A Phosphatidylinositol 4-Kinase Pleckstrin Homology Domain That Binds Phosphatidylinositol 4-Monophosphate*

Pleckstrin homology (PH) domains are found in many proteins involved in signal transduction, including the family of large molecular mass phosphatidylinositol (PI) 4-kinases. Although the exact function of these newly discovered domains is unknown, it is recognized that they may influence enzyme regulation by binding different ligands. In this study, the recombinant PI 4-kinase PH domain was explored for its ability to bind to different phospholipids. First, we isolated partial cDNAs of the >7-kilobase transcripts of PI 4-kinases from carrot (DcPI4Kα) andArabidopsis (AtPI4Kα). The deduced primary sequences were 41% identical and 68% similar to rat and human PI 4-kinases and contained the telltale lipid kinase unique domain, PH domain, and catalytic domain. Antibodies raised against the expressed lipid kinase unique, PH, and catalytic domains identified a polypeptide of 205 kDa in Arabidopsis microsomes and an F-actin-enriched fraction from carrot cells. The 205-kDa immunoaffinity-purified Arabidopsis protein had PI 4-kinase activity. We have used the expressed PH domain to characterize lipid binding properties. The recombinant PH domain selectively bound to phosphatidylinositol 4-monophosphate (PI-4-P), phosphatidylinositol 4,5-bisphosphate (PI-4,5-P2), and phosphatidic acid and did not bind to the 3-phosphoinositides. The PH domain had the highest affinity for PI-4-P, the product of the reaction. Consideration is given to the potential impact that this has on cytoskeletal organization and the PI signaling pathway in cells that have a high PI-4-P/PI-4,5-P2 ratio.

Since the first report of changes in phosphoinositide metabolism in response to light (1), there have been many studies of the metabolism of inositol phospholipids in plants (2)(3)(4). These studies reveal two distinguishing features of phosphoinositide metabolism in higher plants: 1) [ 3 H]PI-4-P 1 is 10 -20-fold higher than [ 3 H]PI-4,5-P 2 , and 2) changes in [ 3 H]PI-4-P are detectable. Conversely, in the most responsive animal systems, the ratio of PI-4-P to PI-4,5-P 2 is ϳ1:1, and there is only a transient change in PI-4,5-P 2 , with little to no change in PI-4-P even though inositol 1,4,5-trisphosphate may increase 40-fold (5). One explanation that we are exploring for these differences is that the synthesis of PI-4,5-P 2 is rate-limiting in plant cells because PI-4-P is sequestered by cytoskeletal or other proteins and not readily available for phosphorylation to PI-4,5-P 2 or other metabolic pathways.
A clear distinction between two structurally different isoforms of the PI 4-kinase has emerged from the cloning and sequencing of PI 4-kinase from yeast (14 -16), human (17,18), rat (19), and bovine (20,21) tissues. The two isoforms differ in size, amino acid sequence homology, and putative location and function within the cell. The smaller type encodes a polypeptide of ϳ110 -125 kDa that is found in the cytosol and is associated with the Golgi apparatus (22). The second type encodes a larger protein of ϳ200 -230 kDa that is membrane-associated (19,22).
A distinctive feature of the group of higher molecular mass PI 4-kinases is that they contain putative PH domains. PH domains are poorly conserved protein modules of ϳ100 amino acids in length (23)(24)(25)(26). These motifs exist in proteins that associate with membranes during signal transduction. PH domains bind a variety of ligands ranging from other signal transduction proteins such as G-protein ␤␥ subunits (27) to polyphosphorylated inositol lipids (28 -31) in vitro. N-terminal regions of the PH domain of phospholipase C-␦1 (PLC␦1) (28), ␤-adrenergic receptor kinase (29), ␤-spectrin (30), pleckstrin, Tsk (T-cell-specific kinase), and Ras GTPase-activating protein (31) bind to inositol phospholipids. The affinity of PH domains for PI-4-P, PI-4,5-P 2 , PI-3-P, PI-3,4-P 2 , and PI 3,4,5-trisphosphate varies with the type of protein (32). This means that rapid cellular changes in the levels of the inositol phospholipids could affect the location and regulation of specific PH domaincontaining proteins. Identifying PH domains and their ligand affinities should increase the understanding of how a protein is regulated. The PH domains of the PI 4-kinases have been described based only on primary sequence homology with other PH domains and have not been characterized biochemically.
