Phosphorylation of a distinct structural form of phosphatidylinositol transfer protein alpha at Ser166 by protein kinase C disrupts receptor-mediated phospholipase C signaling by inhibiting delivery of phosphatidylinositol to membranes.

Phosphatidylinositol transfer protein alpha (PITPalpha) participates in the supply of phosphatidylinositol (PI) required for many cellular events including phospholipase C (PLC) beta and gamma signaling by G-protein-coupled receptors and receptor-tyrosine kinases, respectively. Protein kinase C has been known to modulate PLC signaling by G-protein-coupled receptors and receptor-tyrosine kinases, although the molecular target has not been identified in most instances. In each case phorbol myristate acetate pretreatment of HL60, HeLa, and COS-7 cells abrogated PLC stimulation by the agonists formyl-Met-Leu-Phe, ATP, and epidermal growth factor, respectively. Here we show that phosphorylation of PITPalpha at Ser166 resulted in inhibition of receptor-stimulated PLC activity. Ser166 is localized in a small pocket between the 165-172 loop and the rest of the protein and was not solvent-accessible in either the PI- or phosphatidylcholine-loaded structures of PITPalpha. To allow phosphorylation at Ser166, a distinct structural form is postulated, and mutation of Thr59 to alanine shifted the equilibrium to this form, which could be resolved on native PAGE. The elution profile observed by size exclusion chromatography of phosphorylated PITPalpha from rat brain or in vitro phosphorylated PITPalpha demonstrated that phosphorylated PITPalpha is structurally distinct from the non-phosphorylated form. Phosphorylated PITPalpha was unable to deliver its PI cargo, although it could deliver phosphatidylcholine. We conclude that the PITPalpha structure has to relax to allow access to the Ser166 site, and this may occur at the membrane surface where PI delivery is required for receptor-mediated PLC signaling.

Phosphatidylinositol transfer proteins (PITPs) 1 are a family of lipid-binding proteins that transfer individual molecules of phosphatidylinositol (PI) or phosphatidylcholine (PC) between membrane compartments (1). Originally identified as soluble proteins of ϳ35 kDa, the family of PITP-related proteins has subsequently grown to include three subgroups of proteins all containing a PITP domain (2,3): the classical PITPs ␣ and ␤ (35 kDa), the larger related proteins RdgB␣I and -II (160 kDa), and the soluble RdgB␤ protein (38 kDa). The yeast Sec14p and its related family members form a separate group of proteins that, although they share lipid binding properties and transfer function with the mammalian PITPs, have no sequence or structural similarity (4 -7).
PITP␣ and PITP␤ are 77% identical (and 94% similar) in amino acid sequence and have been implicated in both signaling and membrane traffic (8 -14). Biochemical studies involving reconstitution of cytosol-depleted cell preparations with crude cytosol have consistently identified PITP (␣ and ␤) as a reconstitution factor in phospholipase C (PLC)-mediated PI 4,5-bisphosphate hydrolysis, the synthesis of 3-phosphorylated lipids by phosphoinositide 3-kinases, regulated exocytosis, and the biogenesis of vesicles at the Golgi (8 -16). Although both PITP isoforms can be used interchangeably in the reconstitution assays, they are likely to have some distinct functions in vivo. Deletion of the PITP␣ or PITP␤ genes give distinct phenotypes; deletion of the PITP␣ gene leads to neurodegeneration and early death, while deletion of PITP␤ is embryonically lethal (17,18). Studies using genetic manipulation have failed to clarify the role of PITP␣ in PLC signaling possibly because of the overlapping roles of these isoforms (19,20).
The ability of PITP␣ and PITP␤ to transfer PI makes these soluble proteins ideally suited for regulating the spatial and temporal provision of PI in specific membrane compartments or in the nucleus where it can be phosphorylated by specific lipid kinases (2,21). Recently we have reported the plasma membrane association of both PITP␣ and -␤ isoforms when cells are activated by EGF to stimulate phospholipase C activity (20). Phosphorylated forms of PI play essential roles in many cellular processes. The major requirement is for PI 4,5-bisphosphate at the plasma membrane where it is a substrate for both PLC and phosphoinositide 3-kinases, enzymes whose activity is regulated by cell surface receptors. Additional roles include regulation of ion channels, enzymatic activity of phospholipase D, maintenance of the cytoskeleton, and recruitment of target proteins by interaction with their phosphoinositide-binding domains including PH (pleckstrin homology), PX (pho ϫ homology), FYVE, and ENTH (epsin/N-terminal homology) (21,22). Proteins containing these domains appear to be involved in membrane traffic including endocytosis, exocytosis, and vesicle budding.
Many previous studies have reported that activation of protein kinase C (PKC) attenuates receptor-coupled PLC activity thus providing a negative feedback signal to limit the magnitude and duration of receptor signaling. Despite numerous examples of such regulation, the target for PKC-mediated inhibition has not been identified (Ref. 23; for reviews, see Refs. 24 and 25). These studies include both the G-protein-regulated PLC␤ family and receptor-tyrosine kinase-regulated PLC␥ family. Additionally no single lipid kinase has yet been identified that provides a dedicated pool of PI 4,5-bisphosphate for PLC-mediated hydrolysis. There are altogether four PI 4-kinases (Type II p55 isoforms (␣ and ␤) and Type III PI 4-kinase ␣ and PI 4-kinase ␤) and at least three phosphatidylinositol 4-phosphate 5-kinases (plus splice variants) (26,27). So far there has been no evidence for negative regulation by phosphorylation of any of the lipid kinases by PMA. The signal transducing system and the phospholipases are different depending on the receptor, and the only common molecule that participates in both G-protein-regulated systems and receptor-tyrosine kinases is PITP␣ (10,19,28).
In this study we examined the possibility that phosphorylation of PITP␣ by PKC may be responsible for the negative feedback inhibition. While this study was in progress, van Tiel et al. (29) reported that PITP␣ was phosphorylated on Ser 166 in vitro by PKC. In the present study, we identified Thr 59 (minor) and Ser 166 (major) as two residues that when phosphorylated are negative regulators of receptor-stimulated PLC signaling.
Our results support a model in which a structurally distinct pool of PITP␣ is phosphorylated at the membrane, and this phosphorylated PITP␣ is unable to deliver its PI cargo in exchange for PC at the plasma membrane.

EXPERIMENTAL PROCEDURES
Materials-All standard chemicals were obtained from Sigma. ␥-Labeled [ 32 P]ATP, 32 P i , [ 3 H]inositol, and [ 14 C]acetate were obtained from Amersham Biosciences. Rat brain PKC (catalog no. 539494) was purchased from Calbiochem. This preparation predominantly contains the conventional PKC isoforms ␣, ␤1, ␤2, and ␥. Endoproteinase Glu-C (V8 protease) and trypsin (modified) were sequencing grade and were obtained from Roche Applied Science. The lipids used for exchange were dimyristoyl PC, egg yolk PC, and bovine brain PI all from Sigma.
