Type II Phosphatidylinositol 4-Kinase b is a Cytosolic and Peripheral Membrane Protein that is Recruited to the Plasma Membrane and Activated by Rac-GTP

These findings provide new insight into how phosphoinositide cascades are propagated in cells.


INTRODUCTION
We and another group recently cloned the first type II PtdIns 4-kinase 15;16, which is designated as PI4KIIα. Overexpression studies showed that PI4KIIα behaves like an integral membrane protein and is primarily organelle associated 15;16. Data base analyses reveal that humans have another type II PtdIns 4-kinase, that has a unique N-terminal 100 amino acids sequence and a highly homologous (58% identical and 75% homologous) downstream sequence to that of PI4KIIα. This kinase, designated as PI4KIIβ, was cloned recently by Balla et al. 17. According to this group, overexpressed PI4KIIβ colocalizes with overexpressed PI4KIIα in endosomal vesicles 17. Neither isoform is obviously enriched in the plasma membrane or the Golgi apparatus, and their endogenous distribution was not reported.
The apparent low abundance of type II PI4Kα and β at the plasma membrane is surprising in view of previous biochemical data suggesting that they are enriched there 8. This paradoxical finding also raises questions about whether PIP and PIP2 synthesis is coupled at the plasma membrane 18, and if so, how. One possibility that is modeled after yeast is that PIP is synthesized primarily in the Golgi, while PIP2 is synthesized primarily at the plasma membrane.
The two processes are therefore not necessarily coupled 11;19. Another possibility is that PIP is synthesized in situ at the plasma membrane by PtdIns 4-kinases that are recruited there by an as yet unidentified mechanism. There is already evidence for the plasma membrane and/or internal membrane recruitment of PIP5Ks and PI4KIIIβ by Rho 20, Rac 21;22 and Arf 12;23;24 small GTPases.
In this paper, we compared the intracellular localization of endogenous and overexpressed PI4KIIα and β, and examined the effect of growth factor stimulation and Rho family GTPases on their distribution and kinase activity. Our results are significantly different from those reported recently by Balla et al. 17 for overexpressed PI4-kinases. The differences and possible explanations are discussed at the end of this paper. 6

EXPERIMENTAL PROCEDURES
cDNA Cloning and Transfection-The human PI4KIIβ cDNA was cloned by PCR from the Invitrogen Full-length Genoscope FL1001. The primers used were 5' gac acg cgt tag aag atc cct ccg ag 3' and 5" ggc tct aga tta cta cca gga gga 3'. The cDNA was subcloned into the pCMV5-myc2 or HA vectors between the MluI and XbaI sites. The rat myc-PI4KIIα clone is as described previously 15. Myc-tagged or untagged RacV12 and RacN17 were gifts of A. Hall (MRC, Univ. London). Cells were transiently transfected using Lipofectamine and used between 8-18 hours post transfection. In cotransfection studies with Rac, the Rac expression plasmids were used at between a seven-to three-fold excess over the PtdIns 4-kinase plasmids. GFPexpression plasmids were used in control samples.
Antibody Production-Synthetic peptides encompassing residues 2-17 of human PI4KIIα and residues 22-36 of human PI4KIIβ and conjugated to keyhole limpet hemocyanin via an extra N-terminal cysteine residue were used to immunize rabbits. The antibodies were affinity purified by absorption against peptides attached to nitrocellulose filters 25.
Immunofluorescence LocalizationCells with and without transfection were fixed either in paraformaldehyde and permeabilized with Triton X-100, or with cold methanol as described previously 26. Cells were stained with antibodies and images were collected by confocal fluorescence microscopy. In some cases, cells were stimulated with PDGF (AB, 100 ng/ml for 15 min.) prior to fixation. Cells with distinct plasma membrane staining were scored in randomly chosen fields.
