Vascular Endothelial Growth Factor- and Thrombin-induced Termination Factor, Down Syndrome Critical Region-1, Attenuates Endothelial Cell Proliferation and Angiogenesis*♦

Activation and dysfunction of the endothelium underlie many vascular disorders including atherosclerosis, tumor growth, and inflammation. Endothelial cell activation is mediated by many different extra-cellular signals, which result in overlapping yet distinct patterns of gene expression. Here we show, in DNA microarray analyses, that vascular endothelial growth factor (VEGF) and thrombin result in dramatic and rapid upregulation of Down syndrome critical region (DSCR)-1 gene encoding exons 4–7, a negative feedback regulator of calcium-calcineurin-NF-AT signaling. VEGF- and thrombin-mediated induction of DSCR-1 involves the cooperative binding of NF-ATc and GATA-2/3 to neighboring consensus motifs in the upstream promoter. Constitutive expression of DSCR-1 in endothelial cells markedly impaired NF-ATc nuclear localization, proliferation, and tube formation. Under in vivo conditions, overexpression of DSCR-1 reduced vascular density in matrigel plugs and melanoma tumor growth in mice. Taken together, these findings support a model in which VEGF- and thrombin-mediated induction of endothelial cell proliferation triggers a negative feedback loop consisting of DSCR-1 gene induction and secondary inhibition of NF-AT signaling. As a natural brake in the angiogenic process, this negative pathway may lend itself to therapeutic manipulation in pathological states.

The endothelium is highly malleable cell layer, constantly responding to changes within the extracellular environment and responding in ways that are usually beneficial, but at times harmful, to the organism. Several mediators, including growth factors (e.g. vascular endothelial growth factor (VEGF) 1 ), cytokines (e.g. tumor necrosis factor-␣ (TNF-␣)), and serine proteases (e.g. thrombin), activate gene transcription in endothelial cells, resulting in changes in hemostatic balance, increased leukocyte adhesion, loss of barrier function, increased permeability, migration, proliferation, and successive angiogenesis. The tight control of these processes is essential for homeostasis; endothelial cell activation, if excessive, sustained, or spatially and temporally misplaced, may result in vasculopathic disease. Under normal conditions, the activation signal may be terminated by negative feedback inhibition of downstream transcriptional networks. Such a mechanism has been well established for TNF-␣ (1)(2)(3)(4). In contrast, little is known about the major self-regulatory processes involved in VEGF and thrombin signaling.
Thrombin is a multifunctional serine protease that is involved not only in mediating the cleavage of fibrinogen to fibrin in the coagulation cascade but also in activating a variety of cell types, including platelets and endothelial cells. Thrombin signaling in the endothelium may result in a multitude of phenotypic changes, including alterations in cell shape, permeability, vasomotor tone, leukocyte trafficking, migration, DNA synthesis, angiogenesis, and hemostasis (15). Thrombin signaling in the endothelium is mediated by a family of seventransmembrane G-protein-coupled receptors, termed protease activated receptors (PAR) (16). Currently, four members of the PAR family have been identified (PAR-1 to PAR-4). Of the various PAR family members, PAR-1 is the predominant thrombin receptor in endothelial cells (17). Once activated, PAR-1 is linked to a number of signal intermediates that include, but are not limited to, MAPK, protein kinase C, PI3K, and Akt (15,18).
Given the overlapping functions of VEGF and thrombin both in terms of signaling pathways activated and phenotypic response, we hypothesized that these two mediators are likely to trigger common transcriptional networks in the endothelium. To identify such factors, we compared DNA microarray analyses of VEGF and thrombin-treated human umbilical vein endothelial cells (HUVEC). We found that under both conditions, the most highly induced gene was DSCR-1. The DSCR-1 gene (also known as MCIP-1), designated as such because it resides within the Down syndrome critical region of human chromosome 21, encodes a protein that binds to and inhibits the catalytic subunit of calcineurin (19). In this report, we show that DSCR-1 acts as a "circuit breaker" in VEGF and thrombin signaling, serving in a negative feedback loop to inhibit endothelial cell proliferation and activation as well as angiogenesis. These results provide new insights into endothelial cell signaling and point to DSCR-1 as a potential therapeutic target for anti-angiogenesis and anti-inflammatory therapy.
