Phosphatidylinositol 4-Kinase IIα Is Palmitoylated by Golgi-localized Palmitoyltransferases in Cholesterol-dependent Manner*

Background: Palmitoylation of phosphatidylinositol 4-kinase IIα (PI4KIIα) regulates its function and Golgi localization, and cholesterol depletion delocalizes Golgi PI4KIIα and inhibits activity. Results: Palmitoyl acyltransferases (PATs) that palmitoylate PI4KIIα were identified. Cholesterol extraction inhibited PI4KIIα association with PATs, decreased palmitoylation, and reduced Golgi phosphatidylinositol 4-phosphate. Conclusion: Cholesterol has a critical role in regulating PI4KIIα interaction with PATs and palmitoylation. Significance: This study uncovered a novel mechanism for preferential recruitment of PI4KIIα to Golgi. Phosphatidylinositol 4-kinase IIα (PI4KIIα) is predominantly Golgi-localized, and it generates >50% of the phosphatidylinositol 4-phosphate in the Golgi. The lipid kinase activity, Golgi localization, and “integral” membrane binding of PI4KIIα and its association with low buoyant density “raft” domains are critically dependent on palmitoylation of its cysteine-rich 173CCPCC177 motif and are also highly cholesterol-dependent. Here, we identified the palmitoyl acyltransferases (Asp-His-His-Cys (DHHC) PATs) that palmitoylate PI4KIIα and show for the first time that palmitoylation is cholesterol-dependent. DHHC3 and DHHC7 PATs, which robustly palmitoylated PI4KIIα and were colocalized with PI4KIIα in the trans-Golgi network (TGN), were characterized in detail. Overexpression of DHHC3 or DHHC7 increased PI4KIIα palmitoylation by >3-fold, whereas overexpression of the dominant-negative PATs or PAT silencing by RNA interference decreased PI4KIIα palmitoylation, “integral” membrane association, and Golgi localization. Wild-type and dominant-negative DHHC3 and DHHC7 co-immunoprecipitated with PI4KIIα, whereas non-candidate DHHC18 and DHHC23 did not. The PI4KIIα 173CCPCC177 palmitoylation motif is required for interaction because the palmitoylation-defective SSPSS mutant did not co-immunoprecipitate with DHHC3. Cholesterol depletion and repletion with methyl-β-cyclodextrin reversibly altered PI4KIIα association with these DHHCs as well as PI4KIIα localization at the TGN and “integral” membrane association. Significantly, the Golgi phosphatidylinositol 4-phosphate level was altered in parallel with changes in PI4KIIα behavior. Our study uncovered a novel mechanism for the preferential recruitment and activation of PI4KIIα to the TGN by interaction with Golgi- and raft-localized DHHCs in a cholesterol-dependent manner.

of PI4KII␣ functions (14), the palmitoyl acyltransferases (PATs) that palmitoylate PI4KII␣ have not been identified. Furthermore, although recent studies by others have shown that the enzymatic activity and Golgi localization of PI4KII␣ and its lateral diffusion within membranes are decreased by cholesterol depletion (9,16), the relation between cholesterol and PI4KII␣ palmitoylation has not been examined.
Humans and mice have 23 phylogenetically conserved PATs that contain a signature Asp-His-His-Cys (DHHC) motif (17)(18)(19)(20)(21)(22)(23). Here, we report that PI4KII␣ is palmitoylated by at least six DHHC PATs and establish that the primarily Golgi-localized DHHC3 and DHHC7 are bona fide PI4KII␣ PATs: they colocalized with PI4KII␣ in the TGN; their silencing by RNAi decreased PI4KII␣ palmitoylation, integral membrane association, and Golgi localization; and they associated with PI4KII␣ in pulldown assays. We also show that PI4KII␣ palmitoylation is cholesterol-dependent: cholesterol depletion reduced the association of PI4KII␣ with DHHCs and decreased its palmitoylation, integral membrane association, and Golgi localization. It also reduced the amount of PI4P at the TGN. Significantly, cholesterol depletion did not decrease DHHC Golgi localization, suggesting that PI4KII␣, and not the DHHCs, is the primary cholesterol-dependent target.