Phosphorylation of PITP␣ by PMA Treatment of HL60 Cells-HL60 cells (2 ϫ 10 8 cells) were washed and resuspended in 2 ml of buffer (20 mM Tris, 137 mM NaCl, 3 mM KCl, 1 mM CaCl 2 , and 1 mM MgCl 2 , pH 7.4), and the cells were incubated for 1 h with 1 mCi of 32 P-labeled sodium orthophosphate at 37°C. The cells were washed, resuspended in 4 ml, and allowed to equilibrate for 15 min at 37°C. 900 l of the cells were then added to tubes containing 100 l of PMA (100 nM final concentration), and the cells were incubated for 0, 30, 60, and 300 s at 37°C. The reaction was stopped by addition of 500 l of lysis buffer (1% Nonidet P-40, 50 mM PIPES, pH 6.8) supplemented with phosphatase inhibitors, and insoluble material was removed by centrifugation. PITP␣ was immunoprecipitated using 5F12 monoclonal antibody (30) coupled to Sepharose, and immunoprecipitates were washed five times in lysis buffer. The immunoprecipitates were resolved by SDS-PAGE, Western blotted, phosphorimaged, and probed using the monoclonal antibody 5F12.
Analysis of Receptor-stimulated Phospholipase C in Intact Cells-For HL60 cells, confluent suspension cultures (10 6 cells/ml) were incubated with 300 M dibutyryl cAMP to differentiate the cells for 48 h in Medium 199 in the presence of 1 Ci/ml [ 3 H]inositol. HeLa cells and COS-7 cells were labeled overnight with 2 Ci/ml [ 3 H]inositol in Medium 199 supplemented with 1.5% dialyzed fetal calf serum and 5 g/ml insulin and transferrin. In all cases, the cells were washed with HEPES buffer (20 mM HEPES, 137 mM NaCl, 3 mM KCl, 5.6 mM glucose, 1 mg/ml bovine serum albumin, 10 mM LiCl, 1 mM CaCl 2 , and 1 mM MgCl 2 , pH 7.2) and incubated with agonist. For HL60 cells, the cells were stimulated with fMLP (1 M) for the indicated times. COS-7 cells were stimulated with EGF (100 ng/ml), and HeLa cells were stimulated with ATP (1 mM) for 20 min. Cells were pretreated with PMA (100 nM) for 5 min prior to stimulation with the agonist where indicated. Inositol phosphates were measured as described previously (8).
Production of Polyclonal Antisera to Phosphorylated Ser 166 -Peptide synthesis and immunization were performed by Eurogentec (Seraing, Belgium). The peptide sequences CEDPAKFKSIKTGRGP and CEDPA-KFKpSIKTGRGP (where pS is phosphoserine) were synthesized, and two rabbits were immunized with the latter phosphorylated peptide. (Cysteine was added at the N terminus to aid coupling for the immunization process.) The serum was first purified using a Protein Gagarose column (10 ml). Serum was filtered through a 0.45-m filter, applied to the column at 1 ml/min, and then washed extensively with TBS (25 mM Tris, pH 7.4, 137 mM NaCl, 5 mM KCl), 0.02% NaN 3 . Bound antibody was eluted using 100 mM glycine, pH 2.5, and then neutralized with 1 M Tris, pH 7.4. (The Protein G column was then washed with 6 M urea before reuse). The eluted antibody was concentrated and desalted to TBS, 0.02% NaN 3 (8 ml).
Peptide (10 mg) was coupled to 1 ml of N-hydroxysuccinimideagarose columns (Amersham Biosciences) using the manufacturer's suggested coupling conditions. Two columns were made, a column containing the phosphorylated peptide and a second with the nonphosphorylated peptide. Protein G-purified immunoglobulins were purified using the peptide columns. Initially the Protein G-purified immunoglobulins were applied to the phosphopeptide column at 0.2 ml/min, and the flow-through was then reapplied (to recover losses). The column was then washed extensively with TBS, 0.02% NaN 3 , and the bound immunoglobulins were eluted with 100 mM glycine, pH 2.5 (3 ml) and immediately desalted to TBS/NaN 3 (4 ml). The resultant immunoglobulins were tested on dot blots containing dilutions of the synthetic phospho-and non-phosphopeptides. The antibody was further purified by applying to the non-phosphopeptide column. Briefly 1 ml was applied to the column and allowed to stand at room temperature for 30 min. The non-bound immunoglobulin (containing phosphopeptide antibodies) was washed through with 1 ml of TBS and then reapplied to the column in 2 ϫ 1-ml lots (leaving to stand for 30 min in between). The column was then washed with 1 ml of TBS resulting in 3 ml of nonbound immunoglobulin. This preparation, depleted of antibodies against the non-phosphorylated peptide, was then used for Western blots. The specificity of the phosphoantibody was validated using the following criteria. The antibody was tested using PITP␣ mutants where the phosphorylation site was mutated to alanine ( Fig. 4B) and by competing antibody binding with prior incubation with phosphopeptide and -phosphatase (New England Biolabs) treatment of phosphorylated PITP␣ (data not known).
Production of PITP␣-specific Polyclonal Antisera-Antibodies were raised in two rabbits against a specific PITP␣ peptide C-terminal sequence (CMRQKDPVKGMTADD). Peptide synthesis and immunization was performed by Eurogentec. Cysteine was added at the N terminus to aid coupling for the immunization process. The antisera were checked against recombinant PITP␤ to confirm that they were specific for PITP␣.
Phosphorylation of PITP␣ in Acutely Permeabilized Cells-HL60 cells were acutely permeabilized with 0.6 IU/ml streptolysin O in PIPES buffer (20 mM PIPES, 137 mM NaCl, 3 mM KCl, 1 mg/ml bovine serum albumin, 1 mg/ml glucose, pH 6.8) in the presence of 1 mM MgATP, 2 mM MgCl 2 , and calcium buffered with EGTA at pCa 7 or 5 (31). His-tagged wild-type and T59A PITP␣ proteins were included as indicated. Phospholipase C was stimulated with 10 M GTP␥S (31). PMA (100 nM) was added to the cells for 10 min prior to permeabilization. Following incubation at 37°C for 20 min, cells were removed by centrifugation, and the proteins were recaptured using nitrilotriacetic acid-agarose (Qiagen) for 1 h at 4°C. Bound proteins were extensively washed with PIPES buffer (containing no bovine serum albumin or glucose). Proteins were eluted with 500 mM imidazole, desalted to 20 mM Tris-HCl, pH 7.6, and concentrated. Quantities of each protein were estimated, and equal amounts were examined for phosphorylation by Western blotting.
