Annexin A6-induced Inhibition of Cytoplasmic Phospholipase A2 Is Linked to Caveolin-1 Export from the Golgi*

The molecular mechanisms regulating the exit of caveolin from the Golgi complex are not fully understood. Cholesterol and sphingolipid availability affects Golgi vesiculation events and involves the activity of cytoplasmic phospholipase A2 (cPLA2). We recently demonstrated that high expression levels of annexin A6 (AnxA6) perturb the intracellular distribution of cellular cholesterol, thereby inhibiting caveolin export from the Golgi complex. In the present study we show that in Chinese hamster ovary cells overexpressing AnxA6, sequestration of cholesterol in late endosomes, leading to reduced amounts of cholesterol in the Golgi, inhibits cPLA2 activity and its association with the Golgi complex. This correlates with the blockage of caveolin export from the Golgi in cells treated with methyl arachidonyl fluorophosphonate, a Ca2+-dependent cPLA2 inhibitor. AnxA6-mediated down-regulation of cPLA2 activity was overcome upon the addition of exogenous cholesterol or transfection with small interfering RNA targeting AnxA6. These findings indicate that AnxA6 interferes with caveolin transport through the inhibition of cPLA2.

The molecular mechanisms regulating the exit of caveolin from the Golgi complex are not fully understood. Cholesterol and sphingolipid availability affects Golgi vesiculation events and involves the activity of cytoplasmic phospholipase A 2 (cPLA 2 ). We recently demonstrated that high expression levels of annexin A6 (AnxA6) perturb the intracellular distribution of cellular cholesterol, thereby inhibiting caveolin export from the Golgi complex. In the present study we show that in Chinese hamster ovary cells overexpressing AnxA6, sequestration of cholesterol in late endosomes, leading to reduced amounts of cholesterol in the Golgi, inhibits cPLA 2 activity and its association with the Golgi complex. This correlates with the blockage of caveolin export from the Golgi in cells treated with methyl arachidonyl fluorophosphonate, a Ca 2؉ -dependent cPLA 2 inhibitor. AnxA6-mediated down-regulation of cPLA 2 activity was overcome upon the addition of exogenous cholesterol or transfection with small interfering RNA targeting AnxA6. These findings indicate that AnxA6 interferes with caveolin transport through the inhibition of cPLA 2 .
Annexins are a family of Ca 2ϩ and membrane-binding proteins involved in membrane trafficking and various other processes including signaling, proliferation, differentiation, and inflammation (1)(2)(3). Each annexin consists of a unique N-terminal tail and a common, well conserved C-terminal core domain containing 4 (or 8 for AnxA6 3 ) repeats of a highly homologous 70-amino acid sequence which facilitates their Ca 2ϩ and phospholipid binding.
We and others showed that AnxA6 is located at the plasma membrane, in the endocytic compartment, and in caveolae. AnxA6 has been implicated in endo-and exocytic membrane trafficking pathways and regulates low density lipoprotein (LDL) receptor-mediated endocytosis (4,5), is crucial for LDL degradation and its transport from late endosomes/pre-lysosomes to lysosomes (6), and is recruited to cholesterol-laden late endosomes (7). In addition, AnxA6 stimulates the membrane recruitment of the GTPase-activating protein p120GAP and protein kinase C to modulate the Ras signaling pathway (8,9). These multifunctional features of AnxA6 are most probably a consequence of (a) its dynamic spatiotemporal behavior in a Ca 2ϩ and/or cholesterol-dependent manner but also (b) the promiscuity to interact with a large set of other molecules (10), in particular through the flexible linker region (amino acids 320 -378) which connects the N-and C-terminal AnxA6 repeats (4 repeats each).
Expression of AnxA6 and other annexins varies in cells and tissues as well as in different pathophysiological situations (1,10). We recently compared wild type Chinese hamster ovary cells (CHOwt, contain very low AnxA6 levels) and AnxA6-deficient human epithelial carcinoma A431 cells with stable cell lines expressing high amounts of AnxA6 (CHOanx6, A431anx6). In these studies we showed that high AnxA6 levels result in an accumulation of cholesterol in the late endocytic compartment, leading to reduced amounts of cholesterol in the Golgi and the plasma membrane. This overall imbalance of cellular cholesterol is accompanied by an inhibition of caveolin export from the Golgi complex. Consequently, a significant diminution in the number of caveolae at the cell surface and a strong reduction of cholesterol efflux was observed (11). However, the molecular mechanism(s) regulating the exit of caveolin from the Golgi complex is not fully understood, and recently the participation of syntaxin 6 (12) or PTRF-cavin (13,14) in this process has also been demonstrated.
