Stabilization of Phosphatidylinositol 4-Kinase Type IIβ by Interaction with Hsp90*

Mammalian cells express two isoforms of type II phosphatidylinositol 4-kinase: PI4KIIα and PI4KIIβ. PI4KIIα exists almost exclusively as a constitutively active integral membrane protein because of its palmitoylation (Barylko, B., Gerber, S. H., Binns, D. D., Grichine, N., Khvotchev, M., Südhof, T. C., and Albanesi, J. P. (2001) J. Biol. Chem. 276, 7705–7708). In contrast, PI4KIIβ is distributed almost evenly between membranes and cytosol. Whereas the palmitoylated membrane-bound pool is catalytically active, the cytosolic kinase is inactive (Wei, Y. J., Sun, H. Q., Yamamoto, M., Wlodarski, P., Kunii, K., Martinez, M., Barylko, B., Albanesi, J. P., and Yin, H. L. (2002) J. Biol. Chem. 277, 46586–46593; Jung, G., Wang, J., Wlodarski, P., Barylko, B., Binns, D. D., Shu, H., Yin, H. L., and Albanesi, J. P. (2008) Biochem. J. 409, 501–509). In this study, we identify the molecular chaperone Hsp90 as a binding partner of PI4KIIβ, but not of PI4KIIα. Geldanamycin (GA), a specific Hsp90 inhibitor, disrupts the Hsp90-PI4KIIβ interaction and destabilizes PI4KIIβ, reducing its half-life by 40% and increasing its susceptibility to ubiquitylation and proteasomal degradation. Cytosolic PI4KIIβ is much more sensitive to GA treatment than is the integrally membrane-associated species. Exposure to GA induces a partial redistribution of PI4KIIβ from the cytosol to membranes and, with brief GA treatments, a corresponding increase in cellular phosphatidylinositol 4-kinase activity. Stimuli such as PDGF receptor activation that also induce recruitment of the kinase to membranes disrupt the Hsp90-PI4KIIβ interaction to a similar extent as GA treatment. These results support a model wherein Hsp90 interacts predominantly with the cytosolic, inactive pool of PI4KIIβ, shielding it from proteolytic degradation but also sequestering it to the cytosol until an extracellular stimulus triggers its translocation to the Golgi or plasma membrane and subsequent activation.

canonical phosphoinositide biosynthetic pathway by phosphorylating the D-4 hydroxyl of the inositol head group of PtdIns. The product of this reaction, PtdIns 4-phosphate (PtdIns4P), 2 serves not only as a major precursor in the synthesis of more highly phosphorylated phosphoinositides, including PtdIns 4,5-bisphosphate and PtdIns 3,4,5-trisphosphate, but also has itself been shown to be a regulator of membrane trafficking (7,8). Two types of PI4K (PI4KII and PI4KIII) have been identified in eukaryotes. The type III kinases, which are expressed from yeast to mammals, are further subdivided into ␣ (ϳ230 kDa) and ␤ (ϳ100 kDa) isoforms. Although mammals also express two type II kinases (␣ and ␤, both ϳ55 kDa), Saccharomyces cerevisiae have only one PI4KII ortholog, known as Lsb6p (9,10).
PI4KII␣ and ␤ have conserved catalytic domains but diverse N-terminal regions extending approximately from residues 1 to 90. Unlike PI4KIIIs, which are almost entirely cytosolic, PI4KIIs can associate integrally with membranes by virtue of palmitoylation of multiple cysteines within their catalytic domains (11,12). The palmitoylation motif, CCPCC, is present in both PI4KII␣ and ␤. However, whereas Ͼ90% of PI4KII␣ is palmitoylated and exists in cells as an active, integrally membranebound species, PI4KII␤ is divided almost evenly between cytosolic and membrane-bound pools (13,14). Moreover, almost half of membrane-bound PI4KII␤ is only peripherally associated with membranes, extractable by sodium carbonate at pH 11 in the absence of detergent. Therefore, it appears that only ϳ25-30% of PI4KII␤ is normally palmitoylated in cells. Because palmitoylation is essential for catalytic activity (14), 70 -75% of this isoform may be inactive under resting conditions. Although a small portion (ϳ7%) of PI4KII␤ is recruited to membranes in response to growth factor receptor activation (13), conditions have not yet been found that result in a major redistribution of the kinase from cytosol to membranes. We reported that the different membrane binding properties and palmitoylation states of PI4KII␣ and ␤ are not due to their highly diverse N-terminal regions but instead to relatively slight differences in their C-terminal 160 residues (14).
