Subcellular Locations of Phosphatidylinositol 4-Kinase Isoforms*

Phosphatidylinositol (PtdIns) 4-kinase catalyzes the synthesis of PtdIns-4-P, the precursor of an array of lipid second messengers generated by additional phosphorylation by PtdIns-4-P 5-kinase and PtdIns 3-kinase. PtdIns 4-kinase activity is conserved from yeast to higher eukaryotes. Multiple isoforms of mammalian PtdIns 4-kinase have been purified, and the activities have been detected in almost all subcellular locations. We previously reported the cloning and characterization of the first mammalian PtdIns 4-kinase named PI4Kα (Wong, K., and Cantley, L. C. (1994) J. Biol. Chem. 269, 28878–28884). Alternatively spliced forms of PI4Kα have also been identified from several sources including bovine brain (Gehrmann, T., Vereb, G., Schmidt, M., Klix, D., Meyer, H. E., Varsanyi, M., and Heilmeyer, L. M., Jr. (1996) Biochim. Biophys. Acta 1311, 53–63). Recently we isolated a distinct human PtdIns 4-kinase gene, named PI4Kβ, that encodes an enzyme that is wortmannin sensitive (Meyers, R., and Cantley, L. C. (1997) J. Biol. Chem. 272, 4384–4390). Here we report the locations of these enzymes and provide evidence for other yet unidentified isoforms present in specific organelles. PI4Kα is mostly membrane-bound and located at the endoplasmic reticulum; whereas PI4Kβ is in the cytosol and also present in the Golgi region. Neither of these isoforms accounts for the major type II PtdIns 4-kinase activity detected in the lysosomes and plasma membrane fraction.

The metabolism of phosphoinositides is a key event in transmitting mitogenic and developmental signals in response to a variety of hormones and growth factors. Two signal transduction pathways, each utilizing distinct phosphoinositide derivatives, have been characterized. In one pathway, phosphatidylinositol (PtdIns) 1 -4,5-P 2 is hydrolyzed by receptor-activated phospholipase C enzymes to generate the signaling molecules inositol 1,4,5-trisphosphate and diacylglycerol (4). PtdIns-4-P and PtdIns-4,5-P 2 have also been shown to mediate actin rearrangement by directly regulating actin-binding proteins (5,6). In a distinct pathway, phosphoinositides are phosphorylated at the C-3 position to generated a family of lipid messengers (7). These lipids, PtdIns-3-P, PtdIns-3,4-P 2 , and PtdIns-3,4,5-P 3 are not substrates for phospholipase C but are implicated in mitogenesis (8), intracellular trafficking (9) as well as actin rearrangement (10,11). Feeding into both PtdIns pathways is the precursor PtdIns-4-P, synthesized by PtdIns 4-kinases.
The subcellular distribution of mammalian PtdIns 4-kinases has been studied. The majority of PtdIns 4-kinase activity in human cells is membrane-bound (12). PtdIns 4-kinase activity is detected in most membrane structures, including plasma membrane (13), nuclear envelope (14), lysosome, Golgi apparatus (15), and endoplasmic reticulum (16). PtdIns 4-kinase activity is also detected in coated vesicles, glucose transportercontaining vesicles (17), and several specialized organelles, such as chromaffin granules (18) and secretory vesicles from mast cells (19). In addition, cytosolic PtdIns 4-kinases have been described (20). In view of such a wide subcellular distribution, it has long been speculated that different isoforms, targeted to different intracellular compartments, perform different physiological functions. While there is some evidence suggesting that distinct isozymes of PtdIns 4-kinase may be targeted to distinct organelles and be independently regulated, previous studies have been limited by lack of cDNA clones and isoform specific antibodies.
