Endothelial Cell Confluence Regulates Cyclooxygenase-2 and Prostaglandin E2 Production That Modulate Motility*

Endothelial cells line the vasculature and, after mechanical denudation during invasive procedures or cellular loss from natural causes, migrate to reestablish a confluent monolayer. We find confluent monolayers of human umbilical vein endothelial cells were quiescent and expressed low levels of cyclooxygenase-2, but expressed cyclooxygenase-2 at levels comparable with cytokine-stimulated cells when present in a subconfluent culture. Mechanically wounding endothelial cell monolayers stimulated rapid cyclooxygenase-2 expression that increased with the level of wounding. Cyclooxygenase-2 re-expression occurred throughout the culture, suggesting signaling from cells proximal to the wound to distal cells. Media from wounded monolayers stimulated cyclooxygenase-2 expression in confluent monolayers, which correlated with the level of wounding of the donor monolayer. Wounded monolayers and cells in subconfluent cultures secreted enhanced levels of prostaglandin (PG) E2 that depended on cyclooxygenase-2 activity, and PGE2 stimulated cyclooxygenase-2 expression in confluent endothelial cell monolayers. Cells from subconfluent monolayers migrated through filters more readily than those from confluent monolayers, and the cyclooxygenase-2-selective inhibitor NS-398 suppressed migration. Adding PGE2 to NS-398-treated cells augmented migration. Endothelial cells also migrated into mechanically denuded areas of confluent monolayers, and this too was suppressed by NS-398. We conclude that endothelial cells not in contact with neighboring cells express cyclooxygenase-2 that results in enhanced release of PGE2, and that this autocrine and paracrine loop enhances endothelial cell migration to cover denuded areas of the endothelium.

Angiogenesis, the formation of new blood vessels from preexisting blood vessels, is essential for wound repair, tumor growth, and metastasis (1). Endothelial cell migration from the confluent monolayer of endothelial cells of mature vessels into matrix underlies this process. Prostanoids, collectively E 2 (PGE 2 ), 1 PGF 2␣ , PGD 2 , PGI 2 , and thromboxane A 2 , are cy-clooxygenase products involved in angiogenesis and tumor growth (2,3). Individual prostanoids are recognized by a family of G protein-coupled receptors whose distribution controls the prostanoid signaling axis (4).
There are two isoforms of cyclooxygenase, cyclooxygenase-1 and cyclooxygenase-2, that have both shared and separate functions (5). Cyclooxygenase-1, found in many tissues, typically is constitutively expressed, although it and not cyclooxygenase-2 is induced in uterine endothelial cells in the third trimester of pregnancy when PGI 2 levels and blood flow increase (6). Cyclooxgenase-2 typically is absent from endothelial cells and white blood cells but accumulates to high levels in endothelial cells over several hours in response to IL-1␤ (7), lipopolysaccharide (8), phorbol myristate acetate (8), TNF␣ (9), and oxidized phospholipids (10). Accordingly, cyclooxygenase-2 has numerous transcriptional regulatory elements in its 5Јregulatory region (11,12) and also is subject to post-transcriptional control (13).
Cyclooxygenase-2 is dramatically induced by growth factors, tumor promoters and mitogens, and is aberrantly expressed in tumors, including those of colon (14), breast, and prostate (15). Cyclooxygenases are the targets of non-steroidal anti-inflammatory drugs (NSAIDs), and NSAIDs decrease cancer risk and suppress tumorigenesis in animal models (15). NSAIDs inhibit endothelial cell spreading, migration, and angiogenesis (16), processes controlled by PGE 2 (17), just as genetic ablation of cyclooxygenase-2 blocks the growth of cyclooxygenase-2-replete tumors by suppressing angiogenesis (3). These gene-targeted animals show that it is cyclooxygenase-2 expression in the host stromal cells, including endothelium, and not by the tumor cells themselves that is critical and suggest that host autocrine and paracrine signaling has a role in tumorigenesis. Tumors express high levels of PGE 2 , and pharmacologic suppression of cyclooxygenase-2 activity, and not that of cyclooxygenase-1, depletes PGE 2 and blocks tumorigenesis (18). A similar result was obtained when PGE 2 was depleted with a monoclonal antibody (18), so PGE 2 is one factor controlling angiogenesis and tumor growth.
