The Confluence-dependent Interaction of Cytosolic Phospholipase A2-α with Annexin A1 Regulates Endothelial Cell Prostaglandin E2 Generation*

The regulated generation of prostaglandins from endothelial cells is critical to vascular function. Here we identify a novel mechanism for the regulation of endothelial cell prostaglandin generation. Cytosolic phospholipase A2-α (cPLA2α) cleaves phospholipids in a Ca2+-dependent manner to yield free arachidonic acid and lysophospholipid. Arachidonic acid is then converted into prostaglandins by the action of cyclooxygenase enzymes and downstream synthases. By previously undefined mechanisms, nonconfluent endothelial cells generate greater levels of prostaglandins than confluent cells. Here we demonstrate that Ca2+-independent association of cPLA2α with the Golgi apparatus of confluent endothelial cells correlates with decreased prostaglandin synthesis. Golgi association blocks arachidonic acid release and prevents functional coupling between cPLA2α and COX-mediated prostaglandin synthesis. When inactivated at the Golgi apparatus of confluent endothelial cells, cPLA2α is associated with the phospholipid-binding protein annexin A1. Furthermore, the siRNA-mediated knockdown of endogenous annexin A1 significantly reverses the inhibitory effect of confluence on endothelial cell prostaglandin generation. Thus the confluence-dependent interaction of cPLA2α and annexin A1 at the Golgi acts as a novel molecular switch controlling cPLA2α activity and endothelial cell prostaglandin generation.

The regulated generation of prostaglandins from endothelial cells is critical to vascular function. Here we identify a novel mechanism for the regulation of endothelial cell prostaglandin generation. Cytosolic phospholipase A 2 -␣ (cPLA 2 ␣) cleaves phospholipids in a Ca 2؉ -dependent manner to yield free arachidonic acid and lysophospholipid. Arachidonic acid is then converted into prostaglandins by the action of cyclooxygenase enzymes and downstream synthases. By previously undefined mechanisms, nonconfluent endothelial cells generate greater levels of prostaglandins than confluent cells. Here we demonstrate that Ca 2؉ -independent association of cPLA 2 ␣ with the Golgi apparatus of confluent endothelial cells correlates with decreased prostaglandin synthesis. Golgi association blocks arachidonic acid release and prevents functional coupling between cPLA 2 ␣ and COX-mediated prostaglandin synthesis. When inactivated at the Golgi apparatus of confluent endothelial cells, cPLA 2 ␣ is associated with the phospholipid-binding protein annexin A1. Furthermore, the siRNA-mediated knockdown of endogenous annexin A1 significantly reverses the inhibitory effect of confluence on endothelial cell prostaglandin generation. Thus the confluence-dependent interaction of cPLA 2 ␣ and annexin A1 at the Golgi acts as a novel molecular switch controlling cPLA 2 ␣ activity and endothelial cell prostaglandin generation.
Ca 2ϩ elevation can induce relocation of cPLA 2 ␣ to the specific intracellular membranes in which the downstream AAmetabolizing cyclooxygenase (COX) enzymes are also located (2). There are two isoforms of COX that have been extensively characterized (COX-1 and -2) (9), and more recently an alternative COX-1 splice variant, COX-3, has also been cloned (10). The spatiotemporal co-localization of cPLA 2 ␣ with COX can couple these enzymes to facilitate efficient conversion of AA into prostaglandins (2,11). Chimeric cPLA 2 ␣ mutants specifically targeted to intracellular membranes in which COX does not reside do not couple with COX and drastically reduce prostaglandin production (11). Thus, the subcellular targeting of cPLA 2 ␣ to specific intracellular membranes is essential for the regulation of both AA and prostaglandin production. Despite this, the subcellular targeting of cPLA 2 ␣ in endothelial cells and its functional coupling with downstream COX enzymes has received little attention.
Release of prostaglandins by endothelial cells, the cells lining the luminal surface of all blood vessels, is essential to the control of vascular tone and thrombus formation (12,13). Therefore, regulation of endothelial prostaglandin generation is critical to the maintenance of normal vascular function. Nonconfluent endothelial cells generate much greater levels of AA and prostaglandin than confluent cells (14 -16), which has been attributed to elevated cPLA 2 ␣ activity. Despite this, the actual mechanism of this differential regulation of cPLA 2 ␣ activity has not been defined.
Inhibition of cPLA 2 ␣ activity by the phospholipid-binding protein, annexin A1, and the resulting block in AA metabolite release is a mechanism by which glucocorticoids exert their anti-inflammatory action (17)(18)(19). Annexin A1 is known to inhibit cPLA 2 ␣ activity upon interaction with the CalB domain of cPLA 2 ␣ in vitro (20,21); however, the relevance of this interaction to processes other than inflammation remains unclear.
