Genome-wide Approaches Reveal Functional Vascular Endothelial Growth Factor (VEGF)-inducible Nuclear Factor of Activated T Cells (NFAT) c1 Binding to Angiogenesis-related Genes in the Endothelium*

Background: VEGF-calcineurin-nuclear factor of activated T cells (NFAT) signaling regulates endothelial cell homeostasis. Results: Novel NFAT targets were identified by genome-wide screening and epigenome analysis. Conclusion: Inhibition of the NFAT targets led to blockade of VEGF-mediated functional angiogenesis. Significance: VEGF-NFAT genome-wide occupancy may be useful for the development of target therapies for VEGF-mediated vascular diseases. VEGF is a key regulator of endothelial cell migration, proliferation, and inflammation, which leads to activation of several signaling cascades, including the calcineurin-nuclear factor of activated T cells (NFAT) pathway. NFAT is not only important for immune responses but also for cardiovascular development and the pathogenesis of Down syndrome. By using Down syndrome model mice and clinical patient samples, we showed recently that the VEGF-calcineurin-NFAT signaling axis regulates tumor angiogenesis and tumor metastasis. However, the connection between genome-wide views of NFAT-mediated gene regulation and downstream gene function in the endothelium has not been studied extensively. Here we performed comprehensive mapping of genome-wide NFATc1 binding in VEGF-stimulated primary cultured endothelial cells and elucidated the functional consequences of VEGF-NFATc1-mediated phenotypic changes. A comparison of the NFATc1 ChIP sequence profile and epigenetic histone marks revealed that predominant NFATc1-occupied peaks overlapped with promoter-associated histone marks. Moreover, we identified two novel NFATc1 regulated genes, CXCR7 and RND1. CXCR7 knockdown abrogated SDF-1- and VEGF-mediated cell migration and tube formation. siRNA treatment of RND1 impaired vascular barrier function, caused RhoA hyperactivation, and further stimulated VEGF-mediated vascular outgrowth from aortic rings. Taken together, these findings suggest that dynamic NFATc1 binding to target genes is critical for VEGF-mediated endothelial cell activation. CXCR7 and RND1 are NFATc1 target genes with multiple functions, including regulation of cell migration, tube formation, and barrier formation in endothelial cells.

The endothelium is a highly malleable cell layer that constantly senses and responds to changes in the extracellular microenvironment. Many extracellular mediators modulate gene transcription, resulting in changes in cell migration, proliferation, leukocyte adhesion, and coagulation. Tight control of these processes is essential for maintaining homeostasis. Endothelial cell activation, if excessive, oversustained, or misplaced spatially and temporally, may lead to vascular diseases such as pathological angiogenesis, inflammation, and atherosclerosis. Therefore, an understanding of the molecular pathways regulating endothelial activation may provide novel insights into therapeutic targets (1).
Endothelial cell activation is associated with small G proteinmediated signaling involving RhoA activation, which contributes to vascular permeability and endothelial cell motility. Rnd1, a Rho GTPase, regulates actin filament disassembly in response to various extracellular stimuli (2). Overexpression of Rnd1 resulted in a "rounded" cell shape in fibroblasts, and these morphological changes were due to inhibition of RhoA activation because of the anchoring p190 RhoGAP to the plasma membrane (3). In addition, activated endothelial cells dynamically sense attractive/repulsive chemokines via several G protein-coupled receptors. Specific chemokines, such as stromal derived factor (SDF) 1, CXCL1, and thrombin, regulate endothelial cell integrity and permeability (4,5). Two receptors for SDF-1, CXCR4 and CXCR7, have been identified (6,7). In particular, CXCR7 has been demonstrated recently to be highly expressed on lung and breast cancer cells and tumor vascular endothelia (8,9). Expression of CXRC7 correlates with tumor malignancy and metastasis.

EXPERIMENTAL PROCEDURES
Cell Culture-Human umbilical vein endothelial cells (HUVEC) and human dermal microvascular endothelial cells were purchased from Lonza and cultured in EGM-2 MV medium (Lonza). We used three independent lots of HUVEC, and human dermal microvascular endothelial cells were a mixture from multiple donors. HEK-293 (ATCC, catalog no. CRL-1573), COS7 cells (ATCC, catalog no. CRL-1651), and human primary skin fibroblasts (Lonza) were cultured in DMEM supplemented with 10% FBS and penicillin/streptomycin.
Immunohistochemistry-HUVEC were cultured in EGM-2 MV medium, serum-starved for ϳ18 h, and then pretreated with 1 M cyclosporin A (CsA) or 10 M MG-132 (Calbiochem) for 30 min before stimulation. Cells were treated with 50 ng/ml VEGF for the indicated times, fixed with 4% paraformaldehyde, and permeabilized with ice-cold methanol and 0.1% Triton X-100 in PBS. Samples were incubated with our generated anti-NFATc1 antibody, anti-CXCR7 (GeneTex), or anti-PECAM1 (BD Biosciences), washed with PBS, and then treated with Alexa Fluor 488-or Alexa Fluor 594-conjugated secondary antibodies (Invitrogen). Slides were mounted by ProLong Gold with antifade reagent with DAPI (Invitrogen). Images were captured with CoolsnapHQ charge-coupled device camera (Roper Scientific) at four independent objectives and quantified by examination with Metamorph software (Molecular Devices).
DNA Microarray-HUVEC were serum-starved overnight, pretreated with 1 M CsA for 30 min, and then stimulated with 50 ng/ml VEGF (Peprotec) for 1 h. RNA was extracted with TRIzol (Invitrogen). Preparation of cRNA and hybridization of probe arrays were performed according to the instructions of the manufacturer (Affymetrix). Data were analyzed according to the Minimum information about a microarray experiment (MIAME) rule.
