Notch Signaling Pathway Enhances Bone Morphogenetic Protein 2 (BMP2) Responsiveness of Msx2 Gene to Induce Osteogenic Differentiation and Mineralization of Vascular Smooth Muscle Cells*

Vascular calcification is regulated in a process similar to bone formation. BMP2 (bone morphogenetic protein 2) is essential for osteoblastic differentiation of mesenchymal progenitor cells and thus has been implicated in the development of vascular calcification. Here we examined whether Notch signaling interacts with BMP2 signaling to regulate osteogenic differentiation and mineralization of vascular smooth muscle cells (SMCs). BMP2 alone scarcely induced the expression of alkaline phosphatase (ALP), an ectoenzyme crucially required for active biomineralization, in human aortic SMCs (HASMCs), despite its strong induction in osteoblast precursor MC3T3-E1 cells. Notably, overexpression of the Notch1 intracellular domain (N1-ICD) markedly enhanced BMP2-mediated induction of ALP activity and mineralization of HASMCs. In HASMCs, expression of Msx2 gene, a well documented BMP2 target gene in osteoblasts, was barely induced by BMP2 alone, and N1-ICD clearly enhanced the BMP2-driven Msx2 gene expression. Deletion and site-directed mutation analysis of Msx2 gene promoter revealed that the RBPJk-binding site was necessary for BMP2 responsiveness. Using the RBPJk-deficient cells and siRNA for RBPJk, we showed that RBPJk was required for BMP2 induction of Msx2 gene expression and ALP activity. Moreover, we showed that Smad1, a transcription factor downstream of BMP2 signaling, interacted with N1-ICD to form a complex within the Msx2 promoter. Immunohistochemistry of human calcifying atherosclerotic plaques revealed colocalized expression of Notch1, BMP2, and Msx2. These results indicate that the Notch intracellular domain·RBPJk complex enhances the BMP2-induced Msx2 gene expression by cooperating with Smad1 and suggest that Notch signaling makes vascular SMC responsive to BMP2 and promotes vascular calcification.

Vascular calcification is a common feature and is associated with aging, chronic kidney disease, diabetes, and dyslipidemia (1,2). In coronary arteries, calcium deposits influence atherosclerotic plaque stability and possibly the incidence of acute coronary syndrome, depending on the size and distribution of the deposits (3) In addition, vascular calcification, especially medial artery calcification, decreases compliance of the vessels, which increases the pulse pressure. It therefore increases cardiac work, which eventually leads to left ventricular hypertrophy and diastolic dysfunction (3). Thus, vascular calcification has a profound effect on the risk of cardiovascular mortality and morbidity.
Once considered to be a passive and unregulated process, vascular calcification is now known to be an active and tightly regulated phenomenon, in which a variety of osteogenic regulatory factors are involved (1,2,4). BMP2 (bone morphogenetic protein 2) has been implicated in vascular calcification; BMP2 was expressed in atherosclerotic plaques in accordance with increased expression of osteogenic transcription factors such as Msx2 and Runx2 and with decreased expression of SMC 2 markers (5,6). In mouse models of medial artery calcification, BMP2 and Msx2 gene expression is activated in aortic adventitia (7,8). Furthermore, matrix GLA protein inhibits vascular calcification by inhibiting BMP2 activity (9). Conversely, matrix GLA protein knock-out mice or rats treated with warfarin, an inhibitor of matrix GLA protein, develop severe arterial calcification (10,11). Despite such a crucial role of BMP2 in vascular calcification in vivo, it remains to be determined whether BMP2 is sufficient to induce vascular calcification.
The evolutionarily conserved Notch signaling pathway controls various cell fates by local cell-cell interactions (12). Notch ligands on the cell surface interact with Notch receptors in adjacent cells to cause cleavage of the Notch intracellular domain (NICD). NICD migrates into the nucleus and associates with RBPJk, which activates transcription of target genes (13). We and others reported that the Notch signaling pathway regulates SMC-specific gene expression in vascular SMCs (13)(14)(15). We have recently shown that Notch signaling induces osteogenic differentiation of vascular SMCs in a manner dependent on Msx2 (16), a key regulator of vascular calcification that was originally identified as a homeodomain transcription factor responsible for osteoblast differentiation and mineralization (17,18). Furthermore, Notch components, such as Notch1 and Jagged1, and Msx2 were strongly expressed in atherosclerotic/ fibrocalcific plaques (16).
