(Received for publication, March 28, 1997, and in revised form, May 30, 1997)
From the Department of Cell Biology, Harvard Medical
School and Division of Signal Transduction, Beth Israel Hospital,
Boston, Massachusetts 02115,
Onyx Pharmaceuticals, Richmond,
California 94806, the ** Program in Molecular Medicine and Department of
Biochemistry and Molecular Biology, University of Massachusetts Medical
Center, Worcester, Massachusetts 01605, the ¶¶ Division of
Pediatric Oncology, Dana-Farber Cancer Institute and Department of
Pediatrics, Harvard Medical School, Boston, Massachusetts 02115, the
|| Division of Medical Chemistry Pharmaceutics,
College of Pharmacy, University of Kentucky, Lexington, Kentucky
50536-0082, and the §§ Beirne B. Carter Center for
Immunology Research and Department of Microbiology, University of
Virginia, Charlottesville, Virginia 22908
Pleckstrin homology (PH) and phosphotyrosine
binding (PTB) domains are structurally related regulatory modules that
are present in a variety of proteins involved in signal transduction,
such as kinases, phospholipases, GTP exchange proteins, and adapter proteins. Initially these domains were shown to mediate protein-protein interactions, but more recently they were also found to bind
phosphoinositides. Most studies to date have focused on binding of PH
domains to phosphatidylinositol (PtdIns)-4-P and
PtdIns-4,5-P2 and have not considered the lipid
products of phosphoinositide 3-kinase: PtdIns-3-P, PtdIns-3,4-P2, and PtdIns-3,4,5-P3. Here we
have compared the phosphoinositide specificity of six different PH
domains and the Shc PTB domain using all five phosphoinositides. We
show that the Bruton's tyrosine kinase PH domain binds to
PtdIns-3,4,5-P3 with higher affinity than to
PtdIns-4,5-P2, PtdIns-3,4-P2 or inositol 1,3,4,5-tetrakisphosphate (Ins-1,3,4,5-P4). This
selectivity is decreased by the xid mutation (R28C). Selective binding
of PtdIns-3,4,5-P3 over PtdIns-4,5-P2 or
PtdIns-3,4-P2 was also observed for the amino-terminal PH
domain of T lymphoma invasion and metastasis protein (Tiam-1), the PH
domains of Son-of-sevenless (Sos) and, to a lesser extent, the PH
domain of the -adrenergic receptor kinase. The oxysterol binding
protein and
-spectrin PH domains bound PtdIns-3,4,5-P3
and PtdIns-4,5-P2 with similar affinities. PtdIns-3,4,5-P3 and PtdIns-4,5-P2 also bound to
the PTB domain of Shc with similar affinities and lipid binding was
competed with phosphotyrosine (Tyr(P)-containing peptides. These
results indicate that distinct PH domains select for different
phosphoinositides.
Proteins involved in signal transduction are often composed of regulatory modules such as PH1 domains, Src homology 2 domains, and PTB domains (1, 2). It is believed that these well defined units can, alone or in combination, determine the subcellular localization of a protein by mediating protein/protein interactions and also, as recently described, protein/lipid interactions (3-5).
PH domains were first identified as protein regions that share
similarities with pleckstrin, the major protein kinase C substrate in
platelets (6). They contain approximately 120 amino acids and are found
in several proteins involved in signaling, such as protein kinases
(Btk, Akt, and ark), phospholipases (PLC), and proteins that act as
exchange factors or GTPase-activating proteins for small G-proteins
(Ras-GRF, Dbl, SOS, Ras-GAP, Tiam-1) (7). Despite their relatively low
sequence homology, the three-dimensional structures of the PH domains
are highly conserved (8-15). Interestingly, the recently described PTB
domains of Shc and IRS-1 have structural similarity to PH domains and
may be thought of as a subclass of PH domains with an additional region
that allows binding to tyrosine-phosphorylated proteins (5, 16).
The physiological relevance of phosphoinositide binding to PH domains
is still not completely clear. Several PH domains have now been shown
to bind with relatively high affinity to either PtdIns-4,5-P2 or Ins-1,4,5-P3 (3, 5, 17-24).
