Type I Phosphatidylinositol 4-Phosphate 5-Kinase Directly Interacts with ADP-ribosylation Factor 1 and Is Responsible for Phosphatidylinositol 4,5-Bisphosphate Synthesis in the Golgi Compartment*

Phosphatidylinositol (PtdIns) 4,5-bisphosphate is involved in many aspects of membrane traffic, but the regulation of its synthesis is only partially understood. Golgi membranes contain PI 4-kinase activity and a pool of phosphatidylinositol phosphate (PIP), which is further increased by ADP-ribosylation factor 1 (ARF1). COS7 cells were transfected with a and b forms of PI 4-kinase, and only membranes from COS7 cells transfected with PI 4-kinase b increased their content of PIP when incubated with ARF1. PtdIns(4,5)P2 content in Golgi membranes was nonexistent but could be increased to a small extent upon adding either cytosol or Type I or Type II PIP kinases. However, when ARF1 was present, PtdIns(4,5)P2 levels increased dramatically when membranes were incubated in the presence of cytosol or Type I, but not Type II, PIP kinase. To examine whether ARF1 could directly activate Type I PIP 5-kinase, we used an in vitro assay consisting of phosphatidycholine-containing liposomes, ARF1, and PIP 5-kinase. ARF1 increased Type I PIP 5-kinase activity in a guanine nucleotide-dependent manner, identifying this enzyme as a direct effector for ARF1.

PH domains and many cytoskeletal proteins (2). Many enzymes involved in the synthesis of phosphoinositides have been identified; these include PI 4-kinases ␣ and ␤, which are localized at the Golgi, and two families of PIP kinases, which are mainly cytosolic (3)(4)(5). Although the PIP kinases were previously termed Type I and Type II PtdIns(4)P 5-kinases, recent studies have revealed that the two families selectively phosphorylate different positions on the inositol ring (6). Type I phosphorylates PtdIns(4)P at the 5-position to make PtdIns(4,5)P 2 , and Type II can phosphorylate PtdIns(5)P and PtdIns(3)P at the 4-position to make PtdIns(4,5)P 2 and PtdIns(3,4)P 2 , respectively. Thus, two pathways for the synthesis of PtdIns(4,5)P 2 can be identified (6): PI can be phosphorylated by PI 4-kinase to PtdIns(4)P and subsequently to PtdIns(4,5)P 2 by Type I PIP 5-kinase. An alternative pathway is via phosphorylation of PI to PtdIns(5)P and subsequent phosphorylation by a Type II PIP 4-kinase (see Fig. 6). The enzyme responsible for making PtdIns(5)P could be PIKfyve (7), although a recent study identifies PIKfyve as a Type III PtdIns(3)P 5-kinase, which specifically phosphorylates PtdIns(3)P to make PtdIns(3,5)P 2 (8).
The Golgi apparatus is an integral component of the secretory pathway. ADP-ribosylation factor 1 (ARF1) is a small G protein that is localized to the Golgi and is required for Golgi structure and function. One well established function of membrane-bound ARF1 is to recruit a complex of cytosolic proteins collectively known as coatomers, which, upon oligomerization, result in the formation of "COP1"-coated vesicles. Many downstream effectors of ARF1 have been identified; these include a PIP 2 -dependent phospholipase D (9, 10), arfaptin (11), and the ␤-subunit of coatomer (12). ARF1 has also been shown to regulate PIP 2 levels in cytosol-depleted permeabilized cells (13,14) as well as in Golgi-enriched fractions (15,16). In vitro, Type I PIP 5-kinase can be activated by PA (17), which is the product of phospholipase D. It has therefore been suggested that PIP 2 synthesis by ARF1 is indirect and is dependent on phospholipase D activation (13,14). However, recent studies suggest that ARF1 can stimulate PtdIns(4,5)P 2 synthesis by mediating the recruitment of PI 4-kinase ␤ and an unidentified PIP 5-kinase from the cytosol to the Golgi complex (15). This ability to stimulate PIP 2 synthesis was apparently independent of its activities on coat proteins and PLD (15).
