Thrombin modulates the expression of a set of genes including thrombospondin-1 in human microvascular endothelial cells.

Thrombospondin-1 (THBS1) is a large extracellular matrix glycoprotein that affects vasculature systems such as platelet activation, angiogenesis, and wound healing. Increases in THBS1 expression have been liked to disease states including tumor progression, atherosclerosis, and arthritis. The present study focuses on the effects of thrombin activation of the G-protein-coupled, protease-activated receptor-1 (PAR-1) on THBS1 gene expression in the microvascular endothelium. Thrombin-induced changes in gene expression were characterized by microarray analysis of approximately 11,000 different human genes in human microvascular endothelial cells (HMEC-1). Thrombin induced the expression of a set of at least 65 genes including THBS1. Changes in THBS1 mRNA correlated with an increase in the extracellular THBS1 protein concentration. The PAR-1-specific agonist peptide (TFLLRNK-PDK) mimicked thrombin stimulation of THBS1 expression, suggesting that thrombin signaling is through PAR-1. Further studies showed THBS1 expression was sensitive to pertussis toxin and protein kinase C inhibition indicating G(i/o)- and G(q)-mediated pathways. THBS1 up-regulation was also confirmed in human umbilical vein endothelial cells stimulated with thrombin. Analysis of the promoter region of THBS1 and other genes of similar expression profile identified from the microarray predicted an EBOX/EGRF transcription model. Expression of members of each family, MYC and EGR1, respectively, correlated with THBS1 expression. These results suggest thrombin formed at sites of vascular injury increases THBS1 expression into the extracellular matrix via activation of a PAR-1, G(i/o), G(q), EBOX/EGRF-signaling cascade, elucidating regulatory points that may play a role in increased THBS1 expression in disease states.

Thrombospondin-1 (THBS1) 1 is a large (450 kDa) glycoprotein that is released into the extracellular matrix by several cell types. First discovered in platelets as a thrombin-sensitive protein (1), THBS1 was later shown to be a major component of platelet ␣-granules, comprising some 25% of the total protein (2). Once released, THBS1 exerts various functions on the vascular system under physiological and pathophysiological conditions causing platelet activation (3,4) and adhesion (5) in addition to endothelial cell adhesion (6), promotion of angiogenesis (7), and inhibition of angiogenesis (8), wound healing (9) as well as pathological states such as tumor progression and metastasis (10,11). THBS1 is important not only in cancer progression but also in inflammatory autoimmune diseases such as rheumatoid arthritis (12).
Under normal conditions basal THBS1 expression is negligible. However, increased THBS1 expression has been linked to pathophysiological states such as atherosclerotic lesions (13), vascular smooth muscle cell hyperplasia (14), and hypertension (15). The mechanism by which THBS1 is up-regulated in disease states, however, is not well understood. Here we demonstrate that thrombin, a key player involved in coagulation, inflammation, and wound healing as well as in disease states such as atherosclerosis, can regulate THBS1 expression in endothelial cells.
Thrombin activates a unique class of G-protein-coupled receptors called protease-activated receptors (PARs) by binding to and cleaving the amino-terminal exodomain of the receptor to unmask a new terminus that functions as a tethered ligand, binding intramolecularly to the body of the receptor (16 -18). Four members of this receptor family have been cloned thus far (19). PAR signaling in human platelets is thought to be carried out primarily via PAR-1 and PAR-4, whereas PAR-1 and the trypsin-sensitive PAR-2 (19) are the predominant isoforms expressed in the endothelium (20,21). Coupling to members of the G i/o , G q , and G 12/13 families (22)(23)(24)(25), activation of PAR-1 can lead to modulation of various signaling pathways, which in turn control cell shape and mobility, secretion, integrin activation, and metabolic and transcriptional responses.
High throughput studies on thrombin-induced transcriptional responses in endothelial cells have been limited primarily to human umbilical vein endothelial cell (HUVEC) (26 -29) primary cultures. However, differences between endothelial cell types in cellular function and signaling are evident. For example, macro-and microvascular endothelial cells show heterogeneity and differences both in their signaling activities and protein expression (30 -33). To study changes in transcription that result from thrombin simulation in microvascular endothelial cells, we have chosen to conduct studies in the human dermal microvascular endothelial cell line (HMEC-1). HMEC-1 maintain a similar phenotype to non-transformed HMECs (34) and expresses a complete set of G-proteins that are known to couple to PAR-1 (25). Our experiments on microvascular endothelial cells address the possible roles of THBS1 in processes involving these cells such as neoangiogenesis in cancer and wound healing or tissue repair after or coincidental with inflammation.