Golgi Fractionation-Golgi-enriched membranes were obtained using a discontinuous Membrane AssociationMembrane association was analyzed in two ways. First, microsomal membranes were isolated by a one step fractionation procedure 15. Cells were scraped from the petri dish in a hypotonic lysis solution containing 0.25 M sucrose, 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 2 mM EDTA, 1 mM dithiothreitol (DTT) and a protease inhibitor cocktail. They were lysed by 2 freeze and thaw cycles, and unbroken cells and nuclei were removed by centrifugation at 1,000 x g for 5 min. to obtain a postnuclear fraction. The supernatant was recentrifuged at 190,000 x g for 15 min. to obtain cytosol (supernatant) and microsome (pellet) fractions. The 190,000 x g pellets were washed once in the lysis buffer and recentrifuged. The pellets were boiled and analyzed on SDS-polyacrylamide gels. In some cases, the pellets were further extracted for 10 min. on ice either with 1 M NaCl in lysis buffer, or with 1 M NaCl/0.1 M Na 2 CO 3 (pH 10). Membranes and insoluble material were collected by centrifugation at 190,000 x g for 15 min.
A second procedure employing multiple centrifugation steps was used to examine the partitioning of these kinases in different crude organelle fractions 28. Cells were lysed by freeze-thawing as above, and centrifuged at 19,000 x g for 20 min. at 4 C. The 19,000 x g pellet was resuspended in 100 µl lysis buffer, and overlaid onto a 0.8 ml cushion of 1.12 M sucrose. After centrifugation at 100,000 x g for 1 hr., the membrane layer at the top of the 8 sucrose cushion was collected with a long needle and sedimented at 40,000 x g for 20 min. This fraction was enriched in plasma membrane markers and is designated as PM 28. The 19,000 x g supernatant obtained from the first step was sequentially centrifuged at 41,000 x g for 20 min. and at 180,000 x g for 80 min. to collect the low speed pellet (LSP) and the high speed pellet (HSP). The partitioning of PtdIns 4-kinases in these fractions was determined by analyzing equal fraction of each sample by western blotting with anti-myc antibody.
In vitro PtdIns 4-Kinase Assay-Kinase activity was measured by phosphorylation of

RESULTS AND DISCUSSION
PI4KIIα and β Protein Expression-We cloned the human PI4KIIβ cDNAs, and confirmed that its sequence is almost identical to that in the database. The Genebank accession no. of our clone is AY091514. The antibodies generated by immunizing rabbits with the unique N-terminal sequences are isoform specific, based on western blotting with overexpressed proteins (data not shown). Each recognizes a single band in western blots in the three tissue culture cell lines tested (Fig. 1A). PI4KIIα, which has a slightly smaller predicted mass than PI4KIIβ (54,022 vs 54,744 daltons, respectively), migrates slightly ahead of PI4KIIβ. Unlike PI4KIIβ, PI4KIIα is located primarily in the Golgi, and there is no obvious plasma membrane staining (Fig. 1C). The mitotic cell in the middle of the field has no Golgi staining by either anti-PI4KIIα or anti-TGN46, consolidating the conclusion that the perinuclear PI4KIIα is Golgi-associated. The colocalization of both PtdIns 4-kinase isoforms in the Golgi and their differences in the cytosol and plasma membrane are confirmed by coexpression (Fig. 1D).
Overexpressed PI4KIIα is primarily organelle associated; it is present on the Golgi apparatus and on vesicular structures outside of the perinuclear region as well (Fig. 1D, inset). The latter is consistent with Balla et al.'s report on overexpressed PI4KIIβ 17. Some overexpressed PI4KIIβ is Golgi-associated, but it is much more cytosolic and more obviously plasma membrane associated than myc-PI4KII. The difference in plasma membrane association is confirmed by direct counting of cells in randomly chosen fields. Overexpressed PI4KIIβ is found in the plasma membrane of 39% of the cells, while overexpressed PI4KIIα is found in only 19% of the cells (Table 1).
Subcellular Fractionation-Subcellular fractionation studies also confirmed the 11 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page predominant Golgi localization of PI4KIIα and the differential localization of the two type II PtdIns 4-kinases. We employed a well-characterized protocol that is widely used to study COP1 coat formation 27 and Arf1 dependent recruitment of regulatory proteins to Golgi-enriched membranes 23. PI4KIIα is highly enriched in Golgi fractions 3 and 4 (interface III) (Fig. 2) and the surrounding fractions as well. TGN38 is almost exclusively in fractions 3 and 4, and this tighter distribution profile is consistent with TGN38's primary localization in the TGN. COP1, which cycles between the cytosol and the Golgi, is found in these fractions, as well as in interfaces II (ER-enriched) and I (heavy membranes). The large amount of cytosolic COP1 is consistent with its transient and GTP-dependent association with membranes. PI4KIIβ has a similar distribution, suggesting that it is transiently associated with the Golgi as well. PI4KIIβ appears as a doublet in some of the fractions, possibly due to differential phosphorylation (manuscript in preparation).