Here we demonstrate that annexin A1 acts as a novel regulator of endothelial cell AA and prostaglandin generation upon interaction with cPLA 2 ␣ at the Golgi apparatus of confluent cells.
Biochemistry-Lysate preparation and Western analysis were performed as described previously (16). Briefly, samples (20 g of protein) were resolved for 60 min at 30 mA/gel on 10% SDS-polyacrylamide minigels using a discontinuous buffer system (24). For immunoblotting, protein was transferred onto nitrocellulose membranes for 3 h at 300 mA (25). Membranes were blocked in 5% (w/v) nonfat milk in phosphate-buffered saline for 30 min and then incubated overnight with primary antibody (1:500) at room temperature. After incubation with horseradish peroxidase-conjugated anti-goat IgG (1:3000) for 1 h, immunoreactive bands were visualized using a West Pico enhanced chemiluminescence detection kit (Pierce). Images were captured on a Fuji Film Intelligent dark box II image reader. Band intensities were determined densitometrically using Aida (Advanced Image Data Analyzer) 2.11 software. For immunoprecipitations, confluent HUVECs were lysed in lysis buffer (25 mM Tris-HCl, pH 7.4, 0.5% Nonidet P-40, 150 mM NaCl, 2 mM EGTA, 2 mM EDTA, 1:250 protease inhibitor mixture) for 30 min on ice. The annexin A1 immunoprecipitations were performed in 0.5% CHAPS lysis buffer (10 mM Tris-HCl, pH 7.4, 0.5% CHAPS, 140 mM NaCl, 0.5 mM CaCl 2 , 0.5 mM MgCl 2 ). cPLA 2 ␣ was immunoprecipitated from cell lysates with anti-cPLA 2 ␣ antibodies overnight at 4°C. Supernatants were then incubated with protein G-agarose for 3 h at 4°C. Bead complexes were washed with ice-cold lysis buffer and boiled with SDS-PAGE sample buffer prior to immunoblotting.
For subcellular fractionation, confluent endothelial cells were mechanically disrupted in homogenization buffer (0.25 M sucrose, 5 mM Tris-HCl, pH 7.4, 25 mM KCl, 5 mM MgCl 2 , 4.5 mM CaCl 2 ) in the presence or absence of 1 mM glutaraldehyde and fractionated on a discontinuous sucrose gradient according to Ref. 26. Interfaces were removed, diluted in homogenization media, and collected by centrifugation at 100,000 ϫ g for 60 min. Isolated fractions were analyzed by SDS-PAGE and immunoblotting. Iodixanol gradients were performed essentially as described by Yang et al. (27), except a 10 -30% gradient was utilized. Differential centrifugation enrichment of membrane fractions was performed as described previously by Lamour et al. (28).
Immunofluorescence-Immunofluorescence microscopy was performed as previously described (16). Briefly, cells were grown to the required level of confluence on 0.1% (w/v) gelatincoated coverslips. Cells were then fixed in 10% (v/v) formalin in neutral buffered saline (HT50-1-128; Sigma) for 5 min at 37°C. Prior to fixation, some cells were stimulated for 1 min with 5 M A23187 as previously (29,30). All ensuing steps were performed at 25°C. After permeabilization with 0.1% (v/v) Triton X-100 for 5 min, cells were refixed (5 min), washed with phosphatebuffered saline, and then incubated in 50 mM ammonium chloride for 10 min. Following phosphate-buffered saline washes, nonspecific binding sites were blocked with 5% (v/v) donkey serum for 3 h. Cells were incubated overnight with primary antibody followed by the appropriate secondary antibodies. Finally, coverslips were mounted on microscope slides in Fluoromount-G mounting medium (Southern Biotech).
Microscopy and Quantitation-Deconvolution fluorescence microscopy was performed using an Olympus IX-70 inverted fluorescence microscope (63 ϫ 1.5 oil immersion lens) and DeltaVision deconvolution system (Applied Precision Inc.). Individual optical sections of 0.2 m were generated from 15 iterative cycles of deconvolution. Quantification of co-localization was determined using the IMARIS software suite (Bitplane AG). Gray scale values below 10% of the maximum pixel intensity were eliminated as background. Co-localized pixels were expressed as percentages of the total pixels selected. Some images were captured using an inverted Zeiss LSM 510 META Axiovert 200M confocal microscope.
Determination of Cytosolic Ca 2ϩ Concentration-HUVECs cultured on glass bottom dishes were washed with HEPES/Tyrode's buffer and incubated with 2.5 M Fluo3-AM (Molecular Probes) for 30 min. Subsequently, cells were washed and then incubated with HEPES/Tyrode's buffer (plus 1 mM Ca 2ϩ ) for 20 min. Cells were placed in a heated chamber (37°C) above an Olympus IX-70 inverted fluorescence microscope. Fluctuations in cytosolic Ca 2ϩ were monitored by acquisition of fluorescence images.