The relationship between ARF and PIP 2 is complicated. In addition to stimulating the synthesis of PIP 2 , ARF1 can directly interact with PIP 2 (18) as can the ARF exchange factors and ARF-GTPase activating protein (GAP). Thus, ARNO, the exchange factor for ARF1, has a PH domain that is recognized by PIP 2 (19), and over-expression results in disruption of the Golgi (20,21). The ARF-GAP, ASAP1, has a PH domain and is specifically activated by PIP 2 in the presence of PA (22). In this study, we have used Golgi membranes and recombinant PIP kinases to examine the role of ARF1 in PIP 2 synthesis. ARF1 was found to increase PtdIns(4,5)P 2 levels dramatically by having two effects: it selectively stimulates the activity of the endogenous PI 4-kinase ␤, and it also directly activates Type I PIP 5-kinase. We conclude that ARF1 has multiple downstream effectors, of which Type I PIP 5-kinase is one.

EXPERIMENTAL PROCEDURES
Preparation of Rat Liver Golgi and Pig Brain Cytosol-Golgi-enriched membranes were prepared exactly as described (23) and kept at Ϫ80°C until required. In brief, one liver was homogenized in 6 volumes of 250 mM sucrose in 20 mM Tris, pH 7.4, 1 mM EGTA, and a mixture of protease inhibitors. The post-nuclear supernatant was obtained after centrifugation at 1,000 ϫ g for 10 min at 4°C and was layered over a 1.3 M sucrose cushion and centrifuged at 200,000 ϫ g for 1 h. The primary interface was recovered (see Fig. 5A), and 1.5 volumes of 2 M sucrose were added. This was transferred to a centrifuge tube, overlaid with 1.1, 0.85, and 0.25 M sucrose, and centrifuged at 200,000 ϫ g for 2 h. Three fractions (F1, F2, and F3) were collected and assayed for galactosyltransferase and ARF-stimulated PLD activity as described previously (24) (see Fig. 5A). F1 was highly enriched in galactosyltransferase and is referred to collectively as the Golgi membranes. Cytosol was obtained from pig brain and was kept at Ϫ80°C and used as required.
Recombinant ARNO, ARF1, and PIP Kinases (Type I and Type II)-Myristoylated ARF1 was made in Escherichia coli and purified exactly as described previously (25). Myristoylation was approximately 10% as determined by mass spectroscopy analysis. ARNO was cloned from RBL-2H3 cells and expressed in E. coli as a His-tagged protein fused to the N terminus, purified by standard techniques. PIP kinases were expressed in E. coli as GST-tagged fusion proteins and purified by standard techniques. The GST was cleaved from the enzyme before use. The Type II enzyme used was the human Type II␣ (26), and the Type I enzyme was murine Type I␣ (27,28).
Determination of Polyphosphoinositide Production in Golgi Membranes-ARF1 was recruited to Golgi membranes (2 g/assay tube) by incubating ARF1 (10 M) and GTP␥S (20 M) in Buffer A (25 mM Hepes, 25 mM NaCl, 2.5 mM MgCl 2 , 0.2 M sucrose, 1 mM dithiothreitol, 1 mM MgATP, 1.6 mg/ml creatine phosphate, and 100 g/ml creatine kinase, pH 7.2) for 15 min at 37°C in an assay volume of 200 l. Membranes were ultracentrifuged (120000 ϫ g) for 30 min. For determination of phosphoinositide kinase activity, the Golgi membranes (2 g/assay tube) were incubated for 10 min in Buffer A supplemented with 2 Ci of ␥-labeled [ 32 P]ATP at 30°C in an assay volume of 100 l final. Cytosol (100 g), BSA control (100 g), and recombinant Type I (250 ng) or Type II PIP kinase (625 ng) were also included in the assay as indicated. At the end of the incubation, membranes were quenched with chloroform:methanol, the lipids extracted and analyzed by thin layer chromatography (TLC) (23), and the plates imaged using Fuji phosphorimaging screens. Data were quantitated using AIDA software.