Here, we tested the hypothesis that Notch signaling regulates BMP2-induced vascular calcification. We show that Notch signaling enhances osteogenic conversion and mineralization of vascular SMCs in the presence of BMP2, through a synergistic activation of Msx2 gene expression by Notch and BMP2 signaling pathways.

EXPERIMENTAL PROCEDURES
Materials-DAPT, a gamma secretase inhibitor that abrogates Notch signaling by interrupting cleavage of NICD upon ligand stimulation (14), was obtained from Calbiochem. Recombinant human BMP2 was kindly gifted from Astellas Pharma Inc. SigmaFAST TM BCIP/NBT, a substrate for ALP, was purchased from Sigma-Aldrich.
Cell Culture-Primary human aortic smooth muscle cells (HASMCs) were purchased from Kurabo. Mouse fibroblastic cell line C3H10T1/2 cells and mouse osteoblastic MC3T3-E1 cells were obtained from the RIKEN Cell Bank. These cells were maintained as described previously (16), and cells from passages 7-15 were used. Generation and culture conditions for RBPJk-deficient mouse embryonic fibroblasts (OT11) and the wild-type cell line (OT13) were as described previously (19,20). Cell lines stably expressing GFP or Jagged-1 (termed L-GFP and L-Jag1, respectively) were generated as described previously (20). Briefly, mouse fibroblasts termed L cells were infected with lentivirus-producing Jagged1 or GFP. Protein expression of L-GFP and L-Jag1 was confirmed by fluorescence microscopy and Western blotting as described previously (16).
Measurement of ALP Activity-ALP activity of various cells was measured using LabAssay ALP (Wako Pure Chemical Industries), according to the manufacturer's protocol. ALP activity was normalized to total protein determined with a Bio-Rad protein assay solution.
Von Kossa and Alizarin Red S Staining-The cells and human artery specimens were fixed with 4% formaldehyde. For von Kossa staining, they were exposed to 5% aqueous AgNO 3 . For Alizarin Red S staining, they were exposed to Alizarin Red S solution, the pH of which was adjusted to 4.1-4.3 using 0.5% ammonium hydroxide. To induce mineralization in the HASMC culture (Fig. 2, C and D), culture media were supplemented with P i , so that the media P i concentration was 1.8 mmol/liter. As for mineralization of C3H10T1/2 cells (Fig. 1, G  and H), ␤-glycerophosphate was added to the medium so that its concentration was 10 mmol/liter. Immunohistochemistry and Immunofluorescence-Carotid arteries were obtained by autopsy at Gunma University Hospital. This protocol was approved by the Institutional Review Board of Gunma University Hospital, and informed consent was obtained from the patients and their families. Tissue samples were fixed in 4% buffered formaldehyde and embedded in paraffin. Immunohistochemical staining of sections was performed using the CSA kit (DAKO), according to the manufacturer's protocol using antibodies against smooth muscle ␣-actin (M0851; DAKO), Msx2 (sc-15396; Santa Cruz Biotechnology), BMP2 (sc-6895; Santa Cruz Biotechnology), Runx2 (sc-10758; Santa Cruz Biotechnology), and Notch1 (sc-6014; Santa Cruz Biotechnology). For immunofluorescence, incubation with an anti-BMP2 antibody and anti-Notch1 (ab-27526; Abcam) was followed by incubation with Alexa Fluor 488conjugated anti-goat IgG (A11055; Invitrogen) and Alexa Fluor 555-conjugated anti-rabbit IgG (A21428; Invitrogen), respectively.
Generation of Adenoviruses-A control adenovirus (Ad-LacZ) and an adenovirus expressing the intracellular domain of Notch1 (termed Ad-N1-ICD) were created as described elsewhere (14). Protein expression by Ad-N1-ICD was confirmed by Western blot analysis (data not shown).