PI4KII␣ has been implicated in generating PtdIns4P pools that regulate membrane trafficking from the trans-Golgi network (15)(16)(17) and in late stages of endocytosis (17). Although no specific function has been ascribed to PI4KII␤, its partial redistribution to the plasma membrane in response to growth factor receptor activation (13) and tyrosine phosphorylation in response to activation of the T-cell receptor (18) suggest that it may have a role in cellular signaling. To understand how extracellular stimuli recruit PI4KII␤ to membranes, where it can be palmitoylated and activated, we sought to understand the basis for its distribution between membranes and cytosol. To this end, we attempted to identify binding partners that (a) selectively bind to PI4KII␤ over PI4KII␣ and (b) could sequester PI4KII␤ to the cytosol. The data presented below demonstrate that the molecular chaperone Hsp90 fulfills these criteria. They further demonstrate that the interaction with Hsp90 is required to stabilize the cytosolic pool of PI4KII␤ and that the PI4KII␤-Hsp90 interaction is disrupted by growth factor receptor activation, resulting in a partial redistribution of the kinase to membranes.

EXPERIMENTAL PROCEDURES
Materials and Antibodies-Anti-PI4KII␤ and anti-PI4KII␣ antibodies were generated as described previously (13,19). Monoclonal anti-Myc antibody 9E10 was obtained from the National Cell Culture Center (Minneapolis, MN). Hsp90 antibody (H-114) was from Santa Cruz Biotechnology (Santa Cruz, CA). Hsp70 antibody was purchased from Cell Signaling (Danvers, MA). Primers were obtained from Integrated DNA Technologies (Coralville, IA). Cloning reagents and reagents for mutagenesis were from Invitrogen. Triton X-100 and reagents for electrophoresis and immunoblotting were from Bio-Rad. Geldanamycin (GA) was obtained from InvivoGen (San Diego, CA). Other reagents, including ATP, buffers, protease inhibitors, and MG132 (proteasome inhibitor), were from Sigma.
Cell Culture and Transfection-HEK293, COS, and HeLa cells were maintained in DMEM supplemented with 10% FBS and antibiotics. The cells were replated 1 day prior to transfection. For transient expression of proteins, the cells were transfected for 20 h using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
In-gel Digestion and Protein Identification Using Mass Spectrometry-Silver-stained protein bands were excised from one-dimensional SDS gels and cut further into small cubes (ϳ1 ϫ 1 mm). After in-gel reduction, alkylation, and destaining, the proteins were digested overnight with trypsin (12 ng/l). The tryptic peptides were extracted from gels dried in a vacuum centrifuge and analyzed by electrospray liquid chromatography-tandem mass spectrometry using a nanoscale C18 column coupled in-line with an ion trap mass spectrometer (LCQ Deca, Thermo Finnigan). The instrument was set in a data-dependent acquisition mode, cycling between one full MS scan and Myctagged proteins were expressed in HEK 293 cells, which were then lysed and subjected to immunoprecipitation with anti-Myc antibodies. The 90-kDa band was extracted and identified as Hsp90 by mass spectrometry. The asterisk designates phospho-PI4KII␤, and the double asterisks designate peroxiredoxin I, another PI4KII␤-binding protein identified in this study. Other unique bands in the precipitates were identified by mass spectrometry as fragments of the full-length kinases. B and C, specific association between endogenous and exogenous Hsp90s and PI4KII␤. Myc-PI4KII␣ or Myc-PI4KII␤ was expressed in HeLa cells without (B) or with (C) FLAG-Hsp90␣. The cell lysates were immunoprecipitated (IP) with anti-Myc antibody and blotted with anti-Hsp90 or anti-FLAG to show the specific association of Hsp90 with PI4KII␤ but not PI4KII␣. D, specific association between endogenous Hsp90 and endogenous PI4KII␤. HeLa cell extracts prepared as above were used to immunoprecipitate endogenous PI4KII␣ or PI4KII␤ with their corresponding antibodies. The upper panel shows an immunoblot of the electrophoresed precipitates with anti-Hsp90 antibody. The lower panels show depletion of the kinases from supernatants following immunoprecipitation. Because of co-migration of antibody heavy chains with the kinases on SDS gels, the amounts of precipitated