pEGFP-CERT and pEGFP-PI4KIIα were described previously (Perry and Ridgway, 2006; Waugh et al., 2006). Alexa Fluor488-conjugated and unconjugated giantin antibodies were purchased from Covance (Emeryville, CA). Antibodies against γadaptin and PI4KIIIβ were from BD Biosciences Transduction Laboratories (Mississauga, ON, Canada). An antibody against early endosomal antigen 1 (EEA1) was from Affinity Bioreagents (Golden, CO). Sheep anti-human TGN46 and rabbit anti-mouse TGN38 antibodies were from AbD Serotec (Raleigh, NC) and Novus Biologicals (Littleton, CO), respectively. Antibodies against OSBP were described previously (Ridgway et al., 1998). Anti-CERT immunoglobulin (Ig)Y, horseradish peroxidase-conjugated rabbit anti-IgY, and Opti-Prep density gradient media were purchased from Sigma-Aldrich (Oakville, ON, Canada). Mouse monoclonal antibody (mAb) 1C4 was generated according to standard methods using recombinant rat PI4KIIα expressed in Sf9 cells as immunogen. 1C4 specifically recognizes the PI4KIIα isoform from human, rat, and hamster species. The 1C4 hybridoma line was cultured in RPMI 1640 medium containing 2.5% ultralow IgG fetal calf serum (FCS) (Invitrogen, Paisley, United Kingdom) and supplemented with 5% Doma-Drive (Immune Systems, Bristol, United Kingdom). Culture supernatants were used in indirect immunofluorescence (1:2 dilution) and Western immunoblotting (1:4 dilution). We purchased [³H]serine and [³²P]γ-ATP from PerkinElmer-Cetus (Woodbridge, ON, Canada). PtdIns(4)P, PtdIns, SM, and ceramide were from Sigma-Aldrich or Echelon (Salt Lake City, UT).

CHO cells were cultured in DMEM containing 5% FCS and proline (34 μg/ml) (medium A). In some experiments, cells were cultured in DMEM containing 5% lipoprotein deficient serum (LPDS). In CHO cells, siRNAs against OSBP (siOSBP2) (Perry and Ridgway, 2006), PI4KIIα (5′-GAGACGAGCCCGCUAGUGUUU-3′), and PI4KIIIβ (5′-GCAAG GAGCCUGUGUUCAUUU-3′) (Dharmacon RNA Technologies, Lafayette, CO) were used at a concentration of 30, 60, and 30 nM, respectively, to silence expression of the respective genes by using TransIT TKO transfection reagent (Mirus, Madison, WI). An exact sequence match between siRNAs and the hamster PI4KIIα and PI4KIIIβ mRNAs was confirmed by polymerase chain reaction (PCR) amplification and sequencing of the respective cDNAs. HeLa cells cultured in DMEM with 5% FCS were transfected with siRNAs against human PI4KIIIα (siRNA1, 5′-AGAAAGGGAGUGAGGACUAUU-3′; siRNA2, 5′-GUUUAACCGCAUCGGGCUAUU-3′) and PI4KIIIβ (siRNA1, 5′-GGGAUGACCUUCGGCAAGAUU-3′; siRNA2, 5′-GCUGAUUGCCGCUCGGAAAUU-3′) (Dharmacon RNA Technologies) by using TransIT TKO transfection reagent. Controls consisted of an identical concentration of two different nontargeting siRNA (siNT) with at least four mismatches with known human and mouse genes that were used interchangeably with similar results.

CHO cells stably expressing green fluorescent protein (GFP)-CERT were prepared by transfection with pEGFP-CERT by using Lipofectamine 2000 and selection in medium A containing G418 (600 μg/ml) for 14 d. Cells stably expressing the fusion protein were maintained in medium A with 300 μg G418/ml. OSBP was silenced in CHO cells by stable expression of shOSBP (Perry and Ridgway, 2006) by using a lentiviral vector (Block-iT Lentiviral shRNA Expression; Invitrogen) and selection in blasticidin (10 μg/ml) for 2 wk. Control cells stably expressing shOSBP containing three silent mutations to prevent activity (Perry and Ridgway, 2006) were prepared in an identical manner.

