The First 5 Amino Acids of the Carboxyl Terminus of Phosphatidylinositol Transfer Protein (PITP) (cid:97) Play a Critical Role in Inositol Lipid Signaling TRANSFER ACTIVITY OF PITP IS ESSENTIAL BUT NOT SUFFICIENT FOR RESTORATION OF PHOSPHOLIPASE C SIGNALING*

Phosphatidylinositol transfer protein (PITP) is essen- tial for phospholipase C signaling and for constitutive and regulated vesicular traffic. PITP has a single lipid- binding site that can reversibly bind phosphatidylinositol (PI) and phosphatidylcholine (PC) and transfer these lipids between membrane compartments in vitro . The role of the carboxyl terminus was examined by compar- ing wild-type PITP (cid:97) with PITP (cid:97) in which 5, 10, and 20 amino acids were deleted from the C terminus. (cid:68) 5- and (cid:68) 10-PITP had reduced PI and PC transfer activities compared with wild-type PITP, with the effect on PI transfer being more marked than that on PC transfer. (cid:68) 20-PITP was inactive at all concentrations tested. All three truncated mutants were unable to restore G-pro- tein-mediated phospholipase C (cid:98) stimulation in HL-60 cells. (cid:68) 5- and (cid:68) 10-PITP, but not (cid:68) 20-PITP, inhibited the signaling function of wild-type protein without any ef- fect on lipid transfer in vitro . We conclude that ( a ) the carboxyl terminus of PITP plays a critical role in phos- pholipase C signaling; ( b ) the transfer activity is not the only determining factor that dictates the restorative function of PITP in inositol lipid signaling; and ( c ) the dominant

Phosphatidylinositol transfer protein (PITP) 1 is a ubiquitous and abundant cytosolic protein that was originally identified because of its ability to transfer phosphatidylinositol (PI) and phosphatidylcholine (PC) between membrane bilayers in vitro (1,2). Two isoforms of PITP (␣ and ␤) that show different lipid binding properties have been identified in mammalian cells. PITP␣ has a single lipid-binding site that can reversibly bind PI and PC, with a 16-fold higher affinity for PI than for PC (3,4). In vitro, PITP␣ can transfer PI, PC, and PG to a lesser extent (5), whereas PITP␤ can transfer sphingomyelin in addi-tion (6). In cells, both PITP␣ and PITP␤ participate in phospholipase C (PLC)-mediated signaling (7)(8)(9) and in vesicular traffic (10 -12).
PITP␣ was identified as the major reconstituting factor that allowed restoration of PLC␤ signaling in HL-60 cells (7). A requirement for PITP has also been identified for inositol 1,4,5trisphosphate production by receptors that activate PLC␥1. When activated by the appropriate agonist, both the epidermal growth factor and IgE receptors are dependent on PITP for PLC␥1 signaling in A431 and RBL-2H3 cells, respectively (8,9). The mechanism of PITP function in lipid signaling has been attributed to the lipid binding/transfer properties of PITP, and the delivery of PI to a signaling complex composed of PI 4-kinase, phosphatidylinositol-phosphate 5-kinase, and the receptor has been proposed as its physiological function (7)(8)(9)13).
A separate role for PITP in exocytosis has also been identified (10,12,14). PITP was purified as a reconstituting factor together with phosphatidylinositol-phosphate 5-kinase for restoration of ATP-dependent priming of secretory vesicles for fusion with the plasma membrane in PC12 cells (10,14). Cells of the myeloid lineage including neutrophils and HL-60 cells secrete lysosomal enzymes when activated with Ca 2ϩ and guanine nucleotides. HL-60 cells depleted of cytosolic proteins become refractory to stimulation, and PITP was found to be required to restore secretory competence for Ca 2ϩ -and guanine nucleotide-mediated exocytosis (12). The addition of PITP led to increased synthesis of PI 4,5-bisphosphate, and this function of PITP provides the most likely explanation of how PITP participates both in exocytosis and in PLC-mediating signaling. In PLC signaling, PI 4,5-bisphosphate functions as a substrate, whereas its function in exocytosis is probably due to the recruitment of specific protein(s) required for the exocytotic machinery.