PI4KII␣ and PI4KII␤ could not be determined. The cross-reactive but not precipitated lower bands that appear on occasion (see also   . At the indicated time points, the cells were lysed, and the lysates were electrophoresed and blotted with anti-Myc antibodies to detect kinase and with anti-actin antibodies for normalization. The graph shows the amount of PI4KII␤ remaining after cycloheximide treatment as a percentage of its initial amount, normalized to the amount of actin in each lane. The data represent the averages of two experiments, one measured in duplicate and the other measured in triplicate. B, pulse-chase method for determining the effect of GA on the half-life of PI4KII␤. The cells expressing Myc-PI4KII␤ were treated with or without 2 M GA for 2 h, starved with methionine/cysteine-free DMEM for 1 h, labeled with [ 35 S]methionine/cysteine for 1 h, then chased with unlabeled methionine/ cysteine for 2, 4, or 6 h, and lysed. PI4KII␤ was immunoprecipitated with anti-Myc antibody, electrophoresed, and analyzed by fluorography. The graph displays quantification of results of a single pulse-chase experiment. Na 2 VO 4 , and 0.1% Nonidet P-40. Cell lysates were obtained after centrifugation at 1000 ϫ g for 5 min and were precleared of nonspecifically binding proteins by incubation with rec-Protein G-Sepharose 4B conjugate (Zymed Laboratories Inc.) for 30 min. Myc-tagged proteins were immunoprecipitated by 4 h of incubation with anti-Myc antibody that had been crosslinked to protein G-Sepharose using dimethyl pimelimidate (Pierce). The immunoprecipitates were washed three times with Nonidet P-40 lysis buffer. The samples were eluted with SDS sample buffer and loaded on SDS-polyacrylamide gels. To quantify PI4KII levels in various fractions, the samples were electrophoresed and immunoblotted with anti-Myc, anti-PI4KII␣, or anti-PI4KII␤ antibodies. To estimate relative amounts of proteins, the immunoblots were incubated with 125 I-labeled secondary antibody, scanned by phosphorimaging (Fujifilm BAS 1500), and quantified using the Multi-Gauge V2.3 program (Fuji Photo Film Co.). Alternatively, immunoblotted bands were quantified by densitometry using the Multi-Gauge V2.3 program (Fuji Photo Film Co.).
Preparation of Cytosol and Membranes-Transfected or untransfected cells were washed with PBS and scraped in a solution containing 0.25 M sucrose, 20 mM Tris-HCl (pH 7.5), 0.1 M NaCl, 1 mM EDTA, 1 mM PMSF, a protease inhibitor mixture (10 g/ml each of N ␣ -p-tosyl-L-lysine chloromethyl ester, N ␣ -p-tosyl-L-arginine methyl ester, N ␣ -p-tosyl-L-lysine chloromethyl ketone, leupeptin, and pepstatin A) and phosphatase inhibitors (50 mM NaF, 50 mM Na 4 P 2 O 3 , 1 mM Na 2 VO 4 ). The cells were lysed by two freeze-thaw cycles and passage through a 27.5-gauge needle. The lysates were then centrifuged at 1,000 ϫ g for 5 min to obtain post-nuclear supernatants. The post-nuclear supernatant were then centrifuged at 200,000 ϫ g for 15 min to separate cytosol from membranes. The resulting membrane pellets were homogenized sequentially, first in 0.1 M Na 2 CO 3 (pH 11) to extract tightly bound peripheral proteins and then in a solution containing 1% Triton X-100 (v/v), 20 mM Tris-HCl (pH 7.5), 1 mM EDTA to extract integral proteins.

Analysis of [ 32 P]P i Incorporation into
Lipids-For radiolabeling, transfected COS cells were incubated in phosphate-free medium containing 5% dialyzed FBS and 25 Ci/ml [ 32 P]P i (PerkinElmer Life Sciences) for 4 h. After three brief washes with PBS, labeled cells were scraped into tubes containing methanol:HCl (10:1), and the lipids were extracted with chloroform. The organic phase was collected and washed with an equal volume of methanol and HCl (1:1). Aliquots were spotted onto TLC plates and separated in a solvent system consisting of n-propyl alcohol/H 2 O/NH 4 OH (65:20:15). Radioactive PtdIns4P spots were scanned using FLA-5100 (Fuji Photo Film Co.), and radioactivity of bands was quantified using the Multi-Gauge V2.3 program.
Geldanamycin Treatment of Cells-GA was solubilized in Me 2 SO and added to cell culture medium to the final concentrations indicated in the text and figure legends. Equal volumes of Me 2 SO without GA were used in controls. The cells were treated for 1-24 h as indicated.