AA Release-This technique was performed as previously (16). Briefly, HUVECs were labeled for 24 h with 1 Ci/ml [ 3 H]AA, washed with phosphate-buffered saline, and then incubated with 10 M BEL for 30 min to inhibit background iPLA 2 activity. Cells were then stimulated with 5 M A23187 in serum-free media (plus 0.3% (w/v) fatty acid-free bovine serum albumin). Aliquots of media and cell lysate were counted by liquid scintillation for radioactivity. cPLA 2 ␣ and Endothelial Cell Prostaglandin Generation NOVEMBER 23, 2007 • VOLUME 282 • NUMBER 47

JOURNAL OF BIOLOGICAL CHEMISTRY 34469
Prostaglandin E 2 Generation-HUVECs were cultured to the required cell density in 6-well culture dishes. Cells were washed and then in some cases incubated with 50 M AACOCF 3 and/or 10 M BEL for 30 min prior to treatment with 5 M A23187 in HEPES/Tyrode's buffer with 1 mM CaCl 2 for 15 min. Aliquots of media were assayed for prostaglandin E 2 (PGE 2 ) content using a high sensitivity ELISA (Assay Designs).

RESULTS
Cytosolic Ca 2ϩ Elevation Targets cPLA 2 ␣ to Intracellular Membranes in Subconfluent Endothelial Cells-cPLA 2 ␣ activity is greater in subconfluent endothelial cells than in quiescent, confluent endothelial cells (15,16). We have previously shown that subconfluent and confluent endothelial cells (see supplemental Fig. 1) express equal amounts of cPLA 2 ␣ (16), indicating that mechanisms other than control of cPLA 2 ␣ expression are responsible for confluence-dependent changes in its activity. In response to elevated cytosolic Ca 2ϩ , cPLA 2 ␣ is activated by relocation to intracellular membranes. Recruitment to specific membranes is required for the regulation of cPLA 2 ␣ activity (2, 11); however, the precise membranes to which cPLA 2 ␣ relocates in primary endothelial cells have not been defined. Therefore, we investigated the Ca 2ϩ -induced relocation of cPLA 2 ␣.
In HUVECs, cPLA 2 ␣ was detectable as a 110 kDa band by Western blotting using a well characterized affinity-purified antibody specific to the C-terminal region of cPLA 2 ␣ (16, 31). Additionally, the immunoreactivity was removed by preabsorption of the antibody with the antigenic peptide (supplemental Fig. 2A). By immunofluorescence microscopy, cPLA 2 ␣ was present as both diffuse and structured pools throughout the cytoplasm and nucleus of subconfluent HUVECs (supplemental Fig. 2B), similar to previous observations with endothelial cells and fibroblasts (16,32,33). To study the relocation of cPLA 2 ␣ in response to cytosolic Ca 2ϩ elevation, we used the Ca 2ϩ ionophore, A23187. This agent is ideal for studying confluence-dependent changes in cPLA 2 ␣ relocation, since A23187 raises cytosolic Ca 2ϩ to similar levels in both subconfluent and confluent endothelial cells (supplemental Fig.  2C). Signaling mediated by other agonists that elevate cytosolic Ca 2ϩ varies with endothelial cell density (34,35). In subconfluent HUVECs, upon elevation of cytosolic Ca 2ϩ , cPLA 2 ␣ relocated to the nuclear periphery and less disperse cytoplasmic structures (see supplemental Fig. 2). Relocation occurred rapidly (Ͻ1 min) and cPLA 2 ␣ immunoreactivity co-distributed extensively with calreticulin and ERGIC-53 ( Fig. 1A and supplemental Fig. 2, D-E), consistent with translocation to the endoplasmic reticulum (ER) and ER-Golgi intermediate compartments (ERGIC) (36 -38). Quantitation of this co-localization revealed a 2.5-fold increase in overlap between cPLA 2 ␣ and calreticulin and a 2.3-fold increase in overlap between cPLA 2 ␣ and ERGIC-53 upon A23187 treatment (Fig. 1A). A23187-induced relocation of cPLA 2 ␣ to the ER and ERGIC promoted its interaction with mem-brane substrate, resulting in a 12-fold increase in AA release from subconfluent cells (Fig. 1B). cPLA 2 ␣ Is Coupled to COX-1 and -2 in Subconfluent Endothelial Cells-AA may be converted into prostaglandin H 2 by the action of the COX