Luciferase Assay-C3H10T1/2 cells, OT11 cells, or OT13 cells were transfected with plasmid DNA by a modified calcium phosphate precipitation method as described previously (21). At 12 h after transfection, they were cocultured with L cells (L-GFP or L-Jag1) in the absence or presence of BMP2. After another 48 h, luciferase assays were performed according to the manufacturer's protocol (Promega). Transfection efficiency was determined by determining the proportion of GFP-expressing cells measured by fluorescence microscopy, which was ϳ25-50%.

BMP2 Induces ALP Activity in MC3T3-E1 Cells but Not in
HASMCs-In view of abundant in vivo evidence that BMP2 plays a crucial role in osteoblastic differentiation and vascular calcification, we attempted to determine whether BMP2 alone is sufficient to induce osteogenic conversion of HASMCs in vitro. To this end, ALP activity, a marker of osteogenic differentiation of HASMCs as well as osteogenesis, was determined. First, we used BCIP/NBT, a substrate for ALP, to visualize ALP activity in MC3T3-E1 and HASMCs. ALP activity was strongly induced in BMP2-treated MC3T3-E1 cells in a dose-dependent manner, whereas no apparent ALP induction was observed in BMP2-treated HASMCs (supplemental Fig. S1, A and B).
To determine more precisely the effect of BMP2 on induction of ALP activity, we next measured ALP activity in MC3T3-E1 cells and HASMCs. Consistent with the results in supplemental Fig. S1, A and B, MC3T3-E1 cells showed strong ALP activity in response to wide range of BMP2 concentration, whereas HASMCs did not show increased ALP activity (supplemental Fig. S1C).
Taken together, these results suggest that BMP2 alone is not sufficient to induce osteogenic differentiation of HASMCs in vitro. Accordingly, we hypothesize that, for BMP2 to exert its instructive role in vascular calcification as previously shown in in vivo experiments, an environmental cue is needed to alter the BMP2 responsiveness of vascular SMCs to undergo osteogenic differentiation.
Notch Confers BMP2 Responsiveness on HASMCs and C3H10T1/2 Cells to Induce ALP Activity and Mineralization-Given evidence that Notch and BMP signaling can synergistically regulate the Notch target gene expression (23,24), we next examined whether Notch signaling alters BMP2 responsiveness of HASMCs. As we previously reported, Notch1 is sufficient to induce Msx2 expression in HASMCs (16). To examine a synergistic activation of the Msx2 gene expression by Notch1 and BMP2, HASMCs were transduced with lower MOI of ad-N1-ICD than the previous experiments. As shown in Fig. 1B, we found that Ad-N1-ICD induced ALP activity by ϳ250-fold in the absence of BMP2, and interestingly, Ad-N1-ICD evoked an ϳ1200-fold increase in ALP activity in the presence of BMP2 (p Ͻ 0.0001 versus unstimulated cells; Fig. 1, A and B).
Because ALP is an ectoenzyme crucially required for active biomineralization rather than dystrophic calcification during tissue necrosis (1, 2), we next examined whether such a high expression of ALP induced by Notch and BMP2 leads to matrix mineralization in HASMCs. As we previously reported, Notch1 alone is sufficient to induce mineralization (16). Here we used culture medium with lower P i concentration, as well as lower MOI of ad-N1-ICD, to more precisely evaluate the synergistic effect of Notch1 on BMP2 responsiveness of HASMCs. Stimulation of HASMCs with both N1-ICD and BMP2 resulted in a marked matrix mineralization of HASMCs as assessed by Kossa staining (Fig. 1, C and D).
Previous studies indicate that cells other than HASMCs, such as myofibroblasts and pericytes, also participate in vascular calcification. Therefore, we next determined whether Ad-N1-ICD has a similar effect on BMP2-stimulated cells in terms of induction of ALP activity and matrix mineralization. To this end, we used C3H10T1/2 cells, mouse fetal fibroblasts. Similar to the findings shown in Fig. 1 (A-D), Ad-N1-ICD markedly promoted ALP activity and mineralization in BMP2stimulated C3H10T1/2 cells (Fig. 1, E-H). These results demonstrate that Notch signaling not only induces osteogenic differentiation by itself but also enhances the BMP2 responsiveness of SMCs to further augment this process.