Immunofluorescence Microscopy-To visualize the redistribution of endogenous PI4KII␤, HeLa cells were grown on coverslips for 16 h. After treatment with 2 M GA for 2 h, the cells were washed three times with ice-cold PBS, fixed with 3.7% formaldehyde in PBS for 10 min at room temperature, and permeabilized with 0.2% Triton X-100 in PBS for 5 min at room temperature. After two washes with PBS and blocking in solution containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Tween 20, and 1% BSA, the cells were incubated with anti-PI4KII␤ antibody followed by rhodamine-labeled secondary antibody (Molecular Probes). Fluorescence microscopy was performed with a Zeiss LSM 510 confocal microscope with 63ϫ oil immersion lens.
Half-life Measurement-With the cycloheximide method, the cells expressing Myc-PI4KII␤ were treated for various times with 0.2 mg/ml cycloheximide, in the presence or absence of 2 M GA. Lysates prepared from these cells were electrophoresed and blotted with anti-Myc antibodies to detect kinase and with anti-actin antibodies for normalization. Relative amounts of kinase and actin were estimated using 125 I-labeled secondary antibodies. With the pulse-chase method, COS cells were transfected with Myc-PI4KII␤, and after 20 h they were treated with or without 2 M GA for 2 h. After starving with methionine/cysteine-free DMEM for 1 h, the cells were labeled with [ 35 S]methionine/cystine for 1 h and then chased with unlabeled methionine/cysteine for 2, 4, or 6 h. The cells were lysed and immunoprecipitated with anti-Myc antibody. The precipitated samples were separated by 10% SDS-PAGE and analyzed by fluorography. The results were quantified using ImageQuant software.
Other Procedures-SDS-polyacrylamide gel electrophoresis and immunoblot analysis were carried out by the methods of Laemmli (20) and Towbin et al. (21), respectively. Protein concentrations were determined using the modified Lowry method (22) according to Peterson (23) with BSA as a standard.

Identification of Hsp90 as a Binding Partner of PI4KII␤-To
understand the basis for the distinct subcellular distributions of PI4KII␣ and ␤, we sought to identify proteins that preferentially interact with one isoform over another. Myc-tagged versions of each kinase were expressed in HEK293 cells and immunoprecipitated using anti-Myc antibodies. One electrophoretic band of 90 kDa was consistently enriched in immunoprecipitates of PI4KII␤ compared with those of PI4KII␣ or of an irrelevant expressed protein, GFP (Fig. 1A). This band was extracted from gels and subsequently identified as Hsp90 by mass spectrometry. Apart from several other bands that proved to be fragments of PI4KII␤, the only other band that selectively co-immunoprecipitated with PI4KII␤ was identified by mass spectrometry as peroxiredoxin I (Fig. 1A, **). The potential significance of this interaction is currently being investigated.
The selective binding of Hsp90 to PI4KII␤ was confirmed by immunoprecipitating PI4KII␣ or PI4KII␤ and blotting the electrophoresed precipitates with an anti-Hsp90 antibody (Fig. 1, B  and D). The results show that Hsp90 co-immunoprecipitates with Myc-PI4KII␤ (Fig. 1B) as well as with the endogenous kinase ( Fig. 1D) but not with PI4KII␣. Selectivity was also demonstrated by co-precipitation of expressed FLAG-tagged Hsp90 with Myc-PI4KII␤ but not with Myc-PI4KII␣ (Fig. 1C).
Inhibition of the PI4KII␤-Hsp90 Interaction by Geldanamycin Results in Destabilization of PI4KII␤-To determine the significance of Hsp90 binding to PI4KII␤, we disrupted the interaction using GA, an agent that binds directly to the ATPbinding site of Hsp90 and blocks its association with many client proteins (24,25). Treatment of cells with 2 M GA for 2 h significantly reduced the amount of Hsp90 that co-immunoprecipitates with Myc-PI4KII␤, whereas similar amounts of Hsp70 were present in immunoprecipitates from GA-treated and untreated cells ( Fig. 2A). Hsp70 serves as general chaperone for newly synthesized proteins engaged in co-translational folding and for unfolded proteins. It is also a component of the Hsp90 chaperone complex, participating in the Hsp90 assisted folding cycle (reviewed in Refs. 26 and 27), but GA apparently does not disrupt the association of Hsp70 and its substrates (28).