In Saccharomyces cerevisiae, only two PtdIns 4-kinase isoforms are found in the entire genome. PIK1, the first PtdIns 4-kinase to be cloned, encodes a nuclear-associated PtdIns 4-kinase of 125 kDa that is indispensable for cell growth (21,22). Mutants arrest in G 2 due to defects in cytokinesis. A second yeast gene, STT4, encodes a 200-kDa PtdIns 4-kinase that appears to be cytosolic and is dispensable for growth. Genetic studies of STT4 suggest its involvement in the protein kinase C pathway (23,24). In contrast, the yeast PtdIns 3-kinase, Vps34, is required for vesicle trafficking from the Golgi apparatus to the vacuole (9). Interestingly FAB1, a yeast gene that is homologous to the mammalian PtdIns-4-P 5-kinase, is required for normal vacuole function and morphology (25). Because PtdIns 4-kinase generates PtdIns-4-P, the precursor for subsequent phosphorylation by downstream PtdIns kinases, PtdIns 4-kinases must play either a direct or an indirect role in vesicular trafficking (26).
Previously, we reported the cloning and characterization of the first mammalian PtdIns 4-kinase, named PI4K␣ (1). This protein is highly homologous to the yeast STT4 enzyme. Northern blot analysis of the poly(A) ϩ mRNAs from human tissues and cell lines revealed multiple alternatively spliced transcripts. A 230-kDa PtdIns 4-kinase made from an alternatively spliced form of the rat PI4K␣ has recently been described (27). A bovine gene encoding a 170 -200-kDa PtdIns 4-kinase that shares more than 95% identity with the human PI4K␣ in the overlapping region has also been reported (2). It represents probably another splice variant of PI4K␣. More recently, we isolated another human cDNA that encodes a 110-kDa PtdIns 4-kinase (named PI4K␤) that is more homologous to the yeast PIK1 gene. PI4K␤ is wortmannin-sensitive and may be the same enzyme recently reported to be involved in the hormone sensitive pools of inositol phospholipids (28). To determine the relative roles of these enzymes in producing phosphoinositides for membrane trafficking, hormone sensitive PtdIns turnover, growth factor-dependent PtdIns-3,4,5-P 3 production, and cytoskeletal rearrangement, it is first important to determine the subcellular locations of these enzymes.

EXPERIMENTAL PROCEDURES
Materials-PtdIns, [␥-32 -P] ATP, and silica gel plates were purchased from Avanti (Alabaster, AL), DuPont NEN, and E. Merck (Germany), respectively. Mouse monoclonal antibody 12CA5, reactive to the influenza virus hemagglutinin, was purchased from BabCo. Rabbit polyclonal anti-PI4K␣ antibody 3334 was raised to a peptide corresponding to amino acids 501-512, KPYPKGDERKKA, coupled to keyhole limpet hemacyanin (1). Peptide antibodies were affinity-purified prior to immunoblotting and immunofluorescent staining experiments. Anti-PI4K␤ antibody was raised against a GST-fusion protein that was generated by polymerase chain reaction using oligonucleotide primers that generate amino acids 410 -538 of PI4K␤ (3). Anti-BiP and anti-␥adaptin were purchased from StressGen Biotechnologies Corp. and Sigma, respectively.
Epitope Tagging and Expression of PI4K␣ and PI4K␤-The 2.6kilobase pair PI4K␣ was tagged at the amino terminus with a 9-amino acid epitope (YPYDVPDYA) derived from influenza virus hemagglutinin by insertion into Bluescript encoding the epitope sequence. The tagged cDNA was subcloned into a mammalian expression vector pRC/ CMV (Invitrogen). The construct was transiently transfected into CHO or HeLa cells using LipofectAMINE (Life Technologies, Inc.) according to the manufacture's procedures. The tagged cDNA was also subcloned into a baculoviral expression vector. Recombinant viruses were generated with the linear AcMNPV transfection module (Invitrogen) and were plaque-purified prior to use. The GST-PI4K␤ construct was generated by polymerase chain reaction as described in detail elsewhere (3) and was expressed in Escherichia coli by standard procedures. The fusion protein is predicted to be 120 kDa since it lacks the aminoterminal 82 amino acids of PI4K␤.