Secretory vesicle formation is also dependent on cytosolic proteins, and PITP␣ and PITP␤ were identified as the active components (11). This function of PITP in mammalian cells is analogous to the requirement of yeast PITP (SEC14p) for export of secretory proteins from the Golgi complex (15). Evidence has been presented that suggests that it functions as a lipid sensor that controls the PC content of yeast Golgi membranes (16). Although yeast PITP (SEC14p) shares no primary sequence homology with mammalian PITP, SEC14p can restore the function of mammalian PITPs in PLC signaling, exocytosis, and vesicle formation (9 -12). The function of yeast PITP is not conserved as mutations in SEC14p in the dimorphic yeast Yarrowia lipolytica do not lead to impaired secretion. Instead, it is required for differentiation from the yeast to the mycelial form (17).
In this report, we show that the carboxyl terminus of the PITP molecule plays a critical role in the restoration of PLC signaling. Deletion of 5 amino acids is sufficient to inactivate the protein with regard to PLC-mediated signaling, although lipid transfer is only reduced. In addition, the ⌬5and ⌬10-PITP deletion mutants were found to inhibit restoration of PLC signaling by wild-type PITP, although they did not interfere with the lipid transfer function of PITP in vitro. We conclude that (a) the carboxyl terminus of PITP plays a critical role in PLC signaling; (b) the transfer activity is not the only determining factor that dictates the restorative function of PITP in inositol lipid signaling; and (c) the dominant inhibitory effects of ⌬5and ⌬10-PITP on wild-type PITP in PLC signaling suggest the existence of a receptor for PITP.
Assay for PI and PC Transfer-PI transfer activity was assayed as described previously (7). This assay measures the transfer of [ 3 H]PI from rat liver microsomes to unlabeled liposomes in the presence of transfer protein. Briefly, protein samples were added to tubes containing [ 3 H]PI-labeled microsomes (62.5 g of microsome protein), liposomes (50 nmol of phospholipid; 98 mol % PC:2 mol % PI), and SET buffer (0.25 M sucrose, 1 mM EDTA, and 5 mM Tris-HCl (pH 7.4)) in a final volume of 125 l. After incubation at 27°C for 30 min, microsomes were precipitated by the addition of 25 l of ice-cold 0.2 M sodium acetate (pH 5.0) and removed by centrifugation (12,000 ϫ g for 15 min). A 100-l aliquot of the supernatant was measured for radioactivity.
Assay for PC transfer activity measures the transfer of radioactivity from [ 3 H]PC-labeled liposomes to rat liver mitochondria. The liposomes consisted of 2 mmol of egg yolk PC/ml containing 1 Ci of [ 3 H]PC in SET buffer and were sonicated on ice prior to use. [ 3 H]PC-labeled liposomes (40 nmol) were incubated with transfer protein and rat liver mitochondria (2 mg of protein) in a final volume of 0.2 ml of SET buffer for 30 min at 37°C. The reaction was halted by placing samples on ice, and mitochondria were sedimented by centrifugation at 12,000 ϫ g for 10 min. The sedimented mitochondria were resuspended in 0.5 ml of SET buffer and sedimented by centrifugation at 12,000 ϫ g for 10 min through 0.5 ml of 14.3% sucrose. The pellet was resuspended in 50 l of 10% SDS and boiled for 5 min, and this solution was counted for radioactivity.
Determination of the Lipid Bound to Recombinant PITP␣ Proteins Purified from Escherichia coli-After overnight culture of E. coli cells expressing PITP␣ proteins in the presence of [ 3 H]acetate (20 ml, 15 MBq), expression of PITP␣ proteins was induced with isopropyl-␤-Dthiogalactopyranoside for 4 h at room temperature, and recombinant proteins were prepared as described above. Lipids were extracted from purified proteins and separated by TLC (solvent: chloroform/methanol/ acetic acid/water (75:45:3:0.2, v/v)). Unlabeled standards were added to the samples to aid recovery and for identification on the TLC plate. The TLC plate was stained with iodine to locate the lipids, and the radioactivity was measured after excision of the spots for cardiolipin, PG, and PE. The TLC plate was also analyzed by imaging the radioactivity on a PhosphorImager.
