Cell Confluence-dependent Remodeling of Endothelial Membranes Mediated by Cholesterol*

The plasma membranes of endothelial cells reaching confluence undergo profound structural and functional modifications, including the formation of adherens junctions, crucial for the regulation of vascular permeability and angiogenesis. Adherens junction formation is accompanied by the tyrosine dephosphorylation of adherens junctions proteins, which has been correlated with the strength and stability of adherens junctions. Here we show that cholesterol is a critical determinant of plasma membrane remodeling in cultures of growing cow pulmonary aortic endothelial cells. Membrane cholesterol increased dramatically at an early stage in the formation of confluent cow pulmonary aortic endothelial cell monolayers, prior to formation of intercellular junctions. This increase was accompanied by the redistribution of caveolin from a high density to a low density membrane compartment, previously shown to require cholesterol, and increased binding of the annexin II-p11 complex to membranes, consistent with other studies indicating cholesterol-dependent binding of annexin II to membranes. Furthermore, partial depletion of cholesterol from confluent cells with methyl-β-cyclodextrin both induced tyrosine phosphorylation of multiple membrane proteins, including adherens junctions proteins, and disrupted adherens junctions. Both effects were dramatically reduced by prior complexing of methyl-β-cyclodextrin with cholesterol. Our results reveal a novel physiological role for cholesterol regulating the formation of adherens junctions and other plasma membrane remodeling events as endothelial cells reach confluence.

Endothelial cells perform numerous functions essential to the physiology of vascular tissue. In addition to providing a smooth lining for blood vessels, the endothelium acts as a semipermeable barrier actively modulating the flow of solutes into and out of blood vessels (1). Endothelial cells also secrete factors regulating both the diameter of blood vessels and the clotting of blood cells (2), and the endothelium provides docking sites for both leukocytes (3,4) and cancer cells (5), facilitating their migration from the vasculature into the surrounding tissue. The breakdown and regrowth of endothelia are essential processes in angiogenesis, the growth of new blood vessels, which occurs in response to both wounding and ischemia, and, pathologically, in diabetic retinopathy (6,7) and the neovascularization of tumors (8).
Many functions of the endothelial barrier are replicated in confluent monolayers of endothelial cells. Upon reaching confluence, endothelial cell growth ceases, a process termed contact inhibition (9,10), survival becomes anchorage-dependent (11), and tight junctions and adherens junctions are elaborated between adjoining cells (12)(13)(14). Adherens junctions are composed of integral membrane proteins, termed cadherins, to which are bound the soluble proteins ␤-catenin, ␥-catenin, and pp120, members of the armadillo family (14). Adherens junction proteins have been implicated in critical functions of endothelial and epithelial cells, including the regulation of intercellular permeability (15), the sorting of proteins to specific plasma membrane domains (16), and contact inhibition (17)(18)(19)(20). In addition, E-cadherin expression in tumor cells correlates inversely with their metastatic potential (21)(22)(23).
The mechanisms that regulate the formation of confluent endothelial monolayers are poorly understood. However, in human umbilical vein endothelial cells approaching confluence, vascular endothelial cadherin, ␤-catenin, ␥-catenin, and pp120 all undergo dephosphorylation on tyrosine (24). Tyrosine phosphorylation and dephosphorylation of adherens junctions proteins are likely to represent physiological mechanisms for the regulation of the stability and assembly of adherens junctions, since epidermal growth factor, hepatocyte growth factor (25), tumor growth factor-␣ (26), and thrombin (27) all induce tyrosine phosphorylation of adherens junctions proteins, and exogenous expression of v-Src in fibroblastic and other cell lines disrupts cell-cell junctions (28).