Here we report the effects of thrombin on THBS1 gene expression in HMEC-1 using a microarray-based approach. We found THBS1 to be up-regulated both in mRNA and secreted protein levels in response to thrombin. Additionally, thrombininduced up-regulation of THBS1 was pertussis toxin-sensitive, indicating a G i/o signaling component. Inhibition of protein kinase C also inhibited thrombin-induced THBS1 expression, implicating a G q -mediated signaling component. This pathway demonstrates a potential mechanism by which THBS1 may be up-regulated at sites of vascular injury and in disease states.

EXPERIMENTAL PROCEDURES
Reagents-All cell culture reagents were purchased from Invitrogen. ␣-Thrombin, bisindolylmaleimide I hydrochloride (Bis I), pertussis toxin, and Y-27632 were purchased from Calbiochem. The agonist peptide TFLLRNKPDK (TK) was purchased from GL Biochem (Shanghai) Ltd.
Endothelial Cell Culture-In the present studies a human dermal microvascular endothelial cell line that was transformed using SV-40 was used (HMEC-1; obtained from Dr. E. Ades (Centers for Disease Control, Atlanta, GA)). The cells were maintained in MCDB 131 medium supplemented with 5% fetal bovine serum, penicillin/streptomycin (5000 units/ml; 5000 g/ml), hydrocortisone (500 g/ml), epidermal growth factor (0.01 g/ml), and L-glutamine (2 mM) in an atmosphere of 95% air, 5% CO 2 at 37°C. The cells were seeded at 1 ϫ 10 5 cells/ml and subcultured after detachment with 0.05% trypsin, 0.5 mM EDTA. All of the studies utilized cell passages 15-20. Primary cultures of HUVECs were grown in medium 199 supplemented with 15% fetal bovine serum, endothelial mitogen, penicillin, streptomycin, amphotericin B, and heparin. Passage 2-3 HUVECs were plated in 48-well plates and used after confluent.
HMEC-1 Treatment and RNA Isolation-Cells were grown to 95% confluence, switched to serum-free medium containing 0.03% bovine serum albumin, and incubated for 3 h. Cells were then treated with either 86 nM (10 units/ml) human thrombin or vehicle (phosphatebuffered saline, pH 7.4 (PBS)) (control samples) and incubated for additional 90 min or 6 or 12 h. Afterward, cell monolayers were rinsed three times with ice-cold PBS followed by the addition of the lysis buffer (RLT buffer from RNAeasy Midi kit, Qiagen, Valencia, CA), cell scraping, disruption, and homogenization. Total RNA enriched in mRNA was isolated using the RNAeasy Midi kit following the manufacturer's protocol. Only RNA samples with an A 260 / 280 ratio Ͼ 1.8 and no visible degradation by gel electrophoresis and ethidium bromide staining were used for microarray hybridizations and real-time (RT)-PCR.
Synthesis of Fluorescent cDNA and Hybridization to Microarray Slides-Microarray analyses were performed by the Vanderbilt Microarray Shared Resource on a fee-for-service basis. Briefly, 30 g of total RNA from both control and thrombin-treated samples were reverse-transcribed using SuperScript reverse transcriptase (Invitrogen), incorporating Cy5-dUTP into the control sample cDNA and Cy3-dUTP into the thrombin-treated sample cDNA. The labeled cDNA probes were purified using the QIAquick PCR purification kit (Qiagen). The purified cDNA was then mixed with a hybridization buffer containing 10 g of poly(A) RNA, 25% formamide, 5ϫ SSC (0.75 M NaCl and 0.075 M sodium citrate), and 0.1% SDS, denaturated and hybridized for 16 h at 42°C on a 11,000 human cDNA microarray. The cDNAs imprinted in these arrays are available from Research Genetics. For detailed information about the microarray used in these experiments and the genes represented in the array refer to array.mc.vanderbilt.edu. After hybridization, slides were washed sequentially for 5 min each in 2ϫ SSC, 0.1% SDS, 1ϫ SSC, and 0.1ϫ SSC. After drying, slides were immediately scanned for fluorescence emission from each spot on the array at 532 and 635 nm for Cy3 and Cy5, respectively.
Normalization-Fluorescent intensities in each channel, Cy5 and Cy3, for each spot were quantified using GenPix Pro 5.0 from Axon Instrument, Inc. (Union City, CA). The resulting data were then normalized, filtered, and analyzed using GeneSpring 6.1, Silicon Genetics (Redwood City, CA). Experiments were normalized to a signal ratio, and two-color normalizations automatically displayed all measurements per spot-intensity-dependent (Lowess) if more than 100 genes per region divided by control channel if fewer than 100 genes per region. The cutoff used was 10 in raw data or 20% of data used for smoothing.