FIG. 1. DSCR-1 is the most highly induced early response gene in VEGF-and thrombin-treated HUVEC. A, cluster analysis of early response genes (1 h). B, the average fold induction of the five most highly induced genes in VEGF-, thrombin-, and TNF-␣-treated HUVEC.
Microarray Analysis-HUVEC were serum-starved overnight in medium containing EGM-2 and 0.5% fetal bovine serum and then treated with 50 ng/ml VEGF or 1.5 units/ml thrombin. RNA was harvested and purified with Trizol according to manufacturer's protocol (Invitrogen). Preparation of cRNA and hybridization of probe arrays were performed according to the protocols of the manufacturer (Affimetrix, Santa Clara, CA).
Transfections and Analysis of Luciferase Activity-HUVEC were transfected using FuGENE 6 reagent (Roche Applied Science) as described previously (20). The serum-starved transfected cells were preincubated for 30 min with 1 M cyclosporine A and then incubated with 50 ng/ml VEGF or 1.5 units/ml thrombin for 6 h.
FACS Analysis-Ad-DSCR-1-, Ad-Control-, or Ad-DSCR-1 plus Ad-CA-NFAT-infected HUVEC were scraped, spun down for 5 min at 100 ϫ g, re-suspended in PBS containing with 0.2% Triton X, and stained in 25 g/ml propidium iodide and 50 ng/ml RNase A. Cells were counted on a FACS Calibur using CellQuest software (BD Biosciences), and the percentages of cells in the G 1 , S, and G 2 /M phases of the cell cycle were determined using ModFit LT software (Verity Software House, Topsham, ME).
Immunolocalization Studies-HUVEC were plated onto glass coverslips (Matsunami Glass, Osaka, Japan) in a 6-well plate at a density of 25,000 cells/slide. The cells were treated in the presence or absence of 1 unit/ml thrombin or 50 ng/ml VEGF for 1 h, fixed in ice-cold 3.7% paraformaldehyde for 10 min, washed with PBS, and subsequently incubated with primary anti-NF-ATc antibody (Affinity BioReagents). Following extensive washes in PBS, the cells were incubated with an Alexa-Fluor 594-labeled secondary antibody (Molecular Probes, Eugene, OR) for 1 h. The slides were then washed in PBS, mounted in Crystal/Mount (Biomeda, Foster City, CA) with Hoechst (Sigma) for identification of nuclear localization, and examined by fluorescence microscopy. The degree of nuclear localization was quantified with NIH image.  8, and 11). GAPDH antisense riboprobe was hybridized with total RNA as an internal control. The results are representative of three independent experiments. B, schematic shows the deletion and/or mutant DSCR-1 promoter constructs. C, HUVEC were transiently transfected with the DSCR-1 (Ϫ1664/ϩ83)-luc plasmid and exposed to 50 ng/ml VEGF, 1.5 units/ml thrombin, or 10 ng/ml TNF-␣ for 6 h in the absence (Ϫ) or presence of 1 M CsA. The results show the mean and S.D. of luciferase light units (relative to untreated and minus CsA cells) obtained in triplicates from at least three independent experiments. D and E, HUVEC were transiently transfected with deletion and/or mutant constructs and exposed to 50 ng/ml VEGF or 1.5 units/ml thrombin for 6 h. Each plasmid was co-transfected with pRL-SV40 to normalize for transfection efficiency. Promoter activities were determined relative to the normalized activity from DSCR-1 (Ϫ1664/ϩ83)-luc plasmid in untreated cells (D) or each untreated control (E). F, HUVEC were transiently transfected with (NF-AT) 4 -TATA-luc plasmid and exposed to 50 ng/ml VEGF, 1.5 unit/ml thrombin, or 10 ng/ml TNF-␣ for 6 h. G, HUVEC were transiently co-transfected with DSCR-1 (Ϫ350/ϩ83)-luc and GATA-2 (pMT 2 -GATA2), constitutively active NF-ATx (pSR␣-CANF-AT), or vector alone (pMT 2 or pSR␣), alone or in combination. The means and S.D. values are derived from at least three separate experiments performed triplicate.