In preliminary experiments in our laboratory, we noted a dramatic increase in membrane-bound annexin II (Ax II) 1 as cultures of cow pulmonary aortic endothelial (CPAE) cells reached confluence. In view of recent studies (29) implicating cholesterol in the mechanism of the binding of Ax II to membranes, we reasoned that cholesterol might also increase with endothelial cell confluence. In fact, membrane cholesterol has been reported to increase 4-fold in EA926 endothelial cells upon reaching confluence (30). Cholesterol has been implicated both in the mechanisms of multiple signaling pathways (31)(32)(33) and in the sorting of membrane proteins such as glycosylphosphatidylinositol-anchored proteins in polarized epithelial cells (34,35). These observations suggested to us that cholesterol might act in pathways that establish cell junctions and other properties of confluent endothelial monolayers. Here, we identify several cholesterol-dependent events that remodel the plasma membrane as endothelial cells reach confluence, most importantly the formation of adherens junctions and the dephosphorylation of adherens junctions proteins. These results reveal a novel physiological role for cholesterol in the regulation of endothelial barrier function.

EXPERIMENTAL PROCEDURES
Cell Lines and Reagents-CPAE cells were purchased from ATCC (Manassas, VA), and cultured in minimal essential medium (ATCC number  supplemented with 20% fetal calf serum, penicillin (50 units/ml), and streptomycin (50 mg/ml). Monoclonal antibodies to Ax II, p11, clathrin heavy chain, pp120, and ␥-catenin and a polyclonal antiserum to caveolin 1 were purchased from Transduction Laboratories (Lexington, KY). A monoclonal antibody to phosphotyrosine (clone 4G10), either in solution or conjugated to protein A-agarose, was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). Horseradish peroxidase-conjugated secondary antibodies were purchased from Promega (Madison, WI). Fluorescein isothiocyanate-conjugated antimouse IgG antibody was purchased from Jackson Laboratories (Bar Harbor, ME). Digitonin was purchased from Calbiochem. TMA-DPH was purchased from Molecular Probes, Inc. (Eugene, OR). All other chemicals were purchased from Sigma Chemical.
To prepare membrane fractions by sucrose density gradient centrifugation, the postnuclear supernatants were first centrifuged at 250,000 ϫ g for 15 min in a Beckman TL-100 centrifuge. The cytosolic fractions were removed, and the membrane pellets were resuspended in 0.25 ml of cytosol buffer and recentrifuged. The washed membranes were resuspended in 0.25 ml of cytosol buffer and brought to equal protein concentrations, and equal volumes of membranes were layered onto sucrose step gradients containing 0.61 ml each of 20,25,30,35,40,45,50, and 55% sucrose in cytosol buffer. The gradients were centrifuged for 18 h at 35,000 rpm in a Beckman SW 50.1 rotor. Either 11 or 12 fractions of equal volume were collected manually from the top of the gradient and mixed with 0.35 volumes of a 5ϫ stock of SDS sample buffer.
To compare cytosolic and membrane fractions, the postnuclear supernatants were made equal in protein concentration, and equal volumes were centrifuged at 250,000 ϫ g for 15 min. The cytosolic fractions were removed, and the membranes were brought to their initial volume with cytosol buffer. A 50% volume of SDS gel sample buffer (5ϫ stock) was added to each sample, which was then boiled and analyzed by electrophoresis. No particulates were detected in samples prepared from membranes in this way, and insoluble proteins were virtually recovered in full (e.g. cf. Fig. 1).
In-gel Tryptic Digestion and Mass Spectrometry-Silver-stained polypeptides were excised from a polyacrylamide gel, destained, and tryptically digested in-gel (38). The digest products were eluted with acetonitrile and further concentrated and desalted with ZipTip 18 microtips (Millipore Corp., Bedford, MA). Peptide masses were determined using a Kratos (Manchester, United Kingdom) Analytical Kompact SEQ Reflector matrix-assisted laser desorption ionizationtime of flight mass spectrometer equipped with a curved field reflectron. Peptide masses were searched against a nonredundant protein data base using MS-Fit of Protein Prospector, a program available from the World Wide Web site of the Mass Spectrometry Facility of the University of California at San Francisco. Fragmentation data from individual peptides obtained by postsource decay using the Kratos instrument were searched against nonredundant protein data bases using MS-Tag of Protein Prospector.