Analysis of PIP 2 by HPLC-For analysis by HPLC, the assay was scaled up between 30 -50-fold to obtain enough cpm for HPLC analysis. PIP 2 was first isolated by TLC as described above. The PIP 2 spot was located by imaging the plates for 2 h and was then excised from the plate. The silica containing the PIP 2 was transferred into a tube and the PIP 2 deacylated by monomethylamine treatment; the glycerol backbone was subsequently removed exactly as described (29). The resulting inositol trisphosphates were analyzed by anion exchange HPLC on a Partisil 10 SAX column using a gradient of sodium hydrogen phosphate buffers (pH 3.75) (29). The peaks were identified by co-elution with 3 H-labeled Ins(1,4,5)P 3 and Ins(1,3,4)P 3 . 3 H-Labeled Ins(1,4,5)P 3 was obtained from Amersham Pharmacia Biotech and the Ins(1,3,4)P 3 was prepared from Ins(1,3,4,5)P 4 (NEN Life Science Products) as described (30).
Expression of PI 4-Kinase ␣ and ␤ in COS7 Cells and Stimulation by ARF1-Plasmids expressing FLAG-epitope tagged PI 4-kinase ␣ and ␤ were transfected into COS7 cells, and expression was confirmed by Western blotting with anti-FLAG antibody (4,31). Rapidly growing COS7 cells were trypsinized and harvested, and 2 ϫ 10 6 cells were electroporated (using a Bio-Rad Gene Pulser at 0.25 V, 125 microfarad) with 2 g of plasmid DNA and 20 g of herring sperm carrier DNA. Cells were grown for a further 48 h and then harvested, washed three times with phosphate-buffered saline, sonicated in the presence of protease inhibitors, and centrifuged at 2,000 ϫ g for 10 min. The supernatant was then centrifuged at 120,000 ϫ g for 45 min to separate the membrane and cytosol. Phosphoinositide production was determined as described for Golgi membranes, except that COS7 cell membranes were present at 10 g/assay tube.
In Vitro Assay for Measuring ARF1-stimulated PIP 5-Kinase Activity-Activity was analyzed in 100 l of Buffer A (but without creatine phosphate and creatine kinase). Lipid vesicles were prepared by sonication to give final concentrations of 80 g/ml phosphatidylcholine and 8 g/ml phosphatidylinositol 4-phosphate in the final assay. When the effect of PA was examined, it was included in the pre-sonication mix so as to give 8 g/ml in the final assay. All samples also contained 1 mg/ml BSA, 20 g/ml recombinant rat ARNO, 2 Ci of [ 32 P]ATP as well as 10 M ARF1 and GTP␥S (20 M). 250 ng of Type I PIP 5-kinase was added, and the tubes were incubated at 30°C for 20 min unless indicated otherwise. At the end of this time, samples were quenched with chloroform:methanol, and the production of PIP 2 was analyzed by TLC. GTP␥S binding to ARF1 was measured under identical assay conditions to those described above, except radiolabeled ATP was omitted and 35 S-labeled GTP␥S (1 Ci) was included. Only 4% of the ARF was GTP␥S-loaded, and thus the concentration of active ARF is calculated as 400 nM.

RESULTS AND DISCUSSION
We used Golgi membranes from rat liver that have previously been shown to be enriched in PI kinase activity but are devoid of activity that makes PIP 2 (23). Golgi membranes were pre-incubated with ARF1 and GTP␥S for 15 min to recruit ARF1 to membranes and were subsequently tested for PIP formation. PIP levels in these membranes were nearly double that found in membranes treated with a control protein, with no production of PIP 2 (Fig. 1). When cytosol is also included during the kinase reaction, PIP 2 production is greatly facilitated in membranes pretreated with ARF1/GTP␥S compared with no pretreatment or with GTP␥S alone. The production of PIP 2 only in the presence of cytosol implies that cytosol is the source of the PIP kinases. These data using cytosol as a source of the kinases are similar to those reported by Godi et al. (15), who concluded that ARF1 stimulates the synthesis of PIP 2 in Golgi membranes.