Endothelial cells migrating during angiogenesis necessarily lack a neighboring cell at the leading edge of the nascent tubule. We find that endothelial cells not completely surrounded by neighboring endothelial cells and those not embedded in a confluent monolayer of cells display a characteristic of highly activated cells, cyclooxygenase-2 expression. Re-expression of cyclooxygenase-2, which was down-regulated as cells formed a monolayer, results in enhanced PGE 2 secretion that aids endothelial cell migration to reestablish a confluent monolayer of endothelial cells.

MATERIALS AND METHODS
Reagents-NS-398 was purchased from Biomol (Plymouth Meeting, PA); the monoclonal antibodies against cyclooxygenase-1 and cyclooxy-* This work was supported in part by National Institutes of Health Grants HL44513 and HL44525. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cell Culture and Monolayer Wounding-Human umbilical vein endothelial cells were isolated and cultured as described (19). These cells were allowed to achieve confluence, typically in 3-5 days, and then were serum-starved by culturing for 24 h in media containing 1% human serum before initiating our experiments. Cells were maintained in M199 supplemented with 20% pooled human serum at 37°C in a humidified atmosphere of 95% air and 5% CO 2 . The wound repair model used freshly isolated endothelial cells grown to confluence in 6-well tissue culture plates before the cells were mechanically removed from the plate by dragging a 1000-l pipette tip along the confluent monolayer as the plate rested over a template. Low wounding consisted of seven cell-free lines formed in the monolayer and medium wounding employed 21 parallel wound lines. For the heavily wounded monolayers, the plates were rotated 90°after the first medium wounding, and then an additional 21 parallel lines were etched in the monolayer to form square islands of remaining cells. The monolayers were washed after wounding to remove the debris and fed fresh medium that also contained any agonist to be included in the experiment. Images of the wounded cultures were recorded just after medium wounding (t ϭ 0) and then 24 h later for most experiments. Four fields were recorded for each sample, and the assays were repeated 10 or more times.
Endothelial Cell Migration-Migration of endothelial cells was determined with a modified Boyden chamber assay. Briefly, polycarbonate filter wells (3-m pores, BD Biosciences) were coated with fibronectin (10 g/ml) for 1 h at room temperature and washed once with phosphate-buffered saline. Endothelial cells from confluent monolayers or individual cells from the same isolation plated at a lower density to preclude monolayer formation were freed from the dish by trypsinization. The cells were then washed by low speed centrifugation and resuspended in culture medium containing a reduced concentration (1%) of human serum. The recovered cells (1.5-2.5 ϫ 10 4 ) were added to the upper chamber of a transwell insert, and the filter inserts were incubated in wells of a 24-well culture plate containing 750 l of medium. NS-398, when present, was added to both the upper and lower chambers. Basic fibroblast growth factor (10 ng/ml) was added to the lower chamber as a positive control. After 21-22 h, the cells were stained with calcein AM (4 g/ml for 30 min), and cell migration was quantitated by measuring the fluorescence of the migrated cells in a fluorescent plate reader (Fusion, Hewlett-Packard) using its bottom reading capability. Alternatively, photographs of the migrated cells were recorded by confocal microscopy. Four randomly selected low power (ϫ10) fields were chosen, and the number of cells in each field was counted. The migration response was expressed as fold increase over base line where each condition was assayed in triplicate wells and each experiment was repeated at least twice. Student's t tests (Graph-Pad Instat) showed all changes were significant (p Ͻ 0.05).
Cell Proliferation-Inhibition of the cell cycle at various points was accomplished by growing the endothelial cells to confluence in 20% human serum, washing the cells, and then starving them in 1% human serum for 17 h in the presence of agents that interfere with cell cycle progression at distinct stages. Aphidicolin (5 g/ml) and mimosine (1 mM) block cell cycling during late G 1 /S phase; 5-fluorouracil (10 M) and 1␤-D-arabinofuranosylcytosine (Ara-C; 1 M) interfere with deoxynucleotide synthesis in S phase; and nocodazole (0.5 g/ml) blocks microtubule depolymerization required for M phase. The monolayers were then wounded, or not, in the high wounding pattern and incubated with the stated agents for 8 h before cellular material was collected for analysis of cyclooxygenase-2 by Western blotting.
