The protein kinase C-dependent phosphorylation of serine 166 is controlled by the phospholipid species bound to the phosphatidylinositol transfer protein alpha.

The charge isomers of bovine brain PI-TPalpha (i.e. PI-TPalphaI containing a phosphatidylinositol (PI) molecule and PI-TPalphaII containing a phosphatidylcholine (PC) molecule) were phosphorylated in vitro by rat brain protein kinase C (PKC) at different rates. From the double-reciprocal plot, it was estimated that the V(max) values for PI-TPalphaI and II were 2.0 and 6.0 nmol/min, respectively; the K(m) values for both charge isomers were about equal, i.e. 0.7 micrometer. Phosphorylation of charge isomers of recombinant mouse PI-TPalpha confirmed that the PC-containing isomer was the better substrate. Phosphoamino acid analysis of in vitro and in vivo (32)P-labeled PI-TPalphas showed that serine was the major site of phosphorylation. Degradation of (32)P-labeled PI-TPalpha by cyanogen bromide followed by high pressure liquid chromatography and sequence analysis yielded one (32)P-labeled peptide (amino acids 104-190). This peptide contained Ser-148, Ser-152, and the consensus PKC phosphorylation site Ser-166. Replacement of Ser-166 with an alanine residue confirmed that indeed this residue was the site of phosphorylation. This mutation completely abolished PI and PC transfer activity. This was also observed when Ser-166 was replaced with Asp, implying that this is a key amino acid residue in regulating the function of PI-TPalpha. Stimulation of NIH3T3 fibroblasts by phorbol ester or platelet-derived growth factor induced the rapid relocalization of PI-TPalpha to perinuclear Golgi structures concomitant with a 2-3-fold increase in lysophosphatidylinositol levels. This relocalization was also observed for Myc-tagged wtPI-TPalpha expressed in NIH3T3 cells. In contrast, the distribution of Myc-tagged PI-TPalpha(S166A) and Myc-tagged PI-TPalpha(S166D) were not affected by phorbol ester, suggesting that phosphorylation of Ser-166 was a prerequisite for the relocalization to the Golgi. A model is proposed in which the PKC-dependent phosphorylation of PI-TPalpha is linked to the degradation of PI.

The charge isomers of bovine brain PI-TP␣ (i.e. PI-TP␣I containing a phosphatidylinositol (PI) molecule and PI-TP␣II containing a phosphatidylcholine (PC) molecule) were phosphorylated in vitro by rat brain protein kinase C (PKC) at different rates. From the doublereciprocal plot, it was estimated that the V max values for PI-TP␣I and II were 2.0 and 6.0 nmol/min, respectively; the K m values for both charge isomers were about equal, i.e. 0.7 M. Phosphorylation of charge isomers of recombinant mouse PI-TP␣ confirmed that the PC-containing isomer was the better substrate. Phosphoamino acid analysis of in vitro and in vivo 32 P-labeled PI-TP␣s showed that serine was the major site of phosphorylation. Degradation of 32 P-labeled PI-TP␣ by cyanogen bromide followed by high pressure liquid chromatography and sequence analysis yielded one 32 P-labeled peptide (amino acids 104 -190). This peptide contained Ser-148, Ser-152, and the consensus PKC phosphorylation site Ser-166. Replacement of Ser-166 with an alanine residue confirmed that indeed this residue was the site of phosphorylation. This mutation completely abolished PI and PC transfer activity. This was also observed when Ser-166 was replaced with Asp, implying that this is a key amino acid residue in regulating the function of PI-TP␣. Stimulation of NIH3T3 fibroblasts by phorbol ester or platelet-derived growth factor induced the rapid relocalization of PI-TP␣ to perinuclear Golgi structures concomitant with a 2-3-fold increase in lysophosphatidylinositol levels. This relocalization was also observed for Myc-tagged wtPI-TP␣ expressed in NIH3T3 cells. In contrast, the distribution of Myc-tagged PI-TP␣(S166A) and Myc-tagged PI-TP␣(S166D) were not affected by phorbol ester, suggesting that phosphorylation of Ser-166 was a prerequisite for the relocalization to the Golgi. A model is proposed in which the PKC-dependent phosphorylation of PI-TP␣ is linked to the degradation of PI.
Phosphatidylinositol transfer protein (PI-TP) 1 is a ubiqui-tous protein that has been shown to play an essential role in secretion (vesicle flow) and in phospholipase C-dependent signaling as was established in reconstituted systems (1)(2)(3)(4). In addition, PI-TP may have a role in delivering substrate to the phosphatidylinositol 3-kinase complex (5,6). Two isoforms of PI-TP have been identified, PI-TP␣ and PI-TP␤ (7)(8)(9). Both isoforms transfer phosphatidylinositol (PI) and phosphatidylcholine (PC) between membranes in vitro (10). PI-TP␤ expresses an additional activity for sphingomyelin (7,11). In studies with the reconstituted systems, PI-TP␣ and ␤ behaved similarly. Recently, PI-TP␣ overexpressed in NIH3T3 cells was shown to enhance the constitutive levels of lysophosphatidylinositol (lysoPI) (12). Overexpression of PI-TP␤ in these cells had no effect on lysoPI formation but stimulated the conversion of ceramide into sphingomyelin (13). In addition to these soluble PI-TPs, a membrane-bound form of PI-TP was detected containing a PI-TP␣ homology domain at the N terminus (amino acids 1-257) and six putative membrane-spanning domains. This retinal degeneration B (RdgB) protein was originally identified in Drosophila (14).