Tube Formation Assays-400-l aliquots of type-I collagen gel (Koken, Tokyo, Japan) containing EGM-2-MV medium without basic fibroblast growth factor were used as described previously (22). Briefly, HUVEC infected with Ad-Control or Ad-DSCR-1 were seeded at 1 ϫ 10 5 cells/well and incubated for 24 h in 5% CO 2 . The medium was removed, and HUVEC were covered with 400 l of the gel. Cells were incubated with 1 ml of EGM-2-MV medium in the absence of basic fibroblast growth factor. Two days later, a branched capillary network was visualized under a microscope. Images from at least three different areas in each well were captured by a digital camera under a microscope.
Matrigel Plug Assays-Matrigel (BD Biosciences) containing 50 ng of VEGF and either 10 9 plaque-forming units of Ad-Control, Ad-DSCR-1, or Ad-DSCR-1 plus Ad-CA-NFAT was injected subcutaneously into C57BL6 mice. After 14 days, matrigel plugs were removed for histological sections. Alternatively, the matrigel plugs were thelial cell activation is not an all-or-none response. Indeed, different extra-cellular mediators engage the endothelium in ways that differ from one signal to the next. A major focus is to study the temporal and spatial dynamics of endothelial cell phenotypes. Using DNA microarrays, we carried out a global survey of mRNA in HUVEC treated in the absence or presence of growth factor (VEGF), serine protease (thrombin), or cytokine (TNF-␣). Clustering analyses of the data revealed a far closer relationship between VEGF and thrombin than between other pairings ( Fig. 1A and supplemental Table I). Of the various transcripts that were responsive both to VEGF and thrombin, DSCR-1 was the most highly induced at the earliest time point. At 1 h, VEGF resulted in 22.3-fold induction of DSCR-1, while thrombin induced DSCR-1 by 17.7-fold (Fig.  1B). The effect of VEGF and thrombin was no longer detectable at 24 and 18 h, respectively. Compared with VEGF and thrombin, TNF-␣ treatment of HUVEC resulted in far less induction of DSCR-1 (3.2-fold at 1 h) (supplemental Table I and data not shown).

VEGF and Thrombin Selectively Induce the DSCR-1 Isoform Consisting of Exons 4 -7 in Primary Human Endothelial
Cells-The DSCR-1 gene includes seven exons and six introns. The first four exons are alternative and code for four different isoforms (23,24). A 5Ј promoter regulates expression of the first three isoforms, which are derived from alternative splicing. The most common of these contains exons 1, 5, 6, and 7. An intragenic region between exons 3 and 4 contains an alternative promoter which initiates transcription of the fourth isoform (exons 4, 5, 6, and 7) ( Fig. 2A) (24). To determine which isoform(s) is up-regulated by VEGF and thrombin, we performed RNase protection assays using riboprobes that span either exons 4 and 5 or exons 1 and 5 (Fig. 2B). As shown in Fig.  2C, the addition of VEGF or thrombin to HUVEC resulted in marked up-regulation of the DSCR-1 isoform encoded by exons 4 -7, with maximal levels occurring at 1 h. In contrast, there was no detectable induction of the DSCR-1 isoform encoded by exons 1, 5, 6 and 7 (Fig. 2C, asterisk). Consistent with the DNA microarray data, TNF-␣ resulted in comparatively low levels of DSCR-1 induction. To determine whether DSCR encoding exons 1, 5, 6, and 7 is expressed in endothelial cells and to further test its inducibility, RT-PCR assays were performed using isoform-specific primers and cDNA from control or VEGF-, thrombin-, or TNF-␣-treated HUVEC. Transcripts were detected only after 40 cycles of amplification and were not affected by the addition of extracellular mediators at 1 or 4 h. By comparison, the isoform encoded by exons 4 -7 was detected in thrombinand VEGF-treated HUVEC after just 21 PCR cycles (Fig. 2D). Together, these results suggest VEGF and thrombin and to a far lesser extent TNF-␣ result in the rapid and selective induction of the DSCR-1 isoform containing exons 4 -7.