Preparation of Rat Brain Cytosol Enriched in PITP␣-Rat brains (9.5 g) were homogenized in 19 ml of buffer (20 mM PIPES, pH 6.8, 137 mM NaCl, 3 mM KCl, 5 mM EGTA, 5 mM EDTA) supplemented with 1 ml of protease inhibitors (Sigma catalog no. P-2714). The homogenate was centrifuged to pellet membranes and insoluble material at 100,000 ϫ g at 4°C for 1 h. Cytosol (20 ml) was filtered through a 0.45-m membrane and concentrated in an Amicon pressure filtration device with a 10-kDa membrane to 10 ml (16 mg of protein/ml). PITP was purified from 9 ml of this cytosol (total protein, 144 mg) by gel filtration using a Superdex-75 HR 26/60 column (Amersham Biosciences). 5-ml fractions were collected, and PITP␣-containing fractions were located by Western blotting with PITP␣-and with phospho-Ser 166 -specific antibodies. Fractions enriched in PITP␣ were pooled and concentrated for use in the in vitro phosphorylation assay.
HPLC Analysis of Phosphopeptides-For HPLC analysis of the phosphopeptides, the phosphorylated protein was initially digested with proteases (trypsin or Glu-C). The phosphorylated protein was treated with 1% SDS and 10 mM dithiothreitol and boiled for 3 min. The protein was additionally treated with 4-vinylpyridine (0.5%) for 30 min at room temperature. SDS-PAGE was performed using 4 -12% bis-Tris gels (Novex, Invitrogen) and MES running buffer. The band corresponding to PITP␣ was excised with a clean, sharp scalpel blade and placed into a clean siliconized tube (Sigma). The gel piece was washed sequentially with (i) water, (ii) 50% acetonitrile, (iii) 100 mM ammonium bicarbonate, and (iv) 50% acetonitrile, 100 mM ammonium bicarbonate each for 15 min. The piece was then crushed using an Eppendorf homogenizer. The resulting crushed gel pieces were incubated with 100% acetonitrile for a further 15 min and then dried under vacuum. After drying, 200 l of Digest buffer (50 mM ammonium bicarbonate, 0.025% Zwittergent 3-16) and 2-4 g of sequencing grade protease (trypsin or Glu-C) were added and incubated at 30°C for 30 min. Proteases were solubilized in 1 mM HCl at 1 mg/ml. More Digest buffer was then added to completely cover the gel pieces, and the mixture was left at 30°C for 16 h. Following overnight digestion, the supernatant was removed from the gel pieces and filtered using an Ultra-MC filter unit (Millipore). The total volume of the filtrate was determined, and an equal volume of 0.2% trifluoroacetic acid was added. The peptides were then run on a C 18 Vydac column (catalog no. 218TP54). The column was run at 0.8 ml/min using water, 0.1% trifluoroacetic acid (Pump A) and 100% acetonitrile, 0.1% trifluoroacetic acid (Pump B). Gradient conditions were 0 -90 min, 0 -30% B; 90 -110 min, 30 -50% B; 110 -120 min, 50 -100% B; 120 -130 min, 100% B; and 130 -140 min, 100 -0% B. Fractions (0.5 min/0.4 ml) were collected at 0 -120 min. Radioactivity was determined using Cerenkov counting in a Packard liquid scintillation counter.
Native PAGE-Thrombin-cleaved wild-type and T59A (10 g) proteins were resolved by native PAGE. Novex Tris/glycine native gels (10 -20%), Tris/glycine running buffer, and sample buffer were obtained from Invitrogen. Gels were run at a constant 125 V for 5 h at room temperature. Proteins were additionally examined by SDS-PAGE to check for purity.
Reconstitution of G-protein-stimulated Phospholipase C Activity in Permeabilized HL60 Cells-The reconstitution of PLC activity was measured for each mutant using cytosol-depleted HL60 cells (8). In brief, [ 3 H]inositol-labeled HL60 cells were permeabilized with streptolysin O to remove endogenous PITPs, washed, and stimulated with GTP␥S (10 M) in the presence of Ca 2ϩ (1 M) and 1 mM MgATP. After incubation for 20 min at 37°C, the production of 3 H-labeled inositol phosphates was measured as an indicator of PLC activity.
Transfer and Binding of PI and PC-Transfer of PI by PITP␣ in vitro was measured by monitoring the transfer of [ 3 H]inositol-labeled PI from microsomes to non-labeled liposomes as described previously (28). Transfer and binding of PI and PC using cytosol-depleted permeabilized HL60 cells was conducted exactly as described previously (33).
Culture and Transfection of COS-7 Cells for Microscopy-COS-7 cells were grown in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% heat-inactivated fetal bovine serum and transfected with GFP-PITP constructs by electroporation. The plasmids used were wild-type PITP␣ and the Ala and Glu mutants of Thr 59 and Ser 166 . For electroporation (two pulses of 500 V and 125 microfarads), the cells were mixed with 0.7 g of the plasmid construct (34) and 35 g of herring sperm carrier DNA. Cells were plated onto coverslips and fixed 24 h post-transfection and subsequently imaged by confocal microscopy using a Leica DMRBE microscope including an SP2 confocal head with AOBS (Accousto optic beam splitter) with a 63ϫ oil immersion objective (numerical aperture, 1.2) and the Leica confocal software (LCS).
Size Exclusion Chromatography and Isoelectric Focusing (IEF)-Recombinant PITP␣ was exchanged to the PC form by the inclusion of a 40 M excess of dimyristoyl PC during the thrombin cleavage reaction. The protein was then separated from excess lipid using a size exclusion step. The exchanged PITP␣ protein was analyzed by IEF for exchange (showing a clear shift from the phosphatidylglycerol form to the PC form). The cleaved and exchanged protein was phosphorylated with rat brain PKC and then further resolved using size exclusion chromatography to remove PKC and excess [␥-32 P]ATP. The column was equilibrated with 20 mM Tris, pH 7.6. Fractions were analyzed by SDS-PAGE and phosphorimaged, and the fractions containing phosphorylated PITP were combined and concentrated. Phosphorylated PITP␣ was then analyzed by IEF, and its ability to exchange lipid was examined. A 40 M excess of PI was incubated with the phosphorylated PITP␣ at 30°C for 5 min. The mixture was resolved on Immobiline IEF gels (Amersham Biosciences). IEF gels consisted of a linear pH 4 -7 gradient precast by the manufacturer and utilizing an immobilized pH gradient. The gels were reswollen in water and run on a Multiphor flat bed electrophoresis system at 200 V for 1min, 0 -3500 V for 90min, and 3500 V for 3 h. The gel was stained and phosphorimaged.