Changes in the availability of cellular cholesterol can be directly related to the vesiculation of the Golgi apparatus. Vesiculation and tubulation events regulating membrane traffic and cargo export from the Golgi require the activity of cholesterol-dependent cPLA 2 (15). Therefore, we addressed the possibility of AnxA6 interfering with caveolin export from Golgi membranes through the inhibition of cPLA 2 . In the pres-ent study we show that overexpression of AnxA6 indirectly inhibits cPLA 2 activity and its association with the Golgi complex through the reduction of cholesterol availability. In support of cPLA 2 being involved in caveolin transport, CHOwt treated with the Ca 2ϩ -dependent cPLA 2 inhibitor, methyl arachidonyl fluorophosphonate (MAFP), but not the Ca 2ϩ -independent cPLA 2 inhibitor haloenol lactone suicide substrate (HELSS), mimics the phenotype of AnxA6 overexpression and leads to an accumulation of caveolin in the Golgi complex.
Cell Culture-CHO cells were grown in Ham's F-12, and HeLa and COS-1 cells were grown together in DMEM with 10% fetal calf serum, L-glutamine (2 mM), penicillin (100 units/ml), and streptomycin (100 g/ml) at 37°C, 5% CO 2 . The generation of the stable AnxA6-overexpressing CHO cell line (CHOanx6) was described in detail (4,6). For the intracellular accumulation of cholesterol, cells were treated for 24 h with U18666A (2 g/ml) as described (7). For the inhibition of Ca 2ϩindependent and -dependent cPLA 2 activity, cells were incubated as described with either 5 M HELSS for 30 min or 15 M MAFP for 60 min, respectively (15). For transient transfections, cells (40 -70% confluence) were transfected with 1 g of DNA/ml using Effectene (Qiagen) following the manufacturer's instructions.
Cycloheximide/Cholesterol Treatments-To inhibit protein synthesis in some experiments, 10 g/ml cycloheximide (in 100 mM Hepes, pH 7.5) was added to the media for 90 min. For cholesterol addition, cells were incubated with 30 g/ml cholesterol (premixed for 30 min in DMEM by gentle agitation) for 90 min.
RNAi-mediated Inhibition of AnxA6-To specifically knock down gene expression of human AnxA6, 1-2 ϫ 10 6 HeLa cells were transfected in 2 ml of medium with 10 nM AnxA6 small interfering RNA (Santa Cruz, sc-29688) and 6 l of Lipofectamine 2000 reagent (Invitrogen) according to the instructions of the manufacturer. Experiments were conducted 72 h after transfection when the depletion of AnxA6 was more significant (see Western blot in Fig. 1C). GFP small interfering RNA was used as a negative control.
Release of [ 3 H]AA-Cells were incubated overnight in 12 wells with 0.25 Ci/ml [ 3 H]AA in DMEM, 0.5% fetal calf serum. Cells were washed and incubated with DMEM, 0.2% BSA for 10 min. Then the media and cells were harvested and measured by scintillation counting (15).
Pulldown Assays with Purified GST-AnxA6-Cells were incubated with and without 30 g/ml cholesterol, placed in cold lysis buffer (50 mM Tris, 100 mM NaCl, 1% Triton X-100, 0.1 mM CaCl 2 plus protease inhibitors), scraped, and centrifuged. Then 75 g of GST-AnxA6 and 21 g of GST were incubated for 90 min with glutathione-Sepharose at 4°C in phosphate-buffered saline. After washing with lysis buffer, Sepharose-bound GST-AnxA6 (and GST as control) was then incubated with 600 g of cell extract for 2 h at 4°C, and proteins bound to the column were collected by centrifugation and analyzed by Western blotting.
Immunoprecipitation-CHO and CHOanx6 cells grown on 100-mm dishes were transfected with pEGFP-cPLA 2 . After washing in phosphate-buffered saline and solubilizing by scraping with a rubber policeman in TGH buffer (1% Triton X-100, 10% glycerol, 50 mM NaCl, 50 mM HEPES, pH 7.3, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 1 mM phenylmethylsulfonyl fluoride, 10 mg/ml leupeptin, 10 mg/ml aprotinin) followed by gentle rotation for 10 min at 4°C, lysates were then centrifuged at 14,000 ϫ g for 10 min at 4°C. Supernatants of transfected or not transfected cells were incubated with rabbit polyclonal anti-GFP or polyclonal anti-annexin A6 antibodies for 2 h at 4°C and then for 60 min after the addition of protein A-Sepharose. Immunoprecipitates were washed twice in TGH supplemented with 150 mM NaCl and then once without NaCl. 10% SDS-polyacrylamide gels were used to separate proteins. Proteins were then transferred to Immobilon-P and immunoblotted using anti-annexin A6 or anti-GFP followed by the appropriate peroxidase-conjugated secondary antibody and ECL detection.