Cell Transfection-HeLa, COS-7, and HEK293 cells were used in different experiments. In general, HeLa cells, which express transfected proteins at relatively low levels, were used for immunofluorescence and RNAi studies. COS-7 cells, which express transfected proteins at moderate levels, were used for co-immunoprecipitation studies. HEK293 cells, which express transfected proteins at high levels, were used in palmitoylation screens. Cells cultured in DMEM containing 10% FBS and antibiotics were transfected with epitope-tagged rat PI4KII␣ (15) and mouse DHHC (19) cDNAs using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were assayed after 12-24 h. siRNAs that target different regions of human DHHC3 or DHHC7 were obtained from Santa Cruz Biotechnology, Inc., and Sigma (DHHC3 siRNA, Santa Cruz sc-75158/Sigma SASI_Hs01_00133396; DHHC7 siRNA, Santa Cruz sc-93249/Sigma SASI_Hs01_00033548). siRNAs were transfected into HeLa cells with RNAiMAX (Invitrogen) and used after 48 -68 h.
[ 3 H]Palmitate Labeling-Cells transfected with Myc-PI4KII␣ with or without HA-DHHC were preincubated for 30 min in DMEM with fatty acid-free BSA (5 mg/ml). They were then labeled with 0.5 mCi/ml [ 3 H]palmitic acid (PerkinElmer Life Sciences) for 4 h in DMEM with fatty acid-free BSA (5 mg/ml) and washed with PBS. In some cases (e.g. Fig. 1B), cells were scraped with SDS-PAGE sample buffer (62.5 mM Tris-HCl (pH 6.8), 10% glycerol, 2% SDS, 10 mM DTT, and 0.001% bromphenol blue) and boiled for 2 min (18). Proteins were resolved by SDS-PAGE. Gels were treated with Amplify (GE Healthcare) for 30 min, dried under vacuum, and exposed to Kodak Biomax MS at Ϫ80°C. Parallel gels were subjected to Western blotting. Band intensity was analyzed with ImageQuant 5.2 software. In other experiments, Myc-PI4KII␣ was immunoprecipitated prior to Western blotting or fluorography (see below) (14).
Immunoprecipitation-Cells were homogenized in solution containing 1% Brij 98, 150 mM NaCl, 5 mM MgCl 2 , 25 mM MES (pH 6.5), 10 mM Na 4 P 2 O 3 , 2 mM NaF, 0.1 mM Na 2 VO 4 , and protease inhibitors (14,24). Lysates were centrifuged at 15,000 ϫ g for 15 min at 4°C, and the resulting supernatants were incubated with recombinant protein G-Sepharose B for 30 min to remove nonspecifically bound proteins. The precleared supernatants were incubated with anti-Myc or anti-HA antibody overnight at 4°C and subsequently with protein G-Sepharose for 1 h at 4°C. Beads were collected by centrifugation and washed extensively. Immunoprecipitated proteins were resolved by SDS-PAGE and subjected to Western blotting or fluorography.
Protein and Palmitate Turnover Measurements-Cells transfected with Myc-PI4KII␣ were incubated with methionine/cysteine-free DMEM for 1 h, labeled with 50 Ci/ml [ 35 S]Met/Cys for 1 h, and subsequently chased with unlabeled 2 mM Met/ Cys for 2, 4, and 6 h. In parallel, another aliquot of cells was incubated for 1 h in DMEM with 5% dialyzed FBS, labeled with [ 3 H]palmitate for 4 h, and then chased with unlabeled 100 M palmitate (25). Cells were lysed and immunoprecipitated with anti-Myc antibody. Immunoprecipitated proteins were resolved by SDS-PAGE and analyzed by fluorography.
Immunofluorescence Microscopy-To label proteins in cells, in most cases, cells were fixed with 3.7% formaldehyde in PBS for 10 min at room temperature, permeabilized with 0.1% Triton X-100 in PBS for 5 min on ice, blocked with 1% BSA and 3% donkey serum in PBS, and stained with primary antibodies and rhodamine-or FITC-conjugated secondary antibodies. Cells used for GM130 labeling were fixed and permeabilized in methanol at Ϫ20°C for 10 min. Cells were examined using a Zeiss LSM 510 laser scanning confocal microscope.