FIG. 5-continued
Thrombin is known to signal through protease-activated receptors (PARs). PAR-1 is the most important thrombin receptor in endothelial cells. To demonstrate a potential role for PAR-1 in mediating the thrombin response, we employed RNase protection assays of HUVEC treated in the presence or absence of TRAP, a 14-amino acid PAR-1 agonist peptide. TRAP resulted in a dose-dependent increase in DSCR-1 mRNA levels. Moreover, thrombin-mediated induction of DSCR-1 was completely blocked by pretreatment with the thrombin-specific inhibitor, hirudin (Fig. 3, C and D). Taken together, these findings suggest that VEGF and thrombin mediate DSCR-1 induction in endothelial cells via Flk-1/KDR and PAR-1, respectively. thrombin-mediated induction of DSCR-1 in endothelial cells is similarly dependent on calcineurin, we employed RNase protection assays with the calcineurin inhibitor, cyclosporine A (CsA). VEGF-mediated induction of DSCR-1 was inhibited 64.1% by 10 nM CsA and 93.3% by 1 M CsA (Fig. 4A, lanes  6 -8). Similarly, thrombin stimulation of DSCR-1 was inhibited by pretreatment with 10 nM and 1 M CsA (by 72.6 and 93.1%, respectively) (Fig. 4A, lanes 9 -11). Next, we wished to deter-mine whether the DSCR-1 promoter contained information for transducing the VEGF/thrombin-calcineurin-dependent signal. To that end, the human DSCR-1 promoter (Ϫ1664 to ϩ83) was isolated and coupled to the luciferase reporter gene (Fig. 4B). The resulting DSCR-1-luc plasmid was transiently transfected into HUVEC. As shown in Fig. 4C, VEGF and thrombin induced DSCR-1 promoter activity by 3.2-and 2.6-fold, respectively, whereas TNF-␣ had no such effect. VEGF-and throm- bin-mediated induction of the promoter was inhibited by 1 M CsA (90.7 and 96.3%, respectively). Taken together, these results indicate that both VEGF-Flk-1/KDR and thrombin-PAR-1 signals induce DSCR-1 mRNA and promoter activity through a calcineurin-dependent pathway.
To delineate the promoter elements responsible for mediating the effects of VEGF and thrombin on DSCR-1 expression, a series of 5Ј-deletion constructs was generated and transiently transfected into HUVEC (Fig. 4B). Most notably, a deletion of the promoter region between Ϫ350/Ϫ166 resulted in a loss of response to VEGF and thrombin (Fig. 4D). The region contains neighboring consensus NF-AT and GATA motifs located at positions Ϫ220 and Ϫ254, respectively. To assess the role of each of these DNA elements in mediating VEGF and thrombin stimulation of DSCR-1, a point mutation of either the NF-AT or GATA motif was introduced and the resulting construct transfected into HUVEC. Each mutation profoundly blocked VEGF/ thrombin-mediated the induction (Fig. 4, E and F). To test the capacity of NF-AT and GATA factors to transactivate the DSCR-1 promoter, we carried out co-transfections in HUVEC with DSCR-1-luc and NF-AT and/or GATA expression plasmids. As shown in Fig. 4G, co-transfection of constitutive active NF-AT resulted in marked transactivation of promoter activity (14.6-fold). Co-transfection of GATA-2 resulted in weak but significant induction of promoter activity (1.71-fold). Importantly, GATA-2 and NF-AT interacted synergistically to induce the promoter by 30.8-fold.
VEGF and Thrombin Promote Binding of NF-ATc, NF-ATp, GATA-2, and GATA-3 to the DSCR-1 Promoter-To study the effect of VEGF and thrombin on NF-AT and GATA DNAprotein interactions, we carried out EMSA using radiolabeled oligonucleotide probes that spanned one or both consensus motifs. In the first set of experiments, a radiolabeled wild-type probe, containing the closely aligned GATA and NF-AT binding sites (Fig. 5A), was incubated with nuclear extracts derived from HUVEC treated in the absence or presence of VEGF, thrombin, or TNF-␣. In control untreated cells, EMSA revealed a specific DNA-protein complex (Fig. 5B, arrow). Importantly, the addition of VEGF or thrombin, but not TNF-␣, resulted in marked increase the DNA binding activity, an effect that was abrogated by pretreatment of cells with 1 M CsA (Fig. 5B,  compare lanes 4, 6, and 8 with lanes 5, 7, and 9). These DNA-protein complexes were inhibited by the addition of 100fold molar excess of the unlabeled self-competitor, but not by the same concentration of the unlabeled NF-AT mutant competitor, or unlabeled GATA mutant competitor (supplemental Fig. 1).