Western Blot Analyses-Subconfluent HUVEC were serumstarved for ϳ18 h and then pretreated with 1 M CsA or 10 M MG-132 for 30 min before VEGF stimulation. 1 h after 50 ng/ml VEGF was added, cells were harvested, and these extracts were divided into nuclear and cytosol fractions as described previously (17,18). 15 g of each protein was loaded in 10% SDS-PAGE and immunoblotted with anti-NFATc1 antibody. The PVDF membrane was strippedandreprobedwithanti-LaminA(SantaCruzBiotechnology) and Hsp 90 (Cell Signaling Technology) as nuclear and cytosol internal control markers, respectively.
HUVEC or COS7 cells were transiently transfected with plasmid DNA using FuGENE HD reagents (Promega), and luciferase activity was detected by the Dual-Luciferase assay kit (Promega) as described previously (17). All data were normalized by Renilla luciferase luminescence derived from the cotransfected pRL-SV40 vector (Promega).
Cell Migration Assay-HUVEC were treated with siRNAs for 24 h, incubated with EBM-2 (Lonza) plus 0.5% FBS for 18 h, stimulated by 50 ng/ml VEGF for 1 h, and labeled with PKH26 red fluorescent dye (Sigma-Aldrich). Migration assays were carried using the BD Biocoat angiogenesis system (BD Biosciences). Labeled cells were seeded on the upper chamber (10 5 cells/250 l of EBM-2 plus 5 ng/ml VEGF) and incubated with 750 l of EBM-2 plus 5 ng/ml VEGF in the presence or absence of 100 ng/ml SDF-1 (R&D Systems) in the lower chamber. After 8 or 24 h, migrated cells were visualized under a fluorescent microscope (Nikon) and quantified using a fluorescence detection cell image analyzer (Kurabo).
siRNA Treatment and Scratch Migration Assay-HUVEC were treated with siRNA for 24 h and incubated with EBM-2 plus 0.5% FBS for 18 h. After 50 ng/ml VEGF stimulation for 1 h, the confluent cell layer was scratched by a small tip (1-mm diameter). Resulting cell plates were incubated with 100 ng/ml SDF-1 for 24 h. Cells that migrated into the scratched area were counted under a phase-contrast microscope (Nikon) as described previously (19).
Flow Cytometry Analysis-siRNA-treated HUVEC were harvested by 1 mM EDTA (pH 8.0) and incubated with phycoerythrin (PE)-conjugated anti-CXCR7 or anti-CXCR4 antibody (BioLegend) for 30 min on ice. Samples were washed three times, resuspended with PBS, and then analyzed immediately by a flow cytometer (Merck Millipore).
Permeability Assay-siRNA-transfected HUVEC were seeded on transwell inserts with 8-m pores (BD Biosciences) and then cultured in EGM-2 MV. After reaching confluence in dishes, HUVEC were serum-starved for ϳ18 h and treated with 50 ng/ml hVEGF plus 1% FITC-dextran for 1 h. The lower chamber aliquots were collected and measured with an absorption spectrometer at A 490 .
Aortic Ring Assay-48-well plates were covered with growth factor-reduced Matrigel (BD Biosciences) (100 l/well) and incubated for 30 min at 37°C, 5% CO 2 . Thoracic aortas were excised from C57/BL6j male mice at 6 -8 weeks of age and transferred into a dish with sterilized 1ϫ PBS. After removing the fibroadipose tissue, arteries were sectioned into 1-mm-long cross sections, rinsed five times with EBM-2, and placed on the Matrigel-coated wells. Artery rings were covered with an additional 100 l of Matrigel and incubated in 1.5 ml of EBM-2 plus 0.5% FBS for 24 h. Aortic rings were transfected with siRNA using Lipofectamine RNAi max (Invitrogen) and then cultured with EBM-2 containing 50 ng/ml mouse VEGF (PeproTech) and 2% FBS. Culture media containing siRNA and VEGF were replaced every 2 days. After 10 days, vessels sprouting from the aortas were stained with 1 g/ml Calcein-AM (Invitrogen) and observed using a fluorescent microscope. Total tube length was calculated by image analyzer software (Kurabo). The siRNA sequences against mouse Rnd1 was GAACAGAAAUCCUAG-ACUATT or CGGUAUUGCUGUGCUUUGATT, and that against mouse Cxcr7 was GAUGGUACGCCGUGUUGUATT or GGAAGAUCAUCUUCUCCUATT.
RhoA Pulldown Assay-siRNA-transfected HUVEC were serum-starved for ϳ18 h, stimulated with 50 ng/ml VEGF, and harvested immediately with lysis/binding/washing buffer from the Active Rho pull-down and detection kit (Thermo Scientific). Pulldown using purified GST-Rhotekin and immunoblot analysis for RhoA were performed according to the instructions of the manufacturer (Thermo Scientific).
Coculture Angiogenesis Assay-The coculture system of HUVEC and fibroblasts was purchased from Kurabo. Cells were incubated in medium containing VEGF (10 ng/ml). siRNA treatment was performed twice, on days 1 and 7. Eleven days later, cells were fixed and immunostained following the instructions of the manufacturer (Kurabo). Vessel trees and branches were calculated by an image analyzer (Kurabo).
Statistics-Data are shown as mean Ϯ S.D. p values were calculated by using two-tailed unpaired Student's t test.