Golgi-enriched membranes were isolated on discontinuous sucrose gradients (Balch et al., 1984). In brief, CHO cells were scraped from 100-mm dishes; resuspended in 10 mM Tris-HCl, pH 7.4, 0.25 M sucrose 2 mM EDTA, 2 mM EGTA, 1 mM sodium pyrophosphate, 1 mM NaF, and 200 nM okadaic acid and protease inhibitors (Roche Diagnostics, Mannheim, Germany); and disrupted by 20 strokes in a Dounce homogenizer. The cell homogenate (0.9 ml) was mixed with 0.9 ml of 2.3 M sucrose in a 5-ml centrifuge tube, overlaid with 2.1 ml of 1.2 M sucrose followed by 1.2 ml of 0.8 M sucrose. The gradients were subject to centrifugation for 4 h at 110,000 × g by using an SW55 Ti rotor (Beckman Coulter, Fullerton, CA). Fractions were collected corresponding to the 0.8 M sucrose layer (fraction I), the 0.8 M/1.2 M interface (fraction II), the 1.2 M sucrose layer (fraction II), the 1.2 M/1.3 M interface (fraction IV), and the 1.3 M layer (fraction V). Fraction II is enriched 25- to 40-fold in the activity of SM synthase, a Golgi enzyme, but also contains endosomes (Lagace et al., 1999).

β-Octylglucoside (β-OG)/deoxycholate (DOC) insoluble membranes were fractionated and assayed for PI4K activity as described previously (Waugh et al., 2003). In brief, CHO cells were solubilized in 2 ml of 10 mM β-OG/4 mM DOC, 100 mM Na₂CO₃, 10 mM EDTA, and 10 mM EGTA, pH 11.0, by sonication (6 5-s bursts) on ice. The resulting homogenate was mixed with an equal volume of 90% (wt/vol) sucrose/20 mM Tris-HCl, pH 7.4, in a 12-ml centrifuge tube and overlaid with 4 ml of 35% sucrose (wt/vol) and 5% sucrose (wt/vol). The gradient was subject to centrifugation at 175,000 × g for 16 h by using an SW41 Ti rotor (Beckman Coulter), and fractions (1.3 ml) were collected and assayed for PI4K activity as described below.

CHO cells stably expressing shOSBP or a nontargeting control were homogenized in 0.25 M sucrose, 1 mM EDTA, and 10 mM Tris-HCl, pH 7.4, by 15 passages through a 25-gauge needle and centrifuged for 10 min at 2000 × g. The postnuclear supernatant was then fractionated on 6–26% Opti-Prep density gradients (1.059–1.156 g/ml) by centrifugation at 200,000 × g for 3 h in an SW55 Ti rotor (Urano et al., 2008). Fractions were collected by aspiration and assayed for unesterified cholesterol content by the Amplex Red method according to the manufacturer's instructions (Invitrogen). The distribution of cellular organelles in the gradient was determined by immunoblotting for VAP A (ER), NPC1 (late endosomes), TGN38 (TGN), and caveolin (endosomes/plasma membrane).

PI4K activity was assayed as described, with minor changes (Waugh et al., 1998). In brief, PI4K activity in CHO membrane fractions (2–5 μg of protein) was assayed in 25 mM HEPES, pH 7.4, 20 mM MgCl₂, and 2 μg of phosphatidylinositol (from a 2 mg/ml stock in 0.2% Triton X-100) in a final volume of 100 μl. The reaction was initiated by the addition of 40 μM [³²P]γ-ATP (0.25 μCi/nmol), incubated at 37°C for 5 min, and terminated by the addition of 0.5 M HCl. Lipids were extracted, separated by thin layer chromatography (TLC) in a solvent system of CHCl₃:acetone:MeOH:acetic acid:H₂O (40:15:18:12:10, vol/vol), and ³²P-labeled lipids were visualized by autoradiography. Bands corresponding to [³²P]PtdIns(4)P (determined with an authentic standard) were scraped and quantified by scintillation counting.