Here we show biochemical and molecular evidence for a large molecular mass PI 4-kinase in both carrot (DcPI4K␣) and Arabidopsis (AtPI4K␣), and we show for the first time the affinity of a PI 4-kinase PH domain for specific phosphoinositides. The active enzyme was purified using antibodies raised against the * The work was supported by National Science Foundation Grant MCB 9604285. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

MATERIALS AND METHODS
Cloning of Plant PI 4-Kinase-Total RNA was extracted from carrot cells grown in suspension culture (33) by hot borate/phenol/chloroform extraction (34,35). cDNA was synthesized from 5 g of total RNA by Moloney murine leukemia virus reverse transcriptase (Promega) and primed with random hexamers (Boehringer Mannheim) as described (36). Amplification of cDNA by the polymerase chain reaction was achieved with degenerate oligonucleotides deduced from conserved amino acid sequences of known PI 4-kinases. The conserved regions were used by Nakagawa et al. (19) to clone the rat PI 4-kinase. The sequences of these regions are (V/T)GDDCRQ and HIDFGF(M/V). Arabidopsis codon usage was followed to design the degenerate primers, and EcoRI sites were added to the 5Ј-ends of the primers to facilitate subcloning of PCR products into pBluescript (Stratagene). The sequences of the primers, using the International Union of Biochemistry code for degeneracy, were CGG AAT TCR YTG GWG AYG AYT GY-C GTC AR for the sense primer and CGG AAT TCN ATR AAW C-CR AAR TCD ATR TG for the antisense primer. The PCRs containing 10 l of the cDNA synthesis mixture, 20 -25 pmol of each primer, and 5 units of Taq polymerase (Promega) were amplified by 25 cycles of the following: 97°C for 30 s, 45°C for 1 min, and 72°C for 1 min. The PCR products were resolved by agarose gel (1%, w/v) electrophoresis, gelpurified, digested with EcoRI, and subcloned into pBluescript. Sequence of the cDNA was determined by the dideoxy chain termination method (37) using the Sequenase kit (U. S. Biochemical Corp.). After determining the sequence of the PCR product, gene-specific primers were used to amplify cDNA 3Ј and 5Ј to the original PCR product by the rapid amplification of cDNA ends (RACE) method exactly as described previously (38). The Arabidopsis PI 4-kinase was cloned by using the carrot 5Ј-RACE product as a probe to screen an Arabidopsis YES cDNA library (a gift of Dr. Ralph Dewey, North Carolina State University). The probe was labeled with [␣-32 P]dCTP and random hexamers. The plated library (3 ϫ 10 6 clones) was hybridized to the carrot cDNA at 55°C for 16 h in hybridization buffer containing 6ϫ saline/sodium phosphate/EDTA (SSPE; 1ϫ SSPE ϭ 10 mM NaH 2 PO 4 , 1 mM EDTA, and 149 mM NaCl) and 5ϫ Denhardt's solution (100ϫ Denhardt's solution ϭ 2% (w/v) fatty acid-free bovine serum albumin, 2% (w/v) polyvinylpyrrolidone, 2% (w/v) Ficoll 400, 0.5% (w/v) SDS, and 100 g of calf thymus DNA (Sigma)). The final wash was in 1ϫ SSPE and 0.1% (w/v) SDS at 55°C for 1 h. The nylon membrane was exposed to x-ray film for 24 h at Ϫ80°C. Two hybridization-positive clones were carried through three successive screens. The larger clone was 2.5 kb and was used as a probe to screen an Arabidopsis Zap II cDNA library that had been selected for large cDNAs (39). Again, two positive clones were carried through three sequential screening steps, and the largest clone (3.1 kb) was analyzed for its complete sequence by the Iowa State Sequencing Facility.
Northern Blot Analysis-Total RNA was fractionated by formaldehyde-containing agarose gel electrophoresis (40). The RNA was transferred by capillary transfer and UV-cross-linked to nylon membrane and then hybridized with the Arabidopsis cDNA labeled by random priming with [␣-32 P]dCTP. Hybridization was at 42°C in hybridization buffer containing 50% (v/v) formamide. The final wash of the blot was in 0.1ϫ SSPE and 0.1% (w/v) SDS at 65°C for 1 h. The nylon was then exposed to x-ray film for 3 days at Ϫ80°C.