Immunoprecipitations-Homogenates from either untreated cells or cells treated with cyclodextrin (6%) or cyclodextrin (6%)-cholesterol (0.24%) complexes were prepared as described above, except that the cytosol buffer contained 1 mM Na 3 VO 4 . The homogenates were normalized for total protein, and equal volumes were mixed with 0.11 volumes of a 10ϫ detergent stock (10% Triton X-100, 1.0% sodium deoxycholate, 0.1% SDS, 1.0 M NaCl) in cytosol buffer, incubated at 4°C for 10 min, and centrifuged at 12,000 ϫ g for 10 min. The supernatants, containing 60 g of protein, were incubated with 5 g of monoclonal antibody (clone 4G10; Upstate Biotechnology, Inc., Lake Placid, NY)) conjugated to protein A-agarose overnight at 4°C with gentle agitation. The beads were washed seven times with a solution of 1% Triton X-100, 0.1% sodium deoxycholate, 0.01% SDS, and 0.1 M NaCl in cytosol buffer. Bound proteins were solubilized in SDS gel sample buffer and analyzed by electrophoresis and immunoblotting.
Cholesterol Measurements-Cells were scraped into a minimal volume of 50 mM Tris, pH 7.4, 150 mM NaCl and homogenized by passage through a 27-gauge, 1.25-inch needle 15 times. For each sample, a sufficient amount of cytosol buffer was added to bring the homogenate to 5.5 ml, the protein concentration was measured, and 5.25 ml was centrifuged for 1 h at 4°C in a Beckman SW 50.1 rotor at 45,000 rpm. Lipids were then extracted according to the following modification of Bligh and Dyer (36). The pellets were resuspended in 0.4 ml of TBS, and 1.5 ml of a 2:1 (v/v) mixture of methanol and chloroform was added. The samples were vortexed, 0.5 ml of chloroform was added, the samples were vortexed again, 0.5 ml of water was added, and the samples were vortexed again. The samples were centrifuged at 3000 ϫ g for 2 min, the chloroform (bottom) layer was removed, and 1.0 ml of chloroform was added to the aqueous phase to extract residual lipid. The samples were again vortexed and centrifuged as before, and the second chloroform layer was removed. The two chloroform extracts were pooled, dried under nitrogen, resuspended in 0.1 ml of isopropyl alcohol, and assayed for cholesterol using a kit (catalog no. 139 050) purchased from Roche Molecular Biochemicals. When samples of cholesterol standard solutions in TBS were extracted using this method, they were quantitatively (Ͼ95%) recovered.
Preparation of Cholesterol-Cyclodextrin Complexes-Complexes of cholesterol and methyl-␤Ϫcyclodextrin were either purchased directly from Sigma or prepared by a modification of the procedure of Klein et al. (31). Briefly, 4 ml of a 6% solution of methyl-␤Ϫcyclodextrin was brought to 80°C. 120 l of a 6% solution of cholesterol in isopropyl alcohol was then added dropwise with stirring, resulting in a cholesterol/ cyclodextrin ratio (w/w) of 0.03. After dispersal of the cholesterol, the solution was allowed to cool to 37°C with stirring and either used immediately or lyophilized and reconstituted in cell culture medium.

Incorporation of [ 3 H]Cholesterol by CPAE Cells-Cells were cultured overnight in medium supplemented with [ 3 H]cholesterol (NEN Life
Science Products) at 12.5 Ci/ml. The cells were then rinsed three times with unlabeled medium, returned to the incubator for 1 h more, either treated or not with cyclodextrin (6%) for the indicated time, and harvested as described above. Radiolabel content was quantitated by scintillation counting of postnuclear supernatants, prepared as described above, and normalized for protein.
Measurement of Relative Cell Surface Area-Relative cell surface area was measured by binding of the membrane-impermeable fluorescent dye TMA-DPH (37). CPAE cells were cultured to subconfluent or confluent densities, treated with trypsin (0.13% in PBS) for 10 min at 37°C, centrifuged at 1000 ϫ g for 5 min, resuspended in PBS, centrifuged again, and resuspended in PBS. Cell suspensions were made equal in cell density, and 0.02 volumes of a stock solution of TMA-DPH (0.25 mM in Me 2 SO) was added. The cells were mixed gently, incubated on ice for 1 min, centrifuged as above to remove unbound reagent, and resuspended to their initial volume in PBS. Fluorescence intensity of the bound reagent was measured using excitation and emission wavelengths of 375 and 432 nm, respectively, subtracting the fluorescence of cells suspended at the same density without reagent. The fluorescence intensity of the bound reagent increased linearly with cell density up to 2 ϫ 10 5 cells/ml.