CHO cells were fixed in 4% (wt/vol) paraformaldehyde for 15 min at 20°C and permeabilized in 0.05% (vol/vol) Triton X-100 for 10 min at 4°C (Perry and Ridgway, 2006). Cells were treated with phosphate-buffered saline (PBS) containing 1% (wt/vol) bovine serum albumin, and incubated with primary and Alexa Fluor-conjugated secondary antibodies as indicated in figure legends. Cells were washed twice with PBS and mounted on microscope slides with Mowiol 4-88 (Calbiochem, San Diego, CA). To disperse the Golgi apparatus, cells were treated with nocodazole (2 μg/ml) for 30 min before the fixation step. Fluorescence images were captured using an LSM510/AxioVert 100M inverted microscope (Carl Zeiss, Thornwood, NY) equipped with 63× and 100× oil immersion objectives (numerical aperture [NA] 1.4).

Cholesterol was visualized by incubating cells with filipin (50 μg/ml in PBS) for 30 min followed by mounting in Mowiol 4-88. Images were captured at identical exposure times using an Axiovert 200M fluorescence microscope (Carl Zeiss) equipped with an axioCam HRm charge-coupled device camera, 4,6-diamidino-2-phenylindole filter package, and 63× oil immersion objective (NA 1.4).

GFP-CERT has a complex pattern of cellular localization but translocates to multiple perinuclear Golgi structures in the presence of 25-hydroxycholesterol. To quantify the change in GFP-CERT localization, images of control and oxysterol-treated CHO cells were captured under identical conditions, and ImageJ software (National Institutes of Health, Bethesda, MD) was used to set pixel thresholds for images of solvent-treated control cells. The change in pixel area above threshold values (normalized to cell number) was measured in oxysterol-activated cells to determine the relative increase in GFP-CERT translocation to perinuclear Golgi structures.

CHO or HeLa cells were transiently transfected with PI4KIIIβ or PIKIIα siRNAs for 48 or 72 h and treated with 25-hydroxycholesterol (2 μg/ml) for 6 h, and then sphingolipids were labeled with [³H]serine (10 μCi/ml) for the final 2 h of the oxysterol treatment. Radiolabeled SM, GlcCer, and ceramide were extracted with CHCl₃:MeOH, separated by TLC, and quantified by scintillation counting (Perry and Ridgway, 2006).

Localization of OSBP, CERT, and FAPP2 to the Golgi apparatus (Lagace et al., 1997; Levine and Munro, 2002; Hanada et al., 2003; Wyles and Ridgway, 2004) involves interaction of their PH domains with PtdIns(4)P synthesized by the Golgi-resident kinases PI4KIIα and/or PI4KIIIβ (Balla et al., 2005; Toth et al., 2006; D'Angelo et al., 2007). Because OSBP promotes sterol-dependent CERT translocation and increased SM synthesis in the Golgi apparatus, we reasoned that this could involve OSBP-dependent activation of a PI4K in the Golgi apparatus. To test this, homogenates of control and 25-hydroxycholesterol-treated CHO cells were fractionated on discontinuous sucrose gradients and analyzed for PI4K expression and activity (Figure 1). This protocol separates Golgi-enriched buoyant membranes (fraction II) from denser ER, mitochondrial, and nuclear membranes (fractions III–V) (Balch et al., 1984). Equivalent amounts (based on volume) of each fraction were resolved by SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotted for PI4KIIα, PI4KIIIβ, OSBP, γadaptin, EEA1, giantin, and CERT (Figure 1A). Giantin, a Golgi resident protein, was primarily detected in fraction II. The γadaptin, a component of the activator protein-1 complex involved in clathrin assembly at the TGN (Bonifacino and Traub, 2003), also was enriched in fraction II but was detected in other fraction as well. Although fraction II was enriched in Golgi markers, it also contained endosomes as indicate by enrichment in EEA1 (Stenmark et al., 1996). PI4KIIα in control CHO homogenates was distributed throughout the gradient but was enriched in fraction II. PI4KIIIβ was detected in fraction II, but the majority was in fractions IV and V. In contrast to OSBP, CERT was enriched in fraction II and relatively absent from dense fractions. Oxysterol treatment did not reproducibly affect the distribution of PI4KIIα, PI4KIIIβ, or CERT in any of the sucrose gradient fractions. Some fractions had minor and inconsistent changes in OSBP and EEA1 distribution in response to 25-hydroxycholesterol treatments. Neither OSBP nor PI4KIIIβ was enriched in fraction II of control or oxysterol-treated cells, even though both displayed prominent Golgi staining (see Figure 9), indicating dissociation from membranes during isolation. Fraction II was not particularly enriched in PI4KIIα or PI4KIIIβ (Figure 1A) but contained >30% of total PI4K activity, indicating it represents a highly active fraction (Figure 1B; Waugh et al., 2003). PI4K activity in fraction II was not affected by treatment with 10 μM wortmannin (data not shown), confirming the relative absence of PI4K type III isoforms (Nakanishi et al., 1995).