Localization studies by indirect immunofluorescence and by microinjection of fluorescently labeled PI-TPs have shown that PI-TP␣ is mainly localized in the nucleus and in the cytosol and that PI-TP␤ is mainly associated with the Golgi membranes (7,8). Upon stimulation of Swiss mouse 3T3 fibroblasts by growth factors that activate the PI signaling pathway or by phorbol 12-myristate 13-acetate (PMA), PI-TP␣ became rapidly associated with the Golgi membranes. Under these conditions PI-TP␣ was found to be phosphorylated, suggesting that this modification may be a prerequisite for its association with the Golgi. (15,16). PI-TP␣ was also a substrate for protein kinase C (PKC) in vitro in agreement with the presence of five putative phosphorylation sites: Thr-59, Thr-169, Thr-198, Thr-251, and Ser-166 (15,16).
Two charge isomers of PI-TP␣ are present in tissues and cells of which one isomer carries a PI molecule (PI-TP␣I) and the other a PC molecule (PI-TP␣II) (17). The cellular concentration of PI-TP␣I is about 8-fold higher as compared with that of PI-TP␣II. Because the affinity of PI-TP␣ for PC is about 16-fold lower than the affinity for PI, the relative amounts of the two isomers reflect the accessible pools of PI and PC in the cell. To date, no comparable charge isomers of PI-TP␤ have been detected. The physiological significance of the two charge isomers of PI-TP␣ is not yet known. It has been suggested that the yeast analogue of PI-TP, SEC14p, acts as a sensor of these phospholipid pools in the Golgi inasmuch as that the two charge isomers affect PI and PC metabolism differently (18). Given the ability of PI-TP␣ to be involved in both PI and PC metabolism (12,19), one may expect a regulatory mechanism in the cell to be able to discriminate between PI-TP␣I and II. Because PI-TP␣ is a substrate for PKC, we have investigated whether the phospholipid bound to PI-TP␣ has an effect on the in vitro phosphorylation. It will be shown that PI-TP␣II is more rapidly phosphorylated by PKC than PI-TP␣I. Both charge isomers have one major phosphorylation site (Ser-166), replacement of which with Ala or Asp completely abolished the transfer activity. A model will be presented in which the phosphorylation of PI-TP␣ is linked to the agonist-induced production of lysoPI.

EXPERIMENTAL PROCEDURES
Materials-Egg yolk PC, soybean PI, phosphatidic acid, phosphatidylserine, PMA, ATP, phosphoserine, phosphothreonine, and phosphotyrosine were obtained from Sigma. The pBluescript SK Ϫ vector and the Quickchange site-directed mutagenesis kit were purchased from Stratagene (La Jolla, CA). The oligonucleotides were synthesizsed by Eurogentec, Belgium. The pET-15b vector was obtained from Novagen (Madison, WI). The Escherichia coli strain BL21(DE3) was obtained from Dr. J. H. Veerkamp (Department of Biochemistry, University of Nijmegen, Nijmegen, The Netherlands). Isopropyl-␤-D-thiogalactopyranoside was purchased from Promega (Madison, WI). Ni 2ϩ -High Bond matrix was from Invitrogen (San Diego, CA). [␥-32 P]ATP (3000 Ci/mmol) was obtained from Amersham Pharmacia Biotech. Cellulose TLC plates and TPCK-trypsin were purchased from Merck KGaA. The FuGENE6 Transfection Reagent and the anti-c-Myc monoclonal antibody were from Roche Molecular Biochemicals.
Purification of Protein Kinase C-PKC was purified from rat brain by a modified procedure previously described by Huang et al. (20). Rat brains (20 -40 g of tissue) were homogenized, and the cytosolic fraction was subsequently purified on DEAE-Sepharose, Sephacryl 200 and phenyl-Sepharose columns. The purified enzyme has a specific activity of 200 nmol of phosphate/min/mg protein when assayed with histone IIIs as substrate. The purified enzyme is stable for several months when kept at Ϫ80°C in 50% glycerol and 0.01% Triton X-100.
Purification of Bovine PI-TP␣I and II and of Recombinant Mouse PI-TP␣-PI-TP␣I and II were purified from bovine brain cytosol as described by van Paridon et al. (17). The cDNA encoding mouse PI-TP␣ was expressed in E. coli, and the protein was purified as described by Geijtenbeek et al. (16).