Immunofluorescence Microscopy-Cells were grown on coverslips and fixed with 4% paraformaldehyde, washed, permeabilized with 0.1% saponin, and incubated with primary and secondary antibodies as described elsewhere (19). For anti-cPLA 2 labeling, cells were fixed with cold methanol for 2 min. For double labeling with anti-cPLA 2 and anti-␣-mannosidase II, cells were first fixed with cold methanol followed by 4% paraformaldehyde for 12 min. Finally, samples were mounted in the anti-fading media Mowiol (Calbiochem), and cells were observed using an oil immersion Plan-Apo63x/1.4 objective in an Axio-plan Zeiss microscope. Images were captured with an AxioCam HRc camera and were digitally treated with Axio-Vision 3.1 software. In some experiments a Leica TCS SL laserscanning confocal spectral microscope (Leica Microsystems) with argon and HeNe lasers attached to a Leica DMIRE2 inverted microscope was used. Image analysis was performed with Adobe Photoshop 7.0 software. To quantify the fluorescence intensity staining, images were captured using identical contrast and exposure times. Using NIH image software (Image J), the area to be quantified (the plasma membrane) was selected, the pixel intensity was determined, and the average staining intensity (20 cells from 3 experiments) was calculated. Colocalization of cPLA 2 and ␣-mannosidase II was quantified using the Image J program (National Institutes of Health) and the "Highlighting Colocalization" plugin (Pierre Bourdoncle, Institut Jacques Monod, Service Imagerie, Paris), which highlights the colocalized points of two images. Two points were considered to colocalize if their intensity was higher than 50%, and an image of colocalized points was generated. Then in each cell the Golgi was defined as the region of interest, and the co-localization (%) therein was calculated. In each experiment at least 50 cells per cell line were analyzed. In some experiments the colocalization of mannosidase II with transiently transfected GFP-cPLA 2 was analyzed. The transfection procedures have been described elsewhere (11).
Photobleaching Experiments and Time-lapse Confocal Microscopy-Fluorescence recovery after photobleaching (iFRAP) experiments were carried out in COS-1 cells using a Leica TCS SL laser-scanning confocal spectral microscope (Leica Microsystems) with argon and HeNe lasers attached to a Leica DMIRE2 inverted microscope equipped with an incubation system with temperature and CO 2 control (16). For visualization of GFP, images were acquired using 63ϫ oil immersion objective lens (NA 1.32; 488 nm laser line; excitation beam splitter RSP 500, emission range detection 500 -600 nm). The confocal pinhole was set at 2-3 Airy units to minimize changes in fluorescence due to GFP-tagged proteins moving away from the plane of focus. The whole cytoplasm, except the Golgi of a GFP fusion protein transfected cell, was photobleached using 50 -80 scans with the 488-nm laser line at full power. Pre-and post-bleach images were monitored at 30-s intervals for 40 min. The excitation intensity was attenuated down to ϳ5% of the half laser power to avoid significant photobleaching. The relative loss of fluorescence intensity in the unbleached region of interest and overall photobleaching in the whole cell during the time series were quantified using Image Processing Leica Confocal Software. Background fluorescence was measured in a random field outside of the cells. Fluorescence correction and normalization of GFP-tagged proteins were calculated according to Rabut and Ellenberg (20).
Isolation of Golgi Membranes-A modified method of Balch (21) described in detail by Brügger et al. (22) was used to isolate Golgi membrane fractions. All procedures were carried out at 4°C. In brief, 4 -5 ϫ 10 9 cells were harvested, washed twice with phosphate-buffered saline, twice with BB (breaking buffer; 0.25 M sucrose in 10 mM Tris-HCl, pH 7.4), and finally resus-pended in four volumes of BB. Cells were homogenized by passing 25 times through a ball-bearing homogenizer (Balch homogenizer), which disrupts early and late endosomes, but not Golgi vesicles, and brought to a sucrose concentration of 37% (w/w) by the addition of 62% (w/w) sucrose. 14 ml of sample was overlaid with 15 ml of 35% (w/w) sucrose and 9 ml of 29% (w/w) sucrose (in 10 mM Tris-HCl, pH 7.4) and centrifuged for 2.5 h at 25,000 rpm. Typically, 2 ml of a Golgi-enriched membrane fraction was recovered at the 35-29% interphase. In some experiments, before homogenization cells were preincubated with and without cholesterol for 90 min (30 g/ml). The purity of the isolated Golgi membranes was confirmed by Western blot analysis with markers for Golgi (TGN38, GM130, GMAP210), early and late endosomes (Rab5, Rab7), plasma membrane (Na ϩ K ϩ ATPase), and endoplasmic reticulum (KDEL; data not shown).
