(Received for publication, November 1, 1996, and in revised form, February 26, 1997)
From the Department of Physiology, University College London, London WC1E 6JJ, United Kingdom
Phosphatidylinositol transfer protein (PITP) is
essential 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
comparing wild-type PITP with PITP
in which 5, 10, and 20 amino
acids were deleted from the C terminus.
5- and
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.
20-PITP was inactive at all concentrations tested. All
three truncated mutants were unable to restore G-protein-mediated
phospholipase C
stimulation in HL-60 cells.
5- and
10-PITP,
but not
20-PITP, inhibited the signaling function of wild-type
protein without any effect on lipid transfer in vitro. We
conclude that (a) the carboxyl terminus of PITP plays a
critical role in phospholipase 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 inhibitory effects of
5- and
10-PITP
on wild-type PITP in phospholipase C signaling suggest the existence of
a receptor for PITP.
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 addition (6). In cells, both PITP
and PITP
participate in phospholipase C (PLC)-mediated signaling (7-9) 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 1,4,5-trisphosphate 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 (7-9, 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 Ca2+ 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 Ca2+- 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 5- and
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
5- and
10-PITP
on wild-type PITP in PLC signaling suggest the existence of a receptor
for PITP.
Recombinant
wild-type and mutant PITP proteins were cloned by the polymerase
chain reaction using human PITP
cDNA as template and the
following primers: 5
-GAACTTCTAGATCCCATATGGTGCTGCTCA-3
and
5
-CTCGGGATCCTAGTCATCTGCTGTCATTCC-3
for wild-type PITP,
5
-CATCTGGATCCTATCCTTTCACTGGGTC-3
for
5-PITP,
5
-CCTTTGGATCCCTACTTTTGTCTCATTTCA-3
for
10-PITP, 5
-CCAGGGATCCCTACGTCTCTTCTTCCATCC-3
for
20-PITP,
5
-GCCGGATCCTAGTCATCTGCTGTCATTCCTATCACTGG-3
for K264I,
5
-GCCGGATCCTAGTCATCTGCTGCCATTCCT-3
for T267A, and 5
-GCCGGATCCTAGTCATCTGCTACCATTCCT-3
for T267V. The polymerase chain reaction products were cloned into the
NdeI-BamHI restriction site of the pET14b
expression vector (QIAGEN Inc.) and transformed into XL1-Blue cells
(Stratagene). Positive clones were sequenced, compared with published
human PITP
cDNA sequence, and transformed into DE3 (pLysS) cells
(Promega). Expression of recombinant PITP
was induced with
isopropyl-
-D-thiogalactopyranoside (0.1 mM) for 4 h at room temperature, and bacterial cells were collected by
centrifugation. The pellet was resuspended in buffer containing 50 mM sodium phosphate and 300 mM NaCl (pH
8.0). After freeze-thaw, the samples were centrifuged at 100,000 × g for 30 min at 4 °C. Recombinant proteins were
purified from the supernatants using Ni2+/nitrilotriacetic
acid-agarose resin (QIAGEN Inc.) as described previously (9).
PI transfer activity was assayed as described previously (7). This assay measures the transfer of [3H]PI from rat liver microsomes to unlabeled liposomes in the presence of transfer protein. Briefly, protein samples were added to tubes containing [3H]PI-labeled microsomes (62.5 µg 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 µl. After incubation at 27 °C for 30 min, microsomes were precipitated by the addition of 25 µl of ice-cold 0.2 M sodium acetate (pH 5.0) and removed by centrifugation (12,000 × g for 15 min). A 100-µl aliquot of the supernatant was measured for radioactivity.
Assay for PC transfer activity measures the transfer of radioactivity from [3H]PC-labeled liposomes to rat liver mitochondria. The liposomes consisted of 2 mmol of egg yolk PC/ml containing 1 µCi of [3H]PC in SET buffer and were sonicated on ice prior to use. [3H]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 × g for 10 min. The sedimented mitochondria were resuspended in 0.5 ml of SET buffer and sedimented by centrifugation at 12,000 × g for 10 min through 0.5 ml of 14.3% sucrose. The pellet was resuspended in 50 µl of 10% SDS and boiled for 5 min, and this solution was counted for radioactivity.
Determination of the Lipid Bound to Recombinant PITPAfter overnight culture of
E. coli cells expressing PITP proteins in the presence of
[3H]acetate (20 ml, 15 MBq), expression of PITP
proteins was induced with
isopropyl-
-D-thiogalactopyranoside 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.
50 µg of
purified PITP proteins (45 µl) were incubated with 5 µl of
[3H]PC/phosphatidic acid (70:30 mol %; 120 µM and 0.1 µCi/µl final concentrations) or
[3H]PI (120 µM and 0.0135 µCi/µl final
concentrations) in 20 mM Tris buffer containing 5 mM MgCl2 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
MgCl2 and 60 mM NaCl (pH 7.4) as described
previously (7).
HL-60 cells
were grown and labeled with [3H]inositol as described
previously (13). Labeled HL-60 cells (5 × 107) were
permeabilized with 0.6 IU/ml streptolysin O in PIPES (supplemented with
1 mM Mg2+·ATP 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 Mg2+·ATP, 20 mM LiCl, and 4 mM MgCl2 at
pCa 6. Permeabilized cells (20 µl) were incubated with
recombinant proteins in the presence of 10 µM GTPS in
a final volume of 45 µl 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.
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 µg/ml wild-type PITP was maximal in the PI transfer
assay, and corresponding concentrations of the mutant proteins were
inactive. The 5- and
10-PITP mutants displayed substantial PI
transfer activity when concentrations as high as 200 µg/ml were
tested (Fig. 2A, panel b). In comparison, near normal transfer of PC could be observed at high concentrations of the
5- and
10-PITP mutant proteins. Although
5- and
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
[3H]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 incorporated into wild-type
PITP was predominantly PG, with some PE (8:2 PG/PE) (Fig.
3B). In comparison to wild-type PITP, the 5- and
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 >60%. The results in
Fig. 3B indicate that
5- and
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
10- and
20-PITP mutants to exchange the lipids are significantly decreased.
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 µg 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.
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).
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
5- and
10-PITP were not observed when lipid transfer was examined
in vitro (Fig. 5B).
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 interchangeably 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 three to four times more protein was
required. Despite the residual transfer activity obtained with
5-
and
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
Arg253 and Arg259, 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.