Expression and Purification of the Carrot PI 4-Kinase PH Domain from Escherichia coli-A cDNA encoding the carrot PI 4-kinase PH domain was generated by PCR amplification with primers that flanked either end of the domain sequence. The sequences of the primers were CGG GAT CCC CCC TGG TTA GGC AAC ACA TT (sense) and GGA A-TT CCA ACC TTG AAA ACG CAA GCT T (antisense). The primers contained BamHI (sense primer) and EcoRI (antisense primer) sites on their 5Ј-ends to facilitate directional subcloning into the bacterial expression vector pRSET-A (Invitrogen). The PCR product was gel-purified, digested with BamHI and EcoRI, and ligated into pRSET-A. BL21(DE3) pLys S cells were transformed with the recombinant plasmid, and expression was induced with the addition of isopropyl-␤-Dthiogalactopyranoside (1 mM final concentration) to the cell culture. Bacterial cells expressing the His-tagged PH domain were lysed by sonication and solubilized in 6 M guanidine hydrochloride, and the recombinant polypeptide was purified by metal affinity chromatography using ProBond resin (Invitrogen). Because the PH domain was insoluble, purification was carried out under denaturing conditions with solutions containing 8 M urea. Column fractions were dialyzed sequentially to remove urea and to promote refolding. Tomato eEF-1␣ (a gift from Christine K. Shewmaker, Calgene Inc.) was expressed and purified using the same protocol.
Preparation of Membranes-Total microsomes were prepared from carrot suspension culture cells 5 days after transfer or from whole Arabidopsis thaliana plants. Suspension culture cells were filtered by gravity and homogenized in an equal volume of buffer containing 10 mM KCl, 1 mM EDTA, 1 mM MgCl 2 , 50 mM Tris (pH 7.5), 95 mM LiCl, 2 mM EGTA, polyvinylpolypyrrolidone (0.1 g/g of cells), 8% (w/v) sucrose, 1 mM dithiothreitol, 2 g/ml aprotinin, 0.1 mM phenylmethylsulfonyl fluoride, 1 mg/100 ml leupeptin, and 2 mM benzamidine. Homogenization was with an equal volume of 0.2-mm glass beads in a Virtis homogenizer in 30-s pulses for 2 min at 4°C. Arabidopsis plants were coarsely macerated and ground in a Virtis homogenizer with an equal volume of buffer containing 3 mM EDTA, 2 mM EGTA, 30 mM Tris (pH 7.4), 250 mM sucrose, 14 mM ␤-mercaptoethanol, 2 mM dithiothreitol, 2 g/ml aprotinin, 0.1 mM phenylmethylsulfonyl fluoride, 1 mg/100 ml leupeptin, and 2 mM benzamidine. The homogenate was centrifuged at 2000 ϫ g for 5 min. The resultant supernatant was centrifuged at 40,000 ϫ g for 60 min to obtain a microsomal fraction. Microsomes were resuspended in 30 mM Tris (pH 7.4). F-actin-rich fractions were prepared from the microsomal fraction isolated from 5-day-old carrot suspension culture cells as described previously (10).
AtPI4K␣ Antibody Synthesis-PCR was used to amplify the reading frame of the largest AtPI4K␣ clone. The primers used were CGG-GAT CCG TTC AGT CAC ATA TAT TAG AA (sense) and G GAA T-TC TTA CTT CTC GAT GCC TTG (antisense). An internal BamHI site 45 nucleotides upstream of the region encoding the lipid kinase unique domain and the EcoRI site of the antisense primer allowed the PCR product to be digested with these two enzymes, purified, and ligated into pRSET-B. Expression and purification of the recombinant protein were exactly the same as for the recombinant PH domain described above, except that the protein was not dialyzed. Instead it was concentrated in a Centricon 10 (Amicon, Inc.), and resolved by SDS-PAGE so that ϳ50 g of recombinant protein were present in each lane of the gel. The gel was stained with 0.05% (w/v) Coomassie Brilliant Blue R-250 in water and washed copiously with water until the bands were visible. The bands containing the recombinant protein were excised from the gel and sent to Zeneca LifeScience Molecules for injection into two rabbits (662 and 663). The rabbits were given seven boosts of the recombinant protein over the course of 3 months. Sera from test bleeds and production bleeds were analyzed for cross-reactivity to the recombinant protein by immunoblotting (data not shown).