Preparation of Particulate and Cytosolic Fractions-Adherent parental CHO-IRS or CHO-IRS cells transfected with PI4K␣ were harvested with phosphate-buffered saline containing 1 mM EDTA and 1 mM EGTA. Cells were pelleted and then resuspended in homogenizing buffer (10 mM Tris-HCl, pH 7.4, 20 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml pepstatin). After swelling for 15 min on ice, the cells were Dounce-homogenized (25-35 strokes). Nuclei and unbroken cells were spun out at 1000 ϫ g for 10 min at 4°C. The postnuclear supernatant was centrifuged at 100,000 ϫ g for 1 h in a SW55Ti rotor (Beckman Instruments). The membrane pellet was solubilized with 1% Triton X-100 in homogenizing buffer. The cytosolic and particulate fractions were adjusted to equal volume.
Separation of Subcellular Organelles-Fractionation procedures were performed according to Storrie and Madden (29). Briefly, CHO-IRS cells transfected with PI4K␣ were disrupted by low pressure nitrogen cavitation in 0.25 M sucrose, pH 7.4, containing various proteinase inhibitors. After centrifugation at 1300 ϫ g for 10 min, the postnuclear supernatant was collected and overlaid on a hybrid percoll/metrizamide discontinuous density gradient (5 ml of 6% Percoll, 2 ml of 17% metrizamide, 2 ml of 35% metrizamide). The step gradient was centrifuged at 20,000 rpm for 30 min at 4°C in a SW44Ti rotor (Beckman Instruments). Interfaces were sequentially removed from the top of the gradient as follows: first the top of the gradient, followed by the postnuclear supernatant in 0.25 M sucrose, the sucrose/Percoll interface, the 6% Percoll/17% metrizamide interface where lysosomes sediment, and finally the 17%/35% metrizamide interface where mitochondria sediment.
Assays of Organelle Marker Enzymes-Alkaline phosphodiesterase, cytochrome c oxidase, ␤-hexosaminidase, and ␣-mannosidase II were used as the marker enzymes for plasma membrane, mitochondria, lysosomes, and Golgi apparatus, respectively. Lactate dehydrogenase was used as a cytosol marker. Each of the enzyme assays was performed as described previously (29).
Immunofluorescence Studies of PtdIns 4-Kinases-HeLa cells grown on coverslips were fixed with 4% paraformaldehyde for 20 min. Fixed cells were permeabilized and nonspecific reactive sites were blocked for 30 min at room temperature in phosphate-buffered saline containing 0.1% Triton X-100, 5% normal goat and donkey serum. Cells were then incubated with the appropriate primary PtdIns 4-kinase isoform-spe-cific antibody for 1 h at room temperature. Primary antibodies were detected by species-specific secondary antibodies, namely Cy TM 3-conjugated goat anti-rabbit IgG and fluorescein isothiocyanate-conjugated donkey anti-mouse IgG. ER was identified by staining BiP, a resident ER protein. Golgi was visualized using anti-␥-adaptin. The immunostained cells were observed by either conventional light or confocal microscopy.
PtdIns Kinase Assay and HPLC Analysis-PtdIns kinase assays were performed as described previously (30). Briefly, the reaction mixture contained 0.3% Triton X-100, 50 M ATP, 20 mM HEPES, pH 7.5, 10 mM MgCl 2 , 0.2 mg/ml sonicated lipids, 20 Ci of [␥-32 P]ATP (3000 Ci/mmol; DuPont NEN) per sample. Assays were performed at 37°C for 20 min and then stopped with 25 l of 5 N hydrochloric acid. The lipid was extracted with 160 l of 1:1 (v/v) chloroform:methanol. The organic layer was collected and analyzed by both thin layer chromatography and HPLC as described in detail elsewhere (31).