To determine the identity of proteins in the VEGF-and  (Fig. 5, C and D, lane 4, asterisks), whereas anti-NF-ATp antibody resulted in a comparatively weak supershift (Fig.   5, C and D, lane 5). Preincubation with anti-GATA-2 or GATA-3 antibodies resulted in decreased intensity of DNA binding activity (Fig. 5C, lane 6, and D, lanes 6 and 7, hatched  rectangles).
To more clearly assess GATA DNA-protein interactions, a radiolabeled probe containing the NF-AT mutant (Fig. 5A) was   FIG. 7. DSCR-1 attenuates matrix neo-vascularization and tumor-progression in mice. A, matrigel containing 50 ng of VEGF and either 10 9 plaque-forming units of Ad-Control, Ad-DSCR-1, or Ad-DSCR-1 plus Ad-CA-NFAT was injected subcutaneously into C57BL6 mice. After 14 days, matrigel plugs were removed for analysis of new vessel formation by histological sections and hemoglobin assays. Optical magnifications were ϫ100 (upper panels) and ϫ200 (lower panels). The data are representative of seven independent experiments. B, hemoglobin in the matrigel plugs was measured using the Drabkin's reagents and normalized by the weight of matrigel. Data are expressed as means and S.D., n ϭ 5. *, p Ͻ 0.01 compared with Ad-Control and Ad-DSCR-1. #, p Ͻ 0.001 compared with Ad-DSCR-1 and Ad-DSCR-1 and Ad-DSCR-1 plus Ad-CA-NFAT.
[caret], p Ͼ 0.5 (not significant) compared with Ad-Control and uninfected control mice. C, DSCR-1 reduces B16-melanoma growth in mice. Tumor growth was reduced at 4 days (9.1-fold; p Ͻ 0.04), 6 days (7.7-fold; p Ͻ 0.04), and 8 days (7.4-fold; p Ͻ 0.04) by injection of Ad-DSCR-1. D, gross tumors representative of Ad-Control or Ad-DSCR-1 injected groups immediately after resection. E, tumor mass after 14 days following Ad-Control or Ad-DSCR-1 administrations. Data are expressed as means and S.D., n ϭ 5. *, p Ͻ 0.031 compared with Ad-Control. F, representative sections of B16-melanoma at the edge (panels a and c) and center (panels b and d) of the tumor. incubated with nuclear extracts derived from VEGF-treated HUVEC. These experiments resulted in specific DNA-protein complexes (Fig. 5E, lane 2, arrow), which were inhibited by the addition of 100-fold molar excess of unlabeled NF-AT mutant self-competitor but not inhibited by unlabeled GATA mutant competitor (Fig. 5E, lanes 3 and 4). Furthermore, the com-plexes were inhibited by preincubation with anti-GATA-2 or -3 antibodies but not by anti-GATA-6 antibody (Fig. 5E, lanes  5-7). To further study NF-AT binding, a radiolabeled probe containing the GATA mutant (Fig. 5A) was incubated with nuclear extracts derived from VEGF-treated HUVEC. The mixture resulted in a specific DNA-protein complex, which was  inhibited by the addition of 100-fold molar excess of unlabeled GATA mutant self-competitor but not a NF-AT mutant competitor. The DNA-protein complex was strongly supershifted by anti-NF-ATc antibody and only weakly supershifted by NF-ATp antibody (Fig. 5F, lanes 3-6). Similar results were obtained with nuclear extracts derived from thrombin-treated HUVEC (data not shown). The faster migrating DNA-protein complex includes NF-ATc, as it was specifically supershifted by anti-NF-ATc antibody (Fig. 5, C, D, and F, open arrows). Together with the transfection assays, the mobility shift results suggest that VEGF and thrombin induce DSCR-1 expression through the coordinate binding of NF-ATc and GATA-2/3 to closely positioned NF-AT and GATA motifs in the intragenic promoter.