Antiserum from test bleed 2 of rabbit 662 was purified for IgG on a protein A-Sepharose column (Sigma). Unspecific and His tag-generated antibodies were removed by incubating the purified IgG with an acetone precipitate of E. coli cells expressing His-tagged eEF-1␣ and removing the aggregates by centrifugation (41).
Immunoaffinity Purification of AtPI4K␣-A protein A-Sepharoseantibody affinity column was made by direct coupling with dimethyl pimelimidate (Sigma) as described previously (41). Production bleed antisera from both rabbits (0.5 ml total) were pooled and added to 1 ml of protein A-Sepharose beads in 30 mM Tris (pH 7.4). After coupling the antibody to the beads, the column was washed with 20 bed volumes of 30 mM Tris (pH 8.0). The efficiency of coupling of antibody to protein A beads was analyzed by SDS-PAGE before and after the addition of dimethyl pimelimidate. Heavy chain IgG bands were present at 55 kDa before coupling, but not after.
Arabidopsis microsomes (12.5 mg) were solubilized for 30 min at 4 o C in buffer used to solubilize cytoskeletal proteins (2% (v/v) Triton X-100, 100 mM Tris (pH 7.4), 10 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 150 g/ml leupeptin, and 670 g/ml DNase I) (42). Solubilized microsomes were centrifuged at 40,000 ϫ g for 30 min. The supernatant was incubated with antibody-coupled beads in a 10 ϫ 33-mm column resuspended in 1 bed volume of 100 mM Tris (pH 7.4) at 4°C with shaking for 2 h. The bed was allowed to settle, and the flow-through fraction was collected. The column was washed with 10 bed volumes of 100 mM Tris (pH 7.4). In addition, to prepare the column for elution and to ensure adequate washing, the column was washed with 10 mM phosphate buffer (pH 7.0) until the flow-through fraction had a spectrophotometric absorbance reading of Ͻ0.010 at 280 nm. Bound proteins were eluted with 100 mM glycine (pH 3.0). Fractions of 1 bed volume were collected, immediately neutralized with 0.05 volume of 1 M phosphate (pH 8.0), and analyzed by SDS-PAGE and immunoblotting and for PI 4-kinase activity.
Kinase Assay-Each fraction eluted from the immunoaffinity column was assayed in duplicate (60 l/assay) to determine the PI 4-kinase activity. The reaction mixture contained final concentrations of 7.5 mM MgCl 2 , 1 mM sodium molybdate, 0.5 mg/ml PI, 0.1% (v/v) Triton X-100, 0.9 mM ATP, 30 mM Tris (pH 7.2), and 20 Ci of [␥-32 P]ATP (7000 Ci/mmol) in a total volume of 100 l. Stock PI (5 mg/ml) was solubilized in 1% (v/v) Triton X-100. The reactions were incubated at 25°C for 2 h with intermittent shaking. The reactions were stopped with 1.5 ml of ice-cold CHCl 3 /MeOH (1:2) and kept at 4°C until the lipids were extracted. Lipids were extracted as described previously (9). Extracted lipids were vacuum-dried, solubilized in CHCl 3 /MeOH (2:1), and spotted onto Whatman LK5D silica gel plates that had been completely dried in a microwave oven for 5 min after presoaking in 1% (w/v) potassium oxalate for 80 s. The lipids were separated in either a CHCl 3 /MeOH/NH 4 OH/H 2 O (86:76:6:16) solvent system (9) or a borate/ pyridine-based solvent (43) and quantitated with a Bioscan System 500 Imaging Scanner. The plates were subsequently exposed to a phosphor screen for 2 days and visualized by a Storm PhosphorImager (Molecular Dynamics, Inc.).