Other Biochemical Methods-SDS-polyacrylamide gels were silver-stained using a kit (Bio-Rad catalog no. 161-0449). The antibodies to clathrin, Ax II, and caveolin were diluted 1:5000 for immunoblotting; the antibodies to pp120 and p11 were diluted 1:2000. Horseradish peroxidase-conjugated antibodies were detected on immunoblots using a chemiluminescence reagent from NEN Life Science Products. Immunoreactive bands were digitized using an Agfa (Mortsel, Belgium) Arcus II scanner, and band intensities were measured using Adobe Photoshop (Mountain View, CA). Relative amounts of protein were determined by fitting the band intensities to a standard curve. All polyacrylamide gels and immunoblots are representative of at least three separate experiments. Protein concentrations were determined using a Coomassie Blue-based reagent (Bio-Rad catalog no. 500-0006).

RESULTS
Cell Confluence-dependent Binding of Ax II to Endothelial Membranes-In preliminary studies in our laboratory, we found that membranes prepared from confluent CPAE cells contained dramatically higher levels of Ax II than did membranes from subconfluent cells (Fig. 1A). In contrast, cytosolic fractions from subconfluent and confluent cells contained approximately equal amounts of Ax II, while postnuclear supernatants from confluent cells contained more Ax II than did those prepared from subconfluent cells. Thus, we observed higher levels of Ax II in confluent versus subconfluent cells, and the additional Ax II in confluent cells appeared to be specifically recruited to the membrane, and not the cytosolic, fraction. Fractionation of membranes from subconfluent and confluent CPAE cells indicated that Ax II was entirely recruited to low density membranes (Fig. 1B), consistent with other studies localizing Ax II to low density endosomal and plasma membrane fractions (29).
In CPAE cells, Ax II exists, in part, in a heterotetrameric complex with p11 (39), a member of the S100 family of calciumbinding proteins (40). Like Ax II, p11 was dramatically enriched in membranes prepared from confluent, rather than subconfluent, CPAE cells (Fig. 1A). However, virtually no p11 FIG. 1. Cell confluence-dependent binding of Ax II to CPAE membranes. A, subcellular fractionation of Ax II and p11 from subconfluent or confluent CPAE cells. Postnuclear supernatants (PNS), cytosols (Cyt), and membrane (Memb) fractions were prepared from subconfluent (Ϫ) or confluent (ϩ) CPAE cells and 10% of each fraction was immunoblotted for Ax II, p11, or clathrin. B, membranes prepared from subconfluent (Ϫ) or confluent (ϩ) CPAE cells were fractionated on 20 -55% sucrose step gradients. 20% of each fraction was resolved on a 7.5-17.5% polyacrylamide gradient SDS gel and immunoblotted for Ax II. C, dependence of binding of Ax II and p11 to CPAE membranes on cholesterol. Washed membranes were prepared from confluent CPAE cells, resuspended in cytosol buffer to a concentration of 65 g/ml, and divided into four aliquots. EGTA, Triton X-100, or digitonin was added from a 10ϫ stock to the indicated final concentration, and the aliquots were centrifuged at 250,000 ϫ g for 15 min. The supernatants were removed, and the pellets were resuspended to volume in cytosol buffer. 20% of each sample was immunoblotted for Ax II, p11, and clathrin. D, silver-stained 7.5-17.5% polyacrylamide gradient SDS gels of the supernatants shown in C, electrophoresed under identical conditions, indicating that a polypeptide identified as Ax II (arrow; see "Results") was the most prominent polypeptide extracted with digitonin and that multiple additional polypeptides were extracted specifically with Triton X-100. was detected in cytosolic fractions from either subconfluent or confluent cells, and postnuclear supernatants prepared from confluent cells contained much higher levels of p11, indicating its increased level of expression.