DSCR-1 Inhibits Nuclear Localization of NF-ATc and Tube Formation in Primary Human
Endothelial Cells-We next wished to determine the functional relevance of VEGF-and thrombin-mediated DSCR-1 induction. To that end, we infected HUVEC with IRES-containing adenoviruses expressing both DSCR-1 and EGFP (Ad-DSCR-1) or EGFP alone (Ad-Control). Adenovirus-mediated overexpression of DSCR-1, but not control EGFP, inhibited VEGF-and thrombin-mediated nuclear localization of NF-ATc at 1 h, 94.2 and 92.8%, respectively. (Fig. 6, A and B). In collagen gel assays of HUVEC, VEGF induced the formation of capillary or tube-like structures (compare Fig. 6C, a and b). Compared with Ad-Control, Ad-DSCR-1 infection resulted in marked reduction of tube formation (29.2% of basal level) under the same infection rate and EGFP expression levels (Fig. 6C, panels b-h, and D). Together with the transient transfection and EMSA results, the above findings suggest that VEGF and thrombin each induce the nuclear localization of NF-ATc, cooperative binding of NF-ATc and GATA-2/3 to the DSCR-1 promoter, secondary induction of DSCR-1 mRNA expression, and subsequent attenuation of NF-AT signaling and blood vessel growth.

DSCR-1 Attenuates Cell Proliferation of Primary Human
Endothelial Cells-Angiogenesis involves a complex interplay of myriad cellular functions, including endothelial cell migration and proliferation. To evaluate the effect of DSCR-1 on cell proliferation, we performed FACS analysis with subconfluent HUVEC infected with either Ad-Control, Ad-DSCR-1, or Ad-DSCR-1 plus Ad-CA-NFAT. FACS analysis of Ad-Control-infected HUVEC (at 90% confluence) revealed 67.2, 17.7, and 15.1% in G 0 /G 1 , S, and G 2 /M phases, respectively. Ad-mediated overexpression of DSCR-1 increased the percentage of cells in G 0 /G 1 phase and decreased the fraction in S phase, an effect that was reversed by co-infection with Ad-CA-NFAT (Fig. 6E). Similar findings were obtained with HUVEC at 50% confluence (data not shown). Compared with Ad-Control, Ad-DSCR-1 did not result in increased apoptosis (Fig. 6F), arguing against a toxic effect of DSCR-1 overexpression. Taken together, these findings suggest that the constitutive DSCR-1 expression in endothelial cells induces G 0 /G 1 arrest.

DSCR-1 Blocks Matrix Neo-vascularization and Tumor
Progression in Mice-Having established an inhibitory role for DSCR-1 on cell cycle arrest and angiogenesis in vitro, we wished to evaluate the functional relevance of these findings in vivo. To that end, we investigated the effects of DSCR-1 overexpression on new blood vessel formation in a matrigel plug assay. In these experiments, matrigel containing either Ad-Control, Ad-DSCR-1, or Ad-DSCR-1 plus Ad-CA-NFAT was implanted as subcutaneous plugs into C57BL/6 mice. Fourteen days after implantation, cross-sections from control matrigel plugs demonstrated significant blood vessel formation (Fig. 7A,  left). In contrast, overexpression of DSCR-1 in the matrigel markedly reduced plug vascularity (Fig. 7A, middle). Importantly, the inhibitory effect of DSCR-1 was rescued by coinfection with adenovirus expressing constitutive active NF-AT (Fig. 7A, right). To quantify the extent of neo-angiogenesis, we FIG. 7-continued measured hemoglobin content in the matrigel plugs. Matrigel that contained no virus or Ad-Control demonstrated comparable levels of angiogenesis. However, Ad-DSCR-1 resulted in a significant (67%) reduction of the hemoglobin content. This effect was reversed by co-expression of constitutively active NF-AT (Fig. 7B).
To evaluate whether DSCR-1 has anti-tumor activity, mice were injected subcutaneously with B16-melanoma cells. When tumor size reached 50 mm 3 in volume, the xenografts were injected with Ad-DSCR-1 or Ad-Control. Tumor mass was measured on each subsequent day (Fig. 7C). Compared with control, Ad-DSCR-1 treatment resulted in statistically significant reduction in blood vessel density and tumor size (Fig. 7,  D-F). There was no difference in body weight between the groups (data not shown). Collectively, these findings suggest that DSCR-1, a negative feedback regulator of NF-AT signaling, acts as an inhibitor of angiogenesis.