Filtering-To determine statistically significant data points from the 11,409 genes tested, the data set was filtered using a t test p value test performed by GeneSpring (Silicon Genetics).
GeneSpring correlated the t test values with Student's t distribution chart with n Ϫ 1 degrees of freedom to yield a significance value (p value). Data were then filtered by the p values to include only data that had a p value Յ0.005; this displayed only the normalized mean gene intensities that differed from 1. Any coincidental data that could have passed the first filtration was removed by using a multiple testing correction based on Bonferroni's inequality. This statistical correction helped to further limit the chance of false positives to be no more than p value Յ0.005 by multiplying each nominal p value by N (the total number of genes).
Clustering-A k-means clustering algorithm provided by the Gene-Spring software was implemented to divide the genes of interest into groups based on their expression patterns. The k-means clusters were constructed so that the average behavior in each group was distinct from any of the other groups. Analyzing the time series experiment this way, unique classes of genes that were up-regulated in a time-dependent manner were identified.
The k-means clustering algorithm utilized by GeneSpring first divided the genes of interest into a user-defined number (k) of equal-sized groups based on the order in the selected gene list. Four clusters were chosen. Next, the program created centroids (in expression space) at the average location of each group of genes. Choosing to proceed through 1000 iterations, the genes were reassigned to the group with the closest centroid. The standard correlation algorithm was chosen for the similarity measure. After all of the genes had been reassigned, the location of the centroids was recalculated, and the process was repeated until the maximum number of iterations had been reached.
Promoter Analysis-The Genomatix software Suite (Munich, Germany) was used to predict and analyze promoter regions. Gene2Promoter was used to extract promoter regions for the corresponding genes of each cluster. Common frameworks were then found using the GEMS Launcher task "definition of common framework" using the complete vertebrate matrix library. FrameWoker runs with different quorum constraint parameters were performed; 40 or 30% of input sequences had to contain the framework. ElDorado using "Annotation and Analysis" was subsequently used to analyze the promoter regions of MYC, EGR1 since they did not fall into a common framework.
Literature Mining-Genomatix Bibliosphere was used to mine the literature for co-citations between each of the 65 genes from the filtered list.
Semiquantitative RT-PCR-Universal RT reagents and SYBR Green PCR Master mix were purchased from ABI (Branchburg, NJ). Total RNA (1.0 g) was reverse-transcribed using random hexamers with Universal RT. Reagents were incubated for 10 min at room temperature then 30 min at 42°C followed by 5 min at 99°C and 5 min at 55°C. SYBR Green PCR Master mix was used on an i-Cycler instrument (Bio-Rad) to amplify human cDNAs as well as the cDNA for glyceraldehydes-3-phosphate dehydrogenase (GAPDH) as the internal control. All primers used for real-time semiquantitative PCR were designed using Beacon Designer II (Premierbiosoft). Semiquantitative RT-PCR reactions were carried out in a total volume of 50.0 l containing 5 l of cDNA with the following thermocycling steps: 95°C for 8.5 min, then 45 cycles of 95°C for 15 s, 61.5°C for 1 min, and 95°C for 2.0 min then 55°C for 2 min followed by melting curve data collection with 201 cycles of 0.2°C/8 s temperature escalation and analysis. Specificity and sensitivity of PCR assays were tested with amplification of target cDNAs in serially diluted total RNA samples. Amplicon signals for each target cDNA strongly correlated with serial dilutions of each RNA sample. Data analysis and calculations were done following the 2 Ϫ⌬⌬CT method (35). Each sample was assayed in duplicate. Primer concentration and annealing temperature were optimized for highly specific and reproducible detection of SERPINE1, tissue plasminogen activator (PLAT), THBS1, PAR-1, PAR-2, MYC, EGR1 ␤-actin, and GAPDH RNA by semiquantitative RT-PCR.