Western Blot Analysis-CHOwt, CHOanx6, and HeLa cell lysates, samples from isolated Golgi membranes, and GST-AnxA6 pulldown assays were separated by 10% SDS-PAGE and transferred to Immobilon-P (Millipore) followed by incubation with primary antibodies and the appropriate peroxidase-conjugated secondary antibodies and ECL detection (Amersham Biosciences). Protein content was measured by the method of Lowry et al. (23).

RESULTS
AnxA6 Expression Levels Affect Cholesterol-dependent cPLA 2 Activity-Regulation of membrane traffic and cargo export from the Golgi complex involves cPLA 2 activity (24,25). To address whether the AnxA6-dependent reduction of cholesterol export from late endosomes (11) could be linked to cholesterol-and cPLA 2 -dependent secretory pathways from the Golgi to the plasma membrane, we first compared the activity of cPLA 2 in CHOwt and CHOanx6 and measured the cellular release of radiolabeled arachidonic acid ([ 3 H]AA) as described by Grimmer et al. (15). CHOwt express very low amounts of AnxA6 (4,6,8,9), whereas AnxA6 levels in CHOanx6 are rather similar to AnxA6 expression levels in other commonly used cell lines, such as HeLa (11). As shown in Fig. 1A, both cell lines express comparable levels of cPLA 2 (Fig. 1B). Importantly, these experiments clearly showed a 1.5-2.0-fold reduction of [ 3 H]AA release in CHOanx6 cells (p Ͻ 0.01) (Fig. 1A).
To find out whether AnxA6 down-regulation leads to increased cPLA 2 activity, HeLa cells, which express high levels of AnxA6 (11) (Fig. 1C), were transfected with RNAi targeting AnxA6 (RNAi-AnxA6), and the release of AA was measured. Upon AnxA6 knockdown, the cellular release of AA increased significantly (2.0-fold, Fig. 1C). RNAi-GFP-transfected cells were used as a control and did not show any changes in the release of AA. Then, to address if reduced cPLA 2 activity could be overcome by the addition of exogenous cholesterol in AnxA6-expressing cells, [ 3 H]AA-labeled CHOwt and CHOanx6 cells were incubated with and without cholesterol, and the amount of released radioactivity was determined as described above (Fig. 1D). Similar to the results described in Fig. 1A, release of [ 3 H]AA is reduced in CHOanx6 cells compared with the controls (100 Ϯ 6 and 65 Ϯ 7% for CHOwt and CHOanx6, respectively). In agreement with previous data (15), exogenous cholesterol increased cPLA 2 activity 1.54 Ϯ 0.10and 2.12 Ϯ 0.37-fold in both cell lines, respectively (n ϭ 3). Thus, the addition of cholesterol compensates for the inhibitory effect of AnxA6 on cPLA 2 activity in CHOanx6 cells to result in a release of [ 3 H]AA, which is comparable with that of the controls. It is tempting to speculate that the retention of cholesterol in late endosomes of CHOanx6 cells (11) leads to an impaired supply of cholesterol for cPLA 2 , thereby interfering with Golgi vesiculation.
Impaired Translocation of cPLA 2 to the Golgi Complex in AnxA6-expressing Cells-Then, to address if elevated AnxA6 levels perturb cPLA 2 translocation to the Golgi, we isolated Golgi membranes from CHOwt and CHOanx6 cells and compared the amount of cPLA 2 . Western blot analysis ( Fig.  2A) revealed reduced cPLA 2 levels in the Golgi-enriched membrane fraction of CHOanx6 cells. Because CHOanx6 are characterized by an accumulation of cholesterol in late endosomes leading to reduced cholesterol levels in the Golgi (11), we concluded that reduced cPLA 2 levels in the Golgi of CHOanx6 cells could be due to an ineffective delivery of cholesterol from late endosomes to the Golgi complex. In support of this hypothesis, treatment of CHOwt cells with U18666A, a pharmacological agent to accumulate cholesterol in late endosomes, also resulted in a clear reduction of cPLA 2 in the Golgi fractions without affecting total amounts of cPLA 2 (Fig. 2B, compare the second and fourth lanes). To confirm these findings by means of immunofluorescence microscopy, the staining pattern of cPLA 2 and a Golgi marker (␣-mannosidase II) in CHOwt and CHOanx6 was compared (Fig. 3A). In both cell lines punctuate and in part diffuse staining of cPLA 2 was observed throughout the cell and was more intense in the perinuclear region, in particular in CHOwt cells. In the CHOanx6 cell line, cPLA 2 appeared cytosolic with minor cPLA 2 staining detected within the Golgi complex as judged by colocalization with anti-␣-mannosidase II and anti-cPLA 2 (see Fig. 3A, arrows in panel a-c). Quantification of fluorescence intensity confirmed the increased co-localization (ϳ3-fold) of cPLA 2 with ␣-mannosidase II in CHOwt compared with CHOanx6 cells (Fig. 3B). In addition, when Golgi membranes were isolated from CHOwt or CHOanx6 cells in the presence of cholesterol (90 min) and analyzed by Western blotting, a significant increase of cPLA 2 levels was observed (Fig. 3C). Thus, in CHOanx6 cells the reduced localization of cPLA 2 in the Golgi correlates with a decreased enzymatic activity of cPLA 2 (Fig. 1A).