In the case of the phospholipase C- PH domain, considerable evidence
is provided that PtdIns-4,5-P2 binding to this domain
recruits it to the membrane where substrates reside and that the
product of PtdIns-4,5-P2 hydrolysis,
Ins-1,4,5-P3 binds to the PH domain to dissociate it from
the membrane as a feedback regulatory mechanism (15). In most of the
studies referenced above, the possibility that the various PH domains
might bind to products of PI 3-kinase was not investigated. This is an
important question since unlike PtdIns-4,5-P2 and
PtdIns-4-P, which are constitutively produced in cells,
PtdIns-3,4-P2 and PtdIns-3,4,5-P3 are nominally
absent in quiescent cells and only appear in response to cell
stimulation (25). Thus, these lipids could provide a mechanism for
regulated recruitment of proteins to the membrane. However, even in
stimulated cells the levels of PtdIns-3,4-P2 and
PtdIns-3,4,5-P3 are far below the levels of PtdIns-4-P and
PtdIns-4,5-P2. Therefore, in order for a protein to be
regulated by one of these lipids, it must have a strong selectivity for
the D-3 lipids compared with the affinity for PtdIns-4-P or
PtdIns-4,5-P2. While this manuscript was in review, two
papers have appeared indicating that the PH domain of the Akt/PKB
protein-Ser/Thr kinase selectively binds to and is activated by
PtdIns-3,4-P2 (26, 27). Another paper appeared showing that
the PH domain of the Btk protein-Tyr kinase selectively binds to
PtdIns-3,4,5-P3 (28) and another paper appeared showing that the Btk PH domain selectively binds to Ins-1,3,4,5-P4
(29).
In this study we have compared the ability of six different PH domains and the Shc PTB domain to bind to five different phosphoinositides as well as to inositol phosphates. We find that a subgroup of the PH domains investigated has a relatively high selectivity for binding to PtdIns-3,4,5-P3, while others bind PtdIns-4,5-P2 and PtdIns-3,4,5-P3 with comparable affinity. All of the domains investigated here bound PtdIns-3,4-P2 poorly compared with PtdIns-4,5-P2 or PtdIns-3,4,5-P3, indicating specificity to the binding. These results indicate that distinct PH domains have evolved selectivity for different phosphoinositides to provide discriminatory regulation.
pGEX vectors containing the cDNA
sequences encoding the mouse Btk PH domain (amino acids 6 to 217),
the Shc-PTB domain (amino acids 17-207), the
ark, OSBP,
-spectrin, mSos1 (amino acids 456-569), and Tiam-1 N-terminal PH
domains were expressed in Escherichia coli as GST fusion
proteins by isopropyl-
-D-thiogalactopyranoside induction
(5, 30). Bacterial lysates (prepared as described in Ref. 31) were
incubated with glutathione-Sepharose beads for 2 h at 4 °C and
washed several times with 30 mM Hepes, pH 7.0, 100 mM NaCl, 1 mM EDTA (HNE) containing 0.5%
Nonidet P-40, followed by HNE, without detergent. Point mutations
within the Btk PH domain (R28C) were introduced using the Transformer
site-directed mutagenesis kit (CLONTECH) and
confirmed by DNA sequencing.
Synthetic water-soluble 3H-labeled dioctanoyl-PtdIns-3,4,5-P3 (5 × 107 cpm/µmol) (32) was incubated with Sepharose beads containing approximately 800 nM of GST-Btk PH domain fusion protein or GST alone in HNE-0.02% Nonidet P-40 buffer. After 1 h at room temperature, the beads were separated from the supernatant by centrifugation. The supernatant was then collected, mixed with scintillation liquid, and counted on a Beckman counter. The amount of [3H]PtdIns-3,4,5-P3 bound to the Btk PH domain was calculated by subtracting the amount of free 3H present in the GST-Btk PH domain supernatant from the amount of free 3H present in the control GST supernatant under identical conditions. The beads containing GST alone retained less than 3% of the total counts. A nonlinear least squares fit to the data (Kaleidagraph) was determined using the equation [bound]= Bmax × [free]/(KD + [free]), where [bound] is the concentration of [3H]C8PtdIns-3,4,5-P3 bound, [free] is the concentration of [3H]C8PtdIns-3,4,5-P3 free in solution, Bmax is the saturation binding, and KD is the dissociation constant.