We next tested whether PI4K activity in subcellular fractions from CHO cells was responsive to 25-hydroxycholesterol. Total PI4K activity in homogenates (H) of oxysterol-treated CHO cells was reduced by 20%, and the activity in fractions I, III, IV, and V was similar to that of controls (Figure 1C). However, PI4K activity in the Golgi/endosome-enriched fraction II from oxysterol-treated cells has increased by 40% compared with untreated controls. Increased enzyme activity in fraction II of 25-hydroxycholesterol-treated cells was not the result of increased PI4KIIα or PI4KIIIβ protein expression (Figure 1A). The time course for 25-hydroxycholesterol activation of PI4K activity in fraction II is shown in Figure 1D. Treatment of CHO cells for 60 min caused a maximal 75% increase in PI4K activity in fraction II that declined at 90 min.

To confirm that membrane-associated PI4K activity was activated by 25-hydroxycholesterol, CHO cells were extracted with a β-OG/DOC containing buffer, pH 11.0, and fractionated on discontinuous sucrose gradients (Figure 2). This method separates soluble kinases from a buoyant β-OG/DOC–resistant membrane fraction enriched in PI4KIIα (Waugh et al., 2003). Similar to results with A431 cells (Waugh et al., 2003), PI4K activity in detergent extracts of CHO cells was enriched in fraction 4, which correspond to the 5–35% sucrose boundary (Figure 2A). Next, CHO cells were treated with or without 25-hydroxycholesterol (2.5 μg/ml) for 1 h, solubilized in β-OG/DOC buffer, separated on sucrose gradients, and assayed for PI4K activity (Figure 2B). Compared with untreated controls, oxysterols increased the PI4K activity in fraction 4 by 35% and caused variable activation in fractions 5–7 that had less kinase activity. Immunoblotting of sucrose gradient fractions from control and oxysterol-treated cells confirmed that fraction IV was enriched in PI4KIIα and that oxysterol treatment did not affect expression of the kinase (Figure 2C). β-OG/DOC-resistant membranes were enriched in the ER protein VAP but were devoid of PI4KIIIβ. Thus, two independent fractionation methods show that a membrane-associated PI4K, possibly the IIα isoform, was activated in response to 25-hydroxycholesterol.

To test whether 25-hydroxycholesterol activation of PI4K activity required OSBP, CHO cells were transfected with siNT or siOSBP for 48 h, treated with or without 25-hydroxycholesterol, homogenized, fractionated on sucrose gradients, and assayed for PI4K activity (Figure 3). Compared with nontargeting controls, siOSBP transfection reduced expression by >90% but did not affect PI4KIIα or PI4KIIIβ (Figure 3A). A comparison of 25-hydroxycholesterol-dependent activation of PI4K activity in individual fractions revealed that fraction II from siNT-transfected cells displayed the expected 40% increase in PI4K activity in response to 25-hydroxycholesterol (Figure 3B). However, PI4K activity in this fraction from OSBP-depleted cells was resistant to activation by 25-hydroxycholesterol. Fraction I also displayed a slight increase in PI4K activity that was suppressed by knockdown of OSBP.