Preparation of Recombinant PI-TP␣ I and II-The mouse recombinant PI-TP␣ (recPI-TP␣) purified from E. coli contains one molecule of phosphatidylglycerol (PG) (16). To exchange the PG molecule for a PI or PC molecule, recPI-TP␣ (28 nmol) was incubated with PI (385 nmol) present in unilamellar vesicles consisting of PI:PC (30:70 mol %) or with PC (900 nmol) present in vesicles consisting of PC:phosphatidic acid (70:30 mol %). Vesicles were prepared in 20 mM Tris buffer, pH 7.6, 50 mM NaCl by sonication under nitrogen for 10 min. Vesicles and protein were incubated for 15 min at 37°C. The reaction was stopped by the addition of MgCl 2 to a final concentration of 5 mM. Protein and vesicles were separated on a DEAE-cellulose column that was equilibrated in 20 mM Tris, pH 7.6, 50 mM NaCl, 5 mM MgCl 2 . The recPI-TP␣ eluted in the run-through and the negatively charged vesicles were retarded on the column (17). The protein content was determined using the Bradford assay (21). The products of the exchange reaction were analyzed by isoelectric focussing (16).
Site-directed Mutagenesis of Recombinant PI-TP␣-The PI-TP␣ cDNA cloned into the pBluescript vector (pBlue-wtPI-TP␣) (16) was used for site-directed mutagenesis using the Quickchange site-directed mutagenesis method according to the manufacturer's instruction (Stratagene). Ser-166 was replaced by Ala using the following mutagenic oligonucleotides: sense primer, 5Ј-CCAGCAAAATTTAAGGCT-GTCAAAACAGGACGC-3Ј; antisense primer, 5Ј-GCGTCCTGTTTTGA-CAGCCTTAAATTTTGCTGG-3Ј. The bold nucleotides encode the mutated amino acid (Ser-166 to Ala-166), and the underlined nucleotide is a mutation that does not result in a change in amino acid composition, but it deletes a DraI restriction site. Incorporation of the mutagenic oligonucleotides into the construct (pBlue-PI-TP␣(S166A) was checked by restriction enzyme analysis and by sequencing. Both the mutated and wtPI-TP␣ cDNAs were cloned into the pET-15b expression vector. Expression of these constructs yielded wtPI-TP␣ or PI-TP␣(S166A) fused to an N-terminal peptide containing six histidine residues. A mutant PI-TP␣ in which Ser-166 was replaced with Asp-166 was obtained in a similar way using the above oligonucleotides except that the bold nucleotides were replaced for GAT (sense primer) and ATC (antisense primer).
Purification of wild type and Mutant His 6 -tagged PI-TP␣-The E. coli strain BL21(DE3) was transformed by either the wtPI-TP␣-pET15b or the mutant PI-TP␣-pET-15b construct. A 10-ml culture of each transformant grown overnight in LB medium containing 50 g/ml ampicillin was used to inoculate 1 liter of LB medium (also containing 50 g/ml ampicillin). Bacteria were grown at 37°C. At an A 600 of 0.8, the cultures were induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside and grown for an additional 3 h. His 6 -tagged wtPI-TP␣ or His 6 -tagged mutant PI-TP␣ were purified from these cultures. All further manipulations were performed at 4°C. Bacteria were harvested by centrifugation at 5,000 ϫ g for 30 min. The pellet was resuspended in 50 ml of STE buffer (0.5 M NaCl, 10 mM Tris/HCl, 0.1 mM EDTA, pH 7.5) and sonicated for 5 ϫ 30 s at 80 W output with a macrotip of a Branson Sonifier B12. The homogenate was centrifuged at 17,500 ϫ g. The supernatant was collected and extensively dialyzed against buffer A (18.3 mM Na 2 HPO 4 , 1.3 mM NaH 2 PO 4 , 500 mM NaCl, pH 7.8). The dialyzed supernatant was applied to a Ni 2ϩ -High Bond column (20 ml) and eluted with a linear gradient (200 ml) of 0 -500 mM imidazol in buffer A (flow rate, 0.5 ml/min; fractions, 5 ml). The fractions were assayed for PI-TP␣ by measuring PI transfer activity and by Western blotting using an antibody against PI-TP. Purified His 6 -tagged wtPI-TP␣ and His 6tagged mutant PI-TP␣ were used for the in vitro phosphorylation experiments.
Phospholipid Transfer Activity Assay-PI and PC transfer activities were determined in a continuous fluorescence assay using 2-pyrenyldecanoyl-PI or -PC as substrates. (11,17). Measurements were performed using a fluorimeter (Photon Technology International) equipped with a thermostated cuvette holder and a stirring device.
Phosphorylation of PI-TP in Vitro by Protein Kinase C-PI-TP␣ (0.1-5 g) was phosphorylated in a reaction volume of 60 l containing 20 mM Tris/HCl, pH 7.5, 7.5 mM magnesium acetate, 10 g/ml leupeptin, 10 M ATP, and 1-2 Ci of [␥-32 P]ATP. The Ca 2ϩ /phospholipidindependent phosphorylation was determined in the presence of 1 mM EGTA, and the Ca 2ϩ /phospholipid-dependent phosphorylation was determined in the presence of 1 mM Ca 2ϩ , 96 g/ml phosphatidylserine, and 3.2 g/ml diacylglycerol. The mixture was incubated for 10 min at 37°C, and the reaction was terminated by the addition of 600 l of cold acetone. Bovine serum albumin (1 g) was added, and after 30 min on ice, the precipitated protein was spun down, dissolved in sample buffer (125 mM Tris/HCl, pH 6.8, 5% (w/v) SDS, 12.5% (v/v) 2-mercaptoethanol, and 10% (v/v) glycerol) and analyzed by SDS-PAGE (15% gel) followed by autoradiography.