Immunoprecipitation-Carrot microsomes were solubilized in 1% (v/v) Triton X-100 at 4°C overnight. To preclear unspecific antibodies and antigens, solubilized microsomes were incubated for 1 h on ice with 0.2 volume of preimmune serum from rabbit 662. 0.33 volume of protein A-Sepharose was added and incubated for 1.5 h at 4°C with shaking. The beads were pelleted, and the supernatant was saved and used for subsequent steps. Immunoprecipitation, as described previously (41), was carried out by adding to the supernatant either preimmune serum or partially purified antiserum from test bleed 2 of rabbit 662 (described above). The protein A-antibody-antigen complex was washed three times with 30 mM Tris (pH 7.2) and resuspended in SDS-PAGE sample buffer for subsequent analysis by electrophoresis and immunoblotting.
Immunoblotting-Protein was separated by SDS-PAGE using 8% (w/v) polyacrylamide and transferred onto polyvinylidene difluoride membrane (44). The membrane was incubated for 1 h with anti-AtPI4K␣ antiserum (1:1000) at 25°C. Cross-reactivity was detected by incubation with goat anti-rabbit IgG, F(abЈ) 2 conjugated to horseradish peroxidase (Pierce), and subsequent chemiluminescent detection by exposing the blot to X-Omat AR film (Eastman Kodak Co.) for the amount of time indicated.
Fat Western Blotting-Phospholipids were from Sigma, except for PI-3-P and PI-3,4-P 2 , which were from Matreya, Inc., and NBD-labeled PA, NBD-labeled phosphatidylcholine, and rhodamine-labeled phosphatidylethanolamine, which were from Avanti Polar Lipids. Phospholipids were solubilized in chloroform as stock solutions of 1 mg/ml. A minimum of 10 l containing 0.5, 1.0, or 5.0 g of lipid were spotted onto nitrocellulose (NitroBind, MSI) at a time. The membrane and lipids were dried at 24°C for 1 h. The nitrocellulose was incubated with 3% (w/v) fatty acid-free bovine serum albumin (isolated by cold ethanol precipitation; Sigma A-6003) in TBST (10 mM Tris (pH 8.0), 140 mM NaCl, and 0.1% (v/v) Tween 20) for 1 h and then placed in a solution containing the His-tagged fusion proteins (PH domain or eEF-1␣) diluted in TBST (0.5 g/ml) at 4°C overnight with shaking. The nitrocellulose was washed with TBST three times for 10 min each and then incubated with T7 tag monoclonal antibody (Novagen) to the His-tagged region diluted 1:10,000 in TBST for 1 h at 24°C. The nitrocellulose was washed three times for 10 min in TBST at 24°C and then incubated with goat anti-mouse IgG conjugated to horseradish peroxidase (Pierce) at a titer of 1:30,000 in TBST for 1 h at 24°C. The nitrocellulose was washed again in TBST three times for 10 min and then incubated for 5 min in a 1:1 mixture of peroxidase substrate and luminol/enhancer (Pierce) for subsequent chemiluminescent detection. The nitrocellulose was exposed to X-Omat AR film for 0.5-5 min as indicated.

RESULTS
Cloning of AtPI4K␣ and DcPI4K␣-Using degenerate primers based on highly conserved regions of yeast PIK1 and STT4 PI 4-kinase lipid kinase domains, we amplified a cDNA of 394 nucleotides from carrot RNA. Sequence analysis of this PCR product showed that the deduced amino acid sequence was 50% identical and 64% similar to yeast STT4 and only 35% identical and 54% similar to Arabidopsis PI 3-kinase (AtVPS34). 5Ј-and 3Ј-RACE were used to amplify regions of sequence beyond this initial PCR product. The 3Ј-RACE product was 1.1 kb in size and spanned the rest of the catalytic domain and extended through the 3Ј-untranslated region to the poly(A) tail. The 5Ј-RACE product contained 1.4 kb and had sequence homology to the lipid kinase unique domain and PH domain of the other PI 4-kinases reported in GenBank TM . The 5Ј-RACE product from carrot was used as a probe to screen an Arabidopsis YES cDNA library. Two positives were carried through three sequential screens of the plated phage. The two clones were 2.4 and 2.1 kb in size. The larger clone was partially sequenced and used as a probe to screen another Arabidopsis library that contained large cDNAs. Two clones were carried through this screening process, and the largest was ϳ3.1 kb in size. This larger clone was completely sequenced and is shown in Fig. 1. This clone contains the 3Ј-untranslated region and spans the protein coding region up to ϳ1635 nucleotides upstream of the deduced lipid kinase unique domain. Fig. 2 is a linear representation of the deduced amino acid sequences and the homology between the known PI 4-kinases and the plant sequences reported here. The region of sequence containing the lipid kinase unique domain, PH domain, and catalytic lipid kinase domain is 41% identical to rat and human PI4K␣ and yeast STT4 and only 24% identical to Arabidopsis PI 3-kinase in these regions. Based on this sequence homology alone, this cDNA is predicted to encode a PI 4-kinase in A. thaliana.