To verify and extend our findings on endogenous cPLA 2 , CHOwt and CHOanx6 were transiently transfected with the   full-length cPLA 2 ␣ fused to (GFP-cPLA 2 ) (Fig. 4). To study the localization of GFP-cPLA 2␣ upon elevation of Ca 2ϩ or cholesterol, translocation of the GFP-cPLA 2 fusion protein was elicited by treating the cells with the Ca 2ϩ ionophore ionomycin (2.5 M for 2 min) or with cholesterol (30 g/ml for 90 min). In untreated controls, GFP-cPLA 2␣ distribution was mostly cytoplasmic. However, upon Ca 2ϩ increase cPLA 2␣ translocated from the cytoplasm to the perinuclear region including the Golgi area of both CHOwt and CHOanx6 cells. A similar pattern of translocation was previously observed (18,26). Importantly, and similar to the endogenous cPLA 2 , the addition of cholesterol also showed a translocation into the perinuclear region.
To elucidate the identity of the intracellular membranes targeted by GFP-cPLA 2 , cells expressing GFP-cPLA 2 were again stimulated with ionomycin or cholesterol, fixed, and stained with antibodies for a cis-Golgi marker (␣-mannosidase II). As shown in Fig. 4, the most significant colocalization between GFP-cPLA 2 and the Golgi marker was in CHOwt cells in the presence of Ca 2ϩ (panels g-l). Cholesterol treatment also increased the membrane binding of GFP-cPLA 2 to Golgi membranes (panels m-r). Similar results were obtained in CHOanx6 cells.
In support of the biochemical data (Figs. 1D and 2B), the addition of exogenous cholesterol resulted in a significant translocation of cPLA 2 to the Golgi complex (see Fig. 3A; compare panels f and l, and see quantification in Fig. 3B; Fig. 4, panels m-r), which appeared increasingly vesicular and less compact in the presence of cholesterol (compare panels b and e with h or k) in both CHOwt and CHOanx6 cells. Similarly, and as shown previously by others (27), elevation of Ca 2ϩ using Ca 2ϩ ionophores (ionomycin) induced translocation of cPLA 2 to the Golgi in both cell lines (data not shown). In fact, in most studies translocation is associated with increase of cPLA 2 activity (15,28,29). Interestingly, cholesterol-induced vesiculation of the Golgi complex most likely occurs through a mechanism involving Ca 2ϩ -dependent members of the cPLA 2 family (15).
However, the inhibition of cPLA 2 activity in CHOanx6 cells could also occur via a direct interaction of AnxA6 with cPLA 2 (as demonstrated for annexin A1 (30)), thereby interfering with its enzyme activity and/or translocation to Golgi membranes. To test this hypothesis, we therefore performed pulldown assays utilizing AnxA6 fused to GST (GST-AnxA6) as bait (Fig. 5A) and immunoprecipitation experiments with antibodies against AnxA6 and cPLA 2 (Fig. 5, panels B and C). In support of AnxA6 inhibiting cPLA 2 indirectly, these studies confirmed binding of established AnxA6-binding proteins, such as p120GAP and Raf-1, but did not reveal a direct interaction of AnxA6 and cPLA 2 . Even the addition of exogenous cholesterol did not induce the association of cPLA 2 to AnxA6.

Protein Export from the Golgi Is Blocked in MAFP-treated
Cells-Thus, in CHOanx6 cells, reduced cPLA 2 levels in the Golgi correlate with reduced cPLA 2 activity (Figs. 1 and 2) and an accumulation of cav-1 in the Golgi (11); see also Fig. 6, compare panels a and b. As mentioned above, CHOwt express only residual amounts of AnxA6 (7,9,11), and similar to RNAi approaches, this very low concentration of AnxA6 makes CHOwt an appropriate model system to analyze if inhibition of cPLA 2 activity per se in an AnxA6-independent manner can result in an accumulation of cav-1 in the Golgi. Therefore, CHOwt (and CHOanx6) were incubated with Ca 2ϩ -independent (HELSS) and Ca 2ϩ -dependent (MAFP) cPLA 2 inhibitors, and the staining of cav-1 was compared (Fig. 6). No differences in the subcellular distribution of cav-1 were observed in CHOwt cells after incubation with HELSS (Fig. 6, compare panels a and c). However, in CHOwt cells treated with MAFP, the cav-1 staining was more intense in the Golgi area compared with untreated CHOwt controls (98 Ϯ 1.74% of CHOwt cells showed the accumulation of cav-1 in the Golgi) but similar to the one obtained in CHOanx6 cells (11) (Fig. 6, compare panel  a with e and b).