Primers-Primers used for semiquantitative real time PCR were purchased from Integrated DNA Technologies (Coralville, IA): SER-PINE1 antisense, 5Ј-AAT GTT GGT GAG GGC AGA GAG-3Ј, and  SERPINE1 sense, 5Ј-CAT TAC TAC GAC ATC CTG GAA CTG-3Ј;  PLAT antisense, 5Ј-GGT CTG GAG AAG TCT GTA GAG-3Ј, and PLAT  sense, 5Ј-CCT AGA CTG GAT TCG TGA CAA-3Ј; THBS1 antisense,  5Ј-CTG ATC TGG GTT GTG GTT GTA-3Ј, and THBS1 sense, 5Ј-CCT  GTG ATG ATG ACG ATG A-3Ј; PAR-1  ELISA Assay of THBS1 in HMEC-1 Culture Medium-The 96-well plates were coated with THBS1 (0.1 g/ml) purified from human platelets (Hematologic Technologies, Inc., Essex Junction, VT) overnight at 4°C. The plates were washed twice with PBS containing 0.02% NaN 3 and blocked with 3% bovine serum albumin in PBS for 3 h at room temperature. After the plates were washed twice with PBS containing 0.02% NaN 3 , 100 l of ELISA buffer (PBS containing 1% bovine serum albumin, 0.05% Tween 20, and 0.02% NaN 3 ) containing rabbit antihuman THBS1 polyclonal antibodies (1:1,000 dilution) (Athens Research and Technology, Athens, GA) and either various concentrations (0.01-5 g/ml) of purified human THBS1 (standard curve) or cell culture medium was added to each well and incubated for 3 h at room temperature. Then the plates were washed twice with PBS containing 0.5% Tween 20 and 0.02% NaN 3 and an additional 2 times with PBS containing 0.02% NaN 3 . Phosphatase-conjugated goat anti-rabbit IgG antibodies (1:1,000 dilution) (Kirkegaard and Perry Laboratories, Gaithersburg, MD) diluted in ELISA buffer were added and allowed to incubate for 3 h at room temperature. The plates were then washed as above, and the substrate solution (pNPP Microwell Substrate System; Kirkegaard and Perry Laboratories) was added and incubated for 1-3 min until a yellow color developed. The reaction was stopped by the addition 5% EDTA. Absorbance at 410 nm was determined on a plate reader (Molecular Device, Union City, CA). Each ELISA was carried out in triplicate and repeated at least two additional times. Data are expressed as mean Ϯ S.E.
Transendothelial Electrical Resistance-In vitro barrier dysfunction was monitored by electric cell-substrate impedance sensor (36). Gold electrodes were purchased from Applied BioPhysics (Troy, NY). Wells were coated in 0.1% gelatin before being wetted by culture media. After treatment, cells were seeded at 2 ϫ 10 5 cells/well and allowed to recover 24 h. The small and larger counter electrodes were connected to a phase-sensitive lock-in amplifier. A constant current of 1 A was applied by a 1 V, 4000 Hz AC signal connected serially to 1 megaohm resistor between the small and large counter electrodes. The voltage between the small electrode and the large counter electrode was monitored by a lock-in amplifier, stored, and processed by a personal computer. The same computer controlled the output of the amplifier and switched the measurement to different electrodes in the course of the experiment. Before the experiments the monolayers were serumstarved 18 h. Cells were pretreated with a sufficient concentration of Y-27632 (25 M) for complete inhibition of Rho kinase (ROCK) for 3 h before stimulation by 10 nM thrombin. Pretreatment did not affect monolayer absolute resistance.
[ 32 P] ADP Ribosylation-Complete ribosylation of G␣ i/o by pertussis toxin was determined as previously described (37)(38)(39). Briefly, cells were grown to confluence in 100-mm dishes and treated with 0.1 g/ml pertussis toxin (PTX) for 3.0 h. Plasma membranes were isolated and subjected to a second round of ADP-ribosylation in vitro in the presence of radiolabeled nicotinamide adenine dinucleotide ([ 32 P] NAD) Amersham Biosciences to determine the amount of toxin substrate remaining. After two rinses with phosphate-buffered saline, membranes were harvested in ice-cold TE buffer (25 mM Tris, pH 7.6, 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml pepstatin A, 10 g/ml leupeptin, and 1 g/ml aprotinin). Cell lysate was homogenized by titration seven times through a 25-gauge needle. Cellular debris was removed by low speed centrifugation (500 ϫ g for 10 min). Supernatants were transferred to fresh tubes, and membranes were precipitated by centrifugation (40,000 ϫ g for 60 min). Membranes were resuspended in 120 l of TE buffer, and total protein was determined using BCA protein assay (Pierce) according to the manufacturer's protocols. PTX was activated by incubation in 25 mM dithiothreitol for 30 min at 37°C. 40 g of purified membranes was incubated with 5 g/ml PTX in reaction buffer (2.5 M NAD, 1 mM ATP, 1 mM GTP, 10 mM thymidine, 6 mM MgCl 2 , 2 mM EDTA, 2 mM dithiothreitol, 20 mM Tris, pH 7.6, and 25 Ci/ml [ 32 P]NAD) in a total volume of 200 l. Reactions were allowed to proceed at 37°C for 60 min. Reactions were quenched by the addition of 20 l of ice-cold 100% (w/v) trichloroacetic acid. Membranes were precipitated by centrifugation (12,000 ϫ g for 20 min) and resuspended in 15 l of 4ϫ Laemmli loading buffer. Samples were boiled for 5 min, resolved by SDS-PAGE, and visualized by autoradiography.