PtdIns 4-Kinase Antibody
Specificity-Since the 97-kDa form of PI4K␣ and the PI4K␤ enzyme have similar mobilities on SDS-gels, it is critical to verify that the antibodies raised against the highly divergent regions of these two enzymes indeed do not cross-react. We assessed the isozyme specificity of the PtdIns 4-kinase antibodies by immunoblotting the recombinant PI4K␣ and PI4K␤ that were expressed in Sf9 insect cells and in E. coli, respectively. genes are almost ubiquitously expressed in human tissues (1,3), and since the antibodies raised against these enzymes react with the respective proteins in other mammalian tissues (not shown), we first investigated the ability of these antibodies to blot proteins in CHO cells. The anti-PI4K␣ antibodies did not detect a 97-kDa PI4K␣ isoform in the parental CHO cells ( Fig.  2A, left panel), but strongly reacted with an approximately 180-kDa protein (p180) of the size expected for the high molecular weight alternative splice of PI4K␣ (2). This result is consistent with the ubiquitous expression of the 7.5-kilobase pair message for PI4K␣ (1, 27). The 97-kDa form of PI4K␣ was detected when this cDNA was introduced. The PI4K␤ antibody reacted with an approximately 110-kDa band ( Fig. 2A, right   panel), consistent with the size of this protein previously detected in Jurkat cells (3). The postnuclear supernatants from the PI4K␣ transfected cells were separated into soluble and particulate fractions by centrifugation at 100,000 ϫ g. The 97-kDa PI4K␣ was predominantly associated with the particulate fraction (Fig. 2B, left panel), as was p180. p110 PI4K␤ was found predominantly in the soluble fraction (Fig. 2B, right  panel).
To further investigate the intracellular localization of PtdIns 4-kinase isoforms, cells overexpressing PI4K␣ were fractionated using the hybrid Percoll/metrizamide discontinuous density gradient as described previously (29). This gradient allows the isolation of lysosomes, mitochondria, and partial separation of plasma membrane from cytosol and organelles such as ER and Golgi. Interfaces sequentially collected from the top of the step gradient were adjusted to the same volume. Equal portions of individual fractions were then used for organelle marker enzyme assays, and for Western blot analysis with the anti-PtdIns 4-kinase antibodies. Monitoring the ␤-hexosaminidase activities, we found 85% of the lysosomes sedimented at the interface of Percoll/17% metrizamide as expected (Table I).
The cytochrome c oxidase assays indicated that 69% of the mitochondria sedimented to the 17%/35% metrizamide interface, and 22% remained in the Percoll/17% metrizamide interface. Western blot revealed no significant amounts of anti-PI4K␣ or anti-PI4K␤ reactive protein present in these two fractions. Instead, p97 PI4K␣, PI4K␤, and the anti-PI4K␣ reactive p180 concentrated at the top of the gradient and in the 0.25 M sucrose fraction (Fig. 3), where over 95% of Golgi apparatus and 40% of plasma membrane cofractionated (Table I). About 40% of the plasma membrane marker activity was also present at the sucrose/Percoll interface, but PI4K␣ and ␤ were not detected in this fraction, suggesting that they are not associated with this fraction of the plasma membrane.
The various subcellular fractions were also assayed for total PtdIns 4-kinase activity. Although half of the PtdIns 4-kinase activity was detected in the upper two factions where cytosol, Golgi apparatus, and endoplasmic reticulum partition, and where both PI4K␣ and PI4K␤ fractionate (Fig. 3), a substantial amount of activity was detected in the region where plasma membranes sediment (34%) and where lysosomes sediment (15%) ( Table I). The assay was carried out under conditions in which PtdIns 3-kinase is inactive (0.3% Triton X-100). HPLC analysis of the deacylated products generated under these conditions verified the absence of PtdIns-3-P (Fig. 4). The PtdIns 4-kinase activity was not significantly inhibited by 1 M wortmannin, indicating that the wortmannin-inhibitable PI4K␤ is not a major component of total activity in these fractions (data not shown). Consistent with previous studies, an inhibitory antibody, 4C5G, raised against the 55-kDa type II PtdIns 4-kinase caused about 75% inhibition of the plasma membrane PtdIns 4-kinase (sucrose/Percoll fraction, Table I). This antibody also substantially inhibited PtdIns 4-kinase activity in other fractions. The 55-kDa type II PtdIns 4-kinase was previously shown to be the major PtdIns 4-kinase in red cell plasma membrane (32) and in other cells. Thus, the results in Table I indicate that the membrane bound forms of PI4K␣ and PI4K␤ are in compartments of the cell consistent with Golgi apparatus and/or endoplasmic reticulum, and that the major PtdIns 4-kinase activity in plasma membrane and lysosome is due to a type II PtdIns 4-kinase distinct from the PI4K␣ isoforms and from PI4K␤.
