defects (47). Additionally, we have found that ANKRD1, a TGF␤/Smad3 target in smooth muscle cells (33), was synergistically induced by Notch activation and TGF␤ signaling in endothelial cells. Interestingly, the association between the disruption of ANKRD1 expression and the pathogenesis of total anomalous pulmonary venous return (TAPVR), a congenital heart defect, has also been suggested in a recent study (48). Here we show that endothelium-specific inhibition of Notch activity by the overexpression of dnMAML1 in vivo blocks EndMT and reduces Smad3 expression in cardiac cushion cells of mouse embryonic hearts, and this impingement of Notch on a second distinct signaling pathway may explain the severe cardiac defects seen in targeted Notch1 mutants (6).
Although synergism between Notch and TGF␤ signaling (18) and between Notch and BMP-6 (35) have been previously reported, the molecular mechanism underlying the synergy has not been clearly elucidated. In this study, we show that TGF␤-induced Smad3 occupancy on SBE in the ANKRD1 promoter is not further increased by Dll4-induced Notch activation, suggesting that synergistic induction of ANKRD1 by both TGF␤ and Notch signaling is not mediated by Smad3 occupancy on the SBE of its promoter. Rather, we found that Dll4-induced Notch activation increases TGF␤-induced Smad3 occupancy on the CSL site in the ANKRD1 promoter. Similarly, Smad3 occupancy on both SBE and CSL sites in the HEY1 promoter is induced only when both TGF␤ and FIGURE 8. Effects of TGF␤ and Dll4-induced Notch activation on histone modification. HMEC with or without Dll4 co-culture were left untreated or treated with TGF␤1 for 2 h. Histone H4 acetylation (AcH4) (A) and H3K4Me3 (B) were examined by ChIP followed by qPCR, using anti-acetyl-histone H4 and anti-histone H3 (trimethyl Lys 4 ) antibodies, respectively, with IgG as a negative control. ChIP-qPCR was conducted using primers that amplify the proximal promoter regions and/or 5Ј-ends of the PAI1 (top), ANKRD1 (middle), and HEY1 genes (bottom) (see supplemental Table 4 for primer sequences). The enrichment of these regions was calculated as a percentage of the respective input DNA concentration and expressed as relative signal after normalization against the untreated vector samples (designated as 1). Values are shown as the mean Ϯ S.E. of four independent experiments. *, p Ͻ 0.05.
Notch pathways are activated, in keeping with higher expression of HEY1 induced by combined TGF␤ stimulation and Dll4-induced Notch signaling as compared with either signal alone. In contrast, TGF␤-induced Smad3 occupancy of SBEs in the PAI1 promoter is not further increased by Notch activation, and a CSL binding site is not identified in the PAI1 promoter up to 3 kb from the transcriptional start site. In keeping with Smad3 occupancy on both SBEs and CSL sites, we found that combined TGF␤ stimulation and Dll4-induced Notch activation induces higher histone H4 acetylation in genes whose promoter contains both SBEs and CSL sites (ANKRD1 and HEY1) but not in PAI1 whose promoter contains only SBEs but not CSL binding sites. Since Dll4-induced Notch activation is required for Smad3 occupancy of CSL binding sites in the ANKRD1 and HEY1 promoters and the cooperative induction of histone acetylation surrounding the transcriptional start site of these genes, NICD must play a critical role in recruiting Smad3 to CSL binding sites, probably through the physical interaction between NICD and Smad3, as we have shown (Fig.  5C).
Because both Smad3 and NICD recruit histone acetyltransferases to their transcription complexes (37,38), formation of the Smad3/NICD/CSL complex may recruit histone acetyltransferases more effectively than Smad3 or NICD alone and thereby facilitate increased levels of histone acetylation and more active gene expression. In support of this, we found that combined TGF␤ and Dll4-induced Notch activation induces greater histone acetylation of ANKRD1 and HEY1 but not PAI1. Thus, the synergistic induction of a subset of target genes, such as ANKRD1 and HEY1, by combined TGF␤ stimulation and Notch activation is mediated by contemporaneous Smad3 occupancy on both SBEs and CSL binding sites in their promoters and consequently greater histone acetylation.
In summary, we demonstrate for the first time that Notch signaling differentially affects the expression of R-Smads and consequently modulates the balance between different TGF␤/ Smad signaling pathways in endothelial cells. We also demonstrate that the synergistic induction of a subset of genes by Notch and TGF␤ signaling is attributed to the simultaneous recruitment of Smad3 to both SBEs and CSL sites in the promoter of these target genes and the cooperative induction of histone acetylation. Our findings on the antagonism and synergy between Notch and TGF␤ signaling pathways shed light on the molecular mechanism underlying the functional interaction between these two important pathways.