Phosphorylation of PI-TP␣ in Vivo-Swiss mouse 3T3 cells were cultured in Dulbecco's modified Eagle's medium containing 10% newborn calf serum and buffered with NaHCO 3 (44 mM) in a 7.5% CO 2 atmosphere. Near-confluent cell cultures in 75-cm 2 flasks were labeled for 4.5 h with 1.5 mCi of carrier-free [ 32 P]P i in 6 ml of phosphate-free Ham's F-12 (DF) medium containing 5% newborn calf serum. PMA (100 ng/ml) was present during the last 15 min of the labeling. The membrane-free supernatant was prepared, and the immunoprecipitation procedure was carried out as described by Snoek et al. (15).
Phosphopeptide and Phosphoamino Acid Analysis-After identification by autoradiography, the 32 P-labeled PI-TP␣ bands were excised from the dried gel and eluted as described by Boyle et al. (22). Briefly, the gel slices were homogenized in 50 mM ammonium bicarbonate, pH 7.3-7.6, SDS (final concentration, 0.1%), and 2-mercaptoethanol (final concentration, 1%) were added, and the samples were boiled for 5 min. After incubation of the mixture at 37°C for 2 h, the gel was spun down and the supernatant containing the 32 P-labeled proteins was collected. A second elution with 0.1% SDS and 1% 2-mercaptoethanol was carried out on the gel pellet. Carrier protein (boiled RNase, 10 g) and trichloroacetic acid (final concentration, 12%) was added to the combined supernatant fractions, and the samples were incubated on ice for 1 h. The trichloroacetic acid precipitate was washed with cold ethanol and dried. For phosphoamino acid analysis, the pellet was dissolved in 6 M HCl and hydrolyzed for 1 h at 110°C. The HCl was removed by lyophilization, and the pellet was dissolved in pH 1.9 buffer (glacial acetic acid:formic acid (88%):H 2 O; 78:25:897 v/v/v). A mixture of phosphoserine, phosphothreonine, and phosphotyrosine (1 g of each) was added. The 32 P-labeled phosphoamino acids were separated by twodimensional electrophoresis on 20 ϫ 20-cm cellulose TLC plates. The first dimension was in buffer pH 1.9, and the second dimension was in glacial acetic acid:pyridine:H 2 O (50:5:945 v/v/v), pH 3.5. After electrophoresis the plates were dried, the phosphoamino acids were visualized by staining with 0.2% (w/v) ninhydrin in acetone, and the 32 P-labeled amino acids were identified by autoradiography.
For phosphopeptide mapping the trichloroacetic acid pellet was dissolved in performic acid, and oxidation was performed for 1-2 h on ice. After lyophilization the sample was and incubated with TPCK-trypsin in 50 mM ammonium bicarbonate (200 g/ml) at 37°C for 5 h. The incubation was repeated by the addition of fresh trypsin, and the sample was lyophilized. The phosphopeptides were separated on cellulose TLC plates. In the first dimension electrophoresis was performed using the pH 1.9 buffer; in the second dimension TLC was performed in n-butanol:pyridine:glacial acetic acid:H 2 O (75:50:15:60 v/v/v/v). Radioactive phosphopeptides were identified by autoradiography.
Identification of the Phosphorylation Site-PI-TP␣ was phosphorylated as described above with the following changes. The ATP concentration was 1 mM with a trace of [␥-32 P]ATP, and the incubation time was 2-4 h at 30°C. The proteins were separated by SDS-PAGE, eluted, and precipitated with 10% trichloroacetic acid as described above. The pellet was digested with cyanogen bromide (2.5 mg/ml in 70% formic acid, 50 nmol cyanogen bromide/nmol protein) by incubation for 24 h in the dark at room temperature. After lyophilization, the sample was dissolved in 6 M guanidine HCl in 0.085% trifluoroacetic acid, and the peptides were separated on a reverse phase column C 2 /C 18 (Amersham Pharmacia Biotech, SMART system) with a 0 -60% (v/v) acetonitril. The radioactive peak was collected, and the N-terminal amino acid sequence of the 32 P-labeled peptide was determined by automatic Edman degradation using the 476A protein sequencer (Applied Biosystems).