To label Golgi PI4P, a fixation/permeabilization protocol that was optimized for detecting Golgi PI4P was used (26). All steps were performed at room temperature. Cells on coverslips were fixed with 2% formaldehyde in PBS. After 15 min, coverslips were rinsed in PBS containing 50 mM NH 4 Cl, and cells were permeabilized by incubation with 20 M digitonin in buffer A (20 mM PIPES (pH 6.8), 17 mM NaCl, and 2.7 mM KCl) for 5 min. After washing, permeabilized cells were incubated for 45 min with buffer A supplemented with 5% (v/v) donkey serum and 50 mM NH 4 Cl. Cells were stained with anti-PI4P antibody and rhodamine-conjugated secondary antibody in buffer A with 5% donkey serum. Golgi PI4P intensity was analyzed using ImageJ software.
To assess colocalization, cells coexpressing different tagged proteins were analyzed. For each cell, 10 -12 images at a slice interval of 0.3 m were captured using the Z-stack scanning function of the Zeiss LSM 510 confocal microscope. Colocalization was analyzed using the Zeiss LSM 510 ZEN software. Fluorescent signals within the same voxel were considered to be colocalized and are expressed as percentage colocalization. Ten randomly chosen cells from each group were analyzed in each experiment.
Preparation of Cytosol and Sequentially Extracted Membrane Fractions-A sequential extraction procedure described previously was used (14). All solutions were maintained at 4°C and contained protease and phosphatase inhibitors (50 mM NaF, 50 mM glycerophosphate, 1 mM Na 2 VO 4 , and 10 M microcystin). Cells expressing Myc-PI4KII␣ were washed with ice-cold PBS and scraped from Petri dishes in a solution containing 0.25 M sucrose, 20 mM Tris-HCl (pH 7.5), 0.1 M NaCl, and 1 mM EDTA. Cells were subjected to two freeze-thaw cycles and passed through a 27-gauge needle. Lysates were centrifuged at 1000 ϫ g for 5 min to remove unbroken cells and nuclei. The post-nuclear supernatant was then centrifuged at 200,000 ϫ g for 15 min to separate the cytosol from the membranes. The resulting membrane pellets were homogenized in solution containing 20 mM Tris-HCl (pH 7.5), 1 M NaCl, and 1 mM EDTA to extract weakly bound peripheral membrane proteins (fraction 1, salt extraction). Membranes were collected by another round of centrifugation (200,000 ϫ g for 15 min) and homogenized in 0.1 M sodium carbonate (pH 11) to extract tightly bound peripheral membrane proteins (fraction 2, carbonate extraction). Finally, integrally associated membrane proteins were extracted by homogenization in 1% Triton X-100 (fraction 3, detergent extraction).
Preparation of Low Buoyant Density Membrane Fractions-Low buoyant density detergent-resistant membrane raft fractions were prepared using Brij 98 (14,24). Cells were homogenized in solution containing 1% Brij 98, 150 mM NaCl, 5 mM MgCl 2 , 25 mM MES (pH 6.5), 10 mM Na 4 P 2 O 3 , 2 mM NaF, 0.1 mM Na 2 VO 4 , and protease inhibitors. The lysate (1.5 ml) was mixed with an equal volume of 80% (w/v) sucrose and overlaid with 5.5 ml of 35% sucrose and 3.5 ml of 5% sucrose in buffer without Brij 98. Following centrifugation at 210,000 ϫ g for 16 h at 4°C in an SW40 rotor, 1-ml fractions were collected from the bottom of the tube, and equal aliquots were subjected to SDS-PAGE and immunoblotting.
Cholesterol Depletion and Repletion-Cells were sterol-depleted by incubation with 10 mM methyl-␤-cyclodextrin (M␤CD; Sigma C4555) for 30 min at 37°C in serum-free DMEM (9). Cholesterol repletion was performed by incubating cells with DMEM with 5% lipoprotein-deficient serum with or without 2.5 mM water-soluble cholesterol (complexed with M␤CD; Sigma C4951) for 2 h.
Statistical Analysis-Quantitative results are expressed as means Ϯ S.E. Comparison between groups was performed using one-way analysis of variance.