VEGF Induces Dynamic Calcineurin-mediated NFATc1
Activation in Primary Cultured Endothelial Cells-We reported previously that VEGF and thrombin activates the Ca 2ϩ -calcineurin-NFAT axis, which is a critical intercellular signaling pathway in primary cultured endothelial cells (11). To further study VEGFmediated NFATc1 activation, we generated a mouse monoclonal antibody against the human NFATc1 antigen, amino acids 134 -223, with high affinity (Fig. 1A). Our antibody detected all DNA binding domain-containing NFATc1 splicing isoforms, but not the closely related NFATc2, in a Western blot analysis (Fig. 1, B and C). To determine how quickly NFATc1 was activated after VEGF treatment, we carried out time course immunohistochemistry staining with the anti-NFATc1 antibody in HUVEC. As shown in Fig. 1, D and E, NFATc1 was localized both in the cytoplasm and in the nucleus without VEGF. However, VEGF treatment caused NFAT to translocate to the nucleus, with maximum NFAT nuclear localization at 10 min. NFATc1 remained in the nucleus at 1 h but returned to the cytoplasm by 4 h after VEGF treatment. Moreover, VEGF-mediated NFATc1 activation was dependent on calcineurin activity because CsA treatment largely abrogated VEGF-mediated NFATc1 nuclear translocation (Fig.  1E). To test the possibility of increased turnover/degradation of cytoplasmic NFATc1, we examined the amount of nuclear and VEGF-NFATc1-mediated CXCR7 and RND1 Induction in Endothelium cytoplasmic NFATc1 after VEGF stimulation to quantify changes in protein levels. To evaluate NFATc1 turnover, immunoblot analyses with anti-NFATc1 antibody were performed on nuclear and cytoplasmic fractions of HUVEC treated with or without the proteasome inhibitor MG-132. As a positive control, we also examined p65 nuclear localization. MG-132 treatment abrogated TNF-␣-mediated p65 nuclear localization (data not shown) but did not interfere with VEGF-mediated NFATc1 nuclear localization (Fig. 1F). In contrast, CsA strongly attenuated translocation into the nucleus (Fig. 1F). This suggests that VEGF promptly activates the calcineurin pathway, which leads to NFATc1 nuclear localization in endothelial cells. A, schematic of spliced isoforms of human NFATc1 and the antigen for our generated mouse monoclonal antibody. A.A., amino acid. B, ChIP-Western blot analysis of immunoprecipitates using the anti-NFATc1 antibody from HUVEC extracts. The right column indicates the NFATc1 isoforms, n ϭ 5., M.W. molecular weight. C, human NFATc1 (pCMV-NFATc1), human NFATc2 (pCMV-NFATc2), or mock control (pCMV) were transfected in COS7 cells. Whole cell extracts were prepared and blotted with an anti-NFATc1 or NFATc2 antibody. Anti-␤-actin antibody was used as a loading control, representative of four independent experiments. D, NFAT immunofluorescent staining. NFAT localization in the nucleus versus the cytoplasm was calculated in HUVEC treated with VEGF at the indicated time points. Data are mean Ϯ S.D. of the nuclear localization levels obtained from three independent experiments. *, p Ͻ 0.01 compared with without VEGF. E, immunofluorescent staining of NFATc1 (left column) in mock-and VEGF-treated HUVEC with or without CsA and MG-132. Merged images with PECAM1 and DAPI are shown in the right column. Scale bars ϭ 50 m. F, Western blot analyses with anti-NFATc1 antibody. HUVEC were treated with PBS (mock) or VEGF for 1 h in the presence or absence of CsA and MG-132. Subsequently, nuclear and cytosolic fractions were subjected to 10% SDS-PAGE. LaminA and Hsp90 are shown as nuclear and cytosolic markers, respectively. n ϭ 3.

A Genome-wide Approach Revealed Comprehensive VEGFresponsive Regions by NFATc1 Binding Site-enriched Analysis in Primary Cultured
Endothelial Cells-We next wished to identify globally how VEGF-mediated NFATc1 nuclear localization regulates target gene expression. To that end, we per-formed a cDNA microarray analysis using VEGF-stimulated HUVEC in the presence or absence of CsA or the overexpression of constitutively active, nuclearly localized NFATc1 (CA-NFATc1). A comparison of microarray data from endothelial cells with adenovirally overexpressed CA-NFATc1 or endothe-

VEGF-NFATc1-mediated CXCR7 and RND1 Induction in Endothelium
lial cells treated with VEGF identified 416 genes that were more than 2-fold up-regulated by VEGF treatment, whereas 1678 genes were induced by CA-NFATc1. Among these, 193 genes were commonly induced ( Fig. 2A), and the representative genes are shown in supplemental Table 1. Among the genes that were down-regulated, only nine genes were reduced more than 2-fold by VEGF for 1 h. 196 genes were reduced by overexpression of CA-NFATc1 ( Fig. 2A). Collectively, these data suggest that VEGF signaling via NFATc1 favors gene up-regulation rather than down-regulation.
Subsequently, to globally survey where nuclear localized NFATc1 is bound to the genome, we performed ChIP with an NFATc1 antibody, and then the precipitated genome was sequenced comprehensively (ChIP-seq). To test the accuracy of the ChIP assay with the NFATc1 antibody, we used the previously reported NFATc1-regulated Dscr-1 and Egr-3 promoter regions as positive controls and the non-endothelial MyoD1 promoter region as a negative control. The NFATc1-mediated immunoprecipitated DNA and non-immunoprecipitated whole genome control (input DNA) were used to prepare libraries for deep sequencing and analyzed using massively parallel sequencing. The genome-wide NFATc1 binding regions were calculated by QuEST (see Ref. 20 for details). 4119 regions were identified as NFATc1-associated areas with VEGF treatment from the ChIP-seq. To reconstruct the NFATc1-associated binding regions, we clustered the regions into three sections on the basis of the distance from the transcription start site in the respective genes. As shown in Fig. 2B, half of the NFATc1bound peaks were located at the proximal promoter region within 1 kb from the transcription start site and the 5Ј-UTR. 17% of NFATc1-bound peaks were categorized in the gene body, exon, intron, and 3Ј-UTR. The remaining 33% of the NFATc1 binding sites were at intergenes.