Notch Confers BMP2 Responsiveness on the Msx2 Gene in HASMCs-Given that Msx2 plays a critical role in the osteogenic conversion of vascular SMCs (1,8,25), we tested whether Msx2 expression was regulated by Notch and BMP2 signaling. Consistent with the results in Fig. 1, in HASMCs transduced with LacZ, BMP2 had no effect on Msx2 mRNA and its protein levels (Fig. 2, A and B), whereas transduction of HASMCs with Ad-N1-ICD significantly induced Msx2 mRNA (ϳ7-fold, p Ͻ 0.001 versus LacZ) and protein expres-sion in the absence of BMP2, as we previously reported (16). Notably, semi-quantitative RT-PCR, Western blot analysis, and real time PCR revealed that stimulation with both Ad-N1-ICD and BMP2 resulted in significant increases in Msx2 mRNA (ϳ30-fold, p Ͻ 0.001 versus LacZϩvehicle) and protein levels (Fig. 2, A and B).
We next examined the effect of BMP2 and Notch ligand stimulation on Msx2 expression by performing coculture experiments of C3H10T1/2 cells with either L cells expressing GFP (L-GFP) or L cells expressing Jagged1 (L-Jag1). In C3H10T1/2 cells cocultured with L-GFP, BMP2 had minimal effect on Msx2 expression. Interestingly, BMP2 clearly induced Msx2 expression in C3H10T1/2 cells cocultured with L-Jag1, although the coculture with L-Jag1 by itself had minimal effect on Msx2 expression (Fig. 2C). Consistently, Ad-N1-ICD but not Ad-LacZ markedly induced Msx2 in BMP2-stimulated C3H10T1/2 cells (Fig. 2D). These results indicate that BMP2 induction of Msx2 expression is minimal, if at all present, in the absence of Notch stimulation in HASMCs and C3H10T1/2 cells, and stimulation of these cells with both BMP2 and acti-   Fig. 2 (E and F), Ad-N1-ICD induced ALP activity by ϳ700-fold versus baseline in siGFP-transfected HAMSCs in the presence of BMP2. On the other hand, such an enhanced ALP activity was significantly attenuated in HASMCs transfected with siMsx2. These results indicate that Msx2 mediates the robust induction of ALP activity by BMP2 and Notch1.

Notch Enhances BMP2-induced Msx2 Promoter Activity-
To determine the molecular mechanisms by which Notch increases the BMP2 responsiveness of the Msx2 gene expression, we used luciferase reporter constructs, Msx2-3.2k-Luc, which contains a 3.2-kb fragment (from Ϫ3212 to ϩ1) of the 5Ј-flanking sequence of the murine Msx2 gene, and Msx2-5.1k⌬3.3k-Luc, which contains a 1.8-kb fragment (from Ϫ5082 to Ϫ3298) of the enhancer sequence of the Msx2 gene in front of the SV40 promoter (Fig. 3A). Results showed that although Msx2-3.2k-Luc was unresponsive to either L-Jag1 or BMP2 stimulation (Fig. 3B), either L-Jag1 stimulation or BMP2 treatment up-regulated luciferase activities of Msx2-5.1k⌬3.3k-Luc  lower panel), respectively. C and D, C3H10T1/2 cells were either cocultured with L-GFP or L-Jag1 or infected with an MOI of 80 of either Ad-LacZ or Ad-N1-ICD. They were cultured in the absence or presence of 50 ng/ml of BMP2. On day 2, total RNA was extracted for semiquantitative PCR analysis. E and F, HASMCs were first transfected with siGFP or siMsx2 and the next day were infected with Ad-LacZ or N1-ICD in the absence or presence of BMP2 as for A and B. ALP activity was visualized with BCIP/NBT on day 4 (E) or measured and normalized by protein amount (F). These assays were repeated three times. by ϳ6and 3-fold, respectively (Fig. 3C). Notably, stimulation with both Jag1 and BMP2 induced an ϳ15-fold up-regulation; incremental induction of luciferase activity was abrogated by DAPT, a Notch signaling-specific inhibitor (Fig. 3C). Similarly, Msx2-5.1k⌬3.3k(mRBS)-Luc, which carries a mutated RBPJkbinding site (16), showed increased activity by BMP2, but the increase in luciferase activity was not further enhanced by L-Jag1 stimulation (Fig. 3D). These results suggest that Notch and BMP2 signaling cooperatively activate Msx2 gene expression at the transcription level and that sequence spanning from Ϫ5082 to Ϫ3289 within the Msx2 promoter mediates this effect.