Modes of Membrane Association-The immunofluorescence localization and subcellular fractionation results suggest that PI4KIIα is primarily Golgi associated while PI4KIIβ is both cytosolic and organelle associated. We used a quick one step microsome preparation to compare and characterize the organelle association of endogenous PtdIns 4-kinases. Following high-speed centrifugation, 38% of endogenous PI4KIIβ was recovered in the 190,000 x g light microsome pellet fraction (P, Fig. 3A), while >99% of endogenous PI4KIIα was membrane bound, as reported previously for overexpressed PI4KIIα 15;16.
Significantly, the membrane-associated PI4KIIα and β are differentially sensitive to salt and alkaline extraction (Fig. 3A). Microsome-associated PI4KIIβ was 95% extractable with 1 M NaCl and Na 2 CO 3 /NaCl did not extract more. Thus, PI4KIIβ behaves like a peripheral membrane protein. In contrast, PI4KIIα was resistant to carbonate extraction (Fig. 3A), behaving like an integral membrane protein. 12 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page Similar results were observed with overexpressed PtdIns 4-kinases, except that more myc-PI4KIIβ is microsome-associated (ranging from 20 to 50% in COS7 cells in a total of 5 experiments, and more in 293 cells (data not shown)), and myc-PI4KIIβ was less extractable by high salt or carbonate than the endogenous protein (Fig. 3B). We believe that variable recovery and decreased extractability are due to overexpression and therefore limited the overexpression time to less than 18 hrs.  COS7 cells that were serum starved overnight have some HA-PI4KIIβ staining of the plasma membrane, while cells that were cotransfected with the constitutively active myc-RacV12 (at a 1:4 ratio) have much more pronounced plasma membrane staining (Fig. 4B). As expected, myc-RacV12 is enriched at the plasma membrane where it colocalizes with HA-PI4KIIβ (Fig. 4B) Likewise, PI4KIIα is recruited to the plasma membrane by RacV12 (Fig. 4C). This effect is GTP-dependent, because cells RacN17 does not increase plasma membrane recruitment (Fig. 4C, bottom panels). The increase in plasma membrane staining is corroborated by comparing cells with plasma membrane association (  5A). We employed a differential fractionation protocol that was optimized to establish the Golgi and plasma membrane localization of syntaxin 6 28. The cell lysates were separated into a cytosolic fraction (CYT), a low-speed pellet (LSP) that is enriched in Golgi membranes and early endosomes, a high-speed pellet (HSP) that is enriched for lysosomes and late endosomes and a plasma membrane rich fraction (PM).

Effects of Growth Factor
This fractionation technique confirmed that the two kinases are differentially localized.
47% and 2% of the total PI4KIIβ and PI4KIIα, respectively, are cytosolic (Fig. 5A). Of the membrane bound myc-PI4KIIβ, most (35% of soluble and insoluble enzyme) was found in the HSP fraction which is enriched for lysosomes/late endosomes. Some was associated with the PM fraction (6%) and the Golgi/early endosome-enriched LSP fraction (12%). In contrast, most of the PI4KIIα was in the LSP (68%), and smaller amounts were present in the HSP (11%) and PM (19%) fractions. The ratios of PI4KIIβ in the PM: LSP: HSP fractions are 1:2:6, while that for PI4KIIα is 1:4:1. These results establish that the two isoforms are preferentially associated with different organelles at steady state, and corroborate data obtained by gradient centrifugation (Fig. 2), one step fractionation (Fig. 3) and by immunofluorescence microscopy (Fig. 1). From the large difference in kinases partitioning between the Golgi and plasma membrane, we conclude that PI4KIIα may be primarily responsible for the synthesis of PIP at the Golgi, while PI4KIIβ may have a more important role elsewhere, such as shuttling between the cytosol, endosomes, lysosomes and the plasma membrane. 15 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page We next confirmed that RacV12 increases PI4KIIα and β association with the plasma membrane. As expected, RacV12 is found predominantly in the PM fraction (Fig. 5A). RacV12 increases in PI4KIIβ in the PM (2.5 fold increase) and LSP (1.6 fold increase) fractions, and slightly decreases it in the HSP and cytosol (to 0.7 and 0.8 fold of control, respectively). Some PI4KIIα is recruited to the plasma membrane (1.6 fold increase), but the most dramatic change is a 22-fold decrease in HSP association. The different responses of PI4KIIα and β to RacV12 suggest that the observed redistributions do not merely reflect a generalized change in vesicular trafficking that alters the amount of membranes recovered in each fraction. If this were the case, both kinases would be expected to shift from one compartment to another to a similar extent.