Next, we identified the commonly recognized motif from the whole NFATc1-mediated ChIP genome sequences. As shown in Fig. 2C, the NFATc1 consensus motif, AGGAAA, was identified to be the highest enriched binding element, with an E value of 1.3ϫ e Ϫ194 . In addition to NFAT, CCAAT/enhancer binding protein (C/EBP)-and CREB1-recognized sequences were also found as the second and third most coenriched elements. NFATc1 activation predominantly induced the target genes ( Fig. 2A). Therefore, to determine whether NFATc1 binding correlated with the active chromatin state, we added ChIP-seq analyses using antibodies against H3K4me3 as an active proximal promoter mark and H4Ac as an enhancer mark using HUVEC with or without VEGF treatment. Combination of the enrichment scores for NFATc1 and H3K4me3 indicated that basal H3K4me3 methylation was independent of VEGF and that, in the absence of VEGF, some NFATc1 was bound to these transcriptionally active promoter regions. In contrast, H4Ac-enriched genes were correlated significantly with VEGFmediated gene expression because H4Ac marks were minimally present in untreated HUVEC but increased more than 3-fold after VEGF stimulation (Fig. 2D). These findings suggest that subsets of NFATc1 are constitutively harbored in the H3K4me3-positive region. After VEGF stimulation, more abundant NFATc1 was bound on H4Ac-positive enhancer as well as H3K4me3-positive promoter regions, which allowed maximum activation of VEGF-regulated genes (Fig. 2E).
Binding of NFATc1 to DNA might not always necessarily correlate with changes in gene expression. To determine the association between genome-wide NFATc1 binding and target gene expression, we compared the results of ChIP-seq with those of DNA microarrays of VEGF-or CA-NFATc1-overexpressed HUVEC in the presence or absence of CsA. We selected microarray gene set probes that exhibited significant expression, as defined by a Ͼ100-fold average difference and more than 2-fold up-regulation by VEGF for 1 h, which were reduced more than 30% by CsA treatment. Alternatively, the genes that were reduced more than 2-fold by VEGF but oppositely induced by CsA were selected. 48 total genes met these criteria (Fig. 2F, left panel). Interestingly, more than half of these genes were directly bound by NFATc1 at the proximal promoter region. Moreover, 20 total genes were significantly up-regulated via CA-NFATc1 overexpression (Fig. 2F, right panel). In these clusters, we found Vcam-1, Dscr-1 short isoform (Dscr-1s), and Egr-3 genes, which we reported previously as transactivated by NFAT (11,13,15). Representative NFAT binding sites with epigenetic information are shown in Fig. 2G.
Dynamic NFATc1 Binding on the Promoter Directs VEGFmediated CXCR7 Induction in Primary Cultured Endothelial Cells-Among the NFATc1 positively regulated gene clusters, we decided to further focus our analysis on Cxcr7 and Rnd1 gene expression because the identification of these genes as NFAT-responsive is novel and because they are candidates for VEGF-mediated angiogenic modulators. Cxcr7 mRNA was most highly (4.1-fold) induced at 1 h after VEGF treatment, and this induction was abrogated by the pretreatment with CsA (Fig. 3A). NFAT-mediated CXCR7 transactivation was also validated by adenoviral CA-NFATc1 overexpression in HUVEC (Fig. 3B). Moreover, as shown in Fig. 3C, NFATc1 bound on the H3K4me3-positive proximal promoter region where histone acetylation is associated with open chromatin. Both H4Ac and H3K4me3 binding levels were increased after the VEGF treatment. To determine whether NFATc1 binding is involved in changes in histone modifications, we performed comparative  OCTOBER

VEGF-NFATc1-mediated CXCR7 and RND1 Induction in Endothelium
ChIP-qPCR with anti-H3K4me3 and anti-H4Ac antibodies. After VEGF stimulation, H3K4me3 and H4Ac binding was increased 4.1-and 2.8-fold, respectively, on the Cxcr7 promoter region but not MyoD1 as a negative control (Fig. 3D). Importantly, CsA treatment completely abrogated VEGF-stimulated H4Ac binding but not H3K4me3 binding to the Cxcr7 promoter region (Fig. 3D). Therefore, CXCR7 expression in endothelial cells is regulated by increased NFAT binding connected to histone acetylation. Moreover, we found three NFAT consensus elements around the positive region from NFATc1-mediated ChIP-seq data (Fig. 3C), but there were no other NFATc1-enriched binding regions within the insulator CCCTC-binding factor (CTCF)-divided Cxcr7 locus. The significant but transient Cxcr7 gene induction correlated directly with significantly increased but transient NFATc1 binding on the Cxcr7 promoter (Fig. 3E).
To evaluate the transcriptional mechanism underlying NFATc1-mediated CXCR7 expression, we cloned the 5Ј flanking region (Ϫ2193/ϩ83) of the Cxcr7 gene and inserted it into the luciferase reporter vector, CXCR7-luc. VEGF stimulation of HUVEC containing this vector induced significant promoter activity. This promoter activity was completely abrogated by CsA treatment. Moreover, the NFAT point-mutated promoter failed to respond to VEGF (Fig. 3F). Cotransfection with the NFATc1 expression plasmid and Cxcr7 promoter, but not the NFAT mutated Cxcr7 promoter, resulted in the more than 3-fold Cxcr7 promoter activation compared with the mock control (Fig. 3G). Taken together, the increase in Cxcr7 mRNA and the promoter activity inductions via VEGF signaling indicate that NFATc1 directly regulates Cxcr7 expression.