Palmitoylation of a CCPCC motif in the catalytic region of PI4KIIα mediates cholesterol-dependent activity in the TGN (Barylko et al., 2001; Waugh et al., 2003, 2006; Minogue et al., 2010), suggesting it is a candidate for regulation by OSBP and sterols. We tested whether the PI4K activity in fraction II was due to the IIα isoform by examining sensitivity to manipulation of cholesterol levels using methyl β-cyclodextrin (MβCD) or MβCD complexed with cholesterol as described previously (Waugh et al., 2006). To determine whether altering cholesterol affected Golgi PI4K activity, CHO cells were treated with MβCD for 30 min followed by incubation in lipid-depleted medium with or without cholesterol supplementation (Figure 4A). Fraction II was isolated from these cells, assayed for PI4K activity and compared with control cells that were not pretreated with MβCD but cultured under identical conditions. PI4K activity in fraction II from cholesterol-depleted cells cultured in lipoprotein-deficient serum (LPDS) was slightly reduced. However, culturing cells in media supplemented with MβCD-cholesterol increased PI4K activity by 50% compared with cells that received only LPDS. Similar to previous results using a Golgi/endosomal fraction from A431 cells (Waugh et al., 2006), extraction of cholesterol from fraction II with MβCD caused a dose-dependent inhibition of PI4K activity (Figure 4B). Supplementation of fraction II membranes with MβCD–cholesterol complex increased PI4K activity to a maximum of 1.5-fold at 100 nmol cholesterol/assay (Figure 4C). Interestingly, addition of 60 μM 25-hydroxycholesterol directly to fraction II membranes also stimulated PI4K activity by 40–50% (data not shown).

Cholesterol sensitivity of PI4KIIα suggested it could be associated with a cholesterol-enriched cellular compartment. Consistent with its previously described endosomal/TGN localization (Balla et al., 2002; Salazar et al., 2005; Minogue et al., 2006; Craige et al., 2008), GFP-PI4KIIα was present in perinuclear vesicles in control and 25-hydroxycholesterol–treated cells (Figure 5A). Filipin staining of cholesterol in the endosomes/lysosomes was also similar in control and 25-hydroxycholesterol–treated CHO cells but did not overlap with GFP-PI4KIIα. We also observed that filipin distribution was unaffected by OSBP depletion with siRNA (data not shown), suggesting it is not a good probe for assessing the relationship between cholesterol distribution and PI4KIIα. We also measured whether activation by 25-hydroxycholesterol and OSBP affected the cholesterol content of fraction II. However, cholesterol mass in fraction II isolated from control and OSBP-depleted cells treated with or without 25-hydroxycholesterol was similar (data not shown). Although the sucrose density gradient method is useful for quantifying changes in PI4K activity, it does not have the resolution to identify subtle changes in the cholesterol content of Golgi and endosomal membranes. For this purpose, cholesterol and PI4KIIα distribution in control and OSBP-depleted CHO cells were determined by fractionation of postnuclear supernatants on Opti-Prep gradients (Figure 5, B and C). CHO cells were stably depleted of OSBP by expression of a lentiviral short hairpin RNA (shRNA) that afforded >95% knockdown of OSBP protein compared with nontargeting control (shNT) (Figure 5C). PI4KIIα in shOSBP and ShNT cells was distributed primarily to fractions containing the endosomal marker NPC1 and partially overlapped with TGN38 (Figure 5C). Control shNT CHO cells have a cholesterol peak in fractions 4–6 that comigrated with NPC1and PI4KIIα (Figure 5B). In OSBP-depleted cells, there was a significant shift in cholesterol distribution from fractions 4–6 to fractions 1 and 2 that contained TGN38 and caveolin. Although cholesterol distribution was significantly altered by OSBP silencing, the distribution of PI4KIIα was unaffected. PI4KIIα localization to a TGN/endosomal compartment that is subject to cholesterol depletion when OSBP expression is reduced further supports the conclusion that this kinase is regulated by a sterol- and OSBP-dependent mechanism.