Recently, Harder et al. (29) identified Ax II as one of a small group of proteins extracted from bovine hamster kidney membranes by 0.01% digitonin, which selectively binds to membrane cholesterol but not other membrane components (41). We found that all of the membrane-bound Ax II and p11 in CPAE cells was also solubilized after extraction with 0.01% digitonin (Fig. 1C), as well as with Triton X-100 (1%). Silver-stained gels of the detergent extracts (Fig. 1D) indicated that a polypeptide migrating at the position of Ax II (ϳ35 kDa) was by far the most prominent species extracted with 0.01% digitonin, and mass spectrometry of the products of an in-gel tryptic digest of this polypeptide confirmed it to be Ax II. In addition, numerous polypeptides were extracted by Triton X-100, but not digitonin (Fig. 1D), indicating that digitonin at 0.01% did not solubilize a broad range of membrane components, as expected. Neither digitonin nor Triton X-100 extracted clathrin, consistent with several studies reporting membrane-bound clathrin to be detergent-insoluble (42)(43)(44)(45). EGTA (1 mM) also solubilized most of the membrane-bound Ax II, consistent with the calcium-dependent binding exhibited by Ax II to both liposomes (46,47) and organelles (48).
Dependence of Membrane Cholesterol and Caveolin Distribution on Cell Confluence-The observation that membrane binding of the Ax II-p11 complex could be disrupted by digitonin indicated that this binding depended on cholesterol. Since we had also observed that binding of the Ax II-p11 complex to membranes increased with cell confluence, these observations suggested to us that membrane cholesterol might also increase with confluence in CPAE cell cultures. In fact, membrane cholesterol has been reported to increase 3-4-fold in EA926 endothelial cells upon reaching confluence (30), and we found that CPAE cells exhibited a 3.6-fold increase in membrane choles-terol upon reaching confluence (32 Ϯ 4 versus 9 Ϯ 1 g of cholesterol/mg of protein, n ϭ 3; confluent versus subconfluent, respectively). In contrast, the surface area per g of protein of confluent CPAE cells, measured by binding of TMA-DPH, was only 0.96 Ϯ 0.05-fold (n ϭ 2) that of subconfluent cells, indicating that other plasma membrane lipids did not increase proportionately with cholesterol.
To begin to evaluate the functional significance of this increase, we compared the subcellular distributions of caveolin-1, a cholesterol-associated protein, in subconfluent and confluent CPAE cells. Partial depletion of cholesterol with methyl-␤cyclodextrin has been found to relocalize caveolin-1 and -2 from low density to high density plasma membrane vesicles (33). This redistribution was found to correlate with an inhibition of epidermal growth factor-and bradykinin-stimulated phosphoinositide hydrolysis (33). Since membrane cholesterol increased as CPAE cells reached confluence, we reasoned that caveolin-1 might translocate from membrane fractions of high density to low density with increasing confluence.
Membranes prepared from subconfluent or confluent CPAE cells were fractionated by sucrose density gradient centrifugation, and the distribution of caveolin-1 was examined by immunoblotting (Fig. 2, A and B). Membranes from subconfluent cells contained a single population of caveolin-1-rich vesicles, migrating at a peak density of 44% sucrose. In contrast, membranes from confluent cells contained an additional caveolin-1 peak, migrating at 31% sucrose, containing slightly less than half of the total caveolin-1, which may represent caveolin-rich membranes containing increased levels of cholesterol (Ref. 33; see above). Thus, the increased cholesterol in confluent CPAE cells did, in fact, correlate with the redistribution of a protein in the manner predicted on the basis of its association with cholesterol. In contrast with caveolin-1, immunoblots indicated that neither the distribution nor the amount of clathrin, which was primarily present in high density membrane fractions, changed with cell confluence (Fig. 2A), and silver-stained gels indicated that the overall polypeptide composition of membrane fractions appeared to be largely independent of cell confluence ( Fig. 2A).