RESULTS
Thrombin Consistently Up-regulates the Expression THBS1 and Other Genes in HMEC-1-To identify candidate genes whose expression is modulated by thrombin stimulation in microvascular endothelial cells, analyses using a microarray were performed. The cDNA library used to make the microarray, developed by Research Genetics, consists of ϳ11,000 different human genes. HMEC-1 cultures were stimulated with a saturating concentration of thrombin, 86 nM (10 units/ml). Total RNA was then isolated from treated and control cells at 1.5, 6, and 12 h after thrombin stimulation. Control and treated cDNAs were simultaneously hybridized to the microchip. The raw data obtained was then quantified, normalized, and filtered according to criteria outlined under "Experimental Procedures." Of the ϳ11,000 genes analyzed, only 65, including THBS1, showed consistent and statistically meaningful differences in expression when cells were treated by thrombin. Fig.  1A is a heat-map representation of the expression patterns of the 65 candidate genes (for a complete list of the genes, accession numbers and numeric -fold increases see the supplemental data) All 65 genes identified were up-regulated at each point of the time course. There were no statistically meaningful genes whose expression was suppressed.
Each gene was then clustered into four groups according to expression profiles using a k-means clustering algorithm. When promoters from each gene of the cluster were analyzed and compared, common putative transcription factor binding sites were identified. These transcription factors were then analyzed to find that transcription models consisting of multiple transcription factor binding motifs, which have been empirically shown to function synergistically, are in close physical proximity (10 -100 base pairs) and were identified in Ͼ40% of the promoter regions. Analysis of Cluster-1 identified four models. The most common transcription factor family in the four models was SP1F.
Cluster 4 contained 10 genes, some of which are marked by their functions in the NF-␤ signaling pathway; nuclear factor of light polypeptide gene enhancer in B-cells inhibitor ␣ (NFKBIA), tumor necrosis factor, ␣-induced protein 3 (TN-FAI3), and syndecan 4 (SDC4). Transcription factors such as serum response factor (SRF) and Fos-related antigen-2 (FOSL2) were also expressed as well as the growth factor connective tissue growth factor. Promoter-cluster analysis was performed for the genes of Cluster-4 identifying nine models, NFKB being the most common transcription factor family.

Thrombin Induction of THBS1 Expression Confirmed by RT-PCR-To confirm the microarray results obtained for THBS1
and a subset of other thrombin-induced genes, semiquantitative RT-PCR was preformed. Samples were harvested 6 h after thrombin treatment, the time of maximal THBS1 induction. Thrombin induced THBS1, SERPINE1, PLAT, and connective tissue growth factor by 3.1, 4.2-, 3.0-, and 1.8-fold, respectively (Fig. 1C). These data correspond well with the microarray data of 2.6-, 3.1-, 3.1-, and 1.9-fold, respectively, confirming those results. In addition, expression of both PAR-1 and PAR-2 was analyzed via RT-PCR and found to be induced by 3.3-and 1.8-fold, respectively. These data confirm previous studies that quantified thrombin-induced PAR-1 expression in HMEC-1 using Northern blot analysis (41) and demonstrate that thrombin can induce PAR-2 expression. These results suggest that, although corresponding cDNAs to the PARs are found on the chip used in the present studies, the results were below the strin- gent statistical criteria employed to analyze the data set.
Confirmation of PAR-1-mediated THBS1 Protein Expression-Endothelial cells that express THBS1 continuously secrete it into the extracellular matrix where it can interact with its physiological binding partners. Although the precise mechanism that governs its secretion has not been fully elucidated, a direct correlation between THBS1 transcription and secretion into culture media has long been observed (42,43). To confirm that the increase in THBS1 mRNA correlated with an increase protein expression and to quantify the amount of THBS1 released into the culture medium, a competitive ELISA approach was employed. HMEC-1 cultures were treated with or without 86 nM (10 units/ml) thrombin, and the culture media were harvested as before and assayed as described under "Experimental Procedures" (Fig. 2A). At 4, 8, and 12 h, culture media harvested from thrombin-treated cells contained greater amounts of THBS1 compared with control cells.

Thrombin Induction of THBS1 Expression in HUVECs-
Differences in cellular responses and gene expression between endothelial cell types has been well documented (30 -33). To determine whether THBS1 expression is up-regulated by thrombin stimulation in another endothelial cell type, similar experiments were preformed using HUVECs. HUVEC cultures were stimulated with or without 86 nM (10 units/ml) thrombin, and the -fold increase in thrombin-induced THBS1 expression 24 h post-stimulation was determined by competitive ELISA using thrombin stimulated HMEC-1 as a control (Fig. 2B). Thrombin induced THBS1 expression in HUVECs to a slightly greater extent than in HMEC-1. These data confirm in two different endothelial cell types that thrombin mediates an increase in THBS1 expression.