In many cases, interfering with the interaction of Hsp90 with its client proteins reduces the stability of those clients. To test whether this also applies to PI4KII␤, we analyzed the effect of GA treatment on the cellular levels of PI4KII␤ in a dose-dependent (Fig. 2B) and time-dependent manner (Fig. 2C). After cell exposure to GA, we observed reduction of cellular levels of endogenous PI4KII␤ but not of endogenous PI4KII␣. PI4KII␤ levels were reduced approximately by half following 4 h of treatment with 1 M GA (Fig. 2B). Reduction of PI4KII␤ was evident even after a 1-h exposure to 2 M GA (Fig. 2C). We also tested the sensitivity of PI4KIII␤ to GA, because Flanagan and Thorner (29) reported that Hsp90 co-purifies with the yeast PI4KIII␤ ortholog, Pik1p. Indeed, based on this report, PI4KIII␤ has been classified in the literature as a binding protein of Hsp90 (30, 31), although the function of this interaction was not investigated. Our results indicate that PI4KIII␤ is resistant to GA treatment (Figs. 2B and 4A).
Consistent with the above results, GA was found to shorten the half-life of PI4KII␤ (Fig. 3A). HeLa cells expressing Myc-PI4KII␤ were treated with the protein synthesis inhibitor cycloheximide in the presence or absence of 2 M GA. At various times thereafter, the kinase levels were estimated by immunoblotting. The half-life of PI4KII␤ was reduced from 4 to 1.5 h by GA treatment. We verified that GA also reduced the half-life of Myc-PI4KII␤, from 3 to 1.5 h, using the 35 S-labeling pulsechase procedure (Fig. 3B). This is the first reported measurement of the half-life of a type II kinase, and it demonstrates that, like many other signaling proteins, it turns over relatively rapidly.
Hsp90 Protects PI4KII␤ from Degradation by the Proteasome-GA-mediated reductions in the cellular levels of Hsp90 client proteins are likely due to their enhanced ubiquitylation and proteasomal degradation (32)(33)(34)(35). To check whether Hsp90 protects PI4KII␤ from proteasomal degradation, we examined the effect of treating cells expressing Myc-PI4KII␤ with both GA and the proteasome inhibitor, MG132. In the absence of these agents, the majority of Myc-PI4KII␤ was present in the soluble fraction of cells homogenized in buffer containing 1% Nonidet P-40. The remainder, presumably including aggregated kinase, was recovered in low speed pellets (Fig. 4A). GA treatment for 24 h reduced the amount of PI4KII␤ recovered either in the soluble or insoluble fractions. However, nearly the entire initial pool of kinase distributed to the pellet when cells were simultaneously treated with GA and MG132, indicating that PI4KII␤ is protected from proteasomal degradation but accumulates in an aggregated state. In contrast, the amounts of PI4KIII␤ (or of Hsp90) were not influenced by either GA or MG132. It was previously reported that PDK1, Raf1, and Src also accumulate in the insoluble fraction upon treatment of cells with both GA and MG132 (34,36,37).
We next verified that dissociation of the PI4KII␤-Hsp90 complex by GA enhanced ubiquitylation of the kinase. The cells were co-transfected with HA-ubiquitin and Myc-PI4KII␤ or Myc-PI4KII␣ to allow detection of ubiquitylated kinase with anti-HA antibodies. In the absence of GA, inhibition of the proteasome with MG132 did not appreciably increase the amount of ubiquitylated kinases detected in anti-Myc immunoprecipitates (Fig. 4B). However, co-treatment with GA for 1.5 h resulted in a significant accumulation of ubiquitylated PI4KII␤, whereas the ubiquitylation of PI4KII␣ was unaffected by either MG132 alone or in combination with GA.