Lipid Competition Assay[32P]3-Phosphoinositides were prepared by
incubating 200 µM of a crude brain phosphoinositide
mixture (Sigma; contains 60% PtdIns and phosphatidylserine, 15-20%
PtdIns-4-P, and 15-20% PtdIns-4,5-P2] with purified PI
3-kinase as described previously (33). After 30 min at room
temperature, the reaction was stopped with 10 mM EDTA, and
the lipids were isolated by chloroform:methanol extraction as described
(34). Sonicated 32P-labeled phosphoinositides (prepared as
described above) or synthetic water-soluble 3H-labeled
C8PtdIns-3,4,5-P3 (6500 cpm; 0.93 µM) were
incubated with competitor lipids (unlabeled PtdIns-4,5-P2
(Avanti) or unlabeled synthetic PtdIns-3,4,5-P3 (32)) and
15 µl of Sepharose beads containing approximately 2-20 µg of the
indicated GST fusion protein in the HNE buffer containing 0.02%
Nonidet P-40. The incubation was carried out for 1 h at room
temperature, after which the beads were washed twice with 1 ml of
HNE-0.5% Nonidet P-40 (2-3 min total). The 32P-labeled
lipids that remained associated with the beads were extracted and
resolved by thin layer chromatography (TLC) using 1-propanol:2
M acetic acid (65:35, v/v). The radioactivity migrating with PtdIns-3,4,5-P3 was quantified using a Bio-Rad
phosphorimager. When 3H-labeled PtdIns-3,4,5-P3
was used as a tracer, the washed beads were mixed with scintillation
liquid, and the radioactivity associated with them was measured in a
Beckman scintillation counter. The data were plotted as a percentage of
the control with no additional competitive lipid added. A nonlinear
least squares fit to the data was performed using the equation % bound = 100
n × L/(KI(app) + L), where
n indicates the percent specific binding, L
indicates the concentration of unlabeled lipid added, and
KI(app) is the apparent competitive dissociation
constant for the unlabeled lipid. This equation assumes simple
competitive binding and that the competitive lipid is in excess over
the PH domain concentration (L ~ Ltotal). The apparent competitive dissociation
constant relates to true KI by the equation
KI(app) = KI (1 + L*/KD*), where L* and
KD* are the free concentration and dissociation
constant of the radiolabeled lipid, respectively. Thus,
KI(app) is an overestimation of the true
KI but approaches the true KI at
low concentrations of L* and low concentrations of the
PH domain where L ~ Ltotal. Since the same concentration of PH domain and radiolabeled lipid are
used for a given set of PtdIns-4,5-P2 and
PtdIns-3,4,5-P3 competition experiments, the ratios of the
apparent dissociation constants measured equal the ratios of the true
dissociation constants: e.g.
KI(app)(PtdIns-4,5-P2)/KI(app)(PtdIns-3,4,5-P3) = KI(PtdIns-4,5-P2)/KI(PtdIns-3,4,5-P3).
For the peptide competition for lipid binding to the Shc PTB domain, the tyrosine-phosphorylated peptides SCFTNQGpYFF (from the interleukin-2 receptor) and RENEpYMPMAPQIH (from the polyomavirus middle T protein) were used (pY indicates phosphotyrosine).
Lipid Selectivity[32P]Phosphoinositides were prepared using a 200 µM solution of presonicated crude brain phosphoinositides (Sigma) and a mixture of purified PI 3-kinase, PtdIns 4-kinase, and PtdIns-4-P 5-kinase as described (33). Dried lipids were resuspended in HNE buffer, sonicated, and added to 40 µl of Sepharose beads containing ~5 µg of the Btk PH domain GST fusion protein in HNE-0.02% Nonidet P-40. Samples were incubated and washed as described above. Lipids that remained associated with the beads were extracted, deacylated, and analyzed by HPLC as described (34). The radioactivity associated with each selected lipid was then divided by the radioactivity associated with the same lipid in the starting mixture to determine the fraction of each lipid that bound to a given domain. On average, 8% of the total radioactivity added remained associated with the wild type Btk PH domain beads and 4% with the R28C mutant. The GST beads alone bound only 0.5% of the total lipids, such that the radioactivity associated with each lipid was often undetectable. To compare the relative selectivity of wild type and mutant Btk PH domain for PtdIns-3,4,5-P3, the data were normalized such that if all the lipids bound equally well, each would have a selectivity of 1 (4).
Protein Competition AssayPhosphorylated EGF receptor intracellular domain, including the cytoplasmic tail (TKD 61, a gift from Dr. R. Cerione from Cornell University, Ithaca, NY), was prepared by infecting SF9 cells with baculovirus carrying the TKD 61 sequence. The cells were lysed in a hypotonic solution containing 30 mM Hepes, pH 7.0, 1 mM EDTA, by passing the cell suspension through a 23-gauge needle. The lysate was cleared by centrifugation at 15,000 rpm for 10 min and incubated with 10 mM MgCl2 and 1 mM ATP at 37 °C for 1 h. GST-Shc-PTB-containing beads were incubated with different concentrations of sonicated PtdIns-4,5-P2 or PtdIns-3,4,5-P3 for 1 h at room temperature in the presence of HNE containing 0.02% Nonidet P-40. Phosphorylated TKD 61 was added to the lipid/beads mixture and incubated for 20 min at room temperature. The beads were washed two times with HNE-0.5% Nonidet P-40, resuspended in an equal volume of 2 × Laemmli buffer and incubated for 10 min at 100 °C. Proteins were resolved by 7.5% SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose membrane, and blotted using anti-Tyr(P) antibody (4G10). Blots were developed using the chemiluminescence system (ECL from Amersham Corp.).