DSCR-1 Overexpression in Primary Endothelial Cells Results in Down-regulation of Multiple Pro-angiogenic and Pro-
inflammatory Genes-Finally, to gain insight into the broader role for DSCR-1 in auto-inhibition of VEGF and thrombin signaling, HUVEC were infected with Ad-Control or Ad-DSCR-1, treated in the absence or presence of VEGF or thrombin, and then processed for DNA microarray analyses. From the duplicate arrays, a total of 24 genes (from a total of 8974 genes) were induced by VEGF or thrombin and strongly reduced in the presence of DSCR-1 (Fig. 8A). Notably, 10 genes in this group are associated with cell proliferation and angiogenesis and 11 with inflammation. Interestingly, when the filter was changed to the DSCR-1-mediated up-regulated genes, a total of 16 genes emerged, including the cell cycle inhibitor, p21, anti-angiogenesis factor, ADAMTS1 (28), and anti-coagulation factors t-PA, u-PAR, and thrombomodulin ( Fig. 8A and supplemental Table  II). These results, many of which have been validated in RT-  1, 2, and 4)-or Ad-DSCR-1 (lanes 3, 5, and 6)-infected HUVEC treated in the absence (Ϫ) (lanes 1 and 6) or presence of VEGF (lanes 2 and 3) or thrombin (lanes 4 and 5). Cyclophilin A indicates the internal control. The results are representative of at least two separate experiments. PCR (Fig. 8B), suggest that DSCR-1 works as a key negative feedback regulator responding not only with VEGF-and thrombin-mediated angiogenesis and cell proliferation but also with inflammation and hemostatic balance. DISCUSSION The activation of endothelial cells by extracellular stimuli is a key mechanism underlying the development of many vascular diseases. In the present study, we have employed DNA microarrays to analyze the global gene expression profile of activated primary human endothelial cells and show that VEGF and thrombin, but not TNF-␣, resulted in the rapid and pronounced induction of DSCR-1. Activated DSCR-1, in turn, profoundly attenuated calcineurin-dependent NF-AT signaling (see below). Thus, we hypothesize that DSCR-1 functions as a circuit breaker, ensuring balanced regulation of endothelial cell proliferation, migration, and angiogenesis (Fig. 9).
The VEGF and thrombin signals are transduced through calcineurin-dependent nuclear translocation of NF-AT. The NF-AT family of transcription factors includes four structurally related members, NF-ATp, NF-ATc, NF-AT3, and NF-AT4. NF-ATp and NF-ATc have been detected in endothelial cells (29 -31). Previous studies in non-endothelial cells have shown that NF-AT cooperates with other transcription factors, including AP1 (32), GATA (33,34), cMAF (35), or MEF2 (36), to enhance target gene expression. Among the functional sequelae of calcineurin-NF-AT signaling is immune response, cardiac and skeletal and vascular smooth muscle proliferation and/or hypertrophy (37), cardiac valve development (30,38), differentiation of adipocytes and skeletal muscle (39,40), synaptic plasticity, and cellular apoptosis. Here, we show that VEGF and thrombin induce the nuclear translocation of NF-ATc in HUVEC and that NF-ATc cooperates synergistically with GATA-2/3 to transactivate the DSCR-1 promoter. These data are consistent with our previous studies demonstrating an important role for the GATA family of transcription factors as signal transducers in the endothelium (20).
While previous investigations have established the importance of PI3K and MAPK signaling pathways in mediating endothelial cell survival, proliferation, and angiogenesis, comparatively little is known about the functional role for calciumcalcineurin signaling in these biological processes. Activated GSK-3␤ has been shown to promote the cytosol translocation of NF-AT (41,42) and to negatively regulate endothelial cell proliferation and angiogenesis (43). Together with our results, these data support the notion that both calcineurin and PI3K-AKT-dependent inactivation of GSK-3␤ contribute to nuclear retention of NF-AT and secondary endothelial cell migration, growth, and angiogenesis.