Thrombin-induced THBS1 Expression Is Dose-dependent-To further characterize the ability of thrombin to regulate THBS1 expression, the thrombin dose-response was investigated. Using competitive ELISA as before, Fig. 3 shows the dose-response curve for thrombin-mediated THBS1 protein expression in HMEC-1. The EC 50 value was 4.6 (2.6-7.9) nM with an approximate of 1.9 (1.8-2.1)-fold increase at saturation, at the 95% confidence interval.
Thrombin-induced THBS1 Expression Is PTX-and Protein Kinase C-sensitive-We have previously demonstrated in HMEC-1 that PAR-1 effectively couples to G i/o , resulting in activation of extracellular signal-regulated kinase (41). To determine whether activation of G i/o is necessary for thrombininduced THBS1 expression, HMEC-1 were pretreated with PTX, and the effects on THBS1 expression were determined via competitive ELISA as before. Pretreatment with PTX inhibited the response by ϳ70% (Fig. 4A). The efficiency of PTX treatment was controlled for by a substrate depletion assay. Pretreatment with PTX for 3 h completely inactivates all G␣ i/o subunits (Fig. 4B). These results suggest that activation of G i/o is essential for thrombin-induced THBS1 expression.
Endothelial PAR-1 has been shown to activate ROCK via the G 12/13 -mediated pathway (50). To determine whether ROCK activity is necessary for thrombin-induced THBS1 expression, HMEC-1 were pretreated with the specific ROCK inhibitor Y-27632. Pretreatment with 25 M Y-27632 had no statistically significant effect (Fig. 4A). Y-27632 inhibition of ROCK was controlled for using a barrier function assay. PAR-1-induced endothelial barrier dysfunction is dependent upon ROCK activity; thus, HMEC-1 were grown on gold electrodes, and the transendothelial electrical resistance was measured in response to thrombin. Pretreatment for 3 h with 25 M Y-27632 did not affect the absolute resistance of the monolayer but completely inhibited the transient monolayer barrier dysfunction induced by 10 nM thrombin, Fig. 4C. The data indicate Y-27632 treatment under these conditions completely inhibits

FIG. 2. Confirmation of PAR-1-induced THBS1 protein expression in HMEC-1 and HUVECs.
A, the concentration of secreted THBS1 into the culture media from HMEC-1 was determined via competition ELISA from thrombin (86 nM)or PAR-1-TK (100 M)-stimulated cells and control cells at the time points indicated. Statistical analysis was performed using the two-tailed t test; ** indicated a p value Ͻ0.01 for both thrombin or PAR-1-TK compared with control; *** indicates p values Ͻ 0.001 for both thrombin or PAR-1-TK compared with control. B, the effects of thrombin (86 nM) treatment on THBS1 induction in HUVECs were determined via competition ELISA from cultured media harvested 24 h after stimulation using similarly treated HMEC-1 as a control. Data represent at least three separate experiments. Statistical analysis was performed using the two-tailed t test; ** indicated a p value Ͻ 0.01. All results are shown as the mean Ϯ S.E. ROCK activity. These results suggest G 12/13 activation does not play a role in thrombin-induced THBS1 expression.
PAR-1 is known to activate not only G i/o and G 12/13 , the effects of whose activation is addressed above, but G q as well. In endothelial cells, activation of G q causes an increase in intracellular calcium mobilization and activation of protein kinase C (51, 52), a regulator kinase that has been shown to modulate the expression of transcription factors in endothelial cells (53,54). To address whether PAR-1 activation of G q affects thrombin-induced THBS1 expression, HMEC-1 were pretreated for 30 min with 5 M Bis I before stimulation with or without 86 nM (10 units/ml) thrombin. Secreted THBS1 was quantified using the competitive ELISA as before. Inhibition of protein kinase C with Bis I resulted in ϳ50% inhibition of thrombin-induced THBS1 expression (Fig. 4A). These results suggest that G q -mediated activation of protein kinase C plays a functional role. HMEC-1-THBS1 is a high molecular weight glycoprotein of the extracellular matrix. Increased THBS1 expression has been linked to pathophysiological states such as atherosclerotic lesions (13), hyperplasia (14), hypertension (15), and cancer (11). The mechanism by which THBS1 is up-regulated in disease states, however, is not well understood.