Immunofluorescent Localization of PtdIns 4-Kinase Isoforms-To further discern the intracellular localization of PI4K␣ and ␤, cytoimmunofluorescence studies were performed using both conventional and confocal microscopy. We chose HeLa cells for these studies since they grow better on coverslips and are hence more suitable for immunofluorescent staining experiments. HeLa cells transiently expressing HA-tagged PI4K␣ were double stained with anti-PI4K␣ (Fig. 5A) and anti-HA epitope antibodies (Fig. 5B). The two antibodies yielded identical patterns, suggesting that the staining obtained using anti-PI4K␣ antisera was specific. The anti-HA antibody did not stain non-transfected cells (not shown). The anti-PI4K␣ antibodies revealed punctate staining near the nucleus and in the cytoplasm, a pattern similar to that expected for ER. To determine whether PI4K␣ is present in the ER, we double stained an ER resident protein Bip (Fig. 5C) and PI4K␣ (Fig. 5D) in parental HeLa cells. Overlap of PI4K␣ with the ER marker was clearly observed after merging the two images by confocal microscopy (Fig. 5E). However, the colocalization was not complete; some punctate staining of PI4K␣ in the peripheral regions was not observed to superimpose with BiP staining. As in CHO cells, the major endogenous protein in HeLa cells that reacts with the anti-PI4K␣ antibody is the p180 protein (not shown), so this protein appears also to be located in the ER.
Immunofluorescent studies with anti-PI4K␤ antibody revealed a pattern characteristic of the Golgi apparatus. Specifically, we noted a crescent of staining on one side of the nucleus (presumably the TGN), and some punctate staining within the cytoplasm which may represent the buds from extended TGN tubules (Fig. 5G). The Golgi localization was confirmed by double staining ␥-adaptin, which has previously been shown to be localized in the TGN and the late endosomes ( Fig. 5F) (33,34). In this case, complete colocalization was observed (Fig.  5H). Staining could be eliminated by preincubation of the primary antibody with the immunizing GST-PI4K␤ and not by GST alone (not shown), indicating that the observed signal is generated by PI4K␤-specific antibodies. Our observation that much of the PI4K␤ protein is in the supernatant of a 100,000 ϫ g spin (Fig. 2B) is consistent with several possibilities. First, the lysis and homogenization procedures dislodge some of this protein from the particulate fraction. Second, the cytosolic fraction is partially lost during cell fixation, or last, this fraction is not as easily visualized because of dilution. DISCUSSION We compared the subcellular distribution of known mammalian PtdIns 4-kinases. The results demonstrate the differential localization of different isoforms and provide support for the hypothesis that individual members of this family of PtdIns kinases are targeted to distinct subcellular compartments and hence are performing different cellular functions. PI4K␣ is detected predominantly in the particulate fraction, and PI4K␤ is present in the cytosol and to a lesser extent in the particulate fraction. Fractionation of CHO-IRS cells revealed that neither TABLE I Summary of organelle marker enzyme assays and the distribution of PtdIns 4-kinase isozymes Fractions of subcellular organelles obtained from the hybrid Percoll/metrizamide gradient (see "Experimental Procedures") were assayed for organelle-specific enzyme activities. The percentage of total activity of each organelle marker is presented. Fractions from the same experiment were also assayed for lipid kinase activity. The PtdIns-4-P products generated in individual organelles were separated by TLC, quantified by a PhosphoImager (BioRad), and expressed as the percentage of total activity. The data shown represent the averages of three separate experiments.  the ␣ nor ␤ isoform is located in the lysosomes or mitochondria, and probably not in the plasma membrane. Immunocytofluorescence studies demonstrate that PI4K␣ is present in the endoplasmic reticulum, and PI4K␤ is localized to the Golgi region.