To examine whether the increase in PIP was due to increased PI 4-kinase activity rather than recruitment, we replaced the cytosol with purified PIP kinases. The addition of Type I or Type II PIP kinases increased the level of PIP 2 to a small extent in the untreated Golgi membranes. When Golgi membranes were incubated with ARF1 and GTP␥S for 15 min, re-isolated, and incubated with Type I or Type II PIP kinases, FIG. 1. Type I and Type II PIP kinases make PIP 2 in Golgi membranes. Rat liver Golgi membranes were pretreated with or without ARF1 and/or GTP␥S for 15 min at 37°C, ultracentrifuged, and Golgi membranes recovered. The Golgi membranes were assayed for phosphoinositide synthesis in the presence of an irrelevant protein, BSA (100 g), brain cytosol (100 g), Type I PIP 5-kinase (250 ng) or Type II PIP 4-kinase (625 ng) in a 100-l assay. Phosphoinositide synthesis was monitored by measuring the incorporation of [ 32 P]ATP into PIP and PIP 2 for 10 min at 30°C. After the samples were quenched with chloroform:methanol, the lipids were extracted and analyzed by TLC. The data were quantitated as shown in the histogram; a representative lipid image is shown below. The experiment was conducted in duplicate, and the error bars denote the range of the duplicate samples. One experiment representative of three others is shown. PIP 2 levels increased considerably in the presence of Type I, but not Type II, PIP 5-kinase. (Fig. 1).
To analyze the PIP 2 produced by Type I and Type II enzymes, the assay was scaled up. ARF1-pretreated Golgi membranes were incubated with either cytosol, Type I, or Type II PIP kinases, and the PIP 2 formed was isolated by TLC. For the experiment using Type II, the scale-up factor had to be much greater, using a higher specific activity of ATP, to obtain sufficient counts for HPLC analysis. The lipids were deacylated and deglycerated to inositol trisphosphates prior to analysis by HPLC. When cytosol was used to generate PIP 2 , the predominant product was Ins(1,4,5)P 3 (derived from PtdIns(4,5)P 2 ) (Fig. 2b). When Type I PIP 5-kinase was substituted for cytosol, the predominant product was again Ins(1,4,5)P 3 , indicating that PtdIns(4)P must have been phosphorylated by the 5-kinase (Fig. 2c). When Type II PIP 4-kinase enzyme was used, two isomers of inositol trisphosphates were identified, Ins(1,4,5)P 3 and Ins(1,3,4)P 3 (Fig. 2d). Thus, Type II enzyme makes both PtdIns(4,5)P 2 and PtdIns(3,4)P 2 . These products could be derived only from PtdIns(5)P and PtdIns(3)P because Type II enzyme only phosphorylates the inositol at the 4-position.
PtdIns(5)P has only recently been identified as a minor component of mammalian cells (6). Our results indicate that Golgi membranes contain a pool of PtdIns(5)P and also a smaller pool of PtdIns(3)P that is available for conversion to PtdIns(4,5)P 2 and PtdIns(3,4)P 2 . Alternatively, the relatively smaller size of the PtdIns(3,4)P 2 could be because the Type II enzyme is less active toward PtdIns(3)P. Phosphorylation of these minor lipids by Type II PIP 4-kinase is not regulated by ARF1 treatment. In contrast, ARF1 pretreatment leads to an increase in PtdIns(4,5)P 2 levels, which is derived from PtdIns(4)P phosphorylation by Type I PIP 5-kinase (Fig. 1). ARF1 pretreatment of Golgi has two effects: it increases the activity of an endogenous PI 4-kinase, thus increasing PIP levels, and it causes a dramatic increase in PIP 2 levels.