Tyrosine Phosphorylation of Adherens Junctions Proteins Regulated by Cholesterol-The effects of confluence on the subcellular distributions of caveolin-1 and Ax II indicated that the increased cholesterol seen in confluent cells was associated with cholesterol-dependent modifications of plasma membrane organization. To determine whether this increase was required for any of the functional properties of confluent endothelial monolayers, we examined the effects of cholesterol depletion on adherens junctions. Adherens junctions proteins, including vascular endothelial cadherin, ␥-catenin, ␤-catenin, and pp120, have been found to undergo dephosphorylation on tyrosine as HUVEC cells reach confluence in culture (24). Reduced tyrosine phosphorylation of adherens junctions proteins has been found to correlate with the strength of adherens junctions (28,49,50).
To evaluate a possible role for cholesterol in formation and stability of adherens junctions, we first examined the effect of methyl-␤-cyclodextrin, which partially (60 -70%) depletes cholesterol from CPAE cells (32), on tyrosine phosphorylation of membrane proteins. Incubation of confluent CPAE cells with methyl-␤-cyclodextrin (6%) for 25 min at 37°C reduced cellular cholesterol by 70 -75%, as measured by decreased labeling with [ 3 H]cholesterol. Under these conditions, methyl-␤-cyclodextrin induced dramatic increases in tyrosine phosphorylation of polypeptides of ϳ30, 55, 80, 120, 145, and 170 kDa, as well as a reduction in tyrosine phosphorylation of a ϳ60-kDa polypeptide (Fig. 3A). Many of these changes did not occur when cells were incubated with cyclodextrin-cholesterol complexes (31), indicating that they were specifically due to sequestration of cholesterol by methyl-␤-cyclodextrin.
We tested whether the 80-and 120-kDa proteins, which were tyrosine-phosphorylated as a result of cholesterol depletion, might be ␥-catenin and pp120, respectively. Immunoprecipitates of tyrosine-phosphorylated proteins were prepared from untreated cells and from cells treated with either cyclodextrin or cyclodextrin-cholesterol complexes (Fig. 3, B and C). Immunoblots indicated that the immunoprecipitates from the cyclo-dextrin-treated cells, but not the cells treated with cyclodextrin-cholesterol complexes, were dramatically enriched in both ␥-catenin and pp120, which co-electrophoresed with the 80and 120-kDa polypeptides, respectively. Thus, depletion of membrane cholesterol did, in fact, reverse the tyrosine dephosphorylation of two prominent adherens junctions proteins normally seen in cells having reached confluence.
Since tyrosine phosphorylation of adherens junctions proteins has been found to disrupt adherens junctions (28,42,43), we evaluated by immunofluorescence the effect of methyl-␤cyclodextrin on the distribution of pp120 (Fig. 4). In untreated cultures of confluent CPAE cells, pp120 was clearly concentrated along intercellular junctions. Generally, contacts between adjoining cells appeared continuous, extending along their entire lengths, and very few gaps between cells were observed. In contrast, in confluent cultures treated with methyl-␤-cyclodextrin, immunofluorescence staining of pp120 appeared uniformly diffuse throughout the cytoplasm. In addition, large gaps between adjacent cells were generally observed, indicating retraction of plasma membranes from

FIG. 3. Tyrosine phosphorylation of adherens junctions proteins induced by cholesterol depletion of CPAE cells.
A, confluent CPAE cells were incubated in either growth medium, medium plus methyl-␤-cyclodextrin (6%), or medium plus complexes of methyl-␤-cyclodextrin (6%) and cholesterol (0.18%) for 25 min and harvested. Cells were homogenized in cytosol buffer containing 1 mM Na 3 VO 4 , and membrane fractions were prepared by high speed centrifugation of postnuclear supernatants. Equal amounts of membrane proteins were analyzed by SDS gel electrophoresis and immunoblotted for phosphotyrosine. B and C, confluent CPAE cells were incubated in either growth medium, medium plus methyl-␤cyclodextrin (6%), or medium plus cyclodextrin-cholesterol complexes and harvested. Tyrosine-phosphorylated proteins were immunoprecipitated using a monoclonal antibody (clone 4G10), the immunoprecipitated proteins were divided in half, resolved on 7.5-17.5% polyacrylamide gradient SDS gels and immunoblotted either for phosphotyrosine or for ␥-catenin or pp120, using a mixture of the antibodies to both proteins. The 100-kDa band presumably represents a cross-reactive isoform of pp120, previously identified in endothelial cells (65).
sites of cell-cell contact. Immunofluorescence staining of pp120 in confluent cells treated with cyclodextrin-cholesterol complexes was concentrated at adherens junctions, as in untreated cells, indicating that the disruption of adherens junctions by methyl-␤-cyclodextrin was due specifically to cholesterol depletion.