enzymes. Targeting of cPLA 2 ␣ to the specific intracellular membranes in which COX enzymes are located can lead to the coupling of these enzymes to facilitate efficient conversion of AA into prostaglandins (2,11). Functional coupling does not require the direct interaction of AA synthesizing and metabolizing enzymes but relies on both enzymes being in close apposition. Consequently, AA released by cPLA 2 ␣ is statistically more likely to encounter the required downstream enzyme than if synthesized at a distant site. Despite the central importance of prostaglandins to numerous vascular processes, including angiogenesis (12,13), this functional coupling between AA release and prostaglandin synthesis has never been investigated in endothelial cells. We predicted that recruitment of cPLA 2 ␣ to the ER of subconfluent HUVECs would promote functional coupling, since both COX-1 and -2 are located at the ER of endothelial cells (39). Subconfluent cells were stimulated to elevate intracellular Ca 2ϩ , and the co-distribution of cPLA 2 ␣ with COX-1 and -2 was assessed ( Fig. 1C and supplemental Fig. 2, F and G). Quantitation revealed a 3.2-fold increase in overlap between cPLA 2 ␣ and COX-1 and a 2.3-fold increase between cPLA 2 ␣ and COX-2 upon cytosolic Ca 2ϩ elevation (Fig. 1C). We then assessed the ability of subconfluent HUVECs to generate PGE 2 , a major downstream product of both COX-1 and -2 activity (Fig. 1D). PGE 2 plays a key role in a variety of key vascular processes, such as angiogenesis and the regulation of vascular tone (40,41). Prior to A23187 treatment, PGE 2 generation was minimal (0.35 pg/1000 cells) but rose 24-fold upon cytosolic Ca 2ϩ elevation. Pretreatment of HUVECs with BEL, an inhibitor of iPLA 2 activity (42), had no effect on Ca 2ϩ -induced PGE 2 generation. This suggests that iPLA 2 -mediated AA release is not involved in PGE 2 production in endothelial cells. Inhibition of cPLA 2 ␣ with AACOCF 3 (43) inhibited PGE 2 generation by 88% (Fig. 2D). Thus, Ca 2ϩ -induced PGE 2 generation was FIGURE 2. Sequestration of cPLA 2 ␣ at the Golgi apparatus inhibits membrane recruitment. A, the localization of cPLA 2 ␣ in subconfluent (sub-confl.) and confluent (confl.) HUVECs was determined by immunofluorescence microscopy. B, HUVECs were transfected with either recombinant GFP alone or recombinant cPLA 2 ␣ tagged with GFP to the N terminus (GFP-cPLA 2 ␣) or C terminus (cPLA 2 ␣-GFP). After 48 h, transfected HUVECs were directly fixed, and both GFP and TGN46 were detected by immunofluorescence microscopy. C and D, confluent HUVECs were directly fixed or stimulated with 5 M A23187 for 1 min prior to fixation. cPLA 2 ␣ and either calreticulin (C) or ERGIC-53 (D) were detected by immunofluorescence microscopy. E, quantification of cPLA 2 ␣ co-distribution with calreticulin and ERGIC-53. Percentage co-localization was calculated using the IMARIS computer package as described under "Experimental Procedures" (n ϭ 13, ϮS.E.). *, p Ͻ 0.001 versus subconfluent cells. All results are representative of three separate experiments. All images represent 0.2-m sections through cell nuclei. *, EC nuclei. Scale bar, 25 m. cPLA 2 ␣ and Endothelial Cell Prostaglandin Generation NOVEMBER 23, 2007 • VOLUME 282 • NUMBER 47 almost entirely dependent on cPLA 2 ␣ activity. The maximal capacity of the cells to produce PGE 2 was assessed by incubating cells with excess exogenous AA. Exogenous AA reversed the inhibitory effect of AACOCF 3 but did not elevate PGE 2 generation any higher than that liberated by cytosolic Ca 2ϩ elevation. Thus, specific targeting of cPLA 2 ␣ to the ER/ERGIC and co-localization with the COX enzymes appears to result in maximal conversion of AA into PGE 2 . This demonstrates coupling between these enzymes in endothelial cells and indicates that release of AA by cPLA 2 ␣ is the rate-limiting step in PGE 2 production in endothelial cells.