To avoid the complexity of
interpreting protein binding to lipid bilayers and micelles, a
water-soluble form of PtdIns-3,4,5-P3 (dioctanoyl-PtdIns-3,4,5-P3, abbreviated as
C8PtdIns-3,4,5-P3) was used to investigate PH domain
binding characteristics. Various concentrations of water soluble
3H-labeled C8PtdIns-3,4,5-P3 were incubated
with bacterially expressed GST-Btk PH domain fusion protein immobilized
on beads. Fig. 1A is a plot of
the amount of the ligand bound versus free. The data fitted
a hyperbolic curve indicating a KD of approximately 800 nM, and the stoichiometry was extrapolated to
approximately 0.9 mol of C8PtdIns-3,4,5-P3 bound per mol of
PH domain.
To determine the importance of the diacylglycerol moiety for binding, we compared the abilities of Ins-1,3,4,5-P4 (the head group of PtdIns-3,4,5-P3) and unlabeled C8PtdIns-3,4,5-P3 to displace [3H]C8PtdIns-3,4,5-P3 from the Btk PH domain. The results in Fig. 1B indicate that very high concentrations of Ins-1,3,4,5-P4 (>100 µM) are required to displace the [3H]C8PtdIns-3,4,5-P3. In contrast, unlabeled C8PtdIns-3,4,5-P3 caused 50% displacement at approximately 4 µM. As expected the KI(app) for C8PtdIns-3,4,5-P3 is somewhat higher than the KD determined by direct binding (0.8 µM), since the KI(app) is affected by the total concentration of the PH domain and the concentration of [3H]C8PtdIns-3,4,5-P3 used in the experiment (see "Materials and Methods"). These results indicate that PtdIns-3,4,5-P3 is capable of directly binding to the PH domain of Btk with approximately 40 times higher affinity than Ins-1,3,4,5-P4. Hence, the diacylglycerol moiety plays a significant role in the interaction of PtdIns-3,4,5-P3 with the Btk PH domain.
To measure the relative affinity of the Btk PH domain for
phosphoinositides with long chain fatty acids, Nonidet P-40 micelles containing PtdIns-4,5-P2 or a synthetic dipalmitoyl version
of PtdIns-3,4,5-P3 (C16PtdIns-3,4,5-P3) were
used to compete for [3H]C8PtdIns-3,4,5-P3
binding to this domain (Fig.
2A). A least squares fit to
the data yields a KI(app) of 0.7 µM
for binding of C16PtdIns-3,4,5-P3. Thus,
C16PtdIns-3,4,5-P3 appears to have a 5-6-fold higher
affinity than C8PtdIns-3,4,5-P3 (Fig. 2A versus
Fig. 1B). However, interpretation of this result is complicated by the fact that C16PtdIns-3,4,5-P3 forms a
micelle. Fig. 2A also shows that
C16PtdIns-3,4,5-P3 has a higher affinity for the Btk PH
domain than does PtdIns-4,5-P2, as judged by the concentrations of the two lipids required to displace
[3H]C8PtdIns-3,4,5-P3. The ratio of these two
constants (8 µM/0.7 µM = 11) indicates that
the Btk PH domain has an order of magnitude selectivity for
PtdIns-3,4,5-P3 over PtdIns-4,5-P2. Addition of physiological concentrations of calcium and magnesium did not significantly affect binding of C8PtdIns-3,4,5-P3,
C16PtdIns-3,4,5-P3, or PtdIns-4,5-P2 (not
shown).
Since the [3H]C8PtdIns-3,4,5-P3 utilized as a probe in Fig. 2A is water-soluble and not a natural form of PtdIns-3,4,5-P3, we also investigated binding using [32P]PtdIns-3,4,5-P3 produced enzymatically from natural lipids as the probe. In this experiment both the radiolabeled probe and the competing lipid are in micelles. The results presented in Fig. 2B again show that C16PtdIns-3,4,5-P3 has about a 1 order of magnitude higher affinity for the BTK PH domain than does PtdIns-4,5-P2. The apparent KI values determined in this experiment are higher than those determined in Fig. 2A due to differences in PH domain concentration, probe concentration, and probe affinity, but the ratio of the constants (21 µM/2.5 µM = 8.4) is similar to that found in Fig. 2A. High concentrations of PtdIns (20 µM), PtdIns-3,4-P2 (20 µM), Ins-1,4,5-P3 (1 mM), and Ins-1,3,4-P3 (1 mM) were unable to displace [32P]PtdIns-3,4,5-P3 (not shown).
As a third assessment of the selectivity of the Btk PH domain, we added
this domain to a mixture of radiolabeled versions of all five
phosphoinositides sonicated into the same vesicle containing carrier
PtdIns and phosphatidylserine and determined which lipids were
preferentially retained following stringent washing with Nonidet P-40.
This experiment indicates which lipid has the slowest off rate.
Consistent with the competitive binding experiments,
[32P]PtdIns-3,4,5-P3 was preferentially
retained (Fig. 3A). The fact that both PtdIns-4,5-P2 and PtdIns-3,4-P2 bound
poorly compared with PtdIns-3,4,5-P3 indicates that
phosphates at both the 3 and 5 positions are critical for high affinity
binding. An Arg to Cys mutation at position 28 in the PH domain of Btk
(BtkR28C) causes agammaglobulinemia. In agreement with other studies
(28, 29), we find that this mutation dramatically reduces the
selectivity for PtdIns-3,4,5-P3 over the other lipids by
both direct binding experiments (Fig. 3B) and competition
experiments (not shown).