Exchange of the Endogenous Lipids with PI and PC-50 g of purified PITP␣ proteins (45 l) were incubated with 5 l of [ 3 H]PC/phosphatidic acid (70:30 mol %; 120 M and 0.1 Ci/l final concentrations) or [ 3 H]PI (120 M and 0.0135 Ci/l final concentrations) in 20 mM Tris buffer containing 5 mM MgCl 2 and 60 mM NaCl (pH 7.4). After the solution was incubated at room temperature for 1 h, PITP␣ proteins were repurified with nitrilotriacetic acid spin (QIAGEN Inc.), and the radioactivity was counted.
For exchange of endogenous bacterial lipids for PI or PC prior to use in PLC assays, purified wild-type PITP and ⌬5-PITP were incubated with the appropriate lipid vesicles at a ratio of 1 mg of protein to 2.4 mg of lipid (1:100 molar ratio) and incubated at 4°C overnight. The lipid vesicles were removed by the addition of DE52 in 20 mM Tris buffer containing 5 mM MgCl 2 and 60 mM NaCl (pH 7.4) as described previously (7).
Reconstitution of G-protein-regulated Phospholipase C Activity in Cytosol-depleted HL-60 Cells with PITP␣ Proteins-HL-60 cells were grown and labeled with [ 3 H]inositol as described previously (13). Labeled HL-60 cells (5 ϫ 10 7 ) were permeabilized with 0.6 IU/ml streptolysin O in PIPES (supplemented with 1 mM Mg 2ϩ ⅐ATP at pCa 7) for 10 min to deplete the cells of cytosolic proteins including PITP. The permeabilized cells were washed and finally resuspended in PIPES supplemented with 4 mM Mg 2ϩ ⅐ATP, 20 mM LiCl, and 4 mM MgCl 2 at pCa 6. Permeabilized cells (20 l) were incubated with recombinant proteins in the presence of 10 M GTP␥S in a final volume of 45 l at 37°C for 20 min. The reaction was quenched by the addition of chloroform/ methanol, and inositol phosphates were extracted, separated by passage through Dowex 1-X8 anion-exchange resin, and counted for radioactivity.

C-terminal Truncation of PITP Reduces Its Ability to Trans-
fer Phospholipids-We have constructed three C-terminally truncated forms of PITP␣ (⌬5, ⌬10, and ⌬20) in which 5, 10, and 20 amino acids residues are deleted, respectively (Fig. 1A). The proteins were expressed in E. coli and purified, and the in vitro transfer activities of these recombinant proteins were compared with those of wild-type PITP (Figs. 1B and 2 (A and B)). In addition to these three mutants, we constructed ⌬80-PITP, but this mutant could not be recovered from the soluble fraction of E. coli.
Transfer activity was monitored using a donor membrane that contained radiolabeled PI or PC and an acceptor membrane compartment. The acceptor compartment was reisolated and monitored for the transferred radiolabeled lipid. Deletion of 5 amino acids was sufficient to reduce the transfer activity for both PI and PC (Fig. 2, A and B). 1 g/ml wild-type PITP was maximal in the PI transfer assay, and corresponding concentrations of the mutant proteins were inactive. The ⌬5and ⌬10-PITP mutants displayed substantial PI transfer activity when concentrations as high as 200 g/ml were tested ( Fig. 2A,  panel b). In comparison, near normal transfer of PC could be observed at high concentrations of the ⌬5and ⌬10-PITP mutant proteins. Although ⌬5and ⌬10-PITP retained some ability to transfer PI and PC, ⌬20-PITP could not transfer PI or PC (Fig. 2, A and B).