Dependence of Membrane Cholesterol, Ax II-p11, and Tyrosine Phosphorylation on Cell Density-These observations indicated that the increased cholesterol seen in confluent endothelial monolayers was required for the formation of adherens junctions. This finding, in turn, suggested that this increase in cholesterol might precede formation of adherens junctions. To test this hypothesis, we examined the levels of cholesterol, tyrosine-phosphorylated proteins, Ax II, and p11 in membranes as a function of cell density in CPAE cell cultures (Fig.  5). CPAE cells were plated at increasing densities and harvested after 48 h, by which time only the cells plated at the highest density were confluent; cells plated at lower densities were sparsely distributed, exhibiting only occasional intercellular contacts of minimal extent (not shown). Membrane cholesterol in fully confluent cells was 3.8 Ϯ 0.3-fold (n ϭ 2) higher than in cells plated at the lowest density. Cells plated at intermediate densities exhibited cholesterol levels between these extremes. In particular, cells plated at the second highest density, which were less than 50% confluent, contained as much membrane cholesterol as did fully confluent cells. Membranebound Ax II and p11 exhibited similar dependences on cell confluence as did cholesterol (Fig. 5B), reaching maximal levels at cell densities well below those seen in confluent monolayers. Tyrosine phosphorylation of adherens junctions proteins decreased gradually with increased plating density, and substantial reductions were seen even in subconfluent cultures (Fig.  5C). Thus, increases in membrane cholesterol, Ax II, and p11, and tyrosine dephosphorylation of adherens junctions proteins all occurred well before the formation of extensive intracellular contacts.

Regulation of Cholesterol and Tyrosine Phosphorylation by
Cell Confluence-We have found that tyrosine dephosphorylation of the adherens junctions proteins ␥-catenin and pp120 required membrane cholesterol and that the reversal of these dephosphorylations by depletion of cholesterol was accompanied by the disruption of adherens junctions. As indicated earlier, tyrosine dephosphorylation of adherens junctions was previously found to occur with increasing confluence in endothelial cell cultures (24), and the rephosphorylation of these proteins has been found to destabilize adherens junctions (28,49,50). Our observation that these dephosphorylations occurred well before the formation of cell-cell contacts and adherens junctions (Fig. 5) strongly suggests that they do not occur as a consequence of the binding or clustering of cadherin molecules on adjacent cells and could instead be required for these events. The observation that membrane cholesterol also increased more than 3-fold prior to the formation of confluent monolayers further suggests that cholesterol lies in a regulatory pathway leading to multiple remodeling events at the plasma membrane, including dephosphorylation of adherens junctions proteins, required for plasma membrane differentiation. Such a pathway could be initiated by autocrine factors, secreted at effective concentrations even at relatively low cell densities. It is noteworthy that platelet-derived growth factor, basic fibroblast growth factor, and other growth factors have been found to increase levels of intracellular free cholesterol by the activation of enzymes associated with cholesterol synthesis, uptake, and release from lipoproteins (reviewed in Ref. 51).