These findings suggest that AnxA6 is a component of the molecular machinery involved in the regulation of cPLA 2 . Thus, in cells with high AnxA6 levels, such as CHOanx6, the inhibition of the cholesterol-and Ca 2ϩ -dependent activity of members of the multifactorial cPLA 2 protein family could contribute to an accumulation of cav-1 in the Golgi. Together with the diminution of cholesterol at the cell surface, the sequestration of cholesterol in late endosomes and the significantly reduced Golgi cholesterol in CHOanx6 cells (ϳ38% reduced compared with CHOwt) (11), this could lead to a slower and reduced trafficking of cav-1 from the Golgi to the plasma membrane in CHOanx6 cells.
We previously demonstrated the requirement of cholesterol for caveolin export from the Golgi (16). To validate if Ca 2ϩ -dependent cPLA 2 activity is involved in cav-1 export, we com-  A, pulldown experiments using GST-AnxA6 as bait with and without exogenous cholesterol (ϩChol) were performed to study the potential interaction of AnxA6 with cPLA 2 . CHOanx6 lysates (WCL) were incubated with GST-AnxA6 for 2 h at 4°C. Bound proteins (GST-AnxA6) were collected, separated by SDS-PAGE, and analyzed by Western blotting using antibodies to cPLA 2 ; Raf-1 and p120GAP were used as positive control. Pulldown assays using GST alone (GST) as a negative control did not show binding of cPLA 2 and p120GAP, respectively. B and C, lysates of CHOwt and CHOanx6 cells expressing GFP-cPLA 2 were immunoprecipitated (IP) with anti-GFP and anti-AnxA6 antibodies. The presence of cPLA 2 and AnxA6 in the immunoprecipitates was analyzed by Western blotting. NRS, normal rabbit serum.
pared the export of caveolin from the Golgi complex of COS-1 cells with and without MAFP by inverse fluorescence recovery after photobleaching (iFRAP) experiments (16). To measure and visualize the kinetics of caveolin export and to avoid any interference of GFP-tagged cav-1 with endogenous cav-1, we monitored the localization of GFP-tagged cav-3 (GFP-cav3) (trafficking of GFP-cav-1 and GFP-cav-3 is comparable in commonly used cell lines that do not express cav-3) (16,31). Therefore, COS-1 cells treated with and without MAFP were transfected with GFP-cav3, and the entire cytoplasm except the Golgi region of transfected cells was photobleached. Then, GFP-cav3 trafficking from the Golgi was monitored by confocal microscopy. Results from individual cells were compared, and a representative experiment is shown (Fig. 7A, for quantification see Fig. 7B). As expected, 40 min after photobleaching, 53 Ϯ 2.23% of GFP-cav3 fluorescence disappeared from the Golgi in COS-1 control cells (mobile fraction, for quantification details see methods), but much slower kinetics for GFP-cav3 export (8 Ϯ 2.17%, Fig. 7B) from the Golgi were observed in MAFPtreated COS-1 cells, further supporting Ca 2ϩ -dependent cPLA 2 proteins contributing to promote cav-1 export from the Golgi. In CHOanx6 cells, the inhibitory effect of Anx6 on Golgi export appears specific for caveolin, as kinetics for MAL, a raftassociated integral membrane protein, and EGFR, which is found in lipid rafts and clathrin-coated pits, were identical to CHOwt controls. Interestingly, in these follow-up iFRAP experiments in COS-1 cells, MAFP-mediated inhibition of cPLA 2 not only inhibited export of GFP-cav3 but also GFP-MAL and GFP-EGFR (Fig. 7B). These findings could point at cell-specific differences in CHO and COS-1 cells with respect to the regulation of Ca 2ϩ -dependent cPLA 2 proteins. Alternatively, cPLA 2 proteins might act as common regulators of Golgi vesiculation, and modulators, including AnxA6, might interfere only with the activity of cPLA 2 subpopulations, thereby creating specificity in the regulation of protein export from the Golgi in some cell types. However, these experiments would also allow independent pathways of AnxA6 and cPLA 2 interfering with protein transport through the Golgi complex, and future studies will have to clarify this issue.