PH domains from Sos and Tiam-1 Also Bind Preferentially to PtdIns-3,4,5-P3
We investigated the selectivity of
additional PH domains using the competitive binding assays. Both the PH
domain of Sos and the N-terminal PH domain of Tiam-1 bound
PtdIns-3,4,5-P3 with high affinity and specificity. Using
the [3H]C8PtdIns-3,4,5-P3 displacement assay,
both of these PH domains had apparent KI values for
C16PtdIns-3,4,5-P3 of less than 1 micromolar (Fig. 4,
A and B). As
discussed above, these measurements are underestimates of the true
affinities for C16PtdIns-3,4,5-P3. In these experiments,
the Sos PH domain had about a 5-fold selectivity for
PtdIns-3,4,5-P3 over PtdIns-4,5-P2 (ratio of
apparent KI values = 2 µM/0.4
µM). The Tiam-1 N-terminal PH domain had about a
3-fold selectivity for PtdIns-3,4,5-P3 over
PtdIns-4,5-P2 (ratio of apparent KI
values = 1.5 µM/0.5 µM).
The selectivities of these two domains were also investigated using displacement of micellar [32P]PtdIns-3,4,5-P3 (Fig. 4, C and D). Although the apparent KI values determined by this assay were higher due to the different probes and conditions, these assays confirmed that both PH domains preferentially bound C16PtdIns-3,4,5-P3 over PtdIns-4,5-P2 (Sos PH domain selectivity, 16 µM/7 µM = 2.3-fold: Tiam-1 N-terminal PH domain selectivity, 15 µM/4 µM = 3.75-fold). Thus two different types of binding experiments indicate that both of these domains have between 2-5-fold selectivities for binding to PtdIns-3,4,5-P3 compared with PtdIns-4,5-P2. Both direct binding experiments (not shown) and competition experiments (Fig. 4, A and B) indicated that PtdIns-3,4-P2 binds more weakly than either PtdIns-3,4,5-P3 or PtdIns-4,5-P2 to these two PH domains. Thus, like the Btk PH domain, the Tiam-1 and Sos PH domains have specificity for PtdIns-4,5-P2 over PtdIns-3,4-P2, but have highest affinity for PtdIns-3,4,5-P3.
PH Domains ofThe PH domains of ark, OSBP, and
-spectrin were
also analyzed. The percentage of
[32P]PtdIns-3,4,5-P3 that was competed by 10 µM PtdIns-4,5-P2 was compared with the
percentage that was competed by 10 µM
PtdIns-3,4,5-P3, to give an estimation of the relative
affinity of the different PH domains for these two lipids. The results
are summarized in Table I. In contrast to
the Btk PH domain, the OSBP and the
-spectrin PH domains bind to
PtdIns-4,5-P2 and PtdIns-3,4,5-P3 with similar affinities. The
ark PH domain showed a somewhat higher affinity for
PtdIns-3,4,5-P3 than for PtdIns-4,5-P2, both in
this experiment and in direct binding experiments (not shown). However,
this domain bound PtdIns-3,4,5-P3 more weakly than did the
Btk, Tiam-1 and Sos PH domains.
|
Recently, it was shown
that the Shc-PTB is able to bind to liposomes containing
PtdIns-4,5-P2 (5). To determine whether this domain also
binds PtdIns-3,4,5-P3, 32P-labeled
PtdIns-3,4,5-P3 was incubated with GST-Shc-PTB immobilized on beads, in the presence or absence of Tyr(P)-containing peptides. After washing the beads with Nonidet P-40, approximately 20% of the
total PtdIns-3,4,5-P3 (Fig.
5) remained associated with the beads. In
the presence of 10 µM of a peptide containing the
sequence NQGpY, that binds with high affinity to the Shc-PTB domain
Tyr(P)-binding pocket (35), PtdIns-3,4,5-P3 binding was
inhibited by 85%. A control phosphopeptide (10 µM) had
little effect on PtdIns-3,4,5-P3 binding. These results
suggest that, like PtdIns-4,5-P2,
PtdIns-3,4,5-P3 can directly bind to the Shc-PTB domain,
and this interaction can be competed with a phosphopeptide that binds
with high affinity to this domain.