Both RBPJk-binding Site (RBPJk-BS) and Smad-binding Element (SBE) Are Indispensable for the Enhanced Msx2 Gene
Transcription Induced by Notch and BMP2-In addition to the RBPJk-binding site at Ϫ3794 (16), the Msx2 promoter contains SBE between Ϫ3523 and Ϫ3471 (Fig. 4A), which confers BMP responsiveness on a reporter gene in cultured cells and in transgeneic embryos (22). Therefore, we investigated whether these sites can mediate cooperative regulation of Msx2 gene expression by BMP2 and Notch1. As expected, Msx2-4.6k⌬3.3k-Luc, which contains both the RBPJk-binding site and SBE, showed a synergistically increased activity in response to L-Jag1 stimulation and BMP2 (Fig. 4B). Similarly to Msx2-5.1k⌬3.3k (mRBS)-Luc (Fig. 3D), Msx2-3.7k⌬3.3k-Luc, which lacks the RBPJkbinding site showed increased activity by BMP2, but the increase in luciferase activity was not further enhanced by L-Jag1 stimulation (Fig. 4C). Furthermore, Msx2-4.6k⌬3.3k⌬SBE-Luc, which lacks SBE but contains the binding site for RBPJk, did not show synergistically provoked activity (Fig. 4D). These results suggest that both the RBPJk-binding site and SBE were required for the full activation of Msx2 transcription by Notch1 and BMP2.
RBPJk Mediates the Enhanced Msx2 Expression by Notch and BMP2-To test the hypothesis that RBPJk is required for the enhanced expression of Msx2 gene by Notch1 and BMP2, we employed mouse fibroblast cell line OT11, which carries homozygous mutations at the RBPJk loci, and cells of its parental wild-type cell line, OT13. BMP2 dramatically induced Msx2 gene expression and ALP activity in OT13 cells transduced with N1-ICD (Fig. 5, A and C), as well as in HAMSCs and C3H10T1/2 cells. More importantly, BMP2 stimulation had no effects on the induction of Msx2 gene expression and ALP activity in OT11 cells (Fig. 5, A and C). We next examined whether ligand stimulation of Notch receptor induces Msx2 promoter activity using coculture of OT 11 or OT 13 cells with L-Jag1 or L-GFP cells. As shown in Fig. 5E, coculture of OT13 cells with L-Jag1 cells induced luciferase activity of Msx2-5.1k⌬3.3k-Luc to increase by ϳ2.5-fold in the absence of BMP2, and BMP2 stimulation alone induced this reporter activity 2.5-fold in OT13 cells (Fig. 5E). BMP2 stimulation of OT13 cells cocultured with L-Jag1 induced a 10-fold increase (Fig. 5E). In contrast, neither BMP2 stimulation nor coculture with L-Jag1 induced Msx2 promoter activity in OT11 cells (Fig.  5E). As a consequence, simultaneous stimulation with BMP2 and Jagged1 had little if any effect on the induction of Msx2 promoter activity in OT11 cells (Fig. 5E). Finally, we tried to confirm the effect of RBPJk silencing on osteogenic conversion of HASMCs. As shown in Fig. 5 (B and D), specific knockdown of human RBPJk almost completely abrogated Msx2 expression and ALP activity induction by Notch1 and BMP2. Taken together, these findings support our hypothesis that the N1-ICD⅐RBPJk complex is necessary for the induction of Msx2 expression and ALP activity in response to BMP2 and hence for the synergistic induction by BMP2 and Notch signaling pathways.