Our data suggest that PI4KIIβ is recruited from the cytosolic fraction to the membranes by RacV12, whereas PI4KIIα, which is constitutively membrane associated, is increased at the plasma membrane by redistribution among various membrane compartments.
RacV12 Increases PI4KIIβ, but not PI4KIIα, Activity-The effect of Rac on PtdIns 4kinase activity was tested by immunoprecipitating myc-tagged PtdIns 4-kinases from cells that cotransfected with an excess of untagged RacV12 or RacN17 cDNAs. The myc-PI4KIIβ immunoprecipitated from cells coexpressing RacV12 had a 1.5 fold higher kinase specific activity than those precipitated from cells without Rac, and RacN17 decreased activity to 0.6 of control levels (Fig. 5B). These effects are specific for PI4KIIβ, because Rac-GTP had minimal effect on PI4KIIα activity, even though it changes its intracellular distribution.
We do not know how RacV12 activates myc-PI4KIIβ. The mostly likely possibility is that when PI4KIIβ is recruited to the plasma membrane by Rac, it is exposed to activating enzymes or ligands located at the plasma membrane. In this scenario, in vivo activation of PI4KIIβ at the plasma membrane could be much higher than the 1.5 fold increase detected in the immunoprecipitates, because only 15% of the total PI4KIIβ (Fig. 5A) is at the plasma membrane 16 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page where it can be exposed to the activating ligands/enzymes.

PI4KIIβ activation by
phosphorylation/dephosphorylation is an attractive possibility. However, our preliminary experiments showed that alkaline phosphatase treatment did not alter the activity of either membrane-associated or cytosolic myc-PI4KIIβ kinase (data not shown).

Membrane Association Increases the Activity of Myc-PI4KIIβ-To test the possibility that
PI4KIIβ is activated by membrane association, we compared the specific activity of myc-PI4KIIβ immunoprecipitated from the 190,000 x g pelletable microsomal fraction and from the supernatant fraction obtained from cells grown in serum-containing medium (i.e. activating conditions) (Fig. 6A). The membrane-bound PI4KIIβ is 16 times more active than an equivalent amount of soluble PI4KIIβ in the experiment shown. Similar results were obtained in three independent experiments, and activation can reach as high as 20 fold. Assuming that only 6% of PI4KIIβ in resting cells is membrane-associated (Fig. 5A&B), translocation of the entire cytosolic pool (94%) to the membrane would be predicted to increase overall cellular kinase activity by approximately 15-fold. Thus, translocation of 9% of the total PI4KIIβ upon Rac stimulation should increase activity by 1.4 fold, which is remarkably similar to the 1.5 fold value obtained experimentally (Fig. 5B). These results establish that PI4KIIβ is strongly activated by membrane association.
Membrane associated PI4KIIβ has almost identical kinetic properties as PI4KIIα which is constitutively membrane-associated. They have similar apparent K m for ATP (65 µM vs. 86 µM for PI4KIIβ and α, respectively) and for PtdIns (21 vs. 35 µM) (Fig. 6B). The shallow drop in PI4KIIβ activity at high PtdIns concentration was consistently observed, raising the possibility that this isoform is more sensitive to substrate inhibition than PI4KIIα. The physiological significance of this remains to be explored. PI4KIIβ, like PI4KIIα, was inhibited by micromolar 17 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page adenosine, but not by wortmannin (data not shown), in agreement with Balla et al. 17. PI4KIIβs differential sensitivity to these inhibitors justifies its classification as a bona fide type II PtdIns 4kinase.