NFATc-mediated CXCR7 Expression Leads to Endothelial Migration and Tube Formation-siRNA to NFATc1 significantly abrogated NFATc1 mRNA and protein expression (13) (data not shown), which almost completely abrogated VEGFinduced membrane expression of CXCR7 (Fig. 4A). CXCR7 is a chemokine receptor and has been identified recently as a receptor for CXCL12/SDF-1 ligand (6,8). CXCR7 and the classic SDF-1 receptor CXCR4 are both expressed by endothelial cells (7). SDF-1 is expressed by various cells, including endothelial cells, but is more highly expressed in perivascular tissues from the endothelium (Fig. 4B). To verify CXCR4 and 7 regulation in endothelial cells, we compared VEGF-mediated induced patterns in each mRNA. CXCR4 is constitutively expressed by endothelial cells. In sharp contrast to Cxcr7, Cxcr4 mRNA was up-regulated weakly and slowly after 24 h via VEGF, which did not correlate directly to immediate NFAT activation (Fig. 4, C and D). To further distinguish between CXCR4 and CXCR7, we generated specific siRNA for each (Fig. 4E) and performed endothelial cell migration and tube formation assays. Mild induction of CXCR4 mRNA via VEGF did not lead to the upregulation of membrane-expressed CXCR4 levels within 24 h. Therefore, siRNA against CXCR4 markedly reduced the basal expression of CXCR4, but this reduction was not changed in the presence of VEGF (Fig. 4F). In contrast, siRNA to CXCR7 resulted in Ͼ90% inhibition of VEGF-induced CXCR7 mRNA expression (Fig. 4G). VEGF treatment increased membrane expression of CXCR7, which was abolished by two independent siRNAs (si-CXCR7 oligo1 and 2, Fig. 4H and data not shown).
Under these conditions, si-NFATc1 treatment significantly attenuated cell migration compared with si-control. CXCR4 specific knockdown resulted in ϳ70% reduction of SDF-1-mediated cell migration. Importantly, CXCR7-specific knockdown reduced cell migration toward SDF-1 more prominently (ϳ90%). Combined knockdown of CXCR4 and CXCR7 caused complete abrogation of the response to SDF-1 in endothelial cells (Fig. 5A). Similarly, cell scratch assays also demonstrated that, in the absence of CXCR7 (si-CXCR7), HUVEC sensitivity to SDF-1 was not increased by VEGF (Fig. 5B). These data suggest that both CXCR4 and CXCR7 are important for SDF-1mediated motility in endothelial cells but that NFATc1-mediated CXCR7 is regulated more tightly by VEGF stimulation.
Next, to study the functional relevance of the VEGF-mediated NFATc1-CXCR7 axis, we evaluated angiogenesis activity in vitro. VEGF stimulated regrowth of the endothelial cell frontier region after denudation, where nuclear localization of NFATc1 and concurrent CXCR7 expression were apparent. Without VEGF treatment, NFATc1 nuclear localization and CXCR7 expression were minimal (Fig. 5C). Moreover, coculture of endothelial cells and SDF-1-rich fibroblasts enhanced vascular tree and branch formation. However, NFATc1 knockdown markedly reduced this coculture effect, despite VEGF stimulation (Fig. 5D). Finally, to evaluate NFATc1-CXCR7 activity under the physiological vasculature, we designed siRNA to mouse cxcr7 and then performed aortic ring assays. Cxcr7 siRNA reduced cxcr7 mRNA expression after VEGF treatment by more than 90% (Fig. 5E). VEGF induced neovessel sprouting from the aortic ring treated with control culture medium (Fig. 5F). In contrast, aortic rings in which Cxcr7 was reduced by siRNA failed to induce vessel sprouting from the aorta, even in the presence of VEGF (Fig. 5F). Taken together, these findings suggest that NFATc1 activation via VEGF signaling in endothelial cells causes membrane expression of CXCR7,  3). B, CXCR7 expression in HUVEC infected by Ad-NFATc1 or Ad-lacZ for 24 h (*, p Ͻ 0.001 compared with the control) (n ϭ 4). C, genome browser view around the CXCR7 gene. ChIP-signals in the presence or absence of VEGF are shown with NFATc1 (red), H3K4me3 (yellow), and H4Ac (purple). The number below each ChIP-seq indicates the threshold window on the integrated genome browser (IGB) software. D, HUVEC were treated with VEGF for 1 h. ChIP was performed with antibodies to H3K4me3 (green) and H4Ac (pink). Subsequent qPCR using primers specific to the Cxcr7 (left panel) and MyoD1 (right panel) target loci are as shown under "Experimental Procedures." Fold enrichment was compared with non-immunoprecipitated chromatin (percent input). Data are expressed as mean Ϯ S.E. obtained from three independent experiments. *, p Ͻ 0.05, compared with to without VEGF; #, p Ͻ 0.05 compared with to without CsA (n ϭ 3). E, ChIP-qPCR analysis with NFATc1 antibody on the CXCR7 locus in HUVEC treated with VEGF at the indicated time points (*, p Ͻ 0.001 compared with 0 min of VEGF treatment) (n ϭ 3). F, HUVEC were transfected with CXCR7-luc or CXCR7 (NFAT mut)-luc and then serum-starved, preincubated with 1 m CsA or mock control, and treated with VEGF for 4 h. Data are mean Ϯ S.D. of luciferase light units obtained from six independent experiments. *, p Ͻ 0.001 compared with to without VEGF; **, p Ͻ 0.001 compared with to without CsA. G, CXCR7-luc and the NFATc1 expression plasmid (pCMV-NFATc1) were transiently cotransfected into COS7 cells. Two days later, cells were assayed for reporter activity. Data are mean Ϯ S.D. of relative expression compared with the mock control (pCMV). *, p Ͻ 0.001 compared with the control (n ϭ 6). OCTOBER