Although the results shown here and previously (11) implicate a role for AnxA6 in cPLA 2 inhibition via sequestration of cholesterol in late endosomes, other members of the annexin family could also be involved. In particular, annexin A1 (AnxA1) binds to cPLA 2 (31, 32) and has recently been described to regulate membrane transport in the late endosomal compartment (33). To address the specificity of AnxA6induced effects on cav-1 transport relative to other annexins, we co-transfected COS-1 cells with GFP-cav3 with and without AnxA1 expression vectors and studied if AnxA1 affects cav-3 export from the Golgi complex by means of iFRAP experiments. AnxA1 overexpression did not alter the distribution of cholesterol, as judged by filipin staining (data not shown). In addition, results shown in Fig. 7C clearly demonstrate that the kinetics of GFP-cav3 export from the Golgi are independent of AnxA1 co-overexpression.
Finally, we analyzed the ability of cholesterol upon inhibition of cPLA 2 to induce vesiculation and possibly to re-establish the transport of cav-1 from the Golgi. CHOanx6 cells were treated for 2 h with MAFP with and without cholesterol for 90 min and then stained with cav-1 together with GM130 to visualize the Golgi apparatus. Fig. 8, panels a, d, and g, shows the typical pattern of cav-1 accumulation in the Golgi of CHOanx6 cells. Treatment of MAFP had no apparent effect on overall Golgi structure or the accumulation of cav-1 in the Golgi (compare panels b with e and h). The addition of exogenous cholesterol in the presence of MAFP was not sufficient to restore Golgi vesiculation (see panel h). However, when cells were preincubated with MAFP followed by extensive washing and removal of MAFP, incubation with cholesterol resulted in a dramatic vesiculation of the Golgi apparatus in CHOanx6 cells (panel k). This was accompanied by a redistribution of cav-1 from the Golgi as judged by the appearance of cav-1 staining at the plasma membrane (panel j). These results strongly suggest that the exit of cav-1 from the Golgi complex not only requires cholesterol but also functional cPLA 2 . In fact, exogenous cholesterol seems to target and compete for the same population of Ca 2ϩ -dependent cPLA 2 proteins that is inhibited by MAFP. Therefore, the addition of cholesterol cannot re-establish the exit of caveolin from the Golgi complex when MAFP is present and Ca 2ϩ -dependent cPLA 2 is inhibited (Fig. 8, panel h).

DISCUSSION
Data presented in this study support our hypothesis that the concomitant sequestration of cholesterol in late endosomes together with reduced amounts of cholesterol in Golgi membranes of CHOanx6 cells (11) reduces the activity of cholesterol-and Ca 2ϩ -dependent cPLA 2 , thereby contributing to an accumulation of cav-1 in the Golgi. The retention of cholesterol in late endosomes of CHOanx6 cells (11) leads to an impaired supply of cholesterol for cPLA 2 , thereby interfering with Golgi vesiculation. This is supported by several studies demonstrating that both cPLA 2 and cholesterol are required for the formation of vesicles in the secretary pathway. cPLA 2 has been implicated in Golgi tubulation and vesiculation events (24,25,34). Both constitutive and regulated formation of secretory vesicles requires cholesterol. Depletion of cholesterol inhibits the formation of secretory vesicles from the trans-Golgi network (35). Vice versa, increased cellular cholesterol levels induce pronounced vesiculation and dispersal of Golgi-derived vesicles (15). Along these lines, Stüven et al. (36) showed that cholesterol levels at the Golgi complex appear precisely balanced to allow protein transport to occur.
It has recently been identified that cholesterol-sensitive cPLA 2 activity (15,37) correlates with the increased translocation of cPLA 2 to the Golgi apparatus in response to elevation of cellular cholesterol levels (15). Binding of cPLA 2 to Golgi membranes might then facilitate the generation of lysophospholipids, such as lysophosphatidic acid, and stimulate the vesiculation of the Golgi apparatus through the action of C-terminal-binding protein/brefeldin ␣-ribosylated substrate (34). However, other factors like dynamin (15), the actin cytoskeleton (38), and calmodulin (39) are also involved in this process, suggesting that multifactorial protein-lipid interactions regulate Golgi vesiculation.
Taken together, mechanistically several direct/indirect events seem to converge to block the transport of caveolin in cells expressing elevated levels of AnxA6; first, AnxA6 appears to impair cholesterol transport into Golgi membranes driven by the lysosomal membrane protein NPC1 (11); second, AnxA6-induced reduction of Golgi-cholesterol leads to the loss of cPLA 2 activity through its displacement from the Golgi membranes. It should be noted that other mechanisms involving the direct binding of AnxA6 to phosphatidylserineenriched membrane domains in lysosomes, endosomes, or the plasma membrane, thereby affecting cPLA 2 activity and/or late endosomal cholesterol transport, cannot be completely ruled out yet (40). One possible scenario might involve the mem-   were transfected with GFP-cav3, GFP-MAL, and GFP-EGFR, treated with and without MAFP as indicated, and imaged before and after photobleaching as described above. The loss of fluorescence signal (iFRAP) from the Golgi was quantified. Kinetic parameters (mobile fraction in % ϮS.D.) for each protein with and without MAFP from four independent experiments are given. C, COS-1 cells were co-transfected with GFP-cav3 together with AnxA1 or empty control vector, and the export of GFP-cav3 from the Golgi was determined in iFRAP experiments as described above (for details, see "Experimental Procedures").