Fluorescence Resonance Energy Transfer by Fluorescence Lifetime Imaging Microscopy between EGFP-PITP␣ and BODIPY-labeled Lipids-COS-7 cells were transfected with wild-type PITP␣, and the cells were incubated with BODIPY lipids. Measurements of GFP lifetime were done exactly as described previously (20).

PMA Pretreatment Causes Phosphorylation of PITP␣ and
Inhibits Receptor-stimulated Phospholipase C Activity-PITP␣ has five consensus phosphorylation sites (Thr 59 , Ser 166 , Thr 169 , Thr 251 , and Thr 198 ) for protein kinase C. To examine PITP␣ phosphorylation, HL60 cells, metabolically labeled with 32 P i were treated with PMA. PITP␣ was phosphorylated within 30 s, and the level of phosphorylation was increased at 300 s (Fig. 1A). The G-protein receptor agonist fMLP was also found to stimulate PITP␣ phosphorylation weakly (data not shown). Under conditions in which PITP␣ phosphorylation occurred by PMA pretreatment, fMLP-stimulated inositol phosphate formation was also inhibited (Fig. 1B). The inhibition by PMA was not restricted to fMLP-stimulated HL60 cells where the phospholipase C responsible for inositol phosphate production is PLC␤2 and is activated by ␤␥ subunits. Fig. 1C shows that COS-7 cells stimulated with EGF and HeLa cells stimulated with ATP were similarly inhibited following PMA pretreatment. Thus inhibition is seen for agonists acting via G-proteincoupled receptors through either ␤␥ (fMLP) or ␣ q (ATP) and receptor-tyrosine kinases (EGF).
PITP␣ Is Phosphorylated in Permeabilized Cells Stimulated with GTP␥S to Activate Phospholipase C-PITP␣ has been reported to be phosphorylated at Ser 166 (29), and therefore we generated a peptide antibody specific to phosphorylated Ser 166 . Protein kinase C activation is dependent on diacylglycerol and micromolar levels of Ca 2ϩ that occur downstream to PLC activation. We therefore examined whether PITP␣ would become phosphorylated when cells were incubated with GTP␥S to activate PLC. As a control, PMA was also used. We examined the T59A PITP␣ mutant in parallel. His-tagged PITP␣ proteins were incubated with the permeabilized cells together with GTP␥S and Ca 2ϩ (10 M), conditions in which the PLC is maximally activated (31). The PITPs were captured using the His tag and analyzed by Western blotting using the polyclonal antibody that recognizes PITP␣ and phosphorylated PITP␣ (Fig. 1D). Phosphorylation of both wild-type and T59A proteins occurred under conditions in which PLC activity was stimulated with GTP␥S. Activation of PKC by PMA also phosphorylated PITP␣. It was noted that phosphorylation of T59A was greatly enhanced compared with wild-type PITP␣ (3.7 Ϯ 0.2fold, n ϭ 8; see Fig. 4).
Identification of Ser 166 Phosphorylated PITP␣ in Rat Brain-Attempts to identify phosphorylated PITP␣ with the phosphoantibody in HL60 cells were not possible because of the low concentration of PITP␣ in HL60 cells (2 ng/100 g of cell lysate). In contrast, rat brain has the highest tissue concentration of PITP␣ at 40 ng/100 g of cell lysate. We therefore examined whether Ser 166 phosphorylated PITP␣ could be detected in crude rat brain cytosol. Phosphorylated PITP␣ could only be detected following fractionation of rat brain cytosol by gel filtration. Individual fractions expected to be enriched in PITP␣ were blotted with both a PITP␣-specific antibody and the phospho-Ser 166 antibody. It was consistently observed that phosphorylated PITP␣ elutes slightly earlier than the nonphosphorylated protein (see Fig. 2A). The fractions containing PITP␣ were pooled and used in an in vitro assay using PKC. Recombinant PITP␣ was also analyzed in parallel. Both the partially purified PITP␣ from rat brain cytosol and recombinant PITP␣ could be phosphorylated by PKC at Ser 166 (Fig.  2B).
PITP␣ Is Phosphorylated at Thr 59 and Ser 166 in Vitro-To assess the stoichiometry of phosphorylation, recombinant PITP␣ was phosphorylated with purified PKC using [ 32 P]ATP in vitro. We initially used the His-tagged proteins in the phosphorylation assay and found that the linker region highlighted in italics, MRGSHHHHHHGMASMTGGQQMGRDLYD-DDDKDPMVLLKE, despite not having a consensus PKC phosphorylation site, was phosphorylated at the two serine residues (underlined) by PKC (identified by Edman sequencing). The experiments were therefore repeated with the His tag removed by thrombin cleavage prior to the phosphorylation assay. Cleaved PITP␣ (wild type) was phosphorylated by PKC (Fig. 3A), and the percentage of total PITP␣ that was phosphorylated was low and varied between 1 and 2.4% (average, 1.7%; n ϭ 4).
For identifying additional sites of phosphorylation, the phosphoprotein was digested with trypsin, but chromatography of the resulting peptides by HPLC revealed no distinct peaks labeled with 32 P (Fig. 2A, see inset). There are 39 potential trypsin cleavage sites, and the resultant HPLC profile had unresolved peptides likely resulting from incomplete digestion. We therefore tried digestion with Glu-C, which cleaves C-terminal to glutamate, since PITP␣ has 27 potential cleavage sites. The peptides were analyzed by HPLC, and five peaks with radiolabel were observed (Fig. 3A). Phosphopeptides in these peaks were quantified by Cerenkov counting according to the specific activity of 32 P used for phosphorylation (Fig. 3C). The major peptide fragment (peak 5) was subjected to mass spectrometric analysis by MALDI-TOF mass spectrometry and solid phase Edman degradation (35). No peptide mass that fitted a PITP␣/Glu-C phosphopeptide was detected. In solid phase Edman sequencing, the peptide was put through 34 cycles, and no release of radioactivity was observed. This suggested that the peptide had an N-terminal Glu residue that had reacted with the coupling reagent used to immobilize the peptide. There is one such Glu-C cleavage site in PITP␣, and the resultant fragment (EDPAKFKSIKTGRGPLGPNWKQE) contains Ser 166 and Thr 169 (shown in bold).