Because there was no evidence for a start site in the cDNA sequence and because the homology of this sequence was most similar to the larger structural class of PI 4-kinases, Northern blotting was performed to determine the size of the transcript for this gene. The AtPI4K␣ cDNA hybridized with a single Arabidopsis transcript that is ϳ7 kb in size as shown in Fig. 3. Northern blot analysis of carrot RNA using the 5Ј-RACE product as a probe also revealed a single band at ϳ7.5-8.0 kb (data not shown). These sizes coincide with the size of the transcripts for the larger PI 4-kinases identified in yeast, rat, and bovine tissues.
AtPI4K␣ and DcPI4K␣ Encode Large Molecular Mass PI 4-Kinases-To further substantiate that the cDNA isolated encoded a PI 4-kinase, we generated antibodies to the recombinant polypeptide expressed from the AtPI4K␣ cDNA in E. coli. The polypeptide spanned between the regions just Nterminal to the lipid kinase unique domain down to the C terminus of the catalytic domain. This antibody was used in several different strategies shown in Figs. 4 and 5 to identify the size of the endogenous carrot and Arabidopsis PI 4-kinases.
When partially purified antibody was used to immunoprecipitate the PI 4-kinase from solubilized carrot microsomes, a 205-kDa polypeptide was found in the precipitate that crossreacted with AtPI4K␣ antibodies on Western blots (Fig. 4A). When preimmune serum was used, the 205-kDa polypeptide did not immunoprecipitate (Fig. 4A). More important, the 205-kDa peptide was present in an F-actin fraction purified from the carrot microsomes (Fig. 4B). The immunoblot reveals that despite the presence of many proteins in the fraction, only one band at 205 kDa was recognized by the antibody. This confirms earlier work that showed PI 4-kinase activity enriched in Factin fractions of carrot cells (10,11).
To verify that the 205-kDa protein recognized by the antibodies was a PI 4-kinase, we purified endogenous PI 4-kinase from Arabidopsis microsomes by immunoaffinity chromatography. SDS-PAGE and immunoblot analysis of the eluted fractions are shown in Fig. 5 along with the corresponding PI 4-kinase activity. A polypeptide of 205 kDa eluted from the column and cross-reacted with AtPI4K␣ antiserum (Fig. 5, A  and 5B). Fraction 2, which had the largest amount of the 205-kDa polypeptide, had the highest PI 4-kinase activity as shown by the formation of 32 P-labeled PI-4-P on the image of the TLC plate (Fig. 5C). When the reaction products were analyzed in solvents that separate PI-4-P and PI-3-P (43), only PI-4-P was detected (data not shown).
Recombinant PI 4-Kinase PH Domain Specifically Binds PI-4-P-A putative PH domain was initially described in the human PI4K␣ sequence (17). The PH domain is found in all the large molecular mass PI 4-kinases sequenced thus far and is the distinguishing feature of the family of large PI 4-kinases (Fig. 2). Fig. 6 shows the sequence comparison of this domain in the human, rat, STT4, bovine, carrot, and Arabidopsis sequences. Carrot and Arabidopsis PH domains are 40 -43% identical and 63-66% similar to those found in the other large molecular mass PI 4-kinases.