To compare the relative affinities of the different
phosphoinositides for the Shc-PTB domain, PtdIns,
PtdIns-4,5-P2, or PtdIns-3,4,5-P3 was incubated
with GST-Shc-PTB beads, together with
[32P]PtdIns-3,4,5-P3, in an experiment
analogous to that depicted in Fig. 2B. The results presented
in Fig. 6 indicate that PtdIns does not
displace [32P]PtdIns-3,4,5-P3 binding, even
at 40 µM, while PtdIns-4,5-P2 and
C16PtdIns-3,4,5-P3 compete for
[32P]PtdIns-3,4,5-P3 binding to the Shc-PTB
domain with similar affinities (apparent KI of 25 µM for both lipids).
Phosphoinositide Binding to the Shc-PTB Domain Competes with Protein Binding
The ability of the tyrosine-phosphorylated
peptide to block phosphoinositide binding to the Shc-PTB domain
suggests that the lipid binding and Tyr(P) binding are competitive. To
determine whether phosphoinositides and tyrosine-phosphorylated
proteins are able to simultaneously bind to the Shc-PTB domain, the
ability of PtdIns-4,5-P2 and PtdIns-3,4,5-P3 to
compete with the tyrosine-phosphorylated EGF receptor for Shc-PTB
domain binding was examined. Tyrosine-phosphorylated EGF receptor
intracellular domain was incubated with GST-Shc-PTB beads in the
presence of different concentrations of PtdIns-4,5-P2 or
PtdIns-3,4,5-P3. The EGF receptor that remained associated with the Shc-PTB domain, after an extensive wash, was separated by
SDS-polyacrylamide gel electrophoresis and blotted with anti-Tyr(P) antibody (Fig. 7).
PtdIns-4,5-P2 and PtdIns-3,4,5-P3 were equally able to compete with the tyrosine phosphorylated EGF receptor for
binding to the Shc-PTB domain. 50% competition was observed in the
presence of 10-20 µM of each of these lipids, consistent with their abilities to displace
[32P]PtdIns-3,4,5-P3 (Fig. 5). PtdIns and
PtdIns-3,4-P2 had no effect on EGF receptor binding at
concentrations of up to 80 µM (not shown), indicating
that the binding is specific for PtdIns-4,5-P2 over
PtdIns-3,4-P2.
The data shown in Figs. 5 and 7 suggest that PtdIns-4,5-P2 may play an important role in Shc-PTB domain association with membranes or tyrosine-phosphorylated proteins, but that conversion of PtdIns-4,5-P2 to PtdIns-3,4,5-P3 by PI 3-kinase is unlikely to alter these interactions.
The results presented here demonstrate that different PH domains
have specificity for different phosphoinositides. The domains that we
have investigated fall into two major categories, those with high
affinity and specificity for PtdIns-3,4,5-P3 over
PtdIns-4,5-P2 (Btk, Tiam-1 N-terminal, Sos) and those that
bind PtdIns-3,4,5-P3 relatively weakly and have only slight
or no selectivity for PtdIns-3,4,5-P3 over
PtdIns-4,5-P2 (ark, OSBP and spectrin PH domains and Shc PTB domain). All of these domains had selectivity for
PtdIns-4,5-P2 over PtdIns-3,4-P2 (Fig. 4 and
data not shown) indicating that phosphate at the D-5
position is important for binding. In contrast, results from this
laboratory (26) and another laboratory (27) have demonstrated that the
Akt/PKB PH domain has high affinity and selectivity for
PtdIns-3,4-P2 over PtdIns-4,5-P2. Thus,
different PH domains have evolved the ability to bind to distinct
phosphoinositides, presumably to allow regulation by distinct
extracellular signals.
The specificity of the Btk PH domain for PtdIns-3,4,5-P3 may be of physiological importance. Using a water soluble C8PtdIns-3,4,5-P3 we were able to show that this domain binds monomeric PtdIns-3,4,5-P3 at one mol/mol with a KD less than 1 µM. Competitive binding experiments demonstrated that C16PtdIns-3,4,5-P3 has a 5-6-fold higher affinity than C8PtdIns-3,4,5-P3 and that Ins-1,3,4,5-P4 (the head group of PtdIns-3,4,5-P3) has a 40-fold lower affinity than C8-PtdIns-3,4,5-P3. These results demonstrate that this domain can bind with high affinity to a monomeric lipid and that the diacylglycerol moiety significantly contributes to the binding affinity of the monomer. Similarly, the dynamin PH domain was shown to bind to detergent-solublized PtdIns-4,5-P2 and glycerophosphorylinositol 4,5-bisphosphate, but to have relatively low affinity for PtdIns-4,5-P2 contained in lipid vesicles or to inositol 1,4,5-P3 (36). The apparent higher affinity of the Btk PH domain for C16PtdIns-3,4,5-P3 over C8PtdIns-3,4,5-P3 could be explained by additional contact with the longer chain fatty acid or to a preference for binding micellar lipid over monomeric lipid, due to partitioning into the micelle.