Sequestration of cPLA 2 ␣ at the Golgi Apparatus of Quiescent Endothelial Cells Blocks Its Translocation to Other Membranes-By immunofluorescence microscopy, in confluent endothelial cells, cPLA 2 ␣ was seen to become associated with a reticular juxtanuclear region ( Fig. 2A) corresponding to the Golgi apparatus. Similar results were obtained with recombinant GFPtagged cPLA 2 ␣ (N-terminal and C-terminal linked constructs; Fig. 2B) and with a separate antibody targeted to the C terminus of cPLA 2 ␣ (supplemental Fig. 3). In other cell types, association of cPLA 2 ␣ with the Golgi apparatus promotes AA release (44,45), whereas in confluent endothelial cells, cPLA 2 ␣ activity is inhibited (14 -16). Furthermore, interaction with the Golgi blocked targeting of cPLA 2 ␣ to the ER and ERGIC upon Ca 2ϩ elevation (Fig. 2, C-E). In confluent HUVECs, the co-distribution of cPLA 2 ␣ with calreticulin (Fig. 2C) and ERGIC-53 (Fig.  2D) positive structures was not enhanced upon cytosolic Ca 2ϩ elevation. Quantitation revealed that overlap between cPLA 2 ␣ and calreticulin was only 4% that of subconfluent cells upon cytosolic Ca 2ϩ elevation (Fig. 2E). Overlap between cPLA 2 ␣ and ERGIC-53 was also reduced by 47% relative to A23187treated subconfluent cells (Fig. 2E). The reduction in overlap with ERGIC-53 was not to the extent seen with calreticulin but is probably due to background overlap with ERGIC-53-positive vesicles that have fused with the Golgi apparatus. Most importantly, when quantified, overlap of cPLA 2 ␣ with the ER and ERGIC was not enhanced upon cytosolic Ca 2ϩ elevation (Fig.  2E). Thus, association with the Golgi apparatus sequesters cPLA 2 ␣ away from its intracellular substrate, accounting for the reduced AA release seen at endothelial cell confluence (15,16). Sequestration of cPLA 2 ␣ at the Golgi Apparatus Inhibits Its Functional Coupling with the COX Enzymes-Confluent endothelial cells generate lower levels of prostaglandins than subconfluent cells (14,15). We predicted that the sequestration of cPLA 2 ␣ at confluence would be important to the control of prostaglandin generation. In confluent HUVECs, no significant overlap between cPLA 2 ␣ and COX-1 or -2 was observed prior to A23187 stimulation (Fig. 3, A and B). Furthermore, sequestration at the Golgi apparatus blocks the Ca 2ϩ -induced co-localization of cPLA 2 ␣ with COX-1 and -2 (Fig. 3, A and B). Quantitation revealed that overlap of cPLA 2 ␣ with COX-1 and -2 was only 6 and 22% that of subconfluent cells upon cytosolic Ca 2ϩ elevation (Fig. 3C). As a result, in confluent HUVECs, Ca 2ϩ -induced PGE 2 generation was inhibited by 95.6% relative to subconfluent cells (Fig. 3D). Furthermore, cytosolic Ca 2ϩ elevation did not induce any greater PGE 2 generation than unstimulated controls (Fig. 3D). The low levels of PGE 2 generated from confluent HUVECs were not due to iPLA 2 -or cPLA 2 ␣-mediated AA release, since preincubation with BEL and AACOCF 3 had no effect (Fig. 3D). These results were not due to variations in endogenous COX activity, since both sub-confluent and confluent HUVECs generate similar levels of PGE 2 when supplied with exogenous AA (Fig. 3E). Thus, sequestration of cPLA 2 ␣ at the Golgi apparatus of endothelial cells represents a novel mechanism for the regulation of prostaglandin generation by blocking its targeting to the specific intracellular membranes at which COX-1 and -2 reside.
The Interaction of cPLA 2 ␣ with the Golgi Apparatus is Ca 2ϩ -independent-Treatment of subconfluent HUVECs with A23187 does not enhance the co-distribution of cPLA 2 ␣ with the Golgi-resident protein mannosidase II (46) (Fig. 4A,  ManII). This is despite other reports documenting the Ca 2ϩdependent association of cPLA 2 ␣ with the Golgi apparatus of other cell types (29,47). In quiescent, confluent HUVECs, cPLA 2 ␣ co-distributed extensively with ManII both before and after A23187 treatment (Fig. 4A). Quantitation revealed that the overlap between cPLA 2 ␣ and ManII was not influenced by intracellular Ca 2ϩ elevation (Fig. 4B). Thus, in confluent HUVECs, cPLA 2 ␣ is immobilized at the Golgi apparatus and is insensitive to Ca 2ϩ elevation. The interaction of cPLA 2 ␣ with the Golgi is entirely Ca 2ϩ -independent, since treatment of confluent monolayers with the intracellular Ca 2ϩ chelator, 1,2bis(2-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid acetoxymethyl ester, did not affect the Golgi association of cPLA 2 ␣ (Fig. 4C). Furthermore, the Golgi association of cPLA 2 ␣ was not affected by incubation with the extracellular Ca 2ϩ chelator, EGTA (Fig. 4C), which is known to decrease intracellular Ca 2ϩ levels (7).