Hsp90 Selectively Stabilizes the Cytosolic Pool of PI4KII␤-To explain why Hsp90 binds preferentially to PI4KII␤ over PI4KII␣, we analyzed the interactions using a series of truncated and hybrid kinases. Deletion of the highly diverse N-terminal segments (ϳresidues 1-90, constructs designated as Cat ␣ and Cat ␤) had essentially no effect on Hsp90 binding or GA sensitivity (Figs. 5 and 6B). Therefore, as shown for a number of protein kinases (38), Hsp90 apparently interacts with the cata- lytic domain (residues 91-481) of PI4KII␤. We next asked whether a particular region of the PI4KII␤ catalytic domain determines its GA sensitivity. The catalytic domains of the type II kinases can be divided into two segments: an N-terminal segment (residues ϳ90 -314) having 86% sequence similarity and a C-terminal segment (residues ϳ315-481) having 77% sequence similarity. We generated two hybrids, Cat ␣/␤ (residues 92-314 of PI4KII␣ fused to residues 312-481 of PI4KII␤) and Cat ␤/␣ (residues 91-311 of PI4KII␤ fused to residues 315-478 of PI4KII␣). Cat ␣/␤, like full-length PI4KII␤, bound to Hsp90 and was destabilized by GA. Cat ␤/␣, like PI4KII␣, dis-played neither property (Fig. 5). Interestingly, Cat ␣/␤ and Cat ␤/␣ behave similarly to PI4KII␤ and PI4KII␣ also in terms of their solubilities: Cat ␣/␤ is Ͼ80% cytosolic, whereas Cat ␤/␣ is Ͼ80% integrally membrane-bound (14). The above results raise two possibilities: (a) the C-terminal segment of PI4KII␤ contains a binding site for Hsp90 that is absent from the corresponding region of PI4KII␣ and (b) Hsp90 binds preferentially to PI4KII␤ because a substantial portion of PI4KII␤ is cytosolic because of its C-terminal segment. To address the first possibility, we analyzed a series of truncated mutants of PI4KII␤ lacking the C-terminal 47, 74, or 141 residues. All three of these FIGURE 7. Redistribution of Myc-PI4KII␤ to membranes in response to GA treatment. A, effect of GA on the distribution of PI4KII␤ among cytosolic, peripheral membrane, and integral membrane pools. Cytosolic and membrane fractions were recovered from the supernatants and pellets, respectively, following centrifugation of post-nuclear supernatant (PNS) at 200,000 ϫ g for 15 min. Peripheral membrane proteins were those released from membranes upon treatment with 0.1 M Na 2 CO 3 (pH 11), and integral membrane proteins were those that were then solubilized by 1% Triton X-100. Cells transfected with Myc-PI4KII␤ were treated with 2 M GA for the designated times and fractionated as described under "Experimental Procedures." The lysates were immunoblotted to estimate the relative amounts of PI4KII␤ in each fraction, as shown in the graph. Hsp90, PI4KIII␤, and caveolin were blotted as loading controls. The graph represents the means Ϯ S.D. from three independent experiments. B, PtdIns4P production in cells treated briefly with GA. Cells transfected with vector (mock) or Myc-PI4KII␤ were labeled with [ 32 P]P i for 4 h and incubated with GA for 15 min. The cells were then scraped into methanol:HCl (10:1) to extract lipids, which were then quantified as described under "Experimental Procedures." Phosphorylated lipids were resolved by TLC. The graph shows the relative amounts of radioactive phosphate incorporated into PtdIns4P. The averages of three independent triplicate experiments are shown. The error bars designate the standard deviations. mutants retained their ability to bind to Hsp90 and were sensitive to GA (Fig. 5). In addition, deletion of the region that is most diverse between the catalytic domains of PI4KII␣ and ␤, residues 312-336, had no effect on Hsp90 binding or GA sensitivity. These results support the second possibility that PI4KII␤ requires stabilization by Hsp90 because of its weaker binding to membranes than PI4KII␣, a property that is conferred by its C-terminal 160 residues.
This conclusion was further supported by analysis of a truncated form of PI4KII␣, termed Cat ␣ 91-431 , which lacks its C-terminal 47 residues and consequently distributes predominantly to the cytosol (Fig. 6A). This mutant interacted with Hsp90 (Fig. 6B) and was susceptible to proteolytic degradation in the presence of GA (Fig. 6C), demonstrating that PI4KII␣ contains an Hsp90-binding site that is normally occluded because of its tight association with membranes. Presumably, this tight membrane association also protects PI4KII␣ from proteolytic degradation, although palmitoylation is apparently not required for stabilization, because an unpalmitoylated deletion mutant, PI4KII␣-⌬CCPCC, was not proteolytically degraded upon GA treatment (Fig. 5).
In summary, both PI4KII␣ and ␤ contain an Hsp90 interaction site, which most likely resides in the N-terminal lobe of the catalytic domains. In this respect, PI4KIIs are similar to other Hsp90-binding protein kinases, wherein the N-lobe cores have been shown to be essential for association with Hsp90 (38). PI4KII␣ apparently does not require stabilization by Hsp90 binding because it associates strongly with membranes, even in the absence of palmitoylation. In contrast, PI4KII␤ must bind to Hsp90 and is highly sensitive to its release, perhaps because a substantial portion of this isoform is cytosolic. Apparently, the inactive, cytosolic pool of PI4KII␤ is very unstable unless it associates with its chaperone.