OSBP coordinates cholesterol transport and/or signaling with SM synthesis by activation of CERT activity between the ER and Golgi apparatus (Perry and Ridgway, 2006). PI4KIIα is a good candidate for mediating CERT recruitment to the Golgi apparatus, but other PI4Ks could be involved. Indeed, PI4KIIIβ is abundant in the Golgi apparatus and has a dominant, although apparently not exclusive role in CERT-dependent ceramide transport to the Golgi apparatus under unstimulated conditions (Toth et al., 2006; D'Angelo et al., 2007). To test whether PI4KIIα or PI4KIIIβ was involved in oxysterol regulation of SM synthesis, their expression was silenced in CHO cells with siRNAs followed by metabolic labeling of sphingolipids with [³H]serine in the presence and absence of 25-hydroxycholesterol (Figure 6A). Treatment of siNT-transfected cells with 25-hydroxycholesterol resulted in a 2.5-fold increase in [³H]serine incorporation into SM without affecting [³H]ceramide and [³H]GlcCer levels. CHO cells in which PI4KIIα expression was reduced by >80% had basal [³H]SM synthesis that was similar to siNT controls; however, activation of [³H]SM synthesis by oxysterol was blocked. Silencing of PI4KIIα also reduced synthesis of [³H]GlcCer in control and oxysterol-treated CHO cells, an effect that was not observed after OSBP and CERT depletion (Perry and Ridgway, 2006). In contrast, silencing of PI4KIIIβ did not affect basal or 25-hydroxycholesterol-activated SM synthesis. To confirm these results, oxysterol-activated SM synthesis was quantified in HeLa cells using two different siRNAs directed against PI4KIIα and PI4KIIIβ (Figure 6B). 25-Hydroxycholesterol activation of SM synthesis in HeLa cells is suppressed by OSBP silencing. Moreover, knockdown of PI4KIIα with two different siRNAs suppressed oxysterol-activated SM synthesis by ∼50%, but silencing of PI4KIIIIβ had no affect.

Increased SM synthesis in oxysterol-treated cells is accompanied by Golgi translocation of CERT and delivery of a fluorescent-labeled ceramide analogue from the ER (Perry and Ridgway, 2006). Thus, we tested whether CERT recruitment to the Golgi was also dependent on expression of PI4KIIα. GFP-CERT was stably expressed in CHO cells, and localization to the Golgi apparatus was visualized after depletion of PI4KIIα or PI4KIIIβ by RNA interference (RNAi) and treatment with 25-hydroxycholesterol (Figure 7). siRNAs were effective in knocking down PI4KIIα and IIIβ protein expression by >90% in CHO cells expressing GFP-CERT (Figure 7A). GFP-CERT was in a diffuse and perinuclear compartment in solvent-treated siNT cells but displayed strong Golgi localization in cells exposed to 25-hydroxycholesterol (Figure 7B). Depletion of PI4KIIα did not affect the distribution of CERT in untreated control cells; however, the addition of 25-hydroxycholesterol failed to promote translocation to the Golgi apparatus (Figure 7B). In PI4KIIIβ-depleted cells, the distribution of GFP-CERT was similar to untreated siNT controls and addition of 25-hydroxycholesterol increased the translocation of GFP-CERT to the Golgi apparatus. However, CERT-GFP was slightly dispersed in PI4KIIIβ knockdown cells compared with siNT controls, indicating fragmentation of the Golgi stacks. The quantification of GFP-CERT localization confirmed that siNT- and siPI4KIIIβ-transfected cells treated with 25-hydroxycholesterol had increased perinuclear Golgi staining compared with PI4KIIα knockdown cells (Figure 7C). These result show that, consistent with effects on SM synthesis, PI4KIIα mediates 25-hydroxycholesterol–dependent recruitment of CERT to the Golgi apparatus without affecting its distribution under basal conditions.