Since this interaction has no requirement for Ca 2ϩ , it cannot be mediated by the Ca 2ϩ -dependent lipid binding property of the CalB domain of cPLA 2 ␣. Thus, we hypothesized that association with the Golgi apparatus may occur via a novel mechanism.
cPLA 2 ␣ Interacts with Annexin A1 at the Golgi Apparatus of Confluent Endothelial Cells-There is previous evidence to suggest that cPLA 2 ␣ activity can be modulated by its interaction with a number of binding partners. The CalB domain of cPLA 2 ␣ binds the head domain of vimentin in far Western and co-immunoprecipitation experiments (48). In addition, overexpression of the head domain of vimentin inhibits AA and prostanoid production in rat fibroblasts (48). However, when confluent HUVECs were co-stained for cPLA 2 ␣ and vimentin, no evidence for a co-association was apparent at the Golgi apparatus (Fig. 5A). From yeast two-hybrid and immunoprecipitation experiments, the catalytic domain of cPLA 2 ␣ has also been found to bind p11, a member of the S100 family of calciumbinding proteins (49). In vitro, p11 inhibits cPLA 2 ␣, and in vivo knockdown of p11 elevates AA release in human bronchial epithelial cell lines. However, from immunofluorescence co-staining experiments, no evidence for an association of p11 with cPLA 2 ␣ was found at the Golgi apparatus of confluent HUVECs (Fig. 5A).
The phospholipid-binding protein, annexin A1, is known to interact with the CalB domain of cPLA 2 ␣ in vitro (20,21). In addition, annexin A1 has been implicated in the regulation of cellular cPLA 2 ␣ activity in a number of studies (17,19). We therefore compared the locations of cPLA 2 ␣ and annexin A1 in confluent endothelial cells (Fig. 5A). Results showed that a pool of annexin A1 was located at the Golgi apparatus of confluent endothelial cells and that this pool extensively co-distributed with cPLA 2 ␣ and the Golgi marker protein, TGN46 (Fig. 5, A  and B). In subconfluent endothelial cells, annexin A1 is not enriched at the Golgi apparatus to the same extent as confluent cells (Fig. 5B). Cellular levels of annexin A1 protein do not vary between confluent and subconfluent endothelial cells (supplemental Fig. 4); therefore, an enrichment of annexin A1 at the Golgi apparatus of confluent HUVECs must be a consequence of the cell density-dependent redistribution of annexin A1. Evidence for a physical association between annexin A1 and cPLA 2 ␣ was further obtained upon co-immunoprecipitation of annexin A1 with cPLA 2 ␣ from confluent HUVEC lysates (Fig. 5C). To confirm the specificity of the interaction, immunoprecipitations were also performed after preincubation of anti-cPLA 2 ␣ antibodies with antigenic peptide. Under these conditions, neither cPLA 2 ␣ nor annexin A1 were immunoprecipitated (Fig. 5C). To further assess the confluence dependence of the association between annexin A1 and cPLA 2 ␣, immunoprecipitations of annexin A1 were performed from confluent and subconfluent HUVEC lysates and immunoblotted for the presence of cPLA 2 ␣. Consistent with the immunofluorescence data, only annexin A1 from confluent endothelial cells was associated with cPLA 2 ␣ (Fig. 5D). . cPLA 2 ␣ interacts with annexin A1 at the Golgi apparatus of confluent endothelial cells. A, confluent HUVECs were directly fixed, and then cPLA 2 ␣ and either vimentin, p11, or annexin A1 were detected by immunofluorescence microscopy. B, confluent and subconfluent HUVECs were directly fixed, and then annexin A1 and TGN46 were detected by confocal immunofluorescence microscopy. C, cPLA 2 ␣ was immunoprecipitated (I.P.) from equal volumes of a confluent HUVEC lysate with either anti-cPLA 2 ␣ antibody or anti-cPLA 2 ␣ antibody preadsorbed with antigenic peptide. Immunoprecipitated proteins and resulting supernatant were subjected to SDS-PAGE and immunoblotted (W.B.) for cPLA 2 ␣ and annexin A1. D, equal quantities of either confluent or subconfluent HUVEC lysates were immunoprecipitated with either anti-annexin A1 antibody or control mouse IgG (mse), and bound proteins were subjected to SDS-PAGE and immunoblotted with the indicated antibodies. All results are representative of three separate experiments. All images represent 0.2-m sections through cell nuclei. *, EC nuclei. Scale bar, 25 m.