Regulation of THBS1 Gene Expression by Thrombin in
Although THBS1 can originate from thrombin-induced platelet activation via release from ␣-granules, THBS1 is also expressed and continuously secreted by many cell types of the vascular system including the endothelium (42,43). The effects of thrombin on gene expression and specifically that of THBS1 expression in the surrounding endothelium are less well defined. However, THBS1 has been shown to be up-regulated in vasculature cells in response to glucose (55). To characterize on a more complete scale the effects of thrombin on gene regulation and expression in endothelial cells, a microarray analysis approach was used. Of the potential 11,000 targets, surprisingly only 65 different genes including THBS1 were affected over a time course of 1.5, 6, and 12 h after thrombin treatment ( Fig. 1A and supplemental data). Four distinct expression patterns were found into which the genes were clustered (Fig. 1B). THBS1 along with a subset other genes identified in the microarray studies were confirmed by semiquantitative RT-PCR (Fig. 1C). Interestingly, of the 65 gene identified, all were up-regulated in response to thrombin stimulation. The microarray did not reveal any statistically valid genes whose expression was suppressed.
Similar microarray studies using HUVECs have been performed by others (26). In common, four transcripts were induced by thrombin; small inducible cytokine A2 or monocyte chemotactic protein-1 (CCL2), immediate early response 3 (IER3), and nuclear factor of light polypeptide gene enhancer in B-cells inhibitor-␣ (NFKBIA). Interestingly, these four hits corresponded in magnitude and temporally. However, the remaining 61 genes including THBS1 identified here differ from those induced in other endothelial cells, demonstrating the need to characterize multiple cell types. Although THBS1 was not identified in previous studies using HUVECs, we show here by ELISA that thrombin induces THBS1 expression in HUVECs.
Thrombin Regulation of THBS1 Expression in Both HMEC-1 and HUVECs-Increased levels of secreted THBS1 determined by ELISA correlated with an increase in mRNA both temporally and quantitatively over the 12-h time course ( Fig. 2A). Although endothelial cells express PAR-1, -2, and -3, the results of the PAR-1-specific peptide mimicking those of thrombin indicated that PAR-1 activation is sufficient to mediate thrombin regulation of THBS1 gene expression (open squares compared with open circles Fig. 2A). Thrombin-regulation of THBS1 appears also to be a property of the endothelium, as thrombin induced a ϳ2.7-fold increase in secreted THBS1 levels in culture media harvested 24 h post-stimulation from HUVECs, compared with a ϳ1.7-fold in HMEC-1 controls (Fig. 2B). THBS1 protein expression was also shown to be thrombin dose-dependent (Fig. 3). The EC 50 from the dose-response curve was 4.6 nM. This is similar to the other EC 50 values reported for other PAR-1-mediated responses such as formation of the second messenger inositol phosphate formation in HUVECs (ϳ3.4 nM) (56) and correlates well with the physiologic range of thrombin (57). Taken together, these results suggest that at physiologic concentrations of thrombin near the EC 50 value, endothelial cells respond by up-regulating THBS1 production via a PAR-1-mediated pathway.
THBS1 Promoter Analysis-To identify potential pathways that might lead to PAR-1 induction of THBS1, the promoter region of THBS1 was analyzed. Fig. 5A is a representation of   FIG. 4. Effects of Y-27632, PTX, and Bis I treatment on thrombin-induced THBS1 expression. The concentration of secreted THBS1 in HMEC-1 culture medium was determined as described in Fig. 2. A, cells were pretreated for 3 h with 25 M Y-27632 or 0.1 g/ml PTX or for 30 min with 5 M Bis I before stimulation with 86 nM thrombin. Statistical analysis was performed using the two-tailed t test; ** indicates a p value Ͻ 0.05; ns indicates non-statistically significant. Data represent the average of three independent experiments Ϯ S.E. B, cells were treated with 0.1 g/ml PTX. Membranes were harvested 3 h after treatment. Harvested membranes were subjected to a second round of PTX treatment in the presence of radiolabeled NAD as substrate. Samples were resolved by SDS-PAGE and visualized by autoradiography. The autoradiogram is representative of three independent experiments. C, changes in barrier function monitored using ECIS were determined in HMEC-1 monolayers pretreated for 3 h with PBS (Control) or 25 M Y-27632 then stimulated with 10 nM thrombin. Data represent the average of three independent experiments Ϯ S.E. the predicted promoter site consisting of 600 bp, 500 bp upstream to 100 bp downstream of the transcription start site. The first representation shows all potential binding sites for the 41 different transcription factor families identified.