To identify which PI 4-kinase was regulated by ARF1, we next transfected COS7 cells with FLAG-tagged PI 4-kinase ␣ and ␤ and prepared a crude membrane fraction. PI 4-kinase ␤ was found in both the cytosol and membrane fractions, whereas PI 4-kinase ␣ was mainly membrane-bound. Membranes were examined for ARF1-stimulated PI 4-kinase activity. Despite the expression of PI 4-kinases in the transfected cells (Fig. 3), the incorporation of label into PIP under basal activity was not very different than in untransfected cells. This finding is not surprising, because the p55 enzyme that remains uncloned probably accounts for the bulk of PI 4-kinase activity in most cells (3). Only membranes prepared from PI 4-kinase ␤-transfected cells were able to increase their PIP content when incubated with ARF1 and GTP␥S (Fig. 3). We confirmed by Western blot analysis that the Golgi membranes used in this study had endogenous PI 4-kinase ␤ (data not shown). To observe whether ARF1 could directly regulate the PI 4-kinase ␤ activity, we immunoprecipitated the PI 4-kinase ␤ from the cytosol using the FLAG-tag and incubated the immunoprecipitate with PI:PC (1:10) vesicles in the presence of ARF1, ARNO and GTP␥S. We were unable to show a stimulation of PI 4-kinase ␤ activity by ARF1 in vitro (results not shown); this could be because of incorrect assay conditions or the effect of ARF1 on PI 4-kinase ␤ is indirect.
Only in ARF1-pretreated membranes was the Type I PIP 5-kinase able to increase the PIP 2 content, despite the presence of PIP in the membranes. The increase in PIP 2 stimulated by ARF1 is disproportionate to the increase seen with PIP, suggesting that the increase in PIP 2 is not entirely due to the FIG. 2. HPLC analysis of the PIP 2 formed in Golgi membranes with cytosol, Type I PIP 5-kinase, and Type II PIP 4-kinase. The Golgi membranes were pretreated with ARF1 as described for Fig. 1, except the incubations were scaled up to obtain sufficient counts (cpm) for the analysis. The PIP 2 was separated by TLC, deacylated, and deglycerated to remove the fatty acids and the glycerol backbone. The inositol trisphosphates were analyzed by HPLC as described previously (29). a, Ins(1,4,5)P 3 and Ins(1,3,4)P 3 standards. Golgi membranes were analyzed after incubation with cytosol (b), Type I PIP 5-kinase (c), and Type II PIP 4-kinase (d).
FIG. 3. ARF1 stimulates an increase in PI 4-kinase ␤ activity but not PI 4-kinase ␣ activity. COS7 cells were either untransfected or transfected with plasmid for PI 4-kinase ␣ or PI 4-kinase ␤. The cells were sonicated and membranes prepared. The membranes were either untreated or pretreated with GTP␥S alone or with ARF1 plus GTP␥S for 15 min, washed, and assayed for PIP formation. Expression of both proteins was verified by Western blotting using an anti-FLAG antibody. PI 4-kinase ␣ is 230 kDa, and PI 4-kinase ␤ is 97 kDa. The lower band, at 55 kDa, probably reflects a breakdown product.
increase in PIP levels. This suggested that ARF1 could be having a direct effect on PIP 5-kinase. We developed an in vitro assay using PC vesicles with 10% PIP to examine this possibility. ARF1, ARNO, and GTP␥S were added to these vesicles followed by Type I PIP 5-kinase in the presence of MgATP. ARF1 was able to stimulate PIP 5-kinase activity only when GTP␥S or GTP was present (Fig. 4A). The increase in PIP 2 was dependent on the amount of ARF1 added to the assay (Fig. 4B). Furthermore, we also purified native ARF proteins from brain cytosol (10), which was also active in stimulating PIP 5-kinase activity (data not shown).