Rnd1 Is a Target of the VEGF-NFATc1
Signaling Axis-Rnd1 mRNA was highly (5.8-fold) induced at 1 h by VEGF treatment, and this was strongly attenuated by pretreatment with CsA (Fig.  6A). NFAT-mediated Rnd1 transactivation was also validated by adenoviral CA-NFATc1 expression in HUVEC (Fig. 6B). Importantly, NFATc1 bound on the H3K4me3-positive proximal promoter and the H4Ac-positive 19 kb upstream distal enhancer regions (Fig. 6C). These two NFATc1 binding regions were predicted to belong to the Rnd1 regulatory locus because of the location within two CTCF divisions (Fig. 6C). To determine whether NFATc1 binding on the Rnd1 locus involves changes in histone modifications, we performed ChIP-qPCR with antibodies to H3K4me3 and H4Ac. VEGF stimulation significantly increased the H3K4me3 and H4Ac modification rate. CsA treatment attenuated VEGF-mediated histone acetylation on the Rnd1 enhancer region (Fig. 6D). We found two separated NFAT consensus elements around the proximal Rnd1 promoter region and multiple (eight) NFAT elements in the distal enhancer. Transient RND1 expression correlated with profound but transient NFATc1 binding on the Rnd1 gene locus (Fig. 6E). To evaluate the transcriptional regulation underlying NFATc1-regulated RND1 expression, we isolated the proximal Rnd1 promoter (Ϫ890/ϩ103 bp) region and distal (Ϫ16,066/ Ϫ18,577 bp) region and inserted these regions into a luciferase reporter vector, Rnd1-luc. We failed to detect VEGF-mediated promoter activation when only the proximal promoter was transiently transfected to HUVEC. However, combination of the proximal promoter and the distal enhancer from the Rnd1 locus showed significant reporter up-regulation after VEGF treatment, which was reduced significantly by CsA treatment (Fig. 6F). Moreover, cotransfection of Rnd1-luc with the distal enhancer and NFATc1 expression plasmids resulted in significant (ϳ8.2-fold) up-regulation of the luciferase activity compared with the mock control (Fig. 6G). Collectively, Rnd1 mRNA and the VEGF-stimulated promoter activity suggest that transient NFATc1 binding at the Rnd1 distal enhancer region induces RND1 expression.
NFATc1-mediated Rnd1 Expression Regulates Endothelial Cell Integrity and Neovessel Formation-RND1 activates p190RhoGAP, an inhibitor of Rho family member activity, and is, therefore, an inhibitor of Rho signaling (3). We hypothesized that RND1 induction modulates RhoA activity following the normalization of hyperpermeability. First, we wished to evaluate the correlation between RhoA and Rnd1 activity in endo-thelial cells. Transfection of two independent siRNAs to Rnd1 (si-Rnd1) reduced VEGF-mediated Rnd1 mRNA up-regulation by 91 and 81% (Fig. 7A). VEGF treatment caused a transient increase of RhoA activation (1 min) in the control. Importantly, inhibition of NFATc1 by si-NFATc1 caused basal and VEGFmediated RhoA hyperactivation at both acute (1 min) and sustained (60 min) time points (Fig. 7B). Moreover, knockdown of the NFATc1-direct target, Rnd1, similarly resulted in RhoA hyperactivation (Fig. 7B). Subsequently, to test vessel integrity in cultured endothelia, we performed HUVEC permeability assays in vitro using FITC-labeled dextran and a modified Boyden chamber in which permeability through the endothelial monolayer is measured. HUVEC permeability was increased with si-NFATc1 compared with si-control treatment in the presence of VEGF, resulting in 138-fold higher penetration of dextran. Moreover, two independent si-Rnd1 treatments showed similarly increased higher permeability in the presence of VEGF (Fig. 7C). Collectively, this suggests that NFATc1 and the direct downstream target, Rnd1, regulate RhoA activity and endothelial cell integrity.
RhoA activation is directly linked to vessel migration and extension (21). Therefore, to survey the functional outcome of VEGF-mediated Rnd1 induction in neovessel formation, we carried out cell migration assays using a modified Boyden chamber. Endothelial cell migration was increased more than 40-fold in the presence of VEGF for 24 h (Fig. 7D). Importantly, knockdown of Rnd1 augmented VEGF-mediated cell migration (Fig. 7D). To further characterize VEGF-mediated Rnd1 function, we carried out mouse aortic ring assays in the presence of si-Rnd1. As shown in Fig. 7E, two independent siRNA to mouse Rnd1 resulted in more than 80% reduction of VEGF-stimulated Rnd1 mRNA up-regulation. These siRNAs increased neovessel formation from aortic rings in response to VEGF. Total tube length and the vessel-invaded areas were increased significantly compared with the control in the presence of VEGF (Fig. 7F). Taken together, these findings suggest that VEGF-mediated NFAT activation leads to Rnd1 expression as a direct downstream target, which, in turn, modulates VEGF-mediated RhoA activation.