N1-ICD⅐RBPJk Complex Physically Interacts with Smad1-To determine the molecular mechanism by which BMP2 and Notch1 cooperatively induce Msx2 gene expression in an RBPJk-dependent manner, we tested the possibility that the N1-ICD⅐RBPJk complex formed at the RBPJk-binding site within Msx2 promoter (Msx2 RBPJk-BS) interacts with Smad1, a Smad species that is activated by BMP2, at the Smad-binding element within Msx2 promoter (Msx2 SBE). C3H10T1/2 cells were transfected with FLAG-tagged Smad1 and HA-tagged NICD, and cross-linked chromatin DNA was immunoprecipi-tated with control IgG or indicated antibodies, followed by PCR amplification using mouse Msx2 RBPJk-BS or Msx2 SBE sequence primer. First of all, we confirmed that N1-ICD and Smad1 bound to their cognate binding sequence, Msx2 RBPJk-BS and Msx2 SBE, respectively (Fig. 6, B and C). Further experiments showed that the Msx2 RBPJk-BS was immunoprecipitated with anti-FLAG antibody in cells that received both FLAG-Smad1 and HA-NICD constructs (Fig. 6B). Similarly, Msx2 SBE was immunoprecipitated with anti-HA antibody (Fig. 6C). Furthermore, Msx2 SBE was immunoprecipitated with anti-RBPJk antibody, which recognizes endogenous RBPJk protein, in the presence of both Smad1 and NICD (Fig. 6D). These results suggest that the N1-ICD⅐RBPJk complex physically interacts with Smad1 at the specific DNA sequences within the Msx2 promoter.
Notch1 and BMP2 Are Coexpressed in Human Fibrocalcified Atherosclerotic Plaques-To elucidate the mechanism by which BMP2 induces vascular calcification, we next performed immunohistochemistry of human carotid arteries. Representative immunohistochemistry results are shown in Fig. 7. Typical atherosclerotic plaque was observed (Fig. 7A), and mineral depositions within the plaques (Fig. 7B) were evident, suggesting that the plaques were undergoing a process of atherosclerotic/fibrocalcific calcification. Consistent with previous studies, BMP2 was expressed in the atherosclerotic plaques (Fig.  7D). In addition, Notch1, a receptor of Notch signaling, as well as osteogenic transcription factor Msx2, a common target of BMP2 and Notch signaling, were also observed within the ath- erosclerotic plaques (Fig. 7, E and F). Interestingly, the BMP2positive area was macroscopically colocalized with that of Notch1.
To assess the colocalization of BMP2 and Notch1 expression more accurately, we then performed immunofluorescence. As shown in Fig. 7 (I and J), BMP2-positive cells and Notch1-positive cells were observed at the edge of the plaques. Furthermore, BMP2-positive cells largely overlapped with Notch1positive cells (Fig. 7K).

DISCUSSION
In the present study, we showed several lines of evidence indicating that the cross-talk between BMP2 and Notch signaling pathways induces osteogenic differentiation of vascular SMCs. First, synergistic activation of ALP activity by BMP2 and activated Notch was completely abrogated in RBPJk-deficient cells or siRBPJk-transfected HASMCs. Second, pharmacological inhibitor DAPT nullified the synergistic induction of ALP activity by BMP2 and L-Jag1. Third, mutation and deletion of the RBPJk-binding site within the Msx2 enhancer region abolished the synergistic induction of the Msx2 gene expression by BMP2 and Notch. Fourth, immunohistochemistry revealed colocalization of BMP2 and Notch1 in human calcifying atherosclerotic plaques.