We additionally analyzed the Glu-C-digested HPLC peptides using the phosphospecific Ser 166 antibody (see Fig. 3B). Peaks 3 and 5 were strongly immunoreactive indicating that the phosphopeptides contain phosphorylated Ser 166 . Peaks 1, 2, and 4 were not recognized by the antibody, suggesting that additional sites were phosphorylated. We used the phosphodefective mutants S166A and T59A in the in vitro assay and observed that phosphorylation of S166A was greatly diminished compared with wild type, while phosphorylation of T59A was greater than that of wild type (1.7 Ϯ 0.1-fold, n ϭ 13; Fig.  4A). The phosphorylated sample of PITP␣ T59A was digested with Glu-C, and the peptides were separated by HPLC. Peak 5 (of Fig. 3A) was still present, while the minor peaks 1 and 2 had disappeared (data not shown). These results confirm that PITP␣ is mainly phosphorylated in vitro by PKC at Ser 166 and to a lesser extent at Thr 59 .
Influence of Thr 59 on PITP␣ Conformation Is Important for Ser 166 Accessibility-The low stoichiometry of PITP␣ phosphorylation in vitro is readily explained since Ser 166 is not accessible to PKC in the known PITP␣ structures (5-7). Thus for phosphorylation to occur in vitro at Ser 166 , a minor fraction of the protein must exist in a distinct structural form in solution. The side chain of Ser 166 is located in a small pocket formed by the 165-172 loop and is not solvent-accessible (Fig. 4C). Several other conserved residues in or adjacent to this loop apparently stabilize its conformation. This conformation is essentially identical in all three structures (PI-liganded, PCliganded, and the apostructure). Fig. 4, A and B, compares the phosphorylation of PITP␣, T59A, S166A, and PITP␣⌬5 using radiolabeled ATP and the Ser 166 phosphopeptide antibody. PITP␣⌬5 has a deletion of five C-terminal residues and a higher affinity for membranes (20) and is inferred to have a more relaxed conformation (36). Phosphorylation of the ⌬5 truncation mutant was identical to wild-type protein indicating that this change in conformation does not influence the availability of Ser 166 for PKC-mediated phosphorylation (Fig. 4, A  and B). The data with the C-terminally truncated PITP␣⌬5 clearly support the view that opening the protein up during lipid exchange is not likely to allow access to Ser 166 , and this is confirmed by examination of the apostructure. The only mutant that shows increased phosphorylation is T59A (1.7 Ϯ 0.1-fold, n ϭ 13; Fig. 4A). We noted that T59A showed greatly enhanced phosphorylation when endogenous membrane-associated PKC was used compared with the in vitro phosphorylation by PKC (3.7-versus 1.7-fold, compare Fig. 1D with Fig. 4,  A and B). The greater degree of phosphorylation observed in the permeabilized cell assay could be due to a structural change induced by the proximity of the protein to the membrane.
Since T59A is more highly phosphorylated at Ser 166 in vitro, it must contain a larger population of a structurally distinct form of PITP␣ protein in which Ser 166 is accessible to PKC. Fig.  4D shows native PAGE analysis of both wild-type and T59A PITP␣. In the T59A sample, the PITP␣ resolved into two forms (Fig. 4D, lane 4, indicated by the arrows). One form (lower arrow) corresponded to the wild-type protein, which resolved predominantly as a single species (Fig. 4D, lane 3). Using SDS-PAGE each of the proteins, wild-type and T59A, resolved as a single band (Fig. 4E, lanes 3 and 4, respectively).
Identification of Ser 166 and Thr 59 on PITP␣ as Important Residues for PLC Signaling-To analyze the effects of phosphorylation on PLC signaling and lipid transfer, each of the five putative PKC phosphorylation sites (Ser 166 , Thr 59 , Thr 169 , Thr 198 , and Thr 251 ) were mutated to glutamic acid as a phosphomimetic. Their locations in the PITP␣ structure are shown in Fig. 5A (5, 6). Thr 198 and Thr 59 are located on ␤-strands with Thr 59 very close to the lipid binding site, and Thr 169 and Ser 166 are located in the regulatory loop, while Thr 251 is located in the C-terminal ␣-helix (G-helix). The phosphomimetic mutant proteins and the wild-type protein were expressed in E. coli and examined for their ability to support the production of inositol phosphates following activation of PLC␤2 by ␤␥ subunits in HL60 cells. The purified proteins were analyzed by SDS-PAGE (Fig. 5B). While the purity and expression levels of the wildtype protein PITP␣ and the mutants T198E and T251E were comparable, T59E, S166E, and T169E were less pure and had reduced expression levels compared with wild type. Protein concentrations were therefore estimated from the SDS-poly- acrylamide gel relative to wild-type PITP␣ protein. Only the T59E and S166E PITP␣ mutants were unable to restore Gprotein-stimulated PLC␤-mediated inositol phosphate production, and these mutants were also unable to transfer PI in an in vitro assay (Fig. 5, C and D).
A more extensive analysis of the phosphomimetic mutants S166E and T59E was undertaken, and we included the phosphorylation-defective mutants (S166A and T59A) for comparison. To examine the localization of the mutants, PITP␣ proteins were expressed as GFP fusion proteins and transfected into COS-7 cells (Fig. 6A). GFP-PITP␣ S166E and S166A localized to the cytosol and nucleus similarly to wild-type PITP␣ (20,37). In contrast, GFP-PITP␣ T59E and T59A were excluded from the nucleus. By confocal microscopy, it was observed that the number of cells expressing S166A and S166E was low in comparison to cells expressing wild type, T59A, or T59E. Expression of the GFP fusion proteins was also examined by Western blot. The antibody for PITP␣ identified two bands in the GFP fusion-expressing cells, one band representing the endogenous PITP␣ and the upper band representing the fusion protein. While expression of wild-type PITP␣ and the mutants T59A and T59E was clearly observed, we were unable to observe the expression of the GFP fusions for S166A and S166E. This result is not surprising since the number of GFP fusion-expressing cells represented 1-2% of the total cell population. It was observed that overexpression of S166A-GFP consistently led to increased levels of endogenous PITP␣, and this was not explored further (Fig. 6A).
Thr 59 and Ser 166 mutants (Ala and Glu) were examined for PI and PC binding and transfer using a cell-based assay (33). In this assay, the PITPs were incubated with cytosol-depleted HL60 cells, which act as a source of radiolabeled lipids. Lipid binding is monitored after repurification of the PITPs, and lipid transfer is measured by monitoring transfer to exogenously added liposomes. T59E was unable to transfer PI, while PC transfer was similar to wild type (compare Fig. 6, B and C). This inability to transfer PI was entirely due to loss of PI binding (see Fig. 6, D and G). In comparison, T59A showed reduced PI and PC transfer. A detailed analysis of several independent preparations showed that both PI and PC transfer were reduced to 60 Ϯ 7 and 65 Ϯ 9% (n ϭ 4), respectively, when analyzed at a fixed concentration of 100 g/ml (6). Interestingly PI binding was reduced to 36 Ϯ 11 (n ϭ 4), while PC binding was unaffected. We also examined the PITP␣ mutant T59A for its ability to reconstitute PLC activity, and it showed a modest decrease compared with wild type (data not shown).