Interaction of cPLA 2 ␣ with Annexin A1 Inhibits Prostaglandin Synthesis in Confluent Endothelial Cells-To test the hypothesis that interaction with annexin A1 at the Golgi apparatus is responsible for inhibition of cPLA 2 ␣ activity in confluent endothelial cells, the siRNA knockdown of annexin A1 was performed. In agreement with the hypothesis, knockdown of endogenous annexin A1 resulted in a significant increase in AA-dependent PGE 2 generation relative to HUVECs trans-fected with nontargeting siRNA (Fig. 6, A and B, mock). However, the knockdown of endogenous annexin A1 was not sufficient to elevate PGE 2 generation from HUVECs stimulated in the presence of free AA (Fig. 6C). Thus, the increased ability of siRNA-treated HUVECs to generate PGE 2 was a consequence of an increased ability to release free AA and not due to an effect on downstream aspects of PGE 2 generation. Furthermore, this effect was not a consequence of increased cPLA 2 ␣, COX-1, or COX-2 expression, since knockdown of endogenous annexin A1 had no effect on their protein levels (Fig. 6A). To determine whether the interaction of cPLA 2 ␣ with annexin A1 is responsible for sequestering cPLA 2 ␣ at the Golgi apparatus of confluent HUVECs, the subcellular localization of cPLA 2 ␣ after annexin A1 knockdown was investigated (Fig. 6D). Upon knockdown, endogenous annexin A1 cPLA 2 ␣ becomes distributed throughout the cell in a manner similar to subconfluent cells. Furthermore, in annexin A1 siRNA-treated cells, cPLA 2 ␣ does not appear to co-distribute as extensively with the Golgi marker protein, TGN46, relative to control siRNA-treated HUVECs (Fig. 6D). Quantification of this overlap confirmed that significantly less cPLA 2 ␣ co-distributed with the Golgi apparatus of annexin A1 siRNA-treated cells versus control cells (Fig. 6E). Thus, the association of cPLA 2 ␣ with annexin A1 at the Golgi apparatus represents a novel mechanism for the control of endothelial cell AA release and prostaglandin production.

DISCUSSION
Here we have defined a novel mechanism for the regulation of prostaglandin biosynthesis in endothelial cells. Generation of prostaglandins by the endothelium regulates vascular homeostasis. For example, PGE 2 acts as a potent vasodilator/vasoconstrictor and proangiogenic stimuli (40,41). Here we find that targeting of cPLA 2 ␣ to the ER/ERGIC is required for its functional coupling to the COX enzymes and maximal PGE 2 generation. Furthermore, sequestration of cPLA 2 ␣ at the Golgi apparatus is sufficient to abolish Ca 2ϩ -induced prostaglandin generation in confluent endothelial cells (Fig. 7). This represents a novel molecular switch for the control of endothelial prostaglandin generation and accounts for earlier observations that confluent endothelial cells generate lower levels of prostaglandins compared with nonconfluent endothelial cells (14,15). In this study, we describe the Ca 2ϩ -induced association of cPLA 2 ␣ with the ER/ERGIC of nonconfluent HUVECs. Previously, we demonstrated that intracellular Ca 2ϩ elevation targets cPLA 2 ␣ to membranes distinct from ER/Golgi apparatus of EA.hy.926 cells (31). The EA.hy.926 cell line is a continuous hybrid line formed from the fusion of primary HUVECs and the human lung carcinoma cell line, A549. Consequently, any difference in the localization of cPLA 2 ␣ in EA.hy.926 cells versus their HUVEC donor line must represent characteristics associated with the A549 fusion partner. These differences emphasize the highly cell type-specific nature of the subcellular localization of cPLA 2 ␣ and highlight the potential deficiencies of hybrid cell lines as model cell systems. Sequestration at the Golgi apparatus and inactivation of cPLA 2 ␣ appears to be unique to endothelial cells and is not displayed by any other cell type tested (HeLa, Madin-Darby canine kidney, A549, saphenous vein smooth muscle, and EA.hy.926 cells; data not shown). Furthermore, this mechanism appears to be common to all endothelial cells, since we find that Golgi-localized cPLA 2 ␣ is seen in confluent cell cultures of HUVECs, human dermal microvascular endothelial cells, and HCAEC (not shown). Golgi-localized cPLA 2 ␣ is also seen in endothelial cells in isolated rat mesenteric arteries (data not shown).
It has been suggested previously that confluence-dependent fluctuations in endothelial cell PGE 2 generation are a consequence of differential COX activity (50). However, the previous study was conducted on serum-starved, unstimulated endothelial cells in the absence of intracellular Ca 2ϩ elevation. In their model system, AA must be derived from cPLA 2 ␣-independent sources, and the resultant PGE 2 production is minimal (Ͻ1.4 pg/1000 cells). In addition, we find that total COX activity remains constant independent of endothelial cell density, consistent with previous observations (15). Thus, regulation of prostaglandin generation from endothelial cells is critically dependent on the ability of cPLA 2 ␣ to be targeted to specific intracellular membranes.