BMP2 exerts its remarkable effects on mesenchymal stem cells and commits them to osteoblast differentiation. Msx2, a transcription factor originally found crucial for craniosynostosis, has a function that commits mesenchymal cells to osteogenesis but not to adipogenesis (25). Much like the expression of factors related to osteogenesis, BMP2 and Msx2 are expressed in cells in calcifying atherosclerotic lesions (6). Based on the previous observation that BMP2 activates osteoblast-specific genes such as Runx2 and Dlx2 in C2C12 premyoblasts that finally commit these cells to trans-differentiation into osteogenic cells (26,27), BMP2 has been postulated to elaborate the FIGURE 5. RBPJK protein mediates the enhanced Msx2 expression induced by Notch and BMP2. A and C, RBPJk-deficient cells (OT11) and the wild type (OT13) were infected with either an MOI of 80 of Ad-LacZ or Ad-N1-ICD in the absence or presence of 50 ng/ml BMP2. Expression of Msx2 and ALP activity was evaluated by semiquantitative RT-PCR analyses on day 2 (A) or visualized on day 4 (C). B and D, HASMCs were first transfected with siGFP or siRBPJk and the next day were infected with Ad-LacZ or N1-ICD (an MOI of 20) in the absence or presence of BMP2 (200 ng/ml). Expression of Msx2 and ALP activity was evaluated by semiquantitative RT-PCR analyses on day 2 (B) or visualized on day 4 (D). E, after OT11 cells and OT13 cells were transiently transfected with Msx2-5.1k⌬3.3k-Luc, they were cocultured with either L-GFP or L-Jag1 in the presence or absence of BMP2. Relative luciferase activity compared with that of an empty SV40-luciferase vector is presented. These experiments were repeated three times. osteogenic regulatory programs in the arterial tree. However, the precise effects of BMP2 on the osteogenic differentiation of vascular SMCs remain to be determined. Here, we demonstrate that BMP2 alone is not sufficient to induce ALP activity and Msx2 expression in HASMCs, despite the remarkable induction of these genes in osteoblast MC3T3-E1 (supplemental Fig.  S1). The inability of BMP2 to elicit these responses in HASMCs was not explained by a simple decrease in ligand sensitivity of HASMCs because the same concentration of BMP2 efficiently induced ALP and Msx2 expression in HASMCs transduced with N1-ICD ( Figs. 1 and 2). Furthermore, the addition of neutralizing antibody against BMP2 had no effect on Notch-induced Msx2 expression (data not shown), suggesting that an increase in BMP2 activity or concentration in supernatants of HASMCs expressing NICD does not play a major role in the synergistic induction of osteogenic differentiation of HASMCs.
In addition to well established role of Notch signaling as a determinant of cell fate and the resultant formation of various organ systems (13,28), there have also been several reports describing the cooperative interaction between Notch and other signaling pathways such as BMP. In mouse neuroepithe-lial cells, BMP2 stimulation of NICD-expressing cells showed enhanced expression of the Hes-5 and Hesr-1 genes, which play a role in keeping the progenitors from differentiating into neurons (23). Likewise, BMP2 potently induced the expression of Herp2, which inhibits migration of endothelial cells in the presence of activated Notch (24). Here we reported for the first time cross-talk between Notch and BMP2 signaling in vascular SMCs and C3H10T1/2 cells and that such cross-talk plays a crucial role in osteogenic conversion and mineralization of these cells through enhanced activation of the Msx2 gene, a master regulator of vascular calcification. The important point to note is, however, that the function of Notch signaling and Msx2 is controversial; Notch and Msx2 have been reported to regulate, either positively (25,29,30) or negatively (31)(32)(33)(34), osteoblastic differentiation and skeletal development. Given that Msx2 generally induces ALP activity in C3H10T1/2 and C2C12 and suppresses it in MC3T3-E1, it is assumed that they exert different effects on cellular differentiation depending on the cell type and stage of cell development. In fact, Msx2 activates SMC-specific genes in mesoangioblastic and epithelial cell lines (35) but inhibits such genes in vascular SMCs (36). Our previous data also support this assumption, in which we showed that Notch induces ALP activity in HASMCs and C3H10T1/2 but decreases it in MC3T3-E1 (16). More specifically, we can envisage that cell lineage-specific transcription  factors are responsible for such discrepant regulation, and obviously, further studies are warranted.