brane recruitment of AnxA6 interfering/competing with constitutive trafficking events determined by tethering and/or SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptors) proteins, which eventually could arrest membrane transport (41). We were not able to identify cPLA 2 associated with AnxA6 in pulldown or co-immunoprecipitation experiments. Hence, together with the beneficial effects of exogenous cholesterol on cPLA 2 activity in CHOanx6 cells, a direct inhibitory effect of AnxA6 on cPLA 2 appears unlikely. In contrast, the direct interaction and inhibitory effect of other annexins, including A1, B1, and A5, on cPLA 2 activity has been demonstrated (31,32). Although we have also been unable to identify AnxA6 in the Golgi complex, a direct interaction of AnxA6 with cPLA 2 in the cytoplasm or at the plasma membrane cannot completely be ruled yet, thereby trapping cPLA 2 and making it impossible for cPLA 2 to translocate to the Golgi. Finally, and as a further contribution of this study, we have identified that out of the cPLA 2 protein family involved, the cytosolic group IVA phospholipase A 2 ␣ is sensitive to cholesterol diminution at the Golgi. The alteration of cPLA 2 activity and translocation would ultimately not only reduce vesiculation from the Golgi but also down-regulate the availability of newly synthesized cholesterol acceptors, such as cav-1, for vesicular or non-vesicular export of cholesterol from other compartment, such as late endosomes (42). Therefore, this study also provides evidence of a new player, AnxA6, to be indirectly involved in the molecular mechanisms of the secretory pathway. AnxA6 induces an imbalance of intracellular cholesterol, thereby inhibiting cPLA 2 activity, which ultimately regulates the exit of caveolin from the Golgi complex.
Nevertheless, for the regulation of cPLA 2 activity and localization, additional players like the lipidic environment, the actin cytoskeleton, and the direct interaction of the C2 domain of cPLA 2 with other proteins cannot be excluded. Several proteins have been shown to interact with the C2 domain of cPLA 2 in vitro, including annexins A1 and A5 (43,44), vimentin (45), and PLIP (cPLA 2 -interacting protein) (46). Nevertheless, their role in the regulation of cPLA 2 is still unclear. Future studies, in particular RNAi knockdown experiments, should help to clarify their contribution in modifying cPLA 2 activity and localization.
From the physiological point of view it is evident that the regulation of caveolin transport to the plasma membrane could have critical consequences in various cellular processes, considering the role of caveolin in the formation of caveolae at the cell surface and the importance of these structures in endocytosis, membrane microdomain formation, lipid transport (including cholesterol efflux), and cellular signaling. Most interestingly, in most cell types caveolin expression is strongly linked to the expression of PTRF-cavin, a recently identified protein that is required for caveolae formation (13,14). In fact, CHOanx6 cells were grown on coverslips and incubated with and without MAFP and cholesterol (chol) as indicated. Cells were fixed, permeabilized, and double-labeled for cav-1 and the Golgi marker GM130 (see "Experimental Procedures" for details). When cPLA 2 was inhibited (panels d and g), the Golgi is compact (panels e and h), and cav-1 accumulated in the Golgi complex (panels d and g). After removing the cPLA 2 inhibitor (panel j), the Golgi becomes more vesiculated (panel k), and cav-1 is not retained in the Golgi complex anymore (panel j). Relative fluorescence intensity at the plasma membrane was calculated by the Image J program (National Institutes of Health). The white and black bars relate to the average fluorescence intensity (ϮS.D.) of cav-1 in panel g and j, respectively. Images are representative for three independent experiments. Bar, 10 m. caveolin is expressed in prostate cancer PC3 cells and during development of zebrafish notochord, but both model systems lack caveolae, most likely due to their lack of PTRF-cavin expression (13,14). It is yet unknown if expression levels of other proteins are linked to caveolin function or caveolae formation. However, we recently identified a significant diminution (ϳ60%) in the number of caveolae in EGF receptors overexpressing carcinoma cells (A431) and breast cancer cell lines with high, but not low levels of AnxA6 (11). Future studies will have to clarify if levels of AnxA6 are possibly linked to caveolae formation in development and/or the potential tumor suppressor function of caveolin in some cancers such as breast cancer.