Honda et al. (32) have recently reported that PIP 5-kinase can also be activated by ARF1 but that this required the presence of PA. In these assay conditions, 50 M PtdIns(4)P was used as substrate, and PA was added at equimolar concentrations. We therefore examined whether PA also stimulates PIP 5-kinase activity in the in vitro assay used here. The vesicles used here have PC as the major lipid, and PtdIns(4)P is included at 10% of PC. When PA was included at an equimolar concentration to the PtdIns(4)P, no stimulation of PIP 5-kinase was observed. In addition, the ARF-stimulated PIP 5-kinase activity was completely inhibited (Fig. 4C). This result was unexpected, and therefore the assay conditions of Honda et al. (32) were used to confirm that PA alone could be stimulatory, and moreover, ARF-dependent stimulation required the presence of PA (Fig. 4D). We conclude that the presence of PC inhibits the stimulation of Type I PIP 5-kinase by PA.
To further clarify whether PA is required in PIP 5-kinase activation in Golgi membranes, we monitored the level of ARF1-stimulated PLD activity, which would be the source of endogenous PA. Fig. 5A illustrates the protocol for preparing Golgi membranes. The material at the primary interface is used to purify the Golgi-enriched fraction (F1). Measurements of galactosyltransferase (a Golgi marker) and ARF-stimulated PLD activity indicate that the Golgi fraction has only a minor amount of PLD activity (Fig. 5, B and C). Thus the level of PA generated in the Golgi fractions used here is small. To assess whether this PA is a contributory factor to PIP 5-kinase activation by ARF1, butanol was used to divert the PA to phosphatidylbutanol. As a control we used, butan-2-ol, which does not participate in the transphosphatidylation reaction. Fig. 5D illustrates that ARF1-stimulated PIP 2 production was unaffected by the presence of butanol.
From our data, we conclude that ARF can directly activate PIP 5-kinase and that PLD-derived PA is not required. This conclusion is supported by three observations. In the in vitro assay used here, PA alone does not activate PIP 5-kinase and potently inhibits ARF1-stimulated PIP 5-kinase activity. Furthermore, Golgi membranes used here that show ARF-stimulated PIP 2 production have only minimal ARF-stimulated PLD activity. Finally, butanol (which would remove any PA derived from PLD activity) is without effect on PIP 2 production.
Whether ARF1 directly regulates PI 4-kinase ␤ still remains to be established. Fig. 6 summarizes the pathways that are After incubation for 20 min, samples were analyzed for PIP 2 formation. A, ARF1 in the presence of GTP␥S or GTP stimulates PIP 5-kinase activity; B, activation of PIP 5-kinase activity is dependent on ARF1 concentration. The active concentration of ARF1 was measured using GTP␥S binding as described under "Experimental Procedures." C, effect of PA on ARF1-stimulated PIP 5-kinase activity in vesicles containing PC. Vesicles were composed of PC:PIP (10:1) or PC:PIP:PA (10:1:1). D, effect of PA on ARF1-stimulated PIP 5-kinase activity in vesicles containing substrate alone or substrate with PA (PIP:PA (1:1)). likely to be operative at the Golgi and indicates the reactions that are regulated by ARF1. The ability of ARF1 to stimulate an increase in PtdIns(4,5)P 2 levels suggests that this lipid has a specific function at this location. A potential function of PIP 2 at the Golgi could be in the recruitment of PIP 2 -binding proteins, which could include spectrin (16)-and oxysterol-binding proteins (33), both of which have motifs that recognize PIP 2 . In addition, PIP 2 may also regulate the ARF1 function via its interactions with ARNO, ARF1, and ARF-GAP. Phosphoinositides are also required for secretion from yeast Golgi (34,35), and the yeast PI 4-kinase, Pik1 has been implicated in this. Moreover, it is interesting to note that strains containing a Pik1 mutation exhibit synthetic lethality with mutant arf1 and sec7 alleles (35). PIP 2 could play a similar function in secretion in the mammalian Golgi. Although the precise requirement for PIP 2 at the Golgi still needs to be clarified, it is clear that the need to critically control the levels of PIP 2 at this location is essential. The levels of PIP 2 at the Golgi are kept in check by a PIP 2 5-phosphatase; a deficiency in this enzyme is the cause of Lowe's syndrome (36).