While this work was in progress, Nakagawa et al. (27) reported the cloning and characterization of an alternative splice of rat PI4K␣ gene that generates a protein predicted to be 230 kDa. Gehrmann et al. (2) also isolated a partial bovine cDNA encoding a 170 -200-kDa PtdIns 4-kinase, that is almost certainly another splice variant of PI4K␣. The carboxyl-terminal half of both high molecular weight proteins contains the domains necessary for lipid catalysis, and share greater than 95% identity with the human p97 PI4K␣. We also noted an alternative splice of PI4K␣ from human tissues that is predicted to encode a larger protein (1). Nakagawa et al. (27) found that the 97-kDa PI4K␣ and the higher molecular weight form colocalize, and the later is predominantly associated with the particulate fraction. These results are consistent with our finding that the p180 and p97 cofractionate and colocalize. It is unlikely that they are integral membrane proteins, since neither splice variant contains a predicted transmembrane domain. However, it is clear from our studies that the membrane association does not require the extended amino-terminal sequence of the higher molecular weight form of PI4K␣.
Nakagawa et al. (27) reported that Flag-tagged versions of both forms of PI4K␣ localize to the Golgi membrane in COS cells. In our study, both the exogenously expressed p97 PI4K␣ and the endogenous p180 localize to the ER rather than the Golgi of HeLa cells. The discrepancy cannot be explained by the cell type, since we also obtained ER staining when PI4K␣ was expressed in COS cells (not shown). The possibility that the HA-tag affected localization is unlikely, since the endogenous p180 exhibited the same location as the HA-tagged p97.
By separating intracellular structures on a step gradient and assaying each fraction for PtdIns kinase activity, we have shown that the lysosomal and plasma membrane-associated PtdIns 4-kinase activities are unlikely to be the result of PI4K␤ or any splice variants of PI4K␣. A plausible candidate is the previously described 55-kDa type II PtdIns 4-kinase. Historically, PtdIns 4-kinases are classified into two categories, types II and type III, on the basis of their molecular sizes, and the differences in their sensitivity to adenosine and nonionic detergent (13). The type II PtdIns 4-kinase is dramatically activated by nonionic detergent but is potently inhibited by both adenosine and the 4C5G monoclonal antibody (32). Type II PtdIns 4-kinase has been purified from erythrocyte membrane (32) and also copurified with the epidermal growth factor receptor (35), as well as with the lymphocyte CD4 antigen (36). Additional findings have also demonstrated that it is recruited to and activated by epidermal growth factor receptor upon ligand stimulation (37). We have extensively investigated the possibility that the type II 55-kDa enzyme is a proteolytic product of the 97-kDa PI4K␣ since they share similar enzymatic properties. However, we are unable to detect proteins of 55 kDa resulting from the processing of the 97-kDa PI4K␣. Thus, the 55-kDa type II PtdIns 4-kinase appears to be encoded by a distinct but related gene.
The finding of the differential ER and Golgi localization of PI4K␣ and PI4K␤ suggests a direct role of PI4K␣ and ␤ in ER and Golgi function, as opposed to the ligand-stimulated phosphoinositide turnover at the plasma membrane. The finding that PI4K␤ is wortmannin-sensitive (IC 50 , 140 nM) and is located in the Golgi apparatus should caution future studies employing wortmannin inhibition as the standard of implicating the role of PtdIns 3-kinase in targeting proteins from Golgi apparatus to lysosome. Finally, our results do not exclude other sites of PtdIns 4-kinase function. For example, the colocalization of ER marker was not complete for PI4K␣. We also noticed a significant amount of PI4K␣ in the particulate fraction that could not be extracted with detergent, implicating cytoskeletal association. Future studies with dominant negative forms or gene knockouts will elucidate the roles of these enzymes in mammalian cell function.