To carry out phospholipid transport, PITP has to interact with a phospholipid membrane surface and exchange the endogenous lipid for another. When PITP is purified from mammalian tissues, the two major lipids that are bound to the protein are PI and PC at a ratio of 65:35 (3,4). The recombinant proteins were expressed in E. coli, an organism devoid of both PI and PC. When expressed in E. coli cells, PITP is exposed to PE (84.4%), PG (12.5%), and cardiolipin (2.6%), the main phospholipids present in these cells. We first established which of the endogenous E. coli lipids were associated with purified PITPs.
E. coli cells were grown in the presence of [ 3 H]acetate overnight prior to induction to label the endogenous E. coli lipids. PITP was purified, and the lipids associated with the protein were extracted and analyzed by TLC (Fig. 3A). The lipid incor-porated into wild-type PITP was predominantly PG, with some PE (8:2 PG/PE) (Fig. 3B). In comparison to wild-type PITP, the ⌬5and ⌬10-PITP mutant proteins had a higher proportion of PE, shifting the ratio to 6:4 PG/PE. This change in binding properties was more marked in ⌬20-PITP, where the ratio was nearly 5:5. In addition, the amount of lipid bound to ⌬20-PITP was decreased by Ͼ60%. The results in Fig. 3B indicate that ⌬5and ⌬10-PITP have a similar occupancy compared with wild-type PITP, but only 35% of ⌬20-PITP has a lipid bound to the protein.
The exchange of the endogenous lipids with radiolabeled PI and PC was also examined. The proteins were incubated with PC and PI lipid vesicles and reisolated using the His tag, and the radioactivity associated with the proteins was compared with that of wild-type PITP. The data in Fig. 3C illustrate that although ⌬5-PITP can exchange the endogenous lipids with PI and PC as well as wild-type PITP, the abilities of the ⌬10and ⌬20-PITP mutants to exchange the lipids are significantly decreased.
Deletion Mutants Have No Ability to Restore Inositol Signaling in Cytosol-depleted HL-60 Cells-⌬5-and ⌬10-PITP had reduced PI transfer activity, but near normal PC transfer activity at high concentrations of protein. If lipid transfer was the sole determining factor that was responsible for PITP function in PLC signaling, it would be expected that these proteins may retain some residual capacity to restore signaling in cells when used at high concentrations. We utilized the G-protein-driven PLC␤ signaling in HL-60 cells for this purpose. HL-60 cells were depleted of endogenous PITP, and the washed cells were incubated with wild-type PITP and ⌬5-, ⌬10-, and ⌬20-PITP in the presence of GTP␥S (Fig. 4). Near maximal restoration of PLC signaling occurred around 100 g of PITP/ml. None of the truncated proteins were able to restore PLC-mediated signaling. The maximal concentration that was practical to test in this assay was 1.5 mg/ml, and despite the addition of such a huge concentration, no reconstitution was observed. Instead, at these high concentrations, a slight inhibition of the GTP␥S response was evident.
To exclude the possibility that the bacterial lipids could in any way interfere with the restoration of PLC signaling when added to permeabilized cells, the bacterial lipids were exchanged for PI or PC. Wild-type PITP and ⌬5-PITP were converted into the PC and PI forms and compared with the protein loaded with the bacterial lipid PG. Fig. 4B illustrates that the ability of PITP to function in permeabilized cells in PLC signaling is independent of the nature of the loaded lipid.
Site-directed Mutational Analysis-To identify which specific residues may be important for the lipid binding/transfer properties of PITP, we mutated 2 candidate residues. There are clusters of basic amino acids in the C-terminal region in both PITP␣ and PITP␤, and it was reported that basic amino acid residues are very important for PI 4,5-bisphosphate binding (18). PITP␣ contains a lysine residue at position 264, whereas PITP␤ has a similar basic amino acid, arginine, in this position. In addition to basic amino acids, Alb et al. (19) have reported that replacement of threonine 59 with several different amino acids abolished the PI transfer activity of PITP without any effect on PC transfer activity. Threonine 267 in PITP␣ is a serine in PITP␤, again a conservative change. Although we replaced lysine 264 with isoleucine and threonine 267 with valine or alanine, the site-directed mutants had normal lipid transfer activity (Fig. 2C) and exhibited normal activity in the PLC reconstitution assay (data not shown).