Preparation of Myc-tagged PI-TP␣-The pBlue-wtPI-TP␣, pBlue-PI-TP␣(S166A) and pBlue-PI-TP␣(S166D) constructs were used to generate Myc-tagged PI-TP␣ fusion proteins. The pBlue-PI-TP␣ constructs contained a SacI site upstream of the translational start codon, an NcoI restriction site around the translational start codon, and a BamHI restriction site downstream of the translational stop codon. The SacI and NcoI sites were used to insert the linker encoding the Myc-tagged into the coding sequence. The linker consisted of two oligonucleotide primers carrying a SacI and an NcoI sticky end. The oligonucleotides used were: 5Ј-CATGGAACAAAAACTTATTTCTGAAGAAGATCTGC-3Ј and 5Ј-CATGGCAGATCTTCTTCAGAAATAAGTTTTTGTTCCATGA-GCT-3Ј. The underlined nucleotides of primer 2 represent the NcoI sticky end, and the bold nucleotides represent the SacI sticky end; the remaining bases are complementary to primer 1. The primers (10 M each) were annealed at 60°C for 15 min. After cooling to room temperature, the resulting linker was ligated into the pBlue-PI-TP␣ constructs that were previously digested with SacI and NcoI. The ensuing constructs were digested with SacI and BamHI, and the resulting DNA encoding Myc-tagged PI-TP␣ were ligated into the corresponding sites of the pBK-CMV expression vector. The obtained constructs were denoted as pBK-CMV-Myc-wtPI-TP␣, pBK-CMV-Myc-PI-TP␣(S166A), and pBK-CMV-Myc-PI-TP␣(S166D). Expression of these constructs yields PI-TP␣ fused to an N-terminal peptide containing the 9E10 epitope (the peptide EQKLISEEDL) of the human c-Myc protein (Myc-tagged).
Transfection of NIH3T3 Cells with the Myc-tagged Constructs-NIH3T3 cells were seeded 24 h prior to transfection at 1 ϫ 10 4 cells/cm 2 . Cells were transfected with 2 g of the pBK-CMV-Myc-PI-TP␣ constructs using the FuGENE6 Transfection Reagent kit according to the manufacturer's instruction (Roche Molecular Biochemicals). The following day the cells were reseeded at approximately 5 ϫ 10 3 cells/cm 2 . After another 24 h, G418 (0.4 mg/ml) was added for selection of G418resistant cells. Fresh medium containing G418 was added every 4 days, and resistant clones were identified after 3 weeks of growth.
Immunolocalization-The localization of endogenous PI-TP␣, Myctagged wtPI-TP␣, Myc-tagged PI-TP␣(S166A), and Myc-tagged PI-TP␣(S166D) in serum-starved (semi-quiescent) NIH3T3 cells was determined before and after stimulation with PMA or platelet derived growth factor (PDGF) as described in Ref. 15. Briefly, NIH3T3 cells were made semi-quiescent by replacing the growth medium with Dulbecco's modified Eagle's medium containing 0.5% newborn calf serum. After 2 days the cells were incubated for 15 min at 37°C with PMA (50 ng/ml) or PDGF (20 ng/ml) and fixed with methanol (endogenous PI-TP␣) or paraformaldehyde (Myc-tagged PI-TP␣s). Endogenous PI-TP␣ was visualized by indirect immunofluorescence using a polyclonal antibody directed against PI-TP␣ and goat-anti-rabbit-Cy3 as the second antibody. Myc-tagged PI-TP␣s were visualized using a mouse monoclonal antibody directed against the Myc-tagged and goat anti-mouse tetramethyl rhodamine isothiocyanate as the second antibody.
LysoPI Production in Vivo-The effect of PMA and PDGF on the production of lysoPI in NIH3T3 cells before and after stimulation with PMA or PDGF was determined as described in Ref. 12. Briefly, NIH3T3 cells were cultured in a 6-well plate to 80% confluency. The cell cultures were incubated for 48 h with 2 Ci of [ 3 H]-myo-inositol in HEPESbuffered DF medium without inositol containing 2% dialyzed newborn calf serum. Cultures were washed twice with phosphate-buffered saline and incubated for 10 min at 37°C with DF medium (without inositol) containing 0.3% bovine serum albumin and 10 mM LiCl. Subsequently, PMA (50 ng/ml) or PDGF (20 ng/ml) was added, and the incubation was continued for another 15 min. The cells were washed twice with phosphate-buffered saline and scraped in 1 ml of Ϫ20°C methanol. The [ 3 H]inositol phospholipids were extracted and analyzed as described previously (23).

RESULTS
Kinetic Analysis of Phosphorylation-The PKC-dependent phosphorylation of the charge isomers PI-TP␣I and II from bovine brain was determined in vitro as a function of concentration. From the Lineweaver-Burk plot it was calculated that the V max of the phosphorylation of PI-TP␣I was 2.0 nmol/min and that of PI-TP␣II 6.0 nmol/min (Fig. 1A). The K m of either reaction was comparable (0.65 and 0.72 M, respectively). This implies that the affinity of rat brain PKC for both isomers is the same, yet that PI-TP␣ containing a PC molecule is phosphorylated at a faster rate than the protein containing a PI molecule. To confirm these results we have done similar experiments on mouse recPI-TP␣. RecPI-TP␣ contains a PG molecule that can be readily exchanged for either PI or PC (16). Phosphorylation of the charge isomers of recPI-TP␣ by PKC confirmed that the PC-containing protein was phosphorylated at a faster rate than the PI-containing protein (Fig. 1B). In addition, phosphorylation of the PG-containing recPI-TP␣ was comparable with that of the PC-containing protein (data not shown). In comparison with the bovine PI-TP␣s the mouse recPI-TP␣s were better substrates, with V max values about 1 order of magnitude higher and K m values 1 order of magnitude lower (0.1 M).