Results shown in Figures 6 and 7 indicated that OSBP is an upstream regulator of PI4KIIα. However, PI4KIIIβ and PI4KIIα have been implicated in recruitment of the GFP-PH domain of OSBP to the Golgi apparatus (Wang et al., 2003; Balla et al., 2005), implying that PtdIns(4)P production also could initiate or enhance OSBP interaction with the Golgi apparatus. Whereas previous studies used the isolated OSBP PH domain as a PtdIns(4)P reporter, siRNA silencing was used here to establish the role of PI4KIIα and PI4KIIIβ in 25-hydroxycholesterol–dependent recruitment of endogenous OSBP to the Golgi apparatus. First, the localization of OSBP, PI4KIIIβ, PI4KIIα, and CERT was identified after activation with 25-hydroxycholesterol and fragmentation of the Golgi apparatus into mini-stacks with nocodazole (Godi et al., 2004; Figure 8). In CHO cells (Figure 8A), OSBP did not colocalize with giantin in the cis/medial-Golgi but strongly colocalized with PI4KIIIβ in the cis/medial- (Weixel et al., 2005) or trans-Golgi/TGN (Hausser et al., 2005). OSBP also partially localized with endogenous PI4KIIα in the endosomes/TGN (Figure 9; Salazar et al., 2005; Minogue et al., 2006; Craige et al., 2008), and strongly costaining with GFP-CERT, which is also reported in the TGN (Fugmann et al., 2007). Because antibodies to TGN resident proteins reacted poorly with CHO cells, we confirmed the localization of OSBP and the PI4-kinases with TGN46 in HeLa cells (Figure 8B). Consistent with the implied TGN localization in CHO cells, OSBP and PI4KIIIβ strongly colocalized with TGN46. Perinuclear PI4KIIα also costained with TGN46, but there was still a portion that did not and, based on cell fractionation shown in Figure 5C, probably corresponds to the endosomes.

Next, RNAi was used to determine whether PI4KIIα or PI4KIIIβ was required for translocation of OSBP to the Golgi apparatus in response to 25-hydroxycholesterol (Figure 9). In siNT-transfected CHO cells, OSBP was diffusely localized to the cytoplasm and perinuclear structures but shifted to the Golgi apparatus after addition of 25-hydroxycholesterol for 30 min (Figure 9, A and C). Endogenous PI4KIIα in siNT-transfected cells was distributed in punctate structures and in the perinuclear region, a pattern that was not affected by oxysterol (Figure 9A). OSBP in CHO cells with reduced PI4KIIα expression was localized to the perinuclear Golgi apparatus in the presence of 25-hydroxycholesterol (Figure 9B), indicating that OSBP does not depend on PtdIns(4)P generated by this PI4K for its translocation.

The localization of OSBP was examined in CHO cells depleted of PI4KIIIβ, which has been implicated in Golgi localization of proteins with PtdIns(4)P-specific PH domains (Balla et al., 2005; Toth et al., 2006; Figure 9, C and D). 25-Hydroxycholesterol did not affect the constitutive Golgi localization of PI4KIIIβ (Figure 9C). OSBP translocated to the Golgi apparatus in PI4KIIIβ-depleted cells but tended to be more dispersed than controls, suggesting fragmentation of the Golgi stacks or partial suppression of OSBP translocation (Figure 9D). Similar to results with CHO cells, we observed that suppression of PI4KIIα did not affect the prominent Golgi localization of OSBP in HeLa cells treated with 25-hydroxycholesterol (Figure 9E). OSBP also was localized to the Golgi apparatus in HeLa cells with reduced PI4KIIIβ expression, but again the distribution was slightly fragmented compared with nontargeting controls (Figure 9F). Although we cannot discount other sources of PtdIns(4)P, the lack of involvement of the major Golgi-associated PI4Ks in OSBP recruitment to the Golgi apparatus confirms it is an upstream regulator of PI4KIIα activity.