Three different approaches (Fig. 2, A and B, and Fig. 3) indicate that the Btk PH domain has about a 1 order of magnitude higher affinity for PtdIns-3,4,5-P3 than for PtdIns-4,5-P2. These results are in agreement with another study in which PtdIns-3,4,5-P3 liposome binding to the Btk PH domain was detected by the BIAcore biosensor (28). In that study, PtdIns-4,5-P2 liposomes failed to bind. Since with the BIAcore technique it is difficult to detect binding of ligands with KD values higher than 1 µM, the results are consistent with our finding that PtdIns-3,4,5-P3 binds with a KD of less than 1 µM, while PtdIns-4,5-P2 binds with a KD of about 8 µM. Salim et al. (28) did not investigate binding of inositol phosphates. Another laboratory measured direct binding of inositol phosphates to the Btk PH domain and found that Ins-1,3,4,5-P4 bound more tightly than Ins-1,4,5-P3 or other inositol polyphosphates examined (29). This result is in agreement with our observation that Ins-1,3,4,5-P4 is more effective than Ins-1,4,5-P3 or Ins-1,3,4-P3 in competing for PtdIns-3,4,5-P3 binding. However, the competitive binding constant we observed is much higher than the KD reported by Fukuda et al. (29) (40 nM), and we found that C8PtdIns-3,4,5-P3 bound with a 40-fold higher affinity than Ins-1,3,4,5-P4, suggesting that the natural ligand is the lipid rather than the inositol phosphate. Fukuda et al. (29) did not examine binding to phosphoinositides.
A point mutation in the Btk PH domain that causes agammaglobulinemia (R28C) was found to decrease the affinity for PtdIns-3,4,5-P3 (Fig. 3 and data not shown). This result is in agreement with the results obtained by Salim et al. (28) and suggests that binding to PtdIns-3,4,5-P3 may regulate the function of the Btk protein-Tyr kinase in vivo by affecting its location or activity.
In addition to Btk, several other PH domains selectively bind PtdIns-3,4,5-P3 (Group I in Table II). Included are three PH domains of G-protein exchange factors, Sos, Tiam-1 (N-terminal), and GRP-1. Sos activates Ras, Tiam-1 activates Rac, and Grp-1 is predicted to be an exchange factor for adenosine diphosphate ribosylation factor-1 (ARF-1) family members (37). The results we obtained with Sos do not agree with a previous finding that this domain preferentially bound to liposomes containing PtdIns-4,5-P2 compared with liposomes containing PtdIns-3,4,5-P3 (38). The previous study may be complicated by the fact that PtdIns-3,4,5-P3 perturbs liposome structures.2 The direct binding and competitive binding experiments with radiolabeled lipids clearly indicate a small but significant preference for PtdIns-3,4,5-P3. Our results suggest that local conversion of PtdIns-4,5-P2 to PtdIns-3,4,5-P3 by PI 3-kinase may enhance recruitment of Sos and/or Tiam-1 to the region of the membrane where PI 3-kinase is activated. Although these domains bind PtdIns-3,4,5-P3 with KD values that are submicromolar, they are less selective for PtdIns-3,4,5-P3 over PtdIns-4,5-P2 than is the Btk PH domain. Thus, it is likely that PtdIns-4,5-P2 also plays a role in recruitment of these proteins to the membrane. Deletion of the Sos PH domain was shown to interfere with its exchange activity in vivo (39). Whether phosphoinositide binding to the PH domains of these exchange factors stimulate their exchange activities remains to be determined. Tiam-1 also has a PH domain C-terminal of the exchange domain (40). Deletion of the N-terminal PH domain of Tiam-1 abolished membrane localization while deletion of the C-terminal PH domain did not affect localization (41). We found that the C-terminal PH domain bound relatively weakly to phosphoinositides (not shown). However, it might contribute to membrane binding or activation once Tiam-1 is located at the membrane via the N-terminal PH domain.
|
Grp-1 (general receptor for phosphoinositides-1) was discovered because of its ability to bind to PtdIns-3,4,5-P3 in a expression cloning assay (37). GRP-1 is a member of a family of proteins with N-terminal Sec7 homology domains and C-terminal PH domains. The PH domain was shown to have high affinity and high selectivity for PtdIns-3,4,5-P3 (37). One member of this family, ARF nucleotide binding site opener, was shown to act as an exchange factor for ARF-1 and this activity was stimulated by phosphoinositides in vitro (24). Whether PtdIns-3,4,5-P3 regulates ARF family members in vivo remains to be determined.
We found another group of PH domains that bind
PtdIns-3,4,5-P3 relatively weakly and have little or no
preference for PtdIns-3,4,5-P3 over
PtdIns-4,5-P2. These include PH domains of ark,
spectrin, and OSBP (Table II) and the PTB domain of Shc. These domains
do have specificity for PtdIns-4,5-P2 over
PtdIns-3,4-P2, suggesting that the lipid binding is
relevant for function. Indeed, PtdIns-4,5-P2 has been shown
to act synergistically with
subunits in activation of
ark
(20). Since PtdIns-4,5-P2 is more abundant than
PtdIns-3,4,5-P3 in vivo (25), the results
indicate that these proteins are not regulated in vivo by
products of PI 3-kinase. The PH domain of phospholipase C-
appears
to fit into this group, since it binds PtdIns-4,5-P2 and
PtdIns-3,4,5-P3 with similar affinities and binds
PtdIns-3,4,-P2 more weakly (22).