Immunofluorescence-Endothelial cells were grown to confluence in 8-well glass chamber slides coated with fibronectin or maintained at a lower density to preclude monolayer formation as before. Multiple wound lines were made in each chamber of confluent endothelial cells using a 200-l pipette tip, and the remaining cells were washed and then fed with endothelial cell medium containing 1% human serum. After the specified time, the medium was flicked out of the wells, and chambers were peeled off. Cells on the slides were fixed in 2% paraformaldehyde followed by permeabilization with 0.5% TRI 301 for 5 min, and incubated with cyclooxygenase-2 antibodies (1:1000) overnight at 4°C. The next day, the slides were developed with biotinylated goat anti-mouse immunoglobulin (2 g/ml) for 1 h, followed by Alexa 488-labeled streptavidin for 45 min at room temperature. Propidium iodide (15 g/ml for 5 min) was used to stain the nuclei before the images were recorded by confocal microscopy.
Immunoblot Analysis-Cells were washed twice with phosphatebuffered saline and then with iced cell lysis buffer (20 mM Tris/HCl, 16 mM CHAPS, 0.5 mM dithiothreitol, 1 mM EDTA, 1 mM benzamidine hydrochloride, 1 g/ml leupeptin, 10 g/ml soybean trypsin inhibitor). A cell scraper removed the cellular material, and the lysates were kept on ice for 30 min and then centrifuged at 4°C for 5 min at 10,000 ϫ g. Protein content of the supernatants was quantitated using a BCA protein assay (Pierce). Thirty micrograms of each sample was mixed with sample buffer and resolved in a 10% acrylamide gel by electrophoresis. Proteins in the gel were transferred to polyvinylidene difluoride membranes, and the resulting membranes were blocked overnight at 4°C with 5% dried milk and then incubated with antigen-specific antibodies (cyclooxygenase-2, Cayman Chemical Co; ␤-actin, ICN Biomedicals, Aurora, OH) and further reacted with appropriate secondary antibody. The horseradish peroxidase of the secondary antibody was detected by chemiluminescence (Amersham Biosciences) according to the manufacturer's directions. PGE 2 ELISA-Endothelial cells were grown to confluence, or their cognate subconfluent cultures from the same cord, in 6-well plates and extensively wounded with a 1000-l pipette tip as described above. The residual cells were washed with serum-free endothelial cell medium and fed fresh endothelial cell growth medium containing 1% human serum. After 8 h, this medium was collected, and PGE 2 was quantitated by a sandwich ELISA according to the manufacturer's protocols. Some cultures were supplemented with arachidonic acid by washing the culture with Hanks' buffered saline solution and then adding 20 M arachidonic acid in Hanks' buffered saline solution containing 0.1% human serum for 30 min before the medium was collected for PGE 2 ELISA. The data in these experiments were normalized by cell number.

Cyclooxygenase-2 and PGE 2 Production Are Regulated by
Cell Contact-Inducible cyclooxygenase-2 has a key role in tumor angiogenesis (15,20), and the prostaglandin PGE 2 it produces stimulates angiogenesis (21,22) and endothelial cell adhesion and spreading (17). Endothelial cells primarily exist as a confluent monolayer of quiescent cells ( Fig. 1A) but rapidly spread, migrate, and proliferate in response to mechanical denudation as occurs during invasive procedures. We found that confluent monolayers of endothelial cells cultured for 24 h in a reduced amount of serum expressed trace amounts of cyclooxygenase-2 protein by immunocytochemistry (Fig. 1A, lower left). However, cells cultured in numbers insufficient to form a monolayer displayed a sharply increased level of cyclooxygenase-2 expression (Fig. 1A, lower right). Staining shows the enzyme was associated with punctate intracellular structures and particularly stained the nuclear membrane, as anticipated (23). The amount of cyclooxygenase-2 staining by individual cells not organized into a monolayer was equivalent to that found in confluent endothelial cell cultures after treatment with powerful stimuli, such as inflammatory cytokines (TNF␣ and IL-1␤), lipopolysaccharide, or a phorbol ester (Fig. 1B).