Dissociation of PI4KII␤ from Hsp90 by Exposure to GA Results in Transient Translocation to Membranes and Increased Kinase Activity-Because Hsp90 binds preferentially to soluble PI4KII molecules, it is also possible that Hsp90 sequesters PI4KII␤ to the cytosol. If so, GA treatment may promote redistribution of PI4KII␤ from cytosol to membranes. To detect this redistribution, it was necessary to subject cells to GA treatments shorter than 4 h, before the competing processes of ubiquitylation and proteolysis obscured the analysis. Cells expressing Myc-PI4KII␤ were treated with 2 M GA for 0, 0.25, 1, or 3 h prior to fractionation into cytosolic, peripheral membrane, and integral membrane pools. The longer incubations (1 and 3 h) caused a decrease in PI4KII␤ levels from all three pools. However, the integral membrane-bound pool was much more stable, declining by only ϳ20% after a 3-h GA treatment versus 60 -70% for the cytosolic and peripherally membrane-bound pools (Fig. 7A). This result further supports the view that tight membrane binding stabilizes the type II kinases despite the absence of Hsp90 binding. Exposure of cells to GA for only 15 min was sufficient to disrupt the Hsp90-PI4KII␤ interaction but was too brief to elicit a detectable loss of PI4KII␤. Interestingly, this short GA treatment induces a partial translocation of the kinase from cytosol to membranes, increasing its proportion in peripheral and integral pools by 1.2-and 1.5-fold, respectively (Fig. 7A). The increase in the amount of integral PI4KII␤ observed in GA-treated cells is particularly important, because only this species is palmitoylated and catalytically active (14). Indeed, 15-min GA treatment resulted in a 1.5-fold elevation in [ 32 P]P i incorporation into cellular PtdIns4P (Fig.  7B), corresponding closely to the increase in PI4KII␤ that is integrally associated with membranes.
To show that endogenous PI4KII␤ also responds to GA treatment by relocalizing to membranes, we electrophoresed membrane and cytosol fractions from COS cells that were treated or untreated for 2 h with 2 M GA and blotted the gels with anti-PI4KII␤ antibodies. The endogenous kinase was nearly entirely depleted from cytosol of GA-treated cells (Fig. 8A). The GA-dependent redistribution of endogenous PI4KII␤ to membranes was confirmed by immunofluorescence microscopy (Fig. 8B). Interestingly, the increase in abundance of the kinase on the plasma membrane after GA treatment also occurs when cells are treated with growth factors (13). Therefore, we next asked whether activation of growth factor receptors disrupt the Hsp90-PI4KII␤ interaction. To address this question, COS cells were co-transfected with Myc-PI4KII␤ and FLAG-Hsp90, and FIGURE 8. Redistribution of endogenous PI4KII␤ to membranes in response to GA treatment. A, cytosol (S) and membrane (P) fractions were prepared from untransfected HeLa cells, left untreated or treated for 2 h with 2 M GA, as described in Fig. 7. The samples were electrophoresed, blotted, and stained with anti-PI4KII␤ antibodies (13). B, immunofluorescence localization of endogenous PI4KII␤ in control and GA-treated cells. HeLa cells were left untreated (left panels) or treated with 2 M GA for 2 h (right panels). Fixed and permeabilized cells were stained with anti-PI4KII␤ primary antibodies and rhodamine-labeled secondary antibodies. Similar results were obtained in three separate experiments. their co-immunoprecipitation was measured in response to treatment with EGF (50 ng/ml) or PDGF (100 ng/ml). Under both conditions, the PI4KII␤-Hs90 complex was disrupted to a similar extent as was caused by a 15-min treatment with GA ( Fig. 9).

DISCUSSION
Of the four mammalian PtdIns 4-kinase isoforms, PI4KII␤ is by far the least understood in terms of function or regulation. Unlike the other three isoforms, PI4KII␤ is almost evenly distributed between membranes and cytosol. The two type III kinases are almost entirely cytosolic (10), and the other type II isoform, PI4KII␣, is almost entirely membrane-bound (11).