To determine whether or not the putative PH domain of the carrot PI 4-kinase encoded a lipid-binding polypeptide, we expressed the PH domain as a fusion protein with a polyhistidine tag in E. coli. The recombinant PH domain was purified by affinity chromatography and renatured by dialysis. The ability of the recombinant polypeptide to bind different phospholipids FIG. 1. cDNA and deduced amino acid sequence of AtPI4K␣. The corresponding amino acids are listed below the nucleotide sequence. was studied using a newly developed technique called Fat Western blotting, which is a modification of Western blotting and a previously developed lipid-protein binding assay (45). This involves first binding lipids to nitrocellulose and then using blotting conditions where the lipid is neither hydrolyzed nor lost from the blot. To determine whether or not the phospholipids would remain bound to the nitrocellulose after extensive washing during the procedure, we spotted NBD-labeled phosphatidylcholine and PA as well as rhodamine-labeled phosphatidylethanolamine onto nitrocellulose. The nitrocellulose was incubated in TBST at 25°C with shaking. After 24 h of incubation, there was no apparent loss of the uniform layer of lipid. Additionally, when unlabeled PA, phosphatidylcholine, phosphatidylethanolamine, PI, PI-4-P, and PI-4,5-P 2 were treated in the same way, after extensive washing, they all remained clearly bound to the nitrocellulose based on visualization with iodine (data not shown). Fig. 7A shows that when Fat Western blotting was performed with the recombinant carrot PH domain, the PH domain bound PA, PI-4-P, and PI-4,5-P 2 , but not phosphatidylcholine, phosphatidylethanolamine, or PI. Clearly, the PH domain bound to PI-4-P with higher affinity than to PA or PI-4,5-P 2 at the lowest concentration of lipid. As a control for unspecific binding of protein or the poly-His tag to lipids, we used recombinant eEF-1␣ fused to the same polyhistidine tag as the PH domain. The His-tagged eEF-1␣-expressed protein did not bind any of the lipids (Fig. 7B).
When the 3-phosphoinositides were used, the PH domain did not bind PI-3-P except very weakly at the highest concentration (5 g) of lipid and did not bind PI-3,4-P 2 at any of the concentrations used (Fig. 7C) . Additionally, the recombinant carrot PH domain that had been denatured by boiling did not bind to any of the lipids, including PI-4-P even at the highest concentration of lipid (data not shown). DISCUSSION We have cloned and characterized a 205-kDa PI 4-kinase from higher plants. The deduced amino acid sequences of the plant PI 4-kinase clones are 41% identical and 68% similar to rat and human PI4K␣ lipid kinase unique, PH, and lipid kinase catalytic domains. Using polyclonal antibodies raised against recombinant AtPI4K␣, we verified that this PI 4-kinase isoform is found in an F-actin-rich fraction. More important, in this study, we characterized the PI 4-kinase PH domain and found that it bound selectively to PA, PI-4-P, and PI-4,5-P 2 and did not bind to similarly charged PI-3-P and PI-3,4-P 2 . PI 3,4,5trisphosphate was not analyzed because there is no evidence for the presence of this lipid in plants. Denaturation of the expressed PH domain abolished its ability to bind to these lipids. Together, these data suggest that binding to PA, PI-4-P, and PI-4,5-P 2 is not due simply to an unspecific electrostatic interaction. ␤-Spectrin, a cytoskeletal protein that attaches actin filaments to the plasma membrane, also has a PH domain that binds specifically to phosphoinositides phosphorylated on the D-4 and D-5 positions of the inositol ring (30). Although some PH domains have been shown to bind phosphoinositides phosphorylated on the D-3 position (32), such as that of PKB/ Akt (46), the PH domain of the plant PI 4-kinase has a clear preference for 4-phosphoinositides.
PH domains have poorly conserved amino acid sequences, but the tertiary structures of these motifs are very similar. Of the many PH domains previously identified, only one residue, a tryptophan near the C terminus, is conserved. It has been suggested that the C-terminal regions of the PH domains bind other signaling proteins. Touhara et al. (27) have shown that the C-terminal regions of the PH domains of Ras guanine nucleotide-releasing factor and ␤-adrenergic receptor kinase overlap with a G-protein ␤␥ subunit-binding site. Binding of the PH domain of Btk to G␤␥ also occurs at the C-terminal end and includes the ␣-helix in which the invariant tryptophan occurs (47). Interestingly, the newly characterized DcPI4K␣, AtPI4K␣, and Saccharomyces cerevisiae PLC␦ 2 PH domains do not contain this conserved tryptophan, yet still bind polyphosphoinositides.