Mechanism of Cholesterol Regulation of Tyrosine Phosphorylation-It is not clear how cholesterol levels modulate the tyrosine phosphorylation states of adherens junctions and other membrane proteins. Many signaling proteins can be isolated in low density, cholesterol-rich particulate fragments isolated after extraction of plasma membranes at 4°C with Triton X-100. These fragments are thought to be derived from discrete FIG. 5. Dependence of membrane cholesterol, Ax II/p11, and tyrosinephosphorylated proteins on CPAE cell density. CPAE cells were plated at the indicated densities, grown for 2 days, and harvested. Only cells plated at the highest densities (Ͼ2.5 ϫ 10 4 cells/cm 2 ) had reached confluence. A, cells were homogenized and sedimented to measure the amount of cholesterol per mg of total protein, as described under "Experimental Procedures." Values and errors represent the mean and range of two experiments (B and C). Membrane fractions were prepared in the absence (B) or presence (C) of 1 mM Na 3 VO 4 . 20% of each sample was resolved on a 7.5-17.5% polyacrylamide SDS gel and immunoblotted for Ax II and p11 (B) or phosphotyrosine (C). microdomains within the plasma membrane, since at least some of their constituents are distributed nonrandomly within the planar bilayer (reviewed in Ref. 52). Such microdomains are thought to exhibit a higher degree of order than surrounding phospholipid-rich regions (53). It has been speculated that the clustering of signaling proteins within microdomains is required for the effective transduction of at least some extracellular signals. Thus, depletion of cholesterol with methyl-␤cyclodextrin was found to inhibit epidermal growth factor-and bradykinin-stimulated phosphoinositide hydrolysis (33) and stress-activated MAP kinase, but not Jun kinase, activity in CPAE cells (32). As noted earlier, methyl-␤-cyclodextrin was also found to redistribute caveolin, the epidermal growth factor receptor and G q from low density to high density membrane fractions (33), and we have found that methyl-␤-cyclodextrin fully depleted membrane fractions from sucrose density gradients of Ax II and p11, as well as a small amount of clathrin, 2 consistent with recent studies implicating cholesterol in the internalization of clathrin-coated pits (54). Thus, it is possible that cholesterol levels indirectly regulate tyrosine kinases and phosphatases by localizing them to or excluding them from cholesterol-rich microdomains, perhaps restricting their access to substrates. Alternatively, cholesterol might directly act through proteins with sterol-sensing domains to regulate the levels or activities of kinases, phosphatases, or other signaltransducing proteins. Proteins containing sterol-sensing domains have been found to mediate feedback inhibition of cholesterol biosynthesis (55) and have been proposed to act in the pathway by which the sonic hedgehog protein regulates pattern development in Drosophila (56).
Co-regulation of Ax II-p11 and Membrane Cholesterol-We (Fig. 1, C and D) and others (29) have found that sequestration of cholesterol by low levels of digitonin solubilized Ax II and only a very few other proteins. Our results also indicated that both binding of the Ax II-p11 complex to membranes and expression of the p11 polypeptide exhibited similar dependences on cell density as did membrane cholesterol (Fig. 5). The functional relationship between Ax II-p11 and membrane cholesterol is unclear. While Ax II was found in low density fractions prepared from partially purified plasma membranes ( Fig. 1B; Refs 29 and 57), Ax II has been found not to co-purify with caveolae isolated by immunoadsorption using antibodies to caveolin (57,58), suggesting that Ax II resides in a separate cholesterol-rich plasma membrane compartment. Lateral clustering of annexins on chromaffin granules has been detected by fluorescence resonance energy transfer (59), suggesting that Ax II-p11 forms multimeric complexes, which could support the binding of other protein assemblies. While Ax II-p11 has been implicated in multiple functions at both the intracellular and extracellular faces of the plasma membrane (reviewed in Ref. 60), including regulated exocytosis (61) and the activation of plasmin at the cell surface (62), the mechanisms by which Ax II-p11 interacts with membranes are unclear. While specific protein receptors for Ax II have not been identified, Ax II and other annexins have been reported to form ion channels in liposomes (63), and a structural model has been proposed for the direct insertion of Ax XIII into phospholipid bilayers (64). In addition, Ax II has been reported to bind directly to anionic phospholipids at submicromolar concentrations of calcium (47,48). All of these observations suggest that Ax II-p11 might interact directly with membrane lipids to form scaffolds for macromolecular complexes.
In summary, we have found that endothelial cells undergo a dramatic increase in membrane cholesterol at an early stage in the formation of confluent monolayers, required for specific plasma membrane events including the stabilization of adherens junctions and the tyrosine dephosphorylation of adherens junctions proteins. Further work will be needed to ascertain the full range of functions regulated by these events.