Here we establish that inhibition of cPLA 2 ␣ at the Golgi apparatus of confluent endothelial cells occurs upon interaction with annexin A1. A number of cPLA 2 ␣-binding proteins that negatively regulate cPLA 2 ␣ activity have been previously identified and include the intermediate filament protein vimentin (48,51), the calcium-binding protein p11/S100A10 (49), and the phospholipid-binding protein annexin A1 (20,21). Using co-immunoprecipitation assays, we found no evidence for binding of cPLA 2 ␣ to vimentin or p11/S100A10 in confluent cells (data not shown); however, cPLA 2 ␣ was seen to specifically associate with annexin A1 only at the Golgi apparatus of confluent unstimulated endothelial cells. This interaction was responsible for inhibition of cPLA 2 ␣ activity, since siRNA-mediated knockdown of annexin A1 released cPLA 2 ␣ from the Golgi and significantly elevated AA-dependent PGE 2 production. Thus, interaction of annexin A1 with cPLA 2 ␣ at the Golgi apparatus represents a novel regulatory mechanism for the control of endothelial cell AA release and prostaglandin production. This is consistent with previous studies documenting the importance of annexin A1-mediated inhibition of cPLA 2 ␣ to the anti-inflammatory action of glucocorticoids (17)(18)(19).
In this study, we also describe the cell density-dependent relocation of annexin A1 to the Golgi apparatus of confluent endothelial cells. Consequently, the accumulation of annexin A1 at the Golgi apparatus inactivates cPLA 2 ␣ upon its interaction with annexin A1 and the sequestration of cPLA 2 ␣ away from its intracellular substrates. When endothelial cells reach confluence, they characteristically undergo the contact inhibition of cellular proliferation and exit the cell cycle to form quiescent monolayers. The molecular mechanisms regulating the contact inhibition are not clearly defined but are known to involve the inhibition of growth factor receptor signaling, activation of protein phosphatases, and inactivation of Rac upon the formation of cell-cell contacts (34,52,53). Similar mechanisms may promote the confluence-dependent redistribution of annexin A1 to the Golgi apparatus and subsequent inactivation of cPLA 2 ␣. However, the distinct mechanisms involved in the redistribution of annexin A1 remain to be elucidated.
In subconfluent endothelial cells, cPLA 2 ␣ is distributed throughout the cytoplasm and nucleus of resting cells and In subconfluent cells, cPLA 2 ␣ is free to associate with the ER and ERGIC upon cytosolic Ca 2ϩ elevation. The resulting spatiotemporal colocalization with the COX enzymes then facilitates the conversion of AA into prostaglandins. Upon cell confluence, cPLA 2 ␣ becomes associated with the Golgi apparatus and interacts with the cPLA 2 ␣-inhibitory protein, annexin A1. Consequently, the blocking of membrane targeting and inhibition of cPLA 2 ␣ activity abolishes cPLA 2 ␣-mediated signaling. remains largely inactive. Upon Ca 2ϩ elevation, cPLA 2 ␣ can then interact with intracellular membrane substrates via its Ca 2ϩ -dependent lipid-binding domain (C2 domain). In confluent cells, the Ca 2ϩ -dependent ability of cPLA 2 ␣ to interact with membranes is blocked upon its sequestration at the Golgi apparatus. Sequestration of signaling proteins away from their endogenous substrates is an emerging concept in the regulation of a variety of enzymes. The Golgi apparatus is known to possess an "exoskeleton" of structural proteins, which includes the molecular scaffold Sef. Interaction with Sef sequesters MEK/ERK at the Golgi apparatus and excludes it from nuclear targets (54). Similarly, annexin A1 is also known to form membrane scaffolds (55). However, isolation of endothelial cell Golgi membranes using differential centrifugation, sucrose density gradients, or iodixanol (OptiPrep) gradients does not produce annexin A1-or cPLA 2 ␣-enriched Golgi fractions when isolated from confluent cells lysed in the absence of high Ca 2ϩ concentrations (Ͼ1 mM) (supplemental Figs. 5 and 6, A and B). It is only upon the addition of Ca 2ϩ and the absence of chelators that cPLA 2 ␣ and annexin A1 become associated with membrane-containing fractions. Thus, in confluent endothelial cells, cPLA 2 ␣ may not directly interact with components of the Golgi membrane but may interact with an associated structure or scaffold that is disrupted upon mechanical lysis and subcellular fractionation. The Golgi-localized pool of annexin A1 may represent such a labile scaffold to which cPLA 2 ␣ is immobilized in confluent endothelial cells. An analogous situation has been reported for phospholipase D, which is blocked from accessing its phospholipid substrate by binding to the cytoskeletal, spectrin-related protein, fodrin (56). The next challenge will be to elucidate the molecular details of annexin A1-mediated inhibition of cPLA 2 ␣ in the endothelial cell.