Phosphopeptide Mapping and Phosphoamino Acid Analysis-Two-dimensional analysis of tryptic 32 P-labeled peptides showed that PI-TP␣I, PI-TP␣II and recPI-TP␣ have one major phosphopeptide in common as well as a minor one. The phosphopeptide map of PI-TP␣II showed three additional minor spots (Fig. 2). Phosphoamino acid analysis of the in vitro phosphorylated proteins demonstrated that all three PI-TP␣s are mainly phosphorylated on serine (Fig. 3). The immunoprecipitated PI-TP␣ from PMA-stimulated 32 P-labeled Swiss mouse 3T3 cells was purified by SDS-PAGE. Phosphopeptide mapping and phosphoamino acid analysis of 32 P-labeled PI-TP␣ showed that in vivo PI-TP␣ was exclusively phosphorylated on serine (data not shown). In line with this result, the phosphopeptide map is very similar to that of recPI-TP␣ (Fig. 2).
Determination of the Phosphorylation Site-To determine the PKC phosphorylation site of the in vitro phosphorylated PI-TP␣, this protein was cleaved by cyanogen bromide, and the peptides were separated on a reversed phase column. The 32 P label was detected in one peak that, as shown by automated Edman degradation, represented the cyanogen bromide peptide Lys-105 to Met-191, containing three serine residues (i.e. Ser-148, Ser-152, and Ser-166; data not shown). Given that the phosphoamino acid analysis showed serine as the preferred amino acid residue of phosphorylation and that in the labeled cyanogen bromide peptide only Ser-166 is the predicted PKC phosphorylation site, this serine was replaced with alanine by site-directed mutagenesis. Purified His 6 -tagged wtPI-TP␣ and His 6 -tagged PI-TP␣(S166A) were used as substrates for PKC. As shown in Fig. 4, PI-TP␣(S166A) was not phosphorylated by PKC (lanes 1-4), whereas wtPI-TP␣ was phosphorylated (lanes [5][6][7][8]. To exclude the possibility that incorrect folding was the cause of PI-TP␣(S166A) not being phosphorylated by PKC, we denatured the mutant and wtPI-TP␣ by boiling in SDS and ␤-mercaptoethanol, followed by SDS-PAGE electrophoresis and elution of the protein. PKC-dependent phosphorylation of these denatured proteins gave the same results as obtained with the native proteins (data not shown). These results show unambiguously that Ser-166 in PI-TP␣ is phosphorylated by PKC.
Transfer Activity of the PI-TP␣ Mutants-Purified His 6tagged wtPI-TP␣ was able to transfer PI, whereas His 6 -tagged PI-TP␣(S166A) lacked this activity (Fig. 5). Similarly, His 6tagged PI-TP␣(S166A) lacked PC transfer activity. As a control, Ser-152 was replaced with Ala. The resulting His 6 -tagged PI-TP␣(S152A) had normal PI and PC transfer activity. This indicates that Ser-166 is important for transfer activity. To investigate the possibility that the lack of transfer activity was due to incorrect folding during expression in E. coli, we have carried out refolding experiments on inclusion bodies according to standard procedures. 2 In agreement with the above observations, His 6 -tagged PI-TP␣(S166A) could not be activated, whereas refolded His 6 -tagged wtPI-TP␣ and His 6 -tagged PI-TP␣(S152A) were fully active.
Under optimal conditions 10 -15% of PI-TP␣ was found to be phosphorylated by PKC in vitro (15). In agreement with this relatively low level of phosphorylation, the phospholipid transfer activity was not affected by PKC treatment (data not shown). To further examine the effect of phosphorylation on transfer activity, a mutant of PI-TP␣ was made in which Ser-166 was replaced with Asp to mimic phosphorylated serine. As shown in Fig. 5, PI-TP␣(S166D) was not active in the phospho-  4. Phosphorylation of wtPI-TP␣ and PI-TP␣(S166A). PI-TP␣ was phosphorylated as described under "Experimental Procedures" and subjected to SDS-PAGE analysis (A) and autoradiography (B). Lanes 1 and 2, 1 g of PI-TP␣(S166A); lanes 3 and 4, 2 g of PI-TP␣(S166A); lanes 5 and 6, 1 g of wtPI-TP; lanes 7 and 8, 2 g of wtPI-TP␣; lanes 9 and 10, PKC control. The samples in the odd-numbered lanes were incubated in the absence of Ca 2ϩ , phosphatidylserine, and diacylglycerol, and those in the even-numbered lanes were incubated in the presence of Ca 2ϩ , phosphatidylserine, and diacylglycerol.