There are many differences in the regulation of the PI pathway between species. The sequence changes in the PH domain might contribute to these differences. For instance, the plant PLC␦ isoform that has been cloned does not contain a PH domain (48,49), whereas human PLC␦1 contains one at its N terminus (28), and as mentioned, the S. cerevisiae PLC␦ isoform contains a PH domain at its amino terminus. S. cerevisiae PLC␦, however, does not contain the conserved C-terminal tryptophan. These differences in the PLC PH domains of different species provide an important means for regulation of the isoforms. For example, when the human PLC␦1 PH domain binds PI-4,5-P 2 , the activity of PI hydrolysis is enhanced 9-fold (28). As inositol 1,4,5-trisphosphate is formed, it competes for the PI-4,5-P 2 -binding site and eliminates the PI-4,5-P 2 -enhanced stimulation of activity. Deletions of at least 20 amino acids at the amino-terminal end of the PLC␦1 PH domain abolished the stimulatory effect of PI-4,5-P 2 binding. Enzyme activity, however, was not eliminated by deletion of the entire PH domain. In this case, the PH domain regulates enzyme activity by binding to membrane phospholipids and allowing a close association of the enzyme with its substrate. At the same time, the PH domain provides a mechanism for feedback control of enzyme activity by binding to the product, inositol 1,4,5trisphosphate, which displaces the bound PI-4,5-P 2 and causes the enzyme to fall away from the membrane and its substrate.
For PLC␦1, the PH domain is at the amino-terminal end of the enzyme, removed from the catalytic site. In contrast with PLC␦1, the PI 4-kinase PH domain is positioned directly adjacent to the catalytic domain, which is at the carboxyl-terminal end of the protein. This close proximity of the PH domain to the catalytic domain implies that its role in PI 4-kinase function is different from that in PLC␦1, as might be expected because PI 4-kinase is involved in synthesis and PLC is involved in catabolism of the lipids. 2 B. Amidon and J. Flick, personal communication.
FIG. 6. Amino acid sequence alignment of PI 4-kinase PH domains of Arabidopsis, carrot, yeast STT4, rat, human PI4K␣, and bovine brain PI4K200. Dots indicate gaps between sequence added to optimize alignment. Dashes indicate identical residues compared with one or both plant sequences. The conserved tryptophan (W) is indicated in boldface.

FIG. 7. Fat Western blots of the recombinant carrot PI 4-kinase
PH domain with phospholipids. Phospholipids are indicated above each blot, and the amount of lipid spotted onto nitrocellulose is shown to the left of each row of lipid. A, the blot was incubated with 0.5 g/ml recombinant carrot His-tagged PH domain. The exposure time was 5 s. The blot shown is representative of six experiments that showed the same binding characteristics. B, the blot was incubated with 0.5 g/ml recombinant His-tagged eEF-1␣. The exposure time was 5 min. C, the blot was incubated with 0.5 g/ml His-tagged PH domain. The exposure time was 5 s. The blot shown is representative of two experiments that showed the same binding characteristics. PC, phosphatidylcholine; PE, phosphatidylethanolamine; PIP, PI-4-P; PIP 2 , PI-4,5-P 2 .
More important, the carrot PI 4-kinase PH domain binds PI-4-P more specifically than PI-4,5-P 2 which raises the question as to why the enzyme would bind to its product. One scenario would be that the cytoskeleton-associated PI 4-kinase synthesizes PI-4-P and immediately binds it through the PH domain. In this manner, the PI 4-kinase would sequester PI-4-P in the cytoskeleton and protect it from binding to actinbinding proteins such as gelsolin and profilin. The polyphosphoinositides promote actin polymerization by binding to gelsolin and profilin (50,51). To increase actin polymerization, gelsolin and profilin would have to compete with PI 4-kinase for PI-4-P. Therefore, the PH domain of the PI 4-kinase could be the first site for regulating actin polymerization by the phosphoinositides. Similarly, the lipid transfer proteins, phospholipases, or other kinases would have to competitively bind and displace the PI 4-kinase. If the PI 4-kinase synthesizes PI-4-P and remains bound to PI-4-P, protecting it from further metabolism, then this could account for the large ratio of PI-4-P to PI-4,5-P 2 (10:1) that exists in higher plant membranes (52). In addition, by regulating the biosynthesis of PI-4,5-P 2 , PI 4-kinase would have an important impact on the responsive state of the cell.