Because cytosolic PI4KII␤ is catalytically inactive, its redistribution to membranes is likely to represent a major mechanism of regulation. Indeed, a portion of PI4KII␤ is recruited to membranes in response to stimuli, such as growth factor receptor activation (13). In an effort to explain the subcellular distribution of PI4KII␤, we sought to identify binding partners that could either recruit the kinase to membranes or sequester it in the cytosol. Our results revealed that the molecular chaperone Hsp90 binds preferentially to cytosolic (and weakly membranebound) PI4KII␤, inhibits its association with membranes, and protects it from proteolytic degradation. Moreover, the interaction between PI4KII␤ and Hsp90 was weakened upon treatment of cells with EGF or PDGF, suggesting that Hsp90 is a key regulator of stimulus-dependent PtdIns4P production. Based on our data, we propose a model for the life cycle of PI4KII␤ shown in Fig. 10. According to this model, cytosolic PI4KII␤ associates with Hsp90 for stabilization. An extracellular signal disrupts the interaction, allowing the free kinase to translocate to the membrane, where it may undergo palmitoylation and activation. Short exposure to GA may mimic the effect of growth factors. In contrast to PI4KII␤, we suggest that PI4KII␣ rapidly associates with Golgi membranes after its synthesis on cytosolic ribosomes and then is stably palmitoylated and, hence, constitutively active.
Hsp90 is essential for the maturation, stability, and translocation of a defined set of so-called "client" proteins (reviewed in Refs. 26, 31, 39, and 40). Currently there are more than 100 known Hsp90 clients, and many of those, including steroid hormone receptors, transcription factors, and protein kinases, participate in signal transduction pathways. To our knowledge, this is the first example of a lipid kinase as an Hsp90 client, although PI3K activity is indirectly regulated by Hsp90 (36). The chaperone function of Hsp90 is ATP-driven and involves the coordinated assistance of a variety of co-chaperones, as well FIGURE 9. EGF and PDGF disrupt the interaction between Hsp90 and PI4KII␤. COS cells were transfected with FLAG-Hsp90 alone or together with Myc-PI4KII␤ and starved for 8 h to achieve quiescence. The cells were then treated with GA, PDGF, or EGF at different concentrations for 15 min. Me 2 SO was added as a mock treatment. The cell lysates were prepared in 1% Nonidet P-40 buffer, immunoprecipitated (IP) with anti-Myc antibodies, and blotted with anti-Hsp90 antibody. The inputs were immunoblotted with anti-Hsp90, anti-Myc, and anti-phospho-ERK1/2 antibodies (to demonstrate effectiveness of growth factor treatments). Note the slight increase in ERK1/2 phosphorylation at 5 M GA, as reported (41,42). These data are representative of three independent experiments. con, control.
FIGURE 10. Proposed life cycle of PI4KII␤. Hsp90 complex binds to cytosolic PI4KII␤, either newly synthesized or dissociated from membranes. In response to stimuli, PI4KII␤ is released from the Hsp90 complex and then docks at the membrane through an as yet unidentified docking protein.
Acyltransferases on the membrane palmitoylate peripherally bound PI4KII␤ leading to integral membrane binding and expression of catalytic activity. A portion of membrane-bound PI4KII␤, particularly that which is peripherally associated, recycles to the cytosol, and is recaptured by Hsp90. GA disrupts the interaction between Hsp90 and PI4KII␤. Short exposure to GA mimics the effect of other stimuli, whereas longer exposure to GA gives rise to proteasome-dependent degradation of PI4KII␤.
as of Hsp70 (26,27). Specific inhibitors of the Hsp90 ATPase reaction, such as GA, have been used to disrupt interactions between Hsp90 and its client proteins in cells and to establish the functional significance of these interactions (24). As shown in this study for PI4KII␤, GA treatment has often resulted in enhanced proteolytic degradation and reduction in cellular levels of Hsp90 clients (reviewed in Ref. 41).
Based on our observations, we hypothesize that growth factor-dependent stimulation of PI4KII␤ is tightly and obligatorily linked to its release from Hsp90. There are other examples in the literature of Hsp90 serving as an inhibitor of kinase activity. For example, protein kinase R is activated by dsRNA, which triggers its dissociation from Hsp90 (42). Also, Src is transiently activated upon its release from Hsp90 (43). Thus, it appears that in some cases, interaction with Hsp90 can be inhibitory in the short term, although protective in the long term. At present, we have no information regarding the downstream signaling event(s) that might trigger growth factor-dependent dissociation of PI4KII␤ from Hsp90. The most obvious possibility, stimulus-dependent phosphorylation of PI4KII␤, has not yet been examined, although it is interesting that this modification occurs only in the membrane-associated species (14). Moreover, because there is evidence that Hsp90 kinase interactions can be disrupted by phosphorylation (44), this direction will be pursued in future studies.