FIG. 5. In vitro phospholipid transfer activity of mutant and wtPI-TP␣. The phospholipid transfer activity of the PI-TP␣s was assayed as described under "Experimental Procedures" by measuring the transfer of fluorescently labeled PI (Pyr-PI) from quenched donor vesicles to acceptor vesicles. To the donor vesicles in the cuvette were subsequently added the acceptor vesicles (arrow I), bovine serum albumin (arrow II), and different PI-TP␣s (arrow III). Line 1, wtPI-TP␣; line 2, PI-TP␣(S152A); line 3, PI-TP␣(S166A); Line 4, PI-TP␣(S166D). lipid transfer activity assay, suggesting that the phosphorylation of Ser-166 may constitute a regulatory step.
As shown in Ref. 12, wtPI-TP␣ stimulates lysoPI formation in an in vitro assay. Under comparable conditions PI-TP␣(S166A) had no effect (data not shown). From this we conclude that PI-transfer activity is a prerequisite for PI-TP␣ to stimulate lysoPI formation.
Relocalization and LysoPI Formation-Stimulation of NIH3T3 cells with PMA for 15 min resulted in an extensive relocalization of PI-TP␣ to perinuclear membrane structures (Fig. 6). The same relocalization was observed after stimulation with PDGF (data not shown). This is in agreement with a previous study on the relocalization of PI-TP␣ in Swiss mouse 3T3 cells (15). In the latter study it was shown that these perinuclear structures were Golgi-like. To determine whether phosphorylation of PI-TP␣ was required for the relocalization to the Golgi complex, NIH3T3 cells were transfected with the pBK-CMV-Myc-wtPI-TP␣, -PI-TP␣(S166A), and -PI-TP␣-(S166D) constructs. The localization of the Myc-tagged PI-TP␣s was determined in serum-starved cells using an antibody against the Myc-tagged. Similar to the endogenous PI-TP␣, Myc-tagged wtPI-TP␣ was localized throughout the cytosol and in the nucleus and relocated to perinulcear Golgi structures upon stimulation by PMA (data not shown). Myc-tagged PI-TP␣(S166A) and Myc-tagged PI-TP␣(S166D) were also localized in the cytosol and the nucleus. However, these mutants did not relocalize upon stimulation by PMA, suggesting that phosphorylation of Ser-166 is required for the relocalization to the Golgi. Given that the expression of PI-TP␣ is involved in the regulation of phospholipase A-dependent formation of lysoPI (12), we have determined the cellular amount of lysoPI under the above conditions of PI-TP␣ translocation. Stimulation by PMA and PDGF increased the amount of lysoPI 2-and 3-fold, respectively (Fig. 7). No significant effects were observed on the levels of PI 4-phosphate and PI 4,5-bisphosphate. DISCUSSION Analysis of the amino acid sequence of mammalian PI-TP␣ showed the presence of five possible sequence motifs that are consensus phosphorylation sites for PKC (i.e. Thr-59, Thr-169, Thr-198, Thr-251, and Ser-166) (15). A review on specific phosphorylation sites for various PKC isoforms indicated that Ser-166 is the only putative site (24). In this study we have proven by peptide analysis and site-directed mutagenesis that indeed the phosphorylation of PI-TP␣ by rat brain PKC in vitro was restricted to Ser-166. Stimulation of PKC in Swiss mouse 3T3 cells by phorbol ester (PMA) also phosphorylated a serine residue of PI-TP␣. Given that the tryptic peptide maps of the in vitro and in vivo phosphorylated PI-TP␣s were comparable, we presume that Ser-166 is the site of PKC-dependent phosphorylation in intact cells.