Two independent sets of siRNA generated similar results. Depletion of DHHC3 was confirmed by Western blotting (Fig.  2B) and immunofluorescence labeling (Fig. 3A). Depletion of DHHC7 was established by the decrease in immunofluorescence labeling only (Fig. 3A) because the commercially available anti-DHHC7 antibody did not generate a clear signal in Western blotting.
We also examined the effects of DHHC3 or DHHC7 RNAi on PI4KII␣ intracellular distribution. Depletion of DHHC3 or DHHC7 by RNAi dispersed PI4KII␣ throughout the cytoplasm in small punctae (Fig. 3A). Because we have shown previously that palmitoylation promotes PI4KII␣ association with the TGN (14), these results supported the conclusion that DHHC3 and DHHC7 palmitoylate PI4KII␣ to promote their Golgi targeting.
Unexpectedly, the DHHC3 RNAi cells and, to a lesser extent, the DHHC7 RNAi cells had almost complete loss of perinuclear TGN46 staining (Fig. 3, A and B). The extent of TGN46 dispersal in the DHHC3 RNAi cells was more pronounced than that observed with PI4KII␣ RNAi per se (Fig. 3B). By comparison, staining of the cis-Golgi matrix protein GM130 was much less severely affected than TGN46 in DHHC3-depleted cells (Fig.  3B). Thus, these DHHCs are critically important for maintaining the integrity of the TGN. We propose that this is mediated partly by interference with PI4KII␣ palmitoylation and hence PI4P generation.
To rule out off-target effects due to RNAi, we also examined the effect of expressing DN-DHHCs on the TGN morphology.
In control transfected cells, ϳ65% of the cells had normal compact TGN46 staining (Fig. 3C). In contrast, only 27 and 37% of cells transfected with DN-DHHC3 or DN-DHHC7, respectively, displayed compact TGN staining (Fig. 3C). Taken together, our results established that DHHC3 and DHHC7 are individually necessary for TGN integrity and functions, includ- ing PI4KII␣ Golgi recruitment and palmitoylation. Additional studies will be required to identify other mechanisms whereby DHHCs contribute to TGN integrity.
We examined the requirements for this interaction by using PI4KII␣ mutants. The truncation mutant containing the catalytic domain (amino acids 93-478) (Fig. 4B, Cat) and the embedded CCPCC palmitoylation motif associated with DHHC3, but the non-palmitoylatable SSPSS mutant did not. Thus, PI4KII␣ interacts with DHHC3 and DHHC7, and its CCPCC palmitoylation motif, but not the N-or C-terminal extensions, is required for interaction.
Dependence of PI4KII␣ Palmitoylation on Cholesterol-Although previous studies have shown independently that PI4KII␣ catalytic activity and Golgi localization are dependent on palmitoylation (14) as well as on cholesterol (9,16,34), the relation between palmitoylation and cholesterol has not been examined. To examine this relation, we manipulated cholesterol in intact cells using the M␤CD depletion and repletion protocols that have previously been optimized to examine effects on PI4KII␣ behavior in cells and in membranes (9,16). M␤CD decreased Myc-PI4KII␣ co-immunoprecipitation with HA-DHHC7 (Fig. 4C). Conversely, association was restored by reintroduction of cholesterol complexed with M␤CD. Similar results were obtained with HA-DHHC3 (data not shown). Furthermore, cholesterol depletion/repletion also reversibly impacted PI4KII␣ palmitoylation (Fig. 5A).
Previous independent studies have shown in discontinuous gradient centrifugation experiments that PI4KII␣ (14) and DHHC3 and DHHC7 (17) were partially associated with deter- Endogenous DHHC3 was detected with anti-DHHC3 antibody in Western blots. C, DHHC RNAi effect on PI4KII␣ membrane extractability. Cells were homogenized, and the post-nuclear supernatant (lysate) was analyzed by Western blotting (left). The lysate was centrifuged at 200,000 ϫ g to obtain the cytosolic fraction (c) and membrane pellet. Membranes were extracted sequentially with 1 M NaCl, 0.1 M Na 2 CO 3 , and 1% Triton-X100 (fractions 1-3, respectively) (right). Equivalent amounts of each fraction were analyzed by immunoblotting with anti-Myc antibody. The percentage of total PI4KII␣ in each fraction is indicated below the Western blot. gent-insoluble light membrane fractions (rafts). We confirmed that this is indeed the case in parallel experiments (Fig. 5B). We compared the effects of cholesterol depletion on the extractability of PI4KII␣ and DHHCs. M␤CD decreased the "integral" membrane association of PI4KII␣, but not DHHC3 or DHHC7 (Fig. 5B). These results clearly established that PI4KII␣ is a peripheral membrane protein that becomes "integrally" associated after cholesterol-dependent palmitoylation, whereas DHHC3 and DHHC7 are bona fide transmembrane proteins (35) that are not dependent on cholesterol for membrane insertion.