Analysis of the Ser 166 mutants revealed that both the phosphomimetic and phosphodefective mutants were severely impaired for both PI and PC transfer (Fig. 6E). Since the expression of the Ser 166 mutants in E. coli was greatly reduced, they were only tested at a single concentration for transfer. S166A was also examined in the microsome-liposome assay, which requires less protein, and it was clear that this protein retained some activity. When examined for its ability to restore PLC signaling, the protein also retained some activity (data not shown) unlike S166E (Fig. 5C). Examination of binding revealed that S166A and S166E still bound PI and PC although at a reduced level (Fig.  6, F and G). S166E was more impaired than S166A, and this was reflected in their ability to transfer lipids.

Extent of PITP␣ Phosphorylation Is Independent of the
Bound Lipid Species-A previous study has reported that the PKC-dependent phosphorylation is controlled by the phospholipid species bound to PITP␣ (29). The PC-bound form of PITP␣ was found to be a better substrate than the PI-bound form. In light of the observation that the structures of PITP␣ bound to PI and PC are nearly identical (5, 6), we have re-examined whether the phospholipid species influences phosphorylation by PKC. PITP␣ proteins were exchanged with either egg yolk PC or bovine brain PI and purified by gel filtration to remove the vesicles. It was noted that some of the PITP␣ eluted in the void volume with the vesicles, leading to loss of protein. PI-and PC-loaded PITP␣ were phosphorylated by PKC to a similar extent (Fig. 7A).
Phosphorylated PITP␣ Can Exchange Bound PC for PI but Not Vice Versa-To examine whether phosphorylated PITP␣ could exchange its lipid-bound cargo, we used recombinant PITP␣ loaded with dimyristoyl PC for phosphorylation by PKC.
As observed before with rat brain PITP␣, the phosphorylated PITP␣ eluted slightly earlier compared with the bulk of the protein (Fig. 7B). Fractions enriched with the PC-loaded phosphorylated PITP␣ were pooled (Fig. 7B, i), half the sample was incubated with PI vesicles for lipid exchange, and the samples were analyzed by IEF. The proteins were silver-stained (Fig.  7C, panel a) and also phosphorimaged (Fig. 7C, panel b) to identify the phosphorylated proteins. In the silver-stained gel (Fig. 7C, panel a), the unphosphorylated PC form of PITP␣ focuses at pH 6, and when exchanged with PI, the majority of protein focuses at pH 5.7. This change in behavior is due to PITP␣ acquiring one net negative charge when PC is replaced with PI. In the phosphorimage, it is seen that the phosphorylated PC-liganded PITP␣ is shifted to a lower pH compared with the non-phosphorylated protein (Fig. 7C, compare lanes  marked PC in panels a and b). This shift is due the acquisition of an additional negative charge, this time due to phosphorylation. The phosphorylated PITP␣ (PC form) shows a further shift toward the positive side after it is exchanged with PI (panel b). Thus, phosphorylation of PITP␣ does not prevent exchange of PC for PI.
To examine whether phosphorylation of PITP␣ could affect the exchange of PI for PC, the same protocol was followed as above except that bacterially derived phosphatidylglycerol was exchanged with bovine brain PI. The PI-loaded protein was phosphorylated with PKC and subsequently incubated with dimyristoyl PC vesicles for lipid exchange as described above followed by analysis by IEF. On IEF, the proteins would not enter the gel, suggesting that the highly charged proteins were binding to the vesicles avidly and were not freely mobile. From this behavior we conclude that PI-loaded phosphorylated PITP␣ must not have exchanged its cargo for PC since the phosphorylated PITP␣ (PC form) had no difficulty in entering the gel as seen from Fig. 7C.
Phosphorylation of PITP␣ Perturbs Binding to PC in Intact Cells-Although the phosphoprotein (PC form) was still capable of exchange (PC for PI), the data described above suggested that the PI form of the phospho-PITP␣ was not able to exchange PI for PC. To analyze this behavior by a non-invasive technique we took advantage of a recent method for analyzing lipid-protein interactions using fluorescence lifetime imaging (20). Previous studies have already shown that EGF-stimu-lated inositol phosphate production is dependent on PITP␣ (10). We established that PMA pretreatment of COS-7 cells abolished the EGF-stimulated inositol phosphate production (Fig. 1C). COS-7 cells were transfected with GFP-PITP␣ and labeled with BODIPY PI and PC. As reported previously, upon EGF stimulation a significant reduction in GFP lifetime is observed indicating close proximity between PITP␣ and the lipid cargo (20). In PMA-pretreated cells, only the interaction between PITP␣ and membrane PC is inhibited but not that between PITP␣ and membrane PI (Table I). These results complement the in vitro analysis by IEF that PI-loaded PITP␣, when phosphorylated, is unable to interact with PC, while the phosphorylated PC-loaded PITP␣ is still able to exchange its cargo for PI. DISCUSSION In this study, we found that PITP␣ can be phosphorylated in cells pretreated with PMA, and from an in vitro analysis identified Ser 166 (major) and Thr 59 (minor) as the residues that are phosphorylated. We also showed that PMA pretreatment disrupts both G-protein-stimulated phospholipase C␤ and EGFstimulated phospholipase C␥ activation. Such negative feedback regulation by PMA pretreatment has been observed in many other cell types including Fc⑀R1 on mast cells (38), va-FIG. 5. Analysis of phosphomimetic mutants of PITP␣ for PLC activation and for PI transfer. A, structure of PITP␣ (Protein Data Bank code 1T27) with the phosphorylation sites indicated in purple. Throughout the paper the numbering used is the rat/mouse numbering for ease of comparison with previously published data using mouse PITP␣. Human PITP␣ is shorter by one residue because of a deletion at position 52. B, PITP␣ mutants were purified as described under "Experimental Procedures." 