⌬5-PITP Inhibits the Function of Wild-type PITP in PLC Signaling, but Not in Lipid Transfer in Vitro-The truncated mutants were examined for their ability to inhibit the restorative effects of wild-type PITP (Fig. 5A). ⌬5-PITP inhibited wild-type PITP when added simultaneously to permeabilized HL-60 cells. The concentration of ⌬5-PITP required for inhibition was much lower than the amount of wild-type protein added. Although ⌬10-PITP acted as well as ⌬5-PITP, ⌬20-PITP did not affect the activity of wild-type PITP to restore PLC signaling. The inhibitory effects of ⌬5and ⌬10-PITP were not observed when lipid transfer was examined in vitro (Fig. 5B). DISCUSSION PITP has diverse effects on several cell functions ranging from PLC signaling to membrane traffic. In all cases reported to date, mammalian PITP␣ and PITP␤ can be used inter-changeably with SEC14p, the yeast form of PITP, despite the lack of sequence homology with mammalian PITP. Since the PI binding/transfer activity is the common feature shared by the two mammalian forms of PITP and SEC14p, it must be the relevant activity that determines their abilities to restore PLC signaling and membrane traffic. Rescue of SEC14 defects in yeast with mammalian PITP has also been observed (20). However, PITP␣ only rescues the temperature-sensitive mutations, but not the null mutation. This result clearly indicates that SEC14p and mammalian PITP have distinct as well as overlapping functions.
To identify the structural requirements of PITP, we have systematically deleted the carboxyl terminus of PITP␣ and have compared the lipid binding/transfer characteristics in vitro with the ability of the truncated proteins to restore inositol lipid signaling in HL-60 cells. Deletion of just 5 amino acid residues was sufficient to impair the function of PITP both in lipid transfer and in PLC signaling. In addition, the ⌬5-truncated form of PITP was inhibitory to wild-type PITP and potentially functions as a dominant-negative mutant.
Deletion of 5 residues at the C terminus decreased the lipid transfer activity of both PI and PC. PI transfer activity was affected more than PC transfer activity. For example, ⌬5-PITP retained 30% of PI transfer activity when examined at 10 times the concentration of wild-type PITP, whereas PC transfer activity was comparable to that of wild-type PITP except that FIG. 3. A, PG and PE are associated with wild-type PITP purified from E. coli. Wild-type PITP was purified from E. coli cells grown in the presence of [ 3 H]acetate overnight prior to induction, and the lipids associated with the purified protein were extracted and separated by TLC. The radiolabel associated with the lipids was analyzed using a PhosphorImager. The positions of the lipid standards (PE, PG, and cardiolipin (CL)) are indicated. B, shown is the quantitation of the lipids associated with wild-type PITP and with ⌬5-, ⌬10-, and ⌬20-PITP. Wild-type and truncated forms of PITP were purified from E. coli cells grown in the presence of [ 3 H]acetate overnight prior to induction, and the lipids associated with the purified protein were extracted and separated by TLC. The radioactivity associated with the lipids was determined and is expressed as dpm/mg of protein. For each lipid, the dpm is also expressed as a percentage of the total radioactivity found in PE ϩ PG ϩ cardiolipin. The data from three independent experiments were pooled to calculate the percentage of the total radioactivity (ϮS.E., n ϭ 3). C, shown is the binding of radiolabeled PI or PC to the recombinant proteins. Purified PITP␣ proteins were incubated with [ 3 H]PC/phosphatidic acid or [ 3 H]PI. After the solution was incubated at room temperature for 1 h, PITP␣ proteins were repurified, and the radioactivity was counted. The results are expressed as % of binding observed with wild-type PITP. three to four times more protein was required. Despite the residual transfer activity obtained with ⌬5and ⌬10-PITP, both these deletion mutants had no activity when examined for restoration of G-protein-mediated PLC signaling in HL-60 cells even when 15-fold higher concentrations over wild-type protein were used. The uncoupling of transfer function with reconstitution supports the emerging view that the function of PITP is not just to passively transfer PI from intracellular membranes where it is synthesized to sites of PLC signaling, but to directly participate in the synthesis of PI 4,5-bisphosphate (8,12,13).