Cholesterol depletion decreased PI4KII␣ association with the perinuclear Golgi region and increased PI4KII␣ in cytoplasmic vesicles (Fig. 5C), in agreement with a previous report (9). In contrast, cholesterol depletion did not induce dispersal of DHHC3 and DHHC7 from the Golgi and also did not disrupt perinuclear TGN46 staining to a similar extent, at least at the resolution of the immunofluorescence microscopy (Fig. 5C). These differences suggested that cholesterol depletion induces dispersal of PI4KII␣ from the Golgi by decreasing palmitoylation and not directly as a result of disruption of DHHC localization in the Golgi or disruption of the TGN per se. The impact of cholesterol depletion on PI4P levels in the Golgi was examined by immunofluorescence microscopy using anti-PI4P antibody. This approach was made possible by the availability of a recently optimized fixation/permeabilization protocol that was optimized to preserve Golgi/endomembrane PI4P (26). Cholesterol depletion reduced Golgi PI4P levels by 70%, and Golgi PI4P was restored by cholesterol repletion (Fig.  5D). Taken together, our results established that cholesterol is critical for recruitment of PI4KII␣ to the Golgi by promoting its association with Golgi-localized DHHC3 and DHHC7. Subsequently, palmitoylation and activation of PI4KII␣ increase PI4P generation at the Golgi.

DISCUSSION
S-Palmitoylation regulates the subcellular localization, activity, and raft association of multiple proteins (21-23, 36, 37). Although many proteins are dynamically palmitoylated, others, such as caveolin (29), are stably palmitoylated. Here, we have demonstrated, using pulse-chase experiments, that the palmitoyl groups of PI4KII␣ remain attached to the protein almost throughout its entire life time, indicating that PI4KII␣ is stably palmitoylated. We identified the DHHC PATs that palmitoylate PI4KII␣ and showed that palmitoylation provides a structural signal for PI4KII␣ Golgi targeting, integral membrane anchoring, and catalytic activation and, importantly, that palmitoylation is cholesterol-dependent.
Previous studies have shown that the majority of DHHCs are found in the Golgi or endoplasmic reticulum, but a minority are also present at extra-Golgi sites and on the plasma membrane (38 -41). Here, we have shown that PI4KII␣ can be palmitoylated by at least six DHHC PATs (DHHC2, DHHC3, DHHC7, DHHC14, DHHC15, and DHHC21). The profile of positive PATs is similar, but not identical, to that reported for several other unrelated proteins. For example, although endothelial NOS and stathmin are palmitoylated by DHHC2, DHHC3, and DHHC7 (17,39), endothelial NOS is not palmitoylated by DHHC14 or DHHC15, and stathmin is not palmitoylated by DHHC14 or DHHC21. DHHC2 is found on the plasma membrane and in vesicles that have the characteristics of recycling endosomes (40,41). Because PI4KII␣ is enriched in the Golgi and also in endosomes (7,8,10), the possibility that it may be palmitoylated by DHHC2 at extra-Golgi sites should be explored.
In light of our primary interest in PI4KII␣ palmitoylation in the Golgi, we focused on DHHC3 and DHHC7 in this study. We have shown, using DN-PATs and PAT knockdown by RNAi, that DHHC3 and DHHC7 are bona fide PI4KII␣ PATs that are colocalized and associated with PI4KII␣. Binding is independent of the ability of PATs to palmitoylate because both the WT and palmitoylation-defective DN-DHHCs bind PI4KII␣. However, because the PI4KII␣ SSPSS mutant does not bind to DHHCs, it appears that the CCPCC palmitoylation motif is nevertheless required. Importantly, non-candidate DHHC18 and DHHC23 did not associate with PI4KII␣. These features suggest that PI4KII␣ binds candidate DHHCs specifically in a manner that is dependent on the CCPCC motif, and it becomes palmitoylated by the catalytically active DHHC after physical association. Additional studies will be required to determine whether PI4KII␣ binds DHHC3 or DHHC7 directly or indirectly.