The Akt/PKB PH domain makes up a third category. In contrast to the other PH domains that we have investigated, this domain preferentially binds PtdIns-3,4-P2 (26, 27). In addition, PtdIns-3,4-P2 activates the protein-Ser/Thr kinase activity of Akt, while PtdIns-4,5-P2 has no effect, and PtdIns-3,4,5-P3 causes inhibition.
A final group of PH domains that we investigated bound very poorly to phosphoinositides under the assay conditions that we used (not shown). This group includes the C-terminal Tiam-1 PH domain (discussed above), the dynamin PH domain, and the IRS-1 PTB domain. A previous study indicated that the dynamin PH domain could bind with high affinity to PtdIns-4-P and PtdIns-4,5-P2 only when these lipids were presented in a detergent-solublized form (36). There are several possible interpretations for these results. The proteins that contain these domains may utilize weak phosphoinositide binding at PH domains in conjunction with lipid or protein interaction at other domains for membrane localization. These domains may function exclusively as protein interaction domains or they may bind to vesicles with a specific curvature or to lipids distinct from phosphoinositides. A trivial explanation is that these domains may fail to fold properly in bacteria.
Based on the crystal structure of the PLC PH domain bound to
Ins-1,4,5-P3 (14), residues from the
1-
2 loop and the
3-
4 loop of PH domains are predicted to interact with the
phosphate moieties of bound phosphoinositides. In Table II the PH
domains are divided into four groups based on phosphoinositide
selectivity. This table also includes the sequences at the predicted
1-
2 loops of the PH domains studied here, aligned according to
the crystal structure of PLC
(14). The exclamation marks indicate the basic residues of PLC
(14) and spectrin (42) from this region
that are involved in phosphate binding. The Arg-28 residue of Btk is
indicated by an asterisk. The analogous residue in PLC
(R40)
coordinates the D-5 phosphate of Ins-1,4,5-P3. The effects of the R28C mutation described here and elsewhere (28, 29) suggest that
this residue may coordinate the D-3 phosphate of PtdIns-3,4,5-P3. Recently published studies of the Akt PH
domain are also in agreement with this idea (26). It is likely that other residues in the
1-
2 loop (as well as the
3-
4 loop)
contribute to selectivity in phosphoinositide binding.
In general those PH domains with clusters of basic residues in the
1-
2 region tend to have higher affinity (and selectivity) for
PtdIns-3,4,5-P3. Table II is color-coded so that basic
residues are in blue and acidic residues in red. The Btk, Tiam-1
N-terminal, and GRP-1 PH domains have 8, 11, and 6 basic residues,
respectively, in the
1-
2 region. Most of the other domains with
lower affinity for PtdIns-3,4,5-P3 have 5 or fewer basic
residues. Residues at the
3-
4 and
5-
6 loops were also shown
to make contacts with Ins-1,4,5-P3 (14, 42); however, the
number of positive residues in these regions does not correlate with
high affinity binding to PtdIns-3,4,5-P3 (not shown).
Although the spectrin PH domain has a large
1-
2 loop with many
basic and acidic residues, this protein has a Val at the position
analogous to Arg-28 of Btk. Compared with Btk it binds relatively
weakly to phosphoinositides (and inositol phosphates). The structure of
the spectrin PH·Ins-1,4,5-P3 complex indicates a very
different binding mode from that of PLC
(42).
Finally, it is interesting that phosphoinositide binding to the Shc-PTB domain competes with phosphotyrosine-containing protein binding. This result is analogous to our previous finding that PtdIns-3,4,5-P3 binding to Src homology 2 domains is competitive with Tyr(P) peptides(4). These results suggest that modulation of Tyr(P) binding to adapter domains by phosphoinositides may be a general phenomenon in signaling. However, the Src homology 2 domains investigated had high selectivity for PtdIns-3,4,5-P3, while the Shc-PTB domain did not discriminate between PtdIns-4,5-P2 and PtdIns-3,4,5-P3.
We thank Dr. R. Lefkowitz and Dr. K. Touhara
for providing the GST-ark and GST-OSBP constructs, Dr. Gerry Shaw
for the GST-
-spectrin construct, Dr. Jeremy Thorner for the purified
PtdIns 4-kinase, Brian Duckworth for the purified PtdIns-4-P 5-kinase,
and Dr. R. Cerione for the Baculovirus expression vector encoding the EGF cytoplasmic tail.