5 g of each protein was analyzed by SDS-PAGE to examine the level of purity. C, PITP␣ mutants were assayed for their ability to reconstitute G-protein-stimulated phospholipase C activity in cytosol-depleted HL60 cells. Each mutant was analyzed in duplicate, and the data shown are representative of several (two to four) independent mutant preparations. Error bars represent the range for each point. D, PITP␣ mutants were assayed for transfer of radiolabeled PI from microsomes to liposomes. Each mutant was analyzed in duplicate, and the data shown are representative of several (two to four) independent mutant preparations. Error bars represent the range for each point. WT, wild type; 59E, T59E; 166E, S166E; 169E, T169E; 198E, T198E; 251E, T251E. sopressin-stimulated WRK-1 cells (39), angiotensin-stimulated adrenal glomerulosa cells, and carbachol-stimulated astrocytoma cells (40). The literature on the feedback regulation of phosphoinositide turnover has been reviewed, and in many cases the molecular target for PKC could not be identified (24,25). Previous studies have reported that in fMLP-stimulated HL60 cells (a neutrophil-like cell line), inositol phosphate production due to PLC activation was inhibited by PMA pretreatment (23,41). HL60 cells contain PLC␤2 and PLC␤3, and their activation occurs via G␤␥ subunits exclusively when stimulated with fMLP (42). Knock-out of the PLC␤2 gene reduces inositol phosphate production and Ca 2ϩ signaling by 70%, and while knock-out of both PLC␤2 and PLC␤3 genes completely abrogates the responses, indicating that PLC␤2 is the major phospholipase for inositol phosphate production in neutrophils (42,43). In a recent study, the inhibition of fMLP-stimulated inositol phosphate production by PMA was examined, and it was concluded that phosphorylation and consequent inactivation of PLC␤2 or PLC␤3 isozymes was unlikely (23). Although PLC␤3 can be phosphorylated by PKC at Ser 1105 , this phosphorylation only inhibits G␣ q -stimulated PLC␤3 but not G␤␥-stimulated PLC␤3 (23). This serine residue is not conserved in PLC␤2. PITPs have not been examined previously, and in this study we demonstrated that phosphorylation of PITP␣ could be responsible for this PMA-induced inhibition. Although the extent of phosphorylation of PITP␣ in vitro was low, this should not disqualify PITP␣. The low stoichiometry of PITP␣ phosphorylation observed in vitro is due to inaccessibility of the Ser 166 residue to PKC (Fig. 4C). Since Ser 166 is in an identical environment in the three solved structures of PITP (PC-PITP, PI-PITP, and apoPITP), this suggests this is the predominant form in solution. We suggest that it is in equilibrium with a small fraction of PITP␣ that is structurally distinct such that Ser 166 is accessible. This equilibrium is shifted in the mutant T59A since it is more highly phosphorylated in vitro. Although Thr 59 is a long distance from Ser 166 , perturbations in this region appear to be sufficient to shift the equilibrium to a different structural form.
In permeabilized cells, PITP␣ phosphorylation was observed under conditions in which endogenous PKC is stimulated by PMA and Ca 2ϩ or by GTP␥S, which activates PLC (Fig. 1D). In cells, phosphorylation of PITP␣ will occur at the membrane, which may facilitate structural changes (see Fig. 8). There are precedents for such changes in proteins that are active at the membrane (44). We have identified two tryptophan residues that are essential for PITP␣ to mediate lipid transfer (6). Bulky aromatic tryptophan side chains are known to be able to penetrate the interfacial region, which is composed of the head groups, water, and portions of acyl chain methylene groups that extend toward the interface from the hydrocarbon core (45). They do not penetrate into the hydrophobic core of the membrane. Docking to the membrane, which is mediated by the interaction of these tryptophan side chains, may induce a conformational change. For the phosphorylation sites to become accessible, this may go further than the changes seen in the apostructure, converting PITP␣ into a "molten globule" that allows partial unfolding of the loop and so permitting PKC to gain access. The steroidogenic acute regulatory protein, StAR, which is also a transfer protein like PITP but for cholesterol, is reported to undergo such a transition when it binds to membranes (46 -48). Thus, we anticipate that when PITP␣ docks to the membrane, the protein undergoes extensive changes including relaxation of the 165-172 loop such that the Ser 166 is available for phosphorylation. We cannot discount the possibility that ancillary proteins may aid this process.
Additional evidence that PITP␣, phosphorylated at Ser 166 , has a different structural conformation comes from analysis of its behavior on gel filtration. The phosphorylated fraction of both the native protein from rat brain and recombinant protein phosphorylated in vitro eluted earlier on a size exclusion column. The order of elution of proteins from a size exclusion column is inversely related to size and shape so that earlier FIG. 6. Analysis of the phosphomimetic (T59E and S166E) and phosphorylation-deficient mutants of PITP␣ (T59A and S166A). A, localization of GFP-PITP␣ mutants by confocal microscopy in COS-7 cells and Western blot analysis. COS-7 cells were transfected with GFP-PITP constructs, imaged by confocal microscopy, and analyzed for protein expression by Western blotting of radioimmune precipitation assay bufferextracted total cell lysates (30 g/lane). B and C, transfer of PI (B) and PC (C) with wild-type PITP␣, T59A, and T59E from cytosol-depleted HL60 cells to liposomes. D and F, binding of PI and PC after incubation of His-tagged proteins with cytosol-depleted HL60 cells for 10 min. Proteins were recovered using the His tag, a fraction of the sample was analyzed by SDS-PAGE to monitor protein recovery, and the bound lipids were extracted from the rest of the sample and analyzed by TLC. E, transfer of PI and PC with S166A and S166E at a single concentration (50 g/ml). G, quantitation of PI and PC binding relative to wild-type PITP␣. WT, wild type; 59A, T59A; 59E, T59E; 166A, S166A; 166E, S166E. elution of the phosphorylated protein implies the protein has a higher molecular weight. Since phosphorylation of a single residue does not normally cause such a dramatic change in molecular weight, an increase in hydrodynamic volume arising from a structural change in the protein is the more likely explanation.
Alignment of 39 PITP␣-related sequences identified in mam-mals, fish, amphibians, flies, soil amoebae, red photosynthetic algae, and parasites indicate that Ser 166 is conserved in 36 of 39 sequences (Table II)  . Phosphorylated protein elutes slightly earlier (underlined and labeled i) than the bulk of the protein (overlined and labelled ii). C, the peak fractions containing the phosphorylated proteins (i) were combined and concentrated. Approximately 20 g of protein was resolved by IEF. Panel a, silver-stained IEF gel; panel b, phosphorimage of the IEF gel. Lane PC, PITP␣ loaded with PC and phosphorylated, and lane PI, PITP␣ loaded with PC and phosphorylated, and then incubated with 20 g of PI at 30°C for 5 min. Following phosphorylation PC-loaded PITP␣ has an increased negative charge, and exchange to PI increases the net negative charge further shifting the protein toward the anode. Gels shown are representative of experiments performed three to five times with separate PITP preparations.