Cyclooxygenase-2 protein expression in cells maintained in a confluent state was not detectable by Western analysis (Fig.  2A). In contrast, cells at 40% of the final density of a confluent monolayer expressed this enzyme, and cells seeded even more sparsely to give a density of 20% of confluent cultures contained more cyclooxygenase-2 than this. These samples were normalized for protein content before electrophoresis, which produced equal levels of staining for ␤-actin, and so this cyclooxygenase-2 staining reflects average cellular content. This enzyme was fully functional because endothelial cells not organized into a confluent monolayer made and released about 17 times more PGE 2 than confluent cultures on a per cell basis when provided with exogenous arachidonate (Fig. 2B). Subconfluent endothelial cells continuously produced twice as much PGE 2 as their confluent counterparts even when exogenous arachidonate was not provided to overcome the limiting level of endogenous arachidonate (Fig. 2C). Endothelial cell production of PGE 2 from both confluent monolayers and individual cells primarily was a function of cyclooxygenase-2 activity because the non-steroidal anti-inflammatory drug NS-398 that selectively inhibits cyclooxygenase-2 effectively blocked PGE 2 synthesis and release by endothelial cells cultured under either condition (Fig. 2C).
Wounding Endothelial Cell Monolayers Stimulates Cyclooxygenase-2 Expression Proximal and Distal to the Wound-Endothelial cells in a confluent monolayer clearly differ from individual cells before they organize and establish intercellular communications. We tested whether an abrupt disruption of a confluent monolayer, as occurs during invasive clinical procedures, affects cyclooxygenase-2 expression in the same way as seeding the culture at a low density. To do this, we denuded sections of tightly confluent monolayers by dragging a pipette tip along the plate in a pattern to create three levels of wounding (Fig. 3A). We found by Western analysis (Fig. 3B) that cyclooxygenase-2 was present in the cells remaining after dide counter-stained nuclei. B, endothelial cells unable to form monolayers expressed cyclooxygenase-2 at levels equivalent to agonist-activated monolayers of endothelial cells. Endothelial cells were plated at a density that allows rapid monolayer formation or at lower density not compatible with monolayer formation. Confluent cultures were stimulated with the stated agonist (5 ng/ml IL-1␤, 50 ng/ml TNF␣, 100 ng/ml Escherichia coli lipopolysaccharide (LPS), or 0.5 M phorbol myristate acetate (PMA), or not, for 4 h) before the cells were fixed and stained for cyclooxygenase-2 protein expression and nuclear DNA as above. Subconfluent cells were not exposed to an exogenous agonist. wounding the monolayer and that the level of protein expression increased with the number of lines drawn through the culture. The amount of cyclooxygenase-2 accumulated by wounded cultures was nearly that reached following phorbol ester stimulation. Cyclooxygenase-1 did not change by wounding the monolayer, although phorbol myristate acetate stimulation modestly enhanced the cellular content of this isoform.
We visualized cyclooxygenase-2 expression by wounded monolayers to determine whether cells adjacent to the wound line, which are the cells directly affected by the wounding procedure, were the only cells to express cyclooxygenase-2. We found that each cell adjacent to the wound expressed this enzyme, but many cells distal to the wound also expressed immunoreactive cyclooxygenase-2 (Fig. 3C). Enhanced expression of cyclooxygenase-2 was detected just 2 h after wounding the monolayer, with a further enhancement 4 -8 h after wounding. In Fig. 3C, the lower two panels are composite images that show the cells remaining between two wound lines, with the wounds being the black acellular area at the ends of the picture marked by the introduction of a white dotted line. These images show that the majority of the green fluorescence representing cyclooxygenase-2, and indeed the brightest cells, was found well away from the wound edge, so cyclooxygenase-2 expression occurs in cells not mechanically affected by the wounding or by loss of neighboring cells.