DISCUSSION
Recent technological advancements now allow transcription factor binding regions to be defined and the histone code to be unlocked. In this report, we examined VEGF-stimulated NFATc1 binding on a genome-wide scale in human primary cultured endothelial cells and then comprehensively mapped and compared these signals with epigenetic histone marks and  (n ϭ 3). Black, si-control without VEGF; red, si-control with VEGF; blue, si-CXCR7 with VEGF. OCTOBER

VEGF-NFATc1-mediated CXCR7 and RND1 Induction in Endothelium
whole-genome expression arrays. VEGF treatment of endothelial cells caused NFAT nuclear localization, which led to expression of proangiogenic and proinflammatory genes. At the early time point, around one-third of VEGF stimulated genes were also stimulated by NFAT ( Fig. 2A), suggesting that the calcineurin-NFAT pathway is an important route for VEGF-mediated signal transduction in endothelial cells. ChIP-seq with NFATc1 in the presence of VEGF demonstrated that many of the NFATc1 binding elements were located within proximal promoters, the majority of which were commonly located with proximal H3K4me3, active promoter histone marks (22) (Fig. 2D). These NFATc1 occupancy patterns were unique, considering our previous reports of GATA2-and STAT6-mediated ChIP-seqs in endothelial cells (20,23). GATA2 and STAT6 predominantly occupied distal enhancer regions. These differences might be on the basis of the functional difference of the transcription factors. NFAT is crucial for turning on the VEGF switch, but GATA2 and STAT6 function in endothelial specificity and chronic inflammation, at least in part, via changing the chromatin structure. Generally, compared with methylation, histone acetylation is thought to be more dynamically and accurately correlated to target gene expression levels (24). VEGF-stimulated NFAT activation correlated significantly with H4Ac histone modification (Fig. 2D). Surprisingly, the H3K4me3 mark was also increased moderately by VEGF treatment in our data. Although the mechanism and the functional consequences are still unknown, H3K4me3 levels do not always correlate with gene expression levels. For example, increased H3K4me3 marks in the Rnd1 promoter were important, but not important enough, to activate the Rnd1 core promoter-luc without accompaniment of histone acetylation (Fig. 6F). Genes obtaining bivalent H3K4me3 and H3K27me3 marks in ES cells have been demonstrated previously to be ready for expression, but transcription had not begun (25). Our data, taken together with previous studies, suggest that small amounts of NFAT might already be present in the transcriptionally permissive core promoter and that, after VEGF stimulation, abundant responsive NFAT can bind on H3K4me3-positive as well as H4Ac-positive regions with dynamic histone modifications (Fig. 2E). Our model proposes that gene expression in response to VEGF is amplified rapidly via this sequence. Chromatin is already in an expression-permissive state because of the presence of H3K4me3 methyl marks, and small amounts of NFAT are localized to the promoter region prior to VEGF stimulation. After VEGF treatment, more NFAT translocates to the promoter region, accompanied by a significant increase in H4Ac marks and some increase in H3K4me3, which further promote open chromatin. Together, increased NFAT binding and chromatin modifications promote gene expression.
NFAT nuclear localization and subsequent target gene activation are tightly regulated by calcineurin phosphatase, which is inhibited by the immunosuppressors CsA and FK506. When NFAT is phosphorylated by the priming kinase DYRK1A, NFAT is exported from the nucleus to the cytoplasm (26). In T cells, NFAT directly transactivates genes. Many of these genes function in inflammation or immune responses (27). The majority of NFAT-activated genes in endothelial cells are unique compared with T cells, except Egr-3 and Cox2. The differential activation mechanism between endothelial cells and T cells might be on the basis of the epigenetic microenvironment. Indeed, T cell-specific induced genes, TNF-␣, colony-stimulating factor 1, IL-2, and IL-4 were fully covered with silencer histone markings and never enriched by active promoter markings in primary cultured endothelial cells. Specific regulation of these cell types was similarly observed in our previous study of IL-4-STAT6-mediated gene expression patterns in endothelial cells and T cells (23).
In HUVEC, ChIP-seq with NFATc1 antibody identified DNA enriched for the sequence AGGAAA, a well recognized consensus binding sequence. This is not random enrichment. The associated E value is 1.3ϫ e Ϫ194 (Fig. 2C). HUVEC results were coenriched with C/EBP and CREB as second and third occupied sequences via ChIP-seq with NFATc1 (12). We and others have reported that C/EBP and CREB cobinding with NFAT functioned for maximum transactivation of DSCR-1 and Egr-3 in endothelial cells (15,28). These data suggest that not only the specificity with agonist-receptor interaction but also the epigenetic microenvironment for NFAT would regulate the specific downstream target genes in endothelial cells.