An advantage of the high throughput approach of microarray analysis is the accumulation of large data sets from which patterns of information can be extracted. To increase the likelihood of a potential transcription factor binding site being functional in the current signaling context, the promoter regions of all genes that clustered by expression profile were also analyzed. Given that the genes of each cluster display similar expression profiles within their respective clusters, it is likely that a population of those genes are transcriptionally under similar control. The second representation of Fig. 5A shows only those transcription factor binding sites common to members of the cluster.
The transcription factors were then further analyzed for transcription factor models consisting of multiple transcription factor binding motifs that have been shown empirically to function synergistically, are in close physical proximity (10 -100 base pairs), and were present in Ͼ40% of the promoter regions of the cluster. Two such models were identified for THBS1. The final representation of Fig. 5A shows one model consisting of an EBOX and EGR binding module. The second predicted model consisted of an EBOX and ZBPF module.
To determine which of the two predicted transcription models was more likely to describe thrombin-induced THBS1 expression, the expression patterns of all the members of each transcription factor family predicted by both models were initially screened using the unfiltered results from the microarray analysis. Only transcription factors for the EBOX and EGR families appeared to be thrombin-induced.
Quantitative RT-PCR confirmed prominent transcription factors from each family MYC and EGR1, respectively, are each up-regulated at the point of maximal THBS1 expression (6 h) (Fig. 5B). These results suggest that thrombin-induced THBS1 expression may be regulated first by the expression of these transcription factors. This would be consistent with the observation that transcription factors of the EBOX and EGR families have relatively short half-lives in the cell (58,59).
Similar promoter analyses of the MYC and EGR1 genes have revealed possible models of their transcriptional regulation. This method predicted a well documented NF␤ pathway of MYC expression (68 -74). Analysis also predicted an experimentally determined model involving ETS and SRF transcription factor family regulation of EGR1 via G i/o -mediated activation of extracellular signal-regulated kinase (54,(75)(76)(77)(78)(79).
We have previously shown that thrombin can induce extracellular signal-regulated kinase activity via a PTX-sensitive pathway in HMEC-1 (41). These data are consistent with our present findings that pretreatment of cells with PTX inhibited thrombin-induced THBS1 expression by ϳ70% (Fig. 4, A and   FIG. 5. THBS1 comparative promoter analysis and proposed signaling pathway. The figures in panel A depict the promoter region of THBS1 (top), transcription factor (TF) binding sites common to THBS1 and other members of Cluster-3 (middle), and a predicted transcription promotion model consisting of an EBOX and EGR motif (bottom). B, total mRNA was isolated 6 h after HMEC-1 were stimulated with 86 nM thrombin.
This pathway model predicts that signaling by both G i/o and G q are necessary to increase THBS1 expression. Consistent with that model, we show inhibition of G i/o by PTX treatment inhibited the response. Likewise, inhibition of protein kinase C, whose activity is dependent upon G q -mediated calcium mobilization, by treatment with Bis I also inhibited the response, Fig. 4A.
We also observed that treatment with the ROCK inhibitor Y-27632 had no effect (Fig. 4, A and C). It has been reported by others that phosphorylated MYC negatively regulates THBS1 expression in epithelial cells and that the phosphorylation state is ROCK-dependent (80). Our results are consistent, temporally, with the expression pattern of THBS1. They suggest that G 12/13 activation of ROCK peaks and returns to base line well before MYC expression is achieved, thereby not affecting the phosphorylation state of MYC. This is consistent with the temporal kinetics of Rho activity after thrombin stimulation of endothelial cells.
In addition, thrombin has been shown to activate the transcription factor, Y-box DNA-binding protein B (dbpb) in endothelial cells to induce platelet-derived growth factor ␤ expression (81,82). Although platelet-derived growth factor ␤ was found to be thrombin-induced in the present microarray studies (clustering to group 2), the promoter analysis of this group or any other cluster did not reveal any putative Y-box-containing transcription models.
Induction of an antiangiogenic factor such as THBS1 at the site of vascular injury at first might seem paradoxical. However, one must consider the time delay involved in THBS1 production. Maximal THBS1 transcription is achieved by ϳ6 h (Fig. 1). The importance of negatively regulating angiogenesis after wound healing is completed is apparent. An increase in THBS1 might be one means by which the cell regulates the temporal aspects of angiogenesis in response to vascular injury.
In conclusion, it is clear that expression of THBS1 in endothelial cells is up-regulated by exposure to thrombin. This demonstrates that THBS1 can originate at sites of vascular injury from the endothelium in addition to platelet activation. Given the role THBS1 plays in determining platelet activation, adhesion of the hemostatic plug and the angiogenic fate of the surrounding vasculature, pathways like those which we report here for thrombin induction of THBS1 provide important information about the complex nature of THBS1 regulation and may one day contribute to therapeutic approaches.