Endothelial Cells Release PGE 2 after Wounding, Which Stimulates Cyclooxygenase-2 Expression-Confluent cultures of endothelial cells make and release PGE 2 , and wounding the monolayer increased this secretion (Fig. 4A, upper panel). The amount of PGE 2 released from the cells remaining in the mechanically disturbed monolayer varied with the extent of wounding, and the most heavily scored monolayers released FIG. 3. Cyclooxygenase-2 and PGE 2 production are stimulated by wounding endothelial cell monolayers. A, patterns of monolayer wounding. Endothelial cells were grown to confluence in 6-well plates, and then portions of the monolayer were removed by scraping a pipette tip over the plate to create 7 (low wounding), 21 (medium wounding), or 42 (high wounding) cell-free lines. B, wounding confluent endothelial cell monolayers induced cyclooxygenase-2 (COX-2) re-expression that correlates with the level of wounding. The 6-well plates were washed after wounding and the remaining cells fed with media containing 1% pooled human serum and incubated for 4 h before the cells were lysed, and material was recovered for protein assay and electrophoresis. Equal amounts of protein were loaded on a 10% SDS-polyacrylamide gel, the proteins resolved by electrophoresis, transferred to polyvinylidene difluoride membranes, and immunoblotted with antibodies to cyclooxygenase-1, cyclooxygenase-2, or ␤-actin. NW, non-wounded; LW, low wounding; MW, medium wounding; HW, high wounding. C, spatial expression of cyclooxygenase-2 after wounding. Monolayers of endothelial cells were grown to confluence on glass chamber slides that had been coated with fibronectin before multiple wound lines in a high wound pattern were created with a 200-l pipette tip as above, or not, and then fixed at the stated times. Cyclooxygenase-2 protein expression was detected by immunocytochemistry, and cell nuclei were counterstained with propidium iodide as before. The lower two elongated panels are composite images capturing all the cells between two parallel wound lines that are the black, acellular regions at the extreme right, and left portions of the image marked by a white dotted line. the most PGE 2 . We pretreated the monolayers with NS-398 to inhibit cyclooxygenase-2 activity, and we found that nearly all of the PGE 2 released from wounded monolayers came from this enzyme (Fig. 4A, lower panel). The positive control for cyclooxygenase-2 expression, phorbol myristoyl acetate, also stimulated PGE 2 release from intact monolayers that was also abolished by NS-398 pretreatment. PGE 2 stimulates cyclooxygenase-2 expression in some (24), but not all (12), contact-inhibited cells, and so we questioned whether PGE 2 acted on endothelial cells to induce the ratelimiting enzyme for its synthesis, cyclooxygenase-2. We found that 10 nM PGE 2 , the lowest concentration we tested, effectively stimulated cyclooxygenase-2 expression and that stimulated cyclooxygenase-2 accumulation was maximal at 33 nM (Fig. 4B). This observation, coupled with enhanced secretion of PGE 2 by wounded monolayers, suggested that the supernatants from wounded monolayers should increase cyclooxygenase-2 expression in confluent endothelial cell cultures. We extensively wounded endothelial cell cultures as before, collected the media overlaying these cells 4 h after wounding, and then incubated new, confluent cultures of endothelial cells containing little cyclooxygenase-2 with this media for 4 h. We found that media from wounded cultures, but not that from unmanipulated monolayers, stimulated cyclooxygenase-2 protein expression in target cells maintained as a confluent monolayer (Fig. 4C). The level of this cyclooxygenase-2 expression increased with the level of wounding of the donor culture, just as the amount of secreted PGE 2 increased with wounding density (Fig. 4A). We also found that medium from subconfluent endothelial cell cultures induced cyclooxygenase-2 expression in confluent monolayers to a greater extent than media from confluent cultures (Fig. 4C).
Inhibition of Cyclooxygenase-2 Suppresses Endothelial Cell Migration-Disruption of the integrity of an endothelial barrier causes adjacent cells to fill in the denuded areas by spreading, migration, and proliferation (16). We established that endothelial cells migrated into the areas denuded by the soft pipette tip, which did not score the plate and create a barrier to migration, over time (Fig. 5A). In contrast to the clean edge just after wounding, individual cells had broken away from the monolayer adjacent to the wound and entered the denuded area 24 h after the wounding. This created a disorganized edge of the wound with a few individual cells migrating individually into the denuded zone. Primarily, however, closure of the wound resulted from the entry of groups of adjacent cells that remained in contact with one another. We found extensive closure of the wound by 34 h, with little movement prior to 6 h (not shown) after wounding the monolayer (Fig. 5A). Endothelial cell migration into the denuded region depended on cyclooxygenase-2 activity because NS-398 suppressed cell entry into the acellular area and wound closure (Fig. 5B). The wound edge after NS-398 treatment remained uniform, although there was a modest decrease in the distance between the remaining endothelial cells.