Our results provide a mechanistic model to explain how PI4KII␣ is preferentially targeted to the Golgi for palmitoylation. We have previously proposed a multistep process (14): 1) although non-palmitoylated PI4KII␣ has high intrinsic membrane-binding ability and can potentially bind to all mem- branes, newly synthesized (not yet palmitoylated) PI4KII␣ is nevertheless preferentially recruited to the Golgi by binding to a hypothetical Golgi "docking protein"; 2) PI4KII␣ penetrates the Golgi membrane bilayer by virtue of its amphipathic membrane-anchoring loop; and 3) this positions the CCPCC motif against the membrane and brings it within reach of the Golgi-localized DHHCs. These cysteines are palmitoylated, and the covalently attached palmitate groups provide a strong secondary anchor to kinetically trap PI4KII␣ in the Golgi and in Golgiderived membranes (14).
The data presented here suggest that the DHHCs are the docking proteins and also the PATs for PI4KII␣. The multistep . Cells were homogenized in 1% Brij 98 and subjected to centrifugation in a discontinuous sucrose step gradient. Protein distribution in the collected fractions was monitored by immunoblotting with anti-Myc (PI4KII␣), anti-HA (DHHC), and anti-flotillin (a detergent-resistant membrane marker) antibodies. Right, "integral" membrane association. Cell lysates were centrifuged, and the 200,000 ϫ g supernatant (c) was collected. The pellets were extracted sequentially with NaCl, Na 2 CO 3 , and 1% Triton X-100 (fractions 1-3, respectively) and subjected to Western blotting. C, Golgi localization. Cells were labeled with anti-Myc (PI4KII␣), anti-DHHC3 or anti-DHHC7, and anti-TGN46 antibodies. Scale bars ϭ 10 m. The insets show the enlarged merged images of green and red channels. D, Golgi PI4P. Cells were fixed and permeabilized using a protocol optimized for the detection of Golgi PI4P with anti-PI4P antibody. PI4P intensity in the Golgi region (marked by TGN46 staining) was quantified using ImageJ software, averaged by cell numbers, and is expressed relative to the control value. model can therefore be streamlined considerably. This streamlined model for PI4KII␣ Golgi recruitment and palmitoylation provides new insight into how cholesterol regulates PI4KII␣. Because PI4KII␣ and DHHCs are partially raft-associated and cholesterol promotes PI4KII␣ association with DHHC3 and DHHC7, co-partitioning of PI4KII␣ and DHHCs in raft microdomains should promote their interaction and result in higher levels of PI4KII␣ palmitoylation. This model can explain the previous observation that PI4KII␣ in isolated raft membranes is catalytically more active than that in non-raft membranes (9,16,34). Cholesterol depletion has a minimal effect on DHHC association with the Golgi per se, but it decreases PI4KII␣ association with DHHCs presumably by dispersing raft microdomains to decrease encounters between PI4KII␣ and DHHCs. We also showed for the first time in cells that cholesterol depletion dramatically decreases Golgi PI4P, consistent with a decrease in the interaction of PI4KII␣ with DHHCs and, consequently, its palmitoylation.
In this study, we also reviewed for the first time new information about the role of DHHC3 and DHHC7 in maintaining Golgi integrity. Depletion of either DHHC3 or DHHC7 by RNAi or expression of DN constructs disrupts the TGN to a greater extent than PI4KII␣ RNAi per se. These results strongly suggest that DHHCs regulate Golgi integrity partly, but not entirely, by PI4KII␣ palmitoylation and PI4P generation to maintain the organelle identity of the Golgi. Because DHHC3 and DHHC7 palmitoylate multiple proteins at the Golgi (33,41,42), it is not possible at present to definitively assign a cause and effect relation for any one affected Golgi protein. Our discovery of the unanticipated role of DHHCs in maintaining TGN integrity expands our understanding of DHHC biology.