We identified the RBPJk-binding site 5Ј-GTGGGAA-3Ј at Ϫ3794, which is located in close proximity to the previously reported SBE at Ϫ3523 (Fig. 4). Accordingly, we hypothesize that the synergistic effect of Notch and BMP2 in the regulation of Msx2 gene expression might be caused by cooperation among transcription factors recruited to the RBPJk site and SBE. In fact, ChIP assay showed that Smad1 binds to the RBPJkbinding site in the presence of N1-ICD, and N1-ICD binds to the SBE in the presence of Smad1 (Fig. 6), supporting the hypothesis described above. Such functional cooperation between Notch and BMP2 signaling seems to be compatible with the reports that synergy between the two signaling pathways is attributable to the physical interaction between NICD and Smad (23,24).
Our observation that BMP2 requires Notch signaling to exert its osteogenic effect on vascular SMCs may be relevant to the pathogenesis of vascular calcification. An increasing amount of evidence indicates that macrophage activation contributes to vascular calcification. By using fluorescence reflectance imaging, Aikawa et al. (37) demonstrated that macrophage infiltration precedes osteogenic activity and promotes calcification in aortas of apolipoprotein E-deficient (apoE Ϫ/Ϫ ) mice. In addition, Fung et al. (38) showed that the Notch ligand Delta-like 4 and Jagged1 as well as Notch3 increased in macrophages stimulated with LPS in a Toll-like receptor 4-and NF-B-dependent manner. These results concur with our hypothesis that macrophage-derived BMP2 and Notch ligands send specific signals to SMCs to initiate osteogenic differentiation in the tunica intima of atherosclerotic plaques where the cell-cell contact between macrophages and SMCs takes place. Indeed, immunohistochemistry revealed prominent reactivity against anti-BMP2 and anti-Notch1 antibodies in human atherosclerotic plaques. We found that BMP2, Notch1, and Msx2 were coexpressed in atherosclerotic plaques in high fat-fed apoE Ϫ/Ϫ mice (data not shown), suggesting apoE Ϫ/Ϫ mice as an appropriate model for atherosclerotic/fibrocalcific calcification. Furthermore, given that adventitial BMP2/Msx2 signaling contributes to medial artery calcification, Notch and BMP2 signaling cross-talk is expected to be widely involved in vascular calcification, and this concept should be examined in future studies.
Msx2 was expressed in areas where mineralization had not clearly developed (Fig. 7F). Because Msx2 has been reported to function at the very early phase of bone formation and diminishes in later stages (39), Notch/BMP2-induced Msx2 activation may play a role during the early phase of atherosclerotic/ fibrocalcific calcification. It is interesting to note that Notch1, BMP2, and Msx2 expression was detected within atherosclerotic plaques where smooth muscle ␣-actin was stained negative (Fig. 7C). This is consistent with previous reports demonstrating that vascular calcification is accompanied by downregulation of SMC marker genes (4,40). A potential explanation for this observation is that Runx2 acts as a repressor for SMC marker gene expression in calcifying vascular SMCs. We have recently shown that Runx2 inhibits SMC marker gene expression by inhibiting SRF/myocardin-dependent transcription in HASMCs (41). Indeed, Runx2 stained positively in the lesions where BMP2, Notch1, and Msx2 were expressed but smooth muscle ␣-actin was negative. Speer et al. (42) have convincingly shown that osteochondrogenic precursor-and chondrocyte-like cells in calcified blood vessels are derived from SMCs using SM22-Cre mice, which allow the cells derived from smooth muscle cells to be traced. They showed that SMC-derived osteoblastic cells were negative for SMC lineage markers but were positive for Runx2.
In summary, our data demonstrate that BMP2 requires Notch signaling to induce osteogenic differentiation of vascular SMCs. Synergistic induction of the Msx2 gene through formation of a complex containing NICD⅐RBPJk and Smad1 is the underlying mechanism. These findings provide a novel insight into the role of the Notch signaling pathway in BMP2-driven vascular calcification.