Progression through the Cell Cycle Does Not Control Cyclooxygenase-2 Expression after Monolayer Wounding-Endo-
clooxygenase-2 expression. Endothelial cell monolayers were extensively wounded, washed, and given fresh media with 1% human serum. Alternatively, medium overlaying subconfluent cells was changed at this time to the reduced serum medium. After 4 h, medium from these cultures was transferred to fresh, undisturbed confluent endothelial cell monolayers that had just been serum-deprived by overnight incubation in media containing 1% human serum. These target monolayers were incubated for 4 h with media from wounded or subconfluent cells before the assay was stopped, and material collected for protein analysis and electrophoresis before cyclooxygenase-2, cyclooxygenase-1, and ␤-actin was determined by immunoblotting as before.

FIG. 4. Wounding stimulates cyclooxygenase-2 expression and stimulates PGE 2 release, and PGE 2 is an agonist for cyclooxygenase-2 expression in confluent endothelial cells.
A, endothelial cells were grown to confluence in 6-well plates, and the monolayer was wounded in patterns that produced three levels of injury, treated or not with 30 M NS-398, washed and incubated with fresh media containing 1% pooled human serum, and no exogenous arachidonate, for 8 h. PGE 2 released over this time was quantitated by ELISA as in Fig. 2. Ctl, control; Med, medium. B, PGE 2 stimulates cyclooxygenase-2 (COX-2) protein expression. Confluent monolayers of endothelial cells were incubated with PGE 2 at the stated concentrations for 4 h before cellular protein was collected for electrophoresis and immunoblotting with anticyclooxygenase-2 or ␤-actin antibodies. C, media from wounded monolayers, or subconfluent endothelial cells, contains an agonist for cy-thelial cell monolayers are growth-arrested, and either release from an enforced cell cycle arrest or the addition of the growth factors in serum stimulates cyclooxygenase-2 expression in serum-starved fibroblasts (25). We tested the role of progression through the cell cycle on cyclooxygenase-2 expression in endothelial cells after wounding by including agents that interfere with deoxynucleotide synthesis (Ara-C and 5-fluorouracil) in S phase, that introduce a block at the G 1 /S boundary (aphidicolin and mimosine), or block microtubule cycling during the M phase (nocodazole). We found that wounding increases cyclooxygenase-2 protein in the presence of any of these inhibitors, and so wounding was unlike serum stimulation and was not dependent on cell replication (Fig. 6).
Endothelial Cell Motility Is Enhanced by Cyclooxygenase-2 Activity and PGE 2 -We quantified endothelial cell motility using a modified Boyden chamber by imaging and counting calcein-labeled cells. We found that endothelial cells isolated from confluent cultures migrated through the filter in the absence of an added agonist, and that the addition of the chemoattractant ␤-fibroblast growth factor increased the number of migrating cells (Fig. 7A). Endothelial cells isolated from subconfluent cultures, however, were inherently more motile than their counterparts isolated from a confluent monolayer of cells. In fact, cells from a subconfluent culture were as motile as cells from a confluent culture after being stimulated with ␤-fibroblast growth factor.
We found that endothelial cell motility was suppressed by about half after treating the cells with NS-398 (Fig. 7B). The effect of NS-398 was statistically significant at the 30 M we used in other experiments, but we were unable to further reduce endothelial cell motility with higher concentrations of this cyclooxygenase-2 inhibitor. We conclude that cyclooxygenase products modulate motility but are not essential for cell migration. We inhibited cyclooxygenase-2 activity in cells isolated from subconfluent cultures, which again suppressed a portion of their motility, and then added PGE 2 back to the isolated cells. We found that the addition of PGE 2 partially restored cell motility in cells treated with NS-398 (Fig. 7C). The increase in migration is a chemokinetic enhancement of cell motility rather than direct chemoattraction by a gradient of PGE 2 because PGE 2 was added to both the upper and lower well and because a gradient would not be maintained over the 22 h of the experiment. DISCUSSION We find that individual endothelial cells not organized into a tightly confluent monolayer of cells express cyclooxygenase-2. Expression of this regulatory enzyme then diminishes as the cells form the intercellular contacts made possible by closely opposed cells in confluent monolayers. Cyclooxygenase-2 expression and PGE 2 production were rapidly reestablished after an abrupt loss of neighboring cells when we mechanically wounded cultures of quiescent endothelial cell monolayers. Reactivation of cyclooxygenase-2 expression by wounding leads to increased synthesis and release of PGE 2 to the surrounding media because the selective inhibitor NS-398 completely suppressed enhanced PGE 2 secretion. Juxtacrine signaling, i.e. signaling by a neighboring cell, and paracrine and autocrine signaling by PGE 2 are previously unknown mechanisms used to control expression of cyclooxygenase-2. This mode of regulation is relevant to angiogenesis before perfusion of the new vessel can occur, and is one that will have a major effect on the apparent background expression of cyclooxygenase-2 in cultured endothelial cells.