We newly identified two novel functional NFATc1 endothelial cell targets, CXCR7 and RND1. CXCR7 is now classified as a chemokine receptor that is able to bind SDF-1 and CXCL11 (6). SDF-1 is a chemoattractant for vascular endothelial cells (29). Because we failed to detect SDF-1 as a VEGF-mediated inducible factor in endothelial cells, we thought it likely that SDF-1 expressed by perivascular cells activates CXCR7 in endothelial cells in a paracrine manner. Indeed, smooth muscle  3). B, SDF-1-mediated endothelial cell migration in si-CXCR4 and si-CXCR7 HUVEC. HUVEC were scratched and incubated with EBM-2 plus 100 ng/ml SDF-1 or 100 ng/ml SDF-1 plus 50 ng/ml VEGF. The scratched field is shown by a broken red line. After 24 h, migrated cells were counted under a phase-contrast microscope. Right panel, mean Ϯ S.E. from four optical independent images. #, p Ͻ 0.05 compared with si-control with SDF-1 alone; *, p Ͻ 0.05 compared with si-control with SDF-1ϩVEGF; N.S., not significant (n ϭ 4). C, confluent HUVEC were denudated and recultured with basal conditioning (EBM-2 plus 0.5% FBS) medium. One day later, cells were treated in the presence or absence of VEGF for 2 h. Subsequently, cells were fixed and immunostained with anti-NFATc1 (green) and -CXCR7 (red) antibodies. The dashed rectangle indicates the representative growth frontier region (n ϭ 4). Scale bars ϭ 50 m. D, HUVEC and fibroblasts were cocultured and treated with si-control or si-NFATc1 in the presence of VEGF-containing media. siRNAs were added twice, on days 1 and 7. Eleven days later, cells were fixed and visualized with anti-PECAM1 (green) or anti-CXCR7 (red). Merged images with DAPI are shown on the right. Images are representative of eight independent experiments. Vascular trees and branches were calculated with a cell imager (Kurabo) and are mean Ϯ S.E. *, p Ͻ 0.05 compared with si-control. Scale bars ϭ 50 m. E, real-time qPCR analysis of mCxcr7 expression in mouse endothelial cells. Data are mean Ϯ S.D. derived from three independent experiments. *, p Ͻ 0.001 compared with si-control. F, aortic ring assays in si-control or si-Cxcr7 treatment. The outgrowth of neovessels from the aorta was observed under a phase-contrast microscope, and tube length was calculated using a cell image analyzer. The mean Ϯ S.D. was derived from six independent experiments. *, p Ͻ 0.05 compared with the si-control with VEGF. OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42 cells and fibroblasts showed much higher SDF-1 expression than endothelial cells, which may also potentiate VEGF-stimulated angiogenesis in aortic rings. Importantly, CXCR7 is increasingly reported as being expressed by malignant tumors and tumor endothelial cells (9). Moreover, we demonstrated that CXCR7 is a direct downstream target of the VEGF-NFAT axis in endothelial cells. Consistent with our data of CXCR7 as an NFATc1 target, similar phenotypes were observed between enhancer region (2555bp) promoter region (993bp) 5 Rnd1 locus (47,532,000-47,571,500  NFATc1 null mutation and CXCR7-deficient mice. Ventricular septal defects and heart valve malformation have been reported for each targeted knockout mouse (30,31). It is well known that the interaction between SDF-1 and its unique receptor, CXCR4, is critical for tumor growth and metastasis (7). However, the recent identification of CXCR7 merits an addition to the short list of known SDF-1 receptors. Blockade of CXCR7, as well as CXCR4, in tumor-and VEGF-stimulated endothelial cells may lend itself to antitumor therapy. We identified a second target of VEGF-NFAT signaling, Rnd1. Rnd1 inhibits RhoA activation to regulate actin filament disassembly, which leads to a rounded cell shape (2). VEGFmediated Rnd1 expression was transient. We hypothesized that it functions as a modulator of RhoA activity to prevent hyperactivation, which could cause dysfunctional angiogenesis. In line with our theory, Rnd1 knockdown further exaggerated VEGF-promoted endothelial permeability and migration, at least following the hyperactivation of RhoA. RhoA signaling plays an essential role in VEGF-dependent in vivo angiogenesis and in the initial step of in vitro endothelial cord assembly. Tumor capillary endothelial cells, or targeted deletion of central cavernous malformation, led to hyperactivation of RhoA, which resulted in excessive but dysfunctional angiogenic sprouting (32,33). Normalization of RhoA activity rescues functional angiogenesis. Therefore, VEGF regulation of Rnd1 to normalize RhoA activity would be important for angiogenesis. NFAT induction of Rnd1 to modulate VEGF-induced angiogenesis is similar to our recent findings related to the calcineurin modulator DSCR-1. Both overexpression and null mutation of DSCR-1 cause the attenuation of the VEGF-medi-  4). B, pulldown assay using GST-rhotekin in whole cell extracts derived from siRNA-treated HUVEC. GDP and GTP␥S are the negative and positive control, respectively (n ϭ 5). C, FITC-dextran permeability assay in a Boyden chamber using VEGF-treated HUVEC. Data are mean Ϯ S.D. of FITC concentrations derived from three independent experiments. *, p Ͻ 0.001 compared with si-control. D, VEGF-1-mediated cell migration of HUVEC with Rnd1 knockdown. Right panel, mean Ϯ S.D. migration level derived from triplicate experiments. *, p Ͻ 0.05; #, p Ͻ 0.01 compared with the absence of VEGF and si-control in the presence of VEGF for 24 h, respectively (n ϭ 3). E, real-time qPCR analysis of mRnd1 expression in mouse endothelial cells. Data are mean Ϯ S.D. derived from three independent experiments. *, p Ͻ 0.001 compared with the si-control. F, aortic ring assays in si-control or si-Rnd1 treatment. The outgrowth of neovessels from the aorta was observed under a phase-contrast microscope, and tube length was calculated using a cell image analyzer. The mean Ϯ S.D. were derived from six independent experiments. *, p Ͻ 0.05 compared with the si-control plus VEGF. OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42 ated angiogenesis because of blocking or excessive activation of NFAT, respectively (11,16). Therefore, physiological and transient DSCR-1 and Rnd1 induction seem to be important for proper angiogenesis promotion. Taken together, we demonstrated that the VEGF-mediated calcineurin-NFAT signaling axis is one of the key angiogenesis regulatory pathways. NFAT directs a number of multifunctional downstream targets in endothelial cells, including feedback modulators, chemokine receptors, and transcription factors (Fig. 8).

VEGF-NFATc1-mediated CXCR7 and RND1 Induction in Endothelium
In summary, our findings reveal a mechanism by which VEGF-mediated NFATc1 widely binds the genome on epigenetically coordinated active DNA elements and regulates the expression of genes with functional importance in endothelial cells. Given the critical role of NFAT in endothelial cell activation and dysfunction, the discovery of the new transcriptional circuits via NFAT may provide a foundation for targeted therapies in VEGF-mediated vascular diseases.