Replacement of Ser-166 with an Ala or Asp residue yielded a PI-TP␣ species, which in vitro completely lacked PI and PC transfer activity. For comparison, replacement of Ser-152 with an Ala residue had no effect on the transfer activity. This indicates that the activity of PI-TP␣ is dependent on the presence of Ser-166. This residue may be essential for a correct folding of the active protein during its expression in E. coli. It is possible that concomitant with the expression, Ser-166 may form a hydrogen bond with the polar head group of the PG molecule that is normally present in recombinant wtPI-TP␣ (16). In the case of Ala-166 or Asp-166, this hydrogen bond cannot be formed, resulting in a PI-TP␣ that cannot interact with PG. This interaction may be important for PI-TP␣ to fold properly during its synthesis in E. coli. Reactivation of native PI-TP␣ expressed in E. coli from inclusion bodies was only possible when phospholipid was present in the refolding buffer. Under these conditions of refolding, inclusion bodies of PI-TP␣(S166A) did not yield an active protein. In addition, comparison of holo-and apo-species (with or without a phospholipid ligand, respectively) of native PI-TP␣ indicated that binding of a phospholipid ligand is required for obtaining the proper, more compact structure of holo-PI-TP␣. Apo-PI-TP␣ demonstrated a significant relaxation of the tertiary structure (25). In a previous study it was reported that mutation of the putative PKC phosphorylation site Thr-59 abolished the PI but not the PC transfer activity (26). Based on this observation a model was proposed where the phosphorylation/dephosphorylation of Thr-59 was presented as a key event in the regulation of phospholipid transfer activity in situ. However, because we have not found any evidence for the phosphorylation of a threonine residue, it remains to be established whether Thr-59 plays any role in this process Apart from being essential for proper folding and, hence, for in vitro phospholipid transfer activity, Ser-166 is also the single site of PKC phosphorylation. Very interestingly, the rate of phosphorylation was dependent on the phospholipid ligand bound to PI-TP␣. Both for bovine and mouse PI-TP␣, it was observed that the V max was 2-3-fold higher with PC than with PI as ligand. Because the K m values were comparable, we concluded that the affinity of PKC for PI-TP␣ was not affected by the ligand, yet that the charge difference between PI and PC resulted in a different rate of phosphorylation. It is as yet impossible to determine the phosphorylation of PI-TP␣I and II in situ as conventional methods fail to distinguish between these charge isomers. From the three-dimensional structure of the yeast PI-TP analogue (SEC14p), we know that the polar head group of the bound phospholipid molecule is exposed at the surface of the protein (27). If we assume that this is also the case in the mammalian PI-TP␣, the polar head group of the bound phospholipid could have an effect on the surface characteristics of the protein and hence on the rate of phosphorylation. In crystallization studies of PI-TP␣, we have observed that the conditions at which crystals were obtained were dependent on the bound phospholipid, suggesting a direct effect on the surface charge and conformation. 2 In the present study the recombinant mouse PI-TP␣ was a better substrate for the rat brain PKC as compared with the bovine PI-TP␣. Apparently the minor differences in isoelectric point and primary structure (16) have an effect on the in vitro phosphorylation by rat brain PKC.
In a previous study we have shown that activation of PKC in Swiss mouse 3T3 cells induced a translocation of PI-TP␣ from the cytosol to the Golgi complex (15). A similar translocation to the Golgi was observed when NIH3T3 cells were activated by PMA or PDGF (Fig. 6). We presume that phosphorylation of PI-TP␣ is required to make PI-TP␣ interact with the Golgi. This raises the question of what function PI-TP␣ may fulfill at the Golgi. In previous studies PI-TP␣ was shown to play an essential role in the ATP-dependent formation of transport vesicles from the trans-Golgi network in reconstituted systems (2, 28 -30). It has been proposed that the formation of these vesicles involves the local conversion of PI into metabolites that cause the remodelling of the phospholipid bilayer in the vicinity of a coated bud (28). PI-TP␣ would play a role in this conversion by delivering PI as a substrate to PI kinases (31). However, whether PI-TP␣ fulfills a role in the phosphorylation of PI in intact cells has not been confirmed. Studies with NIH3T3 cells overexpressing PI-TP␣ have indicated that the intracellular levels of lysoPI and further metabolites glycerophosphoinositol, Ins(1)P and Ins(2)P were constitutively increased (i.e. 3-6 fold) most likely as a result of PI-TP␣ acting as a regulator of a PI-specific phospholipase A 2 (12). No evidence was obtained that in these overexpressors PI-TP␣ had any effect on the phosphorylation of PI by activation of PI kinases. Glycerophosphoinositol and its further metabolites have been shown to act as a mitogen in certain cell lines (32) and may possibly contribute to the enhanced growth rate of these PI-TP␣ overexpressors (12). In cells overexpressing PI-TP␤ effects on PI metabolism were not observed. Concomitant with the translocation of PI-TP␣ to the Golgi, activation of wtNIH3T3 cells by PDGF and PMA resulted in a 2-3-fold increase of lysoPI (Fig. 7). In view of our evidence that levels of lysoPI are controlled by the amount of PI-TP␣ expressed, we presume that the PKC-dependent translocation of PI-TP␣ to the Golgi results in more PI becoming available for degradation by the phospholipase. Based on this hypothesis, we propose a model explaining how the phosphorylation of PI-TP␣ results in an enhanced lysoPI formation. As shown in the model presented in Fig. 8, the receptor-controlled activation of PKC leads to a more rapid phosphorylation of the PC-containing PI-TP␣ than of the PI-containing PI-TP␣. The phosphorylated form of PI-TP␣II is then translocated to the Golgi where in view of the high preference for PI, the protein unloads its PC molecule and binds a PI molecule. It can be expected that upon interaction with the Golgi, the phospholipid exchange reaction and dephosphorylation of PI-TP␣ by a protein phosphatase occurs. A similar protein phosphatase was proposed in the model by Alb et al. (26). After its release from the Golgi, PI-TP␣ delivers its bound PI to the PI-specific phospholipase A 2 to be degraded. The effect of the phosphorylation/dephosphorylation of PI-TP␣ is that upon receptor activation at the plasma membrane there is a rapid and controlled increase of PI available for metabolism. The important implication of this model is that the relative amount of PI-and PC-containing PI-TP␣s as controlled by the accessible PI and PC pools in the cell can be shifted toward the PI-containing PI-TP␣ as a result of PKC activation.