TABLE I
PMA pretreatment inhibits EGF-stimulated fluorescence resonance energy transfer between PITP␣ and membrane PC but not with membrane PI GFP-PITP␣-transfected COS-7 cells were incubated with BODIPY-PC (BPC) or BODIPY-PI (BPI) for 15 min at 37°C, stimulated with 35 ng/ml EGF (15 min at 37°C), and fixed with 4% paraformaldehyde. Fluorescence resonance energy transfer results in the shortening of the GFP (donor) fluorescence lifetime that is measured by two parameters, the phase shift ( p ) and relative modulation depth ( m ). The average lifetime () of GFP-PITP␣ (donor) (( p ϩ m )/2) in cells that are only incubated with BODIPY-PC and BODIPY-PI (acceptors) does not vary significantly. The localized GFP lifetimes at the plasma membrane were quantified from three independent experiments. A region of interest containing approximately 50 pixels was used to obtain a localized average lifetime value. The calculated GFP lifetimes of the individual pixels were averaged (ϮS.D.). In PMA-pretreated cells, the EGF-stimulated interaction of GFP with BODIPY-PC but not BODIPY-PI is inhibited. GFP  Phosphorylation of PITP␣ at Ser 166 tant in PITP function and phosphorylation plays a fundamental role in most PITP-related molecules. In the sequence of C. elegans, the equivalent residue is alanine, while in E. cuniculi, it is histidine. Since we observed that substitution of Ser 166 with alanine somewhat reduces both PI and PC transfer and binding, it will be of interest to examine the activity of these PITPs and whether this is compensated by another alteration in the sequence. In addition to the phosphorylation of Ser 166 , our analysis showed that Thr 59 is also phosphorylated. Although it is a minor site in the in vitro studies reported here, the extent of Thr 59 phosphorylation in cells may be greater. In addition, Thr 59 clearly has an indirect role in regulating the accessibility of Ser 166 for phosphorylation. Thr 59 is conserved in all the currently known PITP sequences (6), and we show here that when mutated to a phosphomimetic, the protein effectively became a PC-binding and transfer protein. In the T59E mutant, PI binding was reduced by over 90%, and PI transfer was undetectable. Nonetheless PC binding and transfer were unaffected. A similar pattern has been observed previously where mutation of this residue to a variety of amino acids (T59V, T59E, T59D, T59S, T59N, and T59Q) all led to near complete inhibition of PI transfer but had very little effect on PC transfer. The only mutant that was unaffected was T59A, which was reported to retain full PI transfer activity (49). However, in that study binding of PI or PC was not analyzed (49). In our study, we observed that T59A was compromised for both PI and PC transfer, although only PI binding was substantially reduced (see Fig. 6).
In PITP␣-PC, Thr 59 makes only van der Waals contact (5) with a methyl group of the PC head group, and thus it is not surprising that mutation of this site had a minimal effect on PC binding. In contrast, in PITP␣-PI this residue is involved in binding the inositol head group (6), and mutation to alanine would be predicted to have a more drastic effect on PI binding. This is precisely what the data showed. On the other hand, steric crowding would provide an explanation for the near complete loss of PI binding when Thr 59 was mutated to glutamate. Since this residue is conserved in all the PITP sequences identified so far, it is anticipated that this residue plays a similar function in all of the PITP family. However, analysis of the PITP domain of Drosophila RDGB␣ protein indicates that the T59A mutation, not T59E, leads to a reduction in PI transfer, while PC transfer is unaffected (50). This result is surprising in the light of the PI-bound PITP␣ structure and needs to be revisited. In the mammalian RDGB␣ protein (also known as Nir2), the PITP domain of the wild-type protein localizes to the cytoplasm similar to PITP␣, and mutation of Thr 59 to glutamate causes the protein to translocate to lipid droplets (51). This striking relocalization is certainly not a feature of PITP␣ T59E (see Fig. 6A). Instead, with PITP␣, when Thr 59 was mutated to either alanine or glutamate, the protein was unable to enter the nucleus. This localization is clearly not dictated by the species of lipid bound to PITP␣ but may be due to a structural change that inhibits its nuclear targeting. As discussed above, T59A was a better substrate for PKC, and this is likely due to a structural change in the protein.
Recent reports from studies in cells derived from PITP␣ knock-out mice have been unable to show any effects on PLC signaling. In one study, embryonic stem cells were stimulated with serum and lysophosphatidic acid, and it was reported that inositol 1,4,5-trisphosphate production was not affected (18). In another study fibroblasts were stimulated with vasopressin, and again the authors found no effect on PLC signaling (52). These findings are hampered by the presence of other PI transfer proteins, in particular PITP␤ present in the cells. We have observed the interaction of PITP␤ with the plasma membrane following EGF stimulation (20), suggesting that PITP␤ could fulfill the role of PITP␣ in these knock-out studies. We have previously shown that PITP␤, Sec14p, and Dictyostelium PITPs are able in vitro to substitute the PLC signaling role of PITP␣ (19,34). We are currently using interference RNA to FIG. 8. Model depicting how phosphorylation of PITP␣ causes inhibition (negative feedback) following stimulation of phospholipase C. Following stimulation of phospholipase C by cell surface receptors, production of diacylglycerol (DG) leads to activation of PKC. By docking onto the plasma membrane PITP␣ undergoes a structural change such that the Ser 166 residue is accessible to protein kinase C. Phosphorylated PITP␣ is unable to deliver its PI cargo but not its PC cargo. The disruption in the delivery of PI leads to local depletion of PI 4,5-bisphosphate (PIP 2 ) and cessation of the PLC response. Pretreatment of cells with PMA mimics this process by enhancing phosphorylation of PITP␣ at Ser 166 .
transiently deplete PITP␣ in a variety of cell types including neuronal cells with agonists that work through G-protein-coupled receptors. Using interference RNA we have been able to show that transiently decreased levels of PITP␣ do affect the production of inositol phosphates. 2 Using different cell types we were able to show that disruption of PLC activity is both agonist-and cell type-dependent. We would suggest that these differing effects correlate with the levels of other PI transfer proteins such as PITP␤ present in these cell types as well as the potency of the agonist used.
In conclusion, we identified two residues, Thr 59 and Ser 166 , as phosphorylation sites on PITP␣ that had dramatic effects on PITP␣-dependent phospholipase C signaling. The effect of phosphorylation was to specifically reduce the delivery of PI for PLC signaling and hence decrease inositol phosphate production. Retaining PC transfer is not sufficient for the protein to function in PLC signaling as exemplified by T59E. We anticipate that PITP␤, which also can participate in PLC signaling and is known to go to the plasma membrane during stimulation with EGF (19,20), will be phosphorylated by PKC in a similar manner at both Thr 59 and Ser 166 . Phosphorylation of Ser 166 in PITP␤ by PKC has been reported recently (53