Denudation of vascular endothelium is a consequence of invasive clinical procedures from stent placement to grafting to venipuncture. This constitutes an inflammatory signal for the endothelium as shown by the synthesis of cyclooxygenase-2 and PGE 2 . Endothelial cells remaining after mechanical denudation responded to the extent of monolayer disruption with a graded accumulation of PGE 2 and cyclooxygenase-2. Immunocytochemistry showed cyclooxygenase-2 protein expression by FIG. 5. Closure of monolayer wounds is blocked by NS-398. A, endothelial cell monolayers were wounded with a pipette tip as above and allowed to recover for the stated times. B, endothelial cell monolayers were imaged just after wounding with a pipette tip or maintained in the continuous presence of 100 M NS-398 for 24 h before imaging.
FIG. 6. Endothelial cell monolayers need not transverse the cell cycle to induce cyclooxygenase-2 expression after monolayer wounding. Endothelial cells were grown to confluence in 20% human serum, washed, and serum-deprived for 17 h in media containing just 1% human serum. Agents that interfere with cell cycle progression during S phase (5-fluorouracil and Ara-C), during late G 1 /S phase (aphidicolin and mimosine), and M phase (nocodazole) were present during this period at the concentrations stated under "Materials and Methods." The monolayers were then wounded, or not, in the high wounding pattern and incubated in the continued presence of the stated agents for 8 h before cellular material was collected for analysis of cyclooxygenase-2 (COX-2) expression by Western blotting. SDS-polyacrylamide gels contained 35 g of cellular protein per well.
both cells adjacent and distal to the wound line, which suggests cells distal to the mechanical wound received information regarding the loss of cellular contacts from the cells immediately adjacent to the wound edge. Disrupting an endothelial cell monolayer with a fine scratch results in a wave of Ca 2ϩ excitation from the injured/surviving cells adjacent to the wound to distal cells that is critical for motility (26) and proliferation (27). In addition to this, we find a positive feed-forward loop involving the prostanoid PGE 2 after mechanical disruption of endothelial cell monolayers that aids recovery of monolayer integrity. PGE 2 stimulates endothelial cell cyclooxygenase-2 expression at low nanomolar levels, as recently found for cyclooxygenase 2 accumulation in pulmonary artery smooth muscle cells after bradykinin stimulation (28). PGE 2 accumulated to levels (ϳ6 nM) sufficient to stimulate cyclooxygenase-2 expression after wounding confluent monolayers even in the absence of an exogenous source of arachidonate. We cannot directly support the autocrine/paracrine role of PGE 2 in stim- FIG. 7. PGE 2 from cyclooxygenase-2 enhances endothelial cell migration. A, endothelial cells isolated from subconfluent cultures migrate more slowly than those from confluent cultures. Endothelial cells recovered from either confluent (Conf) or subconfluent (Non-conf) cultures were added to the upper chamber of a transwell over a filter with 3-m pores. After 22 h, the insert was removed, and the cells were stained with Calcein-AM. Cells that had migrated through the filter were imaged by confocal microscopy, and the number of cells in four random fields of low power (ϫ10) were enumerated (bottom). The positive control was 10 ng/ml ␤-fibroblast growth factor (␤FGF). B, NS-398 inhibits some, but not all, spontaneous endothelial cell migration. Endothelial cells were recovered from subconfluent cultures, added to the upper well of a transwell, and then NS-398 was added to the upper and lower migration chambers at the stated concentrations. Cells that had migrated through the filter after 22 h were imaged after staining with the fluorescent dye calcein AM as in A. C, PGE 2 partially overcomes NS-398 inhibition of migration. Endothelial cells were isolated and treated with 100 M NS-398 as above, but some wells additionally contained 5 nM PGE 2 in both the upper and lower wells. Migration was quantitated as in A.