Our data suggest that DSCR functions as an endogenous anti-angiogenic factor. In addition to its effect on nuclear localization of NF-ATc and tube formation in primary endothelial cells, DSCR-1 inhibited endothelial cell migration and promoted G 0 /G 1 cell cycle arrest. Most importantly, DSCR-1 blocked matrix neo-vascularization and tumor progression in mice. Moreover, in repeated microarray studies, overexpression of DSCR-1 resulted in down-regulation of pro-angiogenic factors such as PlGF, angiopoietin-2, BMP-2, and PDGF-B and up-regulation of the growth-arrest factors p21 and anti-angiogenic factor ADAMTS1 (28). It will be interesting to determine whether DSCR-1 expression is modulated in tumor endothe- lium and whether such changes are correlated with tumor growth and/or metastases. From a therapeutic perspective, our results point to a potential benefit of overexpressing and/or inducing the activity of DSCR-1 in tumor endothelium.
In addition to its inhibitory effect on endothelial cell proliferation and angiogenesis, DSCR-1 was also shown to downregulate the expression of a number of activation markers in endothelial cells, including ICAM-1, VCAM-1, tissue factor, interleukin-8, and E-selectin. These results are consistent with previous studies demonstrating a link between calcineurin-NF-AT and a pro-inflammatory response. For example, in endothelial cells, CsA-sensitive induction of NF-AT-mediated has been implicated in the expression of granulocyte macrophage colony-stimulating factor and E-selectin (29). Histamine induces interleukin-8 in HUVEC through a CsA-sensitive, NF-AT-dependent mechanism (44). A previous study in HUVEC demonstrated that VEGF induces tissue factor expression via a calcineurin-NF-AT/AP1-dependent mechanism (31). Taken together, these data suggest that DSCR-1 may exert an autoinhibitory break on Ca 2ϩ -calcineurin-NF-AT-mediated endothelial cell activation.
Such a mechanism is reminiscent of the NF-B-I-B␣ autoinhibitory loop. Cellular activation results in phosphorylationdependent degradation and subsequent ubiquitination of I-B␣. As a result, RelA translocates to the nucleus and partners with other members of the NF-B family to transactivate a multitude of target genes, including pro-inflammatory mediators. In addition, the NF-B family induces the early expression of its inhibitor, I-B␣ (3,4), which serves to dampen further RelA activity. Consistent with these results, we have shown that VEGF, thrombin, and TNF-␣ each results in the rapid induction of I-B␣ in HUVEC, an effect that is blocked by pretreatment with proteasome inhibitors and enhanced by cyclohexamide (data not shown). The NF-AT-DSCR and NF-B-I-B␣ negative feedback loops may function to "fine tune" the desired downstream effect of the transcription factor and signal transducer.
During the preparation of this manuscript, Hesser et al. (45) reported that VEGF, but not basic fibroblast growth factor, induces DSCR-1 mRNA and protein in HUVEC. They showed that this occurs through Flk-1/KDR and involves a Cn-dependent mechanism. Finally, consistent with our results, they demonstrated that DSCR-1 suppressed the expression of inflammatory marker genes in activated endothelial cells (45). However, there are several important differences between the two studies. First, Hesser et al. (45) reported that VEGF and TNF-␣ induce comparable levels of DSCR-1 mRNA at 6 h. In our study, we show that VEGF has a much greater effect compared with TNF-␣ on DSCR-1 mRNA at 1 h. Second, we have demonstrated that thrombin induces DSCR-1 expression, suggesting that the auto-inhibitory loop is not specific to VEGF signaling. Third, and most importantly, we have extended the studies to show that DSCR-1 also inhibits VEGF-mediated effects on endothelial cell migration, cell cycle progression, and tube formation in primary human endothelial cells, as well as neo-vascularization and tumor progression in mice.
An important goal in vascular biology is to understand the molecular mechanisms by which the microenvironment regulates vascular function in space and time. In this study, we have analyzed the effect of VEGF and thrombin signals on endothelial cell gene expression and demonstrated a key role for DSCR-1 as a negative feedback regulator common to both mediators. Based on this knowledge, we believe that DSCR-1 may lend itself to therapeutic manipulation in vasculopathic disease states, including tumor angiogenesis.