It is possible that the deletion of the 5 amino acids from the C terminus affects the lipid binding properties of PITP significantly and hence the loss in transfer function. This was not found to be the case. We first examined which lipids were associated with the full-length recombinant protein purified from E. coli and found that PITP was loaded with PG predominantly (80%) and the remainder with PE. A previous study reported that PITP purified from E. coli had only PG bound to it (21). Their failure to detect PE probably resulted from the use of different techniques to monitor the lipid content. In the study reported here, we radiolabeled E. coli prior to induction so that the protein expressed would incorporate radiolabeled lipid. Analysis by TLC clearly revealed that both PG and PE could associate with PITP. The other study extracted the lipid from PITP and examined the chemical mass of the lipid (21). It is very possible that the low levels of PE were not detectable by mass determination.
Deletion mutants of PITP (⌬5 and ⌬10) were still capable of binding the endogenous E. coli lipid, but had slightly altered ratios of anionic lipids (PG) to zwitterionic lipids (PE). Deletion of 20 amino acids had a more dramatic effect in that the amount of lipid associated was decreased by 60%, and this implies that a substantial amount of PITP does not have a lipid bound to it. In all three deletion mutants, a detectable amount of label in cardiolipin was also observed, suggesting that PITP mutants are less discriminatory than wild-type PITP (Fig. 3B).
Truncation also affected the ability to exchange the endogenous lipids with PI or PC in vitro. But the effects were small, and ⌬5-PITP could exchange PI or PC as well as wild-type PITP. However, their ability to transfer PI or PC was significantly impaired. ⌬20-PITP was completely inactive when examined for PI or PC transfer. Loss of PI transfer with ⌬5-PITP was reduced by 30%, whereas restoration of PLC signaling was completely disrupted taking into account the 10-fold higher concentrations of ⌬5-PITP used for both assays. This uncoupling of transfer with PLC signaling indicates that in addition to transfer, the PITP molecule has additional properties that are required for PLC signaling.
⌬5-PITP inhibits PITP function in cells, but not in the in vitro PI transfer assay. Inhibition by ⌬5-PITP was not of a competitive nature. This suggests that in cells ⌬5-PITP interacts with the "putative" receptor for PITP more strongly and effectively inhibits full-length PITP from exerting its effect. In the in vitro transfer assay, lipid transfer is a passive event and is not dependent on interaction with another protein. The identification of the binding partners for PITP will provide additional insights into the mechanism of PITP function. It should be recognized that in the visual transduction system in Drosophila, PITP is part of a larger entity that is an integral membrane protein (22).
In summary, we have identified the carboxyl terminus of PITP as being critical in PLC-mediated signaling. Deletion of the carboxyl terminus does not lead to complete loss of transfer activity, but does lead to complete loss of signaling in cells. Thus, the transfer activity is not the only determining factor that dictates the restorative function of PITP in inositol lipid signaling. It has been recently reported that limited proteolysis of PITP by trypsin cleaves the carboxyl terminus at Arg 253 and Arg 259 , and this leads to a decrease in PC transfer activity (23). These data are in accord with our results. In addition, our results suggest that in cells PITP interacts with a putative receptor. The identification of the binding partners for PITP will be greatly facilitated by taking advantage of ⌬5-PITP. The ability of the mutant proteins to block signaling will also provide valuable tools in assessing the function of PITP in living cells.