The amount of PGE 2 made in vivo may be enhanced by exogenous sources of arachidonate such as high density lipoprotein (29), but blood flow washing over the endothelium will likely restrict PGE 2 induction of cyclooxygenase-2 expression to those cells that synthesized the PGE 2 . The PGE 2 feedforward signaling loop, however, may extend to paracrine signaling in some compartments that are closed. For example, cyclooxygenase-2 accumulates in the endothelium throughout the central nervous system during acute peripheral inflammation (30) or burn injury (31). This increases PGE 2 levels in cerebrospinal fluid, to 3 and 0.3 nM respectively, and associates with hyperalgesia in these models. PGE 2 may also accumulate during angiogenesis where the endothelial cells at the leading edge are not part of an organized monolayer, and any PGE 2 released before perfusion of the new microvessel is established will be confined to that area.
Endothelial cells express cyclooxygenase-2 in response to diverse soluble agonists ranging from serum (32) to endotoxin (8) to cytokines (7,9) to lipid agonists of peroxisome proliferator-activated receptors (12,33). Cyclooxygenase-2 expression has not been characterized as a response to environmental stimuli such as intercellular contacts, although prostaglandin production from arachidonate by endothelial cells is modulated by the number of population doublings (34), and cell density has been shown to affect the amount of PGE 2 made in response to IL-1 stimulation of an osteoblast-like cell line (35). Cyclooxygenase-2 is also induced in a fibroblastic cell line after readdition of serum to cells synchronized by complete serum starvation (25), but wounding endothelial cell monolayers does not induce cyclooxygenase-2 in this way. Growth factors were not added back to endothelial cell cultures after wounding, and inhibition of the cell cycle at various points did not block the increase in cyclooxygenase-2 induced by disrupting the integrity of the monolayer. Previous work (36,37) shows that incorporation of [ 3 H]thymidine into endothelial cells adjacent to a wound in a monolayer is a late event that occurs well after our experiments ended.
Cells adjacent to the mechanical wound migrate into the area by releasing at least some of their contacts with adjacent cells in the undisturbed portion of the monolayer to migrate as individual cells. Thus, we found a few individual cells in the denuded area 24 h after injuring the monolayer, but we also found a disorganized monolayer edge that extended into the wound space. We can ascribe a role for cyclooxygenase-2 in the migration of endothelial cells into the wound because NS-398 suppressed migration. However, we found it impossible to quantitate the rate of wound edge migration or cellular coverage of the denuded area over time because of the large heterogeneity in the organization of the monolayer edge. Instead, we recovered cells from both confluent and pre-confluent cultures, and we assayed their migration through 3-m filters to find cells obtained from the pre-confluent condition migrated faster than their counterparts isolated from confluent monolayers. We also found that NS-398 suppressed migration, and so the recovered cells mimicked their counterparts in wounded cultures of endothelial cells. We also found that adding PGE 2 back to the NS-398-inhibited cells at least partially overcame the reduction in cell migration caused by the loss of cyclooxygenase-2 activity.
Cyclooxygenase-2 activity of host stromal tissue, e.g. the vasculature and supporting structures, has a profound effect on tumorigenesis (15). Cyclooxygenase-2 products act on tumor cells (3) and underpin the angiogenic response of host vascula-ture induced by tumor cells (38,39). Thus, blocking the synthesis of cyclooxygenase-2-derived PGE 2 with inhibitors (18), genetically deleting cyclooxygenase-2 (3), or sequestering PGE 2 with an antibody suppresses tumor growth (18). Similarly, cyclooxygenase-2 has a critical role in the angiogenesis induced by inflammatory cytokines (22) and inflammatory insults (40). The production of PGE 2 also underlies the angiogenic effects of vascular endothelial cell growth factor and basic fibroblast growth factor (16,41). Accordingly, application of PGE 2 into the connective tissue of rat femoral vessels causes intense vascular sprouting (42). PGE 2 functions as an autocrine and paracrine signal to modulate endothelial cell monolayer formation, and the regulatory enzyme cyclooxygenase-2 is controlled by the integrity of the monolayer.