In conclusion, our results establish that both PtdIns 4-kinases are recruited to the plasma membrane by Rac in a GTP-dependent manner, but PI4KIIβ, a predominantly cytosolic vitro kinase studies showed that immunoprecipitated PI4KIIβ-HA is only about 30% as active as PI4KIIα-HA. 18 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page In contrast, we studied both endogenous and overexpressed PtdIns 4-kinases, and presented new information about significant differences in the localization, regulation and mode of membrane association of PI4KIIα and β. First, immunofluorescence localization studies showed that PI4KIIβ is both cytosolic and organelle associated while PI4KIIα is predominantly Golgi associated. Furthermore, the endogenous and overexpressed forms of both kinases are enriched in the Golgi, and colocalize with two different Golgi markers that are located in the cis and trans Golgi (Fig. 1B). Similar results were obtained using two different methods to fix and permeabilize the cells, suggesting that the localization is unlikely to be due to fixation artifacts.
Second, biochemical fractionation studies showed that PI4KIIβ is predominantly recovered in the cytosol fraction, while PI4KIIα is almost completely Golgi-associated ( Fig. 2 and 5).
Furthermore, PI4KIIβ associates with membrane as a peripheral protein, while PI4KIIα behave as an integral membrane protein (Fig. 3). Third, PI4KIIβ is recruited and activated at the plasma membrane by PDGF and by Rac in a GTP-dependent manner (Fig. 5). Fourth, membrane recruitment stimulates PI4KIIβ activity significantly (Fig. 6A).
What is the basis for the discrepancies between ours and their results? At least some of the differences may be ascribed to the different placement of the epitope tags. Balla's group places the GFP or HA tag at the C-termini; we attach a myc or HA tag to the N-termini. We chose to examine N-terminally tagged expressed PtdIns 4-kinases because we found that deletion of as few as seven residues from the C-terminus of PI4KIIα dramatically decreases its enzymatic activity whereas removal of even 91 residues from the N-terminus had no effect on catalysis (manuscript in preparation). A second potential explanation is the different levels of overexpression used in Balla's and our experiments. We routinely used short overexpression time (between 8 to 18 hours post-transfection), because we found that myc-PI4KIIβ becomes more organelle-associated and insoluble with prolonged overexpression (data not shown), and 19 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page cells that overexpress at a high level (e.g. 293 cells) have more membrane-bound myc-PI4KIIβ kinases than those that express at moderate level (e.g. CV1 or COS7 cells). We also found that overexpressed PI4KIIα becomes increasingly associated with endosomes and lysosomes (data not shown) even though it is initially predominantly Golgi-associated (Fig. 1C & D). Balla's group routinely used cells overexpressing the kinases for 24-48 hours and found that GFPtagged PtdIns 4-kinases partially colocalize with anti-gm130 and GFP fluorescence is also present in surrounding granules that are not positive for anti-gm130. We suggest that their higher level of protein expression increases non-Golgi vesicle staining and obscures authentic Golgi staining. Perhaps the most important consideration here is that, in our studies, immunofluorescence localization and biochemical fractionation of overexpressed kinases merely served to confirm the behavior of endogenous kinases as detected with isoform specific antipeptide antibodies. Finally, we cannot explain the large discrepancies in specific activities of PI4KIIα and β between our two groups. This difference cannot be due to a difference in membrane association, because their overexpressed PI4KIIβ is as pelletable as their PI4KIIα.
Perhaps PI4KIIβ is more readily denatured when expressed at a high level. It remains to be seen if the different placement of the epitope tags can also account for this difference. 20 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page Cells with obvious plasma membrane staining were counted in randomly chosen fields. Between 75-100 cells are counted in each category per experiment, and the data is average of two experiments. 24 Plasma membrane recruitment of type II PtdIns 4-kinases by Rac M2:06860revised2 Page